This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.
In the United States, nearly 80% of family caregivers of people with dementia have at least one chronic condition. Dementia caregivers experience high stress and burden that adversely affect their health and self-management. mHealth apps can improve health and self-management among dementia caregivers with a chronic condition. However, mHealth app adoption by dementia caregivers is low, and reasons for this are not well understood.
The purpose of this study is to explore factors associated with dementia caregivers’ intention to adopt mHealth apps for chronic disease self-management.
We conducted a cross-sectional, correlational study and recruited a convenience sample of dementia caregivers. We created a survey using validated instruments and collected data through computer-assisted telephone interviews and web-based surveys. Before the COVID-19 pandemic, we recruited dementia caregivers through community-based strategies, such as attending community events. After nationwide closures due to the pandemic, the team focused on web-based recruitment. Multiple logistic regression analyses were used to test the relationships between the independent and dependent variables.
Our sample of 117 caregivers had an average age of 53 (SD 17.4) years, 16 (SD 3.3) years of education, and 4 (SD 2.5) chronic conditions. The caregivers were predominantly women (92/117, 78.6%) and minorities (63/117, 53.8%), experienced some to extreme income difficulties (64/117, 54.7%), and were the child or child-in-law (53/117, 45.3%) of the person with dementia. In logistic regression models adjusting for the control variables, caregiver burden (odds ratio [OR] 1.3, 95% CI 0.57-2.8;
When designing mHealth app interventions for dementia caregivers with a chronic condition, it is important to consider caregivers’ perceptions about how well mHealth apps can help their self-management and which app features would be most useful for self-management. Caregiving factors may not be relevant to caregivers’ intention to adopt mHealth apps. This is promising because mHealth strategies may overcome barriers to caregivers’ self-management. Future research should investigate reasons why caregivers with a low education level and low burden of chronic disease and treatment have significantly lower intention to adopt mHealth apps for self-management.
In the United States, more than 11 million family caregivers provide care to a loved one with Alzheimer disease or related dementias [
Previous research supports that family caregivers of people with dementia perform less self-management than noncaregivers and experience worse health and well-being outcomes [
mHealth strategies are effective in improving self-management and health outcomes of persons living with diabetes, mental health conditions, and cancer, among other chronic conditions [
The Technology Acceptance Model (TAM) is a well-known theoretical framework for exploring the factors associated with mHealth app adoption. The TAM was originally developed to explain the intention to adopt software systems [
Prior studies have expanded the TAM to improve its utility and predictive power [
Furthermore, the findings from other studies suggest that caregiving factors may be relevant to caregivers’ intention to adopt mHealth apps. For example, in the context of mHealth apps that support caregiving, caregivers with higher caregiver burden and strain had higher mHealth app use [
In addition, racial and ethnic groups have similar rates of smartphone ownership according to national surveys [
Taken together, although much progress has been made in expanding the TAM, there is still limited knowledge of the factors associated with mHealth app adoption among dementia caregivers with a chronic condition. Caregivers are often burdened to care for their own chronic health conditions, in addition to the multimorbidities of the person with dementia, and therefore have unique barriers to self-management compared with other populations [
Aim 1: to examine the relationships among dementia caregivers’ technological, self-management, and caregiving factors and their intention to adopt mHealth apps for self-management. Hypothesis 1: we hypothesized that technological and self-management factors would be positively, and caregiving factors would be negatively, associated with the intention to adopt mHealth apps for chronic disease self-management, controlling for the caregivers’ multimorbidities, age, gender, and income.
Aim 2: to explore whether the caregivers’ race or ethnicity and education moderate the relationship between the study variables and caregivers’ intention to adopt mHealth apps for chronic disease self-management.
