ABSTRACT

1. Introduction

Lassa fever is an acute viral hemorrhagic illness endemic to West Africa, causing recurrent outbreaks with substantial morbidity, mortality, and pressure on fragile health systems.[1] In Nigeria, seasonal dry‑season surges repeatedly strain frontline capacity for triage, isolation, and effective infection prevention and control (IPC) measures.[2] Early recognition is difficult because initial symptoms overlap with malaria and typhoid, delaying testing and isolation and increasing nosocomial risk.[3]

In resource‑limited settings, symptom‑based screening remains necessary despite its limitations. However, diagnostic estimates vary widely due to reliance on surveillance line‑lists, inconsistent case definitions, and analytic methods that ignore matched designs.[4] These gaps hinder the development of context‑specific algorithms for earlier case identification. There is thus a need to quantify the diagnostic yield of common symptoms, identify independent predictors using matched‑aligned models, and translate findings into pragmatic triage heuristics for frontline use.

This study used a mixed‑methods design integrating a matched case–control dataset and gender‑stratified focus group discussions (FGDs). While full qualitative findings are reported elsewhere, key insights are included to clarify how providers and communities interpret early symptoms and triage decisions, enhancing operational relevance. A triangulation approach was used, in which quantitative predictors were interpreted alongside qualitative insights to improve operational relevance and validity of triage recommendations.

This analysis focuses on clinical predictors, comparing diagnostic and predictive performance using matched (conditional) and unmatched (unconditional) models. The aim is to identify parsimonious, operational early predictors to improve triage, accelerate testing, and strengthen IPC in endemic settings.

2. Literature Review / Theoretical Background

This section summarizes evidence most relevant to clinical prediction of Lassa fever at presentation and outlines the theoretical and analytical approach guiding this study. Although it draws from the broader doctoral thesis from which this paper is derived, it is deliberately scoped to clinical variables and early triage in endemic, resource‑limited settings.

Lassa fever is an acute zoonotic viral hemorrhagic illness caused by Lassa virus (LASV; Arenaviridae), with person‑to‑person transmission occurring in households and health facilities.[5] Ondo State is a persistent Nigerian hotspot; during the 2018 Grade 2 outbreak, the state recorded 161 confirmed cases and 42 deaths (Case Fatality Rate = 26.1%), with cases concentrated around Owo township and seasonality peaking in the dry season (December-April).[6] These contextual features repeatedly stress triage and IPC,[7] making early, symptom‑based identification crucial for reducing diagnostic delay and nosocomial risk.

The initial clinical picture of Lassa fever is nonspecific, typically including fever, headache, fatigue, vomiting, and abdominal pain, and overlaps with malaria and other febrile illnesses.[8] Severe disease can progress to hemorrhage and neurological involvement, while sensorineural hearing loss is a recognized sequela among survivors.[9] This overlap undermines early recognition, delays testing and isolation, and creates opportunities for facility-based transmission when IPC is stretched.

Across outbreak investigations, fever is nearly universal in confirmed cases (90–100%), followed by headache (50–70%), fatigue (40–60%), vomiting (30–50%), and abdominal pain (30–45%).[10] No single symptom provides sufficient diagnostic accuracy; features like vomiting or myalgia may be specific but lack sensitivity, reinforcing the need for multi-symptom approaches.[11] Multicenter analyses and machine‑learning models likewise demonstrate that combinations of symptoms, particularly fever paired with a second feature such as vomiting or fatigue, provide better diagnostic discrimination than individual symptoms alone, even though laboratory confirmation remains essential for definitive diagnosis.[12]

Classical case–control studies identified symptom combinations, such as pharyngitis, retrosternal pain, and proteinuria, as improving diagnostic performance. [13] Still, their generalizability is limited in malaria‑co‑endemic settings where overlapping febrile presentations reduce specificity. More recent work favors multi‑symptom and model‑based approaches, including interaction‑driven algorithms and machine‑learning classifiers that outperform single‑symptom screening.[14] Much of the existing literature, however, relies on routine line lists or on unmatched analytical methods, leading to wide variability in estimated diagnostic utility. Failing to account for matching during analysis can introduce bias and obscure clinically relevant predictors, a concern highlighted in methodological critiques of Lassa fever epidemiology.[15] This study addresses this gap by directly comparing matched (conditional) and unmatched (unconditional) multivariable models on the same dataset, demonstrating how analytic alignment with matched sampling improves the identification and stability of early clinical predictors.

While Lassa fever transmission involves the epidemiological triad of agent, host, and environment,[16] this study centers on the host (patient) dimension, that is, the presenting symptoms and their diagnostic value at first contact. Using the triad only insofar as it structures clinical presentation (host) against plausible exposure windows (agent/contact) helps prioritize parsimonious symptom sets that can trigger testing and pre‑isolation in endemic facilities.

