Research Techniques Made Simple Databases for Clinical Research

Research Techniques Made Simple: Databases for Clinical Research Katrina Abuabara, MD MA David Margolis, MD Ph. D University of Pennsylvania

What Is the Research Question? • The usefulness and validity of databases depend on the research question • Research questions are often framed around an exposure and an outcome – Exposure: environmental, medication, risk factor, or disease – Outcomes: onset of disease (incidence), presence of disease (prevalence), or severity of duration of a disease or symptom

Which Epidemiologic Study Design?

What Type of Electronic Database? • Repurposed data: originally generated for purposes other than clinical research – Administrative claims data – Electronic medical record (EMR) data • Ad hoc data: designed for a particular study, often prospective cohort or registry • Hybrid data: Blends features of the prior two categories

What Type of Electronic Database?

Design of a New Patient Database • Consider measures to optimize data collection, patient follow-up, data quality, and data security • Consider the generalizability of the results when selecting patients • Consider whether to include a valid comparison group

Considerations for the Design of a New Registry or Database 1 1. PCORI Standards Committee (2013) The PCORI Methodology Report. http: //www. pcori. org/content/research-methodology.

Potential Sources of Bias • Bias is a systematic deviation of a study’s result from a true value – Information bias: systematic differences in the accuracy or completeness of data leading to differential misclassification of individuals – Selection bias: systematic differences in the probability of including subjects in the study (or being lost to follow-up) leading to spurious conclusions

Potential Sources of Imprecision • Imprecision may arise from the study size or from the measurement of exposures, confounders, or outcomes • Detailed chart review and physician query may be used to evaluate the validity of measurements • Dermatologic outcomes are not often precisely measured in electronic databases