Abstract / Summary
Background: Understanding protective behavior during infectious disease outbreaks is critical for public health policy. App-based experimental epidemic games (epigames) offer a novel empirical approach, but their external validity and feasibility as rigorous randomized-trial platforms require evaluation. We examined whether real-life health beliefs transferred into an epigame, whether opportunity cost influenced voluntary quarantine, and whether the design could support subsequent behavioral experiments. Methods: We conducted a preregistered 15-day randomized controlled trial using the Epigames smartphone app at the American University of Iraq - Baghdad, with a total of 539 registered participants. The app sampled Bluetooth proximity at sub-minute resolution, simulated respiratory-pathogen transmission through the contact network, and elicited daily quarantine choices. Participants were randomized to a low-cost barrier, receiving 5 points for quarantine and 10 for remaining active, or a high-cost barrier, receiving 5 and 17 points, respectively; a willingness-to-accept pilot informed the contrast. Real-life (S1) and in-game (S2) surveys measured susceptibility, severity, quarantine self-efficacy, and perceived benefits. We assessed belief correspondence with Holm-adjusted Spearman correlations and prospective cumulative-logit models and analyzed choices using participant-day binomial generalized estimating equations (GEE) and participant-level quasi-binomial models. Results: The trial integrated app-based randomization, longitudinal contact measurement, repeated choices, and timed surveys in a naturalistic campus setting. Participants contributed 1,611 participant-days and 1,615 recorded choices. Among 257 participants with matched surveys, all four corresponding S1-S2 beliefs were positively associated (Spearman's {rho}=0.126-0.353; all Holm-adjusted p<0.05), and prospective ordinal models confirmed matched-domain correspondence. Quarantine rates were 9.6% under the low barrier and 8.9% under the high barrier. The high barrier did not detectably alter quarantine overall (GEE OR 0.90, 95% CI 0.58-1.40, p=0.650; quasi-binomial OR 0.93, 95% CI 0.58-1.48, p=0.753), and joint belief-by-group tests were null. Two non-confirmatory patterns warrant prospective investigation: baseline severity showed a small positive association with quarantine rate ({rho}=0.132, raw p=0.0155, Holm p=0.0619), and the estimated high-barrier effect for "skeptical" participant (those with responses at the floor of all S2 belief scales) was large but specification-sensitive (GEE OR 0.060, p=0.039; quasi-binomial OR 0.058, p=0.056). Women were more likely than men ever to quarantine (arm-adjusted OR 2.76, p=0.0002), but there was no gender-by-barrier interaction. Conclusion: This study demonstrates the feasibility of a preregistered epigame RCT combining randomized incentives, passively measured contacts, repeated protective choices, and prospectively measured beliefs. The strongest evidence concerned real-life to in-game belief correspondence; the severity and low-belief barrier patterns are targets for adequately powered, prospectively specified replication. Technical limitations, uneven prompt response, and gameplay that may not have sustained engagement could have attenuated behavioral contrasts. Epigames provide a promising foundation for network-aware behavioral epidemiology. Protocol refinement and multisite replication are needed before inferring real-world behavioral validity.