Given a 1:1 age‑ and sex‑matched design (50 Reverse Transcription Polymerase Chain Reaction (RT‑PCR)–confirmed cases; 50 controls), conditional logistic regression (CLR) is the principal inferential tool to estimate independent clinical predictors while honoring the matched structure and reducing design‑stage confounding.[17] Because outbreak datasets often include sparse strata or require additional adjustment, unconditional logistic regression (ULR) controlling for matching variables serves as a prespecified sensitivity analysis.[18] Model diagnostics (Hosmer–Lemeshow, and Nagelkerke R²) and post hoc power support calibration, explanatory performance, and interpretability. This dual‑model strategy improves robustness and transparency of clinical‑predictor estimates.

In endemic, resource‑constrained settings, symptom‑based heuristics remain operationally necessary.[19] Evidence consistently shows that fever offers the highest sensitivity, while symptoms such as vomiting and myalgia confer greater specificity, though they occur less frequently.[20] Combining these features into a simple triage rule, particularly in patients with non‑resolving malaria, can accelerate Lassa fever testing and isolation and reduce missed cases and nosocomial exposure.[21] This study, therefore, prioritizes clinically parsimonious predictors that are feasible to apply at first contact and aligned with operational realities in Lassa‑endemic health systems.

Despite substantial scholarship on Lassa fever, clinical prediction at first presentation remains constrained by methodological and operational limitations. Symptom‑performance estimates vary widely because many studies rely on routine surveillance line‑lists and heterogeneous analytic approaches, which are prone to missing data, misclassification, and design misalignment.[22] A key methodological gap is the underuse of matched‑aligned modelling in studies that employ matched sampling; failing to account for matching can bias effect estimates and obscure true clinical predictors.[23] Moreover, although several analyses identify statistically significant symptoms, few translate these findings into operational triage rules appropriate for endemic, resource‑limited facilities.[24] To address these gaps, this study quantifies the diagnostic yield of commonly reported symptoms; identifies independent clinical predictors using both conditional logistic regression and unconditional logistic regression as a prespecified sensitivity analysis on the same matched dataset; and proposes a parsimonious, implementable triage cue to support earlier testing and strengthen IPC in Lassa fever–endemic Nigerian settings. These diagnostic and triage challenges have direct implications for intergovernmental actors such as the World Health Organization (WHO), the Africa Centers for Disease Control and Prevention (Africa CDC), the World Bank, and United Nations Children’s Fund (UNICEF), which support early detection, IPC strengthening, and clinical‑algorithm refinement in Lassa‑endemic countries.

3. Methodology

This study adopts a post‑positivist orientation within a critical‑realist perspective, treating clinical phenomena such as symptoms and Reverse Transcription Polymerase Chain Reaction (RT‑PCR) confirmed disease status as real and measurable, while recognising that their expression is shaped by contextual factors such as malaria co‑endemicity and health‑system constraints.[25] In practice, this orientation supports the use of structured quantitative inference to identify clinical predictors, supplemented by targeted qualitative insights to clarify the operational relevance of those predictors at triage.[26]

Data were collected during the 2018 WHO Grade 2 Lassa fever outbreak in Ondo State, Southwestern Nigeria, through a 1:1 matched case–control investigation at the Federal Medical Centre, Owo (FMCO), the designated treatment and referral centre during the emergency period from January through June 2018. Clinical and symptom data were collected using a structured KoBoCollect questionnaire during follow‑up encounters at the FMCO, between January and June 2018. Cases were consecutive patients with RT‑PCR–confirmed Lassa fever who were available for follow‑up, and controls were patients admitted for other unrelated conditions during the same time frame at the same facility who tested negative for Lassa fever and were available for follow-up. Controls were individually matched to cases on sex and age within ±2 years to minimize confounding by demographic structure and improve efficiency. The analytic sample comprised 100 participants: 50 cases and 50 matched controls, drawn from the outbreak response clinical follow‑up at FMCO.

