Scholarlink

<p>Stepped-wedge cluster-randomized trials (SW-CRTs) are increasingly used to evaluate healthcare interventions, yet contamination arising from professional collaboration remains an underappreciated threat to valid inference. We examined spillover effects in a SW-CRT assessing a video-game intervention to increase Advance Care Planning (ACP) billing among hospitalists, which previously showed no direct intervention effect (OR 0.96, 95% CI 0.88–1.06, <i>p</i> = 0.42). Leveraging physician social networks constructed from shared-patient encounters, we quantified diffusion of intervention exposure through professional ties. In a network-adjusted intent-to-treat model excluding within-hospital connections, the intervention effect became positive though not statistically significant (OR 1.08, 95% CI 0.94–1.24, <i>p</i> = 0.26), while spillover from intervened peers showed a strong association with ACP billing (OR 1.20, 95% CI 1.13–1.27, <i>p</i> &lt; 0.001). In an alternative as-treated formulation that included within-hospital ties and modeled exposure as the number of intervened physicians per patient, the intervention effect strengthened and achieved significance (OR 1.11, 95% CI 1.03–1.20, <i>p</i> = 0.006). A persistent positive interaction between intervention and spillover (OR 1.01, 95% CI 1.00–1.02, <i>p</i> = 0.03) indicated that the direct effect was amplified in more highly connected networks. Simulation results supported these findings, demonstrating that omitting spillover biases estimates toward the null, while network-adjusted models recover both direct and indirect effects with unbiased point estimators and coverage probabilities of interval estimators close to the nominal level. Together, these findings underscore the importance of incorporating social network structure into SW-CRT analyses to more accurately capture both the direct and diffusive pathways through which interventions influence behavior.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gaming the system: evaluating spillover in a video game intervention for advance care planning using physician social networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月13日

查看全文
  • Carly A. Bobak,
  • Deepika Mohan,
  • Megan A. Murphy,
  • Amber E. Barnato,
  • A. James O’Malley

摘要

Stepped-wedge cluster-randomized trials (SW-CRTs) are increasingly used to evaluate healthcare interventions, yet contamination arising from professional collaboration remains an underappreciated threat to valid inference. We examined spillover effects in a SW-CRT assessing a video-game intervention to increase Advance Care Planning (ACP) billing among hospitalists, which previously showed no direct intervention effect (OR 0.96, 95% CI 0.88–1.06, p = 0.42). Leveraging physician social networks constructed from shared-patient encounters, we quantified diffusion of intervention exposure through professional ties. In a network-adjusted intent-to-treat model excluding within-hospital connections, the intervention effect became positive though not statistically significant (OR 1.08, 95% CI 0.94–1.24, p = 0.26), while spillover from intervened peers showed a strong association with ACP billing (OR 1.20, 95% CI 1.13–1.27, p < 0.001). In an alternative as-treated formulation that included within-hospital ties and modeled exposure as the number of intervened physicians per patient, the intervention effect strengthened and achieved significance (OR 1.11, 95% CI 1.03–1.20, p = 0.006). A persistent positive interaction between intervention and spillover (OR 1.01, 95% CI 1.00–1.02, p = 0.03) indicated that the direct effect was amplified in more highly connected networks. Simulation results supported these findings, demonstrating that omitting spillover biases estimates toward the null, while network-adjusted models recover both direct and indirect effects with unbiased point estimators and coverage probabilities of interval estimators close to the nominal level. Together, these findings underscore the importance of incorporating social network structure into SW-CRT analyses to more accurately capture both the direct and diffusive pathways through which interventions influence behavior.