Leveraging mHealth and Virtual Reality to Improve Cognition for Alzheimer’s Patients: A Systematic Review

Background: Alzheimer’s Disease (AD) is a global problem affecting 58 million people, expected to reach a prevalence of 88 million people by 2050. The disease affects the brain, memory, cognition, language, and motor movement. Many interventions have sought to improve memory and cognition. mHealth and virtual reality (VR) are two such interventions. Objectives: To analyze studies from the last 10 years with older adults with AD to ascertain the effectiveness of telehealth techniques such as mHealth and VR for memory care. Methods: In accordance with the Kruse Protocol and reported in accordance with PRISMA 2020, five reviewers searched four research databases (PubMed, CINAHL, Web of Science, and ScienceDirect) on 3 August 2022 for studies with strong methodologies that fit the objective statement. Results: Twenty-two studies from 13 countries were analyzed for trends. Four interventions (mHealth/eHealth, VR, mHealth + VR, game console, and telephone) used RCT, quasi-experimental, pre-post, observational, and mixed methods. These interventions improved cognition, memory, brain activity, language, depression, attention, vitality, quality of life, cortical atrophy, cerebral blood flow, neuro plasticity, and mental health. Only three interventions reported either no improvements or no statistically significant improvements. Cost, time, training, and low reimbursement were barriers to the adoption of these interventions. Conclusion: mHealth and VR offer interventions with positive effectiveness for memory care for AD. The long-term effect of this improvement is unclear. Additional research is needed in this area to establish clinical practice guidelines.


Rationale
Alzheimer's Disease (AD) is a growing condition around the word. As we approached the COVID-19 pandemic, AD was the largest killer of older adults: it kills more people than breast cancer and prostate cancer [1]. The prevalence of the disease was calculated in 2021 to be 58 million people, but it is predicted to exceed 88 million by 2050 [1]. Of the dementia population, AD accounts for about 2/3 s [1]. There is currently no cure for AD, and there are only about 10 pharmaceuticals approved to manage the condition. The disease creates plaque on the brain (tau) that eventually affects the communication of 100 billion neurons in the brain, degrading and ultimately destroying these neurons [2]. Early stages of AD is seen as simple forgetfulness of recently learned facts, but late stages of AD affects speech, motor skills, and long-term memory [1]. Researchers and practitioners do not fully understand the etiology and pathogenesis of AD: we can treat the symptoms, but we cannot prevent or cure the disease [3][4][5]. Researchers have searched for decades for interventions to improve symptoms of cognitive decline, and one of these is cognitive training through telemedicine.
Many tests are used to assess impairment and symptoms associated with AD. AD affects cognition, which is a complex process in the brain that involves memory, abstraction and iconic concepts, mental operations, consciousness, search strategies, problem solving, and social context [6]. One common method to measure cognition is the mini-mental state examination (MMSE), which estimates a severity of cognitive impairment through a series of questions organized into seven categories: orientation to time, orientation to place, registration of three words, attention to calculation, recall of three words, language, and visual construction [7]. Given over time, the MMSE can identify rate of decline or document improvement.
Telemedicine is defined as healing from a distance using information communication technology to overcome geographical boundaries and increase health outcomes [8]. mHealth is a subset of telemedicine that leverages mobile technology to deliver some sort of intervention or interaction with a provider. mHealth interventions with patients who have AD suffer from barriers such as cognition, perception, physical ability, frame of mind, speech and language [9]. mHealth design must break steps into very simple, easy to understand modules, must often repeat instructions to keep the attention of the users, and use simple memory tests to avoid overwhelming the user [10]. mHealth has been coupled with other interventions such as transcranial alternating current during cognitive training, but results are not conclusive [11]. Virtual reality (VR) has also entered the area of AD research, specifically in the area of cognitive training. The reason is that VR exercises multiple perception components of psychophysics (visual, tactile, and kinesthetic perceptual sensations) [12]. The proponents of VR like its immersive and adaptable environment. It has been used in the areas of brain damage, poststroke intervention, musculoskeletal recovery, and in cognitive training for AD. This review will focus on the telemedicine-related interventions (mHealth, VR, and serious games) in the area of memory for AD patients. Multiple systematic literature reviews have examined this interaction. Many conclude that telemedicine can assess cognition, monitor activity, and improve communication with provider teams [13]. Telemedicine can positively affect mood, function, and quality of life, but its effect on cognition is unclear [14].
A systematic literature review and meta-analysis was published in 2022 that analyzed 16 Randomized Controlled Trials (RCTs) [15]. The meta-analysis focused on a smaller set of studies. It found that serious games are as effective as no intervention or passive interventions at improving executive functions. It concluded that conventional exercises were just as effective. The reviewers felt their group for analysis was too small for final conclusions.
A systematic literature review was published in 2022 that analyzed 28 studies over 10 years [9]. It evaluated several aspects of mHealth. It found positive perceptions of the users of mHealth (both AD patients and their caregivers). The caregivers attributed positive effect of mHealth interventions on their physical and mental health; however, effectiveness was not evaluated.

