Methods: We used PCR positive cases notified to public health as the gold standard (public health diagnosed). We extracted records with ICD-9 codes of 033, 033.0, 033.1, 033.8, and 033.9 from physician billing and hospitalization databases (physician diagnosed BP). Public health records were deterministically linked with billing and hospitalization data using ULIs. Numbers and proportions of true positive (TP: public health AND physician diagnosed), false positive (FP: only physician diagnosed) and false negative (FN: only public health diagnosed) cases were estimated. Sensitivity (SE) was calculated as TP/TP+FN; positive predictive value (PPV) as TP/TP+FP. Physician inaccuracy was calculated as FN/TP+FN. Analysis is ongoing as data are received for negative predictive value and specificity.
Results: From public health OR physician diagnosis 7711 cases of BP were identified. Of these, 13% were TP, 27% FN, and 60% were FP. SE of physician diagnosis was 32.5% (95% CI: 30.9% - 34.1%) and PPV was 18.0% (95% CI: 17.3% – 19.4%). Among 5587 physician diagnosed cases, ICD-9 code 033 (whooping cough) was used in TP and ICD-9 code 033.0 (B. pertussis) was used in FP (86% each). Of the 2124 FN cases 71% were physician diagnosed as “acute respiratory infection” (ICD-9 codes 460-466). Physicians incorrectly diagnosed 67% of public health diagnosed BP cases.
Conclusion: Our study demonstrates low sensitivity of physician diagnosis for BP. It also shows that 67% of Alberta physician diagnosed cases of BP are misdiagnoses. Additional lab negative data is essential to estimate negative predictive value and specificity.
S. J. Drews, None
L. Svenson, None
M. L. Russell, None