Predictors of Dropout in a Digital Intervention for the Prevention and Treatment of Depression in Patients With Chronic Back Pain: Secondary Analysis of Two Randomized Controlled Trials

Background

Depression is a common comorbid condition in individuals with chronic back pain (CBP), leading to poorer treatment outcomes and increased medical complications. Digital interventions have demonstrated efficacy in the prevention and treatment of depression; however, high dropout rates are a major challenge, particularly in clinical settings.

Objective

This study aims to identify the predictors of dropout in a digital intervention for the treatment and prevention of depression in patients with comorbid CBP. We assessed which participant characteristics may be associated with dropout and whether intervention usage data could help improve the identification of individuals at risk of dropout early on in treatment.

Methods

Data were collected from 2 large-scale randomized controlled trials in which 253 patients with a diagnosis of CBP and major depressive disorder or subclinical depressive symptoms received a digital intervention for depression. In the first analysis, participants’ baseline characteristics were examined as potential predictors of dropout. In the second analysis, we assessed the extent to which dropout could be predicted from a combination of participants’ baseline characteristics and intervention usage variables following the completion of the first module. Dropout was defined as completing <6 modules. Analyses were conducted using logistic regression.

Results

From participants’ baseline characteristics, lower level of education (odds ratio [OR] 3.33, 95% CI 1.51-7.32) and both lower and higher age (a quadratic effect; age: OR 0.62, 95% CI 0.47-0.82, and age²: OR 1.55, 95% CI 1.18-2.04) were significantly associated with a higher risk of dropout. In the analysis that aimed to predict dropout following completion of the first module, lower and higher age (age: OR 0.60, 95% CI 0.42-0.85; age²: OR 1.59, 95% CI 1.13-2.23), medium versus high social support (OR 3.03, 95% CI 1.25-7.33), and a higher number of days to module completion (OR 1.05, 95% CI 1.02-1.08) predicted a higher risk of dropout, whereas a self-reported negative event in the previous week was associated with a lower risk of dropout (OR 0.24, 95% CI 0.08-0.69). A model that combined baseline characteristics and intervention usage data generated the most accurate predictions (area under the receiver operating curve [AUC]=0.72) and was significantly more accurate than models based on baseline characteristics only (AUC=0.70) or intervention usage data only (AUC=0.61). We found no significant influence of pain, disability, or depression severity on dropout.

Conclusions

Dropout can be predicted by participant baseline variables, and the inclusion of intervention usage variables may improve the prediction of dropout early on in treatment. Being able to identify individuals at high risk of dropout from digital health interventions could provide intervention developers and supporting clinicians with the ability to intervene early and prevent dropout from occurring.

Predicting Dropouts From an Electronic Health Platform for Lifestyle Interventions: Analysis of Methods and Predictors

BACKGROUND

The increasing prevalence and economic impact of chronic diseases challenge health care systems globally. Digital solutions can potentially improve efficiency and quality of care, but these initiatives struggle with nonusage attrition. Machine learning methods have been proven to predict dropouts in other settings but lack implementation in health care.

OBJECTIVE

This study aimed to gain insight into the causes of attrition for patients in an electronic health (eHealth) intervention for chronic lifestyle diseases and evaluate if attrition can be predicted and consequently prevented. We aimed to build predictive models that can identify patients in a digital lifestyle intervention at high risk of dropout by analyzing several predictor variables applied in different models and to further assess the possibilities and impact of implementing such models into an eHealth platform.

METHODS

Data from 2684 patients using an eHealth platform were iteratively analyzed using logistic regression, decision trees, and random forest models. The dataset was split into a 79.99% (2147/2684) training and cross-validation set and a 20.0% (537/2684) holdout test set. Trends in activity patterns were analyzed to assess engagement over time. Development and implementation were performed iteratively with health coaches.

RESULTS

Patients in the test dataset were classified as dropouts with an 89% precision using a random forest model and 11 predictor variables. The most significant predictors were the provider of the intervention, 2 weeks inactivity, and the number of advices received from the health coach. Engagement in the platform dropped significantly leading up to the time of dropout.

CONCLUSIONS

Dropouts from eHealth lifestyle interventions can be predicted using various data mining methods. This can support health coaches in preventing attrition by receiving proactive warnings. The best performing predictive model was found to be the random forest.

The Continued Use of Mobile Health Apps: Insights From a Longitudinal Study

BACKGROUND

Mobile health (mHealth) apps that support individuals pursuing health and wellness goals, such as weight management, stress management, smoking cessation, and self-management of chronic conditions have been on the rise. Despite their potential benefits, the use of these tools has been limited, as most users stop using them just after a few times of use. Under this circumstance, achieving the positive outcomes of mHealth apps is less likely.

OBJECTIVE

The objective of this study was to understand continued use of mHealth apps and individuals' decisions related to this behavior.

METHODS

We conducted a qualitative longitudinal study on continued use of mHealth apps. We collected data through 34 pre- and postuse interviews and 193 diaries from 17 participants over two weeks.

RESULTS

We identified 2 dimensions that help explain continued use decisions of users of mHealth apps: users' assessment of mHealth app and its capabilities (user experience) and their persistence at their health goals (intent). We present the key factors that influence users' assessment of an mHealth app (interface design, navigation, notifications, data collection methods and tools, goal management, depth of knowledge, system rules, actionable recommendations, and user system fit) and relate these factors to previous literature on behavior change technology design. Using these 2 dimensions, we developed a framework that illustrated 4 decisions users might make after initial interaction with mHealth apps (to abandon use, limit use, switch app, and continue use). We put forth propositions to be explored in future research on mHealth app use.

CONCLUSIONS

This study provides insight into the factors that shape users' decisions to continue using mHealth apps, as well as other likely decision scenarios after the initial use experience. The findings contribute to extant knowledge of mHealth use and provide important implications for design of mHealth apps to increase long-term engagement of the users.