Lu Wang
Abstract
Informal caregivers of people living with Alzheimer’s disease or related dementias (PLWD) face challenges like obtaining personalized information and monitoring PLWD’s health. Rapid advancements in technology, especially in sophisticated and controversial areas like artificial intelligence (AI), prompted our study to assess AI’s potential and challenges in supporting the needs of informal caregivers of PLWD. Caregiving activities require dynamic, often unpredictable, and sometimes emotionally draining tasks that deal with a large amount of information. We conducted a systematic review to understand what AI technology has been developed to support informal caregivers of PLWD. We collected 920 papers from ACM Digital Library, IEEE Xplore, and PubMed. Screening and eligibility evaluation resulted in 16 papers for full-text review. We present which documented needs of informal caregivers have been explored by the existing research, and the contexts of the AI solutions including interfaces, data, and algorithms, as well as their effectiveness, challenges, and limitations.
1 INTRODUCTION
Informal caregivers refer to individuals providing unpaid care for a friend or relative who requires special attention [14, 18, 60]. According to 2023 Alzheimer’s Disease Facts and Figures, more than 11 million Americans as informal caregivers provide unpaid care for people with Alzheimer’s disease or related dementias (PLWD) [8]. Alzheimer’s disease, a fatal degenerative condition caused by damage to nerve cells (neurons) in the brain, is currently incurable and has no known treatments to halt, prevent, or cure it [7, 8, 62]. It is the most common contributor to dementia, which, as an overall term for a particular group of symptoms, refers to difficulties with memory, language, problem-solving, and other thinking skills [7, 8]. However, informal caregivers often don’t have formal training, knowledge, and skills about caregiving activities, thus facing the challenges of time- and resource-consuming caregiving activities to maintain quality care for PLWD [60, 68].
A scoping review extracted 8 themes with 18 specific needs of informal caregivers of PLWD, covering the needs related to caregiving tasks, relationships with formal service providers, housing, juggling responsibilities, financial costs, personal health, family relationships, and planning [55]. Telephone-based, video-based, and computer-based information and communication technologies (ICT) [36, 41] and assistive technology, such as wearable sensors [25], timers [27], alarms [51], clocks [48], and calendars [6] have been developed to support informal caregivers of PLWD in caregiving tasks [43, 72], personal health [28, 36, 41], family relationships [40, 63], and planning [15, 28]. There is also growing interest in the CHI community to investigate informal caregivers’ perspectives and needs, such as examining the roles of informal caregivers when engaging PLWD with a videogame-based ICT system [71] or online activities [53], and supporting informal caregivers in communication and relationship enhancement with PLWD through interactive sound players [29], cooperative games [46], competition and skill games [46], communication systems with postcards [70], and conversational agents [77]. Researchers found that the responsibilities of informal caregivers can vary from moment to moment, depending on the conditions of PLWD [53]. As PLWD’s conditions progress from mild or focal neuropsychological deficits to global deficits in multiple domains such as speech production, attention, or other challenges [77], the needs of informal caregivers of PLWD and their caregiving tasks are not discrete states or unchanging conditions [53]. This dynamic nature necessitates that informal caregivers flexibly adapt to the shifting demands of caregiving tasks [53]. Besides, cultural factors and genders of informal caregivers and care recipients affect the dynamics and complexity of the caregiving process [23]. With the integration of new technologies, for example, artificial intelligence (AI), into daily lives, which introduces new opportunities and challenges in the context of caregiver support, continuous evaluation and adaptation of new technologies are essential to fully meet the complex and changing needs of informal caregivers.
AI is defined as “a new generation of technologies capable of interacting with the environment by (a) gathering information from outside (including from natural language) or from other computer systems; (b) interpreting this information, recognizing patterns, inducing rules, or predicting events; (c) generating results, answering questions, or giving instructions to other systems; and (d) evaluating the results of their actions and improving their decision systems to achieve specific objectives” [19]. It has been applied in healthcare across a variety of scenarios [61], including health services management [1, 2, 11, 17, 22, 61], predictive medicine [32, 33, 35, 49], clinical decision-making [31, 37, 44], and patient data and diagnostics [17, 33, 56, 57, 58, 64, 73], demonstrating how the healthcare system can benefit from AI-supported functionalities, for example, heterogeneous health information analysis [1, 2, 11, 61, 69], real-time information updates [17, 22], automatic optimization [17, 22], and personalized treatment plans [2, 16, 61]. Such benefits could help informal caregivers in terms of personalized information retrieval, managing caregiving activities, and monitoring PLWD.
To understand how AI technology has been and can be better applied and designed to support informal caregivers of PLWD, this study aims to delve into existing literature to comprehensively understand both the met and unmet needs of informal caregivers of PLWD in the studies on AI solutions, paving the way for future research to address the gaps and limitations of current research. Besides, we applied a framework of informal caregivers’ needs [55] as guidance to examine existing AI solutions and specifically, to investigate the following research questions:
• | RQ1: What needs of informal caregivers of PLWD have or have not been investigated by AI solutions? | ||||
• | RQ2: What are the contexts of AI solutions for supporting the needs of informal caregivers of PLWD in terms of interfaces, data, and algorithms? | ||||
• | RQ3: What are the effectiveness, challenges, and limitations of these AI solutions? | ||||
Our contributions include: 1) providing an overview of existing research on AI solutions for informal caregivers of PLWD, 2) informing the gaps in existing research in terms of the limitations of current research and the unaddressed needs of informal caregivers of PLWD, 3) discussing the challenges and opportunities of AI solutions in supporting informal caregivers of PLWD, and 4) providing a research agenda for future work.
2 METHODS
For the methods of this systematic review, we followed the phases as recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [24, 52], shown in Figure 1: literature search, screening filter, eligibility evaluation, and analysis. We collected papers from multiple specialized databases, ACM Digital Library, IEEE Xplore, and PubMed, each focusing on specific disciplines [30]. Appendix A shows more details of the process.
Figure 1: Diagram showing the adapted PRISMA.
| Themes | Needs | [3] | [4] | [38] | [76] | [20] | [42] | [67] | [13] | [47] | [59] | [65] | [5] | [21] | [26] | [50] | [74] | Total |
| 1. Caregiving | 1-a. Physical/nursing care | ● | ● | 2 | ||||||||||||||
| 1-b. Household work | ● | 1 | ||||||||||||||||
| 1-c. Supervision/ support | ● | ● | ● | ● | ● | ● | ● | ● | ● | 9 | ||||||||
| 1-d. Coordination | 0 | |||||||||||||||||
| 1-e. Help received from others (informal & formal) | 0 | |||||||||||||||||
| 2. Relationship with Formal Service Providers | 0 | |||||||||||||||||
| 3. Housing | 0 | |||||||||||||||||
| 4. Juggling Responsibilities | ● | 1 | ||||||||||||||||
| 5. Financial Costs | 0 | |||||||||||||||||
| 6. Personal | 6-a. Physical health | 0 | ||||||||||||||||
| 6-b. Emotional health | ● | ● | ● | ● | 4 | |||||||||||||
| 7. Relationships | 7-a. With care recipient | ● | 1 | |||||||||||||||
| 7-b. With family | ● | 1 | ||||||||||||||||
| 8. Planning | 8-a. Crises planning | 0 | ||||||||||||||||
| 8-b. Future planning | ● | 1 | ||||||||||||||||
| 8-c. Information about dementia and dementia care | ● | ● | 2 | |||||||||||||||
| 8-d. Information about professional support and formal services | ● | 1 | ||||||||||||||||
| 8-e. Information about legal regulation in caring | 0 | |||||||||||||||||
Table 1: An Overview of Investigated and Uninvestigated Needs based on the Framework of Needs of Caregivers of PLWD [55]
2.1 Search Query
We tailored our strategies to each database. Considering the broad applicability of AI across various technologies, we incorporated solution-related keywords exclusively in the PubMed search to acquire more relevant literature. We included customer-oriented solution keywords as well, such as "telehealth" and "telemedicine," since these solutions could also embed AI when delivering services to clients. We excluded such keywords for databases in technology, ACM Digital Library, and IEEE Xplore. This approach yielded 96 papers from ACM Digital Library, 186 from IEEE Xplore, and 655 from PubMed, encompassing literature published within the time frame from January 1st, 2013, to December 31st, 2022.
