Chikungunya fever is an arboviral disease caused by the Chikungunya virus (CHIKV). In Sub-Saharan Africa, distinguishing CHIKV from other acute febrile illnesses is challenging because presentations overlap with malaria, dengue, typhoid and other endemic infections. Laboratory confirmation (RT-PCR, ELISA, IgM assays) is often limited in low-resource settings. The study aimed to develop an evidence-based clinical case definition for acute CHIKV infection usable at primary health care level during an outbreak context in the Democratic Republic of the Congo (DRC).
This investigation was a cross-sectional study embedded in routine outpatient services at Centre Hospitalier Lukunga, Binza Ozone health zone, Kinshasa. Data and specimens were prospectively collected between June and November 2019 during an ongoing CHIKV outbreak. The study was exploratory; no formal sample size calculation or pre-specified null hypothesis testing was performed.
Patients aged 3 years and older presenting with symptoms suggestive of acute CHIKV infection were enrolled. Clinical examination variables and blood samples were collected. Laboratory confirmation of CHIKV infection used real-time reverse-transcription polymerase chain reaction (RT-PCR) and enzyme-linked immunosorbent assay (ELISA). Malaria rapid diagnostic tests (RDTs) were performed to detect co-infection. Data collection focused on symptoms, including fever, joint pain distribution (for example knee and shoulder pain), rash and other systemic features.
Two principal analytic approaches were used to identify symptom patterns predictive of acute CHIKV infection. First, Classification and Regression Tree (CART) analysis with 10-fold cross-validation was applied to derive a simple decision-tree rule. Second, predictive logistic regression models with a lasso penalty were fitted to select and shrink predictors and assess multivariable performance. Model discrimination was assessed using the Area under the Receiver Operating Characteristics Curve (AUC) with cross-validation.
A total of 132 patients were analysed. Laboratory testing identified acute CHIKV infection in 40.2% of participants. CART analysis selected knee pain and shoulder pain as potential predictors but the resulting tree classified 67.4% of patients correctly while missing 71.7% of confirmed cases. The CART model demonstrated poor discrimination with a cross-validated AUC of 53% (95% CI 43–63%).
Penalized logistic regression produced a model with an overall accuracy of 61.1% and a cross-validated AUC of 62% (95% CI 52–72%). Although better than CART on these metrics, the logistic model did not reach levels of sensitivity and specificity deemed acceptable for clinical case finding at primary care level.
An exploratory analysis excluding participants with malaria co-infections yielded similar CART results, indicating that malaria co-detection did not explain the poor performance of the symptom-based tree. In the subgroup analyses, logistic regression failed to identify reliable predictors when malaria co-infections were removed.
The study provides empirical data on the clinical presentation of CHIKV in an urban Sub-Saharan African outpatient setting during an outbreak. Typical symptoms of CHIKV, including abrupt fever and symmetric joint pain, were observed; specific joint pain locations (knee, shoulder) emerged as candidate clinical signals. However, both simple decision-tree and penalized multivariable approaches failed to produce a symptom-based algorithm with adequate discriminatory performance for routine use at primary health facilities.
Several contextual factors likely contributed to limited predictive value of symptoms: overlap with other febrile illnesses common in the region, variable clinical presentations across cases, and the inherently limited sensitivity of symptom-only algorithms when laboratory confirmation is unavailable. The study highlights the trade-off between simplicity (practical algorithms for primary care) and diagnostic accuracy in heterogeneous outbreak settings.
No symptom-based model developed in this cohort achieved sufficient sensitivity and specificity to serve as a standalone clinical case definition for acute CHIKV infection in this DRC outbreak setting. While particular joint pain patterns had some association with confirmed infection, the overall diagnostic performance was inadequate. The authors emphasize the need for improved access to serological and molecular testing at peripheral levels and recommended integrating local epidemiological context into clinical decision-making. For surveillance and outbreak response, reliance on clinical algorithms alone may lead to under- or overestimation of cases.
As an exploratory, facility-based cross-sectional study, the authors did not perform formal sample size estimation. The analysis included 132 patients, and results reflect that sample and the outbreak context in Kinshasa during 2019. Data sharing is restricted by consent and ethics requirements; the source reports that raw data cannot be publicly released but may be requested through institutional review board processes. Additional studies with larger samples, diverse settings and enhanced laboratory access are needed to refine clinical criteria and improve front-line case detection for CHIKV.