2020/05/30 by Charles F. Manski, Aleksey Tetenov, Manski, Charles F. +1
Economics, Econometrics and Finance · Mathematics · #Advanced Causal Inference Techniques #Econometrics (econ.EM) #FOS: Biological sciences #FOS: Computer and information sciences #FOS: Economics and business #Health Systems, Economic Evaluations, Quality of Life #Methodology (stat.ME) #Quantitative Methods (q-bio.QM) #Statistical Methods in Clinical Trials
paper · pdf · doi:10.48550/arxiv.2006.00343
openalex publication_date 2020/05/30 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
As the COVID-19 pandemic progresses, researchers are reporting findings of\nrandomized trials comparing standard care with care augmented by experimental\ndrugs. The trials have small sample sizes, so estimates of treatment effects\nare imprecise. Seeing imprecision, clinicians reading research articles may\nfind it difficult to decide when to treat patients with experimental drugs.\nWhatever decision criterion one uses, there is always some probability that\nrandom variation in trial outcomes will lead to prescribing sub-optimal\ntreatments. A conventional practice when comparing standard care and an\ninnovation is to choose the innovation only if the estimated treatment effect\nis positive and statistically significant. This practice defers to standard\ncare as the status quo. To evaluate decision criteria, we use the concept of\nnear-optimality, which jointly considers the probability and magnitude of\ndecision errors. An appealing decision criterion from this perspective is the\nempirical success rule, which chooses the treatment with the highest observed\naverage patient outcome in the trial. Considering the design of recent and\nongoing COVID-19 trials, we show that the empirical success rule yields\ntreatment results that are much closer to optimal than those generated by\nprevailing decision criteria based on hypothesis tests.\n