WS-0010 - Use Of External Data In Pharmacoepidemiology

Til Stürmer, Michele Jonsson Funk, Olaf Klungel, Kaatje Bollaerts

Live Session with Q&A
September 22, 2020
10:00 am EDT
Confirm Attendance

Background: Limitations of available data is one of the principal problems in pharmacoepidemiology. Multiple methods have been proposed to mitigate bias due to uncontrolled confounding, exposure, and outcome misclassification. Some of the most effective tools leverage external data.

Objectives: The symposium will provide an overview of methods related to external data, orient the audience to existing tools/resources available to incorporate external data, and generally encourage the audience to use external data to improve study validity.

Description: Three 15’ - 20’ presentations followed by 5’ Q&A each followed by ~20’ podium discussion.1. "Validation studies and data linkages" Administrative claims and other routinely collected data provide the foundation for many drug utilization, safety, and effectiveness studies. These databases provide a rich source of timely health care information on large, well‐defined populations yet lack potentially important information. The increasing ability to link data sources provides researchers with opportunities to add clinical detail to well defined cohorts. The presentation will cover issues related to ongoing linkages of US Medicare claims with electronic health records (EHR) and provide examples on how such linked data can be used to address a variety of issues, including uncontrolled confounding. 2. "Methods to control for unmeasured confounders based on validation data" The presentation will provide an overview of commonly used methods to address unmeasured confounders in the design and analysis phase and provide recommendations for proper application in pharmacoepidemiology. Methods that utilize additional information on confounders collected in a validation study (e.g., external adjustment, propensity score calibration, two-stage sampling, and multiple imputation) will be discussed highlighting the importance of using both statistical evidence and substantial clinical knowledge for the interpretation of study results. 3. "Multiple-bias analysis based on parameters assessed in external data" The use of quantitative bias analyses is advocated by several key-opinion leaders and international health authorities including regulatory agencies. Such bias analyses involve bias parameters for which the values can be chosen to be either wide-ranging and hypothetical (‘Which bias levels are needed to reach a certain value?’) or evidence-based (‘Which result would be obtained given a certain realistic level of bias?’). Traditional deterministic bias analyses can be improved by randomly sampling from the distributions of multiple bias parameters to obtain a distribution of the adjusted effect measure.