📄 medicine

Unexplained neuroinvasive disease as an early-warning signal of West Nile virus transmission in Spain, 2016-2019

This study demonstrates that analyzing unexplained viral encephalitis, meningitis, and acute flaccid paralysis cases using a Bayesian shared-component model can effectively identify hidden West Nile virus transmission hotspots in Spain, offering a potential early-warning system for viral circulation before clinical cases are officially recognized.

Original authors: Javier Del-Águila-Mejía, Beatriz Fernández-Martínez, Shirin Taheri, Noemí López-Perea, Josefa Masa-Calles, Jordi Figuerola, Juan de Mata Donado-Campos, Diana Gómez-Barroso

Published 2026-07-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Javier Del-Águila-Mejía, Beatriz Fernández-Martínez, Shirin Taheri, Noemí López-Perea, Josefa Masa-Calles, Jordi Figuerola, Juan de Mata Donado-Campos, Diana Gómez-Barroso

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine Spain as a vast, complex garden. For years, gardeners (public health officials) have been worried about a specific, invisible pest: the West Nile Virus (WNV). This virus is carried by mosquitoes and can cause severe brain inflammation in humans. However, the virus is tricky. Most of the time, it hides in the shadows, infecting birds and horses, and only occasionally jumping to humans. When it does jump, it often looks exactly like other common illnesses, making it hard to spot until a big outbreak happens.

The researchers in this paper asked a clever question: "If we can't see the virus directly, can we see the 'footprints' it leaves behind in the hospital?"

Here is a simple breakdown of their study, using everyday analogies:

1. The Mystery of the "Unexplained" Patients

Every year, hospitals in Spain treat many people for severe brain infections (like encephalitis or meningitis). Doctors run tests, but in many cases, they can't find the specific germ causing the sickness. These are labeled as "Unexplained Viral Encephalitis" (VEUC).

Think of these patients as people walking into a hospital with a broken leg, but the X-ray shows nothing. The doctors know something is wrong, but they can't name the cause. The researchers wondered: Could some of these "broken legs" actually be caused by the West Nile Virus, even if the doctors didn't know it yet?

2. The Detective Work: Mapping the "Ghost"

The team looked at data from 2016 to 2019. They created two maps:

Map A: Where the West Nile Virus was officially confirmed in humans, horses, or birds.
Map B: Where there were "extra" cases of those unexplained brain infections.

They used a special statistical tool (a Bayesian Shared-Component Model) that acts like a transparent overlay. Imagine placing a clear sheet over Map A and another over Map B. The tool looks for areas where the "hot spots" (areas with high numbers) on both maps line up perfectly.

3. The Big Discovery: A Hidden Pattern

The study found that the two maps did overlap significantly.

The Known Hotspots: The virus was already known to be active in the southwest and northeast of Spain. The "unexplained" brain infections were also high in these exact same areas. This made sense.
The Surprise: The "unexplained" brain infections were also high in the southeastern Mediterranean coast. At the time of the study (2016–2019), the virus had not been officially detected there yet.

The Analogy: Imagine you are looking for a leak in a house. You see water damage (the unexplained illness) on the ceiling in a room where you haven't found the pipe yet. The water damage is your early warning signal that a leak (the virus) is likely happening, even if you haven't seen the water dripping from the pipe yet.

4. Why Did This Happen? (The Ingredients)

The researchers checked if environmental factors explained this overlap. They found that the "shared pattern" was driven by the perfect recipe for the virus to thrive:

Weather: Warm temperatures and specific rainfall patterns.
Nature: The presence of the right mosquitoes (Culex perexiguus), birds (which carry the virus), and horses.
Human Activity: Areas where people live and farm.

When they added these environmental factors to their model, the "shared pattern" between the virus and the unexplained illnesses became even clearer. It confirmed that the unexplained illnesses weren't just random; they were happening in the exact places where the virus should be circulating.

5. The "Noise" vs. The "Signal"

It's important to note that not all unexplained brain infections were caused by West Nile. The study found a separate "noise" pattern in the north of Spain that had nothing to do with the virus. This is like having a broken leg from a car accident (unrelated to the virus) versus a broken leg from a fall caused by a slippery floor (related to the virus). The researchers successfully separated the "virus-related" signal from the "other causes" noise.

The Bottom Line

The paper concludes that unexplained neurological diseases can act as an early-warning system.

If a specific area has a sudden spike in "mystery brain infections," it might mean the West Nile Virus is already circulating there, waiting to be officially detected. By paying attention to these "unexplained" cases, health officials could potentially spot the virus earlier, before a major outbreak occurs, allowing them to test for it and warn the public sooner.

In short: The study suggests that when we see a cluster of "mystery" brain illnesses, we should suspect the West Nile Virus might be the invisible culprit, even before we have a confirmed lab test to prove it.

Technical Summary: Unexplained Neuroinvasive Disease as an Early-Warning Signal of West Nile Virus Transmission in Spain, 2016-2019

Problem Statement
West Nile virus (WNV) is a rapidly expanding mosquito-borne pathogen in Europe, with Spain experiencing recurrent outbreaks and increasing geographical expansion. While most WNV infections are asymptomatic, a small fraction progress to West Nile neuroinvasive disease (WNV-ND), presenting as meningitis, encephalitis, or acute flaccid paralysis. Clinically, WNV-ND is indistinguishable from other viral neurological infections of unknown etiology (VEUC). Consequently, in regions where WNV is not actively suspected or routinely tested, a significant portion of WNV transmission may remain concealed within the burden of unexplained neurological cases. The study addresses the hypothesis that unrecognized WNV transmission contributes to the burden of VEUC and that these two phenomena share a common geographical structure driven by social-ecological conditions.