Data collection employed a structured KoBoCollect instrument to capture demographics and the presence or absence of 17 pre‑specified presenting symptoms (fever, headache, fatigue, vomiting, abdominal pain, muscle pain, chest pain, joint pain, nausea, anorexia, diarrhoea, cough, sore throat, convulsion/seizure, acute hearing loss, dehydration, and rash).[27] Additional epidemiological and environmental variables collected during the outbreak response were not included in this analysis, which is limited to clinical predictors at first contact. All data were de‑identified, range‑checked, and cross‑validated for completeness and internal consistency. Matched and unmatched versions of the dataset were prepared in SPSS version 25 to support design‑concordant and sensitivity analyses using the same source data.[28]

The analytic approach proceeded in two stages. First, symptom frequencies among cases were summarized, and standard diagnostic metrics, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) for each symptom, were calculated using RT‑PCR as the reference standard, to assess the screening value of individual symptoms in a resource‑limited endemic setting.  Second, independent clinical predictors of case status were identified using a dual‑model strategy. Conditional logistic regression (CLR) served as the primary inferential framework because it preserves the 1:1 age‑ and sex‑matched structure; in SPSS version 25, this was implemented using the Cox regression procedure with a constant time, a validated equivalence for matched case–control analysis. Unconditional logistic regression (ULR), adjusting for matching variables, was employed as a pre‑specified sensitivity analysis where matched strata were sparse or additional covariate flexibility was required.[29]

Candidate symptoms were admitted to multivariable models using a liberal bivariate threshold (p < 0.25) to reduce type II error, given the modest sample size. Final models report adjusted odds ratios with 95% confidence intervals, and comparative model performance was assessed using the −2 Log Likelihood, supplemented by calibration and explanatory diagnostics (Hosmer–Lemeshow and Nagelkerke R² in the unconditional models.[30] The two approaches, CLR and ULR, were compared to evaluate whether estimates were consistent and whether the matched design produced superior model fit, as expected in matched case–control studies. This dual‑model strategy ensured both design‑concordant inference (via CLR) and robustness assessment (via ULR). Post hoc power calculations documented in the parent thesis indicated high achieved power for the effect sizes observed for key predictors, guiding interpretation of rarer symptoms that yielded wide confidence intervals.

Because this study’s purpose is operational, supporting earlier recognition and safer IPC at the point of care, concise qualitative insights were incorporated where they clarified the real-world meaning of statistical signals. In 2025, six gender‑stratified FGDs were conducted with Lassa fever cases, matched controls, and healthcare workers at FMCO. For this analysis, only those insights that illuminate how persistent fever after 48-72 hours of antimalarial therapy or the onset of vomiting influence perceived severity and triage decisions were integrated, with full qualitative findings reported elsewhere.

All analyses were conducted on de‑identified records under an outbreak‑response data‑use agreement. Data collection was conducted under the authority of the Ondo State Lassa Fever Emergency Operations Centre and the WHO Incident Management System. Research analysis followed the ethical approvals granted by the Ondo State Ministry of Health Research Ethics Committee (NHREC/Ondo SMH‑HREC/02/01/2025; Protocol Number OSHREC/03/11/2025/1110), including written informed consent for FGD participants.

Given the operational and policy implications of early diagnosis, the analytic approach aligns with the priorities of intergovernmental organizations, including WHO, Africa CDC, the World Bank, and UNICEF, which support surveillance strengthening, early‑warning systems, and improvements in frontline clinical triage in Lassa fever–endemic countries.

4. Findings and Discussion

The demographic characteristics of the 50 cases and 50 controls were broadly comparable across most variables. Age did not differ significantly between groups: the mean age of cases was 36.9 ± 17.2 years and controls 36.3 ± 16.6 years, with no statistically significant difference in either the unmatched (p = 0.855) or matched analysis (paired mean difference = 0.62 ± 3.06 years; 95% CI: −0.25 to 1.49; p = 0.158). Gender distribution was identical by design (56% male, 44% female in both groups). Participants were also similar with respect to marital status, occupation, tribe, religion, and local government area. The only demographic variable showing a statistically significant difference in the matched analysis was educational level: controls had a higher proportion of tertiary‑educated individuals (40% vs. 24%), while cases were more likely to have no formal education (12% vs. 6%). Yoruba ethnicity predominated in both groups, and the majority of participants resided in Owo LGA (82% of cases; 84% of controls).

Across the 50 RT‑PCR–confirmed cases, the most frequently reported symptoms were fever (41/50; 82.0%), headache (30/50; 60.0%), fatigue (19/50; 38.0%), vomiting (17/50; 34.0%), and abdominal pain (16/50; 32.0%) Less common symptoms included while muscle pain (20.0%), chest pain (14.0%), joint pain (14.0%), nausea (8.0%), anorexia (8.0%), diarrhea (6.0%), cough (6.0%), sore throat (6.0%), convulsion/seizure (4.0%), and acute hearing loss (2.0%). No rash was recorded. These frequencies reflect the widely documented non‑specific early presentation of Lassa fever in malaria‑co‑endemic settings and define a realistic set of candidate symptoms for triage in resource‑limited facilities.[31]

