Multinational characterization of neurological phenotypes in patients hospitalized with COVID-19

Neurological complications worsen outcomes in COVID-19. To define the prevalence of neurological conditions among hospitalized patients with a positive SARS-CoV-2 reverse transcription polymerase chain reaction test in geographically diverse multinational populations during early pandemic, we used electronic health records (EHR) from 338 participating hospitals across 6 countries and 3 continents (January–September 2020) for a cross-sectional analysis. We assessed the frequency of International Classification of Disease code of neurological conditions by countries, healthcare systems, time before and after admission for COVID-19 and COVID-19 severity. Among 35,177 hospitalized patients with SARS-CoV-2 infection, there was an increase in the proportion with disorders of consciousness (5.8%, 95% confidence interval [CI] 3.7–7.8%, pFDR < 0.001) and unspecified disorders of the brain (8.1%, 5.7–10.5%, pFDR < 0.001) when compared to the pre-admission proportion. During hospitalization, the relative risk of disorders of consciousness (22%, 19–25%), cerebrovascular diseases (24%, 13–35%), nontraumatic intracranial hemorrhage (34%, 20–50%), encephalitis and/or myelitis (37%, 17–60%) and myopathy (72%, 67–77%) were higher for patients with severe COVID-19 when compared to those who never experienced severe COVID-19. Leveraging a multinational network to capture standardized EHR data, we highlighted the increased prevalence of central and peripheral neurological phenotypes in patients hospitalized with COVID-19, particularly among those with severe disease.


Supplementary Statistical Methods and Equations
For each neurological ICD code (3 alphanumeric characters), we reported the total count across all contributing healthcare systems and all countries (Y) as well as the proportion of patients hospitalized with COVID-19 who had each ICD code at each healthcare system (and each country), both before admission (propbefore) and after admission (propafter) (eEq. 1).
We next compared the prevalence of each neurological ICD code (first three characters before decimal point) and disease category by contributing healthcare system, before and after admission date, between patients who ever met the 4CE criteria of severe COVID-19 (see main text) and those who did not. For each ICD code X, we computed the expected number of ever-severe patient cases D-with: where ? denotes the observed number of never-severe patient cases with code X, and ? denotes the observed number of ever-severe patient cases with code X (eEq. represents the proportion of patients without severe disease but with neurological code X. We then performed an enrichment analysis to examine the difference in proportions of ever-severe disease across neurological ICD codes. Specifically, we calculated the enrichment of each neurological ICD code by dividing the observed number of severe cases by the expected number of severe cases and reported a value of log2 enrichment (LOE) and its 95% confidence interval (eEq. 4).

LOE =
We estimated the 95% confidence interval of the LOE using the Poisson model method. Finally, we computed the p values using Fisher's exact test and corrected for multiple hypothesis testing with Benjamini-Hochberg's false discovery rate (FDR) procedure. We considered a result with pFDR<0.05 statistically significant.

Exploratory Analysis of ICD-9 Data
We analyzed separately the neurological phenotypes among the subset of the patients with ICD-9 codes (eTable 2), as a minority of the 4CE contributing healthcare systems in the USA (7 healthcare systems partially submitting ICD-9 codes related to procedure-related ICD-9 codes) and Italy (5 healthcare systems exclusively submitting ICD-9 codes) reported neurological ICD-9 codes (eTable 3). Because 7 of these contributing healthcare systems reported both ICD-9 and ICD-10 data and given the aggregate data format, we could not ascertain the exact number of patients with ICD-9 data. We separated this analysis from the main ICD-10 analysis because one-to-one mapping from ICD-9 to ICD-10 codes was not available for all codes.
As with the ICD-10 data, there was increased prevalence of "disorders of consciousness and other neurological conditions" in patients after admission date when compared to before admission date (eFig. 2, eFig. 3). These differences appear to be driven by the contributing healthcare systems in the USA. However, there was no statistically significant difference when examining the change in prevalence of individual neurological conditions after admission date. The smaller sample size might explain the difference from the ICD-10 data results.
Finally, there was a significantly lower proportion of patients with "other and ill-defined cerebrovascular disease" (ICD-9 437: RDDafter=-42%) among patients with severe disease in the period after admission date. However, we must interpret these findings with great caution given the limitations of the ICD-9 data and the inconsistency with the larger sample size of the ICD-10 data. Note: When aggregating the summary statistics at a site, counts below a certain obfuscation threshold (see healthcare system-specific obfuscation parameters) are masked as 0 to preserve system-specific privacy and to reduce the risk of patient re-identification. eTable 2. Neurological disease categories, corresponding ICD-9 codes, and their descriptions for the exploratory analysis. We note that for the four 3-character neurological ICD-9 codes that did not have a one-to-one mapping to corresponding ICD-10 codes, we manually grouped them into different neurological groups: V41: Problems with special senses and other special functions (Vision/smell/taste) eTable 3. Statistically significant associations of neurological conditions (ICD-9 codes) after admission and severe disease status (pFDR < 0.05). Log2 enrichment (LOE) and 95% confidence interval for each ICD-10 code (left) and the absolute difference between the observed (▴) and expected (•) number of severe cases (right) after admission. A purple positive LOE value for an ICD-9 code indicates a statistically significantly higher proportion of severe cases with the given ICD-10 code when compared to the never-severe cases. Conversely, a green negative LOE value indicates a statistically significantly lower proportion of severe cases with the given ICD-10 code compared to the neversevere cases. Neurological ICD-10 codes are ordered based on the expected number of severe cases after admission date across all sites. The results are generally consistent between the US sites and the non-US sites, except for the following: (1) ICD-10 code R43 (Disturbances of smell and taste) displays opposite directions between the US sites (higher proportion of severe cases with R43) and the non-US sites (lower proportion of severe cases with R43); (2) The US sites have several significant findings that are not significant among the non-US sites, largely due to the smaller number of sites outside the US. Overall, the findings from the subgroup analyses between US and non-US sites are consistent with the findings from the pooled analysis (Fig. 4, main text). eFigure 3. Prevalence of each ICD-9 code by site and country before and after admission date. Sites refer to the contributing healthcare systems.

eFigure 4. Prevalence of neurological phenotypes among all patients by ICD-9 code.
Sites refer to the contributing healthcare systems.
(A) Difference in prevalence of each neurological ICD-9 code by site and country, calculated as after admission date -before admission date (Eq. 2). The absolute values of prevalence are displayed in eFig. 3. (B) Per ICD-9 code, total counts of patients at all sites (left) and average proportion of patients (right) before and after admission date. Mean prevalence estimates across sites are shown as circles and their 95% confidence intervals as bars. ICD-9 codes are ordered based on the mean prevalence difference between before and after admission date.