For ACM Digital Library and IEEE Xplore, our search strategy combined two components with "AND" between: (1) caregivers-related keywords, carer* OR caregiver*, and (2) Dementia-related keywords, dementia OR Alzheimer’s. We also constrained the document types to be Research Articles, Proceedings, and Journals for ACM Digital Library and Conferences and Journals for IEEE Xplore. For PubMed, we applied a search query combining three components with "AND" between them: (1) caregivers-related keywords, carer* OR informal caregiver* OR family caregiver*, (2) Dementia-related keywords, dementia OR Alzheimer’s, and (3) solutions-related keywords, AI OR artificial intelligence* OR system* OR techn* OR application* OR tool* OR agent* OR telemedicine OR tele medicine OR telehealth OR tele health OR telemonitor* OR tele monitor* OR ehealth OR e-health OR mhealth OR m-health. We constrained Free full text and Full text for the collection.
2.2 Screening Filter and Eligibility Evaluation
Among the 937 total papers from all three databases, we filtered out 86 papers due to duplicates, research protocols, or having no abstract, resulting in 851 papers for screening. The inclusion criteria require that the paper should focus on (1) informal caregivers as the target users; (2) Alzheimer’s disease, dementia, or mild cognitive impairments; and (3) AI-integrated solutions, including keywords related to (a) AI algorithms, such as machine learning, unsupervised learning, deep learning, recommender systems, recommendation approaches, and task-related terms (e.g., classification), (b) products known to be AI-embedded, such as Google Home, Alexa, and Siri; and (c) intelligence, for example, “smart” and “aware”.
Two researchers with research experience in informal caregiving for PLWD initially divided the 851 papers, encountering 55 overlaps for criteria (1) and (2), achieving Kappa scores [45] of 0.924 and 0.845, respectively. This process left 384 papers meeting the first two criteria. Subsequently, two researchers knowledgeable about AI split these 384 papers, noting 20 overlaps for criterion (3). After further filtration, the remaining 32 papers achieved a Kappa score of 1. These 32 papers underwent title and abstract eligibility evaluation by four researchers. Full-text reviews were then conducted to assess compliance with inclusion and exclusion criteria, with another researcher, who possesses research experience in informal caregiving for PLWD as well as knowledge of AI, verifying any papers not meeting these criteria. Ultimately, 16 papers remained after the full-text review.
2.3 Data Extraction and Analysis
To answer RQ1, we applied the framework of the needs of informal caregivers of PLWD [55]. Appendix A.4 introduces the development history of the framework and definitions of each need. To extract the related information from the papers for RQ2 and RQ3, the research team coded two randomly selected papers together to establish consistency in the data extraction process for each code, including the PLWD’s stages, demographics of participants involved in the study, such as gender, age, and race, research goals, the informal caregivers’ needs investigated by the study, interfaces of the AI solutions, data acquired by the studies, AI algorithms, evaluation outcomes, and identified challenges and limitations. Then, four researchers equally divided the 16 papers and independently coded them. Among the four researchers, one lead researcher went through all 16 papers to confirm the validity of the codes.
3 FINDINGS
Out of the sixteen papers, seven papers (43.75%) included informal caregivers [3, 4, 20, 38, 42, 67, 76], among which two studies also involved PLWD [3, 4], two studies also involved other participants [38, 76], and the rest three studies only involved informal caregivers [20, 42, 67]. Four papers (25%) involved participants other than informal caregivers or did not inform whether the participants were informal caregivers [13, 47, 59, 65]. Five papers (31.25%) didn’t have information of participants [5, 21, 26, 50, 74], among which four papers did not involve human participants [5, 21, 50, 74], and one paper did not report the profiles of the participants [26]. PLWD’s stages and the demographics of participants were largely missing in these studies and more information is shown in Appendix B.
| Themes | Needs | Sensors | Self-reports | Camera videos | Trusted resources | Frames of soap opera | Simulated data |
| 1. Caregiving | 1-a. Physical/nursing care | ❉ ☞ ❤[42] | ☞[47] | ||||
| 1-b. Household work | ❤[38] | ||||||
| 1-c. Supervision/support | ❉ ☞ ❤[42] | ❉ ☞[3, 4, 59] | ❉ ☞[59] | NR[74] | ❉[5] | ||
| 4. Juggling Responsibilities | ❉ ☞ ❤[42] | ||||||
| 6. Personal Health | 6-b. Emotional health | ❤[76] | |||||
| 7. Relationships | 7-a. With care recipient | ❉ ☞ ❤[42] | |||||
| 7-b. With family | ❉ ☞ ❤[42] | ||||||
| 8. Planning | 8-b. Future planning | NR[67] | |||||
| 8-c. Information about dementia and dementia care | |||||||
| 8-d. Information about professional support and formal services | ❉[50] | ❉[50] | |||||
❉ - Interfaces: Software Applications; ☞ - Interfaces: Wearable Devices; ❤ - Interfaces: Smart Devices at Home; ❦ - Interfaces: Smart Homes; NR - Interfaces Not Reported
Table 2: An Overview of Data Acquired and Interfaces Developed by the Studies
❉ - Interfaces: Software Applications; ☞ - Interfaces: Wearable Devices; ❤ - Interfaces: Smart Devices at Home; ❦ - Interfaces: Smart Homes; NR - Interfaces Not Reported
3.1 Needs of Informal Caregivers of PLWD Investigated and Uninvestigated by AI Solutions (RQ1)
As the Table 1 shows, of the 16 AI solutions for supporting informal caregivers of PLWD, the investigated needs were supervision/support (9, 56.25%), emotional health (4, 25.00%), physical/nursing care (2, 12.50%), information about dementia and dementia care (2, 12.50%), household work (1, 6.25%), juggling responsibilities (1, 6.25%), relationships with the care recipient (1, 6.25%) and with family (1, 6.25%), future planning (1, 6.25%), and information about professional support and formal services (1, 6.25%). The uninvestigated needs of informal caregivers of PLWD by AI solutions were coordination of formal services, help received from others (in-formal & formal), relationship with formal service providers, housing, financial costs, physical health, crises planning, and information about legal regulation in caring. This may stem from the difficulties in investigating the solutions with multiple stakeholders, establishing connections with relevant institutions like hospitals and legal departments, and explaining AI for decision-making.
To support informal caregivers of PLWD in supervision/support, researchers applied AI to infer and predict critical events of PLWD, such as agitation [3, 4], anxiety [59], behavioral symptoms [74], and falls [13]. Researchers also applied AI to constantly monitor PLWD [42, 65]. Helmy et al. investigated how AI can help manage PLWD’s wandering behaviors by allowing informal caregivers to designate certain activities (or zones) as dangerous and triggering alert (with activity information) when trespassing was detected [26]. Andersen et al. also developed an AI-based application to involve caregivers and ordinary citizens who volunteered to locate and offer help to PLWD when PLWD got lost [5].
To support informal caregivers of PLWD in emotional health, Goins et al. applied AI algorithms to assess caregivers’ burden based on survey data [20, 21]. Researchers also applied AI to provide emotional health support through smart dogs [42] and humanoid social robots [76]. Moreover, Nakazawa et al. applied AI to analyze the first-person videos recorded by caregivers and evaluated caregivers’ skill levels of tender dementia-care techniques to provide physical/nursing care support [47]. Lyu et al. applied AI to help informal caregivers monitor PLWD’s health conditions through wearable devices while providing companionship through smart dogs [42]. Yuan et al. developed a robot to deliver information about dementia and dementia care [76] and Li et al. applied Amazon Alexa-based devices to provide personalized education and guidance for informal caregivers of PLWD [38].
