Background: In studies of the association between antibiotic-resistant infection and mortality, the importance of controlling for the underlying severity of illness is well recognized. However, it is unclear when the severity of illness should be assessed. Controlling for severity of illness on the day the culture specimen is obtained may underestimate the true association between resistance and mortality.
Objective: To assess the impact of calculating the Acute Physiology and Chronic Health Evaluation (APACHE) II score at different time points on the association between antimicrobial resistance and mortality.
Methods: We used an existing data set from a study that investigated the association between fluoroquinolone resistance and mortality. The APACHE II score was calculated at 3 time points: the day the culture specimen was obtained, 1 day before the culture specimen was obtained, and 2 days before the culture specimen was obtained. Separate multivariable models were constructed using the 3 different APACHE II scores. These models were compared qualitatively.
Results: Of 91 total subjects, 51 were infected with a fluoroquinolone-resistant strain and 40 with a fluoroquinolone-susceptible strain. The median APACHE II score for all subjects was 13 (95% confidence interval [CI], 11-15) when calculated on the day the culture specimen was obtained, 12 (95% CI, 11-13) when calculated 1 day before, and 11 (95% CI, 10-13) when calculated 2 days before the culture specimen was obtained. Of 91 subjects, 12 (13.2%) died. The 3 multivariable models (each with the APACHE II score calculated on a different day) were not substantively different; the adjusted odds ratio for the association between fluoroquinolone-resistant infection and mortality varied only from 1.38 to 1.65 in the 3 models.
Conclusions: APACHE II scores calculated at different time points relative to obtainment of the culture specimen did not differ substantively. Furthermore, when the adjusted association between fluoroquinolone resistance and mortality was assessed, there were no substantive differences across multivariable models that incorporated APACHE II scores calculated at different time points.