External validation of a clinical prediction score for melioidosis among patients with community-acquired pneumonia: a substudy of the prospective multicentre MISSION cohort.

Abstract

Background: Pulmonary melioidosis may be clinically indistinguishable from other causes of community-acquired pneumonia (CAP), delaying appropriate microbiological investigation and antimicrobial therapy. A four-variable clinical prediction score incorporating monsoon-season presentation, diabetes mellitus, shock at presentation, and extrapulmonary focal abscesses was previously developed to identify melioidosis among patients with CAP. We aimed to externally validate this score in a prospective multicentre cohort and explore whether its performance could be improved by adding readily available predictors.

Methods: This was a nested substudy of the prospective, multicentre, hospital-based MISSION study conducted across 15 tertiary-care centres in India between 1 January 2023 and 31 December 2025. Hospitalised adults with diabetes mellitus or chronic kidney disease who met predefined CAP criteria were eligible. Participants were excluded if one or more score variables could not be reliably ascertained at presentation, defined as clinical information available before microbiological confirmation. The original score was applied without refitting. Discrimination was assessed using the area under the receiver operating characteristic curve (AUC), and sensitivity and specificity were calculated at prespecified cut-offs. An exploratory modified score was developed using independent predictors identified through multivariable logistic regression.

Results: Of 1,418 participants in the parent cohort who met the predefined CAP syndrome definition, 78 were excluded because one or more score variables could not be reliably ascertained at presentation. The final validation cohort included 1,340 participants, of whom 87 (6.5%) had culture-confirmed melioidosis CAP. The original score demonstrated moderate discrimination, with an AUC of 0.718 (95% CI 0.654-0.781). At a cut-off of ≥4, sensitivity was 72.4%, and specificity was 56.4%; at ≥5, sensitivity was 48.3%, and specificity was 86.4%. Extrapulmonary focal abscesses (adjusted odds ratio [aOR] 20.56, 95% CI 10.82-39.05), diabetes mellitus (aOR 2.94, 95% CI 1.36-6.37), male sex (aOR 2.39, 95% CI 1.32-4.33), high-risk occupation (aOR 2.37, 95% CI 1.43-3.93), and shock at presentation (aOR 2.30, 95% CI 1.31-4.05) were independently associated with melioidosis CAP. An exploratory modified score incorporating these variables had an apparent AUC of 0.770 (95% CI 0.711 to 0.828). At a cut-off of ≥3, sensitivity was 63.2% and specificity was 81.4%.

Conclusions: The original four-variable score demonstrated moderate discrimination when externally validated among high-risk hospitalised adults with CAP. A cut-off of ≥4 may support screening, whereas ≥5 provides greater specificity but misses more than half of melioidosis cases. The exploratory modified score showed modestly higher apparent discrimination, and requires independent external validation before clinical use.