Lyu et al. found that smart dogs can also help informal caregivers of PLWD with juggling responsibilities by providing more freedom for informal caregivers, enhancing relationships with family, and relationships with care recipients [42]. Li et al. applied AI-based Alexa devices to provide recommendations to informal caregivers on food, nutrition, and cooking for PLWD [38]. Moreover, Oliva et al. investigated applying AI to provide tailored recommendations for professional support and formal services based on user preferences and personal details, healthcare professional recommendations, and caregivers’ opinions through explicit ratings given to interventions and actions such as reviewing or sharing [50]. Sunmoo et al. also used AI models to examine the relationship between demographics such as race and outcomes of social work services, thus assisting in future planning of service adoption [67].
| Themes | Needs | Algorithm Types | Evaluation Methods | |
| 1. Caregiving | 1-a. Physical/nursing care | OpenFace Library [47] Amazon Rekognition [47] | ||
| 1-b. Household work | Amazon Alexa [38] | ■ □ [38] | ||
| 1-c. Supervision/support | Android Activity Recognition API [26] | ▲ | F [3, 4, 59] | ||
| 4. Juggling Responsibilities | Adaptive Network-based Fuzzy Inference System [42] | ■ [42] | ||
| 6. Personal Health | 6-b. Emotional health | Scaled Conjugate Gradient Backpropagation Neural Network [20] Neural Network | ▲ | F [20] | |
| 7. Relationships | 7-a. With care recipient | Adaptive Network-based Fuzzy Inference System [42] | ■ [42] | |
| 7-b. With family | Adaptive Network-based Fuzzy Inference System [42] | ■ [42] | ||
| 8. Planning | 8-b. Future planning | Decision Tree [67] | ▲ | F [67] | |
| 8-c. Information about dementia and dementia care | Amazon Alexa [38] | ■ □ [38] | ■ □ [38] | |
| 8-d. Information about professional support and formal services | Hybrid Filtering Approach with a Content-based Algorithm and Rule-based Filtering [50] | \(\varnothing\) [50] | ||
NR - Algorithm Not Reported; CNN - Convolutional Neural Network; LSTM - Long Short-term Memory Network; SVM - Support Vector Machine; PCA - Principal Component Analysis; MIL - Multiple-Instance Learning
▲ - Metric-based AI Solution Performance Evaluation | F: F-score; ACC: Accuracy; P: Precision; R: Recall; O: Other Metric Measurements; ■ - Usability Study Evaluation involving Informal Caregivers; □ - Usability Study Evaluation NOT involving Informal Caregivers; ■ □ / ■ ■ - Multiple Rounds of Usability Study Evaluation; \(\varnothing\) - No Evaluation was Included
Table 3: An Overview of Algorithms and Evaluation Methods Applied by the Studies
NR - Algorithm Not Reported; CNN - Convolutional Neural Network; LSTM - Long Short-term Memory Network; SVM - Support Vector Machine; PCA - Principal Component Analysis; MIL - Multiple-Instance Learning
▲ - Metric-based AI Solution Performance Evaluation | F: F-score; ACC: Accuracy; P: Precision; R: Recall; O: Other Metric Measurements; ■ - Usability Study Evaluation involving Informal Caregivers; □ - Usability Study Evaluation NOT involving Informal Caregivers; ■ □ / ■ ■ - Multiple Rounds of Usability Study Evaluation; \(\varnothing\) - No Evaluation was Included
3.2 Contexts of AI Solutions to Support the Needs of Informal Caregivers of PLWD (RQ2)
3.2.1 Interfaces developed.
Among the sixteen AI solutions for informal caregivers of PLWD, four studies did not introduce the interfaces of their solutions [20, 21, 67, 74]. As Table 2 shows, there were four main types of interfaces: 1) software applications, 2) wearable devices, 3) smart devices at home, and 4) smart homes. Multiple interfaces could be applied by one study. Software applications included the digital platforms available on smartphones, tablet computers, and web browsers to provide tailored non-pharmacological educational interventions to caregivers and PLWD [50], applications to annotate or monitor symptoms of PLWD [3, 4, 42, 59], mobile app: Alzimio to monitor PLWD’s activity zones [26], and the application to involve volunteers for PLWD’s wandering behaviors [5]. Wearable devices were either worn by PLWD so that informal caregivers could monitor PLWD’s conditions, which included smartwatches [3, 4, 42], wearable sensors [13, 59], wearable cameras [59], and stethoscope [42], or worn by informal caregivers such as wearable cameras to evaluate the caregiving skills [47]. Smart devices at home included smart dogs [42], humanoid social robot [76], and Alexa-enabled devices [38]. Smart homes referred to the one built in the Barry Lam Hall at National Taiwan University [65].
3.2.2 Data acquired.
As shown in Table 2, we found researchers worked with six main types of data: sensor data (8, 50.00%), self-reported data (7, 43.75%), camera video data in real-time (2, 12.50%), data from trusted online resources including scientific literature and authorized websites (3, 18.75%), public media such as still frames of soap operas (1, 6.25%), and simulated data (1, 6.25%).
Sensor data was mainly used to support the needs of informal caregivers of PLWD in supervision/support, emotional health, relationships with the care recipient and family, and juggling responsibilities. Utilized sensor data included physiological data, such as motion, pulse, electroencephalography, cardiopulmonary condition, galvanic skin response, heart rate [13, 42, 59], body movement data captured by wearable sensors, for example, accelerometers affixed to certain specific positions of the body (e.g. waist, wrist, ankle, arm, hip) [3, 4, 13, 59, 65], location data captured by Bluetooth, gyroscope, GPS, and WiFi [13, 26, 65], and environmental sensor data of light, sound, and temperature [21].
Self-reported data was collected for supervision/support, emotional health, future planning, and information about professional support and formal service. Self-reported data captured in the studies included reports of the symptoms of PLWD such as agitation [3, 4, 21] and anxiety [59], informal caregivers’ self-assessed health status [20, 21], the frequency, severity, count, and distress of caregivers caused by PLWD [20], concerns of caregivers [20], perceived support from close persons of caregivers [20], demographics [20, 67], informal caregivers’ preferences, personal details, and opinions given to interventions [50], behavioral and physio-psycho-social variables, and the outcome of social work service [67].
Trusted online resources were collected to provide support for household work, emotional health, information about dementia and dementia care, and information about professional support and formal services. Such data utilized in the studies included data from scientific literature [38, 50], the United States Department of Agriculture [38], the National Heart Foundation of Australia [38], the National Institute on Aging [38], the National Institute for Health and Clinical Excellence [50], American Association of Retired Persons [76], Family Caregiver Alliance [76], Alzheimer’s Association [38, 76], Dementia.org [76], Alzheimer’s.net [38], Mayo Clinic [38, 76], and Dementia Carer site [50].
Videos in real-time captured by wearable cameras can be used to evaluate the skill levels of tender dementia-care techniques for physical/nursing care [47] and to record the number of faces in the present environment for supervision/support [59]. Frames from the soap opera Emmerdale episodes focusing on the dementia storyline were applied to recognize behavioral symptoms of PLWD [74]. Simulated data of the behavior of real users who opened the app during a time window was also utilized to test the system [5].
3.2.3 Algorithms applied.
Table 3 shows the algorithms applied to support the needs of informal caregivers of PLWD. These AI algorithms included established AI libraries and APIs, supervised learning, and unsupervised learning algorithms. Among the 16 AI solutions, two studies did not explicitly mention what algorithms they applied: Anderson et al. claimed that they applied OpenFaaS which included different AI techniques for online and offline anomaly detection [5] and Yuan et al. indicated their humanoid social robot applied AI [76]. Three studies applied established AI APIs or libraries: Helmy et al. utilized activity recognition API to recognize PLWD’s activities [26]; Nakazawa et al. used OpenFace library and Amazon Rekognition for face recognition [47]; Li et al. applied Amazon Alexa for automatic speech recognition and natural language processing [38].
There was a diversity of algorithms applied to support the needs of informal caregivers of PLWD. Supervised learning algorithms were applied for the classification and prediction of agitation [3, 4], anxiety [59], falls [13], and activities as well as conditions of PLWD [42, 65, 74], eye contact detection [47], and the burden of caregivers [20, 21], as well as the likelihood of resolution of issues by social work services [67]. These algorithms included neural network algorithms [3, 20, 21, 42, 47, 59, 74], support vector machine [4, 59, 74], decision trees [59, 67], axis-parallel rectangle for multiple instance learning [4], boosting bag-level decision stumps [4], XGBoost [13], finite state machine [65], and random forest [65]. Unsupervised learning algorithms were applied to analyze the data [47, 65] and provide recommendations [50], which included principal component analysis [47], Dirichlet process mixture model [65], N-gram model [65], dynamic Bayesian network [65], and a hybrid filtering approach consisting of a content-based algorithm and rule-based filtering [50].
3.3 Effectiveness, Challenges, and Limitations of AI Solutions in Supporting Informal Caregivers of PLWD (RQ3)
3.3.1 Effectiveness of existing AI solutions.
For the evaluations of the effectiveness of these 16 AI solutions, shown in Table 3, two studies did not evaluate their AI solutions [21, 50]. Ten studies evaluated the performances of AI solutions based on metric measurements [3, 4, 5, 13, 20, 26, 59, 65, 67, 74]. Four studies conducted usability testing [38, 42, 47, 76], with two of them conducting multiple rounds of testing [38, 76]. Three of the four studies conducted usability testing with informal caregivers [38, 42, 76], while two of the four studies involved participants other than informal caregivers in their usability testing [38, 47].