We conducted a cross-sectional, correlational study and collected data in English and Spanish using computer-assisted telephone interviews and a web-based survey, both of which used the same web-based REDCap (Research Electronic Data Capture [
Using G*Power version 3.1.9.2 (Heinrich Heine University) and effect sizes from a recent study [
All study procedures were approved by the Johns Hopkins Medicine Institutional Review Board (IRB). The study survey was created and piloted with content experts. After entering it into REDCap, it was piloted on the web and over the phone with community members to ensure that the skip patterns, survey flow, and instructions were appropriate before implementation. As part of the survey, the team provided pictures of an evidence-based self-management mHealth app for persons with diabetes to standardize the caregivers’ conception of an mHealth app [
Data were collected in English from June 2019 to August 2020 and in Spanish from July 2020 to August 2020 (see
In addition, the team recruited on the web by posting advertisements on a Johns Hopkins University online news center and on social media (Google, Facebook, and YouTube) and by sending recruitment emails through a web-based research registry (ResearchMatch). These methods included an anonymous link to the eligibility screening survey. Interested individuals could click the link, complete the eligibility survey, and begin the web-based survey, if eligible. All eligible participants received information on the study purpose, procedures, risks, and benefits and consented to participate through IRB-approved oral or web-based consent. Data were stored in the REDCap database, to which only authorized, IRB-approved team members with password-protected accounts had access. All participants who completed the study survey were remunerated with a US $10 gift card.
The theoretical framework guiding the study was an expanded TAM, which included the factors relevant to caregivers and their self-management [
To measure the sociodemographic variables, we used questions from the US Census and national surveys. Income was captured with a well-validated question of financial strain (“How hard is it for you to pay for the very basics like food, housing, medical care, and heating?”) [
Revised Technology Acceptance Model guiding the study.
We operationalized the independent variables (perceived usefulness and perceived ease of use) and dependent variable (intention to adopt) using adapted versions of the 3 original TAM scales [
Social influence was measured using the Social Influence Scale developed when the TAM was expanded [
We measured caregiver burden using the 12-item short-form version of the Zarit Burden Interview (ZBI), which has been widely used in dementia caregiving research and found to have good internal consistency, test-retest reliability, and strong correlations with the full ZBI [
Finally, caregivers’ burden of chronic disease and treatment was operationalized using the Illness Intrusiveness Ratings Scale (IIRS). This 13-item instrument measures the degree to which a disease and its treatment disrupt one’s life and activities [
Some web-based surveys were anonymous. Thus, fake or fraudulent survey responses were potential issues that could affect research integrity [
Next, we examined the data for missing,
We used descriptive statistics (mean, median, and SD) to summarize the variables and examined the distributions of independent and dependent continuous variables. We also examined the correlation matrix of bivariate associations between the independent variables and the dependent variable. All TAM variables had left-skewed distributions, with 70.1% (82/117) of the participants choosing values above neutral (somewhat agree and higher; Table S1 in
We applied a data-driven and theoretical approach to dichotomize the TAM variables into high and low groups. Specifically, we used an approximate median split (55/45) and theoretical cutoff points for people who moderately agreed to strongly agreed that they intended to adopt mHealth apps and perceived mHealth apps as useful and easy to use. We used a similar approach for the social influence variable (people who more than somewhat agreed). The self-management and caregiver burden variables were normally distributed; thus, we dichotomized these variables at their medians. Finally, caregiving time was dichotomized into high (≥21 hours/week) and low (<21 hours/week), following a published cutoff score [
For hypothesis testing, each independent variable was individually regressed onto the outcome, controlling for age, gender, income, and multimorbidity, which have been associated with technology adoption in prior studies [
For moderation testing, we used the final model from the aim 1 analyses. Subsequently, we dichotomized race or ethnicity into White, non-Hispanic and people of color and education at its median (16 years). We created interaction terms for each dichotomized independent variable: race or ethnicity and education. All statistically significant interaction terms (
The study team recruited 498 people interested in the study (
CONSORT (Consolidated Standards of Reporting Trials) flow chart of study recruitment.
Sociodemographic characteristics of the study sample of dementia caregivers living with a chronic health condition (N=117).