When benchmarked against RT‑PCR negative controls (Table 1), fever showed the highest sensitivity (82.0%) but only moderate specificity (56.0%) and a PPV of 65.1%, indicating value for initial screening but limited utility as a stand‑alone rule‑in feature in malaria‑co‑endemic settings.[32] Vomiting, by contrast, demonstrated high specificity (94.0%) and PPV (85.0%) despite low sensitivity (34.0%), making it a more reliable rule‑in indicator, consistent with multi‑center assessments.[33] Muscle pain, though infrequent, also showed very high specificity (98.0%) and PPV (90.9%), paralleling patterns observed in symptom‑cluster analyses.[34]

Furthermore, fatigue exhibited an intermediate diagnostic profile (sensitivity 38.0%; specificity 84.0%; PPV 70.4%), consistent with national outbreak summaries.[35] Rare symptoms such as seizures and acute hearing loss were too uncommon to support front‑line diagnostic use, reflecting earlier observations in classical clinical descriptions.[36] Taken together, these accuracy patterns support a parsimonious triage approach, combining a high‑sensitivity cue (fever) with a high‑specificity cue (vomiting) to prompt early Lassa fever testing and pre‑isolation at initial clinical contact. This tiered configuration reflects a pragmatic balance between maximizing early case capture and minimizing unnecessary isolations, an approach well-suited to overstretched, resource‑limited endemic facilities.

Table 1. Diagnostic performance of reported symptoms for laboratory‑confirmed Lassa fever (n=100; 50 cases, 50 controls).

PPV = Positive Predictive Value, NPV = Negative Predictive Value, † For nausea, anorexia, diarrhea, chest pain, joint pain, sore throat, convulsion/seizure, acute hearing loss, dehydration, cough, and rash, estimates were limited by low prevalence among cases and/or controls; thus, they were not shown

In multivariable analyses (Table 2), fever and vomiting consistently emerged as independent predictors of Lassa fever across model specifications, with the matched design outperforming the unmatched framework. In the conditional logistic regression, which preserved the age–sex matching, fever showed a strong association with case status (AOR 7.9; 95% CI: 1.7–36.3; p = 0.008), and vomiting remained significant (AOR 4.1; 95% CI: 1.1–16.5; p = 0.041), while fatigue was borderline after adjusting for other symptoms. These patterns align with previous clinical analyses reporting fever and vomiting as the most robust predictors of early Lassa fever.[37] In the unconditional model, adjusted for age and sex, the direction of associations remained consistent, with fever and vomiting retaining significance and fatigue becoming significant, illustrating the sensitivity of estimates to model specification, a point noted in recent analytic critiques.[38] Model fit strongly favored the matched analysis (−2LL 41.5 vs 104.5), underscoring the importance of design‑concordant modelling for accurate inference in matched case–control studies, as recommended in epidemiologic methods literature.[39]

Table 2. Independent predictors of Lassa fever: conditional (matched) vs. unconditional (unmatched) models.

Notes: Conditional model adjusted for headache, abdominal pain, and muscle pain. Unconditional model adjusted for matching variables (age, sex) and included the following candidate symptoms based on bivariate screening: headache, nausea, abdominal pain, chest pain, anorexia, and convulsion

Convergence with qualitative insights sharpened the operational meaning of the quantitative signals. Gender‑stratified FGDs with patients, controls, and healthcare workers at FMCO revealed a widely shared frontline heuristic that aligns closely with the statistical findings. Across groups, participants repeatedly described “fever that refuses to go down after malaria drugs” as the point at which suspicion should shift from malaria to Lassa fever. One healthcare worker explained: “When a child or adult has taken malaria treatment for two or three days, and the fever is still high, that is when we start thinking Lassa.” Similarly, patients echoed this perspective, with one participant noting, “I kept treating malaria, but the fever was not going away. That was when the doctor said it could be Lassa.”

Vomiting was also identified as a key escalation trigger in the qualitative data. As one nurse put it: “Once the fever is not responding and the person starts vomiting, we move quickly because that is a red flag.” Another participant described the combination of symptoms succinctly: “Fever that is stubborn and then vomiting, it means we should isolate first before we make a mistake.”

These frontline narratives mirror the quantitative pattern that positions fever as the sensitive screening trigger and vomiting as the more specific rule‑in feature, together forming a pragmatic cue for triage in endemic, resource‑limited settings. The alignment between numerical findings and lived clinical reasoning strengthens the operational validity of the proposed triage heuristic.

