Background The BTS CAP audit is the largest national audit of adult CAP management. It relies on acute trusts entering data collected retrospectively from patients’ notes. Cases are selected for entry if they were admitted between 1^st Dec 2012 and 31^st Jan 2013 with a diagnosis of CAP confirmed by appropriate radiological changes. Participating institutions are required to submit a minimum of 20 cases. The audit also provides data on mortality. The national data shows a high mortality rate (18.2%) for CAP. However, the mortality data from our own institution was unexpectedly higher at 28.2%.

Aim To determine if the high mortality rate from our institution is the result of selection bias.

Methods We compared the outcomes of 39 audit cases entered from our institution, a large teaching hospital serving a semi-rural population, with all other cases of CAP admitted over the same period not entered into the audit. Proportions were compared using chi-square tests and continuous variables using Kruskal-Wallis test.

Results During the two month audit period, 124 cases of CAP were identified of whom 39 (31.5%) were entered into the audit. There was no significant difference in age between those entered (77.1 yrs ± SD 11.0) and those not entered (70.1 yrs ± 18.5). However the inpatient mortality rate was significantly higher in those entered into the audit than those who were not (28.2% Vs 10.6%, p = 0.01).

Conclusions These results show that selection bias accounts for the apparently high mortality rates. Although the notes of all patients admitted with CAP were requested for the audit, on reviewing the methods used by the audit department, it is apparent that patients whose notes are most readily available are collected first for audit. Deceased patients’ notes are more easily accessed by the audit team; since the national audit requires only a proportion of patients to be entered, this group are over-represented. If other institutions have similar practices, the national audit will over-estimate mortality from CAP. Case acquisition bias could be reduced by collecting cases prospectively, or by entering all cases of CAP over a shorter predefined time period.