Body Temp and Bacterial Infection in Nursing Home Residents
Body Temp and Bacterial Infection in Nursing Home Residents
From March through May 2011, 12 North Carolina NHs were enrolled in a study of infections. During the 3 months after enrollment, all episodes in which a resident of a participating NH received treatment with a systemic antibiotic for a presumed new NH-acquired illness were identified.
To learn more about the role of body temperature in decisions to prescribe, trained research nurses audited NH records, gathering data on temperature and other resident characteristics (depending on the volume of cases in a month) of all cases or a random sample of 18 prescribing episodes meeting the study's inclusion criteria. Episodes were excluded if the prescription was prophylactic (e.g., before a dental procedure or to prevent recurrent UTIs) or had been prescribed during a hospitalization. Among the data collected were the highest temperature recorded on the day the antibiotic was prescribed and the last three temperatures recorded at times when the resident had not been ill (routinely obtained, nonillness temperatures). Route of temperature was rarely recorded but, according to staff in all participating NHs, was routinely oral.
Data were available on 1,959 antibiotic-prescribing episodes for a presumed infection, which occurred in 1,039 residents in the study NHs. Of these, three episodes were excluded because the listed illness temperature was in a range considered likely to be erroneous (<95ºF or >106ºF), 21 episodes were excluded because no nonillness temperature was listed, and 77 episodes were excluded because no illness temperature was available, yielding an analytical sample of 1,858 prescribing episodes in 1,007 NH residents.
For each resident in the final analytical sample, mean nonillness temperature was calculated for only the first prescribing episode (to avoid potential sample nonindependence), yielding 1,007 resident mean normal temperatures. Descriptive statistics were then calculated using SAS procedures PROC UNIVARIATE and PROC FREQ (SAS Institute, Inc., Cary, NC). As part of that analysis, two definitions were used to determine the prevalence of fever in the study sample: a population-based definition of fever, defined as 2 standard deviations (SDs) above the mean nonillness temperature for the 1,007 NH residents in the sample, and an individualized definition of fever, defined as 2 SDs (using the SD for the population mean) above the average of the three most recent nonillness temperatures for a given resident.
Next, how often episodes of presumed infection for which antibiotics were prescribed were associated with a temperature elevation that met these empirically based definitions of fever was examined. The proportion of illness-day temperatures associated with fever according to these empirically based definitions was also compared with the proportion classified as febrile according to a previously developed definition of fever and the Infectious Diseases Society of America (IDSA) definition of fever. For the IDSA definition, it was possible to evaluate only two of the three criteria, because the data collection included only the highest temperature and not multiple temperatures on the date antibiotics were prescribed. Because each episode of presumed infection may be treated as an independently occurring event, analyses of illness-day temperature data include all 1,858 prescribing episodes.
Methods
From March through May 2011, 12 North Carolina NHs were enrolled in a study of infections. During the 3 months after enrollment, all episodes in which a resident of a participating NH received treatment with a systemic antibiotic for a presumed new NH-acquired illness were identified.
To learn more about the role of body temperature in decisions to prescribe, trained research nurses audited NH records, gathering data on temperature and other resident characteristics (depending on the volume of cases in a month) of all cases or a random sample of 18 prescribing episodes meeting the study's inclusion criteria. Episodes were excluded if the prescription was prophylactic (e.g., before a dental procedure or to prevent recurrent UTIs) or had been prescribed during a hospitalization. Among the data collected were the highest temperature recorded on the day the antibiotic was prescribed and the last three temperatures recorded at times when the resident had not been ill (routinely obtained, nonillness temperatures). Route of temperature was rarely recorded but, according to staff in all participating NHs, was routinely oral.
Data were available on 1,959 antibiotic-prescribing episodes for a presumed infection, which occurred in 1,039 residents in the study NHs. Of these, three episodes were excluded because the listed illness temperature was in a range considered likely to be erroneous (<95ºF or >106ºF), 21 episodes were excluded because no nonillness temperature was listed, and 77 episodes were excluded because no illness temperature was available, yielding an analytical sample of 1,858 prescribing episodes in 1,007 NH residents.
For each resident in the final analytical sample, mean nonillness temperature was calculated for only the first prescribing episode (to avoid potential sample nonindependence), yielding 1,007 resident mean normal temperatures. Descriptive statistics were then calculated using SAS procedures PROC UNIVARIATE and PROC FREQ (SAS Institute, Inc., Cary, NC). As part of that analysis, two definitions were used to determine the prevalence of fever in the study sample: a population-based definition of fever, defined as 2 standard deviations (SDs) above the mean nonillness temperature for the 1,007 NH residents in the sample, and an individualized definition of fever, defined as 2 SDs (using the SD for the population mean) above the average of the three most recent nonillness temperatures for a given resident.
Next, how often episodes of presumed infection for which antibiotics were prescribed were associated with a temperature elevation that met these empirically based definitions of fever was examined. The proportion of illness-day temperatures associated with fever according to these empirically based definitions was also compared with the proportion classified as febrile according to a previously developed definition of fever and the Infectious Diseases Society of America (IDSA) definition of fever. For the IDSA definition, it was possible to evaluate only two of the three criteria, because the data collection included only the highest temperature and not multiple temperatures on the date antibiotics were prescribed. Because each episode of presumed infection may be treated as an independently occurring event, analyses of illness-day temperature data include all 1,858 prescribing episodes.
Source...