Prognostic performance of CURB-65 and procalcitonin for mortality in community-acquired pneumonia in a Peruvian hospital: a retrospective cohort study

Pneumonia remains a leading global public health challenge as one of the main causes of death from infectious diseases. CURB-65 and procalcitonin (PCT) are established tools for assessing hospitalisation needs and severity in community-acquired pneumonia (CAP), respectively. To evaluate the prognostic performance of combining CURB-65 and PCT (mCURB65) for predicting 30-day in-hospital mortality and intensive care unit (ICU) admission in patients hospitalised for CAP in a Peruvian hospital. This retrospective cohort study analysed medical records of patients aged ≥ 18 years hospitalised for CAP in Lima, Peru, between 2020 and 2022. mCURB65 was compared with CURB-65 and PCT using receiver operating characteristic (ROC) area under the curve (AUC) analysis. The incremental value of mCURB65 over CURB-65 was quantified as the change in AUC (ΔAUC) with 95% confidence intervals (CI) and p values derived using DeLong’s method. Calibration, decision curve analysis, and sensitivity analyses for the competing risk of death before ICU admission were also performed. Of 314 patients, 49 (15.6%) died within 30 days and 35 (11.1%) were admitted to the ICU. For 30-day mortality, the AUC was 0.74 (95% CI: 0.66–0.81) for CURB-65 and 0.80 (0.73–0.87) for mCURB65 (ΔAUC 0.066, 95% CI: 0.028 to 0.104; p < 0.001). For ICU admission, AUCs were 0.73 (0.64–0.82) and 0.76 (0.67–0.85) (ΔAUC 0.035, 95% CI: −0.007 to 0.078; p = 0.099). For the composite outcome of ICU admission or 30-day mortality, AUCs were 0.74 (0.68–0.81) and 0.80 (0.74–0.87) (ΔAUC 0.059, 95% CI: 0.025 to 0.092; p < 0.001). PCT cut-offs were 0.185 ng/mL for mortality and 0.165 ng/mL for ICU admission and the composite outcome. Decision curve analysis showed a higher net benefit for mCURB65 across threshold probabilities of approximately 8–39% for mortality and 10–49% for the composite outcome. Adding PCT to CURB-65 significantly improved discrimination for 30-day mortality and for the composite outcome in patients hospitalised for CAP, and yielded a higher net benefit across clinically plausible decision thresholds. The improvement for ICU admission alone was not statistically significant. External validation is required before clinical implementation.