Length of look-back periods in studies using administrative data: a scoping review

Administrative data, including electronic health record (EHR) and claims data, are increasingly used to generate real-world evidence for epidemiologic research. Unlike clinical trials, which assess baseline conditions at the initial visit, observational studies using these data must define an appropriate look-back period to identify baseline characteristics. However, to our knowledge, no established guidelines exist on the optimal length of look-back periods by clinical outcome. We conducted a scoping review, with a lens on methodological studies, to assess the frequency and variability of look-back periods used in epidemiologic studies. We searched PubMed for studies that used look-back periods using EHRs, claims data, or international registries. Our search yielded 262 studies spanning a broad range of clinical outcomes. Look-back periods varied substantially, ranging from 12 hours to 39 years, depending on the clinical outcome. The most common clinical areas included pharmacology, cardiovascular disease, cancer, and health care utilization. Our review underscores the need for tailored look-back periods based on the clinical context; we suggest a one-size-fits-all approach may be inadequate. Research should focus on the impact of look-back periods on less-studied outcomes. This review provides a resource for researchers to make informed decisions on the appropriate look-back periods in epidemiologic studies using administrative data.