Abstract
BACKGROUND: Delayed diagnosis of bloodstream infection (BSI) occurs in > 20% of older patients, with misdiagnosis in 35%. Our objective was to develop and validate a clinically useful screening tool to identify older patients with a high probability of having a BSI. METHODS: Hospitalized patients > 80 years old with BSI (n = 105/group) were evaluated for the tool development in this retrospective matched case-controlled study (learn cohort). The tool was validated in different retrospectively matched case and control patients > 80 years old (n = 120/group) and 65 to 79 years old (n = 250/group) (test cohort). Binary logistic regression was used to develop a screening tool using laboratory and clinical parameters that were significantly associated with BSI (P < 0.05; adjusted odds ratio (OR) > 1); and Classification and Regression Tree (CART) analysis was used to identify parameter breakpoints. Performance metrics were used to evaluate and validate the tool. RESULTS: The significant parameters associated with BSI were maximum temperature (Tmax)(> 37.55C)(OR = 42.575), neutrophils (> 7.95)(OR = 1.923), a change in level of consciousness (LOC) (Yes = 1, No = 0)(OR = 1.571), blood urea nitrogen (BUN)(> 10.05)(OR = 1.359), glucose (> 7.35)(OR = 1.167), albumin (< 33.5)(OR = 1.038) and alanine aminotransferase (ALT) (> 19.5)(OR = 1.005). The optimal screening tool [Ln (odds of BSI) = - 150.299 + 3.751(Tmax) + 0.654(neutrophils) + 0.452(change in LOC) + 0.307(BUN) + 0.154(glucose) + 0.038(albumin) + 0.005(ALT)] had favorable performance metrics in the learn and test cohorts (sensitivity, specificity and accuracy of 95% in the learn cohort and 77, 89, and 81% in the total test cohort); and performed better than using only temperature and neutrophil count. CONCLUSIONS: The validated tool had high predictive value which may improve early identification and management of BSI in older patients.