Accuracy of ICD-9 codes for acute stroke in hospital discharge records

Background Discharge ICD-9-CM (International Classification of Diseases, 9th Revision, Clinical Modification) codes have been used to identify patients with acute stroke for epidemiological, quality of care, and cost studies. However, outcome research based on data routinely collected for administrative purposes requires a preliminary assessment of the accuracy of codes in order to avoid measurement bias. The aim of this study was to determine the accuracy of the primary ICD-9-CM codes for stroke discharged from the Stroke Unit (SU) and from other wards of Sant’Orsola-Malpighi Hospital of Bologna (Italy). We hypothesized that coding in patients discharged from the Stroke Unit would be more accurate than in other wards. Methods We identified all patients who were discharged in 2007 from the Sant’Orsola-Malpighi Hospital of Bologna with a discharge diagnosis of stroke. We used highly specific ICD 9-CM codes for ischemic stroke (433.x1, 434.x1,436, 99702) and intracerebral hemorrhage (431). We reviewed medical records of a random sample of 274 cohort members (SU, N=117, other wards, N=157). Sample size was determined using power analysis. International criteria for case definition were used to review clinical charts. Results Chart review identified 19 patients (6.9%) who did not meet study criteria for acute ischemic or hemorrhagic stroke. The proportion of patients not meeting diagnostic criteria differed between SU and other wards (2/117, 1.7% vs. 17/157, 10.8%, Fisher exact test, p=0.002). Thus the overall PPV of the primary discharge diagnosis for identifying acute stroke was 92.7%, with a difference of about 9% between the SU and other wards (98.3% vs. 89.2%). Conclusions The PPV of incident strokes both in the SU and in other wards is higher than in other reports in the literature. Although an unknown percentage of strokes are missed, this group of proven acute stroke patients can be reliably used for outcome studies.