For performance evaluations based on metric measurements, researchers reported different measurements. Researchers applied F-score to evaluate the performance of AI solutions for agitation prediction (weighted F1:0.89 [3], mean F1:0.85 [3], F-score:0.69-0.73 [4]), anxiety detection (F1:0.82-0.96 [59]), caregiver burdens assessment (F1:0.96-1 [20]), and the prediction of resolutions through medical care-related social services (F-score:0.86 [67]). Accuracy was used to evaluate the performance of AI solutions for fall detection (1 [13]), agitation prediction (0.76-0.87 [4]), activities recognition of PLWD for informal caregivers (0.315-0.524 [74]), and the prediction of resolutions through medical care-related social services (0.819 [67]. Researchers employed AUC as measurements for agitation prediction (0.78-0.92 [4]), and the prediction of resolutions through medical care-related social services (0.82 [67]). Precision (0.921-0.991 [65] or 0.296-0.546 [74]) and recall (0.975-0.995 [65]) were also applied to evaluate the AI solutions for activities recognition of PLWD. Researchers also adopted other measurements such as the confidence to achieve low delay (less than 30 seconds) of the activities detection when accuracy above 0.95 (65% confidence [26]) and the load tests for wandering issues of PLWD that the backend system was able to handle around 5000 users reliably without an increase in response time [5]. However, determining the effectiveness of these AI solutions poses a challenge, as effectiveness is highly context-dependent. Even though the metric measurements attained high scores, this does not necessarily imply the solutions effectively support informal caregivers of PLWD.
Through usability studies with informal caregivers using mixed methods, researchers received positive user experience feedback in terms of usability, readability, convenience, and accuracy of the smart dog, reduced frequency of wandering or agitation behavior of PLWD and physical discomfort during caregiving, and improved sleep quality, physical health, and the relationship with other family members and PLWD [42]. The other robot was also a high-quality user interface in terms of pleasing appearance, clear voice, appropriate gestures and movements, attractive tablet screen designs and layout, and high perceived usefulness [76]. However, participants provided a comparatively low response in terms of meeting the expectations, and one caregiver participant had doubts about using robots, believing that robots were only for PLWD [76]. Through prototype testing and data analysis, researchers proved the validity of using video-based AI systems and analysis to evaluate the skill progression of the tender dementia-care techniques [47] and investigated failed cases of dialogues with the voice assistant [38].
3.3.2 Challenges and limitations identified by researchers in using AI to support informal caregivers of PLWD.
Challenges identified by researchers included 1) running sensing systems continuously without disrupting users’ daily activities on affordable devices used every day (i.e., smartphones) [26]; 2) getting the ground truth due to sparsity, unpredictability, and variations in targeted behavior [4] with confounding patterns as well as inter-person and intra-person variability [3]; 3) developing standard algorithms that can adequately and concisely recognize behavioral symptoms while minimizing false alarms to keep the system reliable, usable and generalizable [21, 26, 74]; 4) dealing with the real-time processing of massive data so that activities and zones can be detected accurately in a timely fashion [13, 26]; 5) data computing including device and data calibration, data transformation, dealing with missing data and biases in subjective data [21], and integrating the information from the ambient sensors and wearable sensors simultaneously [65]; 6) recruiting older adults to participate in research studies involving technologies such as robots [76]; and 7) ethical issues when collecting or processing personal data and health information delivered to the end-user [50].
Limitations included 1) the support for informal caregivers of PLWD in late stages was still insufficient [42]; 2) difficulties for older users that it took longer for them to get accustomed to the system [42]; 3) limited accuracy that the system occasionally gave false alarms due to errors in identifying the severity of the pneumonia symptoms of PLWD [42]; 4) limited interaction capabilities such as verbal communication due to hardcoded questions and follow-up comments [76] and difficulties to match an undefined utterance to a slot [38]; 5) trade-offs in recommendation strategies including cold-start problem, sparsity problem, generalization problem, overspecialization problem, and insensitive to preference changes [50]; 6) limited generalizability of the study findings [67] including evaluation within a laboratory setting [38], small sample sizes [67, 76], short study duration [67] and limited measures and analyses [67].
4 DISCUSSION
AI outperforms ICT and assistive technologies by providing data-driven support such as assisting decision-making [31, 37, 44] and providing personalized recommendations [2, 16, 61]. Previous studies have utilized these advantages, employing AI to support informal caregivers of PLWD in caregiving tasks such as physical/nursing care and supervision/support of PLWD, personalized support for household work, and personalized information seeking for planning. However, there are gaps in existing research.
A critical concern is the gap in understanding how the profiles of informal caregivers and the conditions of PLWD influence technology adoption and experience. According to the concept of Human-Centered Artificial Intelligence (HCAI), which emphasizes AI that not only empowers and enables human users but also openly communicates its values, biases, limitations, and ethical considerations [12], existing AI solutions for informal caregivers of PLWD, while possessing advantages over ICT and assistive technologies, often fall short in terms of transparency and ethical engagement. This becomes evident due to the largely missing demographics of informal caregivers of PLWD and the conditions of PLWD in existing research.
A lack of user-centric design (HCD) was identified in existing research. HCD is an approach to design that prioritizes the needs, motivations, emotions, behavior, and perspectives of individuals in the development process [9].” Involving anticipated end-users in the design of AI solutions prompts inquiries such as “How do researchers determine what aspects of AI can or should be co-designed?” “How does the act of participation in AI design happen?” “What do researchers do with co-designed AI artifacts?” and “How do we facilitate collaboration between co-designers and researchers with expertise across different applications domains?” [78]
Most existing studies on AI solutions for informal caregivers of PLWD have not been tested with the actual end-users—informal caregivers of PLWD—warranting the effectiveness of existing AI solutions in meeting their needs. This gap highlights the necessity for real-world testing to adhere to the principles of HCAI. The effectiveness of AI solutions depends on the context, and the relationships between metric evaluations and the user experience remain uncertain, especially considering humans have thresholds in signal detection and discrimination [10]. Will accuracy impact significant differences in experience, or is there any threshold for the metric measurement to allow a good user experience? Moreover, the varying starting levels and trajectories of trust people place when interacting with different forms of AI [19] further complicates the matter. Understanding these variances is essential for developing effective and trustworthy AI tools.
The unaddressed complexities and requirements of caregivers’ needs present significant challenges. These needs include coordination of formal services, help received from others (informal & formal), relationship with formal service providers, housing, financial costs, physical health, crises planning, and information about legal regulation in caring. Unlike supervision/support, which embeds AI concealed from informal caregivers to monitor PLWD’s condition, some of these unaddressed needs necessitate highly explainable AI solutions that are transparent and comprehensible to caregivers, particularly in collaborative decision-making domains such as housing and financial planning. Opportunities for progress and innovation may arise from gaining a deeper understanding of these needs and involving designers, developers, and informal caregivers of PLWD for AI experience design [66, 78].
5 LIMITATIONS AND FUTURE WORK
Due to the rapid expansion of research in this field, studies published since January 2023 were not incorporated. Using the identical search query, we retrieved 18 records from ACM, 45 records from IEEE, and 11 records from PubMed, published between January 2023 and March 2024. We conducted a preliminary assessment and found that these studies were in line with our findings. We will conduct a comprehensive analysis of these papers in subsequent phases of our research.
To better support informal caregivers of PLWD with their needs, future research should focus on designing and developing more caregiver-centered AI solutions. This involves a deeper understanding of the nuanced needs of caregivers across different demographics and PLWD at various stages, exploring the yet uninvestigated needs of informal caregivers, involving caregivers in the design and evaluation process of AI solutions, overcoming the limitations of current AI applications, addressing identified challenges, and ensuring that they not only meet practical needs but also resonate ethically and empathically with their users.
6 CONCLUSION
This study conducted a systematic review to explore the investigated and uninvestigated needs of informal caregivers of PLWD by AI solutions. The investigated needs were supervision/support, emotional health, physical/nursing care, information about dementia and dementia care, household work, juggling responsibilities, relationships with the care recipient and with family, future planning, and information about professional support and formal services. The uninvestigated needs of informal caregivers of PLWD by AI solutions were coordination of formal services, help received from others (in-formal & formal), relationship with formal service providers, housing, financial costs, physical health, crises planning, and information about legal regulation in caring. We examined the solution design in terms of interfaces, data, and algorithms concerning the investigated needs. We also presented the effectiveness, challenges, and limitations of meeting informal caregivers’ needs. Future research should consider the dynamics of informal caregivers’ needs and develop more caregiver-centered AI solutions.
