Abstract
Background: National clinical databases based on coding data are commonly used to compare pneumonia related outcomes across institutions in the UK. However, variation in the accuracy of coding exists, and the effect of this on outcomes is unknown.
Methods: Institutions participating in the national British Thoracic Society community acquired pneumonia (CAP) audit identified patients that were assigned an International Classification of Diseases (ICD) diagnostic code of pneumonia (J12-J18), admitted between 1st December 2014 and 31st January 2015. Individuals with and without a clinico-radiographic diagnosis of CAP, excluding those with nosocomial infection and who were immunocompromised, were compared using multivariable logistic regression.
Results: Of 105 institutions, the median proportion of cases with a clinico-radiographic diagnosis of CAP out of all coded cases of pneumonia was 61.7% (Interquartile Range, 50.7-80.9). Although older and with more comorbid disease, patients comprising the non-pneumonic cohort had a lower 30-day inpatient mortality than those with clinico-radiographic CAP (15.1% vs 17.0%). Following adjustment in the multivariable model, these former individuals had a 35% lower odds of inpatient death at 30 days (adjusted Odds Ratio 0.65, 95%CI 0.53-0.80, p<0.001).
Conclusion: There was considerable variation between institutions regarding the accuracy of coded cases of CAP. Given the lower mortality in the non-pneumonic cohort, this discrepancy in coding accuracy across institutions in the UK is likely to have an impact on hospital standardised mortality ratios (HSMRs) as trusts with a higher proportion of miscoded cases have a lower pneumonia attributable mortality.
- Copyright ©the authors 2016
