Real‑world evidence is getting smarter

Glynn Dennis argued on April 2, 2026 that the quiet shift is not just faster query writing but rebuilding how patient groups are defined: cohort construction is now being treated as a sequence of explicit, validated steps rather than a single opaque operation, and he used a Type 2 diabetes example — patients on metformin with two high body‑mass readings — to show how a simple clinical idea can map to hundreds of discrete codes. (medcitynews.com) Commercial life‑sciences vendors and analytics teams are rolling these methods into products that automate patient selection and preserve reproducibility; examples include Aetion’s evidence software, IQVIA’s AI rapid‑cohort tools, Deep 6 AI’s patient‑matching platform, Dataize’s clinical‑data abstraction product, and Manifold’s cohort explorer — a “cohort” here meaning a defined group of patients chosen for analysis. (aetion.com) (iqvia.com) (genomeweb.com) (dataize.io) (manifold.ai) The technical pressure point these tools target is the way clinical information is encoded: diagnoses, procedures, drugs and labs live in standardized code sets (for example, ICD‑10 for diagnosis codes, CPT/HCPCS for procedures and billing, NDC for drug identifiers, and LOINC for lab tests), and those code sets were designed for billing or documentation, not precise research queries — that makes mapping a plain‑language clinical definition into reproducible code lists error‑prone. (medcitynews.com) AI does three concrete things to fix that: it performs semantic mapping — translating clinical concepts into the right set of codes across vocabularies; it enforces temporal logic — applying time windows and sequencing rules so, for example, prior medication exposure is checked before an outcome; and it extracts facts from unstructured physician notes using natural‑language processing (NLP), which turns free‑text chart entries into structured fields that can be queried reproducibly. (aws.amazon.com) (dataize.io) Those capabilities are already commercial: IQVIA advertises “AI‑powered Rapid Cohorts” to speed feasibility and patient counts, Aetion moved its evidence platform into the AWS marketplace to scale adoption (March 31, 2025), Deep 6’s matching tech was acquired and folded into Tempus’ stack, and Dataize says its abstraction tool can cut manual chart‑review time substantially — the RWE market for AI‑enabled tools is also being tracked by analysts, who project double‑digit growth over the coming decade. (iqvia.com) (prnewswire.com) (genomeweb.com) (dataize.io) (insightaceanalytic.com)