流感疫苗接种彩票激励的效果：来自大规模随机试验与因果森林分析的证据
Effects of Lottery Incentives for Influenza Vaccination: Evidence from a Large-Scale Randomized Trial and Causal Forest Analysis
NBER Working Papers · 2026 · [{"name": "Kelsey Moran", "affiliation": []}, {"name": "Gail M. Rosenbaum", "affiliation": []}, {"name": "Amir Goren", "affiliation": []}, {"name": "Michelle N. Meyer", "affiliation": []}, {"name": "Christopher F. Chabris", "affiliation": []}, {"name": "Joseph J. Doyle Jr.", "affiliation": []}]
中文摘要
经济激励可能促进疫苗接种，但其成本效益取决于能否精准识别反应最为敏感的患者。我们在2021—2022年流感季期间，针对某一体化医疗系统中的57,579名成年患者开展了一项随机对照试验。即将就诊的患者被随机分配至无接触对照组、简单提醒组、1美元现金激励组，或最高奖金为5,000美元的刮刮乐彩票激励组。任一主动干预均使患者在预约就诊时的疫苗接种率提高2个百分点（较对照组均值提高8%），并使其在整个流感季的接种率提高1.5个百分点。三个主动干预组之间的差异不显著，表明相较于单纯提醒，小额经济激励并未带来额外收益。利用电子健康记录和地区层面特征，我们采用因果森林估计条件平均处理效应。预测的反应程度存在显著差异：处于最高四分位组的患者对激励作出反应的可能性是最低四分位组患者的3—4倍，而提醒的效应则相对同质。这些发现表明，低成本提醒能够提高流感疫苗接种率，而机器学习方法有助于识别对小额经济激励反应最为敏感的患者。
Abstract
Financial incentives may encourage vaccination, but their cost-effectiveness depends on targeting patients who are most responsive. We conducted a randomized controlled trial with 57,579 adult patients in an integrated health system during the 2021–2022 influenza season. Patients with an upcoming appointment were randomized to a no-contact control, a simple reminder, a $1 cash incentive, or a scratch-off lottery ticket incentive worth up to $5,000. Any active intervention increased vaccination at the scheduled appointment by 2 percentage points (an 8% increase over the control mean) and by 1.5 percentage points over the full influenza season. The three active arms did not differ significantly, indicating that small financial incentives provided no additional benefit beyond a reminder. Using electronic health records and area-level characteristics, we used causal forests to estimate conditional average treatment effects. Predicted responsiveness varied substantially: patients in the top quartile were 3–4 times more likely to respond to incentives than those in the bottom quartile, whereas reminder effects were relatively homogeneous. These findings suggest that low-cost reminders can increase influenza vaccination, while machine-learning methods can help identify patients most responsive to small financial incentives.