Investigators reviewed medical records from four hospitals in Brazzaville for adult admissions meeting WHO criteria for severe malaria between January 2020 and December 2024. The initial pool included 523 hospitalized patients; 296 met the study’s inclusion criteria and were analyzed. The cohort skewed young (median age 25.5 years, IQR 20–40) and low income: 82.1% were classified as having a low socioeconomic status. Women constituted 51.7% of the sample. Median hemoglobin measured 10.55 g/dL. These baseline characteristics frame the clinical context in which the authors probed determinants of hospital stay length.
Short, direct data points appear here to orient the reader. The study source did not report additional demographic breakdowns such as comorbidities, HIV status, or pregnancy-specific subgroups; those details were not available in the preprint.
The study team dichotomized hospital length of stay (LOS) into short (<5 days) versus prolonged (≥5 days). Median LOS in the cohort was 4 days (IQR 3–5). That single median captures typical care duration in this setting, while the dichotomous threshold defined the outcome used in regression models. The authors used multivariable logistic regression with a backward stepwise procedure to identify independent predictors of prolonged LOS. The preprint does not report alternative LOS cutoffs or time-to-event analyses; no survival curves or hazard ratios appear in the source document.
Data indicated several independent clinical predictors of a prolonged hospital stay. Delay to consultation increased the odds of prolonged LOS: adjusted odds ratio (aOR) 1.25 per day of delay (95% CI 1.09–1.44). Isolated anemia—that is, anemia listed among severity criteria without other concurrent severity features—was associated with higher odds of prolonged LOS (aOR 7.07, 95% CI 1.23–40.55). Neurologic manifestations also tracked with longer hospitalizations: convulsions carried an aOR of 5.86 (95% CI 1.81–18.97), prostration had an aOR of 4.42 (95% CI 1.21–16.2), and impaired consciousness showed an aOR of 4.29 (95% CI 1.17–15.7). The regression model used isolated hemoglobinuria as the reference severity category, per the methods described.
Short sentences present the core results. Longer sentences explain model framing: investigators adjusted for multiple clinical severity features and constructed the models using a backward stepwise approach, but the preprint did not list every candidate variable entered into the initial model nor provide model fit statistics or post-estimation diagnostics.
The largest effect sizes arose when anemia occurred alongside other severity criteria. The study team reported combined-category associations: anemia with prostration had an aOR of 8.58; anemia combined with jaundice or hemoglobinuria yielded an aOR of 16.21; and impaired consciousness paired with anemia produced the highest reported aOR, 33.31. These combined measures suggest multiplicative or at least additive risk for prolonged hospitalization when hematologic compromise coexists with neurologic or hemolytic manifestations. The source does not present stratified survival curves, nor does it report absolute risk increases associated with these combinations—only adjusted odds ratios are provided.
The authors argued that anemia is a potentially modifiable determinant of prolonged hospitalization for adults with severe malaria. They recommended three operational priorities: earlier screening for anemia on presentation, improved access to safe blood transfusion services, and interventions to reduce delays to care. Each recommendation is linked directly to observed associations—delay to consultation independently predicted prolonged LOS, and anemia (alone or combined) markedly amplified the odds of extended admission. The preprint did not, however, provide specific screening protocols, transfusion thresholds, or implementation strategies; it did not report on transfusion rates or associated outcomes in the cohort.
The article is a medRxiv preprint and has not undergone peer review; readers should treat the results as preliminary. The study was cross-sectional and based on chart review, so causal inference is limited and the timing or sequence of interventions relative to outcomes cannot be firmly established. The authors noted that formal IRB approval was not required because records review preserved anonymity, per the Research Ethics Committee for Health Sciences (CERSSA); they also stated that the deidentified dataset is available from the corresponding author on request.
Several important questions remain unanswered in the source. The preprint did not report transfusion practices, transfusion-associated adverse events, or whether correction of anemia shortened LOS in individual patients. The content did not include detailed comorbidity data, stratified analyses by age or sex beyond basic demographics, or cost analyses quantifying the hospital-resource impact of prolonged stays. No prospective or interventional data were presented to show that the recommended operational changes would reduce LOS. Those gaps define clear next steps for research and implementation studies in this setting.