Related Experiment Videos

Validation of diagnoses of peptic ulcers and bleeding from administrative databases: a multi-health maintenance

Susan E Andrade1, Jerry H Gurwitz, K Arnold Chan

  • 1Meyers Primary Care Institute-Fallon Healthcare System and University of Massachusetts Medical School, Worcester, MA 01605, USA.

Insights

This study evaluated International Classification of Diseases, 9th revision (ICD-9) codes for peptic ulcers and gastrointestinal bleeding. Many codes showed low positive predictive values (PPVs), indicating potential inaccuracies in automated health databases.

Area of Science:

  • Medical Informatics
  • Gastroenterology
  • Health Services Research

Background:

  • Automated health plan databases are increasingly used for research.
  • Accurate diagnostic coding is crucial for reliable data analysis.
  • The validity of International Classification of Diseases, 9th revision (ICD-9) codes for gastrointestinal conditions needs evaluation.

Purpose of the Study:

  • To assess the positive predictive values (PPVs) of ICD-9 codes for peptic ulcers and upper gastrointestinal bleeding.
  • To determine the accuracy of diagnoses recorded in automated health plan databases.

Main Methods:

  • Utilized data from automated health plan records and medical chart abstractions from eight large health maintenance organizations (HMOs).
  • Evaluated PPVs for specific ICD-9 codes related to duodenal ulcer, gastric/gastrojejunal ulcer, peptic ulcer, and gastrointestinal hemorrhage.
  • Confirmed diagnoses through surgery, endoscopy, X-ray, or autopsy.

Main Results:

  • Overall, only 23% (207 of 884) of cases of peptic ulcers and upper gastrointestinal bleeding were confirmed.
  • PPVs varied significantly by diagnosis: duodenal ulcer (66%), gastric/gastrojejunal ulcer (61%), peptic ulcer (1%), and gastrointestinal hemorrhage (9%).
  • PPVs were consistent across different HMOs and geographical regions.

Conclusions:

  • The accuracy of ICD-9 codes for peptic ulcers and upper gastrointestinal bleeding in automated health plan databases is often low.
  • There is a critical need to validate diagnostic codes from automated health databases.
  • Findings highlight the importance of rigorous data accuracy assessment in health services research.

Related Concept Videos