Fall prevention intervention technologies: A conceptual framework and survey of the state of the art

Julian Hamm; Arthur G. Money; Anita Atwal; Ioannis Paraskevopoulos

doi:10.1016/j.jbi.2015.12.013

Hamm, Julian;Money, Arthur G.;Atwal, Anita;Paraskevopoulos, Ioannis;

2016-02-01 00:00:00

Journal of Biomedical Informatics 59 (2016) 319–345 Contents lists available at ScienceDirect Journal of Biomedical Informatics journal homepage: www.elsevier.com/locate/yjbin Methodological Review Fall prevention intervention technologies: A conceptual framework and survey of the state of the art a a,⇑ b c Julian Hamm , Arthur G. Money , Anita Atwal , Ioannis Paraskevopoulos Department of Computer Science, Brunel University London, UK Department of Clinical Sciences, Brunel University London, UK Department of Computing and Information Systems, University of Greenwich, UK article i nfo abstract Article history: In recent years, an ever increasing range of technology-based applications have been developed with the Received 16 October 2015 goal of assisting in the delivery of more effective and efﬁcient fall prevention interventions. Whilst there Revised 14 December 2015 have been a number of studies that have surveyed technologies for a particular sub-domain of fall pre- Accepted 20 December 2015 vention, there is no existing research which surveys the full spectrum of falls prevention interventions Available online 7 January 2016 and characterises the range of technologies that have augmented this landscape. This study presents a conceptual framework and survey of the state of the art of technology-based fall prevention systems Keywords: which is derived from a systematic template analysis of studies presented in contemporary research lit- Falls prevention erature. The framework proposes four broad categories of fall prevention intervention system: Pre-fall Technology-based interventions prevention; Post-fall prevention; Fall injury prevention; Cross-fall prevention. Other categories include, Conceptual framework Application type, Technology deployment platform, Information sources, Deployment environment, Collaborative care User interface type, and Collaborative function. After presenting the conceptual framework, a detailed Healthcare Self-care survey of the state of the art is presented as a function of the proposed framework. A number of research challenges emerge as a result of surveying the research literature, which include a need for: new systems that focus on overcoming extrinsic falls risk factors; systems that support the environmental risk assess- ment process; systems that enable patients and practitioners to develop more collaborative relationships and engage in shared decision making during falls risk assessment and prevention activities. In response to these challenges, recommendations and future research directions are proposed to overcome each respective challenge. 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/). 1. Introduction demand for health and social care services resulting in a cost of £1.8 billion per year to the National Health Service (NHS) in the Falls prevention within the home environment has been a topic UK [6]. of research for over 30 years [1] and is recognised as an important Falls prevention activities are carried out across a range of health issue within the United Kingdom (UK), Europe, North Amer- health disciplines including occupational therapy, physiotherapy, ica and Australia [2]. The frequency of falls increases with age, general practice, nursing, geriatric, gerontology health and social often as a result of physical, functional, and cognitive impairments care [7–9]. There is evidence in the falls prevention research liter- which are likely to emerge as a result of advanced ageing [3]. Con- ature which suggests that in excess of 50% of potential falls relating sequently, it is estimated that 30% of older adults aged 65 and over to older adults are avoided as a result of ongoing falls prevention fall at least once a year [4]. One in ﬁve falls result in bone fractures interventions [10]. There is a range of clinically established preven- and the need for specialist medical attention [5]. Fall related frac- tion interventions that target fall related risk factors [1]. A number tures may cause disabilities and in some extreme cases premature of recent meta analyses, and systematic reviews considered a com- death among older adults, which has a signiﬁcant impact on prehensive range of falls prevention intervention studies for pre- venting falls in community-dwelling older people [11–15]. Fig. 1 presents a diagrammatic summary of the key categories of inter- Corresponding author at: Department of Computer Science, Brunel University, vention that are considered in these reviews and serves as a London, Middlesex UB8 3PH, UK. Tel.: +44 (0)1895 266 758; fax: +44 (0)1895 269 high-level overview of the key areas in which falls prevention E-mail addresses: [email protected] (J. Hamm), Arthur.Money@brunel. research has been undertaken in recent years. ac.uk (A.G. Money), [email protected] (A. Atwal). http://dx.doi.org/10.1016/j.jbi.2015.12.013 1532-0464/ 2016 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). 320 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 fall to avoid long lie syndrome. As such, there is little research evi- Exercise Education interventions Supervised Unsupervised dence of education interventions as a single component interven- Fall prevention tion that reduces the risk and rate of falls [11]. interventions Home assessments are carried out and assistive equipment is pre- Home assessment / Fall risk assessment Assistive equipment scribed to reduce falls within the home environment. Typically, home assessments involve clinicians visiting the older adult’s Technology-based interventions home to assess the suitability of the home environment in relation to the mobility of the patient. Clinicians then propose adaptations, Fig. 1. Overview of falls prevention interventions. often via the installation of assistive equipment, in order to facili- tate independent living and to mitigate any potential fall risks, which could arise during performing activities of daily living In recent years, one popular approach to falls prevention has (ADLs). Accordingly, reviews in the falls literature have revealed been to explore ways of targeting the restoration of muscle that home assessments and adaptations as a single intervention strength and balance for prevention of fall risks [16,17]. Exercise do not, in general, signiﬁcantly reduce the risk of falling. They do, interventions are becoming an increasingly popular approach to however, have some positive effect for those who are at higher risk falls prevention and there is an extensive body of evidence sug- of falling [8,11]. Furthermore, identifying environmental risks and gesting that these interventions can be effective in reducing falls adapting the living environment accordingly may reduce fall risks and the risk of falling [18]. There are many issues, however, with among older adults signiﬁcantly [24]. By deﬁnition, assistive regards to adherence and acceptance of the range of existing exer- equipment are systems or specialist devices prescribed by clini- cise interventions. Supervised one-to-one interventions with the cians, that provide functional support to older adults to help with patient and the practitioner are resource intensive in terms of cost mobility, which would otherwise been proven difﬁcult to do and and time, whilst supervised group exercise interventions require maximises independent living and reduces falls. Assistive equip- older adults to be able to travel to the location of exercise classes. ment includes grab rails, walking frames, hoists, raised toilet seats, Furthermore, there are many issues with regards to adherence and stair rails, raised chairs and beds within the patient’s home acceptance of existing unsupervised home-based exercise interven- [25–30]. Notwithstanding the beneﬁts of the assistive equipment tions, partly due to the lack of interactivity and personalisation provision, there are issues which often persist with the use of that the paper-based exercise interventions typically use in these equipment as it is not always adopted successfully. Consequently, settings [19]. As such, 3D technology and games are increasingly research evidence indicates that more than 50% of home modiﬁca- being seen as a potential means of improving adherence by provid- tions and equipment are rejected [31–33]. As a result, there has ing patients with more tailored and interactive exercise programs been an increase in functional decline, leaving older adults vulner- to engage with [20,21]. able to the risk of falling. Equipment abandonment is often associ- Fall risk assessment is an approach used to assess a number of ated with a number of factors such as lack of knowledge about the risk factors, speciﬁcally mobility issues and physiological factors equipment’s use, involving the users in the decision making pro- that include muscle strength and balance, stability, posture and cess, their attitude towards the equipment, and a lack of ﬁt of gait reaction time. There are many tests (e.g. Berg balance scale, the equipment between service users and their environment Timed Up and Go, Turn 180 test) that have been developed to [32,34–36]. screen older people for fall risks in the community or in a clinical Technology-based interventions have been deployed in a wide setting [22]. These tests are widely known with research evidence range of falls prevention contexts and include diagnosing and that supports their effective use in predicting fall risks to uncover treating fall risks [37–39], increasing adherence to interventions issues that may lead to falls. Older adults who are exposed to fall [40–42], detecting falls and alerting clinicians in case of falls risks such as gait and balance abnormalities, admitted into hospital [43–45]. Technology is also seen as having the potential to play a for medical attention as a result of falling are at high risk of falling. key role in enabling older adults to self-assess, which is in line with Consequently, they are offered a multifactorial fall risk assessment the personalisation agenda within the UK, giving older adults the that is administered by clinicians in a clinical setting, or within a opportunity to perform self-assessments for assistive equipment specialist fall service. Such assessments are a part of multifactorial provision [46–50]. With an increasing pressure and demand on risk assessment or a singular assessment. It is crucial that older the NHS and with limited spending budgets, partly due to an adults who are at high risk of falling are identiﬁed using the fall unprecedented increase of life expectancy resulting in an ageing risk assessment tests so that targeted falls prevention interven- population [51], there is a need to ﬁnd new ways of providing care tions can be prescribed. Conducting such assessments has included to enable patients to provide effective self-care and further steps high cost equipment in specialist fall services. However, 3D tech- towards recognising patients as experts of their own care by giving nology and games have shown promise as a low cost solution to them the chance to provide their own care [52]. Innovations in augment traditional fall risk assessments and to account for low technology are seen as key to reducing costs and lessening the bur- adherence rates of self-assessment of fall risks done at home [23]. den on the healthcare system, whilst also improving the quality Education interventions are developed to increase knowledge and effectiveness of care provided [48], thus enabling patients to about falls prevention and educate patients regarding their risk engage in the effectiveness of self-care to improve clinical out- of falling and falls prevention strategies based on the available comes. Encouraging the adoption of technology, however, has been evidence-based literature. This type of intervention, as a single a primary area of focus, particularly among the older population. component, is often part of a multifactorial falls prevention pro- There are contributing factors that include usability for the older gramme, which leads to positive outcomes such as behavioural adult cohort [53], exploring older users’ perceptions and beliefs change, decreased fear of falling and increased mobility. Education [54], intuitive interactions [55], and multisensory feedback [56], interventions typically take the form of fact sheets with evidence- which play a central role in motivating older adults to engage in based materials. These inform their readers about the preventive clinical interventions. These should be catered for if technological measures to reduce falls, or checklist to help to identify fall hazards interventions are to be adopted by older adults. Therefore, deploy- in the home and to take preventive measures such as change of ing usable and effective information and communication technolo- behavioural patterns. In addition, patients are also offered informa- gies (ICT) in areas of assisted healthcare, speciﬁcally falls tion regarding where they can seek help and assistance in case of a prevention, within the home has the potential to enable older Consensus J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 321 adults to maintain their independence and engage in unsupervised conceptual model, Sections 7 survey the falls technology systems interventions, remotely monitored by clinicians. There is, however, such as that of pre-fall, post-fall, fall injury and cross fall preven- an urgent need to explore the extent to which technology has been tion systems found in the literature to date, respectively. Section 8 developed for the falls prevention domain and to identify the areas discusses challenges of existing falls technology systems and rec- in which work is still required to respond positively to the broad ommends future research directions based on the gaps that exist range of challenges presented by this domain. Technology-based based on the survey of the state of the art in falls prevention tech- interventions have been identiﬁed as having valuable potential in nology research. Conclusions are drawn in Section 9. the applied sub-domains highlighted in Fig. 1; exercise, fall risk, education and home assessment. However, relatively little research 2. Research method has surveyed the extent to which technology has actually been applied to each of the sub-domains and the provision of collabora- This section provides a detailed explanation of the methods tive care, speciﬁcally the emerging patient–practitioner paradigm employed for this study. The steps taken to develop the conceptual within the context of falls prevention. Furthermore, little research framework and carry out the survey of the state of the art are pre- has covered the extent to which opportunities to support fall inter- sented in Fig. 2 and are described in more detail throughout this ventions have been explored respectively and the extent to which section. patients are being enabled to deliver effective self-care to improve clinical outcomes. A number of systematic reviews have been carried out in the 2.1. Literature search strategy falls prevention domain, some of these include: (1) general reviews [15,57], (2) exercise interventions [13,58], (3) fall risk assessment Initially, a number of survey papers were sourced to gain back- [59,60] and technology-based interventions [61]. Although a num- ground knowledge of the research area. Part of the search strategy ber of technology-based systematic reviews have been presented used for ﬁnding existing research was derived from reading previ- in the literature to date, such reviews tend to focus mainly on ous survey papers such as [15,58,63–65]. This provided candidate speciﬁc sub-domains of a much broader context of technology- search terms, keywords speciﬁc to the falls technology domain. based interventions. To the best of our knowledge, there is no The literature search strategy was a two-phase process. In Phase existing research which surveys and categorises across the full falls 1, electronic search and manual search was performed using elec- prevention intervention landscape, the types of existing tronic databases (IEEE Xplore, ACM, Pubmed, Web of Science, technology-based fall prevention systems, their key collaboration BioMed Central and ScienceDirect) to scan for papers that contain functions, the technologies they exploit, and the speciﬁc types of the search terms derived from the falls technology survey papers falls prevention interventions they support. Furthermore, there is that had already been considered. For each paper, a manual scan little existing research which, as a result of taking this holistic of the title and abstract was conducted, and then the paper was view, identiﬁes the areas of clinical practice, which appear to be included if it was considered relevant (the inclusion criteria is well catered and identiﬁes areas which require more attention. speciﬁed in the next section). In Phase 2, each paper’s reference list, In light of the need to better understand the state of the art of found from the electronic search, was manually scanned in order to the falls prevention technology landscape, this paper provides a identify other potentially relevant studies. Thus, the snowballing comprehensive review and a conceptual falls prevention technol- technique [66] was used in phase two to pursue additional papers ogy framework, which was developed as a result of carrying out from citation counts and the list of references in each paper, essen- a survey of the range of fall technology systems presented in the tially performing forward and backward searches. All searches con- literature. Section 2 outlines the research methods used to conduct ducted are based on a full screening of the studies, which were the literature survey. Section 3 presents the conceptual framework published between January 2010 and December 2014. The follow- and its component parts are explained. Through presenting the ing search strings were used in the electronic databases: Literature search strategy Phase 1 Phase 2 Phase 1Phase 2 Snowballing technique Electronic search Scan reference list Manual search Citation counts Inclusion and exclusion criteria Developing the conceptual framework Thematic analysis Thematic analysis Coding Splitting and joining Splitting and joining frame Overarching Overarching Literature Literature themes themes spreadsheet spreadsheet Electronic databases Literature Literature dataset dataset Sub-themes Sub-themes Conceptual Conceptual framework framework Fig. 2. Literature survey research protocol adapted from Afzal et al. [62]. The literature search strategy includes resources and the necessary steps used to survey the evidence and criteria for including studies. The developing the conceptual framework consists of the method and protocol used to construct the conceptual framework from the literature dataset. 322 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 Falls prevention technology AND older adults OR elderly. applications that aimed to: aid in fall risk assessment and/or Falls systems AND patients OR older adults OR elderly. prevention activities, respond to falls, or aid in reducing the risk Falls management AND patients OR older adults. of falling with or without the support of clinicians. Incomplete Falls prevention assistive technology AND patients OR older studies and studies written in another language other than English adults. were excluded. To ensure that the literature dataset reﬂected Falls prevention approaches AND patients OR older adults. recent developments in the ﬁeld whilst remaining manageable, all studies that appeared in the period 2010–2014 were included, Search terms that were used in this review were purposely kept any studies that were outside this time period were excluded from general to avoid potential bias in identifying a candidate dataset of the sample. Studies that did not involve the use of technology for studies which represents the state of the art. To enhance the falls prevention activity were also excluded from the corpus. Each search, Boolean operators were used so that synonyms of search study reference list was scanned for additional studies that met terms were included when carrying out automated searches. Pre- the inclusion criteria. liminary searches were conducted to identify search terms from existing reviews and to combine those search terms that derived 2.2. Developing the conceptual framework from the reviews. Fig. 3 presents the list of electronic databases used, the number of studies retrieved from the searches carried The conceptual framework was derived from surveying and out using search terms with for each respective electronic database, analysing the literature dataset identiﬁed from deploying the liter- the duplicate papers removed, and the total number of papers that ature search strategy presented in Fig. 2.A thematic analysis of the were deemed relevant. literature dataset was then performed in order to review and cat- The inclusion and exclusion criteria were used to identify egorise the studies that were included in the literature sample. appropriate studies, which proposed technology-based systems/ Thematic analysis is a qualitative analysis method for searching, Fig. 3. Literature search strategy. Including the search strategy used to search through the falls prevention technology literature, inclusion criteria set to include relevant studies, and results obtained from the search. J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 323 analysing and representing the overarching themes and sub- allocated to the appropriate code name and extracts of text from themes that emerge from textual datasets [67]. Consequently, the studies that ﬁt each concept were identiﬁed. The dataset the themes and sub-themes and their observed interrelated struc- was examined iteratively, to further develop themes and sub- ture which emerged as a consequence of carrying out a thematic themes. This was achieved via a process of splitting and joining analysis on the literature dataset were articulated via the incre- together of themes and associated text that was related to themes mental development of a conceptual framework that represents and sub-themes. At this point, a list of themes and sub-themes the state of the art of the falls prevention technology landscape. were used to classify each study in the dataset within the coding The following steps were taken to analyse the literature dataset frame. Several iterations of this reﬂective process were carried and develop the conceptual framework. Initially, all falls preven- out until the themes and sub-themes reﬂected the representative tion technology studies were added into a spreadsheet (used as literature dataset. Any inconsistencies were rectiﬁed, arriving at a a data management tool for primary studies that met the inclusion consensus pool of themes and sub-themes that formed the concep- criteria), making up the dataset. After studies were added, the tual framework in Fig. 4. The resulting conceptual framework rep- individual studies listed in the dataset were initially examined resents the falls prevention technology landscape according to the and overarching themes that emerged from the dataset were literature dataset which was analysed. A detailed description of recorded in the literature spreadsheet, which served as a coding the conceptual framework and its component parts (themes and frame for carrying out the thematic analysis. Each theme was sub-themes) is now provided in the next section. Falls prevention technology systems in practice Pre-fall prev. intervention systems #1 Post-fall prev. intervention systems #2 Fall injury prev. intervention systems #3 Fall risk factors Fall risk factor Intrinsic Fall related Intervention types Intervention types Intervention types injuries Extrinsic Functional assessment Activity monitoring Physical activities Cognitive assessment Fall detectors Cognitive training Environmental Medical assistance Education assessment Cross falls prev. intervention systems #4 Fall risk factors Combination of intervention types Technology Deployment #5 Systems Application type Static Interactive Game Virtual reality Platform Desktop Game console Smart-phone Information sources Sensor location Sensor purpose Context User Bespoke Repurposed Co-opted Deployment environment Home Nursing home Hospital Interface type Multimodal Interaction Natural user interfaces Touchscreen Collaboration Asynchronous Synchronous Patient Practitioner Fig. 4. Conceptual model of falls prevention technology. 