Juebin Gu, Yufei Dong, Mulin Liu, Lin Lin
Routinely available clinical indicators provided only modest discriminatory ability for early identification of blood culture-positive BSI. Bloodstream pathogen ddPCR testing provided additional diagnostic information with a modest and statistically non-significant increase in AUC but significant improvement in risk reclassification. ddPCR may serve as a complementary tool for early BSI risk stratification rather than a replacement for blood culture.
BACKGROUND: Bloodstream infection (BSI) is associated with substantial morbidity and mortality. Early identification remains challenging because blood culture results are often unavailable during initial clinical decision-making. This study evaluated routinely available clinical indicators for identifying blood culture-positive BSI and investigated whether bloodstream pathogen droplet digital PCR (ddPCR) testing provides incremental diagnostic value.
METHODS: This retrospective study was conducted at a tertiary teaching hospital. A conventional clinical indicator cohort of 814 hospitalized patients was used to develop a prediction model for blood culture-positive BSI. An independent ddPCR cohort of 336 patients was used to evaluate the incremental value of ddPCR testing. Model discrimination, reclassification, exploratory clinical utility, and internal validation were assessed.
RESULTS: In the conventional cohort, age (OR 1.021, 95% CI 1.001-1.041, P = 0.038) and ln(PCT) (OR 1.441, 95% CI 1.203-1.727, P < 0.001) were independently associated with blood culture-positive BSI. The conventional clinical model achieved an area under the curve (AUC) of 0.696 (95% CI 0.631-0.762). In the ddPCR cohort, incorporation of ddPCR results modestly increased the AUC from 0.716 (95% CI 0.614-0.818) to 0.764 (95% CI 0.665-0.863). Although the increase was not statistically significant by the DeLong test (P = 0.152), significant improvements were observed in integrated discrimination improvement (0.041, P = 0.006) and category-free net reclassification improvement (0.786, P < 0.001). Exploratory decision curve analysis indicated higher net benefit for the combined model.
CONCLUSIONS: Routinely available clinical indicators provided only modest discriminatory ability for early identification of blood culture-positive BSI. Bloodstream pathogen ddPCR testing provided additional diagnostic information with a modest and statistically non-significant increase in AUC but significant improvement in risk reclassification. ddPCR may serve as a complementary tool for early BSI risk stratification rather than a replacement for blood culture.