Objectives: To provide empirically based recommendations for incorporating body temperature into clinical decision-making regarding diagnosing infection in nursing home (NH) residents.
Design: Retrospective.
Setting: Twelve North Carolina NHs.
Participants: NH residents (N = 1,007) with 1,858 randomly selected antibiotic prescribing episodes.
Measurements: Maximum prescription-day temperature plus the three most recent nonillness temperatures were recorded for each prescribing episode. Two empirically based definitions of fever were developed: population-based (population mean nonillness temperature plus 2 population standard deviations (SDs)) and individualized (individual mean nonillness temperature plus 2 population SDs). These definitions were used along with previously published fever criteria and Infectious Diseases Society of America (IDSA) criteria to determine how often each prescribing episode was associated with a "fever" according to each definition.
Results: Mean population nonillness temperature was 97.7 ± 0.5 ºF. If "normal" were defined as less than 2 SDs above the mean, fever would be defined as any temperature above 98.7 ºF, and the previously published fever cutpoints and the IDSA criteria are 4.8 SDs above this mean. Between 30% and 32% of the 1,858 prescribing episodes examined were associated with temperatures more than 2 SDs above the population mean nonillness temperature, whereas only 10% to 11% of episodes met the previously published and IDSA fever definitions.
Conclusion: Clinicians should apply empirically based definitions to assess fever in NH residents. Furthermore, low fever prevalence in residents treated with antibiotics according to all definitions suggests that some prescribing may not be associated with acute bacterial infection.