324 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 3. A conceptual framework of falls prevention technology measure to take. Cognitive assessment includes tests performed to assess cognitive abilities and reduce the progression of cognitive The conceptual framework of the state of the art for falls pre- impairments, which typically lead to falls. Delivering this particu- lar intervention provides opportunities for clinicians to determine vention technology is presented in Fig. 4. The model is divided which preventive interventions are most appropriate to be carried between falls prevention technology systems in practice (illustrated out thereafter and thus, address the intrinsic risk factors identiﬁed in the top part of the ﬁgure), which looks at the various falls pre- as a result of the assessment. Environmental assessment involves vention interventions in practice. The second part of the model systems developed to assess extrinsic risks that impact on older considers technology deployment, which presents the range of falls adults’ ability to function independently within their living envi- technology systems proposed in the literature, the information ronment. This type of assessment aims to remove environmental sources they exploit, the types of user interface which they present hazards that obscure older adults’ ability to perform ADLs and rec- and their respective collaborative functions. ommend equipment to aid mobility and reduce fall risks in the home. 3.1. Falls prevention technology systems in practice Fall injury prevention intervention systems (FIPIs) focus attention on patients who are likely and expected to experience falls in the There are a wide range of falls prevention interventions and future (see Fig. 4, point #3). Primarily, the aim of many such sys- associated systems, which aim to overcome falls and the risk of tems is to detect falls when they occur and to prevent/minimise falling. Pre-falls prevention intervention systems (Pre-FPIs) are tech- the injuries that may occur after the event of falling. FIPIs, there- nology applications that focus on supporting patients who have fore, often aim to detect falls in order to prevent fall related injuries not yet experienced a fall, but may be considered to be at risk of rather than address the risks that lead to falls. There are three main falling (see Fig. 4, point #1). They take a pro-active approach via intervention types used to tackle these risks. Activity monitoring the development of applications, which support the delivery of tar- monitors patient movements obtrusively or unobtrusively whilst geted physical activities, exercises and education programmes that they perform ADLs and attempts to identify abnormalities, other- increase awareness of fall risks and help develop strategies to iden- wise not apparent. Fall detectors, attempt to distinguish fall events tify and overcome environmental fall hazards and the complica- from everyday activity signatures, so as to detect fall events when tions that may arise after having a fall. Cognitive training they occur. Medical assistance involves the provision of support programmes are also deployed to encourage older adults to engage provided by clinicians after a fall. in activities that stimulate their cognition, hence slowing down the Cross fall prevention intervention systems (CFPIs) are technology onset of age-related cognitive decline. Cognitive decline occurs as a applications which attempt to support and deliver a combination natural part of the ageing process and can impact on functional of pre-fall, post-fall and fall injury prevention interventions (see ability and therefore lead to increased risk of falls [68–70]. Fall risk Fig. 4, point #4). factors that Pre-FPIs aim to overcome, include intrinsic risk factors CFPIs propose technology applications which attempt to deliver that relate to natural ageing changes that affect older adults’ phys- system functionality across two or more groups of intervention ical ability, vision, balance, muscle strength and changes to their types i.e. Pre-FPI, Post-FPI and FIPI. An example of a CFPIs that cognition. Lack of mobility could also result in loss of muscle includes Post-FPIs and FIPIs is that of Shi et al. [73] who develop strength and balance impairments, leading to functional decline a smart-phone application which assesses fall risks using tradi- and resulting in a fall [71]. Extrinsic risk factors include factors that tional clinical tests and detects falls after they have occurred in are external to older adults’ physical health, functional ability and order to prevent fall-related injuries. Another example which com- cognition. These include, for example, environmental hazards that bines intervention types of Pre-FPIs and Post-FPIs is that of Silva are apparent within older adults’ home environment [72] such as et al. [37] who assess older adults for intrinsic risks and provide poor lighting, wet ﬂoor surfaces, loose rugs, slippery handrails, an exercise regime of dancing as a type of physical intervention and seating, toileting and bathing furniture, which is not optimally to enhance the uptake and adherence to exercising more often in set up or ﬁtted with suitable assistive equipment for an individ- the older adult population, particularly those who are prone to ual’s mobility needs or to carry out ADLs safely. falls, in an attempt to and reduce those intrinsic risks such as func- Post-fall prevention intervention systems (Post-FPIs) are applica- tional decline and a decline in muscle strength. tions of technology which focus on individuals who have already experienced a fall and aim to help assess and deliver interventions 3.2. Technology deployment to reduce the future risk of repeated falling episodes (see Fig. 4, point #2). The strategies employed by Pre-FPI and Post-FPI often The systems presented in the falls prevention domain host a share similarities, i.e. applications that support the delivery of range of application types and are deployed on a range of hardware exercise and education programmes with a view to overcoming platforms (see Fig. 4, point #5). Application type refers to the range shared intrinsic and extrinsic fall risk factors. However, the cohort of applications which are presented to support fall interventions. and motivation for delivery of these interventions may be some- Interactive applications allow the user to interact with the applica- what different in that Pre-FPI takes a pro-active approach and tion in some manner, whereas static offers no form of interaction Post-FPI supports the delivery of more re-active interventions. between the user and the system. For example, most fall preven- Thus, much of Post-FPIs initially involve fulﬁlling a diagnostic tion injury applications are static as their main purpose is to collect assessment function, whereby the cause of the fall, which triggered data and alert when a fall has occurred. Games are interactive appli- the post fall intervention, is identiﬁed along with other intrinsic cations that make up another group of falls prevention systems and extrinsic fall risks. There are a range of intervention types that which are typically played by patients with the goal of educating are used to carry out functional assessment and cognitive assessment and increasing awareness of fall risks, or to engage the user in exer- of post-fall patients to assess intrinsic risk factors. Functional cise and physical activity which is designed to improve mobility assessment involves screening the patients’ physical movement and hence reduce the risk of falling. Virtual reality (VR) applications for risk factors. As such, this includes older adults performing present simulated 3D interactive environments that allow the user intentional physical activities in order for a range of assessment to navigate through these environments and receive feedback in tests to be performed to gather fall risk behaviour data, which real-time based on multimodal user input. Physical activity inter- helps to determine the type of risk and the appropriate preventive ventions are also often augmented by VR applications to engage J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 325 users in physical activity and fall related physical exercise. With tems is natural user interfaces, which provide patients with a natu- regards to the platforms that falls prevention technology systems ralistic way of interacting with fall prevention systems. This are deployed upon; game consoles are self-contained platforms in typically requires users’ natural movements to be monitored and which speciﬁc game applications are utilised by falls prevention to serve as inputs, gathered via wearable or environmental sensors systems so as to deliver falls prevention related games. For inter- that are used to control fall prevention systems. This serves as an vention types such as physical activities, the game consoles and intuitive way of interacting with the system, particularly when sensor devices such as Nintendo Wii and Microsoft Kinect are often considering that fall prevention systems typically strive to allow used [74–77]. Desktop computers are another common platform users to engage in an unrestricted manner and monitor the user’s that systems are often deployed on. In recent years, smart-phones natural movements within their respective living environments. have shown promise as an ideal candidate for the deployment of Non-interactive interface is an invisible interface, which relies on falls prevention applications partly due to advanced processing intermediary sensor devices to source data from older users and capability, integrated sensors and communication facilities that to save that data to a centralised system, with no feedback pro- such devices now host. A tablet is a mobile touchscreen platform, vided or interaction with the end-users. The other common inter- which includes inertia measurement units, sensors (accelerometer, face used by fall prevention systems is a touchscreen interface, gyroscope, GPS), camera and touchscreen display (requiring touch which enables users to interact with fall prevention systems gestures to interact), replacing the traditional devices such as a deployed on smart-phones by providing touch gestures to touch keyboard and mouse. an object on the screen. This interface is an evolution of the periph- Information sources relate to the range of inputs that systems eral devices such as a keyboard and mouse that were used to inter- use to sense the users and the living environments they monitor act with objects on the screen. Although touchscreens are in order to provide falls prevention system functions. Sensor loca- inherently used for fall prevention systems as they are deployed tion speciﬁes where the sensors are located, either often as wear- on smart-phones, they are not part of sourcing of physiological able sensors on the user or within the context of the environment data from users, but rather a means to operate low level tasks. in which the falls prevention system is being used. With regards Users of the fall prevention systems consist of patients and practi- to context, this may be for example in the form of sensors tioners interacting with the systems. Patients who use fall preven- (camera-based and ﬂoor sensors) installed in the living environ- tion systems tend to be older adults, i.e. people over the age of ment which feed information back to the system about the user’s 65 years who experience advanced age changes, age related health interactions with that environment. Sensor purpose considers the decline, and age related declines in physical and functional abili- sensors used by falls prevention systems as belonging to one of ties. Practitioners are professionals (e.g. occupational therapists, three discrete groups: bespoke, repurposed and co-opted. Bespoke physiotherapists, nurses, carers, social workers, general practition- sensors are developed speciﬁcally for falls prevention systems, ers, accident and emergency staff) who deliver care to older adults which often gather physiological data from users. For example, in the hospital or community. Collaboration represents the means Uzor et al. [41] propose a small sensor which included a big switch by which practitioners work in partnership with patients to deliver to turn the power on and off, light emitting diode (LED) light to an intervention. Asynchronous collaboration relates to activities show power on and a velcro strap case to enable users to attach that are performed in real-time, however, the response to these the sensor to their body to interact with the falls prevention exer- activities do not occur in the time in which they occurred. For cise games. Repurposed sensors are sensors, which were originally example, in case where an older adults’ movement data is gathered developed for a different function, but have since been adapted for through the use of fall injury prevention interventions and if a fall use within the falls prevention context. For example, Kayama et al. event is detected an alert is sent to health care clinicians informing [76] utilise the Microsoft Kinect which was originally developed them of a fall. In this particular scenario, there is a time lag for gaming, however, due to the natural gesture-based interaction between the time of the fall event and the health response to a fall. paradigm this technology supports, the Kinect is repurposed to On the other hand, synchronous refers to when users’ movement provide the platform for an application that promotes the uptake data is gathered in real-time and the response of the movement of a gesture-sensitive falls prevention exercise game. Co-opted sen- data is also given in real-time in the form of visual feedback or sors are typically built into popular devices. For example, the biofeedback depending on the fall prevention systems that the accelerometer and gyroscope that is often built into self- patient is engaging with. For example, Reed-Jones et al. [79] utilise contained smart-phones. These may be used to obtain movement the Wii to improve balance and mobility in older people. The Wii data in order to perform falls prevention interventions, as with Fit game was used in conjunction with the Wii balance board, the study of Ferreira et al. [78], which exploit the smart-phone which served as an input device to source movement data from platform with the built-in sensors available (e.g. the gyroscope, older users to provide real-time visual feedback during game play accelerometer and magnetic sensors) to detect movement by in order to engage users and to better achieve precise body control attaching the smart-phone to the user’s body. Deployment environ- as part of the exercise training. ment reﬂects the range of living environments in which fall preven- In the following sections, the conceptual framework of falls tion technologies are typically designed to be deployed as speciﬁed prevention technology presented in this section is used to survey in the surveyed literature sample. There are three key deployment the systems that have been proposed in the literature. Section 4 environments which fall prevention systems are designed to be reviews pre-falls prevention intervention systems; Section 5 deployed within: the patient’s own home living environment; the reviews post-falls prevention intervention systems; Section 6 reviews hospital environment, typically for hospitalised patients; and falls injury prevention intervention systems; and Section 7 within the nursing home environment which may also take the reviews cross-prevention intervention systems. Table 1 provides form of an assisted living/sheltered housing environment, whereby a list of abbreviations and terms used throughout the review residential care is provided to older adults considered to be at risk sections. of falling. Interface type refers to the form of user interface that each respective falls prevention system provides to its users. Multimodal 4. Pre-fall prevention intervention systems interaction considers the mechanisms that enable users to interact with fall prevention systems, whether the user is the patient or the Pre-fall prevention intervention systems (pre-FPIs) focus on sup- practitioner. A common interface type used in fall prevention sys- porting the prevention of falls by targeting risk factors, which if 326 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 Table 1 on older adults’ balance by encouraging patients to engage in danc- List of abbreviations for terms in the framework. ing activities. Abbreviation Meaning The systems presented in [76,77,91,96,97,100] focus on allevi- ating functional ability deﬁcits and cognitive impairments (Fun Async Asynchronous + Cog), which are typically targeted via the use of game applica- Bal Balance impairments Bs Bespoke sensor tions. As such, cognitive impairments are considered to impact Co Co-opted on the patients’ functional ability. Some systems attempt to mea- Cog Cognitive impairments sure the extent to which cognition impacts upon functional ability. C Context For example, Pisan et al. [77] integrate cognitively demanding DC Desktop Computer Eh Environmental hazards tasks within a virtual environment, such as solving maths prob- Fun Functional ability deﬁcit(s) lems in a ‘‘simpliﬁed stroop test” whilst performing stepping exer- G Game cises within an immersive virtual environment. The aim is to GC Game Console measure the patient reaction time whilst stepping, in order to He Home environment uncover the severity of balance impairments whilst multitasking. Hs Hospital Nii Non-interactive interface Hilbe et al. [91] focus on patients in hospitals and nursing homes NUI Natural User Interface who are cognitively impaired. Patients are monitored to establish Rp Repurposed sensors whether they leave their beds and, if so, the clinicians are informed S Static so as to avoid falls in patients who are considered to be at high risk. Sm Smart-phone Sync Synchronous Kayama et al. [76] and Mirelman et al. [96,97] address the reduc- Ts Touch screen tion of the dual-task ability, cognition, and balance impairments U User-worn by executive function and delivering dual-task training as it is VR Virtual Reality believed that such activity improves cognitive function. Dual tasks include users engaging in problem solving tasks and performing tai chi exercises simultaneously within an immersive virtual reality present, are known to be the cause of falls. Table 2 provides a sum- environment. Finally, Schoene et al. [100] propose a game mary of Pre-FPIs considered in this literature survey and which deployed on a game console that includes stepping and balance make up the sole focus of this section. control tasks to improve reaction time in order to improve physical and cognitive abilities of community-dwelling older adults. 