Sociodemographic characteristic | Values | |||
Age (years), mean (SD) | 52.7 (17.4) | |||
Education (years), mean (SD) | 16 (3.3) | |||
|
||||
|
Female | 92 (78.6) | ||
|
Male | 25 (21.4) | ||
|
||||
|
White or non-Hispanic | 54 (46.2) | ||
|
Black or African American | 31 (26.5) | ||
|
Hispanic or Latino | 17 (14.5) | ||
|
Asian | 11 (9.4) | ||
|
Native American | 2 (1.7) | ||
|
Multiple | 2 (1.7) | ||
|
||||
|
Married or living as married | 58 (49.6) | ||
|
Never married | 34 (29.1) | ||
|
Widowed, divorced, or separated | 23 (219.7) | ||
|
Refused to answer | 2 (1.7) | ||
|
||||
|
Not at all or not very difficult | 53 (45.3) | ||
|
Somewhat difficult | 48 (41) | ||
|
Very or extremely difficult | 16 (13.7) | ||
|
||||
|
Value, mean (SD) | 4 (2.5) | ||
|
|
|||
|
|
Hypertension | 55 (47) | |
|
|
Depression | 50 (42.7) | |
|
|
Hyperlipidemia | 40 (34.2) | |
|
|
Rheumatoid arthritis or osteoarthritis | 39 (33.3) | |
|
|
Asthma | 32 (27.4) | |
|
|
Migraine or chronic headache | 32 (27.4) | |
|
|
Mental health condition | 27 (23.1) | |
|
|
Diabetes (type 1 and 2) | 24 (20.5) | |
|
|
Cataracts | 23 (19.7) | |
|
||||
|
Child or child-in-law | 53 (45.3) | ||
|
Grandchild | 22 (18.8) | ||
|
Spouse or significant other | 22 (18.8) | ||
|
Other (family member) | 12 (10.3) | ||
|
Friend | 8 (6.8) | ||
|
||||
|
No | 106 (90.6) | ||
|
Yes | 10 (8.5) | ||
|
Refused to answer | 1 (0.9) | ||
Time spent caregiving per week (hours), mean (SD) | 27.3 (30.8) |
aNine most common chronic conditions in the sample.
On average, the caregivers were aged approximately 53 years (SD 17.4), with an age range of 19-88 years. Most of the sample consisted of women (92/117, 78.6%), and more than half were minorities (63/117, 53.8%). Approximately half of the caregivers were married or living as married (58/117, 49.6%), and 29.1% (34/117) had never married. On average, the caregivers had completed 16 (SD 3.3) years of education, and more than half of the caregivers (64/117, 54.7%) reported that it was somewhat difficult to extremely difficult to manage on their income. Of the 117 participants, 53 (45.3%) were the child or child-in-law of the person with dementia, with an even proportion of caregivers being the spouse or significant other (22/117, 18.8%) or grandchild (22/117, 18.8%). The caregivers had, on average, 4 (SD 2.5) chronic health conditions, with a range of 1-15 (
The web-based survey respondents were, on average, approximately 20 years younger (t115=–7.81;
In bivariate associations, the intention to adopt mHealth apps was significantly associated with perceived usefulness (
After controlling for age, gender, income, and multimorbidity, we found that perceived usefulness (odds ratio [OR] 31, 95% CI 10-94;
The final aim 1 model is presented in
Final multiple logistic regression models for aims 1 and 2.
Variablea | Adjusted odds ratio (95% CI) | ||
|
Aim 1 final modelb | Aim 2 final modelc | |
|
|||
|
Multimorbidity | 0.87 (0.7-1.1) | 0.93 (0.74-1.2) |
|
Age (years) | 1.03 (0.99-1.1) | 1.03 (0.99-1.1) |
|
Gender | 0.6 (0.16-2.2) | 0.41 (0.1-1.6) |
|
Income | 0.66 (0.21-2.1) | 0.63 (0.16-2.4) |
|
|||
|
Perceived usefulness | 15 (4.3-51)d | 23 (5.6-97)d |
|
Perceived ease of use | 3.1 (0.95-10) | 2.4 (0.67-8.7) |
|
Social influence | 1.9 (0.64-5.5) | 1.8 (0.58-5.7) |
|
Burden of chronic disease or treatment (IIRSe) | 2.5 (0.68-9.2) | 0.31 (0.038-2.5) |
|
|||
|
Education | —f | 0.24 (0.034-1.6) |
|
IIRS × education | — | 31 (2.2-430)g |
aMeasurement of variables is as follows (variable: measurement)—multimorbidity: chronic disease counts from the Centers for Medicare & Medicaid Services Chronic Condition Warehouse [
bAim 1 final model statistics: Hosmer-Lemeshow test,
cAim 2 final model statistics: Hosmer-Lemeshow test,
d
eIIRS: Illness Intrusiveness Ratings Scale.
