Factors Associated With the Experience of Cognitive Training Apps for the Prevention of Dementia: Cross-sectional Study Using an Extended Health Belief Model

mHealth Apps for Dementia, Alzheimer Disease, and Other Neurocognitive Disorders: Systematic Search and Environmental Scan

BACKGROUND

Lifestyle behaviors including exercise, sleep, diet, stress, mental stimulation, and social interaction significantly impact the likelihood of developing dementia. Mobile health (mHealth) apps have been valuable tools in addressing these lifestyle behaviors for general health and well-being, and there is growing recognition of their potential use for brain health and dementia prevention. Effective apps must be evidence-based and safeguard user data, addressing gaps in the current state of dementia-related mHealth apps.

OBJECTIVE

This study aims to describe the scope of available apps for dementia prevention and risk factors, highlighting gaps and suggesting a path forward for future development.

METHODS

A systematic search of mobile app stores, peer-reviewed literature, dementia and Alzheimer association websites, and browser searches was conducted from October 19, 2022, to November 2, 2022. A total of 1044 mHealth apps were retrieved. After screening, 152 apps met the inclusion criteria and were coded by paired, independent reviewers using an extraction framework. The framework was adapted from the Silberg scale, other scoping reviews of mHealth apps for similar populations, and background research on modifiable dementia risk factors. Coded elements included evidence-based and expert credibility, app features, lifestyle elements of focus, and privacy and security.

RESULTS

Of the 152 apps that met the final selection criteria, 88 (57.9%) addressed modifiable lifestyle behaviors associated with reducing dementia risk. However, many of these apps (59/152, 38.8%) only addressed one lifestyle behavior, with mental stimulation being the most frequently addressed. More than half (84/152, 55.2%) scored 2 points out of 9 on the Silberg scale, with a mean score of 2.4 (SD 1.0) points. Most of the 152 apps did not disclose essential information: 120 (78.9%) did not disclose expert consultation, 125 (82.2%) did not disclose evidence-based information, 146 (96.1%) did not disclose author credentials, and 134 (88.2%) did not disclose their information sources. In addition, 105 (69.2%) apps did not disclose adherence to data privacy and security practices.

CONCLUSIONS

There is an opportunity for mHealth apps to support individuals in engaging in behaviors linked to reducing dementia risk. While there is a market for these products, there is a lack of dementia-related apps focused on multiple lifestyle behaviors. Gaps in the rigor of app development regarding evidence base, credibility, and adherence to data privacy and security standards must be addressed. Following established and validated guidelines will be necessary for dementia-related apps to be effective and advance successfully.

Understanding Health Beliefs and Health Behaviors in Older Adults at Risk for Alzheimer's Disease

Background

There are significant public health benefits to delaying the onset of Alzheimer's disease (AD) in individuals at risk. However, adherence to brain healthy behaviors is low. The Health Belief Model proposes that specific beliefs are mediators of behavior change.

Objective

To characterize health belief measures from the Science of Behavior Change Research Network (SBCRN) in an older adult population and associations between health beliefs, AD risk, and current health behaviors.

Methods

A total of 172 individuals from the Rhode Island AD Prevention Registry participated. SBCRN health belief measures included assessments of future time perspective, self-efficacy, deferment of gratification, and consideration of future consequences. Outcome measures included individual AD risk index score, dementia risk awareness, and lifestyle behaviors including physical, cognitive, and social activity.

Results

Participants who were older had higher scores for AD risk, lower future time perspective, and lower generalized self-efficacy (all at p < 0.001). Higher generalized self-efficacy was related to increased physical activity (p < 0.010). Higher future time perspective (p < 0.001) and generalized self-efficacy (p = 0.48) were associated with lower AD risk score. Subjective cognitive decline (SCD) was associated with lower self-efficacy, ability to delay gratification, and a less expansive future time perspective.

Conclusions

Greater self-efficacy and perceived future time remaining were associated with lower AD risk and greater engagement in physical activity. SCD was associated with health beliefs that may negatively affect engagement in positive brain health behaviors. Assessment of and psychoeducation about these intrapersonal health belief constructs may be important targets for behavioral interventions to reduce AD risk.

Text Dialogue Analysis for Primary Screening of Mild Cognitive Impairment: Development and Validation Study

BACKGROUND

Artificial intelligence models tailored to diagnose cognitive impairment have shown excellent results. However, it is unclear whether large linguistic models can rival specialized models by text alone.

OBJECTIVE

In this study, we explored the performance of ChatGPT for primary screening of mild cognitive impairment (MCI) and standardized the design steps and components of the prompts.

METHODS

We gathered a total of 174 participants from the DementiaBank screening and classified 70% of them into the training set and 30% of them into the test set. Only text dialogues were kept. Sentences were cleaned using a macro code, followed by a manual check. The prompt consisted of 5 main parts, including character setting, scoring system setting, indicator setting, output setting, and explanatory information setting. Three dimensions of variables from published studies were included: vocabulary (ie, word frequency and word ratio, phrase frequency and phrase ratio, and lexical complexity), syntax and grammar (ie, syntactic complexity and grammatical components), and semantics (ie, semantic density and semantic coherence). We used R 4.3.0. for the analysis of variables and diagnostic indicators.

RESULTS

Three additional indicators related to the severity of MCI were incorporated into the final prompt for the model. These indicators were effective in discriminating between MCI and cognitively normal participants: tip-of-the-tongue phenomenon (P<.001), difficulty with complex ideas (P<.001), and memory issues (P<.001). The final GPT-4 model achieved a sensitivity of 0.8636, a specificity of 0.9487, and an area under the curve of 0.9062 on the training set; on the test set, the sensitivity, specificity, and area under the curve reached 0.7727, 0.8333, and 0.8030, respectively.

CONCLUSIONS

ChatGPT was effective in the primary screening of participants with possible MCI. Improved standardization of prompts by clinicians would also improve the performance of the model. It is important to note that ChatGPT is not a substitute for a clinician making a diagnosis.