4.1. Fall risk factors The Pre-FPIs presented in [74,98] both focus on reducing extrin- sic risk factors, in addition to intrinsic risk factors. For example, Bell Pre-FPIs target fall risk factors that may be considered as a func- et al. [74] use a desktop-computer-based game and user-worn sen- tion of two distinct categories: intrinsic risk factors [18,19,40–42, sors to reduce impaired mobility via engaging users in exercise 75–97,99–109]; and extrinsic risk factors [74,98]. With regard to tasks and a gaming narrative which educates the player on envi- intrinsic risk factors, functional ability deﬁcits are the sole focus of ronmental fall risk factors such as clutter, placement of furniture, a number of studies [74,78,80,83,85,92,107]. In these examples, a and the dangers of spills on different types of ﬂooring. Otis and range of technologies is used to proactively mitigate observed def- Menelas [98] present a smart-phone application which is the only icits in functional ability. The study, for example, by Visvanathan system that focuses solely on reducing extrinsic risk factors. It con- et al. [107] monitors the physical activity of patients who are hos- siders the environmental conditions in which older adults function pitalised and considered to be at a high risk of falling as a result of and notiﬁes them of potential risks. The environment is scanned functional decline. This is achieved via the use of wearable sensors for slippery surfaces and steep slope by means of a smart shoe with and a sensor network that detects signs of potential risks as a result built-in sensors. of physically impaired patients moving around the hospital room without aid. De Morais and Wickstrom [85] develop a serious game 4.2. Intervention types based on tai chi, to help improve the stability of those who exhibit balance impairments and impaired mobility. Initially, older adults Intervention types used for preventing fall risks in are given a demonstration of pre-recorded tai chi activities at the [18,19,41,42,74–108] are typically administered either by practi- start of the game and are then required to mimic those movements tioners or self-administered by patients. Physical activities are during gameplay. intervention types targeted by [18,19,40–42,75,78–95,98,99,101– Functional ability deﬁcits and balance impairments are the sole 108], to mitigate these intrinsic risk factors. Studies [18,19,40– focus of many studies [18,19,40–42,75,78,79,81,82,84,86–90,92– 42,74–90,92–97,99–108] all explore the value of VR and gaming 95,99,101–106,108], which provide technology-based interven- technologies as a more interactive and engaging platform for tions to enable patients to retain their balance and improve patients to engage in exercise activity compared with more tradi- functional abilities in order that physical activities can be per- tional approaches. For example, Chao et al. [75] investigate the bar- formed safely within their normal living environments. For exam- riers that lead to a lack of adherence to falls rehabilitation exercises ple, Uzor et al. [41] and Williams et al. [42] use 3D visualisation and issues concerning older adults’ behaviour towards exercising. technologies and games to increase adherence rates and engage- Their resulting system included the application of the self- ment with home-based exercises with the aim of improving mus- efﬁcacy theory to enhance exercise behaviour to engage older cle strength and balance. Another example of this is provided by adults in physical activities to increase adherence rates of exercise Hardy et al. [90], who propose an exergame (i.e. exercise game) programmes. The system made use of the Wii which provided both to reduce balance and gait impairments, thus encouraging older visual and audio feedback based on users performance during the adults to exercise by providing a game that requires movements game to encourage users to exercise whilst still using the original similar to that of activities found in evidence-based exercise pro- idea and purpose of the game to entertain users. Silveira et al. grammes. Although many systems augment evidence-based exer- [101] explore the barriers to physical activities such as varying cises, some systems encourage users to engage in less structured adherence rates to exercise programmes, behaviour towards phys- exercise activities such as dancing. Lange et al. [93], for example, ical activities and lack of social company whilst exercising. The use an off-the-shelf game to help reduce impairments that impact proposed system is developed to speciﬁcally increase exercise J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 327 Table 2 Pre-fall prevention interventions. Pre-fall prevention system Pre-fall prevention interventions Fall risk factors Intervention types Systems Information sources Interface type Intrinsic Extrinsic Physical Cognitive Education Application Platform Sensor Sensor Deployment Multimodal Collaboration activities training type location purpose environment interaction Bailey and Buckley [80] Fun X DC U Bs He NUI Sync Bainbridge et al. [81] Fun X G GC C Rp NUI Sync Bell et al. [74] Fun Eh X X G GC U Bs Ns NUI Sync Bieryla et al. [82] Bal X VR + G GC C Rp He NUI Sync Chao et al. [75] Fun + Bal X VR + G GC C Rp Ns NUI Sync Chou et al. [83] Fun X S Sm U + C Co + Bs Nii + Ts Async de Bruin et al. [84] Fun + Bal X VR + G GC C Rp He NUI Sync De Morais and Wickstrom [85] Fun X G DC U Bs NUI Sync Doyle et al. [86] Fun + Bal X VR DC U Bs He NUI Sync Duclos et al. [87] Bal X VR + G GC C Rp NUI Sync Ferreira et al. [78] Fun + Bal X G Sm U Co He NUI + Ts Sync Geraedts et al. [19] Fun X VR Sm U Bs He NUI + Ts Sync Gerling et al. [88] Bal X G GC C Rp He + Ns NUI Sync Grifﬁn et al. [89] Fun X G GC C Rp NUI Sync Hardy et al. [90] Fun + Bal X G GC C Rp He NUI Sync Hilbe et al. [91] Fun + Cog S DC C Bs Hs + Ns Nii Aync Horta et al. [92] Fun X Sm U Co NUI + Ts Async Jorgensen [18] Fun + Bal X VR + G GC C Rp He NUI Sync Kayama et al. [76] Fun + Cog X X G DC C Bs NUI Sync Lange et al. [93] Bal X G DC C Bs He NUI Sync Majumder et al. [94] Fun X S Sm U Co Nii + Ts Async Ferrari et al. [95] Fun X S DC U Bs Hs Nii + Ts Async Mirelman et al. [96] Fun + Cog X X VR DC C Bs NUI Sync Mirelman et al. [97] Fun + Cog X X VR DC C Bs NUI Sync Otis and Menelas [98] Eh X S Sm U Co Nii Async Pisan et al. [77] Fun + Cog X X G DC C Bs NUI Sync Rajaratnam [99] Fun + Bal X VR + G GC C Rp He NUI Sync Reed-Jones et al. [79] Fun + Bal X VR + G GC C Rp He NUI Sync Schoene et al. [100] Fun + Cog X X G GC C Bs He NUI Sync Silveira et al. [101] Fun + Bal X G T Co He Ts Sync Singh [102] Fun + Bal X VR + G GC C Rp He NUI Sync Smith [103] Fun + Bal X VR + G GC C Rp He NUI Sync Sparrow et al. [104] Fun + Bal X S DC He Nii sync Taylor et al. [105] Bal X VR + G GC C Rp He NUI sync Taylor et al. [40] Fun X G GC C Bs NUI Sync Uzor et al. [41] Fun + Bal X VR + G DC U Bs He NUI Sync van Diest et al. [106] Fun + Bal X G DC C Rp He NUI Sync Visvanatha et al. [107] Fun X S DC U Bs Hs Nii Sync Williams et al. [42] Fun + Bal X G DC C Bs NUI Sync Young [108] Fun + Bal X VR + G GC C Rp NUI Sync 328 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 adherence rates and behaviour by involving users in social groups to suitability, the Wii game console is a relatively low-cost solution to stimulate participation with training regimes and integrating and has the capability to simulate an array of physical activities; the system into their daily routine. It also provides feedback on hence it has become a popular repurposed platform used in in-game performance and remote contact to supervise older adults attempting to overcome the issue of uptake of and adherence to during their exercise. The systems presented in [76,77,96,97,100] falls-related exercise interventions. Bainbridge et al. [81] examine use a combination of both cognitive training and physical activities the efﬁcacy of a Wii Fit game for reducing balance impairments intervention types to reduce fall risks. For example, Pisan et al. among community-dwelling older adults. Although the results in [77] present a balance training game that uses Microsoft Kinect this study suggest that the Wii Fit game program can be an effec- to enable older adults to interact with the proposed game. The tive intervention for clinicians to prescribe to patients, it also game involves a series of stepping exercises where squares appear reports that further research is needed to optimise its effectiveness randomly on the screen and the user is required to step on the and to better target the types of movement necessary to reduce fall squares as quickly as possible while solving basic arithmetical risks. The most common sensor devices used with the Nintendo problems. The results from this study revealed that performance Wii are colloquially referred to as ‘‘Wii-motes”, which are hand- during the stepping exercises decreases when participants engage held sensor devices with built-in infrared and accelerometer sen- with physical and cognitive tasks simultaneously, indicating that sors and are similar in size to a TV remote control. The Wii users could potentially be at high risk of falling when multitasking. balance board, with pressure sensors, is often used to monitor Schoene et al. [100] use exergames to address the issue of lack of and assess patients balance. Williams et al. [42], Bell et al. [74] adherence to exercise programmes in light of improving older and Schoene et al. [100] use the Wii balance board with the Wii adults’ balance, stepping ability, cognition and other factors associ- Fit game to assess its feasibility for improving the balance of older ated with falling. This exergame consists of a dancing gameplay, adults who had fallen previously, based on clinically established which provide instructions to perform dance moves using a step balance assessment tools such as the Berg Balance Score (BBS), pad, with the aim of train balance, reaction and attention. Educa- Tinetti Test, Falls Efﬁcacy Scale – International (FES-I), and Timed tion and physical activities are intervention types in [74] which Up and Go Test (TUG). Pre-FPI systems presented in [19,86,96,97] are used to reduce both intrinsic and extrinsic risk factors of falling. are VR applications. The use of the Wii balance board device Bell et al. [74], for example, investigate the beneﬁts of utilising the appears to reduce the fall-related risks based on the outcome mea- Nintendo Wii game console for preventing falls in assisted-living sures of balance and functional ability, as reported in the studies. environments. Participants in this study engaged in exercise train- However other systems, speciﬁcally [18,41,75,79,82,84,87,99,102, ing with the use of the Wii combined with falls prevention educa- 103,105,108], are all interactive virtual reality and game applica- tion sessions. The fall prevention education sessions focus tions which typically provide the user with a means of interacting particularly on reducing clutter, arrangement of furniture in the with the application by the system responding to the user’s phys- living area, positioning of the rug, ﬂooring and spills within the ical state, where aspects of the system are manipulated by their home environment, lighting, and staircase and bathroom safety, movement. presented in the format of checklists. A number of Pre-FPIs [19,78,83,92,94,98,101] are deployed on smart-phone platforms. As a result of advancements in smart- 4.3. Systems phones, they are an ideal technology for tackling an issue like fall prevention, as data can be obtained via built-in sensors. An exam- Pre-FPIs take the form of a range of application types and are ple of the use of smart-phones is that of Ferreira et al. [78] who deployed on a range of platforms. The application types presented propose a smart-phone-based falls prevention system operating in [83,91,94,95,98,104] are all static; they are essentially data col- based on user movement which was then translated to movements lection tools which issue an alert to notify users of potential fall performed for exercises in a serious game application. The main risks as a consequence of abnormal walking/behavioural patterns, purpose of the study is to increase adherence of older adults which are collected from sensors. For example, Majumder et al. exercising within their home. Majumder et al. [94] propose a fall [94] propose a system which includes a feature extraction tech- prevention system for identifying abnormal gait patterns in real- nique to conduct an analysis of walking patterns collected in time to predict an imminent fall and prevent it from occurring real-time to determine whether there is a potential risk of the user by notifying the user on the likelihood of a fall occurring. This falling. This system does not involve any notable form of interac- system was deployed on a smart-phone and used the embedded tion, as it simply analyses and sends alerts based on the data that sensors. Horta et al. [92] and Majumder et al. [94] propose a is collected from patients. Otis and Menelas [98] develop a proto- smart-phone-based solution to obtain movement data from older type of an instrumented shoe with embedded sensors actuators adults in real-time to inform users of abnormal walking pattern which are positioned in certain parts of the shoe to collect data, behaviour identiﬁed by the system, thus helping to avoid the categorise the fall risk status of the environment, and then broad- occurrence of falls. This data is also shared with other stakeholders, cast this in real-time to a smart-phone application. Horta et al. [92] such as clinicians or carers. The remaining system [101] is propose a smart-phone-based system using built-in sensors to col- deployed on a tablet. Silveira et al. [101] develop a tablet-based lect physiological data from older adults to inform them of any exercise intervention system as it provides a touchscreen display abnormal behaviour in their walking pattern. Chou et al. [83] rather than keyboard and mouse and is reported to be more intu- detect the position of patients from that of lying-to-sit and alert itive in providing feedback based on in-application performance. the user with a warning that there is a high risk of falling while get- The Pre-FPIs presented in [18,40,75,79,81,82,84,86–90,99,100,102, ting out of bed. Once the transitions of the patients have been 103,105,108,109] are repurposed game consoles. Bell et al. [74] and detected, a notiﬁcation is sent to clinicians in order to provide care Lange et al. [93] investigate the utility of the Wii game console for and prevent bedside falls. preventing falls, particularly to educate older adults on exercise All of the game applications presented in [40,42,74,76–78,80,81, training and the environmental hazards that often contribute to 85,88–90,93,100,101] make use of the Wii games console to detect falls. There are also Pre-FPIs [41,42,74,76,77,80,85,91,93,96,97,104, user movements in real-time and enable users to interact with 106,107] that are deployed on desktop computers. This is exempli- games and control in-game avatars. These studies explore the ﬁed in the study conducted by Uzor et al. [41] who develop both a effects of such an interaction paradigm and evaluate its suitability game and VR application for desktop computer platform using to the fall and the prevention intervention domain. With regards bespoke sensors to control the system. J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 329 4.4. Information sources from older adults in real-time. Otis and Menelas [98] propose a smartshoe to track the movement of patients and collect informa- The Pre-FPIs presented in [18,19,40–42,74–108] all use informa- tion from smart-phone sensors to the developed application. tion sources to enable the patient and/or practitioner to interact Smart-phones are considered an ideal tool for falls prevention with the systems in some manner. There are, however, differences due to their self-containing nature, size, portability and that they in the information sources and the way in which information is can also be used to communicate with other sensors making the sourced from the user of the system. Sensor location comprises of applications more wide-reaching. Finally, Chou et al. [83] develop two distinct categories, namely, are context and user. Context sen- a system to detect the position of patients from lying to sit and sors are the main devices used in [18,40,42,75–77,79,81,82,84, alert the user with a warning that there is a high risk of falling 87–91,93,96,97,99–106,108] to source information from patients while getting out of bed. Once the transitions of the patients have unobtrusively, without the need for users to wear a device to inter- been detected, a notiﬁcation is sent to alert clinicians in order to act with VR or game applications. Kayama et al. [76] utilise Micro- provide care and prevent a bedside fall. soft Kinect as an input device, deployed in the environment, to enable older adults to interact with a game application. Taylor 4.5. Interface types et al. [40] utilise the Nintendo balance board as an input device where the user stands on the board to interact with the game. Natural user interfaces [18,40–42,74–77,79–82,84–90,93,96,97, Hardy et al. [90] and Grifﬁn et al. [89] utilise Nintendo balance 99,100,102,103,105,106,108] enable users to interface with sys- board to improve balance by controlling in-game avatar and to tems when performing physical activities during game-play and move virtual objects in order to achieve the game objective and collect ambulatory/behavioural data from users unobtrusively. to physically engage the patient as part of an intervention. Finally, Mirelman et al. [96] augment treadmill exercise training with VR Mirelman et al. [96] and Mirelman et al. [97] use pressure on the technology to improve functional ability and cognitive function, treadmill to capture physical movement of older adults performing thereby reducing falls. Users perform exercises on the treadmill; physical activities. On the other hand, user-worn sensors in [19, those movements are then translated into inputs in a virtual envi- 41,74,78,80,85,92,94,95,98,107] require users to wear them in ronment which present users with obstacles, as well as other chal- order to obtain the movement and translate that motion to control lenges, that they have to overcome. Feedback (visual and auditory) the system for clinical use. For example, Uzor et al. [41] built a is presented to users based on errors that are made and tasks suc- bespoke sensor device that was used to enable patients to control cessfully completed. Systems presented in [19,78] use touchscreens the game and was considered less intrusive than other devices and natural user interfaces, which are a specialised way of interact- such as Microsoft Kinect and Wii Remote, and also ideal due to ing with technology-based interventions to reduce fall risks. its size to attach it to speciﬁc parts of the body to capture the Although this type of interaction does not involve nor measure movement. Bailey and Buckley [80] utilise bespoke sensors to col- any physiological parameters, it enables touch input in order to lect data from older adults performing ADLs as an attempt to operate some systems. It is a required action to interact with some understand the cause of falls. systems. Ferreira et al. [78] propose a falls prevention game that Sensor purpose refers to the type of sensing devices used to cap- use embedded sensors on smart-phone to enable users to interact ture data from users and consists of bespoke, repurposed and co- with the serious game application via the use of the built-in touch- opted sensors. Bespoke sensors [19,40–42,74,76,77,80,85,86,91,93, screen. Non-interactive interfaces [91,98,104,107] enable interven- 95–97,100] are custom-built sensors developed speciﬁcally for fall tions to be administered without an interactive interface to prevention and deployed within the living environment or worn by engage users. For example, Sparrow et al. [104] propose an auto- older adults. For example, Hilbe et al. [91] propose a ‘‘Bed-exit” mated home-based exercise programme that provide voice alarm used to reduce bedside falls. The pressure sensors were response for real-time guidance whilst older adults performed designed and integrated on the side rails of the patient’s bed to their exercises. The programme is administered over the telephone track their attempt to get out of bed. The side rail is in a certain with no interactive form of feedback or interface present to guide position so that if pressure is detected from the pressure sensor, or engage users in a way that feedback is given of their perfor- with the value exceeding the threshold, an alarm is sent to clini- mance during exercises. The remaining systems [83,94,95] use cians (e.g. nurses) in order to prevent a fall from occurring. both non-interactive interface and touchscreens for Pre-FPI systems Williams et al. [42] use the Wii balance board as an input device to perform fall prevention activities, such as gathering of data via with the Wii Fit game and balance assessment tools to improve built-in sensors and to use the platforms touchscreen to initiate the balance of older adults who are vulnerable to fall risks. Com- the activities or to visualise analysis of the data that prevent fall mercially available repurposed sensors are used to interact with risks. Chou et al. [83], for example, use sensors integrated into falls prevention exercise games. For example, Pisan et al. [77] the patient’s bed to detect when an attempt is made to leave the and Kayama et al. [76] utilise Microsoft Kinect with a game devel- bed without aid. This system does not require any form of interac- oped for older adults at risk of falling. The game measures changes tion, as it is a monitoring tool for clinicians to prevent hospitalised to patients’ functional and cognitive abilities by carrying out patients from attempting to leave the bed. Once the alarm is trig- physical and cognitive tasks simultaneously, as reduction in gered, the system on the smart-phone receives the alarm signal multi-tasking is known to be a predictive factor of a risk of falling. and clinicians are notiﬁed by a text message alert, which gives In particular, using Kinect is ideal as it is a cost-effective means of details of data received from the bed sensors, such as codes that obtaining data from patients unobtrusively without the need to indicate posture position. wear or to control handheld devices. In terms of collaboration, the systems presented in [18,19,40–42, Co-opted smart-phone sensors are now enabling applications 74–82,84–90,93,96,97,99–108] enable synchronous collaboration such as fall prevention, detection and monitoring patients and engagement between patients and clinicians via a range of [110,111]. The pre-fall prevention systems presented in interface types. In the study by Marisa Ferrari et al. [95], clinicians [19,78,83,92,94,95,98,101] made use of built-in sensors on smart- supervise participants in an exercise training with the use of the phones, which lend themselves well to tracking user movement Nintendo Wii in a nursing home. Users were provided with imme- in order to achieve outcomes of fall interventions. An example of diate feedback of their in-game performance to improve their func- a smart-phone application is that of Horta et al. [92] who use tional ability and balance. Although it is not made clear if patients built-in sensors on smart-phones to capture physiological data were involved in the decisions made in this intervention, the fact 330 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 that both practitioners and patients are engaging in the intervention hence the need to further explore their use to tackle both at the same time provides an opportunity for patients to be seen as intrinsic and extrinsic risk factors. The advancements of more equal partners in their own care. Conversely, the remaining smart-phones have increasingly becomeaportabledevicewith studies [83,91,92,94,95,98] are considered as asynchronous in that unprecedented computational power similar to that of desktop response from sourced movement data does not occur in real- computers. time. The studies of Hilbe et al. [91], Majumder et al. [94] and Mar- One issue that stands out is the extent to which these systems isa Ferrari et al. [95] monitor older adults physical activities in an allow patients and practitioners to collaborate. Whilst some of attempt to predict the likelihood of falling. Data such as abnormal- Pre-FPIs support asynchronous collaboration (i.e. not real-time ities in walking patterns and critical patients leaving their bed are collaboration) [19,78,79,82,92,94,98–100,102,104], the remaining sourced from patients to prevent falls. As these systems monitor studies [18,40–42,74–77,80,81,83–85,89–91,93,95–97,101,107] to improve health outcomes such as reduced fall risks, clinicians are synchronous, thereby giving an opportunity for patient and only intervene when the data collected suggests that the patient practitioner to collaborate as they are in an environment during is at high risk of falling. No feedback is provided, as the purpose the intervention. Given the long-term goal of health care delivery, of these systems is simply to unobtrusively collect data that reﬂects particularly within the UK, to increase the extent to which patients ADLs, rather than perform activities to improve functional ability are more equal partners in the delivery of care [47–49,113],it and balance to undertake ADLs. seems many Pre-FPIs are creating opportunities to support patient engagement in care and the decision making that is required whilst 4.6. Discussion providing this care. Pre-FPIs provide a useful way of preventing the onset of risks 5. Post-fall prevention intervention systems and treating fall risks using intervention types to reduce: func- tional ability deﬁcits [40,74,80,81,85,89,92,94,95,107]; functional Post-fall prevention intervention systems (Post-FPIs) are typically ability deﬁcits and balance impairments [18,41,42,75,78,79,82,84, used in the ﬁrst instance to screen patients for fall risks after they 86–88,90,93,99,101–106,108]; and functional ability deﬁcits and have experienced a fall. Fall assessments are traditionally