fVariables not tested in aim 1 analyses.
g
After exploring moderation, race or ethnicity did not significantly change the relationship between the independent variables and the outcome. The only statistically significant interaction term associated with the caregivers’ intention to adopt mHealth apps was education and burden of chronic disease or treatment (OR 31, 95% CI 2.2-430;
In the final model with the interaction terms included, perceived usefulness (OR 23, 95% CI 5.6-97;
The aim of this study was to explore the barriers to and facilitators of the intention to adopt mHealth apps for self-management among dementia caregivers with a chronic condition. In our study of 117 caregivers, we found that perceived usefulness explained 52% of the variance and was the strongest predictor of caregivers’ intention to adopt mHealth apps for their self-management. Furthermore, after controlling for perceived usefulness, other independent variables were no longer significantly associated with the intention to adopt mHealth apps. None of the caregiving variables were significantly associated with the caregivers’ intention to adopt mHealth apps in any model. We also found that caregivers with a high education level and greater burden of chronic disease and treatment had a significantly greater intention to adopt mHealth apps for their self-management than those with a low education level and low burden of chronic disease and treatment.
Perceived usefulness has consistently been a strong predictor of the intention to adopt mHealth solutions among older adults and persons with a chronic condition [
Although only perceived usefulness was statistically significant, perceived ease of use was clinically meaningful because caregivers who believed that mHealth apps were easy to use had 2.4 times greater intention to adopt them. This finding may not have reached statistical significance because of insufficient sample size or the age of our caregiving sample, reflecting a younger, more tech-savvy generation. For example, we only included caregivers who owned or had access to mobile devices. Previous studies have found that younger adults have higher mobile device ownership and mobile app use, as well as better technology skills than older generations [
In our caregiving sample, social influence had a larger, although statistically nonsignificant, OR of 1.8 (95% CI 0.58-5.7). Existing studies on the significance of social influence with regard to health-related technology adoption have been mixed. Some studies report that social influence is a significant facilitator of health-related technology adoption among general consumers [
This discrepancy in the findings may be related to differences in population, type of technology, or sample demographics. For example, our sample consisted of caregivers who were predominantly English-speaking, middle-aged, and the child or grandchild of the person with dementia. Compared with our caregiving sample, the sample in the study by Dai et al [
In our study, social influence reflected subjective norm (perceptions that people who are important in your life believe that you should perform an action) from the Theory of Reasoned Action [
Furthermore, our sample size (n=117) was smaller than the samples in the studies by Dai et al (n=350) [
Caregiver burden and the hours spent caregiving did not contribute significantly to explaining the intention to adopt mHealth apps among family caregivers. Although these 2 caregiving factors negatively impact caregivers’ self-care [
We found that the burden of chronic disease and treatment was not significantly associated with caregivers’ intention to adopt mHealth apps. Our study finding conflicts with that of existing studies. Other researchers have found that perceived disease threats were significantly associated with the intention to adopt mHealth solutions among persons with a chronic condition [
In our sample, the caregivers’ education and burden of chronic disease and treatment interacted to produce a greater and significant effect on their intention to adopt mHealth apps. The OR (31, 95% CI 2.2-430) should be interpreted with caution because of the smaller number of caregivers in the high and low groups (Table S3 in
Interpreted in the context of existing research, our study offers new insights into the factors related to caregivers’ intention to adopt mHealth apps for self-management. However, additional research is still needed to maximize mHealth app adoption in this population. Furthermore, the diversity of populations, mHealth strategies, and study findings substantiate the importance of user-centered design and the development of mHealth solutions with the end users as key stakeholders [
This study has some limitations. We recruited a convenience sample using community-based (Baltimore, Maryland) and web-based methods. Thus, our results may not be generalizable to all family caregivers of people with dementia, such as those who lack access to the internet or social media. However, web-based recruitment methods enabled us to reach a larger caregiving population across the United States, which may also improve the external validity of the findings. In addition, this study was cross-sectional; thus, relationships are associative, not causal. Another limitation is that only 1 Spanish-speaking caregiver completed the survey, although 11 were eligible. We speculate that this was due to the caregivers’ difficulties with navigating the REDCap survey, which does not allow researchers to change the language of prebuilt, English-only survey buttons and functionalities. The attrition of Spanish-speaking caregivers occurred when they navigated to a different part of the survey with nonmodifiable, English-only REDCap buttons. Future researchers should consider this critical limitation of the REDCap platform.
