AI-Supported Prevention Program Linked to Lower Stress and Higher Psychological Capital
An AI-supported digital prevention program from PREVERY was associated with measurable gains in self-reported mental and health-related outcomes in a prospective before-and-after study. The most consistent signal was a rise in psychological capital alongside lower perceived stress during the active intervention period, followed by partial rebound in stress after the program ended. In a field crowded with wellness apps that promise broad benefit but rarely quantify change, this preprint offers a clearer clinical pattern: participants appeared to feel more capable, more informed, and less stressed while the program was underway.
The study enrolled 253 participants, with 113 included in the final analysis. Outcomes were assessed at baseline, after six weeks, and again three months after the interventional phase using a Numerical Analog Scale, the Perceived Stress Scale, and the PsyCap Scale. Researchers also tracked self-reported current health status, health literacy, health awareness, self-efficacy, and motivation. The design cannot establish causality the way a randomized trial could, but the repeated measures structure still captures a useful bedside question: can a digital prevention program shift how people perceive their stress load and health behavior capacity over time?
Stress Fell During Active Use, Then Rose Again After the Program
The clearest temporal pattern involved perceived stress. PSS scores decreased from baseline to six weeks, then increased again by the three-month follow-up after the intervention phase ended. Every pairwise comparison was statistically significant. That pattern is clinically believable. During an active coaching period, participants may receive repeated prompts, structured reflection, and behavior-oriented feedback that reduce cognitive overload and restore a sense of control. Once the external scaffold is removed, stress can drift upward again if the underlying life demands remain unchanged.
A single sentence captures the most important dynamic in the dataset: the stress benefit did not simply appear, it also faded.
The psychological capital result ran in the opposite direction. CPC increased significantly from baseline to six weeks and remained elevated at three months, with no significant drop between the post-intervention and follow-up assessments. Psychological capital typically reflects optimism, resilience, hope, and confidence in one’s ability to act. In practical terms, an intervention that strengthens these constructs may help patients sustain healthier choices even after intensive digital support ends. The durability of the CPC signal therefore matters more than a transient score shift, because sustained self-belief is often what separates short-lived engagement from actual behavior change.
Health Literacy and Self-Efficacy Improved With Use
Several self-reported health constructs also improved over time. Current health status, health literacy, health awareness, and self-efficacy all showed significant time effects, with higher values at three months than at baseline. Health awareness rose as early as the six-week time point, while self-efficacy showed gains both by six weeks and again between six weeks and three months. Motivation did not change significantly.
Those findings fit the logic of digital health coaching. Education alone rarely changes behavior, but repeated exposure to personalized prevention content can help users recognize patterns, name symptoms, and make small decisions with greater confidence. When an app combines behavioral prompts with AI-supported structure, it may function less like a static information tool and more like a remote behavioral companion. The mechanism is plausible: better health literacy can reduce uncertainty, and reduced uncertainty can translate into stronger self-efficacy. Once people believe they can act effectively, they are more likely to sustain preventive routines such as sleep regulation, stress management, movement, or symptom monitoring, even though the exact behaviors were not specified in the preprint.
The absence of a significant motivation effect is equally informative. Motivation is often volatile and context dependent; it may not move in lockstep with awareness or confidence. A person can become more informed and feel more capable without necessarily reporting a broader increase in drive. In prevention research, that distinction is important because capability and readiness to act do not always travel together.
A Digital Prevention Model Aimed at Strained Health Systems
The study was framed against rising burden from diseases related to chronic stress. That backdrop gives the results practical relevance. Health coaching is effective in principle, but individual coaching is expensive and time consuming, which limits scale. A digital model can extend contact frequency without requiring proportional clinician time, and AI support may allow more tailored nudges than a purely static program. In that sense, PREVERY’s platform is less a substitute for human care than a force multiplier for low-intensity prevention.
The trial was registered as DRKS00037999, and ethics approval came from the Ethics Committee of the German Sport University Cologne on 25 September 2025. Even with that oversight, the conflict-of-interest profile deserves attention. Two authors were co-founders of PREVERY UG and had a financial interest in the app’s commercial success, while the study was designed in part to support certification by the Central Review Board for Prevention. Authors with a conflict were not involved in data analysis. Those disclosures do not invalidate the findings, but they do make independent replication especially important before anyone treats the results as generalizable evidence.
What the Preprint Can and Cannot Support
Because this is a before-and-after observational study, it is vulnerable to regression to the mean, expectancy effects, and the simple fact that people who start and stay in a digital program may differ from those who drop out. The final analysis included 113 of 253 enrolled participants, which leaves open the possibility of selection bias. The outcomes were self-reported rather than biologically verified, so the study measures perceived health and perceived capability, not hard clinical endpoints. No data here show reduced hospitalizations, fewer stress-related diagnoses, or lower medication use.
Still, the pattern is coherent. Psychological capital improved and stayed elevated, while perceived stress fell during the active phase and then partially returned after the intervention ended. Self-reported health literacy, awareness, self-efficacy, and overall health status all moved in a favorable direction. For digital prevention, those are not trivial signals. They suggest that structured AI-supported coaching may improve the psychological prerequisites for behavior change, even if motivation itself remains stubbornly unchanged.
The more difficult question now is not whether a digital prevention tool can nudge survey scores over a few months. It is whether those nudges can be translated into durable health behavior, maintained engagement, and clinically meaningful outcomes in a larger, independently tested cohort.