PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
April 5, 2026Cancer Research0 citations

Parsimonious Electronic Health Record Model for Pancreatic Cancer Risk Stratification

Abstract 1378: Development and validation of a parsimonious electronic health record model for pancreatic cancer risk stratification.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Can a parsimonious EHR-based Cox model accurately predict incident pancreatic ductal adenocarcinoma in adults?

Population

Adults ages ≥40 years from U.S. EHR and claims database and UK Biobank. Training cohort N=4,836,428…

Design

Cohort

Follow-up

3 years

Key result

A parsimonious 19-predictor electronic health record model predicted 3-year incident pancreatic cancer with an AUC of 0.75 and a hazard ratio of 7.63 for the highest risk percentile.

Authors

LMLucas A. MavromatisVZViktor ZlatanicEAEmil Agarunov

Discussion

Loading...

Member takes

Overview

May facilitate early PDAC detection via EHR risk stratification; extends prior models with cross-national generalizability.

Key Points

  • To develop and validate a parsimonious risk stratification model for pancreatic cancer using electronic health records.
  • Utilized Optum Labs DataWarehouse for national EHR data.
  • Developed a Cox model predicting incident pancreatic cancer in adults ≥40 years.
  • Employed elastic net with 10-fold cross-validation to select risk predictors from EHR data.
  • Assessed model performance with a 3-year AUC and calibration metrics across diverse health systems.
  • 14,405 patients developed pancreatic cancer in the training cohort (mean age 60.4), with an incidence rate of 56 per 100,000 person-years.
  • The model included 19 key predictors, such as chronic pancreatitis and type 2 diabetes.
  • 3-year AUC was 0.75, demonstrating strong discrimination in both training and validation cohorts.
  • In the UK Biobank, the model maintained a decent AUC of 0.71, indicating good generalizability.

Structured PICO

Can a parsimonious EHR-based Cox model accurately predict incident pancreatic ductal adenocarcinoma in adults?

P
Population
Adults ages ≥40 years from U.S. EHR and claims database (Optum Labs DataWarehouse) and UK Biobank. Training cohort N=4,836,428 (mean age 60.4); validation cohort N=5,607,398 (mean age 60.2); UK Biobank validation N=498,754.
I
Intervention
Parsimonious EHR-based Cox prediction model using 19 predictors (including chronic pancreatitis, prior cancers, type 2 diabetes, elevated AST, current smoking, male sex) for pancreatic cancer risk stratification.
O
Outcome
Incident pancreatic ductal adenocarcinoma (PDAC) prediction performance (assessed by 3-year AUC and calibration).

A parsimonious EHR-based risk model demonstrates strong discrimination and generalizability for predicting 3-year incident pancreatic cancer risk across U.S. and UK cohorts.

Cite This Study

Mavromatis et al. (2026) studied this question. A parsimonious 19-predictor electronic health record model predicted 3-year incident pancreatic cancer with an AUC of 0.75 and a hazard ratio of 7.63 for the highest risk percentile.

synapsesocial.com/papers/69d1fd9ca79560c99a0a3c64https://doi.org/10.1158/1538-7445.am2026-1378
View Full Paper
Ask AI
Bookmark
Share