Neurological diagnoses in hospitalized COVID-19 patients associated with adverse outcomes: A multinational cohort study

Abstract

Few studies examining the patient outcomes of concurrent neurological manifestations during acute COVID-19 leveraged multinational cohorts of adults and children or distinguished between central and peripheral nervous system (CNS vs. PNS) involvement. Using a federated multinational network in which local clinicians and informatics experts curated the electronic health records data, we evaluated the risk of prolonged hospitalization and mortality in hospitalized COVID-19 patients from 21 healthcare systems across 7 countries. For adults, we used a federated learning approach whereby we ran Cox proportional hazard models locally at each healthcare system and performed a meta-analysis on the aggregated results to estimate the overall risk of adverse outcomes across our geographically diverse populations. For children, we reported descriptive statistics separately due to their low frequency of neurological involvement and poor outcomes. Among the 106,229 hospitalized COVID-19 patients (104,031 patients ≥18 years; 2,198 patients <18 years, January 2020-October 2021), 15,101 (14%) had at least one CNS diagnosis, while 2,788 (3%) had at least one PNS diagnosis. After controlling for demographics and pre-existing conditions, adults with CNS involvement had longer hospital stay (11 versus 6 days) and greater risk of (Hazard Ratio = 1.78) and faster time to death (12 versus 24 days) than patients with no neurological condition (NNC) during acute COVID-19 hospitalization. Adults with PNS involvement also had longer hospital stay but lower risk of mortality than the NNC group. Although children had a low frequency of neurological involvement during COVID-19 hospitalization, a substantially higher proportion of children with CNS involvement died compared to those with NNC (6% vs 1%). Overall, patients with concurrent CNS manifestation during acute COVID-19 hospitalization faced greater risks for adverse clinical outcomes than patients without any neurological diagnosis. Our global informatics framework using a federated approach (versus a centralized data collection approach) has utility for clinical discovery beyond COVID-19.

Fig 1. Study design and federated learning approach.

A. We constructed Cox proportional hazard models to evaluate clinical endpoints in acute COVID-19 patients with concurrent neurological diagnoses. Patients were followed up to 90 days after the first acute COVID-19 hospital admission. Models were adjusted for pre-existing comorbidity burden and prior neurological conditions as well as baseline demographics, including age group, sex, race/ethnicity. B. The analysis plan was provided as a standardized R package and containerized with Docker to facilitate local deployment at each participating healthcare system. Cox proportional hazards statistics (summary.coxph) were extracted from the analysis at each healthcare system and included in a random-effects meta-analysis to pool the summary statistics. NNC: No Neurological Condition; CNS: Central Nervous System diagnosis; PNS: Peripheral Nervous System diagnosis.

Fig 2. Demographic profile for each participating healthcare system arranged by country.

Cohort-wise breakdown of the number of patients, age range, sex, severity status, mortality outcome, readmission status, and race at each healthcare system for each of the following neurological status during acute COVID-19 hospitalization: no neurological condition (NNC), central nervous system (CNS) diagnosis, and peripheral nervous system (PNS) diagnosis. Healthcare systems are arranged by country in descending order by the number of included participating healthcare systems. The stacked bar charts indicate the normalized distribution of age and race. The nested pie-charts are stratified by the neurological status with the darker portion representing the proportion of patients having the value of the binary variable for the given column header.

Fig 3. Frequency of neurological diagnosis codes by age group.

For each age group, we report the total number and proportion of patients who had the associated ICD-10 code. Neurological diagnoses are listed in descending order of overall frequency. Please refer to S1 Fig for the incidence of severe COVID-19 status and mortality as stratified by concurrent neurological status in adults and children. S5 Table details the total counts and percentages of patients with ICD-10 (and ICD-9 codes) as stratified by adult and pediatric populations.

Fig 4. Relative risk of a neurological diagnosis in the adult patient population during acute COVID-19 hospitalization for each pre-admission health condition.

We calculated the relative risks (with 95% confidence intervals) of any central nervous system (CNS) diagnosis (A) and any peripheral nervous system (PNS) diagnosis (B) during acute COVID-19 hospitalization for each pre-existing health condition (in the Elixhauser Comorbidity Index) by dividing the proportion of patients with the condition who developed a neurological diagnosis (CNS or PNS), by the number of patients without the condition who developed a neurological diagnosis. Pediatric patients were excluded from the analysis due to their low frequency of pre-admission health conditions. S3–S4 Tables provide detailed descriptions of the ICD codes comprising each component of the Elixhauser Comorbidity Index.

Fig 5. Pre-admission health conditions with the highest risk for a central nervous system diagnosis during acute COVID-19 hospitalization.

Each petal plot represents the normalized distribution (%) of patients with a pre-admission health condition (as components of the Elixhauser Comorbidity Index: e.g., neurological disorders, paralysis, or psychoses) at each healthcare system for each neurological status during acute COVID-19 hospitalization: no neurological condition (NNC), central nervous system (CNS) diagnosis, and peripheral nervous system (PNS) diagnosis. Each nested petal represents a healthcare system. The colors within each petal are sorted based on their value: the outermost color indicating the neurological status with the highest portion of patients and the innermost color indicating the neurological status with the lowest portion at each healthcare system. With n_k indicating the number of patients from healthcare system k for condition c, and n₁, n₂, and n₃ indicating the number of patients from the NNC, PNS, and CNS group, respectively, we summed patients at each system as n_k = n₁+n₂+n₃. Missing petals indicate no patient for the pre-admission health condition at a healthcare system (n_k = n₁ = n₂ = n₃ = 0). A petal containing only one color indicates that patients with a given pre-admission health condition c at that healthcare system all had the same neurological status during acute COVID-19 hospitalization (e.g., n_k = n₁ or n_k = n₂ or n_k = n₃). Using the pre-admission health condition ‘paralysis’ as an example, all patients with pre-admission paralysis at UCLA had a CNS diagnosis during acute COVID-19 hospitalization. Pediatric patients were excluded from the analysis due to low frequency of children with pre-admission health conditions.

Fig 6. Covariate-adjusted Kaplan-Meier survival analysis to evaluate the time to event of each clinical endpoint stratified by neurological status during acute COVID-19 hospitalizations in adults.

At each healthcare system, we estimated covariate-adjusted Kaplan-Meier time to event curves for each health outcome and neurological group. Specifically, for each outcome, we fit the Cox proportional hazards model to the patient cohort and estimated each patient’s survival rate, which is 1 minus event rate. Importantly, the survival rate was estimated for each patient by holding each patient’s covariates constant except for the neurological status. Thus, we estimated the survival rate for each neurological group, independent of the effect of additional covariates. Lastly, for each neurological group, we averaged all patients’ estimated survival rates to generate the overall survival curve. Survival curves from each healthcare system were combined using a random-effects meta-analysis weighted by the inverse of the variance derived at each participating healthcare system. Two healthcare systems (NUH and UKFR) were excluded from the meta-analysis due to their low frequency of neurological diagnoses (< 1% of adult patients). For the discharge outcome, we demarcated the median hospital stay in days for each neurological group. Due to a lower event rate for mortality, we demarcated the survival probability of the 90^th percentile for both the CNS and NNC groups. As the PNS group had <10% mortality, its survival probability was not demarcated. The table depicts the estimated total number of patients across all healthcare systems who were at risk at 0, 30, 60 and 90-day timepoints, where day 0 is the index date of the first COVID-19 hospitalization. Risk = the total number of patients who were still at risk for the event at a given time point. Event = total cumulative number of patients who experienced discharge or mortality by a given time point.