Based on the knowledge base previously developed for the integrated geriatric care model [28], the current study progressed by employing the subsequent step of the IM framework: program production. The SMART system was designed to tackle 138 care problems spanning eleven domains (e.g., risk for falls, related to environmental hazards) and generate personalized interventions (e.g., guidance on age-appropriate home environment modifications) in accordance with the characteristics (e.g., preferences, living situations) of older people. For instance, if the SMART system detects that an older adult has been bedridden for extended periods and remains in a single position for over 4 h (risk factors), it will diagnose a care problem of “Risk for pressure ulcers, related to prolonged periods of unrelieved pressure”. The corresponding short-term and long-term care goals would be to reduce the duration of pressure on bony prominences and maintain skin integrity, respectively. Based on the comprehensive assessment results, if the individual is unable to reposition by himself/herself and is primarily cared for by the daughter, the personalized intervention would be “Frequent repositioning” with the specific implementation approach being “Remind the caregiver to reposition the older individual every two hours”. A summary of the 138 care problems and examples of care problems along with corresponding personalized interventions were provided in Additional Tables S1 and S2, respectively.
A multidisciplinary team comprising professionals in nursing informatics, nursing research, gerontological care, software engineering, and interface designing was established to collaboratively advance the development of the SMART system. We held weekly meetings throughout all phases to promptly resolve problems and issues raised during the development phase to ensure smooth ongoing progress. The study, encompassing both the system development and usability testing, obtained approval from the Institutional Review Committee of the Capital Medical University (Approval No. 2015SY49). The flow diagram for the development of the SMART system and usability testing was presented in Fig. 1.
The flow diagram for the development of the SMART system and usability testing
Phase 1: development of the SMART system
Functional design
According to human neural reflexes, the Sensors and Scales collected data from older individuals through wearable devices, periodic self-assessments of older people, and assessments by others (e.g., family members, nannies, community nurses, etc.). The collected data were uploaded automatically to the Remote Cloud and Management Center via Wifi or 5G networks for comprehensive analysis to diagnose care problems and formulate customized interventions, which were subsequently distributed to daily caregivers (mainly family members and nannies) or professional care providers (e.g., doctors and nurses in the medical care system who are responsible for medical assistance, domestic workers in the social care system who mainly assist with daily living activities) within the Total Care System as appropriate. The Mobile Phone Autonomous Response System was a set of simple algorithms that were pre-integrated into the SMART system to deal with simple but urgent care problems, similar to the coordinate reflex actions within the spinal cord, such as the withdrawal reflex from a painful stimulus. The process could enhance coordination among various caregivers and professional care providers and ensure high-quality, coordinated, and integrated older care (see Fig. 2).
The overall architecture of the SMART system. App, application; CARE, Continuity Assessment Record and Evaluation
According to the overall objectives and functional design of the SMART system, the research team drafted key functions and modules for the Cloud Platform, Care Receiver App, and Professional Care Provider App using XMind software version 8 (XMind, Ltd). Based on the initial draft, the multidisciplinary team collaborated in brainstorming to produce a more detailed document to guide the development of the SMART system (see Additional Table S3).
Architecture and database design
To enhance the flexibility, scalability, and maintainability of the SMART system, we adopted the MicroService architecture for development. The approach allowed us to decompose the SMART system into small and single-responsibility units that can be developed, operated, deployed, scaled, and managed independently [30]. The MicroService architecture not only can reduce task complexity and improve development efficiency but also enable zero-downtime releases in satisfying older people’s evolving care needs [31].
The Cloud Platform was designed to be compatible with mainstream browsers following a Browser/Server (B/S) architecture [32], while the Care Receiver App and Professional Care Provider App adopted a Client/Server (C/S) architecture with the application programming interface (API) gateway for access [33]. Besides, both Apps were programmed to be accessible on Android-based mobile phones to ensure their widespread popularity and affordability in China [34]. Once stable on Android-based mobile phones, we will extend compatibility to iOS-based mobile phones.
SQL databases were used to store, process, capture, retrieve, and manage substantial amounts of fully structured data, while NoSQL databases were employed for semi-structured and unstructured data [35]. The SMART system stored information such as older people’s basic characteristics and comprehensive geriatric assessment results, with data from each module stored in either Oracle or MongoDB as appropriate (see Fig. 3).
The system architecture of the SMART system
Security measures design
The SMART system transmits sensitive data, such as personal identification and health information, making it susceptible to external and internal threats. Multiple protective measures were therefore applied to ensure system security. Data access control was achieved through user authentication services and password protective mechanisms, including verifying password strength, limiting login attempts, and periodically changing passwords [36]. To secure data transmission and storage, mutual authentication HTTPS were employed to encrypt all sensitive data [37]. Additionally, database backup and recovery functions were included to ensure data availability at any time. The development of the SMART system adhered to international and national data standards (Additional Table S4).
User interface and visualization design
The user interface (UI) of the Care Receiver App was meticulously crafted based on six user-friendly design principles (structure, simplicity, diverse presentation, feedback, consistency, and tolerance) to improve older people’s perceived usefulness and ease of use with the Care Receiver App, following the Technology Acceptance Model [38]. Axure RP8 (Axure Software Solutions, Inc., USA) was employed to draft the module interfaces with functional buttons annotated to assist interface designers in understanding the functionality of each icon. Subsequently, the interface designers used Flinto version 26.0.5 (Flinto Inc., Australia) to document the planned UIs and their interactions. Distinctive design elements such as contrasting color blocks, large fonts, prominent functions, and unique icons, were used to signify various functions.