Another limitation is that the study was originally powered for linear regression. As our data violated the assumptions of linear regression, we needed to use logistic regression. This change increased the models’ degrees of freedom and reduced the power to detect differences among groups. Post hoc power analyses indicated that our study had 80% power to detect an OR of 3 or higher to be statistically significant at α=.05. Thus, we may be making a type II error with some of the independent variables in our final model (such as perceived ease of use and social influence). However, in scatterplot matrices, we did not observe a linear relationship between the caregiving factors and the outcome, thus reinforcing our finding that the caregiving variables may not be relevant to caregivers’ intention to adopt mHealth apps.
In our sample of caregivers with one or more chronic conditions, the perceived usefulness of mHealth apps was the strongest and most significant variable associated with their intention to adopt mHealth apps for self-management. Although ease of use and social influence were not statistically significant, they were clinically significant with larger ORs. Future research is needed to determine which app features are most useful for caregivers’ self-management, estimate effect sizes for sample size calculations, and systematically review how relationships vary by population or type of mHealth strategy.
Our findings also support the theory that the caregiving factors may not influence caregivers’ intention to adopt mHealth apps for self-management. Thus, mHealth solutions may overcome the barriers to caregivers’ self-management. Furthermore, caregivers with a high education level and greater burden of chronic disease and treatment have a higher likelihood of intending to adopt mHealth apps for self-management. Future research should explore the mechanisms by which education and self-management may interact.
Engaging dementia caregivers as stakeholders throughout the process of mHealth app conception, design, and testing can promote their adoption of mHealth apps. This process of user-centered design ensures that these apps are useful and easy to use, addresses factors relevant to caregivers, and builds support systems that encourage adoption.
CHERRIES (Checklist for Reporting Results of Internet E-Surveys) checklist.
Additional data on study variables and cross-tabulations for interaction term.
Checklist for Reporting Results of Internet E-Surveys
Illness Intrusiveness Ratings Scale
Institutional Review Board
odds ratio
Research Electronic Data Capture
Technology Acceptance Model
Zarit Burden Interview
The authors would like to thank Chloe Kwon, MSN, RN, for helping to create the Spanish REDCap survey, recruiting participants, and collecting and cleaning data; Hannah Parks, BSN, RN, for help with Spanish translation and collecting data; and Emerald Rivers, MSN, RN, for creating content for the study Facebook page. The authors would also like to thank Welldoc BlueStar for providing pictures of their evidence-based diabetes self-management app for inclusion in the study materials.
This study was supported by grants from the National Institute of Nursing Research (F31NR018373-01A1; KJWM), National Center for Advancing Translational Sciences (TL1TR003100-01; Mary Catherine Beach), Sigma Foundation for Nursing (Sigma/Doris Bloch Research Award), and Scholl Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
KJWM designed and conducted the study, cleaned and analyzed the data, interpreted the results, and drafted the first version of the manuscript. CB led the power analyses and contributed to the data analysis and interpretation of the results. HRH assisted with designing and conducting the study, analyzing data, and interpreting the results. All authors contributed to the conception and design of the study, revised the manuscript for important intellectual content, and approved the final version.
None declared.