Cerebrospinal Fluid Testing for Neuroinvasive West Nile Virus and Measures to Improve Guideline Adherence

Abstract

Background

Diagnosis of West Nile virus (WNV) neuroinvasive disease, an important cause of neurologic disability in endemic areas, requires appropriate testing given its nonspecific presentation. Guidelines recommend cerebrospinal fluid (CSF) testing of WNV-specific immunoglobulin M (IgM) in all patients, with additional reverse transcription-polymerase chain reaction (RT-PCR) testing only in those who are unable to mount a humoral response due to significant immunosuppression. WNV testing is known to be underutilized, but provider choice between the appropriate IgM test and lower sensitivity RT-PCR test when both are readily available has not been described.

Methods

We analyzed testing patterns for suspected neuroinvasive WNV in a retrospective analysis of 1304 adult patients in a Boston hospital network who underwent CSF testing of WNV-specific IgM or WNV RNA using RT-PCR between 2016 and 2023. Both CSF IgM and RT-PCR were available to order during the duration of the study. Relevant clinical, laboratory, and demographic data were extracted from the electronic health record (EHR).

Results

The median age in our cohort was 63 years, and 46% of patients were female. Sole testing with CSF RT-PCR occurred in 73% of patients, and WNV testing guidelines were adhered to in only 26% of cases. Elevated CSF protein, CSF lymphocytic pleocytosis, admission to a neurology service, immunocompetence, race, and hospital site were significantly associated with improved adherence. WNV tests were positive in 26 patients, and patients whose CSF testing patterns adhered to guidelines were 12 times more likely to receive a WNV diagnosis.

Discussion

The existence of guidelines for appropriate diagnostic testing may not be sufficient to encourage appropriate CSF WNV IgM testing when the RT-PCR test, which is not recommended in immunocompetent patients, remains readily available for ordering. Differential guideline adherence based on institutional factors suggests that systems changes may play an important role in shaping provider behavior around appropriate use of diagnostic testing.

Background

West Nile virus (WNV), a flavivirus transmitted by Culex mosquitoes, is the most common arboviral infection in North America with roughly 2000 cases annually in the United States.1,2 Although most West Nile virus infections are asymptomatic or present as West Nile fever, West Nile neuroinvasive disease carries a 10% mortality rate and accounts for the majority of WNV-related deaths despite representing less than 1% of cases.3 The clinical presentation of WNV neuroinvasive disease closely mimics other causes of viral meningitis, encephalitis, or myelitis, necessitating appropriate utilization of diagnostic testing for accurate diagnosis.4

Diagnosis of WNV neuroinvasive disease requires detection of either viral ribonucleic acid (RNA) by reverse transcription-polymerase chain reaction (RT-PCR) or WNV-specific antibodies, typically immunoglobulin M (IgM) in cerebrospinal fluid (CSF) or serum from a patient with a compatible presentation. However, current guidelines, including those from the United States Centers for Disease Control and Prevention (CDC) since 2010, do not recommend using WNV RT-PCR in CSF in immunocompetent patients due to rapid viral clearance within 3-4 days of symptom onset, leading to low sensitivity in these patients (reported between 10 and23%).5–10 Rather, the CDC recommends testing CSF for WNV IgM in all patients, as antibodies are typically detectable by day 3 or 4 of symptom onset and persist for weeks to months.6,8 In patients with severely impaired humoral immunity, specifically those receiving B-cell depleting therapies, CSF RT-PCR is more sensitive than CSF IgM (93% vs 24%).11 To improve diagnostic yield, CDC guidelines recommend testing CSF with both RT-PCR and IgM in patients up to 12 months from B-cell depleting medications, which is extrapolated in other professional guidelines to include transplant patients.8,12,13

Incomplete CSF testing for infectious pathogens likely contributes to the low rates of etiologic diagnosis, often less than 50% in U.S. based studies, in aseptic meningitis and encephalitis.14–17 Testing for WNV is inconsistent: during outbreaks in Texas and Arizona only 37–40 % of meningitis or encephalitis patients underwent evaluation for WNV in CSF or serum.15,18 Notably, the Arizona study did not report whether testing used RT-PCR or IgM, while in the Texas study, RT-PCR was used in only three cases. Provider testing practices in settings where both RT-PCR and IgM testing are available are unknown. As clinicians are

faced with expanding options to test for neuroinfectious diseases, it is imperative to understand whether providers adhere to diagnostic guidelines when multiple testing modalities are available.