ACKNOWLEDGMENTS
This project has been funded in part by NSF IIS #2144880.
A MORE DETAILS OF RESEARCH METHODS
A.1 Databases
For the databases applied in this study, ACM Digital Library provides access to more than 3 million publications in computing [39] while IEEE Xplore has more than 6 million documents in electrical engineering, computer science, and electronics [75]. PubMed comprises more than 36 million citations for biomedical literature, facilitating searching across several U.S. National Library of Medicine literature resources including MEDLINE, PubMed Central, and Bookshelf [54]. These three databases will provide related literature from technology-oriented work to PLWD caregivers-oriented work.
A.2 Data collection
To improve the search strategy, the asterisk symbol “*” within a search query functions as a wildcard character, allowing matching an unlimited number of characters within a search term. This facilitates a search with broader and more adaptable outcomes. Among the 937 total papers from all three databases, we filtered out 17 duplicates, 68 research protocols, and 1 paper that did not have an abstract. The remaining 851 papers were evaluated for screening.
A.3 Inclusion and exclusion criteria
We discussed the inclusion and exclusion criteria among the research team through multiple trials of eligibility evaluation of 20 random samples per trial among two researchers to screen the titles and abstracts. Based on the inclusion criteria for screening, 819 papers were excluded and 32 papers were identified for the eligibility evaluation of titles and abstracts. The excluded examples include studies that only support PLWD, those in which caregivers are involved in the development phase but positioning PLWD as the target users, studies discussing other diseases, such as kidney disease, or diseases with risks of causing dementia (e.g. stroke and Parkinson’s disease), and papers without specific AI-related keywords, such as assistive technology, internet-based supportive interventions, and tablet-based technology.
| Themes | Needs | Definitions |
| 1. Caregiving | 1-a. Physical/nursing care | Informal caregivers provide physical/nursing caregiving work and support. |
| 1-b. Household work | Informal caregivers do work in or around the caregiving environment (home). | |
| 1-c. Supervision/support | Informal caregivers remind/monitor/help PLWD to do the tasks. | |
| 1-d. Coordination | Formal services coordinate with each other. | |
| 1-e. Help received from others (informal & formal) | Informal & formal parties provide help with the caregiving tasks | |
| 2. Relationship with Formal Service Providers | Informal caregivers feel comfortable asking the medical personnel, that is hospital staff or (PLWD’s) personal doctor, for information, or for more information. | |
| 3. Housing | Informal caregivers have concerns about the condition of their house as it relates to caregiving. | |
| 4. Juggling Responsibilities | Informal caregivers manage to juggle their responsibilities, commitments, and caring for PLWD. | |
| 5. Financial Costs | The financial cost of care and problems. | |
| 6. Personal | 6-a. Physical health | Informal caregivers’ specific medical health conditions. |
| 6-b. Emotional health | Informal caregivers’ specific mental health conditions or feelings. | |
| 7. Relationships | 7-a. With care recipient | The areas of tension between informal caregivers and PLWD regarding the care informal caregivers provide. |
| 7-b. With family | The relationships between informal caregivers and family members other than PLWD. | |
| 8. Planning | 8-a. Crises planning | Informal caregivers have any plans in place for crises. |
| 8-b. Future planning | Informal caregivers have any plans in place for the future care of PLWD. | |
| 8-c. Information about dementia and dementia care | Informal caregivers have needs in information about dementia and dementia care. | |
| 8-d. Information about professional support and formal services | Informal caregivers have needs in information about professional support and formal services | |
| 8-e. Information about legal regulation in caring | Informal caregivers have needs in information about legal regulation in caring. | |
Table A1: A Framework of the Needs of Informal Caregivers of PLWD
A.4 The framework of the needs of informal caregivers of PLWD
For the framework applied in this study as guidance to examine the AI solutions, researchers investigated the needs of informal caregivers of PLWD in a review of 31 articles and got 18 specific needs covering 8 themes [55] by building upon the C.A.R.E. (Caregivers’ Aspirations, Realities, and Expectations) Tool, a psycho-social assessment tool. To develop the C.A.R.E. Tool, researchers identified and understood caregiver’s situations including their problems, strengths, and needs, and examined the tool in terms of interrater reliability, content, and construct validity through focus groups [34]. Table A1 shows the updated 18 specific needs from the 8 themes with definitions retrieved from the original work [34] and the review [55], including physical/nursing care, household work, supervision/support, coordination of services, help received from others (informal & formal), relationship with formal service providers, housing, juggling responsibilities, financial costs, physical health, emotional health, relationship with the care recipient, relationship with family, crises planning, future planning, information about dementia and dementia care, information about professional support and formal services, and information about legal regulation in caring [55].
B PARTICIPANTS INVOLVED IN THE STUDIES
Figure B1: Pie charts showing the overview of participants and stages of PLWD involved in the studies.
Although these studies aimed to support informal caregivers of PLWD, some of them did not involve informal caregivers to evaluate their solutions. As Figure B1a shows, among the 16 AI solutions to support informal caregivers, 4 studies (25% of the studies) did not involve participants at all [5, 21, 50, 74] because they were mainly in the stage of architecture designing. These studies presented the architectures of the AI solutions [5, 21, 50] and tested the system through established datasets [74]. One study (6.25% of the studies) did not report participants due to no direct contact with participants but the system was tested in a test bed and the field for several months using systematic sets of scenarios of usages such as users’ on-foot, in-vehicle, on-bicycle, and still activities [26]. For the rest 11 studies, 2 studies (12.5% of the studies) involved both informal caregivers and PLWD [3, 4]; 2 studies (12.5% of the studies) involved both informal caregivers and other participants [38, 76]; 3 studies (18.75% of the studies) involved only informal caregivers [20, 42, 67]; 4 studies (25% of the studies) involved participants other than informal caregivers and PLWD or without clearly informing whether the participants were informal caregivers [13, 47, 59, 65].
As Figure B1b shows, 75% of the studies did not report the stages of PLWD. Two studies (12.5% of the studies) focused on mild to severe dementia [3, 4]. Two studies (12.5% of the studies) reported PLWD with other diseases, thereby adding complexity to their caregiving situations. One study focused on PLWD in the early stage with type 2 diabetes [38] and the other focused on PLWD in the middle to the late stage with or without having suffered from pneumonia [42].
As Table B1 shows, existing studies varied in sample sizes with missing demographics. The sample size ranged from 1 to 607. One study using survey methodology did not report the sample size at all, only informing the percentage of different responses [42]. The most frequently reported participant demographics in the studies were gender and age, each accounting for 31.25% of the studies. Female participants were generally more than male participants [13, 38, 67, 76]. The age of the participants ranged from 25 to 85+. Three studies (18.75% of the studies) reported races, consisting of White [20, 76], African American [20, 76], and Hispanic [67]. Three studies (18.75% of the studies) reported the careers of participants such as faculty, graduate students, former paid in-home caregivers, and humanitude caregivers of different levels of experts. One study reported the education of participants and one study reported the income of participants.