Prototypes development and iteratively testing
The SMART system was developed using Java/. Ne for its popularity, power, verbosity, and ease of maintenance, with Linux and UNIX as the operating environment. The software engineers used Android Studio to develop the alpha version of the SMART system based on the functional design and system architecture provided by the research team [39]. The research team then randomly invited ten home-dwelling older individuals from the preliminary needs assessment [28] to test the alpha version within an iterative design framework. Details of inclusion and exclusion criteria as well as the characteristics of the ten older adults are summarized in Additional Method 1 and Table S5. Through ongoing feedback from older people, the team continuously streamlined the delivery of care problems and personalized interventions. Following the agile approach, software engineers iteratively identified and solved technical issues throughout the development process to promote adaptive planning, evolutionary development, high-quality delivery, and continuous improvement [40]. After fixing technical issues in the alpha version, a beta version was formulated and tested within the research team until all modules worked well.
Phase 2: usability testing of the care receiver App within the SMART system
The usability testing of the SMART system’s beta version adhered to the International Standards Organization (ISO) standard 9241–11, which defines usability as the extent to which users can use an App to achieve specific objectives with efficiency, satisfaction, and effectiveness in a specified context of usage [41]. The usability testing was registered in the Chinese Clinical Trial Registry (Registration number: ChiCTR-IOR-17010368) on 12/01/2017. Since older individuals are the core users of the SMART system, we only tested the usability of the Care Receiver App among older people.
Study design and participants
To improve sample-selection efficiency, we conducted a cross-sectional study in a geriatric ward of a comprehensive hospital in Beijing, China, which was accredited as a comprehensive facility with dedicated geriatric wards and capable of serving a diverse population of older patients, from December 2020 to February 2021. Older adults were included consecutively if they: (1) were aged 60 or older; (2) were about to be discharged from the hospital and return home; (3) possessed normal communication and interaction abilities; (4) obtained at least a primary school education; (5) had an Android-based smartphone for Internet access; (6) expressed willingness to participate. Older individuals with dementia or other mental illness were excluded. All participants provided written informed consent on enrollment.
Measures
Demographic information of the older participants was collected using a predefined questionnaire, encompassing age, gender, education, monthly income, mobile phone use experience, and daily mobile phone use time. The perceived usability of the Care Receiver App among older individuals was measured using the Health Information Technology Usability Evaluation Scale (Health-ITUES) [42]. The Health-ITUES comprises 20 items from 4 domains: impact (3 items), perceived usefulness (9 items), perceived ease of use (5 items), and user control (3 items). Each item is rated from 1 (strongly disagree) to 5 (strongly agree) on a 5-point Likert scale. The total scores range from 20 to 100, with higher scores indicating better usability. This tool supports customization at the item level to align with specific tasks and expectations of the health systems. The Chinese version of the Health-ITUES demonstrates satisfactory reliability with a Cronbach’s α coefficient of 0.74–0.90 [43]. To accurately reflect the perceived usability of the Care Receiver App among older people, the Health-ITUES was customized and distributed to them after completing the assigned tasks.
Testing process
Once receiving approval from the relevant departments, we screened older individuals according to the predefined inclusion and exclusion criteria. After a detailed explanation of the study’s purpose, significance, and procedures, we assisted the older participants in downloading and installing the Care Receiver App. Training materials, including instructional videos and user manuals, were made available until they felt confident in using the App. Subsequently, the older participants were encouraged to use the App for 24 h to complete the assigned tasks before filling out the customized Health-ITUES. The tasks included independently viewing assessment reports at least once, inputting medication information at least once, recording blood sugar or blood pressure at least once, recording diet or exercise at least once, and accessing health information at least once. For paper questionnaires, we conducted thorough checks to identify any omissions and verify incomplete or ambiguous information on the spot, thereby ensuring data integrity. For electronic questionnaires, data completeness was verified within 12 h.
Sample size calculation
To achieve adequate statistical power, the sample size should be 5–10 times the number of items [44]. With a total of 20 items in the Health-ITUES version designed for older adults, the study necessitated a minimum of 100 participants.
Statistical analysis
The enrolled older participants were classified into two groups based on a threshold of 3: those with mean item scores above 3 constituted the Positive Usability Group, while those with mean item scores of 3 or below defined the Negative or Neutral Usability Group. Detailed variable types and value assignment methods were provided in Additional Table S6. Continuous data was tested for normal distribution using the one-sample Kolmogorov–Smirnov test and expressed as medians with interquartile range (IQR) or mean ± standard deviation (SD) as appropriate. For the between-group comparison, student-t test was employed for continuous variables with normal distribution, while Mann–Whitney U test was utilized for non-normally distributed continuous data to avoid the influence of non-normal distribution and small sample size issues [45]. Categorical variables were expressed as frequencies and proportions (%), with comparison conducted using chi-square or Fisher’s exact test.
Variables with a two-tailed p < 0.10 in the univariate analysis were considered to be entered into the bootstrapped forward stepwise logistic regression model. The odds ratio (OR) and their 95% confidence interval (CI) were used to assess the independent contribution of each variable. Model performance was evaluated by multiple measures including Nagelkerke R2 to determine the proportion of variance accounted for, accuracy derived from the confusion matrix to reflect prediction correctness, and the Hosmer–Lemeshow test to evaluate the model calibration. Additionally, the Receiver Operating Characteristic (ROC) curve was illustrated and the Area Under the Curve (AUC) was computed to evaluate the model discriminatory ability. Statistical analysis was conducted using SPSS software version 26.0 (IBM Corp, Armonk, NY) with statistical significance set at two-sided p < 0.05.
link