con- cognitive impairments [76,77,91,96,97,100]. However, limited ducted within controlled environments, such as within a spe- attention is given to reducing both functional ability deﬁcits cialised falls clinic environment or alternatively within and extrinsic risk factors [74] or focusing solely on reducing extrin- uncontrolled environments. The latter are often carried out as lon- sic risk factors [98]. A considerable number of Pre-FPIs have gitudinal assessments, where older adults are remotely monitored focused their efforts on alleviating intrinsic risks, with limited over a period of time in order to identify activity signatures that effort invested into providing support for overcoming extrinsic fall correspond to fall risks. Although Pre-FPIs have yielded many ben- risks and the process of provision of assistive equipment in order to eﬁts for promoting health promotion activities to prevent the onset mitigate some of these extrinsic fall risk factors. This is despite the of fall risks, similar fall risk factors are diagnosed and treated via provision of specialist assistive equipment being one of the key Post-FPIs. Table 3 provides a summary of Post-FPIs and their interventions used to mitigate fall risks associated with functional respective characteristics. decline. The consensus of the fall technology literature reviewed indi- cates the increasing popularity and reusability of VR and game 5.1. Fall risk factors applications which aim to address the limitations of clinical inter- ventions, particularly adherence and uptake issues. Based on Post-FPIs and technologies that focus on preventing falls by results of studies presented in [40,42,74,76–78,80,81,85,88–90, screening and assessing for fall risks may also be considered as a 93,100,101], games are often proposed as an adjunct to traditional function of two fall risk factor categories: intrinsic risk factors interventions and are not typically designed to replace existing [39,114–130]; and extrinsic risk factors [38]. With regards to intrin- interventions. It seems that users are motivated by the use of exer- sic risk factors, functional ability deﬁcits are the sole focus of assess- cise games as they can provide feedback on performance, thus cre- ment in [114–119,121–123,125–130]. Majumder et al. [119] detect ating a more stimulating and entertaining experience. Employing abnormalities in gait patterns, which are considered to be a com- such technology reduces travel costs for older adults who travel mon cause of falling in the older adult population. Users are noti- to rehabilitation centres [112] and increases patient motivation ﬁed of the likelihood of falling based on data collected and to engage with proposed falls prevention intervention pro- classiﬁed to determine whether or not the patterns of ADLs are grammes. From the corpus of research reviewed [18,19,40–42, abnormal. Redmond et al. [121] provide an unsupervised continu- 74–108], it seems that there are limited research efforts that utilise ous fall risk assessment for older adults who live independently VR technology and games to augment fall education interventions with a high risk of falling and who have been selected for clinical aimed at reducing extrinsic fall risks with the exception of [74,98], intervention. Staranowicz et al. [127] and Weiss et al. [128] present which also lack the use of such technology. systems which monitor older adults’ gait in real-time from ADLs Surprisingly, given the ever-increasing ubiquity of smart- and collate the motion data in order to predict falls so that older phones with patients and clinicians, a relatively small number of adults can receive early intervention. Zijlstra et al. [130] and Pre-FPI are deployed on smart-phones [19,78,83,92,94,98,101] Greene et al. [118] present approaches to monitor and assess fall compared to the majority, which are on desktop computers [41,42, risks for older adults performing clinical tests which emulate ADLs. 74,76,77,80,85,86,91,93,96,97,104,106,107] or repurposed game Riva et al. [123] and Soaz and Daumer [126] analyse gait patterns consoles [18,40,75,79,81,82,84,86–90,99,100,102,103,105,108, to determine the association between features extracted from gait 109]. A possible explanation for this may be that game consoles patterns with a history of falls in order to target older adults who more naturally possess the requirements and functionality that are in need of clinical interventions. Almer et al. [114] develop a can be more readily repurposed for the function of developing framework which was evaluated with a series of assessment tests rehabilitation exercise intervention applications. Nonetheless, (2-Minute Walk, Sit-to-Stand 5 and Timed Up and Go) by recoding smart-phones have shown promise in deploying Pre-FPIs, espe- movement data and using feature extraction techniques to deter- cially to help capture physiological data [18,19,40–42,74–79,8 mine fall risks in the movement data. Cuddihy et al. [117] monitor 1–100,102,103,105–108], but also in a much broader sphere, gait in older adults and notiﬁes caregivers of any changes as they J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 331 Table 3 Post-fall prevention interventions. Post-fall prevention Post-fall prevention interventions system Fall risk factors Intervention types Systems Information sources Interface type Intrinsic Extrinsic Functional Cognitive Environmental Application Platform Sensor Sensor Deployment Multimodal Collaboration assessment assessment assessment type location purpose environment interaction Almer et al. [114] Fun X S Sm U Co He + Hs Nii + Ts Async Barelle and D. Fun X S DC C BS He Nii Async Koutsours [115] Brell et al. [116] Fun X S DC C BS He Nii Async Cuddihy et al. [117] Fun X S DC C Rp He Nii Async Du et al. [38] EH X S DC C BS He Nii Async Garcia et al. [39] Fun X X VR DC C Rp – Nii Sync + Cog Greene et al. [118] Fun X S DC U BS He + Hs Nii Async Majumder et al. [119] Fun X S Sm U Co He Nii + Ts Async Rawashdeh et al. [120] Fun X X VR DC U BS Hs NUI Async + Cog Redmond et al. [121] Fun X S DC U BS He Nii Async Regterschot et al. [122] Fun X S DC U BS He Nii Async Riva et al. [123] Fun X S DC U BS . Nii Async Schoene et al. [124] Fun X X G DC C Rp He + Hs NUI Sync + Cog Singh et al. [125] Fun X G DC C He NUI Sync Soaz and Daumer [126] Fun X S DC U BS He + Hs Nii Async Staranowicz et al. [127] Fun X S Sm C Co He + Hs Nii + Ts Async Weiss et al. [128] Fun X S Sm U Co He Nii + Ts Async Zhang et al. [129] Fun X S DC U BS He Nii Async Zijlstra et al. [130] Fun X S DC U BS He Nii Async 332 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 also carry out assessments remotely and take preventive measures presented in [39,120,124,125] are purpose-built or repurposed to to alleviate fall risks. Singh et al. [125] and Barelle et al. [115] pre- screen older adults for intrinsic risks through game-play or simula- sent an approach in extracting gait features from the walking pat- tion. Singh et al. [125] provide a balance game using the Nintendo terns of older adults’ to perform early diagnosis of functional Wii balance board to measure agility and balance in older adults. decline for a more accurate estimation of fall risks. Post-FPIs pre- This system can also be used to reduce balance impairments as it sented in [39,120,124] address both functional ability deﬁcits and empowers older adults to train frequently by providing visual feed- cognitive impairments via the use of dual tasking. Schoene et al. back based on movement performed, converting real-life move- [124] introduce a device that serves as a proxy to measure older ment into virtual movement as part of an in-game narrative. The adults with severe cognitive and physical impairments. Reaction majority of the systems presented in [38,39,115–118,120–126, time of stepping ability is used to predict potential risk of falls. 129,130] are deployed a on desktop computer platform, whilst The remaining Post-FPIs [38] focuses solely on the extrinsic risk fac- the remaining systems in [114,119,127,128] are deployed on a tors. Du et al. [38] develop a robot to screen older adults’ living smart-phone platform. environment for typical environmental hazards such as, to name a few, poor lighting, unstable furniture, lack of equipment in the 5.4. Information sources bathroom and then provides that information to clinicians. The Post-FPIs presented in [38,39,114–130] exploit a range of 5.2. Intervention types information sources to gather data from patients. In particular, users one of the key sensor locations in [114,118–123,126,128–130] to Post-FPIs intervention types consist of functional assessment, cog- gather information relating to the users’ physical movement and nitive assessment and environmental assessment. In particular, to reduce the progression of fall risks. For example, Weiss et al. functional assessment is the main intervention type presented in [128] use a bespoke wearable sensor to collect long-term gait pat- [114–119,121–123,125–130] in order to determine intrinsic risk terns from older adults performing their ADL routine in order to factors such as functional ability deﬁcits. Majumder et al. [119] capture properties and characteristics of fall risks in a real-life set- develop a smart-phone-based fall assessment system to monitor ting and complement conventional performance-based tests. Pro- abnormal gait patterns from older adults performing physical viding such a solution not only enables fall risks to be assessed activities that are constituted as ADLs. The gait patterns are col- remotely, but it also uncovers useful information regarding the lected from users over a period of time from walking and carrying quality and quantity of ambulation performed by older adults in out ADLs. Staranowicz et al. [127] propose a system which moni- the home. Conversely, context is exploited in [38,39,115–117,124, tors the walking patterns of older adults’ during their ADLs at 125,127] to source information unobtrusively from patients. Bare- home and identiﬁes functional decline via the use of an autono- lle et al. [115] develop an ICT-based home care system to monitor mous robot. The systems proposed in [39,120,124] use cognitive patients with gait impairments in order to diagnose early potential assessment and functional assessment to assess functional ability fall risks before a fall occurs and to respond with appropriate inter- deﬁcits, balance and cognitive impairments. Patients are encour- ventions. The system enables independent living at home by use of aged to conduct physical activities and cognitively demanding biomechanics data and indicators of gait impairments recorded in tasks to determine fall risks. Garcia et al. [39], for example, present a schedule agreed with medical staff and patients based on health a Kinect-based system to gather timing of movement to measure status and ADLs. the reaction time of stepping ability tasks. This is referred to in this Sensor purpose consists of sensors tailored for reducing falls or study as choice stepping reaction time (CSRT) and it is used to pre- built-in sensors repurposed for technology-based interventions. dict falls. As such, this approach measures physical abilities includ- In particular, systems presented in [114,119,127,128] use ing strength and balance, and cognitive abilities such as attention co-opted smart-phone sensors as wearable devices to collect data and speed of processing. The remaining system presented in [38] from users. For example, Staranowicz et al. [127] and Weiss et al. conducts environmental assessment for fall risks. Du et al. [38] [128] use accelerometer and gyroscope sensors on smart-phones develop a robotic system that screens the patient’s home, operated to assess users’ gait patterns for any abnormalities to notify users remotely by clinicians, with the robot being navigated around the of potential falls. Users are not required to wear the smart-phone home whilst checking for fall hazards. Essentially, this system on any particular part of their body, however, it has to be on their automates home assessments that are typically conducted by person as the built-in sensors gather acceleration and movement clinicians. data while users are walking. Almer et al. [114] present a smart- phone-based falls assessment application. The application collects 5.3. Systems data from built-in sensors such as accelerometer and gyroscope; it was evaluated using clinical assessment tests: the ‘‘2-Minute Post-FPIs presented in [38,114–119,121–123,126–130] are all Walk”, ‘‘Sit-to-Stand” and ‘‘Timed Up and Go” (TUG). Bespoke sen- static systems, i.e. they do not provide the user with an interactive sors are used in [38,115,116,118,120–123,126,129,130] to identify interface. Cuddihy et al. [117] propose a static system that mea- risk factors. For example, Greene et al. [118] propose bespoke body sures gait based on ADLs performed by older adults. The system worn sensors to gather older adults movement data. The sensors requires no form of interaction as its main function is to unobtru- are attached to the older adults’ body whilst performing the berg sively collect data in order to analyse gait patterns. Riva et al. [123] balance scale (BBS) and the TUG tests. PostFPIs presented in Soaz and Daumer [126] and Greene et al. [131] use wearable sen- [39,117,124] use repurposed sensors. For example, Singh et al. sors, with no interface, to assess users’ physical activities and bal- [125] use the Wii balance board as an input device to interact with ance to predict the potential risk of falling. Robinovitch et al. [132] a balance game to assess older adults balance. use video cameras to record footage to identify falls or the environ- Deployment environment refers to the range of environments in mental and behavioural factors that lead to falls. Almer et al. [114] which Post-FPIs are deployed within, namely; the home environ- and Majumder et al. [119] develop static applications to collect ment, hospital setting or nursing home setting. The systems pre- fall-like data from users to assess the data for fall-related beha- sented in [114,118,124,126,127] are deployed within the home viour. The applications used smart-phone sensors to extract fea- environment and hospitals. The Post-FPIs presented in [38,115–117, tures that are fall-like behaviour; however, the interface on the 119,121,122,125,128–130] are all deployed solely within the home smart-phone was not used. Conversely, game and VR applications environment. Brell et al. [116] conduct clinical tests using a robot J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 333 to collect data from patients performing ADLs in their home in et al. [120] develop a system that senses patients’ movement and order to diagnose fall risks. Other systems are deployed solely posture data from sensors attached to different parts of the body within the hospital setting [39,120]. Rawashdeh et al. [120], for (wrist, ankle and chest). The data is sent to a base station in which example, propose a virtual 3D avatar system which reﬂects it is processed in real-time. The processed data is used to animate a the movement of hospitalised patients that are prone to falls. There 3D avatar that mirrors patients movement. Clinicians respond to are no Post-FPIs that are deployed solely within the nursing home abnormalities on the 3D avatar that indicate that patients are at environment. risk of falling. However, no data is directly fed back to patients as its sole purpose is to monitor movement. 5.5. Interface type 5.6. Discussion Post-FPIs use natural user interfaces [120,124,125] to enable the Post-FPIs assess patients for intrinsic and extrinsic fall risks user to interact with the system. Schoene et al. [124] propose a using physical, cognitive and environmental assessment interven- game which requires natural interactions such as foot movements tions. After reviewing these systems, the following observations to interact with the proposed game application. Singh et al. [125] are drawn with regards to assessing fall risks. The majority of the provide an interactive interface to engage older adults to exercise post-fall prevention systems assesses intrinsic risk factors such independently and frequently without therapist intervention. The as functional ability deﬁcits [114–119,121–123,125–130]; func- interface provides visual feedback during the intervention to tional ability deﬁcits and cognitive impairments [39,120,124], with enhance compliance to exercise more often and real-time feedback limited attention given to extrinsic risk factors [38] which can also with regards to users’ ability to maintain their balance. Post-FPI result in serious fall injuries. Post-FPIs use a range of intervention systems presented in [38,39,115–118,121–123,126,129,130] are types to assess fall risks. Functional assessments [114–119,121– non-interactive interface which provide a one-directional ﬂow of 123,125–130] were solely used to assess functional ability to data, without presenting feedback of the sourced data to patients. determine risks of falling, whereas [39,120,124] use both func- Regterschot et al. [122] use sensors to identify changes in mobility tional assessments and cognitive assessments to assess multifactor and fall risks to perform clinical tests, without providing an inter- risks. active medium to engage users in fall prevention interventions. While Post-FPI systems play a crucial role in reducing the risk of Zhang et al. [129] use a pendant-worn sensor to detect chair trans- falling, particularly assessing fall risks, few systems address extrin- fers in order to unobtrusively assess fall risks in a non-interactive sic factors [38]. In fact, only one system, [38] has focused on assess- way. Older adults wear the pendant around their neck like a neck- ing the home environment for extrinsic risks. This system involves lace for continuous monitoring. ADLs, particularly chair transfers, a robot to assess the patient’s home. A clinician is able to operate performed by older adults are the system’s sole input to conduct the system remotely by navigating the robot around the home fall risk assessments. A system’s interface is often driven by the whilst going through a checklist of factors. Despite the apparent platform it is deployed on. Systems presented in beneﬁts, the system is not fully autonomous, which makes it prone [114,119,127,128], for example, use a combination of non- to handling errors that can affect its reliability, in addition to still interactive interface and touchscreen built into smart-phones as needing clinicians time to conduct the assessment tasks remotely. the sensors collect data from patients, who use the touchscreen The consensus of the Post-FIP systems indicates that majority of to activate in-system functions. Almer et al. [114] develop a systems are static in nature and offer no means for users to interact smart-phone application with non-interactive interface to conduct with the systems [38,39,114–119,121–123,126–130]. Therefore, it assessment tests by obtaining motion data from patients using appears, that limited efforts are spent on systems which provide an built-in sensors such as accelerometer and gyroscope, and touch- interactive means to assess fall risks [120,124,125]. Another screen for users to authenticate into the system and to display challenge yet to be explored in this domain is the patient–clinician the system’s status and users information. The developed iOS collaboration. The majority of systems provide synchronous collab- application requires little interaction as it displays the current user oration [38,39,114,115,117–120,122,123,125–130] where data is and information about the device. The main engine, running the sourced from older adults and the response is provided in real- assessment tests, is deployed in the background, indicated by a time. The rest of the studies reviewed here present systems that green light, whilst the tests being performed are also displayed. are asynchronous [116,121,124], meaning that data sourced from The systems presented in [39,124,125] enable synchronous collabo- older adults is not clinically assessed at the time it was performed. ration between patients and practitioners. Garcia et al. [39] and Singh et al. [125] carry out assessments through patients perform- ing physical activities. Patients are presented with real-time feed- 6. Fall injury prevention intervention systems back of balance scores and progress made during the assessment programmes. This real-time feedback component improves com- Fall injury prevention intervention systems (FIPIs) aim to detect pliance for patients to regularly assess for fall risks. The remaining and respond to falls after they have occurred and prevent or min- systems [38,114–123,126–130] provide asynchronous collabora- imise fall related injuries that may occur as a consequence of falling. tion. Zijlstra et al. [130] monitor patients with mobility issues Unlike Pre-FPIs and Post-FPIs, they do not typically focus on over- and fall risks whilst they perform sit-to-stand movements during coming the intrinsic/extrinsic risk factors that may lead to a fall chair transfers. Data is sourced from patient to provide a longitudi- occurring, but rather focus on responding to a fall after it has nal proﬁle of changes to ambulation and mobility issues during occurred. These systems typically aim to monitor patient activity transfers in order to determine potential fall risks. No interface is with the goal of providing a channel of communication between presented to users, however, sensors are located on the torso of older adults and clinicians. There are three main intervention types the patients and in and around their home furniture in order to that these systems target. Activity monitoring involves monitoring measure power exertion and movement. Collecting longitudinal patient movements either obtrusively or unobtrusively while they datasets such as [116] enables patients’ movement data to be con- perform ADLs to identify abnormalities in patient daily occupa- sidered over a period of time. Conducting assessments in patients’ tions. Fall detectors monitor patient activity in order to identify homes enables both parties to collaborate to some extent, as the discrete occurrence of a fall. In the event of abnormalities or patients are assessed remotely by clinicians without both parties the occurrence of a fall, clinicians can be alerted via an alert for having to physically be in the same environment. Rawashdeh medical assistance after a fall has occurred [63]. Many systems have 334 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 been proposed over the years and categorised based on their bounding box which calculates the velocity of width, height and sensors, underlying algorithms and computational techniques used depth in order to establish if the activity performed by users is a to detect this phenomenon. fall or an ADL. An alarm is sent to clinicians in order to provide Table 4 summarises fall injury prevention systems proposed in immediate assistance to fallers. the falls technology domain. Mehner et al. [156] detect falls automatically in order to provide rapid medical support so that fall related injuries will be reduced 6.1. Fall risk factors and to physically help older adults off the ground, particularly