| Sample size (n) | Stakeholders | Demographics | References | ||||||
| Total N | Subgroup size | Gender (F/M) | Age(Mean±SD) | Race | Education | Income | Career | ||
| 16 | 8 | PLWD | NA | F(80.5±1.1); M (80.7±7.3) | NA | NA | NA | NA | [3] |
| 8 | Informal CG | NA | NA | NA | NA | NA | NA | ||
| 20 | 10 | PLWD | 5/5 | F(80.6±4.3); M (78.7±6.8) | NA | NA | NA | NA | [4] |
| 10 | Informal CG | NA | NA | NA | NA | NA | NA | ||
| 6 | 1 | Informal CG | 1/0 | NA | NA | NA | NA | NA | [38] |
| 5 | Others: Research team | NA | NA | NA | NA | NA | faculty (1); graduate students (4) | ||
| 10 | 9 | Informal CG (7) + Others: older adults (2) | 5/4 | Age:50-65 (1); Age: 65-75 (3); Age: 75-85 (4); Age: 85+ (1) | White(5); African American (4) | Some college or higher | NA | Doctors (2); NA (7) | [76] |
| 1 | Others: Dementia CG | 1/0 | Age: 65-75 | NA | NA | NA | Former paid in-home CG | ||
| 607 | 607 | Informal CG | NA | NA | Caucasian (569); Afican Americans (38) | NA | NA | NA | [20] |
| NA | NA | Informal CG | NA | Age: 25-78 | NA | NA | NA | NA | [42] |
| 72 | 72 | Informal CG | 60/12 | 57.3± 11.7 | Hispanic | NA | 81.2% had an annual income of fewer than 40,000 USD | NA | [67] |
| 27 | 22 | Others: Healthy subjects | 17/5 | NA | NA | NA | NA | NA | [13] |
| 5 | Others: NA | NA | NA | NA | NA | NA | NA | ||
| 12 | 12 | Others: Humanitude CG | NA | NA | NA | NA | NA | Care experts (2); middle-level CG (7); novice CG (3) | [47] |
| 1 | 1 | Others: NA | NA | NA | NA | NA | NA | NA | [59] |
| 6 | 6 | Others: NA | NA | NA | NA | NA | NA | NA | [65] |
| NA | NA | NA | NA | NA | NA | NA | NA | NA | [5, 21, 26, 50, 74] |
CG: Caregivers; F: Female; M: Male; NA: Not available
Table B1: An Overview of Demographics of Participants Involved in the Study
CG: Caregivers; F: Female; M: Male; NA: Not available
Supplemental Material
Available for Download
- Syed Sibte Raza Abidi and Samina Raza Abidi. 2019. Intelligent health data analytics: A convergence of artificial intelligence and big data. Healthcare Management Forum 32, 4 (2019), 178–182. https://doi.org/10.1177/0840470419846134 arXiv:https://doi.org/10.1177/0840470419846134PMID: 31117831.Google Scholar
Cross Ref
- Zeeshan Ahmed, Khalid Mohamed, Saman Zeeshan, and XinQi Dong. 2020. Artificial intelligence with multi-functional machine learning platform development for better healthcare and precision medicine. Database 2020 (2020), baaa010.Google Scholar
- Ridwan Alam, Martha Anderson, Azziza Bankole, and John Lach. 2018. Inferring physical agitation in dementia using smartwatch and sequential behavior models. In 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI). IEEE, Las Vegas, NV, USA, 170–173. https://doi.org/10.1109/BHI.2018.8333396Google ScholarCross Ref
- Ridwan Alam, Azziza Bankole, Martha Anderson, and John Lach. 2019. Multiple-Instance Learning for Sparse Behavior Modeling from Wearables: Toward Dementia-Related Agitation Prediction. In 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE, Berlin, Germany, 1330–1333. https://doi.org/10.1109/EMBC.2019.8856502Google ScholarCross Ref
- Nicklas Sindlev Andersen, Marco Chiarandini, and Jacopo Mauro. 2021. Wandering and getting lost: the architecture of an app activating local communities on dementia issues. In 2021 IEEE/ACM 3rd International Workshop on Software Engineering for Healthcare (SEH). IEEE, Madrid, Spain, 36–43.Google ScholarCross Ref
- Cathrine Arntzen, Torhild Holthe, and Rita Jentoft. 2016. Tracing the successful incorporation of assistive technology into everyday life for younger people with dementia and family carers. Dementia 15, 4 (2016), 646–662.Google ScholarCross Ref
- Alzheimer’s Association. 2022. 2022 Alzheimer’s disease facts and figures. Alzheimer’s & Dementia 18, 4 (2022), 700–789. https://doi.org/10.1002/alz.12638Google ScholarCross Ref
- Alzheimer’s Association. 2023. 2023 Alzheimer’s disease facts and figures. Alzheimer’s & Dementia 19, 4 (2023), 1598–1695. https://doi.org/10.1002/alz.13016Google ScholarCross Ref
- Jan Auernhammer. 2020. Human-centered AI: The role of Human-centered Design Research in the development of AI. In DRS International Conference 2020. Design Research Society, Brisbane, Australia, 1315–1333.Google ScholarCross Ref
- Kenneth R Boff, Lloyd Kaufman, and James P Thomas. 1986. Handbook of perception and human performance. Vol. 1. Wiley-Interscience, New York, NY, USA.Google Scholar
- Federico Cabitza, Raffaele Rasoini, and Gian Franco Gensini. 2017. Unintended consequences of machine learning in medicine. Jama 318, 6 (2017), 517–518.Google ScholarCross Ref
- Tara Capel and Margot Brereton. 2023. What is Human-Centered about Human-Centered AI? A Map of the Research Landscape. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (
, ) (CHI ’23). Association for Computing Machinery, New York, NY, USA, Article 359, 23 pages. https://doi.org/10.1145/3544548.3580959Google ScholarHamburg , Germany,Digital Library
- Ian Chetcuti, Conrad Attard, and Joseph Bonello. 2022. Data processing using edge computing: A case study for the remote care environment. In 2022 IEEE 21st Mediterranean Electrotechnical Conference (MELECON). IEEE, Palermo, Italy, 720–725.Google ScholarCross Ref
- National Research Council 2010. Informal caregivers in the United States: prevalence, caregiver characteristics, and ability to provide care. In The role of human factors in home health care: Workshop summary. National Academies Press (US), Washington, D.C., USA, 117–144.Google Scholar
- Kieren J Egan, Ángel C Pinto-Bruno, Irene Bighelli, Marla Berg-Weger, Annemieke van Straten, Emiliano Albanese, and Anne-Margriet Pot. 2018. Online training and support programs designed to improve mental health and reduce burden among caregivers of people with dementia: a systematic review. Journal of the American Medical Directors Association 19, 3 (2018), 200–206.Google ScholarCross Ref
- Frank Emmert-Streib and Matthias Dehmer. 2018. A machine learning perspective on Personalized Medicine: an automized, comprehensive knowledge base with ontology for pattern recognition. Machine Learning and Knowledge Extraction 1, 1 (2018), 149–156.Google ScholarCross Ref
- Alexander L Fogel and Joseph C Kvedar. 2018. Artificial intelligence powers digital medicine. NPJ digital medicine 1, 1 (2018), 5.Google Scholar
- Pierre Gérain and Emmanuelle Zech. 2019. Informal caregiver burnout? Development of a theoretical framework to understand the impact of caregiving. Frontiers in psychology 10 (2019), 466359.Google Scholar
- Ella Glikson and Anita Williams Woolley. 2020. Human trust in artificial intelligence: Review of empirical research. Academy of Management Annals 14, 2 (2020), 627–660.Google ScholarCross Ref
- Hilda Goins and Mohd Anwar. 2022. Data-driven Assessment of Dementia Caregivers’ Burden: One size does not fit all. In 2022 Fourth International Conference on Transdisciplinary AI (TransAI). IEEE, Laguna Hills, CA, USA, 17–21.Google ScholarCross Ref
- Hilda Goins, SeyyedPooya HekmatiAthar, Grace Byfield, Raymond Samuel, and Mohd Anwar. 2020. Toward data-driven assessment of Caregiver’s burden for persons with dementia using machine learning models. In 2020 IEEE 21st International Conference on Information Reuse and Integration for Data Science (IRI). IEEE, Las Vegas, NV, USA, 379–384.Google Scholar
- Damien Gruson, Thibault Helleputte, Patrick Rousseau, and David Gruson. 2019. Data science, artificial intelligence, and machine learning: opportunities for laboratory medicine and the value of positive regulation. Clinical biochemistry 69 (2019), 1–7.Google Scholar
- Francisco J Gutierrez, Sergio F Ochoa, and Julita Vassileva. 2016. Identifying opportunities to support family caregiving in Chile. In Proceedings of the 2016 chi conference extended abstracts on human factors in computing systems. Association for Computing Machinery, New York, NY, USA, 2112–2118.Google ScholarDigital Library