Two operational implications follow directly. First, a simple triage rule, “non‑resolving malaria within 48–72 hours plus fever and/or vomiting, then test for Lassa fever and isolate,” is supported by empirical evidence demonstrating high sensitivity for fever and high specificity for vomiting, as well as by provider and community heuristics documented in qualitative studies. In the FGDs, clinicians frequently described persistent fever as the point at which malaria anchoring should be abandoned, and vomiting as the symptom that “tells us to move quickly,” reinforcing its operational value. Adoption of this cue is likely to reduce time‑to‑testing and time‑to‑isolation, thereby limiting nosocomial exposure during periods when health‑facility IPC capacity is stretched.

Second, the matched analytic framework demonstrated superior calibration and interpretability, reflecting the benefits of preserving the study’s age‑ and sex‑matching structure. Epidemiologic guidance similarly emphasizes that matched sampling requires conditional estimation to avoid biased effect estimates, with unmatched models best reserved for robustness checks when strata are sparse.[40] Programs using matched surveillance or outbreak datasets should therefore prioritize design‑concordant conditional logistic regression for deriving early clinical‑predictor evidence.

Other symptoms contributed to incremental but not decisive diagnostic value. Fatigue showed mixed performance, borderline in the matched model but significant in the unmatched sensitivity analysis, reflecting its non‑specific nature and the variability reported in other outbreak analyses.[41] Although not sufficiently discriminatory on its own, fatigue may enhance prediction when combined with fever and/or vomiting, particularly in simple point‑of‑care algorithms designed for resource‑limited settings.[42] By contrast, rare features such as convulsions or acute hearing loss occurred too infrequently to support reliable inference at initial triage; this aligns with earlier observations that such symptoms are more relevant to severity assessment than to early case detection.[43]

The study’s strengths include the use of RT‑PCR as the reference standard, ensuring high diagnostic validity; the age‑ and sex‑matched design, analyzed using both conditional and unconditional logistic frameworks; and a comprehensive assessment of diagnostic performance across commonly reported symptoms. These features, all derived from a high‑burden, real‑world outbreak context, enhance the external relevance of the findings for clinical decision‑making in comparable endemic settings.[44]

However, several limitations merit consideration. The modest sample size underrepresents rare symptoms, contributing to wide confidence intervals for infrequent clinical features; recall and reporting variation may have influenced symptom ascertainment; and the single‑center design may constrain generalizability to settings with different ecological or patient‑mix profiles.[45] Despite these constraints, effect sizes were consistent and clinically coherent, and model fit clearly favored the matched approach, strengthening confidence in the central conclusions and their operational relevance for frontline triage and early identification of Lassa fever cases.

Looking ahead, algorithmic refinement should evaluate symptom‑combination scores, for example, fever plus vomiting plus or minus fatigue, and, where feasible, incorporate low‑cost objective measures such as vital signs or basic laboratory parameters (such as platelet count, transaminases), provided such additions do not undermine usability in low‑resource triage environments.[46] Pragmatic implementation studies in emergency units and fever clinics are needed to assess the real‑world impact of the proposed triage cue on diagnostic yield, time to isolation, and IPC outcomes, while also monitoring for potential unintended effects such as excess isolation or workflow disruption.[47] Such evaluations should be explicitly attentive to equity in access to testing, given well‑documented socioeconomic and structural disparities in Lassa fever care‑seeking and case detection.[48]

In summary, during a WHO Grade 2 outbreak in an endemic setting, fever and vomiting emerged as robust, independent early predictors of Lassa fever when analyzed within a matched case–control framework, consistent with patterns reported in prior outbreak analyses. Their complementary diagnostic profiles, with fever offering high sensitivity and vomiting providing high specificity, support a simple, operational triage rule: “non‑resolving malaria within 48–72 hours plus fever and/or vomiting, then test for Lassa fever and isolate.” Embedding this cue into frontline practice represents a feasible, evidence‑based strategy to accelerate case identification and reduce transmission risk, particularly in resource‑limited facilities confronting overlapping febrile illnesses.

5. Implications for the Intergovernmental Action

Intergovernmental organizations can translate these findings into near‑term system gains by harmonizing a simple, time‑bound triage cue across endemic facilities and underwriting the basic inputs required to act on it. The message is straightforward: when a febrile illness fails to improve after 48–72 hours of antimalarial therapy, and the patient has fever and/or vomiting, clinicians should test for Lassa fever and initiate isolation. This heuristic is grounded in evidence showing fever’s high sensitivity and vomiting’s high specificity, as well as the superior predictive performance of models that respect age/sex matching. Converting this into practice requires institutions such as the WHO, the Africa CDC, and national programs to issue standardized job‑aids, brief refresher materials, and triage prompts for emergency units and outpatient clinics, ensuring frontline teams share a clear escalation threshold and reduce malaria anchoring.[49]