in cases where older adults knock their head and are unconscious All FIPIs presented in [43–45,133–171] focus on minimising fall or not able to seek help. He et al. [145] classify data sourced from related injuries that may occur as a result of experiencing a fall. For monitoring older adult activity. The motion is then split into ﬁve example, Abbate et al. [43] propose a system to monitor patient sub-patterns which include vertical and horizontal motion, lying, movement automatically using an artiﬁcial neural network and sitting, standing, and falls in order to accurately detect falls by feature extraction machine learning technique which alerts emer- employing a feature extraction technique. Once a fall is detected, gency services and other preloaded emergency contacts after a fall an automatic multimedia messaging service (MMS) is sent to pre- has been detected. Mastorakis and Makris [155] present a system loaded contacts which include the patient’s GPS coordinates and using an algorithm which uses a large training dataset which auto-generated image of a Google map pinpointing the location includes falls data and a range ADLs in order to more accurately of the fall. Cabestany et al. [138] automatically detect falls both detect falls when they occur and avoid false positives. Ferrari inside and outside the living environment and then sends an alert et al. [143] monitor and track patient movement in hospital, if a fall to a call centre, informing them that a fall has occurred. Shim et al. is detected, the system automatically sends an alarm to clinicians [164] monitor patients in bed in order to detect bed-side falls in an when patients attempt to leave their beds without aid or displays assisted living environment. If patient movement is detected an increase in activity levels. Transmitting this type of data to around the bed, it is then necessary to consider it as a potential fall nurses in the hospital enables them to provide care and assistance and a caregiver is then notiﬁed. Lee and Carlisle [151] provide a when it is needed. Zhang et al. [169] put forward a system that mechanism that sources data about the activities of older adults; unobtrusively detect falls that occur at night-time where the older if a fall occurs, it is then reported to emergency services. Terroso adult is unconscious and hence may ﬁnd it difﬁcult to move with- et al. [166] detect a fall, either in or out of the home and send an out aid. An alarm is generated to inform clinicians of a fall. Cao automatic message to family members and other stakeholders et al. [44] prevent injuries which occur as a result of lying on the involved in the patient’s care. Ren et al. [161] enable caregivers ﬂoor after a fall for a long period of time, this is achieved by of patients to have access to a centralised cloud server, where data patients wearing the smart-phone and using the built-in sourced from patients activities are transmitted to it. This enables accelerometer sensor to detect falls when they occur. A threshold patients to receive medical attention, in some cases, prior to a fall algorithm was developed which can be adapted to the patient occurring or soon after the event. A fall is detected by distinguish- demographic information, such as age, gender, height and weight ing ADLs from simulated activities that are stored in the cloud ser- to increase the detection accuracy of motion outputted from ver as fall events. Finally, Sahota et al. [162] reduce bedside falls of patients’ movement. patients in hospital by monitoring their activities. If the patients leave the bed, an alarm is triggered and sent to the nursing team 6.2. Intervention types of the hospital, providing the location of the patient who has fallen. The FIPIs presented in [43–45,133–171] use a full range of inter- 6.3. Systems vention types i.e. activity monitoring, fall detector and medical assis- tance to detect and reduce fall related injuries occurring. For Application types of all FIPIs presented in [43–45,133–169,171] example, Abbate, Avvenutia, Bonatesta, Colaa, Corsinia and are static typically offering no form of rich interaction or visual Vecchioa [40] and Abbate, Avvenuti and Light [133] develop ﬁlter- feedback based on the activity monitored by these systems. Martín ing techniques to distinguish falls from ADLs in order to speciﬁ- et al. [154], Dai et al. [139], Tang et al. [165] and Fang et al. [142] all cally identify falls when they occur, gather data proﬁles about gather data from older adults through mechanisms which detect older adult’s movements, and automatically send alerts to clini- falls based on the movements being made by older adults in cians in the event of a fall. Cao et al. [44] monitor older adults’ real-time. There is no interface present, as the sole purpose is to activities and acceleration of movement to determine if the older detect falls and send an automatic message or call automatically adult is currently experiencing a fall. If a fall is identiﬁed, an SMS to preloaded emergency contacts. FIPIs appear to be heterogeneous message is sent to the older adult’s carer provide immediate assis- with regards to the devices, systems and techniques that underpin tance to their client. Bagnasco et al. [136] classiﬁes three different them. One of the main objectives of successful FIPIs is to distin- fall types which are front fall, backward fall and lateral fall, which guish fall events from ADLs. This has proven to be an on-going occur as a result of performing ADLs. This approach increases the challenge and often the primary point of focus of contemporary accuracy of identifying fall events so that they can receive ade- systems. Much effort has been invested in improving classiﬁcation quate support from clinicians. Kepski and Kwolek [147,148],Yu algorithms and detection techniques to be able to consistently dis- et al. [45] and Koshmak et al. [149] all provide a means of either tinguish fall events from ADLs; however, this is perhaps at the obtrusively or unobtrusively monitoring older adults within their expense of focusing attention on developing more interactive, ana- living environment to identify falls. Once a fall has been identiﬁed, lytical and informative user interfaces for such systems. an alarm is triggered for caregivers to provide medical support to FIPIs are deployed on a range of platforms including desktop older adults who have fallen. Laguna and Finat [150], Werner computer [134–136,143,144,148,152,153,157–164,168,169]; et al. [168] and Koshmak et al. [149] monitor older adults move- smart-phone [44,45,133,137,139–142,145,146,149,154–156,165– ment remotely and detect falls when they occur. Paoli et al. 167,171] platforms. Unlike other system categories, there appear [157] and Leone et al. [152] provide notiﬁcations to caregivers in to be no FIPIs deployed on game console platforms. An example the case of a fall and enable them to have authorised access to of systems deployed on a smart-phone platform is that of Abbate monitor older adults. Mastorakis and Makris [155] monitor older et al. [43] and Abbate et al. [133] who develop a fall detection adults activities to identify a fall by using an algorithm with a 3D system on a smart-phone where patients are required to have J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 335 Table 4 Fall injury prevention interventions. Fall injury prevention Fall injury prevention interventions systems Fall risk factors Intervention types Systems Information sources Interface types Fall related Activity Fall Medical Application Platform Sensor Sensor Deployment Multimodal Collaboration injuries monitoring detector assistance type location purpose environment interaction Abbate et al. [43] X X X X S DC + Sm C + U Co + BS He Nii + Ts Async Abbate et al. [133] X X X X S Sm C + U Co He Nii + Ts Async Albert et al. [134] X X X X S DC + Sm U ES He Nii + Ts Async Aud et al. [135] X X X X S DC C BS He Nii Async Bagnasco et al. [136] X X X X S DC U BS He Nii Async Busching et al. [137] X X X X S Sm U Co + BS He Nii + Ts Async Cabestany et al. [138] X X X X S DC + Sm C + U Co + BS He Nii + Ts Async Cao et al. [44] X X X X S Sm U Co He Nii + Ts Async Dai et al. [139] X X X X S Sm U Co He Nii + Ts Async Della Toffola et al. [140] X X X X S Sm C + U Co He Nii + Ts Async Fahmi et al. [141] X X X X S Sm U Co He Nii + Ts Async Fang et al. [142] X X X X S Sm U Co + BS He Nii + Ts Async Ferrari et al. [143] X X X X S DC U BS He Nii Async Fourlas and Maglogiannis X X X X S DC U BS He Nii Async [144] He et al. [145] X X X X S Sm U Co He Nii + Ts Async He et al. [146] X X X X S Sm U Co He Nii + Ts Async Kepski and Kwolek [147] X X X X S DC + Sm C + U Co + BS He Nii Async Kepski and Kwolek [148] X X X X S DC C Rp He Nii Async Koshmak et al. [149] X X X X S Sm U Co He Nii + Ts Async Laguna and Finat [150] X X X X S DC + Sm U Co He Nii + Ts Async Lee and Carlisle [151] X X X X S DC + Sm U Co + BS He Nii + Ts Async Leone et al. [152] X X X X S DC C BS He Nii Async Li et al. [153] X X X X S DC C BS He Nii Async Martín et al. [154] X X X X S Sm U Co He Nii + Ts Async Mastorakis and Makris X X X X S Sm C Co He Nii Async [155] Mehner et al. [156] X X X X S Sm U Co He Nii + Ts Async Paoli et al. [157] X X X X S DC C BS He Nii Async Papadopoulos and Crump X X X X S DC U BS He Nii Async [158] Planinc and Kampel [159] X X X X S DC C He Nii Async Rantz et al. [160] X X X X S DC C Rp He Nii Async Ren et al. [161] X X X X S DC U BS He Nii Async Sahota et al. [162] X X X X S DC C BS He Nii Async Shieh and Huang [163] X X X X S DC C He Nii Async Shim et al. [164] X X X X S DC C BS He Nii Async Tang et al. [165] X X X X S Sm U BS He Nii + Ts Async Terroso et al. [166] X X X X S Sm U Co He Nii + Ts Async Viet et al. [167] X X X X S Sm U Co He Nii + Ts Async Werner et al. [168] X X X X S DC C BS He Nii Async Yu et al. [45] X X X X S Sm C Co He Nii Async Zhang et al. [169] X X X X S DC C BS He Nii Async 336 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 the smart-phone on their person as a wearable device. The system via Zigbee, a communication protocol that creates a wireless per- consists of three fundamental components: (1) a device that col- sonal area network with low-powered devices. On the other hand, lects physiological data via wearable and environmentally embed- systems presented in [44,45,133,139–141,145,146,149,150,154– ded sensors; (2) a ﬁltering technique to process the sensor data to 156,166,167,171] use co-opt smart-phone sensors to detect falls. distinguish it between falls and ADLs; and (3) communication of an For example, Abbate et al. [43] develop a fall detection system on alert in the event of a fall occurring. smart-phones that track the movement of patients, identiﬁes a fall and then automatically sends a notiﬁcation to emergency services. 6.4. Information sources Although there are beneﬁts in using smart-phone sensors, for example, the built-in accelerometer and gyroscope to obtain infor- FIPIs exploit a range of information sources in order to detect mation from the patient, it is recognised that users have to be will- falls. Sensor location deﬁnes where the sensors are located to source ing to wear the device, which can be considered intrusive. information typically located either on the user or are embedded Alternatively, there are sensors repurposed to suit detecting falls. within the context or environment in which the falls are being Repurposed sensors in [148,160] have various forms, such as cam- detected. The sole location of sensors used by the FIPIs presented era, pressure, and audio or are devices that are brand speciﬁc, for in [44,134,136,137,139,141–146,149–151,154,156,158,161,165– example, the Microsoft Kinect. For example, Kepski and Kwolek 167,171] is on the user. For example, Cao et al. [44] require users [148] develop a fall detection system which repurposed the Kinect to wear a smart-phone device on their body, repurposing the as an input device to source information from older adults func- built-in accelerometer in order to monitor the movement of the tioning in their living environment. The systems presented in patient. Another example, that of Terroso et al. [166] uses a wear- [43,137,138,142,147,151] use co-opted sensors and bespoke sensors able accelerometer sensor which sends data on patient movement and are considered as distributed systems. to the smart-phone application and server. The sensor communi- Deployment environment refers to the living environment in cates to the smart-phone application via Bluetooth in order to which FIPIs are deployed. Home environment [43–45,133–169,171] enable the analysis to be carried out, the geographical location relate to FIPIs that are developed to detect fall events among to be logged via the smart-phone GPS sensor, and the message to community-dwelling older adults. For example, Della Toffola et al. be sent. Fang et al. [142] propose an android-based fall detection [140] develop a robotic system to be deployed in the patient home. system which requires users to either attach the smart-phone to This system monitors older adults and responds rapidly to fall their chest, waist or thigh to detect the signiﬁcant change in accel- events that occur in the home. A sensor network, which is part of eration in order to accurately detect a fall. Fahmi et al. [141] use the system, is used to determine where the patient has fallen in built-in accelerometer and orientation sensors which are built into the home. The nodes within the network are connected via wireless the mobile device to measure the position and acceleration of the signal which sends an alert to the robot in case of a fall. The robot user to understand a range of fall characteristics and accurately then communicates the alert to the clinicians who intervene with detect a fall when it occurs. medical assistance. Context is the main information source used in [45,135,148,152, 153,155,157,159,160,162–164,168,169]. Data is unobtrusively 6.5. Interface types collected from patients and is arguably less intrusive than approaches which place sensors on the body of the patient. For All FIPIs use multimodal interaction which comprises the way in example, Kepski and Kwolek [148] use the Kinect to detect falls which information is collected from users, how they control the in a living environment. This approach enables older adults to be system, and the in-built touch mechanisms that are embedded into tracked in 3D, and is a low-cost solution. Although environmen- handheld devices. Non-interactive interface are used in [43–45,133– tally embedded sensors unobtrusively source information based 171] with no interface presented to the user, but uses sensor on user movement, the system is limited by spatial coverage and devices to source information from users and is employed to con- is unable to monitor patient movement wherever they go, unless trol the fall prevention system. For example, Kepski and Kwolek the environment is instrumented by a number of sensors; although [148] develop a fall detection system using the Kinect as an input this may address such a challenge, doing so is often not practical. sensor device that was used to source information from users in Yu et al. [45] use a single camera to monitor community- the living environment. Although in a gaming context the Kinect dwelling older adults in their home in order to detect a fall based provides users with feedback based on their performance, this sys- on posture. However, in some scenarios wearable sensors and con- tem is a non-game application and is used purely for monitoring text sensors are used to improve the level of accuracy of detecting user movements, hence no speciﬁc input required nor an interface falls. The systems proposed in [43,133,138,140,147] all use both presented for users to control the system. Shieh and Huang [163] the user and context as an information source. For example, Cabes- develop a video-like surveillance system which uses cameras to tany et al. [138] propose a small sensor device which require users monitor high risk locations within the home to capture daily to attach around their waist or hip to achieve optimal accuracy movement performed by users. This system is uni-directional as using a developed algorithm. The device is developed so that it is no user interaction is required as multiple cameras are unobtru- easy for users to wear while performing ADLs. The context-based sively deployed throughout the living environment for a wider device is deployed as a bed sensor to detect falls that occur in coverage and to collect vision data from users in order to detect instances where users are not wearing their sensor. If a fall has falls. Li et al. [153] focus on detecting a fall using acoustics in the occurred, an alarm will be generated. If there is no physical living environment and automatically sends a notiﬁcation to the response from the user, a notiﬁcation from the sensors is sent to caregiver when the detected fall occurs. Della Toffola et al. [140] the smart-phone application and a message is then sent to emer- monitor ADLs of older adults who are at risk of falling. The system gency services. detects falls, but recognises that fall detection is prone to false pos- All fall injury prevention systems use devices that have differ- itives, and hence, a robot is deployed in the environment in an ent sensor purpose to sense activity signatures that represent fall attempt to address these issues and to intervene if a fall is events. Bespoke sensors are used in [135,136,143,144,152,153,157, detected. An alarm is sent out to emergency services, caregivers 158,161,162,164,165,168,169] to identify fall events. For example, and clinicians. Rantz et al. [160] deploy a camera-based interven- Bagnasco et al. [136] design a bespoke wearable device that trig- tion to prevent falls in hospital rooms, preserving patients’ privacy gers an alarm after a fall, then transmits the data to a base station and unobtrusively capturing activities that lead to a fall and notify J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 337 clinicians of a fall. The systems presented in [43,44,134,137–139, rather than providing interactive applications that engage patients 142,145,146,149–151,154,156,165–167,171] use both non- in interventions that would reduce fall risks. In most FIPI studies, interactive interface and touchscreen to enable users to use the there is a singular focus on reducing fall-related injuries as such systems via movement and touchscreen gestures. An example of injuries happen in the event of fall. this is He et al. [145] who present a fall detection system on the Android smart-phone. The embedded sensors on the smart- 7. Cross falls prevention interventions phone are utilised to collect information on the user’s movement. If a fall is detected an alarm message with the time and the Cross falls prevention intervention systems (CFPIs) target the patient’s location is sent out to clinicians and other preloaded con- full range of interventions covered by Pre-FPIs, Post-FPIs and FPIs, tacts. This system has a natural user interface, as the sensors are thus providing an integrated approach to the delivery of falls pre- embedded into the smart-phone; the user is required to attach vention interventions to patients. Table 5 presents a summary of CFPIs proposed in the research literature. the smart-phone to their waist, as required for the built-in accelerometer. On the other hand, the touchscreen function gives 7.1. Fall risk factors the user the option of disabling the system by closing the applica- tion or an alert is automatically sent to practitioners. All CFPIs solely target intrinsic fall risk factors [37,73,172–174], Collaboration between patients and practitioners in FIPIs occurs as a result of data being sent to practitioners as a consequence of a with the exception of [73] which targets both intrinsic and extrin- fall event being detected. However, the collaboration between the sic fall risk factors. With regard to CFPIs that solely target intrinsic two parties is asynchronous [43–45,133–171]; these systems do risk factors [37,172–174], functional ability deﬁcits are the sole not offer any real-time communication functions to the practi- focus of these studies. They are also considered in the study of tioner in order to communicate with the patient immediately after Shi and Wang [73], which also addresses extrinsic factors. An the fall has occurred. All systems simply alert the practitioner that example of a study which focused on intrinsic risk factors is that a fall has occurred, but do not provide any further scope for com- by Silva et al. [37] who propose a game to assess older adults’ gait munication, within the bounds of the system, after the alert has in order to delay onset of strength and functional decline. Similarly, been sent. An example of this is Abbate et al. [43] and Dai et al. Chen and Gwin [172] and Ranasinghe et al. [174] focus on intrinsic [139] who develop systems which enable asynchronous collabora- factors such as poor postural transition, gait, history of falling, and other fall-related risk activities, which affect one’s functional abil- tion between the faller and clinician as a notiﬁcation is sent to the clinician in case of a fall. Even if an older adult has fallen, an oppor- ity and ultimately may lead to falls. tunity for collaboration does not occur in real-time, but rather The cross falls prevention interventions presented in [73] focus when clinicians respond by providing medical assistance to on both intrinsic and extrinsic risk factors, especially environmental patients. hazards. This is exempliﬁed in the work undertaken by Shi and Wang [73] who develop a smart-phone application that provided 6.6. Discussion tips to increase awareness of fall hazards in the home. The tips include illustrations of exercises and ideas about how to improve FIPIs are all commonly used to detect falls and prevent fall- the home so as to avoid environmental hazards such as poor light- related injuries with the use of intervention types such as activity ing in the hallway, kitchen and bathroom. monitoring, fall detector and medical assistance, which all are interdependent. The falls prevention technology literature appears 7.2. Combination of intervention types to be saturated with systems developed to monitor activity, detect falls, and send an alert if a fall is detected. Despite the abundance of A few systems [73,173] use intervention types often associated FIPI systems in the research literature and the signiﬁcant beneﬁts with Pre-FPIs and FIPIs to prevent the onset of fall risks or identi- in the deployment of such systems, there are a number of chal- fying risks to avoid fall-related injuries. For example, Shi and Wang lenges that may potentially impact their use in practice. Accurate [73] increase awareness of environmental fall risks, assess and detection of falls is one such challenge particularly distinguishing detect fall risks when they occur and alert older adults and carers the kinematic differences between ADLs and fall events is an on- to take preventive measures in a timely fashion. Cortés et al. [173] going area of research. Preserving users’ privacy is also considered develop assistive technology to increase independence around the challenging [160]. Repurposed camera sensors have an