Objective: To pilot-test a simple checklist designed to improve coding of acute myocardial infarction (AMI) in hospital discharge abstracts.
Background: Health records technologists review hospital charts to code discharge diagnoses according to the International Classification of Diseases, 9th revision (ICD-9). Many studies have suggested that there is a high false positive rate in coding AMI, ie, ICD-9 410, on hospital discharge abstracts.
Patients and methods: The checklist required either at least two of suggestive symptoms, diagnostic electrocardiographic changes, or diagnostic rise in serum cardiac enzymes; or confirmation by autopsy. First case of use was confirmed-typical time to complete the checklist was 3 to 4 mins. Then 16 Ontario community hospitals were recruited to apply the checklist on a blinded basis to 1000 randomly drawn in-patient records-10% were audited for another study to confirm AMI; and 90% were originally coded with 'most responsible diagnosis' (MRD) of AMI, other cardiovascular diagnoses and various noncardiac conditions. Percentage agreement (95% CI) between the checklist and the confirmed or coded diagnosis was analyzed; coding of AMI as a secondary diagnosis was examined in further analyses.
Results: One hospital withdrew for logistical reasons; the final useable sample from 15 hospitals was 943 records. The checklist correctly identified 100% of AMIs independently confirmed for another study; usual coding identified 89.7% of cases (70 of 78; 95% CI 80.8 to 95.5). For cases not confirmed, but where the physician had nonetheless diagnosed AMI, six of 11 charts were miscoded as AMI in hospital records; none were miscoded by the checklist. For records with AMI as MRD, 11.6% (44 of 380; 95% CI 8.5 to 15.2) were classified as false positives by the checklist. Where an AMI was coded as a secondary diagnosis, 52.9% (36 of 68; 95% CI 40.5 to 65.2) met the checklist criteria for AMI. Finally, among records where the MRD was other than AMI, 6.8% (38 of 563; 95% CI 4.8 to 9.2) met checklist criteria for AMI during admission, but 94.7% had an ICD-9410 code as a secondary diagnosis.
Conclusion: A simple checklist can be very easily applied, has extremely high sensitivity for confirming the presence of AMI, and identifies a clinically significant proportion of charts with false positive codes for AMI. Conversely, these findings support the high sensitivity (low false negative rates) of conventional coding practices for AMI in Canadian hospital records, be it as a primary or secondary diagnosis (eg, 95% detection rate). Usual coding, combined with the checklist for tentative ICD-9 410 diagnoses, would improve the accuracy of Canadian hospital records.