Because the cue is only actionable if facilities are equipped to operationalize it, partners such as UNICEF Supply Division and World Bank health‑security financing channels should facilitate a rapidly deployable IPC bundle, masks, gloves, chlorine, disposable gowns, and basic screening partitions, alongside strengthened specimen‑referral logistics that reduce time‑to‑testing once the cue is triggered. These actions align closely with the quantitative pattern observed (fever and vomiting as complementary predictors) and with qualitative accounts from providers who already recognize “non‑resolving malaria” as a practical red flag in routine care.[50]

A second priority for intergovernmental support is to institutionalize design‑concordant analytics as a program standard, ensuring that outbreak learning reliably improves triage practice. State and national analytic units can be supported in pairing-matching datasets by age and sex and in prioritizing conditional logistic regression for clinical‑predictor analyses, in line with established epidemiologic guidance on matched case–control designs.[51] Publishing brief field summaries that compare model fit and calibration across conditional and unconditional specifications would strengthen analytic transparency and interpretation. This is not merely an academic concern: in this study, the matched model demonstrated superior fit and interpretability, echoing prior observations that design‑aligned modelling yields more reliable predictor estimates in Lassa fever analyses.[52] Such clarity provides programs with a stronger basis for updating triage cues and for communicating to clinicians when and why to escalate from anti-malarial to Lassa fever testing and pre‑isolation, a recurrent theme in FGDs where providers highlighted the need for evidence‑based thresholds.

Short‑cycle micro‑grants enabling 6–12‑week post‑surge analytic rounds, using a minimal and standardized dataset, including a fever‑trajectory field and key symptom flags, would create a sustainable feedback loop in which local evidence continually refines frontline heuristics each season.[53] This iterative, design‑concordant analytic practice would help embed evidence‑informed triage in routine Lassa fever surveillance systems across the endemic region.

Intergovernmental actors can also support pragmatic implementation research evaluating the proposed triage cue in real clinical settings, using simple, actionable endpoints such as diagnostic yield, time‑to‑isolation, health‑worker exposures, and personal protective equipment (PPE) consumption, while monitoring for unintended effects, such as over‑isolation.[54] Where feasible, these studies can test low‑cost “triage‑plus” additions, such as vital signs or basic laboratory markers, provided they do not compromise usability in low‑resource environments.[55]

In parallel, risk‑communication partners should align community messaging with clinical thresholds. Families should be encouraged to seek reassessment when fever does not improve within two to three days of antimalarial treatment, and the combination of vomiting with persistent fever should be presented as a clear indication for isolation and testing, rather than a basis for stigma or delay. [56]These communication cues mirror qualitative testimonies from the FGDs, where participants repeatedly described “fever not going after malaria drugs” as the moment suspicion should shift, and emphasized that “vomiting on top of the fever means it is time to act quickly.” Embedding these insights into routine messaging increases the likelihood that patients will present early enough for the triage cue to be effective, improving both case detection and IPC outcomes in endemic settings.

In conclusion, if intergovernmental agencies do only a few things well, including enshrining the time‑bound fever‑plus‑vomiting cue in national standard operating systems (SOPs) and job‑aids, pre‑positioning the protective basics that enable facilities to act on it safely, and institutionalizing matched‑design analytics so that each surge sharpens the rule, then measurable improvements should be visible within a single dry‑season cycle. These improvements include shorter time‑to‑testing, quicker isolation of true positives, fewer health‑worker exposures attributable to delayed recognition, and the establishment of a routine cadence of matched‑analysis briefs that keep triage guidance current and evidence‑aligned. Such outcomes respond directly to the empirical patterns observed in Ondo State, and are achievable with modest, well‑timed funding and coordinated implementation, transforming a statistically validated triage signal into safer, faster clinical practice across endemic health systems.

6. Conflict of Interest

The author states that there is no conflict of interest.

7. Acknowledgment

I am grateful to the Ondo State Ministry of Health and the Lassa Fever Emergency Operations Centre for facilitating access to outbreak data and field operations during the 2018 response. I thank the Federal Medical Centre, Owo (FMCO) leadership, clinicians, surveillance officers, and data collectors for their collaboration, as well as the patients, controls, and healthcare workers who participated in interviews and focus groups. I also acknowledge that this work formed part of my doctoral research, and I appreciate the guidance of my doctoral supervisor and EUCLID faculty advisors, whose mentorship strengthened the study’s design and analysis.

.

References

Asogun, Danny, Bosede Arogundade, Faith Unuabonah, et al. ‘A Review of the Epidemiology of Lassa Fever in Nigeria’. Microorganisms 13, no. 6 (2025): 1419. Accessed March 21, 2026. https://doi.org/10.3390/microorganisms13061419.