advantage home and to help alleviate the burden on caregivers and family over wearable sensor devices as image processing techniques can members. The proposed walking aid has embedded sensors, which be applied to preserve users’ privacy. This also offers an unobtru- collects the usage data and sends it to clinicians. Two of the ﬁve sive way of sourcing information and creates a means of monitor- CFPIs [172,174] use intervention types that are often employed ing patients to verify whether or not they have fallen. Repurposed by Post-FPIs and FIPIs. For example, Chen and Gwin [172] and camera sensors are only able to monitor predeﬁned spaces within Ranasinghe et al. [174] propose systems that monitor older adults’ the living space, however, this can be a beneﬁt in comparison to physical activities to identify fall risks. Once a fall has occurred wearable sensors as, they can source data directly from users with- clinicians are sent a notiﬁcation to either reduce the potential risk out users having to attach a device to their body [148]. or to assist in the case of a fall. Silva et al. [37] assess the intrinsic Whilst the risk of falling cannot be eradicated due to the inevi- risk factors such as functional ability deﬁcits, and more speciﬁcally table nature of falls occurring as a result of ageing, effective fall walking patterns, for quality and to provide a form of exercising prevention measures can be implemented to help minimise the (for example, dancing), to encourage physical activity to counter risks from the outset. The majority of the research efforts in falls the potential risk of falling. preventing technology have focused on using and developing machine learning techniques and optimising algorithms to 7.3. Systems enhance the sensitivity and speciﬁcity of accurate fall detection when they occur, with limited efforts expended on the design The application types employed by [73,172–174] are all static and interface functionality of these systems. The consensus of and the remaining system [37] is an interactive game application, the FIPI studies indicate that these systems provide static applica- which provides a form of interaction and feedback to the user. tions with the sole purpose of detecting falls when they occur, Ranasinghe et al. [174] propose a static system to be used within 338 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 hospitals and residential care homes. Users are required to wear a fall risks. The remaining systems [37,73] use co-opted of smart- device to enable monitoring of movement in the environment. Shi phones sensors to source information from users. For example, and Wang [73] develop a smart-phone application, using built-in Silva et al. [37] require users to wear their smart-phone as a wear- sensors, which did not require any form of user interaction as it able device to provide movement to control and interact with the was merely a data collection tool coupled with suggesting knowl- game. Shi and Wang [73] also exploit built-in smart-phone sensors edge tips on reducing environmental risks within the home and to monitor older adults in order to detect fall events as and when how to get into a recovery position after a fall to prevent any they occur. adverse effects thereafter. Chen and Gwin [172] design a wearable In terms of deployment environment, the majority of CFPIs sensor device, which collects data from users’ performing ADLs and [37,73,173] are deployed within the users’ home environment. Silva an algorithm to detect fall events. There is no interactivity required et al. [37] propose a hybrid system to be deployed within the older for this device to function, as the device’s function is to collect data. adults’ home to enable clinicians to administer clinical tests and Finally, Cortés et al. [173] develop a system to assist users living monitor adherence to unsupervised exercises. Cortés et al. [173] independently by attaching sensors on assisted equipment for clin- deploy a system, within the patients’ home to reduce fall risks icians to monitor the usage of the equipment. Silva et al. [37] and increase assisted living via the use of artiﬁcial intelligence develop a game-based application, which require users to interact and robotic solutions. Conversely, the study by Ranasinghe et al. with the dancing game using the built-in sensors on a smart-phone [174] deploys technology-based interventions to reduce risks in to enable older adults to interact with the game application, ensur- hospitals for older adults who are admitted into acute care, espe- ing that the physiological data of users matches the movement cially if they are cognitively impaired, which therefore warrants required in the dancing game. The game then provides the user the need to monitor them during their stay in hospital or residen- with visual and audio feedback. tial care. Two out of ﬁve systems [37,73] are deployed on a smart-phone platform. Shi and Wang [73] and Silva et al. [37] develop systems 7.5. Interface types that use built-in smart-phone sensors to source data directly from users. Systems in [174] are deployed on the desktop computer plat- The collaboration which is afforded by CFPIs [37,73,172–174],as form. Ranasinghe et al. [174] develop a desktop application used to with other fall prevention systems, is asynchronous. Patients (typi- identify activity signatures of fall risks in real-time from wireless cally in an unobtrusive way) generate physiological and movement sensors and deployed within the living environment, to process data which is sent to clinicians who respond with medical assis- and store the data and alert clinicians to address a fall risk. tance or establish the likelihood of users falling. Shi and Wang [73] develop a game to increase levels of engagement with 7.4. Information sources home-based exercises to reduce fall risks and enable clinicians to monitor those risks and carry out clinical tests. Ranasinghe et al. The information sources in [37,73,172–174] support a wide [174] enable nurses to respond with help to patients who attempt range of fall prevention activity. Sensors are often used to gather to transfer on and off items of furniture, such as the toilet and bed information from various sources and therefore have different sen- unassisted without caregivers’ help, which could lead to falls. sor locations. The majority of CFPIs [37,73,172,174] source informa- All systems [37,73,172–174] use a particular form of multimodal tion directly from users. Chen and Gwin [172], for example, interaction to interact with technology-based interventions. Natu- propose a device that automatically assesses and detects fall risk ral User Interface appear to be a common form of interaction factors by older adults wearing the device on their body to contin- [37,73,172–174] as it enables users to perform physical activities uously monitor physical activities. The device gathers acceleration and gestures to control an in-game avatar and various objects in data in 3D space and plots this to X, Y and Z axes, reﬂecting the a virtual environment. CFPIs in [37] uses touchscreen and natural body pose of the user respectively. Ranasinghe et al. [174] propose user interface for users to manipulate the systems by performing a wearable sensor. Users are required to attach it to a piece of gestures to interact with the touchscreen on a smart-phone. The clothing to enable the device to monitor their activities in real- system responds to those gestures by providing feedback to users. time to classify high risk tasks. Silva et al. [37] present a smart- In the study of Silva et al. [37], the user attaches the smart-phone phone-based system where a accelerometer sensor is used to to their body so that the system can track their dance moves. Users source data from older adults. The smart-phone is attached to their are provided with both audio and visual feedback, as data their lower back so that the system recognises physical activities being dance moves is transmitted to the movement of a character during performed during game play. The remaining system, that of Cortés gameplay. Shi and Wang [73] develop a smart-phone application et al. [173] uses context in which user functions coupled with for the user to interact with via the built-in touchscreen. Chen movement to collect data. The same study proposes a system that and Gwin [172] require natural gestures and movement from older uses wearable devices to capture data from users and the context adults to operate their system, enabling clinicians to monitor to ascertain the state of users in order to respond with support. patients remotely. Sensors are also attached to assistive equipment, such as walking aids, to help keep track of movement and general use of the equip- 7.6. Discussion ment and to keep clinicians informed. All sensors capture data directly from users or the context in The CFPIs presented in this section [37,73,172–174] deploy a which they function. With regards to sensor purpose, a range of sys- full range of techniques typically associated with Pre-FPIs, Post- tems [172–174] use bespoke sensors, which are developed speciﬁ- FPIs and FIPIs to assess, detect, and respond to fall risks. As a result cally for gathering data obtrusively or unobtrusively to track of combining these techniques, multiple fall risks are responded to, users’ movement. For example, Ranasinghe et al. [174] utilise a often allowing for more comprehensive interventions to be pro- wearable sensor device to unobtrusively monitor older adults in vided compared with systems that target one particular interven- real-time, preserving their privacy, and enabling clinicians to tion type. The only cross fall prevention system that reduces source data from them remotely. Chen and Gwin [172] and Cortés both intrinsic and extrinsic risks is [73], which enhances awareness et al. [173] both propose bespoke devices which require older of environmental fall hazards supplemented with guidance of how adults to attach the devices to their body. Cortés et al. [173] to conduct exercise movements to increase adherence. All cross- built-in sensors into assistive equipment to enable monitoring of prevention systems [37,73,172–174] use a natural user interface, J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 339 enabling users to interact with computerised content by perform- ing a range of natural gestures [175]. Movement data is analysed, for signatures that correspond to fall risks, via the use of computa- tional techniques and advanced processing capabilities. CFPIs that are deployed on smart-phones [37,73,172–174] exploit their inher- ent natural user interface and touchscreen interface. From the sys- tems presented here [37,73,172–174], it can be inferred that patients play a major role in their care by engaging with these interventions via the use of sensors and the communication func- tions of the systems. Perhaps the complexity of these systems and the increased overhead required to design and deploy such sys- tems are the reasons that only a small number of such systems have been presented in the research literature to date. 8. Challenges and future research directions In summary, taking a broader view of the typical functions that each category of system fulﬁls, the majority of Pre-FPIs [18,19,40– 42,74–79,81,82,84–90,93,96,97,99–103,105,106,108] deploy 3D technology and games as a means to augment evidence-based exercises, focusing on intrinsic fall risk factors, such as functional ability deﬁcits and balance impairments. A large number of Pre- FPIs [18,19,41,78–80,82,84,86,90,93,99–106] are deployed within the home environment to overcome issues of non-compliance with exercising and eradicate the travelling costs to rehabilitation centres. Most of the post-FPI systems are static [38,39,114–119, 121–123,126–130]. However, the remaining systems provide an interactive means of engaging older adults during fall risk assess- ment programmes [120,124,125]; speciﬁcally focusing on intrinsic fall risks. Post-FPI systems [38,114–122,124–130] are also often deployed within the home environment. With regards to FIPIs, the majority of these systems focus on falls detection, often via a database of simulated fall behaviour to help distinguish between fall events and ADLs. All approaches, to some extent, detect falls obtrusively or unobtrusively focusing on older patients. Technology-based falls prevention research has tended to focus on detecting falls as a result of its inevitable occurrence, particu- larly in older people. Nevertheless, Pre-FPI and Post-FPI systems have shown promise in reducing the onset of fall risks, rather than injuries that occur in the event of a fall. CFPI systems provide a comprehensive fall prevention approach as they include a combi- nation of intervention types of Pre-FPI, Post-FPI and FIPI [37,73,172–174]. Smart-phone features are strongly being used across all system types. These portable, low cost, and increasingly ubiquitous devices are being used as a solution in deploying fall prevention systems, which is in line with a growing number of older adults now becoming more familiar with smart-phones [176], and consequently one can assume that smart-phones will continue to be part of future fall prevention systems. Effective management of falls is a complex endeavour, particu- larly when considering the multiple intrinsic risks, namely social and physical factors and extrinsic risks such as slippery surfaces, poorly ﬁtted or abandoning assistive equipment, poor lighting, unsafe stairs and loose rugs [2,29]. It is recognised that in order to reduce the risk of falling, particularly in an older adult popula- tion, targeting extrinsic risk factors is equally as important as tar- geting intrinsic risk factors [2]. The effective management of fall risks in order to enable older adults to live independently within their homes for longer is seen as being extremely beneﬁcial to the patient in terms of maintaining independence and quality of life [20]. Despite the key role extrinsic fall risk factors play in ensuring that the goal of independent living is realised, and that fall risk factors are suitably managed, it is apparent that extrinsic risk factors are rarely considered and targeted by contemporary fall prevention intervention systems. Of the 104 fall prevention Table 5 Cross falls prevention interventions. Cross fall prevention Cross falls prevention interventions systems Fall risk factors Combination of intervention types System Information sources Interface type Intrinsic Extrinsic Pre-fall Post-fall Fall injury Application Platform Sensor Sensor Deployment Collaboration Multimodal prevention prevention prevention types location purpose environment interaction Chen and Gwin [172] Fun X X S DC U Bs – Async Nii Cortés et al. [173] Fun X X S DC C + U Bs He Async Nii Ranasinghe et al. [174] Fun X X S DC U Bs Hs Async Nii Shi and Wang [73] Fun Eh X X S Sm U Co He Async Nii + Ts Silva et al. [37] Fun X X G Sm U Co He Async NUI + Ts 340 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 systems [18,19,37–45,73–108,114–130,133–169,172–174], only 4 home-based exercises, and give practitioners the ability to systems [38,73,74,98] target extrinsic risk factors. FIPI systems in monitor patient’s physical health remotely. particular, by deﬁnition, do not target extrinsic factors at all, as their sole focus is to detect falls so as to reduce fall-related In response to Pre-FPI challenges, the following research direc- injuries. tions and recommendations are proposed: There is a need for new research to explore how technology- based applications can be applied to better address and manage Recommendation 1: Identify new opportunities and develop new extrinsic fall risk factors. Furthermore, when exploring the extent technology-based applications to support patients and practition- to which existing systems facilitate the process of collaboration ers in their efforts to overcome extrinsic risk factors. One promis- and shared decision making between patient and practitioner, it ing area of technology that may provide opportunities to seems that the majority of systems do not invest signiﬁcantly into overcome this challenge may be found within the interactive delivering such functionality. Pre-FPIs appear to be the category 3D virtual reality and gaming domain. For example, interactive delivering the largest proportion of systems which offer syn- 3D gaming applications which simulate the range of extrinsic chronous communication between patient and practitioner and fall risks that occur at a patient’s home may help to improve hence patient–practitioner collaboration [18,19,40–42,74–82, patient’s awareness of risks and encourage the development 84–90,93,96,97,99–108]. The collaboration which is supported, of strategies to overcome these risks if they occur in real-life. however, is synchronous and hence does not optimally support However, it is important that such solutions are applied in a real-time patient–practitioner discussions/interactions about the meaningful way in order to target extrinsic risk factors that fall risks encountered or indeed how these may be better managed relate to patients personal home environment in which they and overcome. In the rare cases that the system does facilitate function. This is particularly important when considering the asynchronous patient–practitioner communications notion of ageing-in-place, which focuses on enabling patients [83,91,92,94,95,98], the system functionality does not tend to to remain in their home for longer. Therefore, addressing actively support and facilitate shared decisions to be made about extrinsic risk factors via the use of technology could reduce fall the patient’s care or enable the patient to provide input into the events that occur as a result of multiple risk factors or solely decisions made about their care. based on extrinsic risk factors. As a consequence of carrying out this survey, a number of Recommendation 2: Develop technology-based applications challenges have emerged which should be addressed by the falls which enable and support fall prevention intervention education prevention technology research domain. and promotion activity. Taking a pro-active approach to educat- ing patients, who may still be at low risk of falling, around fall 8.1. Challenges to Pre-FPIs risks is likely to increase their awareness of potential risks and encourage behaviour change that may reduce their risk of fall- Challenge 1: Lack of research effort focused on reducing extrinsic ing in the future. Given the distinct lack of applications which risk factors, which are of equally major concern for patients who take such an approach, coupled with the potential beneﬁts, exhibit intrinsic risks and live independently. In many instances, there is a need for more focused technology-based research in falls occur as a result of multiple risk factors including intrinsic this area. Furthermore, for those who have been prescribed and extrinsic risk factors. Many interventions prevent both assistive equipment to help with performing daily activities types of risks in order to increase the effectiveness of prevent- and reduce fall risks, educating patients on the need for equip- ing a fall. Although there are Pre-FPIs that have produced ment might be developed to increase adherence and successful promising results for addressing intrinsic fall risk factors to uptake of assistive equipment in the home to help with mobil- date, there are a limited number of systems that reduce both ity issues and the onset of fall risks. Interactive 3D gaming and functional ability deﬁcits and extrinsic fall risks and solely virtual reality simulations of fall risks again, may offer promis- reduce extrinsic fall risks. ing platforms to deliver educational interventions. Educational Challenge 2: Lack of fall education interventions used in Pre-FPIs interventions deployed on mobile platforms such as smart- to reduce fall risks. There are a small number of Pre-FPIs that phones and tablet-based applications may also be an area of use fall education interventions to reduce fall risks, not least potential opportunity for such applications, particularly given as a singular intervention system. While the vast majority of the popularity and ever increasing ubiquity of such devices. Pre-FPIs that utilise 3D technology and games for preventing intrinsic fall risks has shown promising results, there is an 8.2. Challenges to Post-FPIs absence of using such technology to augment fall education interventions, with the exception of two [74,98]. These two sys- Challenge 3: Current systems do not consider or support the deliv- tems speciﬁcally provide advice for patients to avoid environ- ery of environmental assessment interventions to reduce fall risks. mental hazards. Bell et al. [74], as well as focusing on The majority of the systems produce personalised applications reducing functional ability deﬁcits, also addressed environmen- by sourcing information obtrusively or unobtrusively, using tal risks, noted down in paper-based form, such as decreasing sensors, directly from the patient’s physical movement in accor- clutter, furniture, spills and the impact these could have on dance with clinical assessment tests. Although these systems older adults in their living environment. Otis and Menelas enable patients to self-assess their functional abilities and cog- [98] look at speciﬁc characteristics of the environmental condi- nitive function, there is little consideration given to assessing tions in which older adults function and notiﬁed them of a the environment in which the patients function, with the potential risk of falling. Despite this, Pre-FPI systems enable exception of one system [38]. This is particularly important as patients to self-manage and reduce falls by engaging in unsu- systems proposed in the literature are directing their efforts pervised health promotion activities. Fully realising the to ageing-in-place, independent living and remote assessment patient–practitioner collaboration paradigm is a challenge, as but they do not take into account the fall hazards that may be patients are not given the opportunity to be involved in any apparent within the patient’s home environment. decision-making or interventions that reduce extrinsic factors Challenge 4: Existing Post-FPI systems do not enable patients and of falling. Furthermore, Pre-FPI systems are of major beneﬁt in practitioners to interact and collaborate whilst fall risk assessments that they provide an intuitive way for patients to engage in are carried out using Post-FPIs. The majority of Post-FPI systems J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 341 provide remote synchronous but static communication mecha- user-worn sensors, which require effort in reminding users to nisms and hence do not provide patients or practitioners with a wear the user-worn sensors. means of interacting with each other whilst using these sys- Challenge 8: The majority of FIPIs are static and provide no form of tems. Typically, systems produce static reports on the prede- user interaction. In most studies, developing machine learning ﬁned criteria the system is set up to report on, with no option techniques and optimised algorithms has been the focus in for the patient to provide additional contextual