Matt Harmon

52 Chapter three Multivariable logistic regression was used to establish the independent association of hypothermia with 90-day mortality. The Acute Physiology and Chronic Health Evaluation (APACHE) IV score was included in the model to adjust for severity of disease at ICU admission. Age, body mass index (BMI), admission type and source of infection were a priori considered potential clinically relevant confounders. Next, risk factors for hypothermia from logistic regression analyses were investigated as confounders for mortality. Significant variables were retained in the model, based on 10% change-in-estimate. In order to determine whether hypothermia was associated with biomarker response irrespective of severity of disease, hypothermic patients were 1:1 matched to nonhypothermic patients by APACHE IV scores, using ‘optimal matching’ with R-package “MatchIt”. P<0.05 was considered statistically significant. Results Epidemiology of hypothermic sepsis The selection of study patients is presented in Supplemental Figure 1. From a total of 525 patients, 186 (35.4%) patients were hypothermic during the first 24 hours of admission. Patient characteristics are shown in table 1. Mean body temperature in the first 24 hours was significantly lower in hypothermic versus nonhypothermic patients (median 36.3°C and 37.3°C respectively). Mean age in hypothermic patients was significantly higher and BMI was lower. Hypothermic patients suffered more frequently from cardiovascular disease including chronic cardiovascular insufficiency, hypertension and cerebrovascular disease. Hypothermic patients were most often admitted from the emergency department. Also, a significantly higher proportion of patients with hypothermia had a urinary tract infection. We observed no differences in causative organisms (Supplemental Table 1). Hypothermic patients were more seriously ill, as reflected by higher APACHE IV and Sequential Organ Failure Assessment (SOFA) scores and increased incidence of AKI (and requirement of renal replacement therapy). In line with this, patients with hypothermia had higher maximum white blood cell counts, longer prothrombin times and increased creatinine and lactate levels. Risk factors for hypothermic sepsis Multivariable analysis was performed to determine whether patient factors were independently associated with hypothermia. The initial model contained age, BMI, cerebrovascular disease, chronic cardiovascular insufficiency, hypertension, chronic renal insufficiency, site of infection and admission origin (Supplemental