About Us |
Committed to improving the transparency, reproducibility and validity of longitudinal healthcare database research.
About Us |
|
Relevance in healthcareReplication is a cornerstone of the scientific method. Historically, public confidence in the validity of healthcare database research has been low. Drug regulators, patients, clinicians, and payers have been hesitant to trust evidence from databases due to high profile controversies with overturned and conflicting results. This has resulted in underuse of a potentially valuable source of real-world evidence.
|
ApproachREPEAT conducts Meta-Research (“research on research”) to evaluate how current research practices could be improved to increase confidence in the credibility of evidence from longitudinal healthcare databases. REPEAT uses empirical data to guide efforts to improve standards, guidelines and policies. regarding research conduct and reporting.
|
|
to quantify the current state of reproducibility via direct replication of healthcare database studies
Reproducibility (direct replication) in healthcare database studies is achieved when independent investigators are able to recreate analytic cohorts and obtain the same findings by applying the same design and operational decisions to the same large healthcare data source. Hypothesis: A substantial proportion of cohort characteristics and measures of association in published healthcare database studies will not be able to be reproduced with a reasonable margin of error due to lack of transparency reporting on key scientific decisions. to evaluate the robustness of evidence currently found in healthcare database studies
Robustness in healthcare database studies is the variability in study findings when study parameters or assumptions are modified. It is assessed by making alternative plausible design choices (conceptual replication), changing assumptions regarding unmeasured confounding, quantitative and probabilistic bias as well as use of positive and negative control outcomes. Hypothesis 1: In studies that can be closely reproduced, findings will vary in their robustness when alternative plausible design and analysis choices are implemented. Hypothesis 2: In studies that have clear design flaws, the findings and implications of those findings will be very different (even reversed) after those shortcomings are corrected. |
REPEAT
Division of Phamacoepidemiology & Pharmacoeconomics [DoPE] Brigham & Women's Hospital and Harvard Medical School 1620 Tremont Street Suite 3030 | Boston, MA 02120 | [email protected] | 617-278-0930 |