Esan, Adebimpe, George Adejo, Nnamdi Okomba, Afeez A. Soladoye, Nicholas Aderinto, and David B. Olawade. ‘AI-Driven Diagnosis of Lassa Fever: Evidence from Nigerian Clinical Records’. Computational Biology and Chemistry 120 (February 2026): 108627. Accessed March 21, 2026. https://doi.org/10.1016/j.compbiolchem.2025.108627.

Ilori, Elsie A., Yuki Furuse, Oladipupo B. Ipadeola, et al. ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’. Emerging Infectious Diseases Journal – CDC Volume 25, no. Number 6 (2019). Accessed March 21, 2026. https://doi.org/10.3201/eid2506.181035.

Isere, Elvis Efe, Temioluwa Fuwape, Gboyega Adekunle Famokun, et al. ‘Epidemiological Pattern of Lassa Fever Outbreak in Ondo State, Southwest Nigeria, 2014 to 2019’. Open Journal of Epidemiology 11, no. 01 (2021): 92–100. Accessed March 24, 2026. https://doi.org/10.4236/ojepi.2021.111009.

Kirkwood, Betty R., Jonathan A. C. Sterne, and Betty R. Kirkwood. Essential Medical Statistics. 2nd ed. Blackwell Science, 2003.

Marreiros, Maria Inês. ‘Epidemiology of Infectious Diseases: The Epidemiological Triad’. Arcadia, 18 June 2023. Accessed March 21, 2026. https://www.byarcadia.org/post/epidemiology-of-infectious-diseases-101-the-epidemiological-triad.

McCormick, J. B., I. J. King, P. A. Webb, et al. ‘A Case-Control Study of the Clinical Diagnosis and Course of Lassa Fever’. Journal of Infectious Diseases 155, no. 3 (1987): 445–55. Accessed March 21, 2026. https://doi.org/10.1093/infdis/155.3.445.

Merson, Laura, Josephine Bourner, Sulaiman Jalloh, et al. ‘Clinical Characterization of Lassa Fever: A Systematic Review of Clinical Reports and Research to Inform Clinical Trial Design’. PLOS Neglected Tropical Diseases 15, no. 9 (2021): e0009788. Accessed March 21, 2026. https://doi.org/10.1371/journal.pntd.0009788.

Ochu, Chinwe Lucia, Lorretta Ntoimo, Ikenna Onoh, et al. ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’. Scientific Reports 13, no. 1 (2023): 6545. Accessed March 21, 2026. https://doi.org/10.1038/s41598-023-33187-y.

Olayinka, Adebola T., Kelly Elimian, Oladipupo Ipadeola, et al. ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’. PLOS Global Public Health 2, no. 8 (2022): e0000191. Accessed March 21, 2026. https://doi.org/10.1371/journal.pgph.0000191.

Pal, Mahendra, Kirubel Paulos Gutama, Leena Gowda, and Pratibha Dave. ‘Lassa Fever: An Emerging and Re-Emerging Fatal Viral Disease of Public Health Concern’. American Journal of Public Health Research 10, no. 4 (2022): 143–46. Accessed March 21, 2026. https://doi.org/10.12691/ajphr-10-4-2.

Pearce, Neil. ‘Analysis of Matched Case-Control Studies’. BMJ, 25 February 2016, i969. Accessed March 21, 2026. https://doi.org/10.1136/bmj.i969.

Yun, Nadezhda E., and David H. Walker. ‘Pathogenesis of Lassa Fever’. Viruses 4, no. 10 (2012): 2031–48. Accessed March 21, 2026. https://doi.org/10.3390/v4102031.

[1] Mahendra Pal et al., ‘Lassa Fever: An Emerging and Re-Emerging Fatal Viral Disease of Public Health Concern’, American Journal of Public Health Research 10, no. 4 (2022): 143–46, accessed March 21, 2026, https://doi.org/10.12691/ajphr-10-4-2.

[2] Danny Asogun et al., ‘A Review of the Epidemiology of Lassa Fever in Nigeria’, Microorganisms 13, no. 6 (2025): 1419, accessed March 21, 2026, https://doi.org/10.3390/microorganisms13061419.

[3] Laura Merson et al., ‘Clinical Characterization of Lassa Fever: A Systematic Review of Clinical Reports and Research to Inform Clinical Trial Design’, PLOS Neglected Tropical Diseases 15, no. 9 (2021): e0009788, accessed March 21, 2026, https://doi.org/10.1371/journal.pntd.0009788.

[4] Adebola T. Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’, PLOS Global Public Health 2, no. 8 (2022): e0000191, accessed March 21, 2026, https://doi.org/10.1371/journal.pgph.0000191.