- Neal R Haddaway, Matthew J Page, Chris C Pritchard, and Luke A McGuinness. 2022. PRISMA2020: An R package and Shiny app for producing PRISMA 2020-compliant flow diagrams, with interactivity for optimised digital transparency and Open Synthesis. Campbell Systematic Reviews 18, 2 (2022), e1230.Google ScholarCross Ref
- Lamiece Hassan, Caroline Swarbrick, Caroline Sanders, Angela Parker, Matt Machin, Mary P Tully, and John Ainsworth. 2017. Tea, talk and technology: patient and public involvement to improve connected health ‘wearables’ research in dementia. Research involvement and engagement 3, 1 (2017), 1–17.Google Scholar
- Jad Helmy and Ahmed Helmy. 2016. The alzimio app for dementia, autism & alzheimer’s: Using novel activity recognition algorithms and geofencing. In 2016 IEEE international conference on smart computing (SMARTCOMP). IEEE, St. Louis, MO, USA, 1–6.Google ScholarCross Ref
- Torhild Holthe, Rita Jentoft, Cathrine Arntzen, and Kirsten Thorsen. 2018. Benefits and burdens: family caregivers’ experiences of assistive technology (AT) in everyday life with persons with young-onset dementia (YOD). Disability and Rehabilitation: Assistive Technology 13, 8 (2018), 754–762.Google ScholarCross Ref
- Jenny Hopwood, Nina Walker, Lorraine McDonagh, Greta Rait, Kate Walters, Stephen Iliffe, Jamie Ross, and Nathan Davies. 2018. Internet-based interventions aimed at supporting family caregivers of people with dementia: systematic review. Journal of medical Internet research 20, 6 (2018), e216.Google ScholarCross Ref
- Maarten Houben, Rens Brankaert, Gail Kenning, Inge Bongers, and Berry Eggen. 2022. Designing for everyday sounds at home with people with dementia and their partners. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–15.Google ScholarDigital Library
- Duncan Hull, Steve R Pettifer, and Douglas B Kell. 2008. Defrosting the digital library: bibliographic tools for the next generation web. PLoS computational biology 4, 10 (2008), e1000204.Google Scholar
- Azra Ismail, Divy Thakkar, Neha Madhiwalla, and Neha Kumar. 2023. Public Health Calls for/with AI: An Ethnographic Perspective. Proc. ACM Hum.-Comput. Interact. 7, CSCW2, Article 354 (oct 2023), 26 pages. https://doi.org/10.1145/3610203Google ScholarDigital Library
- Fei Jiang, Yong Jiang, Hui Zhi, Yi Dong, Hao Li, Sufeng Ma, Yilong Wang, Qiang Dong, Haipeng Shen, and Yongjun Wang. 2017. Artificial intelligence in healthcare: past, present and future. Stroke and vascular neurology 2, 4 (2017), 230–243.Google Scholar
- LD Jones, D Golan, SA Hanna, and M Ramachandran. 2018. Artificial intelligence, machine learning and the evolution of healthcare: A bright future or cause for concern?Bone & joint research 7, 3 (2018), 223–225.Google ScholarCross Ref
- Janice Keefe, Nancy Guberman, Pamela Fancey, Lucy Barylak, and Daphne Nahmiash. 2008. Caregivers’ aspirations, realities, and expectations: The care tool. Journal of Applied Gerontology 27, 3 (2008), 286–308.Google ScholarCross Ref
- Konstantina Kourou, Themis P Exarchos, Konstantinos P Exarchos, Michalis V Karamouzis, and Dimitrios I Fotiadis. 2015. Machine learning applications in cancer prognosis and prediction. Computational and structural biotechnology journal 13 (2015), 8–17.Google Scholar
- Frank Ho-yin Lai, Elaine Wai-hung Yan, Kathy Ka-ying Yu, Wing-Sze Tsui, Daniel Ting-hoi Chan, and Benjamin K Yee. 2020. The protective impact of telemedicine on persons with dementia and their caregivers during the COVID-19 pandemic. The American Journal of Geriatric Psychiatry 28, 11 (2020), 1175–1184.Google ScholarCross Ref
- Min Hun Lee and Chong Jun Chew. 2023. Understanding the Effect of Counterfactual Explanations on Trust and Reliance on AI for Human-AI Collaborative Clinical Decision Making. Proc. ACM Hum.-Comput. Interact. 7, CSCW2, Article 369 (oct 2023), 22 pages. https://doi.org/10.1145/3610218Google ScholarDigital Library
- Juan Li, Bikesh Maharjan, Bo Xie, and Cui Tao. 2020. A personalized voice-based diet assistant for caregivers of Alzheimer disease and related dementias: system development and validation. Journal of Medical Internet Research 22, 9 (2020), e19897.Google ScholarCross Ref
- ACM Digital Library. 2023. About the ACM Digital Library. Association for Computing Machinery. Retrieved August 17, 2023 from https://dl.acm.org/aboutGoogle Scholar
- Xin Yao Lin, Herman Saksono, Elizabeth Stowell, Margie E. Lachman, Carmen Castaneda-Sceppa, and Andrea G. Parker. 2020. Go&Grow: An Evaluation of a Pervasive Social Exergame for Caregivers of Loved Ones with Dementia. Proc. ACM Hum.-Comput. Interact. 4, CSCW2, Article 151 (oct 2020), 28 pages. https://doi.org/10.1145/3415222Google ScholarDigital Library
- Robert J Lucero, Elizabeth A Fehlberg, Aditi GM Patel, Ragnhildur I Bjarnardottir, Renessa Williams, Karis Lee, Margaret Ansell, Suzanne Bakken, Jose A Luchsinger, and Mary Mittelman. 2019. The effects of information and communication technologies on informal caregivers of persons living with dementia: a systematic review. Alzheimer’s & Dementia: Translational Research & Clinical Interventions 5 (2019), 1–12.Google Scholar
- Mei-Jing Lyu and Shyan-Ming Yuan. 2020. Cloud-based smart dog music therapy and pneumonia detection system for reducing the difficulty of caring for patients with dementia. IEEE Access 8 (2020), 20977–20990.Google ScholarCross Ref
- Keshini Madara Marasinghe. 2016. Assistive technologies in reducing caregiver burden among informal caregivers of older adults: a systematic review. Disability and Rehabilitation: Assistive Technology 11, 5 (2016), 353–360.Google ScholarCross Ref
- Varoon Mathur, Caitlin Lustig, and Elizabeth Kaziunas. 2022. Disordering Datasets: Sociotechnical Misalignments in AI-Mediated Behavioral Health. Proc. ACM Hum.-Comput. Interact. 6, CSCW2, Article 416 (nov 2022), 33 pages. https://doi.org/10.1145/3555141Google ScholarDigital Library
- Mary L McHugh. 2012. Interrater reliability: the kappa statistic. Biochemia medica 22, 3 (2012), 276–282.Google Scholar
- Diego Muñoz, Stu Favilla, Sonja Pedell, Andrew Murphy, Jeanie Beh, and Tanya Petrovich. 2021. Evaluating an app to promote a better visit through shared activities for people living with dementia and their families. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–13.Google Scholar
- Atsushi Nakazawa and Miwako Honda. 2019. First-person camera system to evaluate Tender Dementia-care skill. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops. IEEE, Seoul, Korea, 0–0.Google ScholarCross Ref
- Lisa Newton, Claire Dickinson, Grant Gibson, Katie Brittain, and Louise Robinson. 2016. Exploring the views of GPs, people with dementia and their carers on assistive technology: a qualitative study. BMJ open 6, 5 (2016), 1–7.Google Scholar
- Kee Yuan Ngiam and Wei Khor. 2019. Big data and machine learning algorithms for health-care delivery. The Lancet Oncology 20, 5 (2019), e262–e273.Google ScholarCross Ref
- Luis Oliva-Felipe, Cristian Barrué, Atia Cortés, Emma Wolverson, Marco Antomarini, Isabelle Landrin, Konstantinos Votis, Ioannis Paliokas, and Ulises Cortés. 2018. Health Recommender System design in the context of CAREGIVERSPRO-MMD Project. In Proceedings of the 11th PErvasive Technologies Related to Assistive Environments Conference. Association for Computing Machinery, New York, NY, USA, 462–469.Google ScholarDigital Library
- Annakarin Olsson, Maria Engström, Kirsti Skovdahl, and Claudia Lampic. 2012. My, your and our needs for safety and security: relatives’ reflections on using information and communication technology in dementia care. Scandinavian journal of caring sciences 26, 1 (2012), 104–112.Google Scholar
- Matthew J Page, Joanne E McKenzie, Patrick M Bossuyt, Isabelle Boutron, Tammy C Hoffmann, Cynthia D Mulrow, Larissa Shamseer, Jennifer M Tetzlaff, Elie A Akl, Sue E Brennan, 2021. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. International journal of surgery 88 (2021), 105906.Google Scholar
- Anne Marie Piper, Raymundo Cornejo, Lisa Hurwitz, and Caitlin Unumb. 2016. Technological caregiving: Supporting online activity for adults with cognitive impairments. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 5311–5323.Google ScholarDigital Library