detail which order to increase the accuracy of fall detection, however little may be useful for interpreting the data in a more personalised consideration has been given to the interface functionality of and appropriate way. Post-FPI systems therefore would beneﬁt the systems. There is a lack of interactive applications to engage from offering more collaborative functions that provide an patients during interventions that could reduce fall risks. How- opportunity to enable patients, to some extent, to collaborate ever, much less effort regarding the interaction has been with clinicians and help interpret the data that these systems explored in this intervention. All systems attempt to alleviate generate. fall related injuries that occur after a fall, these injuries are more severe upon the impact. In response to these challenges, the following future research direction recommendations are proposed: In response to these challenges, the following recommenda- tions are proposed: Recommendation 3: Incorporate environmental assessment inter- ventions into Post-FPI systems. Whilst falls often occur as a result Recommendation 5: Develop, deploy and evaluate FIPI systems of multiple fall risks, it appears that Post-FPIs would beneﬁt under real-life conditions. Falls are a complex phenomenon and greatly from assessing extrinsic risk factors by incorporating are yet to be fully understood. Patients’ physiology in relation environmental assessment interventions into Post-FPIs. to real-life falls differs, which makes gathering simulated fall Recommendation 4: Develop Post-FPIs which allow patients and like behaviour problematic and the robustness of which may practitioners to engage and collaborate with each other as part be considered to be questionable. Therefore, if FIPI systems of the assessment process. From the falls prevention systems are to be ecologically valid, accurate and reliable such systems reviewed, it appears that providing patients with an interactive would need to be evaluated within real-world settings. means could help to increase compliance to fall risk assess- Recommendation 6: New approaches to deploying camera-based ments by presenting real-time feedback and a mechanism that digital video footage of patients within their home environment, supports richer interactions and collaboration between patients whilst also protecting and preserving privacy of patients must be and practitioners. developed. Some promising avenues via which this may be achieved lie within the image processing and face recognition 8.3. Challenges to FIPIs research domain. For instance, there needs to be more develop- ment of algorithms that dynamically remove and selectively Challenge 5: Existing FIPIs are often unable to demonstrate effec- scramble or distort image detail at the point of capture, which tive and reliable differentiation between fall events and daily activ- may be considered to potentially compromise the patient’s pri- ities in order to accurately detect falls, particularly within real-life vacy. Furthermore, providing clear prompts of when cameras settings. As such, much effort has been expended on developing are monitoring to reassure users that their privacy is not being algorithms and computational techniques to improve the level breached at other times may help with the acceptance of such of sensitivity and speciﬁcity in accurately detecting falls via technology. Developing techniques to selectively activate cam- user-worn or camera-based sensors. This still, however, eras or broadcast footage, only when a potential fall is detected, remains an on-going research challenge. There are a small num- could also be a potential solution for preserving user privacy. ber of FIPI systems that have been evaluated with real-life falls Recommendation 7: Invest effort into developing hybrid sensor due to ethical reasons, however, the remaining systems are networks to detect falls. Instrumenting the patients’ living envi- unable to demonstrate the effectiveness and reliability of the ronments with multiple types of sensors has shown promise in proposed system in real-life settings. Hence this raises issues the research literature as a potential solution and addresses relating to the ecological validity of the proposed systems. In drawbacks of certain sensors by installing another. Camera sen- overcoming such issues, most FIPI systems simulate fall like sors that are used by fall prevention systems are deployed in behaviour in order to gather signatures of fall events in a data- the patient’s environment to detect fall events or fall related base to increase detection accuracy. injuries, however such sensors are limited in coverage, which Challenge 6: Preserving the privacy of patients when using FIPI in some cases, depending on the location of the fall, render it systems that utilise cameras to detect falls. While the use of cam- ineffective. However, the advantage of using cameras is that eras as an alternative to user-worn sensors provides an unob- users are not required to wear any sensors on their body. trusive way of monitoring patients, there still remains the User-worn sensors are not limited in covering the environment, challenge of user’s privacy being breached. but require users to attach a device on their body in order to Challenge 7: FIPI systems that use cameras to monitor patients detect falls. only cover a limited space within the monitored environment. Recommendation 8: Develop systems which support richer and Using cameras as an alternative to user-worn sensors presents more engaging mechanisms for user interaction. Little effort no restriction of where it is installed, however, the camera seems to have been invested into considering the user interface devices are limited in covering a certain amount of space in design of FIPIs, or indeed the speciﬁc user-centred interaction its view. Instrumenting the environment with multiple cameras requirements of older adult users and clinicians. Systems do may increase space coverage of the environment, but will not appear to make any signiﬁcant attempt to develop system increase cost and in some instances it may not be feasible to interfaces that support patient/practitioner collaboration and do so. Monitoring patients and detecting falls using camera sen- interactive information sharing. Therefore, investing effort into sors still remains an on-going research challenge, with limited user-centred design of system interfaces is likely to improve the coverage and effort in optimising image processing techniques level of engagement and acceptance of such systems, which in to mask user privacy. Also, older adults may forget to wear turn is likely to impact on their longer term success. 342 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 order to make it more of a comprehensive prevention, due to the 8.4. Challenges to CFPIs collaborative nature of existing systems that are deployed in older adults home to self-administer assessments. As such, post-fall pre- Challenge 9: CFPI systems face similar challenges to other fall pre- vention systems enable older adults to self-assess for intrinsic vention systems in that there is a lack of effort in reducing extrinsic risks, which in turn enable clinicians to conduct their assessments risk factors. Fall risks are categorised as intrinsic and extrinsic, remotely by deploying a system in the patients’ homes. However, both of which are equally of major concern and become a risk extrinsic risks and personalising the home to aid mobility and to older patients who live independently. reduce fall risks by self-assessment has yet to be explored. Challenge 10: CFPIs incorporate intervention techniques associ- Fall injury prevention systems appear to be prominent in the ated with Pre-FPIs, Post-FPIs and FIPIs as a comprehensive preven- literature amongst other systems as it is focusing mainly on detect- tion that can target multiple fall risk factors. The majority of CFPIs ing falls. As such, falls are inevitable and detecting falls when they prevent multiple fall risks by utilising Pre-FPIs, Post-FPIs and occur to prevent fall-related injuries is essential, however, other FIPIs intervention techniques. Although multiple fall risks are areas of preventing fall risks are of major concern. responded to, developing CFPI systems is an overly complex To address and overcome the challenges faced by pre-fall, post- task which brings with it signiﬁcant time and cost overheads. fall, fall injury and cross fall prevention systems, this study has proposed a range of recommendations for fall prevention systems. In response to these challenges, the following recommenda- It is proposed that future fall prevention systems would beneﬁt tions are proposed: even more from addressing extrinsic risks, particularly how equip- ment could be successfully adopted by clinicians conducting home Recommendation 9: Develop CFPIs which support patients and assessments effectively and older adults being able to self-assess practitioners in their efforts to overcome extrinsic risk factors. their needs for assistive equipment in the absence of clinicians in Recommendation 10: Develop pragmatic CFPI systems which the home. To this end, exploring how home furniture are accu- reduce multiple fall risks whilst also minimising the development rately measured by stakeholders involved in home assessments and deployment overhead associated with such systems. Combin- could ensure the correct ﬁt of equipment in the home, which could ing intervention techniques to target multiple fall risks often lead to successful uptake of and adherence to using equipment. provides more effective falls prevention as fall events occur as Providing an innovative way of educating fall prevention to older a result of multiple risk factors. However, the challenge is to adults using fall hazards typically found in the home has been sug- identify CFPIs which also minimise the resource overhead gested as a potential area of research. Moreover, systems would required for developing these comprehensive solutions. beneﬁt from focusing on enabling patients to self-assess and pro- vide self-care against fall risks and to enable collaboration for 9. Concluding discussion shared-decision making between patients and practitioners. This paper presents a conceptual falls prevention technology Conﬂict of interest model, which includes fall prevention interventions, the informa- tion sources they exploit and their collaboration functions. The The authors declare that they have no competing interests. conceptual model of falls prevention technology was derived from and used to survey a range of fall prevention technology systems that have been proposed within the literature in a speciﬁed time Acknowledgment period. Fall prevention interventions were found to belong to one of four system sub-types; pre-fall prevention (mitigating the early The authors of this study would like to thank The Royal Society stages of fall risks through health promotion), post-fall prevention for grant Ref: RG130826 awarded on the e-Gap programme. (assessing fall risks), fall injury prevention (reduces post-fall inju- ries) and cross-prevention (combination of multiple interventions References used to reduce fall risks) used in practice. The application types are categorised (static, interactive, games and virtual reality) and [1] D.A. Sleet, D.B. Moffett, J. Stevens, CDC’s research portfolio in older adult fall the platforms (desktop computer, game console and smart-phones) prevention: a review of progress, 1985–2005, and future research directions, J. Saf. Res. 39 (3) (2008) 259–267. in which they are deployed. The fall prevention technology sys- [2] W.H.O. Ageing, L.C. Unit, WHO Global Report on Falls Prevention in Older Age, tems that exploit information sources were also categorised (user World Health Organization, 2008. and context) and the purpose of sensors used to source information [3] C. McKay, K.E. Anderson, How to manage falls in community dwelling older (bespoke, repurposed and co-opted) and deployment environments adults: a review of the evidence, Postgrad. Med. J. (2010), http://dx.doi.org/ 10.1136/pgmj.2009.093468. where the systems are installed (home, nursing home and hospital). [4] P.A. Logan et al., Community falls prevention for people who call an The interface type that each system used were also categorised emergency ambulance after a fall: randomised controlled trial, BMJ 340 (natural user interface and touchscreen) and their respective collab- (2010). [5] S.R. Cummings, L.J. Melton, Epidemiology and outcomes of osteoporotic oration functions (synchronous and asynchronous) which occurred fractures, Lancet 359 (9319) (2002) 1761–1767. between older adults and clinicians either ofﬂine, sharing an inter- [6] D. Oliver, Development of services for older patients with falls and fractures face during an intervention or online sourcing data remotely. in England: successes, failures, lessons and controversies, Arch. Gerontol. Geriatr. 49 (2009) S7–S12. Although pre-fall prevention systems has shown promise in [7] N.A. Lannin et al., Feasibility and results of a randomised pilot-study of pre- reducing intrinsic risk factors, there is a lack of research energy discharge occupational therapy home visits, BMC Health Serv. Res. 7 (1) which has been expended on reducing extrinsic risks, partly from (2007) 42. [8] L.H. Kraskowsky, M. Finlayson, Factors affecting older adults’ use of adaptive using education interventions to increase fall hazard awareness, equipment: review of the literature, Am. J. Occup. Ther. 55 (3) (2001) 303– which is mainly a component of a multifactorial intervention. This is due to the sole focus of such systems aiming to increase adher- [9] A. Hammond, What is the role of the occupational therapist?, Best Pract Res. ence to and uptake of exercise interventions. Clin. Rheumatol. 18 (4) (2004) 491–505. [10] P. Kannus et al., Prevention of falls and consequent injuries in elderly people, Post-fall prevention systems are prominent for augmenting tra- Lancet 366 (9500) (2005) 1885–1893. ditional clinical tests used to assess functional abilities and cogni- [11] L.D. Gillespie et al., Interventions for preventing falls in older people living in tion, but there appears to be beneﬁts to reducing extrinsic risks in the community, Cochrane Database Syst. Rev. 9 (11) (2012). J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 343 [12] S.E. Lamb et al., Development of a common outcome data set for fall injury [42] M.A. Williams et al., EXercising with Computers in Later Life (EXCELL)-pilot prevention trials: the Prevention of Falls Network Europe consensus, J. Am. and feasibility study of the acceptability of the Nintendo WiiFit in Geriatr. Soc. 53 (9) (2005) 1618–1622. community-dwelling fallers, BMC Res. Notes 3 (1) (2010) 238. [13] C. Sherrington et al., Effective exercise for the prevention of falls: a systematic [43] S. Abbate et al., A smartphone-based fall detection system, Pervasive Mob. review and meta-analysis, J. Am. Geriatr. Soc. 56 (12) (2008) 2234–2243. Comput. 8 (6) (2012) 883–899. [14] S. Gates et al., Multifactorial assessment and targeted intervention for [44] Y. Cao, Y. Yang, W. Liu, E-FallD: a fall detection system using android-based preventing falls and injuries among older people in community and smartphone, in: 9th International Conference on Fuzzy Systems and emergency care settings: systematic review and meta-analysis, BMJ 336 Knowledge Discovery (FSKD), IEEE, 2012. (7636) (2008) 130–133. [45] M. Yu et al., A posture recognition-based fall detection system for monitoring [15] Y.S. Choi et al., Developing a multi-systemic fall prevention model, an elderly person in a smart home environment, IEEE Trans. Inform. Technol. incorporating the physical environment, the care process and technology: a Biomed. 16 (6) (2012) 1274–1286. systematic review, J. Adv. Nurs. 67 (12) (2011) 2501–2524. [46] W. Tadd et al., Dignity in Practice: An Exploration of the Care of Older Adults [16] C.A. Chase et al., Systematic review of the effect of home modiﬁcation and fall in Acute NHS Trusts, NIHR SDO Report, 2011. prevention programs on falls and the performance of community-dwelling [47] DoH (Ed.), Equity and Excellence: Liberating the NHS (White Paper), The older adults, Am. J. Occup. Ther. 66 (3) (2012) 284–291. Stationery Ofﬁce, London, UK, 2010. [17] A.C. Pighills et al., Environmental assessment and modiﬁcation to prevent [48] DoH (Ed.), Liberating the NHS: No Decision about me without me, The falls in older people, J. Am. Geriatr. Soc. 59 (1) (2011) 26–33. Stationery Ofﬁce, London, UK, 2012. [18] M.G. Jorgensen et al., Efﬁcacy of Nintendo Wii training on mechanical leg [49] DoH (Ed.), Health and Social Care (Independent Living) Bill, The Stationery muscle function and postural balance in community-dwelling older adults: a Ofﬁce, London, UK, 2011. randomized controlled trial, J. Gerontol. Ser. A: Biol. Sci. Med. Sci. 68 (7) [50] A.G. Money et al., Using the Technology Acceptance Model to explore (2013) 845–852. community dwelling older adults’ perceptions of a 3D interior design [19] H.A. Geraedts et al., Adherence to and effectiveness of an individually tailored application to facilitate pre-discharge home adaptations, BMC Med. Inform. home-based exercise program for frail older adults, driven by mobility Decis. Mak. 15 (1) (2015) 73. monitoring: design of a prospective cohort study, BMC Public Health 14 (1) [51] R. Jones, Recurring outbreaks of an infection apparently targeting immune (2014) 570. function, and consequent unprecedented growth in medical admission and [20] S. Uzor et al., Falls prevention advice and visual feedback to those at risk of costs in the United Kingdom: a review, Br. J. Med. Med. Res. 6 (8) (2015) 735– falling: study protocol for a pilot randomized controlled trial, Trials 14 (1) 770. (2013). 79–79. [52] A.G. Money et al., Patient perceptions and expectations of an anticoagulation [21] T.E. Shubert, Evidence-based exercise prescription for balance and falls service: a quantitative comparison study of clinic-based testers and patient prevention: a current review of the literature, J. Geriatr. Phys. Ther. 34 (3) self-testers, Scand. J. Caring Sci. (2015). (2011) 100–108. [53] T. Heart, E. Kalderon, Older adults: are they ready to adopt health-related [22] K. Bennell, F. Dobson, R. Hinman, Measures of physical performance ICT?, Int J. Med. Informatics 82 (11) (2013) e209–e231. assessments: Self-Paced Walk Test (SPWT), Stair Climb Test (SCT), Six- [54] A. Nawaz et al., Assessing seniors’ user experience (UX) of exergames for Minute Walk Test (6MWT), Chair Stand Test (CST), Timed Up & Go (TUG), balance training, in: Proceedings of the 8th Nordic Conference on Human– Sock Test, Lift and Carry Test (LCT), and Car Task, Arthrit. Care Res. 63 (S11) Computer Interaction: Fun, Fast, Foundational, ACM, 2014. (2011) S350–S370. [55] K. Gerling et al., Full-body motion-based game interaction for older adults, in: [23] A. Ejupi et al., A kinect and inertial sensor-based system for the self- Proceedings of the SIGCHI Conference on Human Factors in Computing assessment of fall risk: a home-based study in older people, Hum.-Comput. Systems, ACM, 2012. Interact. (2015), accepted for publication. [56] A. Darekar et al., Efﬁcacy of virtual reality-based intervention on balance and [24] L.Z. Rubenstein, Falls in older people: epidemiology, risk factors and mobility disorders post-stroke: a scoping review, J. Neuroeng. Rehabil. 12 (1) strategies for prevention, Age Ageing 35 (Suppl. 2) (2006) ii37–ii41. (2015) 46. [25] S.R. Nyman, C.R. Victor, Older people’s participation in and engagement with [57] I.D. Cameron et al., Interventions for preventing falls in older people in care falls prevention interventions in community settings: an augment to the facilities and hospitals, Cochrane Database Syst. Rev. 12 (2012). Cochrane systematic review, Age Ageing 41 (1) (2012) 16–23. [58] S. Patel et al., A review of wearable sensors and systems with application in [26] A.G. Society et al., Guideline for the prevention of falls in older persons, J. Am. rehabilitation, J. Neuroeng. Rehabil. 9 (12) (2012) 1–17. Geriatr. Soc. 49 (5) (2001) 664–672. [59] K.L. Perell et al., Fall risk assessment measures an analytic review, J. Gerontol. [27] A.D. Naik, T.M. Gill, Underutilization of environmental adaptations for Ser. A: Biol. Sci. Med. Sci. 56 (12) (2001) M761–M766. bathing in community-living older persons, J. Am. Geriatr. Soc. 53 (9) [60] J.T. Chang et al., Interventions for the prevention of falls in older adults: (2005) 1497–1503. systematic review and meta-analysis of randomised clinical trials, BMJ 328 [28] J. Pynoos, D. Sabata, I. Choi, The role of the environment in fall prevention at (7441) (2004) 680. home and in the community, in: Falls Free: Promoting a National Falls [61] M.A. Habib et al., Smartphone-based solutions for fall detection and Prevention Action Plan, 2005, pp. 41–54. prevention: challenges and open issues, Sensors 14 (4) (2014) 7181–7208. [29] J.T. Chang, D.A. Ganz, Quality indicators for falls and mobility problems in [62] W. Afzal, R. Torkar, R. Feldt, A systematic review of search-based testing for vulnerable elders, J. Am. Geriatr. Soc. 55 (s2) (2007) S327–S334. non-functional system properties, Inf. Softw. Technol. 51 (6) (2009) 957– [30] S.R. Lord, H.B. Menz, C. Sherrington, Home environment risk factors for falls in 976. older people and the efﬁcacy of home modiﬁcations, Age Ageing 35 (Suppl. 2) [63] R. Igual, C. Medrano, I. Plaza, Challenges, issues and trends in fall detection (2006) ii55–ii59. systems, Biomed. Eng. Online 12 (1) (2013) 66. [31] L. Nygård et al., Reﬂecting on practice are home visits prior to discharge [64] C. Tacconi, S. Mellone, L. Chiari, Smartphone-based applications for worthwhile in geriatric inpatient care?, Scand J. Caring Sci. 18 (2) (2004) 193– investigating falls and mobility, in: 5th International Conference on 203. Pervasive Computing Technologies for Healthcare (PervasiveHealth), IEEE, [32] J.K. Martin et al., The impact of consumer involvement on satisfaction with 2011. and use of assistive technology, Disabil. Rehabil.: Assist. Technol. 6 (3) (2011) [65] H. Hawley-Hague et al., Older adults’ perceptions of technologies aimed at 225–242. falls prevention, detection or monitoring: a systematic review, Int. J. Med. [33] L.N. Gitlin, Why older people accept or reject assistive technology, Informatics 83 (6) (2014) 416–426. Generations 19 (1) (1995) 41–46. [66] S. Jalali, C. Wohlin, Systematic literature studies: database searches vs. [34] T. Wielandt, J. Strong, Compliance with prescribed adaptive equipment: a backward snowballing, in: Proceedings of the ACM-IEEE International literature review, Br. J. Occup. Ther. 63 (2) (2000) 65–75. Symposium on Empirical Software Engineering and Measurement, ACM, [35] S. Federici et al., Successful assistive technology service delivery outcomes 2012. from applying a person-centered systematic assessment process: a case [67] H. Joffe, L. Yardley, 4 Content and Thematic Analysis, Research Methods for study, 2015. Clinical and Health Psychology, 2004, p. 56. [36] S. Federici, M.J. Scherer, S. Borsci, An ideal model of an assistive technology [68] M. Montero-Odasso et al., Gait and cognition: a complementary approach to assessment and delivery process, Technol. Disabil. 