Propensity scores and preventive drug use in the elderly

摘要 包括药物、手术干预和疫苗接种在内的医学治疗在艾滋病毒/艾滋病中发挥着至关重要的作用。 预防和治疗老年人的慢性病和传染病。提供证据基础的研究 这些治疗的益处和危害包括实验研究（随机对照试验， 随机对照试验）和非实验性研究（观察性、流行病学或最近的真实世界证据）。 非实验性研究为老年人提供了至关重要的证据，特别是那些有多种疾病的人。 多种药物治疗的慢性疾病，因为此类成人很少被纳入RCT;这对于 种族边缘化的老年人，历史上在审判中代表性不足。这种非实验性研究 通常涉及混杂因素，老年人的许多混杂因素-包括虚弱-可能难以在 真实世界的数据。如果不对这种混杂因素进行调整，可能会扭曲结果，导致次优甚至 有害的治疗决定（如果非实验性研究是唯一可用的证据）或 昂贵的RCT治疗没有好处。及时评估老年人的治疗获益和危害 真实世界（相对于实验）的设置需要开发、改进和应用方法， 减少非实验性研究中的混杂偏倚。 自2005年以来，我们在R 01/R56 AG 023178和AG 056479的资助下讨论了这一主题。这 这些努力导致了方法的实质性进步，以提高非实验研究的有效性， 倾向评分法。使用经验数据和广泛的模拟，我们开发了新的 减少混杂的分析技术，包括倾向评分校准和排除患者 与预测相反。我们以口头报告、海报和 讲习班/研讨会和112篇同行评审的出版物，包括23篇在顶级流行病学期刊（AJE和 流行病学），顶级药物流行病学杂志（PDS）23篇，医疗保健10篇，以及几篇顶级药物流行病学杂志（PDS） 医学期刊（JAMA、JNCI、Diabetes Care和JAGS）。 在我们过去15年工作的基础上，提高非实验方法评估的有效性 老年人治疗的预防效果，我们提出了重要的和创新的目标， 非实验性研究的严谨性我们将使用来自经验的激励性例子来解决这些目标。 数据和广泛的模拟，以提高真实世界的证据的有效性。我们将利用我们的 跨学科的专业知识和访问与其他数据源链接的医疗保险数据，以改善现实世界 使用倾向分数的研究。这一建议是及时的：链接数据变得越来越可用， 监管决策中的真实世界证据是一个中心讨论点。这项工作将直接告知 临床相关的治疗决策，提供RCT无法提供的可靠证据，并改善个体 和老年人的健康。

ABSTRACT Medical treatments, including drugs, surgical interventions, and vaccinations, play a crucial role in the prevention and treatment of chronic and infectious disease in older adults. Studies providing the evidence base for the benefits and harms of these treatments include both experimental studies (randomized controlled trials, RCTs) and nonexperimental studies (observational, epidemiologic, or, more recently: real-world evidence). Nonexperimental studies provide critically important evidence for older adults, especially those with multiple chronic conditions with polypharmacy, as such adults are rarely included in RCTs; this is doubly important for racially marginalized older adults that are historically underrepresented in trials. Such nonexperimental studies often involve confounding, and many confounders in older adults-including frailty-can be difficult to capture in real-world data. Failure to adjust for this confounding can distort findings and lead to suboptimal or even harmful treatment decisions (if nonexperimental studies are the only evidence available) or the conduct of costly RCTs of treatments without benefits. Timely assessment of treatment benefit and harm in older adults in real-world (versus experimental) settings requires development, refinement, and application of methods to reduce confounding bias in nonexperimental studies. We have addressed this topic since 2005 with funding from R01/R56 AG023178 and AG056479. This effort has led to substantial advances in methods to improve the validity of nonexperimental research applying propensity score methods. Using both empirical data and extensive simulations, we have developed novel analytic techniques to reduce confounding, including propensity score calibration and the exclusion of patients treated contrary to prediction. We have disseminated our results in the form of oral presentations, posters, and workshops/symposia and 112 peer-reviewed publications, including 23 in top epidemiologic journals (AJE and Epidemiology), 23 in the top pharmacoepidemiologic journal (PDS), 10 in Medical Care, and several in top medical journals (JAMA, JNCI, Diabetes Care, and JAGS). Building on our work over the last 15 years to increase the validity of nonexperimental methods to assess the preventive effects of treatments in older adults, we propose significant and innovative aims that improve the rigor of nonexperimental research. We will address these aims using motivating examples from empirical data and extensive simulations to improve the validity of real-world evidence. We will take advantage of our interdisciplinary expertise and access to Medicare data linked with other data sources to improve real-world studies using propensity scores. This proposal is timely: linked data are becoming more available and the role of real-world evidence in regulatory decision making is a central discussion point. This work will directly inform clinically relevant treatment decisions, provide reliable evidence unavailable from RCTs, and improve individual and population health for older adults.