High-Throughput Observational Evidence Generation Using Linked Electronic Health Record and Claims Data

Background: Many consequential treatment decisions involve off-label or head-to-head choices in complex, comorbid populations routinely excluded from randomized trials, or other prospective analyses. For these decisions, comparative evidence is often entirely absent. Although observational data include these patients, findings are difficult to synthesize because studies differ in cohort definitions, confounder measurement, follow-up periods, and reported outcomes. Prior systems for large scale evidence generation have largely stopped at data preparation, limiting the usefulness of their outputs to decision makers. Methods: We developed a high-throughput evidence-generation workflow using linked EHR and claims data. A prespecified causal-measurement architecture was applied consistently across scenarios, including three post-index follow-up windows through two years; 28 comorbidities; 14 healthcare resource utilization categories; 30 laboratory measures with 57 binary thresholds; 43 adverse-event categories; and evaluated 1038 clinically important outcomes. Scalable collaborative targeted learning (C-TMLE) generated confounding-adjusted, actionable treatment comparisons, with unadjusted estimates reported for transparency. Results: Across 135 clinical scenarios, the workflow generated 210,584,509 outcome evaluations. Each evaluation represented an outcome, follow-up window, treatment contrast, population stratum, and estimator, accompanied by diagnostic information. These results were synthesized into 7,500 narrative summaries and underwent structured clinical and statistical quality control. Conclusions: Standardized, high-throughput workflows can move evidence generation beyond fragmented individual studies toward comprehensive evidence packages. By making treatment-effect heterogeneity visible across clinically meaningful subgroups, this shared evidence base can support precision medicine and reduce redundant stakeholder-specific studies.