Our study aimed to quantify adherence to CDC guidelines for CSF testing in neuroinvasive WNV at two tertiary hospitals that give providers the choice to order the recommended first-line (IgM) or not recommended first-line test (RT-PCR). The electronic health record (EHR) at both institutions allows providers to order either WNV IgM or RT-PCR from CSF at their discretion. However, when using an EHR-based lumbar puncture order set, only the WNV RT-PCR option is displayed – potentially steering providers away from the guideline-recommended first-line IgM test and toward less sensitive RT-PCR testing. Secondly, we aimed to understand whether guideline adherence influenced neuroinvasive WNV diagnosis rates and explore factors impacting adherence. The study’s setting allowed for a natural experiment: the two hospitals, located within 5 miles of one another, share residents and serve similar patient populations, but differ in their processes for ordering WNV IgM testing. At Hospital A, ordering WNV IgM requires completing a paper form to submit via email for State Laboratory approval, while Hospital B allows for ordering without additional documentation or approval. We hypothesized that Hospital B, with fewer barriers to sending guideline-recommended IgM testing, would show higher adherence to CDC testing guidelines.

Methods

Study Cohort

Our study was approved by the Institutional Review Board (IRB #2019P003215), which granted a waiver of informed consent for a retrospective review of patient clinical records. We identified consecutive patients at Massachusetts General Hospital and Brigham and Women’s Hospital in Boston, MA, with a clinical suspicion of WNV neuroinvasive disease – defined as a provider ordering a CSF test for WNV detection from a lumbar puncture. We identified patients through a search of the institutional Research Patient Data Registry (RPDR) for patients 18 years or older who had a CSF WNV assay ordered at either hospital between 2016 and 2023. The initial study year reflected the point of complete implementation of EPIC and Sunquest at the institutions. The CSF WNV tests that we used for the search and inclusion criteria were: CSF WNV RT-PCR, CSF WNV IgG/IgM antibodies, or CSF Massachusetts State Lab Arboviral Panel (which included WNV IgM antibodies). Data on serum WNV testing were requested; however, serum testing was not an inclusion criterion given our primary focus on CSF testing patterns.  We additionally queried a separate internal database of all clinical laboratory test results for the same above-listed WNV tests during the same period as a second check to ensure cohort comprehensiveness (“Crystal Reports” SAP Business Objects BI Platform 4.2 Support Pack 7). There was no predetermined study size as we wished to include as many patients as met criteria.

We included all patients aged 18 years or greater at time of testing who had a WNV identification test ordered from the CSF as part of inpatient or emergency department encounters between January 2016 and December 2023. Patients were included even if this test did not produce a result due to sample collection or processing issues. We included these patients because our primary aim was to explore provider ordering behavior rather than test characteristics. Twenty-three patients were excluded, in whom the test order was cancelled, and the CSF sample was never sent by the ordering team, suggesting it was ordered in error. 

Data collection and definitions

Demographic, clinical, and laboratory data were obtained using RPDR, with manual chart extraction (CMQ, KS) when needed for data elements not provided in the bulk query.

Patients were considered immunosuppressed based on CDC and professional guidelines if they had an ICD-10 code for a history of a solid organ or hematopoietic stem cell transplant (see Supplementary Table 1), or were prescribed B-cell depleting medications (rituximab, ocrelizumab, or ofatumumab) in the twelve months preceding CSF testing. Two investigators (CMQ, KS) manually verified immunosuppression status for 100% of patients (171/171) identified by the above algorithm and for 10% (113/1133) of randomly selected immunocompetent patients who were correctly identified. Admitting service was defined by the billing attending listed for the index encounter and cross-checked against the hospital physician directory. Services were grouped into four categories: neurology (neurology floor teams and neurosciences intensive care units), internal medicine (general medicine, medical intensive care units, and non-oncology medical subspecialties), oncology, and other (surgical services, emergency medicine, etc.); in final analyses these were grouped into neurology and non-neurology services. We defined the testing site by encounter location (Hospital A or Hospital B) where the test was ordered; these labels are used hereafter to maintain confidentiality.