26 (1) (2014) 27–38. understanding brain function and the risk of falling, J. Am. Geriatr. Soc. 60 [37] P.A. Silva et al., Using the smartphone accelerometer to monitor fall risk while (11) (2012) 2127–2136. playing a game: the design and usability evaluation of dance! Don’t fall, in: [69] S. Springer et al., Dual-tasking effects on gait variability: the role of aging, Foundations of Augmented Cognition, Springer, 2013, pp. 754–763. falls, and executive function, Mov. Disord. 21 (7) (2006) 950–957. [38] R. Du et al., Robotics enabled in-home environment screening for fall risks, in: [70] A. Zijlstra et al., Do dual tasks have an added value over single tasks for Proceedings of the 2014 Workshop on Mobile Augmented Reality and Robotic balance assessment in fall prevention programs? A mini-review, Gerontology Technology-based Systems, ACM, 2014. 54 (1) (2007) 40–49. [39] J.A. Garcia et al., Assessing the Kinect’s capabilities to perform a time-based [71] A.F. Ambrose, G. Paul, J.M. Hausdorff, Risk factors for falls among older adults: clinical test for fall risk assessment in older people, in: Entertainment a review of the literature, Maturitas 75 (1) (2013) 51–61. Computing – ICEC 2014, Springer, 2014, pp. 100–107. [72] S. Hignett, T. Masud, A review of environmental hazards associated with in- [40] M.J. Taylor et al., Nintendo Wii as a training tool in falls prevention patient falls, Ergonomics 49 (5–6) (2006) 605–616. rehabilitation: case studies, J. Am. Geriatr. Soc. 60 (9) (2012) 1781–1783. [73] Y. Shi, Y. Shi, X. Wang, Fall detection on mobile phones using features from a [41] S. Uzor, L. Baillie, D. Skelton, Senior designers: empowering seniors to design ﬁve-phase model, in: 9th International Conference on Ubiquitous Intelligence enjoyable falls rehabilitation tools, in: Proceedings of the SIGCHI Conference & Computing and 9th International Conference on Autonomic & Trusted on Human Factors in Computing Systems, ACM, 2012. Computing (UIC/ATC), IEEE, 2012. 344 J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 [74] C.S. Bell et al., Effects of Nintendo Wii on quality of life, social relationships, [102] D. Singh et al., Effects of balance-focused interactive games compared to and conﬁdence to prevent falls, Phys. Occup. Ther. Geriatr. 29 (3) (2011) 213– therapeutic balance classes for older women, Climacteric 16 (1) (2012) 141– 221. 146. [75] Y.-Y. Chao et al., The feasibility of an intervention combining self-efﬁcacy [103] S.T. Smith, T.A. Davies, J. Lennox, Step Training System: an ICT solution to theory and Wii Fit exergames in assisted living residents: a pilot study, measure and reduce fall risk in older adults, in: 35th Annual International Geriatr. Nurs. 34 (5) (2013) 377–382. Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), [76] H. Kayama et al., Effect of a Kinect-based exercise game on improving IEEE, 2013. executive cognitive performance in community-dwelling elderly, in: 7th [104] D. Sparrow et al., Increases in muscle strength and balance using a resistance International Conference on Pervasive Computing Technologies for training program administered via a telecommunications system in older Healthcare (PervasiveHealth), IEEE, 2013. adults, J. Gerontol. Ser. A: Biol. Sci. Med. Sci. (2011) glr138. [77] Y. Pisan, J.G. Marin, K.F. Navarro, Improving lives: using Microsoft Kinect to [105] L.M. Taylor et al., Activity and energy expenditure in older people playing predict the loss of balance for elderly users under cognitive load, in: active video games, Arch. Phys. Med. Rehabil. 93 (12) (2012) 2281–2286. Proceedings of the 9th Australasian Conference on Interactive [106] M. van Diest et al., Suitability of Kinect for measuring whole body movement Entertainment: Matters of Life and Death, ACM, 2013. patterns during exergaming, J. Biomech. 47 (12) (2014) 2925–2932. [78] B.N. Ferreira, V. Guimaraes, H. Sereno Ferreira, Smartphone based fall [107] R. Visvanathan et al., Framework for preventing falls in acute hospitals using prevention exercises, in: 15th International Conference on e-Health passive sensor enabled radio frequency identiﬁcation technology, in: Annual Networking, Applications & Services (Healthcom), IEEE, 2013. International Conference of the IEEE Engineering in Medicine and Biology TM [79] R.J. Reed-Jones et al., WiiFit plus balance test scores for the assessment Society (EMBC), IEEE, 2012. of balance and mobility in older adults, Gait Posture 36 (3) (2012) 430– [108] W. Young et al., Assessing and training standing balance in older adults: a 433. novel approach using the ‘Nintendo Wii’Balance Board, Gait Posture 33 (2) [80] C. Bailey, V. Buckley, Recruiting and retaining older persons within a home- (2011) 303–305. based pilot study using movement sensors, Health Soc. Care Commun. 19 (1) [109] S. Albiol-Pérez et al., Use of the Wii balance board system in vestibular (2011) 98–105. rehabilitation, in: Proceedings of the 13th International Conference on [81] E. Bainbridge et al., The effects of the Nintendo Wii Fit on community- Interacción Persona-Ordenador, ACM, 2012. dwelling older adults with perceived balance deﬁcits: a pilot study, Phys. [110] A.S.M. Mosa, I. Yoo, L. Sheets, A systematic review of healthcare applications Occup. Ther. Geriatr. 29 (2) (2011) 126–135. for smartphones, BMC Med. Inform. Decis. Mak. 12 (1) (2012) 67. [82] K.A. Bieryla, N.M. Dold, Feasibility of Wii Fit training to improve clinical [111] S. Mellone et al., Smartphone-based solutions for fall detection and measures of balance in older adults, Clin. Interv. Aging 8 (2013) 775. prevention: the FARSEEING approach, Z. Gerontol. Geriatr. 45 (8) (2012) [83] W.-C. Chou et al., Design and assessment of a real-time accelerometer-based 722–727. lying-to-sit sensing system for bed fall prevention, in: IEEE International [112] J. Garcia Marin, K. Felix Navarro, E. Lawrence, Serious games to improve the Conference on Systems, Man, and Cybernetics (SMC), IEEE, 2013. physical health of the elderly: a categorization scheme, in: CENTRIC 2011, [84] E.D. de Bruin, Feasibility of strength-balance training extended with The Fourth International Conference on Advances in Human-oriented and computer game dancing in older people; does it affect dual task costs of Personalized Mechanisms, Technologies, and Services, 2011. walking?, J Novel Physiother. (2011). [113] A. Darzi (Ed.), High Quality Care for All NHS Next Stage Review Final Report, [85] W.O. de Morais, N. Wickstrom, A serious computer game to assist Tai Chi Department of Health, London, 2008. training for the elderly, in: IEEE 1st International Conference on Serious [114] S. Almer et al., Assessment test framework for collecting and evaluating fall- Games and Applications for Health (SeGAH), IEEE, 2011. related data using mobile devices, in: Computers Helping People with Special [86] J. Doyle et al., BASE – an interactive technology solution to deliver balance Needs, Springer, 2012, pp. 83–90. and strength exercises to older adults, in: 4th International Conference on-NO [115] C. Barelle, E. Vellidou, D. Koutsouris, The BIOTELEKINESY home care service: a PERMISSIONS Pervasive Computing Technologies for Healthcare holistic concept to prevent fall among the elderly based on ICT and computer (PervasiveHealth), IEEE, 2010. vision, in: 10th IEEE International Conference on Information Technology and [87] C. Duclos et al., Dynamic stability requirements during gait and standing Applications in Biomedicine (ITAB), IEEE, 2010. exergames on the Wii Fit system in the elderly, J. Neuroeng. Rehabil. 9 (2012) [116] M. Brell et al., A mobile robot for self-selected gait velocity assessments in 28. assistive environments: a robotic driven approach to bring assistive [88] K.M. Gerling, J. Schild, M. Masuch, Exergame design for elderly users: the case technologies into established homes, in: Proceedings of the 3rd study of SilverBalance, in: Proceedings of the 7th International Conference on International Conference on PErvasive Technologies Related to Assistive Advances in Computer Entertainment Technology, ACM, 2010. Environments, ACM, 2010. [89] M. Grifﬁn et al., Using the Nintendo Wii as an intervention in a falls [117] P.E. Cuddihy et al., Radar walking speed measurements of seniors in their prevention group, J. Am. Geriatr. Soc. 60 (2) (2012) 385–387. apartments: technology for fall prevention, in: Annual International [90] S. Hardy, S. Göbel, R. Steinmetz, Adaptable and personalized game-based Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) training system for fall prevention, in: Proceedings of the 21st ACM , IEEE, 2012. International Conference on Multimedia, ACM, 2013. [118] B.R. Greene et al., Body-worn sensor based surrogates of minimum ground [91] J. Hilbe et al., Development and alarm threshold evaluation of a side rail clearance in elderly fallers and controls, in: Annual International Conference integrated sensor technology for the prevention of falls, Int. J. Med. of the IEEE Engineering in Medicine and Biology Society, EMBC, IEEE, 2011. Informatics 79 (3) (2010) 173–180. [119] A.J.A. Majumder et al., iPrevention: towards a novel real-time smartphone- [92] E.T. Horta et al., Real time falls prevention and detection with biofeedback based fall prevention system, in: Proceedings of the 28th Annual ACM monitoring solution for mobile environments, in: IEEE 15th International Symposium on Applied Computing, ACM, 2013. Conference on e-Health Networking, Applications & Services (Healthcom), [120] O. Rawashdeh et al., Development of a low-cost fall intervention system for 2013. hospitalized dementia patients, in: IEEE International Conference on Electro/ [93] B. Lange et al., Development of an interactive stepping game to reduce falls in Information Technology (EIT), IEEE, 2012. older adults, Int. J. Disabil. Hum. Dev. 10 (4) (2011) 331–335. [121] S.J. Redmond et al., Automatic segmentation of triaxial accelerometry signals [94] A.J.A. Majumder et al., SmartPrediction: a real-time smartphone-based fall for falls risk estimation, in: Annual International Conference of the IEEE risk prediction and prevention system, in: Proceedings of the 2013 Research Engineering in Medicine and Biology Society (EMBC), IEEE, 2010. in Adaptive and Convergent Systems, ACM, 2013. [122] G.R.H. Regterschot et al., Sensitivity of sensor-based sit-to-stand peak power [95] D. Marisa Ferrari et al., A pilot study testing a fall prevention intervention for to the effects of training leg strength, leg power and balance in older adults, older adults: determining the feasibility of a ﬁve-sensor motion detection Gait Posture 39 (1) (2014) 303–307. system, J. Gerontol. Nurs. 38 (1) (2012) 13–16. [123] F. Riva et al., Estimating fall risk with inertial sensors using gait stability [96] A. Mirelman et al., V-TIME: a treadmill training program augmented by measures that do not require step detection, Gait Posture 38 (2) (2013) 170– virtual reality to decrease fall risk in older adults: study design of a 174. randomized controlled trial, BMC Neurol. 13 (1) (2013) 15. [124] D. Schoene et al., A novel video game-based device for measuring stepping [97] A. Mirelman et al., Treadmill training with Virtual Reality to decrease risk of performance and fall risk in older people, Arch. Phys. Med. Rehabil. 92 (6) falls in idiopathic fallers: a pilot Study, in: International Conference on Virtual (2011) 947–953. Rehabilitation (ICVR), IEEE, 2011. [125] D.K. Singh et al., Participating in a virtual reality balance exercise program [98] M.-D. Otis, B. Menelas, Toward an augmented shoe for preventing falls related can reduce risk and fear of falls, Maturitas 73 (3) (2012) 239–243. to physical conditions of the soil, in: IEEE International Conference on [126] C. Soaz, M. Daumer, A feasibility study for the integration of 3D Systems, Man, and Cybernetics (SMC), IEEE, 2012. accelerometry in fall risk assessment, Biomed. Tech. 57 (2012) 1. [99] B. Rajaratnam et al., Does the inclusion of virtual reality games within [127] A. Staranowicz, G.R. Brown, G.-L. Mariottini, Evaluating the accuracy of a conventional rehabilitation enhance balance retraining after a recent episode mobile Kinect-based gait-monitoring system for fall prediction, in: of stroke?, Rehabil Res. Pract. (2013). Proceedings of the 6th International Conference on PErvasive Technologies [100] D. Schoene et al., A randomized controlled pilot study of home-based step Related to Assistive Environments, ACM, 2013. training in older people using videogame technology, PLoS One 8 (3) (2013) [128] A. Weiss et al., Does the evaluation of gait quality during daily life provide e57734. insight into fall risk? A novel approach using 3-day accelerometer recordings, [101] P. Silveira et al., Tablet-based strength-balance training to motivate and Neurorehabil. Neural Repair 27 (8) (2013) 742–752. improve adherence to exercise in independently living older people: a phase [129] W. Zhang et al., Chair rise transfer detection and analysis using a pendant II preclinical exploratory trial, J. Med. Internet Res. 15 (8) (2013). sensor: an algorithm for fall risk assessment in older people, in: 36th Annual J. Hamm et al. / Journal of Biomedical Informatics 59 (2016) 319–345 345 International Conference of the IEEE Engineering in Medicine and Biology [153] Y. Li et al., Acoustic fall detection using a circular microphone array, in: Society (EMBC), IEEE. Annual International Conference of the IEEE Engineering in Medicine and [130] W. Zijlstra et al., A body-ﬁxed-sensor-based analysis of power during sit-to- Biology Society (EMBC), IEEE, 2010. stand movements, Gait Posture 31 (2) (2010) 272–278. [154] P. Martín et al., Multi-agent system for detecting elderly people falls through [131] B.R. Greene et al., Evaluation of falls risk in community-dwelling older adults mobile devices, in: Ambient Intelligence-Software and Applications, Springer, using body-worn sensors, Gerontology 58 (5) (2012) 472–480. 2011, pp. 93–99. [132] S.N. Robinovitch et al., Video capture of the circumstances of falls in elderly [155] G. Mastorakis, D. Makris, Fall detection system using Kinect’s infrared sensor, people residing in long-term care: an observational study, Lancet 381 (9860) J. Real-Time Image Proc. (2012) 1–12. (2013) 47–54. [156] S. Mehner, R. Klauck, H. Koenig, Location-independent fall detection with [133] S. Abbate, M. Avvenuti, J. Light, MIMS: a minimally invasive monitoring smartphone, in: Proceedings of the 6th International Conference on PErvasive sensor platform, IEEE Sensors J. 12 (3) (2012) 677–684. Technologies Related to Assistive Environments, ACM, 2013. [134] M.V. Albert et al., Fall classiﬁcation by machine learning using mobile phones, [157] R. Paoli et al., A system for ubiquitous fall monitoring at home via a wireless PLoS One 7 (5) (2012) e36556. sensor network and a wearable mote, Expert Syst. Appl. 39 (5) (2012) 5566– [135] M.A. Aud et al., Developing a sensor system to detect falls and summon 5575. assistance, J. Gerontol. Nurs. 36 (7) (2010). [158] A. Papadopoulos, C. Crump, B. Wilson, Comprehensive home monitoring [136] A. Bagnasco, A.M. Scapolla, V. Spasova, Design, implementation and system for the elderly, in: Wireless Health 2010, ACM, 2010. experimental evaluation of a wireless fall detector, in: Proceedings of the [159] R. Planinc, M. Kampel, Introducing the use of depth data for fall detection, 4th International Symposium on Applied Sciences in Biomedical and Pers. Ubiquit. Comput. 17 (6) (2013) 1063–1072. Communication Technologies, ACM, 2011. [160] M.J. Rantz et al., Automated fall detection with quality improvement [137] F. Busching et al., Fall detection on the road, in: IEEE 15th International ‘‘rewind” to reduce falls in hospital rooms, J. Gerontol. Nurs. (2013) 1–5. Conference on e-Health Networking, Applications & Services (Healthcom), [161] L. Ren, Q. Zhang, W. Shi, Low-power fall detection in home-based 2013. environments, in: Proceedings of the 2nd ACM international workshop on [138] J. Cabestany et al., FATE: one step towards an automatic aging people fall Pervasive Wireless Healthcare, ACM, 2012. detection service, in: Proceedings of the 20th International Conference Mixed [162] O. Sahota et al., REFINE (REducing Falls in In-patieNt Elderly) using bed and Design of Integrated Circuits and Systems (MIXDES), IEEE, 2013. bedside chair pressure sensors linked to radio-pagers in acute hospital care: a [139] J. Dai et al., PerFallD: a pervasive fall detection system using mobile phones, randomised controlled trial, Age Ageing 1 (2013) aft155. in: 8th IEEE International Conference on Pervasive Computing and [163] W.-Y. Shieh, J.-C. Huang, Falling-incident detection and throughput Communications Workshops (PERCOM Workshops), IEEE, 2010. enhancement in a multi-camera video-surveillance system, Med. Eng. Phys. [140] L. Della Toffola et al., Development of a platform to combine sensor networks 34 (7) (2012) 954–963. and home robots to improve fall detection in the home environment, in: [164] J. Shim et al., The development of a detection system for seniors’ accidental Annual International Conference of the IEEE Engineering in Medicine and fall from bed using cameras, in: Proceedings of the 5th International Biology Society, EMBC, IEEE, 2011. Conference on Ubiquitous Information Management and Communication, [141] P. Fahmi, V. Viet, C. Deok-Jai, Semi-supervised fall detection algorithm using ACM, 2011. fall indicators in smartphone, in: Proceedings of the 6th International [165] Y. Tang et al., PPCare: a personal and pervasive health care system for the Conference on Ubiquitous Information Management and Communication, elderly, in: 9th International Conference on Ubiquitous Intelligence & ACM, 2012. Computing and 9th International Conference on Autonomic & Trusted [142] S.-H. Fang, Y.-C. Liang, K.-M. Chiu, Developing a mobile phone-based fall Computing (UIC/ATC), IEEE, 2012. detection system on android platform, in: Computing, Communications and [166] M. Terroso et al., Active assistance for senior healthcare: a wearable system Applications Conference (ComComAp), IEEE, 2012. for fall detection, in: 8th Iberian Conference on Information Systems and [143] M. Ferrari et al., Clinical feasibility trial of a motion detection system for fall Technologies (CISTI), IEEE, 2013. prevention in hospitalized older adult patients, Geriatr. Nurs. 33 (3) (2012) [167] V.Q. Viet, G. Lee, D. Choi, Fall detection based on movement and smart phone 177–183. technology, in: IEEE RIVF International Conference on Computing and [144] G.K. Fourlas, I. Maglogiannis, Human movement detection using attitude and Communication Technologies, Research, Innovation, and Vision for the heading reference system, in: Proceedings of the 7th International Future (RIVF), IEEE, 2012. Conference on PErvasive Technologies Related to Assistive Environments, [168] F. Werner et al., Fall detection with distributed ﬂoor-mounted accelerometers: ACM, 2014. an overview of the development and evaluation of a fall detection system [145] Y. He, Y. Li, S.-D. Bao, Fall Detection by built-in tri-accelerometer of within the project eHome, in: 5th International Conference on Pervasive smartphone, in: IEEE-EMBS International Conference on Biomedical and Computing Technologies for Healthcare (PervasiveHealth), IEEE, 2011. Health Informatics (BHI), IEEE, 2012. [169] Z. Zhang et al., Design of an unobtrusive wireless sensor network for [146] J. He, C. Hu, Y. Li, An autonomous fall detection and alerting system based on nighttime falls detection, in: Annual International Conference of the IEEE mobile and ubiquitous computing, in: IEEE 10th International Conference on Engineering in Medicine and Biology Society, EMBC, IEEE, 2011. and 10th International Conference on Autonomic and Trusted Computing [170] J. Leake et al., Is photoplethysmography-derived pulse shape useful for fall (UIC/ATC) Ubiquitous Intelligence and Computing, IEEE, 2013. detection?, in: Proceedings of the 7th International Conference on PErvasive [147] M. Kepski, B. Kwolek, Fall detection on embedded platform using Kinect and Technologies Related to Assistive Environments, ACM, 2014 wireless accelerometer, in: Computers Helping People with Special Needs, [171] J. Dai et al., Mobile phone-based pervasive fall detection, Pers. Ubiquit. Springer, 2012, pp. 407–414. Comput. 14 (7) (2010) 633–643. [148] M. Kepski, B. Kwolek, Unobtrusive fall detection at home using Kinect sensor, [172] B.-R. Chen, J. Gwin, Fall detection and risk of falling assessment with in: Computer Analysis of Images and Patterns, Springer, 2013. wearable sensors, in: Proceedings of the Conference on Wireless Health, [149] G.A. Koshmak, M. Linden, A. Loutﬁ, Evaluation of the android-based fall ACM, 2012. detection system with physiological data monitoring, in: 35th Annual [173] U. Cortés et al., AI based fall management services – the role of the i-walker in I- International Conference of the IEEE Engineering in Medicine and Biology DONTFALL, in: Advances in Artiﬁcial Intelligence, Springer, 2013, pp. 395–406. Society (EMBC), 2013. [174] D.C. Ranasinghe et al., Towards falls prevention: a wearable wireless and [150] M.A. Laguna, J. Finat, Remote monitoring and fall detection: multiplatform battery-less sensing and automatic identiﬁcation tag for real time monitoring java based mobile applications, in: Ambient Assisted Living, Springer, 2011, of human movements, in: Annual International Conference of the IEEE pp. 1–8. Engineering in Medicine and Biology Society (EMBC), IEEE, 2012. [151] R.Y. Lee, A.J. Carlisle, Detection of falls using accelerometers and mobile [175] W.A. König, R. Rädle, H. Reiterer, Squidy: A Zoomable Design Environment for phone technology, Age Ageing 40 (6) (2011) 690–696. Natural User Interfaces, ACM, 2009. [152] A. Leone, G. Diraco, P. Siciliano, Detecting falls with 3D range camera in [176] A. Atwal, A. Money, M. Harvey, Occupational therapists’ views on using a ambient assisted living applications: a preliminary study, Med. Eng. Phys. 33 virtual reality interior design application within the pre-discharge home visit (6) (2011) 770–781. process, J. Med. Internet Res. 16 (12) (2014).

http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

Journal of Biomedical Informatics Unpaywall

http://www.deepdyve.com/lp/unpaywall/fall-prevention-intervention-technologies-a-conceptual-framework-and-nWOEQQZyy6

Fall prevention intervention technologies: A conceptual framework and survey of the state of the art

Loading next page...

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher: Unpaywall
ISSN: 1532-0464
DOI: 10.1016/j.jbi.2015.12.013
Publisher site: See Article on Publisher Site

Abstract

Journal

Journal of Biomedical Informatics – Unpaywall

Published: Feb 1, 2016

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Fall prevention intervention technologies: A conceptual framework and survey of the state of the art

Fall prevention intervention technologies: A conceptual framework and survey of the state of the art

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Fall prevention intervention technologies: A conceptual framework and survey of the state of the art

Fall prevention intervention technologies: A conceptual framework and survey of the state of the art

References

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies