With the paradigm shift in modern healthcare, mobile healthcare (mHealth) has emerged as a core instrument in public health policy. Existing policies are premised on the ‘Engagement-Based Model,’ which postulates that higher levels of user engagem...
With the paradigm shift in modern healthcare, mobile healthcare (mHealth) has emerged as a core instrument in public health policy. Existing policies are premised on the ‘Engagement-Based Model,’ which postulates that higher levels of user engagement lead to improved health outcomes; however, recent reports of discrepancies between quantitative engagement and clinical outcomes have fueled ongoing academic debate. This study defines this phenomenon as the ‘Engagement-Outcome Paradox.’ It aims to verify the reality of this paradox using empirical data from public mHealth programs in South Korea and to deeply explore the operating mechanisms of digital health interventions by identifying the mediating effect of ‘health behavior change’ underlying the phenomenon.
To this end, the study analyzed integrated data from 768 participants over three years (2019, 2023, 2024) from the ‘Songpa-gu Public Health Center Mobile Healthcare Program’ in Seoul. The research model critically examines the limitations of the conventional linear assumption that ‘app engagement’ leads directly to ‘health outcomes,’ proposing an alternative pathway where ‘app engagement (independent variable)’ indirectly influences the improvement of ‘health outcomes (dependent variable)’ through ‘health behavior change (mediator).’ Analytical methods included paired t-tests, Analysis of Covariance (ANCOVA), Structural Equation Modeling (SEM), and Bootstrapping. In particular, a rigorous analytical approach was adopted by including pre-test values as covariates to control for baseline effects and regression to the mean.
The major research findings are as follows. First, the program significantly improved the average health level of the participant group (Hypothesis 1 supported). After 24 weeks, the participant group showed statistically significant improvements (p < .001) in four indicators: systolic blood pressure, diastolic blood pressure, waist circumference, and HDL cholesterol. This suggests that public mHealth programs are effective policy tools for managing key cardiovascular risk factors. However, triglycerides showed a significant increase (p = .041), indicating limitations in inducing dramatic changes in metabolic indicators solely through short-term lifestyle interventions. Additionally, while the program’s intervention focused on encouraging autonomous ‘physical activity and dietary control,’ this result can be interpreted as reflecting the characteristics of triglycerides, which react sensitively to short-term dietary factors such as carbohydrate or alcohol intake.
Second, the study confirmed the existence of the phenomenon where quantitative input (app usage) does not directly translate to qualitative outcomes (health improvement), namely the ‘Engagement-Outcome Paradox’ (Hypothesis 2). The analysis revealed that, with the exception of waist circumference, the level of quantitative app engagement (frequency of mission completion) did not have a significant effect on health outcome improvements across the five major clinical indicators. This implies that the linear assumption of “quantity of engagement equals magnitude of outcome,” commonly accepted in the policy field, does not apply to actual clinical indicators.
Third, regarding the verification of the mediating effect of health behavior change, app engagement significantly predicted positive health behavior changes (Hypothesis 3a supported); however, the pathway from these behavioral changes to improvements in clinical health indicators was not statistically significant (Hypothesis 3b rejected). Consequently, the mediating effect of health behavior change was rejected for all indicators. This study interprets these findings as a gap between subjective survey data (behavior) and objective biometric data (outcomes), as well as the potential complex operation of unmeasured ‘third paths,’ such as psychological and social support, beyond the measured behavioral variables. It also suggests the possibility that the outcomes observed over the 24-week period may be temporary ‘acute effects’ driven by external stimuli rather than the result of behavioral habituation.
Based on these analysis results, this study offers the following policy suggestions. First, policy performance indicators should be reorganized from a focus on simple quantitative input to a ‘qualitative process’ that evaluates the sustainability and intensity of behavior. Second, a ‘Hybrid High-Intensity Intervention’ model that incorporates professional medical intervention should be introduced for high-risk metabolic syndrome groups. Third, to overcome the limitations of subjective surveys, a precision monitoring and analysis system utilizing raw data from wearable devices should be established. Finally, the study raises the necessity of Longitudinal Study with Follow-up observations to verify the long-term causality between behavior change and outcomes.