- PubMed. 2023. PubMed Overview. National Library of Medicine. Retrieved August 17, 2023 from https://pubmed.ncbi.nlm.nih.gov/about/Google Scholar
- Francine NFR Queluz, Emily Kervin, Lori Wozney, Pamela Fancey, Patrick J McGrath, and Janice Keefe. 2020. Understanding the needs of caregivers of persons with dementia: a scoping review. International psychogeriatrics 32, 1 (2020), 35–52.Google ScholarCross Ref
- Wullianallur Raghupathi and Viju Raghupathi. 2014. Big data analytics in healthcare: promise and potential. Health information science and systems 2 (2014), 1–10.Google Scholar
- Alvin Rajkomar, Jeffrey Dean, and Isaac Kohane. 2019. Machine learning in medicine. New England Journal of Medicine 380, 14 (2019), 1347–1358.Google ScholarCross Ref
- Sherri Rose. 2018. Machine learning for prediction in electronic health data. JAMA network open 1, 4 (2018), e181404–e181404.Google Scholar
- Richard Rother, Yingnan Sun, and Benny Lo. 2019. Internet of things based pervasive sensing of psychological anxiety via wearable devices under naturalistic settings. In Living in the Internet of Things (IoT 2019). IET, London, UK, pp. 6.Google Scholar
- Mark Schurgin, Mark Schlager, Laura Vardoulakis, Laura R Pina, and Lauren Wilcox. 2021. Isolation in Coordination: Challenges of Caregivers in the USA. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–14.Google ScholarDigital Library
- Silvana Secinaro, Davide Calandra, Aurelio Secinaro, Vivek Muthurangu, and Paolo Biancone. 2021. The role of artificial intelligence in healthcare: a structured literature review. BMC medical informatics and decision making 21 (2021), 1–23.Google Scholar
- Mark A Smith. 1998. Alzheimer disease. International review of neurobiology 42 (1998), 1–54.Google ScholarCross Ref
- Vimal Sriram, Crispin Jenkinson, and Michele Peters. 2019. Informal carers’ experience of assistive technology use in dementia care at home: a systematic review. BMC geriatrics 19 (2019), 1–25.Google Scholar
- Joshua D Stein, Moshiur Rahman, Chris Andrews, Joshua R Ehrlich, Shivani Kamat, Manjool Shah, Erin A Boese, Maria A Woodward, Jeff Cowall, Edward H Trager, 2019. Evaluation of an algorithm for identifying ocular conditions in electronic health record data. JAMA ophthalmology 137, 5 (2019), 491–497.Google Scholar
- Chun-Fang Su, Li-Chen Fu, Yi-Wei Chien, and Ting-Ying Li. 2018. Activity recognition system for dementia in smart homes based on wearable sensor data. In 2018 IEEE Symposium Series on Computational Intelligence (SSCI). IEEE, Bangalore, India, 463–469.Google ScholarCross Ref
- Hariharan Subramonyam, Colleen Seifert, and Eytan Adar. 2021. Towards A Process Model for Co-Creating AI Experiences. In Proceedings of the 2021 ACM Designing Interactive Systems Conference (Virtual Event, USA) (DIS ’21). Association for Computing Machinery, New York, NY, USA, 1529–1543. https://doi.org/10.1145/3461778.3462012Google ScholarDigital Library
- YOON Sunmoo, Alexandra Mendes, Louis Burgio, Mary Mittelman, Ilana Dunner, Jed A Levine, Carolina Hoyos, Dante Tipiani, Mildred Ramirez, Jeanne A Teresi, 2022. Application of Machine Learning Techniques to Examine Social Service Needs Among Hispanic Family Caregivers of Persons with Dementia. Studies in health technology and informatics 295 (2022), 507.Google Scholar
- Charlotte Tang, Yunan Chen, Karen Cheng, Victor Ngo, and John E Mattison. 2018. Awareness and handoffs in home care: coordination among informal caregivers. Behaviour & Information Technology 37, 1 (2018), 66–86.Google ScholarCross Ref
- Anja Thieme, Danielle Belgrave, and Gavin Doherty. 2020. Machine learning in mental health: A systematic review of the HCI literature to support the development of effective and implementable ML systems. ACM Transactions on Computer-Human Interaction (TOCHI) 27, 5 (2020), 1–53.Google ScholarDigital Library
- Myrte Thoolen, Francesca Toso, Sebastiaan TM Peek, Yuan Lu, and Rens Brankaert. 2022. LivingMoments: Bespoke social communication for people living with dementia and their relatives. In Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–18.Google ScholarDigital Library
- David Unbehaun, Konstantin Aal, Daryoush Daniel Vaziri, Peter David Tolmie, Rainer Wieching, David Randall, and Volker Wulf. 2020. Social technology appropriation in dementia: investigating the role of caregivers in engaging people with dementia with a videogame-based training system. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–15.Google ScholarDigital Library
- David Unbehaun, Daryoush Daniel Vaziri, Konstantin Aal, Rainer Wieching, Peter Tolmie, and Volker Wulf. 2018. Exploring the Potential of Exergames to affect the Social and Daily Life of People with Dementia and their Caregivers. In Proceedings of the 2018 chi conference on human factors in computing systems. Association for Computing Machinery, New York, NY, USA, 1–15.Google ScholarDigital Library
- Griffin M Weber, Kenneth D Mandl, and Isaac S Kohane. 2014. Finding the missing link for big biomedical data. Jama 311, 24 (2014), 2479–2480.Google Scholar
- Zachary Wharton, Erik Thomas, Bappaditya Debnath, and Ardhendu Behera. 2018. A vision-based transfer learning approach for recognizing behavioral symptoms in people with dementia. In 2018 15th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS). IEEE, Auckland, New Zealand, 1–6.Google ScholarCross Ref
- IEEE Xplore. 2023. About IEEE Xplore. Institute of Electrical and Electronics Engineers. Retrieved August 17, 2023 from https://ieeexplore.ieee.org/Xplorehelp/overview-of-ieee-xplore/about-ieee-xploreGoogle Scholar
- Fengpei Yuan, Sharon Bowland, Lauren Proctor, Jordis Blackburn, Namrata Mukherjee, Robert Bray, Ruth Palan Lopez, Kristina Wick, and Xiaopeng Zhao. 2022. Robot-assisted Psycho-education to Enhance Alzheimer’s Caregiver Health. In 2022 IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE). IEEE, Arlington, VA, USA, 57–65.Google Scholar
- Tamara Zubatiy, Kayci L Vickers, Niharika Mathur, and Elizabeth D Mynatt. 2021. Empowering dyads of older adults with mild cognitive impairment and their care partners using conversational agents. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, New York, NY, USA, 1–15.Google ScholarDigital Library
- Douglas Zytko, Pamela J. Wisniewski, Shion Guha, Eric P. S. Baumer, and Min Kyung Lee. 2022. Participatory Design of AI Systems: Opportunities and Challenges Across Diverse Users, Relationships, and Application Domains. In Extended Abstracts of the 2022 CHI Conference on Human Factors in Computing Systems (New Orleans, LA, USA) (CHI EA ’22). Association for Computing Machinery, New York, NY, USA, Article 154, 4 pages. https://doi.org/10.1145/3491101.3516506Google ScholarDigital Library
Index Terms
- Artificial Intelligence Systems for Supporting Informal Caregivers of People Living with Alzheimer's Disease or Related Dementias: A Systematic Review
Recommendations
Diagnosis of Alzheimer's disease using Naive Bayesian Classifier
In the world of modern medicine, though there is lot of medical achievements, some diseases still continue to pest the human race. Unfortunately, dementia is one such disease. All over the world, a large number of people are suffering from dementia. ...
Everyday Patient-Care Technologies for Alzheimer's Disease
The Alzheimer's Association has stimulated collaboration and scientific advancement in preventing and treating Alzheimer's disease by bringing together leading researchers developing technological approaches offering promise in this area. This year, the ...
Emotion Work in Caregiving: The Role of Technology to Support Informal Caregivers of Persons Living With Dementia
CSCWThe CSCW and HCI community has increasingly examined ways to support informal caregivers through technology, given the increasing burden and its consequences on caregivers' emotional and physical health. In this paper, we interviewed 12 informal ...
Comments