BEYOND ICD CODES: USING PATHOLOGY DATA TO IDENTIFY BIOMARKER-DEFINED COHORTS AND CLINICALLY MEANINGFUL SUBGROUPS IN REAL-WORLD DATA
Author(s)
Jelena Pavlovic, MSc, PhD, MD1, Ivana Prokic, MSc, PhD, MD1, Ellie Iob, PhD1, Nikita Jeswani, BA, MBA2, Qurinus Voorham, PhD3.
1PHARMO Institute for Drug Outcomes Research, Utrecht, Netherlands, 2Lumanity, London, United Kingdom, 3Dutch national pathology databank (PALGA), Houten, Netherlands.
1PHARMO Institute for Drug Outcomes Research, Utrecht, Netherlands, 2Lumanity, London, United Kingdom, 3Dutch national pathology databank (PALGA), Houten, Netherlands.
OBJECTIVES: This study explored the feasibility of using Dutch pathology (PALGA) data to reliably select populations, define outcomes, or stratify by disease severity or prognosis using linked real-world data (RWD) from the PHARMO Data Network, a Dutch population-based database which includes longitudinal healthcare records for research.
METHODS: Pragmatic text search terms were assessed in the most recent PALGA data to identify populations poorly captured by using diagnostic codes alone, summarized as report counts. Three disease areas were assessed from 1 January to 30 June 2025: (i) metabolic dysfunction associated steatotic liver disease (MASLD, previously NAFLD)/ metabolic dysfunction associated steatohepatitis (MASH, previously NASH), (ii) primary membranous nephropathy using kidney biomarkers, i.e., phospholipase A2 receptor (PLA2R)/ thrombospondin type-1 domain-containing 7A (THSD7A); and (iii) ulcerative colitis (UC).
RESULTS: For liver pathology, 864 reports containing “NAFLD” or “MASLD” search terms and 3,529 reports containing “NASH” or “MASH” were identified. Fibrosis staging terms (“fibrose stadium”, “F0-F4”, or “stadium 0-4”) appeared in 13,637 reports. For kidney biomarkers, text searching returned 1,174 reports. For UC, a query (“inflammation” AND “colon”) returned 18,000 reports, which could be further optimized to enable pathology-based confirmation of inflammation.
CONCLUSIONS: Optimized PALGA text search algorithms enable a pathology-confirmed diagnosis and severity/prognosis subgroups (e.g., fibrosis stage). Linking PALGA with routinely collected PHARMO data (e.g., disease-specific medication use, GP electronic medical records and hospital administrative data, laboratory results such as estimated glomerular filtration rate, creatinine, albumin, proteinuria for kidney pathology; liver enzymes and lipid profiles for liver pathology; and inflammation markers for UC), provide opportunities for a high-validity cohort identification, outcome definitions and clinically meaningful severity/prognosis stratifications in burden of illness, drug utilization, post-authorization safety and effectiveness studies, and outcomes research including validation.
METHODS: Pragmatic text search terms were assessed in the most recent PALGA data to identify populations poorly captured by using diagnostic codes alone, summarized as report counts. Three disease areas were assessed from 1 January to 30 June 2025: (i) metabolic dysfunction associated steatotic liver disease (MASLD, previously NAFLD)/ metabolic dysfunction associated steatohepatitis (MASH, previously NASH), (ii) primary membranous nephropathy using kidney biomarkers, i.e., phospholipase A2 receptor (PLA2R)/ thrombospondin type-1 domain-containing 7A (THSD7A); and (iii) ulcerative colitis (UC).
RESULTS: For liver pathology, 864 reports containing “NAFLD” or “MASLD” search terms and 3,529 reports containing “NASH” or “MASH” were identified. Fibrosis staging terms (“fibrose stadium”, “F0-F4”, or “stadium 0-4”) appeared in 13,637 reports. For kidney biomarkers, text searching returned 1,174 reports. For UC, a query (“inflammation” AND “colon”) returned 18,000 reports, which could be further optimized to enable pathology-based confirmation of inflammation.
CONCLUSIONS: Optimized PALGA text search algorithms enable a pathology-confirmed diagnosis and severity/prognosis subgroups (e.g., fibrosis stage). Linking PALGA with routinely collected PHARMO data (e.g., disease-specific medication use, GP electronic medical records and hospital administrative data, laboratory results such as estimated glomerular filtration rate, creatinine, albumin, proteinuria for kidney pathology; liver enzymes and lipid profiles for liver pathology; and inflammation markers for UC), provide opportunities for a high-validity cohort identification, outcome definitions and clinically meaningful severity/prognosis stratifications in burden of illness, drug utilization, post-authorization safety and effectiveness studies, and outcomes research including validation.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
EPH62
Topic
Epidemiology & Public Health, Study Approaches
Topic Subcategory
Disease Classification & Coding
Disease
Gastrointestinal Disorders, Urinary/Kidney Disorders