Background: Respiratory rate (RR) is an independent predictor of in-patient mortality, intensive care unit admission and cardiac arrest across a variety of conditions. It is also an integral component of many risk prediction calculators (such as the pneumonia severity index). Thus, accurate assessment of RR is necessary to recognize disease severity and prognosis among hospitalized adults. Several small, single-center studies largely conducted in the emergency department have shown that the recorded RR is inaccurate compared to validated, objective measurements. However, it is unclear whether this is also true among hospitalized adults. We sought to assess patterns of variation in RR in hospitalized medical inpatients as a proxy for accuracy since, if recorded accurately, it would have a normal distribution like other physiologic parameters such as heart rate (HR).

Methods: We analyzed data from a retrospective observational cohort of consecutive medical hospitalizations among adults from November 1, 2009 to October 31, 2010 from 6 hospitals – a mix of teaching, non-teaching, community and county hospitals. The minimum and maximum RR and HR for each day of hospitalization were extracted from the EHR. We assessed patterns of RR using frequency plots, descriptive statistics and the coefficient of variation (100*standard deviation/mean) on the day of admission, subsequent days of hospitalization and for clinical subgroups expected to have substantial variation in RR.

Results: We included 36,966 hospitalizations among 28,511 unique patients, representing 220,665 unique hospital days.  Mean age was 61.7 years old, 54% were female, and 40% were non-white and the median length of stay was 4 days. The distribution of the recorded RR was not normally distributed as compared to HR, a physiologic variable that is objectively measured (Figure). RR was clustered at 18 and 20 with little variation. The minimal RR equaled the maximal value in 26% of patient days. The maximum RR equaled 18 or 20 in 75% of hospital days. There was less variation in the minimum RR compared to the maximum value on admission, as well as over the duration of the hospitalization (CV of 0.2 for the first day, 0.15 for two days prior to discharge, and 0.12 for day of discharge; Table). Even among those hospitalized for cardiopulmonary diseases or who had O2 sat<92% who would be expected to have abnormal breathing, there was very little variation in RRs compared to other patients. There was no difference in variation between age and sex groups.

Conclusions: Among a large, diverse multicenter cohort of adults hospitalized for a broad range of medical conditions, we found little variation in the recorded RR on admission, throughout hospitalization and among those with cardiopulmonary compromise. The clustering of values suggests that the recorded RR is likely an estimate, with 18 and 20 used in place for ‘normal’. Inaccurate recording of the RR may lead to misclassification of disease severity and prognosis, and potentially jeopardize patient safety.