Objectives
The purpose of this review is to analyze the effectiveness of telemedicine-related interventions (mHealth, VR, and serious games) to improve cognition for older adults suffering from Alzheimer's Disease or mild cognitive impairment (MCI) using published literature from the last 10 years. Secondary outcomes will be memory, language, mood, vitality, attention, brain waves, and other conditions measured and reported in the literature. Our review will be different from previous reviews. We will use a larger group of articles for analysis than the former review [15], and it will analyze effectiveness, different from the latter review [9].

Eligibility Criteria
Articles eligible for this review required older adults (>50) with early-stage Alzheimer's Disease or MCI as participants, published in the last ten years, published in peer-reviewed journals, and used strong methods such as RCT or true experiments. Other methods were accepted such as quasi-experimental, mixed method, quantitative, and qualitative.

Information Sources
We searched in four well-known databases: PubMed (MEDLINE), Complete Index of Nursing and Allied Health Literature (CINAHL), Web of Science, and Embase's ScienceDirect. We conducted the search on 3 August 2022. We also performed a journal-specific search of Healthcare. MEDLINE was excluded from all but PubMed. We eliminated reviews from our search to not confound the results. We used only published literature to ensure it was peer reviewed.

Search Strategy
We visited the U.S. Library of Medicine's website to use the Medical Subject Heading's (MeSH) indexing database. Using MeSH, we created a Boolean search string to combine key terms. We used the same search sting in all databases: (mhealth OR telemedicine OR "virtual reality" OR "serious games") AND ("Alzheimer disease" OR dementia) AND memory. Due to differences in filter options in each database, we could not use the exact same filters, but we used similar filter strategies. In CINAHL, we filtered by date, full-text, humans, English language, academic journals, excluded MEDLINE, and excluded reviews. In ScienceDirect, we filtered by date, excluded MEDLINE, and excluded reviews and conference proceedings. In Web of Science, we filtered by date, excluded reviews, and excluded MEDLINE. This practice eliminated most duplicates.

Selection Process
In accordance with the Kruse Protocol, we searched key terms in all databases, filtered results, and screened abstracts for applicability [16]. At least two reviewers screened each abstract, and at least two reviewers analyzed each article for data extraction and thematic analysis.

Data Collection Process
The Kruse Protocol standardized an Excel spreadsheet for data extraction and analysis. We used a series of three consensus meetings to finalize the group of articles for analysis, identify themes in the literature, and perform additional analysis on the data extracted.

Data Items
In accordance with the Kruse Protocol, we collected the following fields of data: database source, date of publication, authors, title of study, participant population, experimental intervention, results (compared to a control), medical outcomes, sample size, bias within study, effect size (Cohen's d), sensitivity, specificity, F1, country of origin, statistics used, patient satisfaction, effectiveness, barriers to adoption, strength of evidence, and quality of evidence. Results were reported in comparison to a control group. Outcomes and effectiveness are highly similar fields, but they are designed for different audiences (providers and administrators). A provider might not be as concerned as length of stay or cost savings as much as direct medical outcomes (e.g., improvement in cognition), but the administrator is.
The primary outcome for this study is cognition, as measured by the MMSE or similar tool such as Addenbrooke

Study Risk of Bias Assessment and Reporting Bias Assessment
Not only did reviewers note observations of bias in each study, but we also assessed the strength and quality of each study using the Johns Hopkins Nursing Evidence Based Practice tool (JHNEBP) [17]. The overall ratings of quality from the JHNEDP provided us with an assessment of the applicability of the cumulative evidence.We considered the instances of bias in how to interpret the results because bias can limit external validity [18].

Effect Measures
Because we accepted mixed methods and qualitative studies, we were unable to standardize summary measures, as would be performed in a meta-analysis. Measures of effect are summarized in tables for those studies in which it was reported. Measures of effect can be reported as Cohen's d, Wald's W, Eta 2 , sensitivity, or specificity. Effects vary based on the statistic used, but they usually follow small (0.0-0.2), medium (0.21-0.79), large (0.8 or higher). An average effect size (ES) can be calculated through a weighted average by using the sample size.

Synthesis Methods
We performed a thematic analysis of the data combining observations (observed multiple times) into themes [19]. We calculated the frequency of occurrences and reported the findings in a series of affinity matrices. This frequency reporting states the probability of finding that theme in the group for analysis, and it provides confidence in the data analyzed. Although thematic analyses are usually reserved for qualitative studies, there is a pattern in the literature for systematic literature reviews to utilize this technique to help synthesize data extracted [20][21][22].

Additional Analyses and Certainty Assessment
Using the standardized spreadsheet, we sorted by intervention and theme to identify interactions. Some interventions appear more effective than others. Sensitivity and specificity were tabulated where reported. Figure 1 illustrates our study selection process. Four databases and one focused journal search were conducted with a standardized Boolean search string. The initial 1096 results were filtered to remove duplicates. At the end of the filtering exercise, 869 records were screened using filters on each database. This exercise removed 812 articles. The resulting 57 were retrieved for a full analysis for eligibility. Several more were filtered out (protocols, conference papers, and those that were not germane to our research objective). The remaining group for analysis was 22.