[5] Nadezhda E. Yun and David H. Walker, ‘Pathogenesis of Lassa Fever’, Viruses 4, no. 10 (2012): 2031–48, accessed March 21, 2026, https://doi.org/10.3390/v4102031.

[6] Elvis Efe Isere et al., ‘Epidemiological Pattern of Lassa Fever Outbreak in Ondo State, Southwest Nigeria, 2014 to 2019’, Open Journal of Epidemiology 11, no. 01 (2021): 95–96, accessed March 24, 2026, https://doi.org/10.4236/ojepi.2021.111009.

[7] Elsie A. Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’, Emerging Infectious Diseases Journal – CDC Volume 25, no. Number 6 (2019): 1066–74, accessed March 21, 2026,  https://doi.org/10.3201/eid2506.181035.

[8] Merson et al., ‘Clinical Characterization of Lassa Fever’.

[9] Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[10] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[11] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[12] Adebimpe Esan et al., ‘AI-Driven Diagnosis of Lassa Fever: Evidence from Nigerian Clinical Records’, Computational Biology and Chemistry 120 (February 2026): 108627, accessed March 21, 2026, https://doi.org/10.1016/j.compbiolchem.2025.108627; Chinwe Lucia Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’, Scientific Reports 13, no. 1 (2023): 6545, https://doi.org/10.1038/s41598-023-33187-y.

[13] J. B. McCormick et al., ‘A Case-Control Study of the Clinical Diagnosis and Course of Lassa Fever’, Journal of Infectious Diseases 155, no. 3 (1987): 445–55, https://doi.org/10.1093/infdis/155.3.445; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[14] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’; Esan et al., ‘AI-Driven Diagnosis of Lassa Fever’.

[15] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[16] Maria Inês Marreiros, ‘Epidemiology of Infectious Diseases: The Epidemiological Triad’, Arcadia, 18 June 2023, accessed March 21, 2026, https://www.byarcadia.org/post/epidemiology-of-infectious-diseases-101-the-epidemiological-triad.

[17] Neil Pearce, ‘Analysis of Matched Case-Control Studies’, BMJ, 25 February 2016, i969, accessed March 21, 2026, https://doi.org/10.1136/bmj.i969; Betty R. Kirkwood et al., Essential Medical Statistics, 2nd ed (Blackwell Science, 2003).

[18] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’; Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’.

[19] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[20] Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[21] Merson et al., ‘Clinical Characterization of Lassa Fever’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[22] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[23] Pearce, ‘Analysis of Matched Case-Control Studies’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[24] Merson et al., ‘Clinical Characterization of Lassa Fever’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[25] Merson et al., ‘Clinical Characterization of Lassa Fever’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[26] Pearce, ‘Analysis of Matched Case-Control Studies’; Kirkwood et al., Essential Medical Statistics.

[27] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[28] Pearce, ‘Analysis of Matched Case-Control Studies’.

[29] Pearce, ‘Analysis of Matched Case-Control Studies’; Kirkwood et al., Essential Medical Statistics.

[30] Kirkwood et al., Essential Medical Statistics.

[31] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[32] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[33] Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[34] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[35] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’.

[36] McCormick et al., ‘A Case-Control Study of the Clinical Diagnosis and Course of Lassa Fever’.

[37] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[38] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’; Pearce, ‘Analysis of Matched Case-Control Studies’.

[39] Kirkwood et al., Essential Medical Statistics; Pearce, ‘Analysis of Matched Case-Control Studies’.

[40] Pearce, ‘Analysis of Matched Case-Control Studies’; Kirkwood et al., Essential Medical Statistics.

[41] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[42] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[43] McCormick et al., ‘A Case-Control Study of the Clinical Diagnosis and Course of Lassa Fever’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[44] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[45] Merson et al., ‘Clinical Characterization of Lassa Fever’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[46] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[47] Merson et al., ‘Clinical Characterization of Lassa Fever’.

[48] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[49] Merson et al., ‘Clinical Characterization of Lassa Fever’.

[50] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[51] Pearce, ‘Analysis of Matched Case-Control Studies’.

[52] Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’; Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.

[53] Merson et al., ‘Clinical Characterization of Lassa Fever’.

[54] Merson et al., ‘Clinical Characterization of Lassa Fever’; Ilori et al., ‘Epidemiologic and Clinical Features of Lassa Fever Outbreak in Nigeria, January 1–May 6, 2018’.

[55] Olayinka et al., ‘Analysis of Sociodemographic and Clinical Factors Associated with Lassa Fever Disease and Mortality in Nigeria’.

[56] Ochu et al., ‘Predictors of Lassa Fever Diagnosis in Suspected Cases Reporting to Health Facilities in Nigeria’.