BACKGROUND: Case validation in studies in databases reduces disease misclassification and increases internal validity.
OBJECTIVES: To assess the validity of ICD-9-CM hospital discharge codes to identify cases of upper gastrointestinal complications (UGIC) in Friuli Venezia Giulia (FVG), Italy.
METHODS: Potential cases were identified using site- and lesion-specific and nonspecific ICD-9-CM hospital discharge codes for UGIC. Specific codes were: 531, gastric ulcer; 532, duodenal ulcer; 533, peptic ulcer; and 534, gastrojejunal ulcer. Nonspecific codes were: 578.0, hematemesis; 578.1, melena; and 578.9, GI hemorrhage unspecified.Validation was conducted for: a) a random sample of 108 potential cases with primary codes 531 and 532; b) all potential cases with primary codes 533, 534, and 578.x (n=1947); and c) a random sample of 458 (20%) potential cases with a code in secondary position. Hospital records were abstracted by trained medical personnel and reviewed by at least two independent researchers.
RESULTS: 2513 potential cases were selected for validation. Medical records were obtained for 2473 (98.4%) cases. The positive predictive value (PPV) was 94.3% for codes 531 and 532; 79.5% for code 533; 83.1% for code 534; and 40.2% for nonspecific codes (578.x). PPV was 34.7% for codes in secondary position. For the combination of UGIC codes in secondary position with a primary code for peritonitis (codes 567.2 and 567.8), PPV was 88.9%, and for the combination with a primary code for acute posthemorrhagic anemia (code 285.1), PPV was 79.2%.Based on these results 3,031 final cases of UGIC were identified. Codes 531 and 532 contributed 67.4% of the total number of cases; codes 533 and 534 contributed 4.7%; nonspecific codes contributed 23.0%; and codes for peritonitis and acute posthemorrhagic anemia contributed 4.9%.
CONCLUSIONS: Complete identification of UGIC cases requires the use of specific and nonspecific codes in primary and secondary position. Validation of nonspecific and secondary codes is necessary to avoid disease misclassification. Retrieval of hospital medical records in FVG is almost complete.