Study Characteristics
Following the PRISMA 2020 checklist, characteristics for each study were systematically extracted and tabulated to include the following data fields: participants, intervention, comparison (to control or other group), observation, study design (PICOS). The standard PICOS table summarizes study characteristics in a manner commensurate with

Risk of Bias in and across Studies
Reviewers exercised the JHNEBP quality assessment tool to identify strength and quality of evidence. Reviewers also made notes of other observations of bias throughout the data extraction. The JHNEBP tool identified 16/22 (73%) of Strength I due to the use of strong methodologies such as RCT and true experiment. Four others (18%) were identified as Strength II due to either quasi-experimental or a pre-post with a control group. Only 2/22 (9%) were identified as Strength III because of the use of observational or mixed methods methodologies. The JHNEBP tool also identified 16/22 (73%) as Quality A due to the use of adequate control groups and sample sizes, and for reporting consistent results. Only 6/22 (27%) were identified as Quality B. No studies were identified as less than Strength III or Quality B.
Reviewers also identified other incidents of bias. [18] There were 22 observations of selection bias, which threatens the internal validity of the studies. These observations stemmed from limiting the population to one region or one country. Reviewers also noted four observations of sample bias, which threatens the external validity of the studies. These observations were noted where the population was a majority of one race or gender. There were two observations of design bias, which threatens the internal validity of the study. These were noted when there seemed to be a significant flaw in the methodology (e.g., short intervention time). Table 2 summarized the results of individual studies. This table shows the themes identified in the literature. In multiple occasions, there were multiple observations of the same theme identified in the same study. This was an artifact of collapsing observations of a similar nature into one theme. An observation-to-theme match can be found in Appendix A. Other observations incident to the data extraction can be found in Appendix B (sample size, bias, effect size, country of origin, statistics used, patient satisfaction, and the JHNEBP strength and quality of evidence).

Results of Syntheses, Additional Analysis, and Certainty of Evidence
We conducted a thematic analysis of the literature to make sense of the data extracted. Through this process, observations noted multiple times became themes. Not all observations were fit into themes: Some remained as individual observations. These themes and observations are reported by category in affinity matrices with frequency distributions. Frequencies do not imply importance-instead they identify the probability the theme was identified in the group of articles analyzed.
Future research should focus on the improvements in cognition, memory, and brain waves to identify the duration of the improvements. The studies analyzed did not imply the results would be long term. Both mHealth and VR offer some good interventions to provide temporal relief and improvement of AD symptoms. Only three studies identified no improvement or no statically significant improvement [37,39,44]. The rest identified improvements in at least one area. Future considerations should focus on the interventions with the largest reported improvements. In this review, those would be mHealth, eHealth.
The results of this review should provide options for providers and care givers who want to see an improvement in one area or another. The results of these studies are positive. However, providers do face several barriers to the adoption of these interventions. The cost to acquire the equipment would not currently be reimbursed with current treatment codes. It would help to codify some of these interventions into critical practice guidelines. An existing CPG would have a better chance of being reimbursed. After acquiring the equipment, the provider would need to train the staff and the users of the equipment for each intervention. The provider and staff would need additional time to operate the equipment, administer and analyze the measurement tests like the MMSE, and EEG. These barriers are not compelling, but they present significant stumbling blocks to universal adoption.

Limitations
To control for sample bias, we queried four well-known databases, and we used every article that emerged from the abstract screening step. We chose only four databases, but others may have identified additional studies with additional interventions. We also limited the search to published articles that had been peer reviewed. This publication bias may have prevented us from identifying other interventions with various margins of success. To control for confirmation bias, we had multiple reviewers participate in every step: screening, data extraction, and analysis. To control for design bias, we stuck with a published protocol aligned with more than 40 published systematic literature reviews.

Conclusions
mHealth and VR offer promising interventions to help memory and cognition for those who suffer from AD. Several interventions show temporary improvement in cognition, memory, and brain activity. The mHealth and eHealth interventions seem to affect a larger scope of measurable criteria, and they may be easier to implement without complicated VR apparatus. Several barriers stand in the way of universal adoption. Additional reimbursement mechanisms would enable providers to adopt these interventions or test them under different circumstances. The AD patients and their caregivers look for answers and an improvement in the AD symptoms. With additional development, mHealth and VR might provide some viable solutions.
Author Contributions: Conceptualization, methodology, and editor C.S.K.; All authors participated in abstract screening, data extraction, and interpretation of results; writing C.S.K. and K.S. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.

Protocol and Registration:
This review was conducted in accordance with the Kruse Protocol for writing a systematic review. It was reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA). This review is registered with PROSPERO: CRD42021266730.

Informed Consent Statement: Not applicable.
Data Availability Statement: Data from this study can be obtained by asking the lead author.

Conflicts of Interest:
The authors declare no conflict of interest. Table A1. Observation-to-theme conversion (Intervention, Results, and Medical Outcomes).