Sex was defined by assigned sex at birth; age was based on the date of CSF testing. Race was defined by patient report and categorized as white, black, other, or not reported. To analyze financial demographics, patients were categorized by insurance type based on the primary payer of the CSF-testing admission: private/Medicare versus public/safety net programs (including those initially uninsured). The WNV season was June to November, covering 99% of reported U.S. cases from 1999-2023.¹⁹ Lumbar puncture hospital day was defined as the difference between the calendar days of CSF testing and hospital admission. CSF cell counts were categorized as normal (total nucleated cell count <5), lymphocytic pleocytosis (cell count ≥5 with ≥50% lymphocytes), or pleocytosis without lymphocyte predominance (cell count ≥5 with <50% lymphocytes).

Guideline adherence was defined as ordering CSF WNV IgM for immunocompetent patients, and both CSF WNV IgM and RT-PCR for immunocompromised patients (Figure 1A).⁸ Excess testing (which was still considered adherent) was defined as ordering both CSF WNV IgM and RT-PCR in immunocompetent patients. Diagnosis of WNV neuroinvasive disease was defined as the presence of a positive CSF RT-PCR, CSF IgM, or serum IgM result in patients who underwent CSF testing for WNV. WNV tests were categorized as positive, negative, or no result (e.g., testing refused due to inadequate or improper sample, insufficient clinical history, or laboratory or transport errors).

Data Analyses

The primary outcome measure was the proportion adherent to CDC WNV CSF testing guidelines. Secondary measures included the proportion of positive test results, incidence of neuroinvasive WNV diagnosis, and the proportion of tests yielding no result. Data were analyzed using R (version 4.3.3) and STATA (version 14) and expressed as counts and percentages, or medians with interquartile ranges (IQR) for non-normally distributed variables. Continuous variables were analyzed using Wilcoxon rank-sum tests for two-group comparisons and Kruskal-Wallis tests for greater than two groups; analyses of categorical variables used Pearson’s χ² tests. No post-hoc pairwise tests were performed. Logistic regression models were adjusted for prespecified demographic variables of sex, age, race, and insurance type, along with variables identified as significant in univariate analyses. Statistical significance was defined as p-value <0.05, as we tested one primary hypothesis in a single regression model. Analyses followed STROBE cross-sectional reporting guidelines.

Data Availability

Anonymized data can be made available upon reasonable request to any qualified investigator with a valid IRB, through the Brain Data Science Platform (bdsp.io).

Data Description

The dataset contains fully de-identified, row-level information on 1,304 adult (≥18 y) inpatients who underwent cerebrospinal-fluid (CSF) WNV testing at two tertiary‐care hospitals in the Mass General Brigham (MGB) network between January 1, 2016, and December 31, 2023. It was created for the study “Cerebrospinal Fluid Testing for Neuroinvasive West Nile Virus and Measures to Improve Guideline Adherence” (accepted, Neurology).

per HIPAA Safe-Harbor, all direct identifiers have been removed; ages > 89 were top-coded to age “90+” mentioned in csv files as "?90"

Files include the main analytic table (wnv_csf_testing.csv), a subject-level metadata file (subject_info_WNV.csv), a variable dictionary (variables.csv), and a categorical codebook (value_codes.csv)

General metadata
The subject_info_WNV.csv file contains one record per patient with demographic and high-level clinical attributes (see mapping below). All identifiers are replaced with a unique BDSP subject ID.

Analytic data
The wnv_csf_testing.csv file holds CSF and serum laboratory values, WNV test orders & results, and derived guideline-adherence flags. Each column is described in variables.csv; categorical encodings are resolved via value_codes.csv.

Mapping of subject_info_WNV.csv (Metadata File)

Usage Notes

The dataset is provided as a single archive "wnv_csf_dataset.zip" whose contents are plain-text CSV files (comma-separated, UTF-8). CSV is a widely supported, open format that can be imported into R, Python/pandas, Excel, Stata, and most BI tools without special software.

wnv_csf_dataset.zip
├── subject-info.csv      # subject-level demographics & dates (human readable)
├── wnv_csf_testing.csv       # main analytic table (1 row / patient)
├── variables.csv             # variable dictionary (name, type, label…)
├── value_codes.csv           # categorical codebook

Release Notes

Data collection and sharing is performed under Institutional Review Board (IRB) approvals and data sharing agreements among participating hospitals, with waiver of the requirement for informed consent. The data is generated as part of usual patient care. All data is de-identified. 

Ethics

Our study was approved by the Institutional Review Board (IRB #2019P003215), which granted a waiver of informed consent for a retrospective review of patient clinical records

Conflicts of Interest

There are no conflict of interests to be declared by any author of the manuscript.