Methodology
The authors analyzed data from Spanish municipalities between 2016 and 2019, a period preceding the major 2020 WNV outbreak which heightened clinical awareness.

Data Sources:
- VEUC: Hospital discharge records (CMBD) for patients >5 years old with primary or secondary diagnoses of viral encephalitis, meningitis, or acute flaccid paralysis of unknown cause (ICD-10 codes). Cases were aggregated by municipality, and Standardized Incidence Ratios (SIR) were calculated. Municipalities with SIR > 1 were classified as having VEUC excess.
- WNV: Presence/absence indicators for WNV in humans, equines, and birds were derived from national surveillance networks (RENAVE and Ministry of Agriculture). A municipality was marked as "1" if WNV was detected in any species during the study period.
- Covariates: A national WNV risk index (based on a One Health framework integrating vectors, reservoirs, and environment) and a suite of social-ecological covariates (atmospheric, hydrological, biotic, land-use, and anthropogenic factors) were utilized.
Statistical Approach:
- A Bayesian shared-component model (SCM) was fitted within a hierarchical framework. This model decomposes the spatial variation of two binary outcomes (VEUC excess and WNV presence) into:
  1. A shared spatial component ( $\lambda_i$ ) common to both diseases.
  2. Disease-specific components (structured and unstructured random effects) unique to each outcome.
- The model was fitted using R-INLA with a data-stacking strategy.
- A stepwise modeling strategy was employed: a null model (M1), a model incorporating the WNV Risk Index (M2), and subsequent models adding atmospheric-hydrological (M3), biotic-anthropogenic (M4), and finally a joint model (M5) with all significant predictors.

Key Results

Shared Spatial Pattern: The baseline model (M1) revealed a pronounced shared spatial component linking VEUC excess and WNV presence. This pattern identified three elevated risk zones: southwestern Spain, northeastern Spain, and unexpectedly, the southeastern Mediterranean coast. The latter region showed no documented WNV cases during 2016–2019 but exhibited a strong shared signal.
Asymmetric Coupling: The shared component loaded substantially more heavily on WNV than on VEUC ( $\beta \approx 5.0$ ), indicating an asymmetric relationship where WNV presence is a stronger driver of the shared signal than VEUC excess.
Role of the WNV Risk Index: Including the WNV Risk Index (M2) significantly improved model fit and compressed the shared spatial field, suggesting that ecological suitability explains a large proportion of the shared geography. The index was strongly associated with VEUC excess (OR 11.4).
Covariate Drivers:
- Atmospheric-Hydrological: Variables such as vapour pressure and precipitation seasonality explained a large proportion of the shared variation.
- Biotic-Anthropogenic: Culex perexiguus suitability was the strongest determinant for WNV, while human footprint was the strongest predictor for VEUC excess (independent of WNV).
VEUC-Specific Signal: A distinct VEUC-specific spatial pattern persisted across all models, particularly in northern Spain and the Ebro valley. This component remained stable even after introducing ecological covariates, suggesting that a significant fraction of unexplained neurological disease is driven by factors independent of WNV transmission (e.g., urban development or healthcare access).

Key Contributions

Spatial Overlap Identification: This is the first population-wide study demonstrating a spatial overlap between documented WNV occurrence and unexplained neuroinvasive disease in Spain.
Syndromic Surveillance Utility: The study provides evidence that unexplained neurological disease can serve as a syndromic signal for WNV circulation, particularly in areas where the virus has not yet been detected in humans or animals.
Methodological Framework: The application of a Bayesian shared-component model successfully disentangled common ecological drivers from disease-specific factors, validating the approach for identifying "silent" transmission zones.
Validation of Ecological Models: The findings support the predictive power of ecological risk models (like the WNV Risk Index) by showing that areas predicted to be suitable for WNV also exhibit excess unexplained neurological disease, even prior to confirmed case detection.

Significance and Claims
The paper claims that focusing WNV monitoring on hotspots of unexplained neurological disease could provide an early warning of virus circulation before clinical WNV cases are formally recognized. The authors argue that the shared spatial signal is not merely a statistical artifact but reflects the convergence of climatic suitability, vector distribution, and host communities required for viral circulation.

Crucially, the study posits that this syndromic signal offers a distinct advantage over traditional surveillance: it does not rely on the prior detection of human cases. Instead, it leverages routinely collected health data to identify areas where transmission may already be occurring but remains undetected due to lack of specific testing. The authors note that subsequent WNV detections in the southeastern Mediterranean coast by 2025 are consistent with their 2016–2019 findings, supporting the utility of this approach as a proactive surveillance tool.

The authors maintain a modest tone regarding causality, acknowledging that the associations are ecological rather than causal. They highlight limitations, such as the inability to distinguish true undiagnosed infections from variations in diagnostic intensity within hospital records. However, they conclude that integrating ecological risk and routine health data within a single spatial framework offers a valuable strategy for prioritizing laboratory testing and preventive interventions in the context of One Health surveillance.