Long-term neurologic outcomes of COVID-19

The neurologic effects of acute COVID-19 are well-documented, but a comprehensive understanding of long-term neurologic sequelae remains incomplete. Previous studies have limitations, including small sample sizes, short follow-up periods (less than 6 months), focus on hospitalized patients, and narrow selection of neurologic outcomes. A comprehensive evaluation of long-term neurologic outcomes at 12 months post-infection, encompassing the full spectrum of acute COVID-19 severity (non-hospitalized, hospitalized, and ICU admission), is necessary to inform post-acute care strategies and healthcare resource allocation. This study addresses this gap by leveraging the extensive data available within the US Department of Veterans Affairs (VA) national healthcare databases. The VA system offers a uniquely comprehensive dataset, allowing for a longitudinal observational study with a large cohort of COVID-19 survivors and two control groups: a contemporary control group (VA users without COVID-19 during the pandemic) and a historical control group (VA users from 2017, pre-pandemic). The use of inverse probability weighting aims to balance the cohorts and provide robust estimates of the risks and burdens of a predefined set of neurologic outcomes at 12 months post-infection, stratified by the severity of the initial COVID-19 infection.

Existing literature on post-acute COVID-19 neurologic complications demonstrates a range of findings, mostly limited to hospitalized patients and short follow-up periods. Studies have highlighted increased risks of stroke, cognitive impairment, and other neurological disorders. However, a comprehensive picture including a broader spectrum of neurological conditions and considering non-hospitalized individuals across a longer time frame has been lacking. The present study aims to fill this critical knowledge gap by examining a much larger and more diverse population and extending the observation period to 12 months to assess the long-term burden of neurological complications after a COVID-19 infection.

This study used a longitudinal observational design using the US Department of Veterans Affairs (VA) healthcare databases. Three cohorts were created: a COVID-19 cohort (154,068 individuals), a contemporary control cohort (5,638,795 VA users without COVID-19 during the pandemic), and a historical control cohort (5,859,621 VA users from 2017, pre-pandemic). The date of the first positive COVID-19 test served as T0 for the COVID-19 cohort. For the control groups, a random T0 was assigned to match the distribution in the COVID-19 cohort. Follow-up ended on December 31, 2021 for the COVID-19 and contemporary control groups and December 31, 2019 for the historical control group. Individuals were included if they survived 30 days after T0. Inverse probability weighting (IPW) was used to adjust for potential confounding variables. The IPW was calculated using multinomial logistic regression to estimate the probability of group membership (COVID-19, contemporary control, historical control) given a set of covariates. Covariates included predefined variables (age, sex, race, etc.) and algorithmically selected high-dimensional covariates (from diagnoses, medications, and lab results). Stabilized weights were truncated at 30 to mitigate the effect of extreme weights. Cause-specific hazard models, treating death as a competing risk, were used to estimate hazard ratios (HRs) of incident neurologic outcomes. Burdens per 1,000 persons at 12 months were estimated, and excess burdens were calculated as the difference between COVID-19 and control groups. Subgroup analyses were conducted based on age, race, sex, obesity, smoking status, Area Deprivation Index (ADI), diabetes, chronic kidney disease, hyperlipidemia, hypertension, and immune dysfunction. Sensitivity analyses were conducted using different covariate sets (predefined only) and a doubly robust method (weighting and covariate adjustment). Positive and negative outcome controls and negative exposure controls were also employed to assess the robustness of the findings. Analyses were performed using SAS and R.

The study found a significantly increased risk of a wide range of neurologic disorders in individuals with COVID-19 compared to both contemporary and historical control groups. These included: * **Cerebrovascular disorders:** Increased risk of ischemic stroke (HR 1.50), transient ischemic attacks (TIAs) (HR 1.62), hemorrhagic stroke (HR 2.19), and cerebral venous thrombosis (HR 2.69). * **Cognition and memory disorders:** Increased risk of memory problems (HR 1.77) and Alzheimer's disease (HR 2.03). * **Disorders of peripheral nerves:** Increased risk of peripheral neuropathy (HR 1.34), paresthesia (HR 1.32), dysautonomia (HR 1.30), and Bell's palsy (HR 1.48). * **Episodic disorders:** Increased risk of migraine (HR 1.21), epilepsy and seizures (HR 1.80), and headache disorders (HR 1.35). * **Extrapyramidal and movement disorders:** Increased risk of abnormal involuntary movements (HR 1.41), tremor (HR 1.37), Parkinson-like disease (HR 1.50), dystonia (HR 1.57), and myoclonus (HR 1.42). * **Mental health disorders:** Increased risk of major depressive disorders (HR 1.44), stress/adjustment disorders (HR 1.39), anxiety disorders (HR 1.38), and psychotic disorders (HR 1.51). * **Musculoskeletal disorders:** Increased risk of joint pain (HR 1.34), myalgia (HR 1.83), and myopathy (HR 2.76). * **Sensory disorders:** Increased risk of hearing abnormalities/tinnitus (HR 1.22), vision abnormalities (HR 1.30), loss of smell (HR 4.05), and loss of taste (HR 2.26). * **Other neurologic disorders:** Increased risk of dizziness (HR 1.44), somnolence (HR 1.67), Guillain-Barré syndrome (HR 2.16), encephalitis/encephalopathy (HR 1.82), and transverse myelitis (HR 1.49). The overall hazard ratio for any neurologic sequela was 1.42 (95% CI 1.38, 1.47), with a burden of 70.69 (95% CI 63.54, 78.01) per 1,000 persons at 12 months. These increased risks were observed across all subgroups examined (age, race, sex, etc.) and even in individuals who did not require hospitalization during the acute phase of COVID-19. The risks increased with the severity of acute infection (non-hospitalized < hospitalized < ICU admission). Findings were consistent across comparisons with both contemporary and historical control groups and were robust in sensitivity analyses.

This large-scale study provides strong evidence for a substantial and persistent increased risk of various neurologic disorders following COVID-19 infection. The findings extend previous research by including a broader range of neurologic outcomes, a longer follow-up period, and consideration of non-hospitalized individuals. The consistency of results across different control groups and sensitivity analyses strengthens the conclusions. The increased risk across all subgroups underscores the widespread impact of COVID-19 on neurological health. The significant burden of these long-term neurological complications has considerable implications for healthcare systems and public health policies. These findings highlight the need for ongoing surveillance, improved post-acute care strategies, and further research to understand the underlying mechanisms of these neurological sequelae and potential preventative or mitigative interventions. The substantial long-term impact on individual lives and healthcare systems necessitates proactive responses.

This study demonstrates a significantly increased risk of a wide array of neurologic disorders lasting up to 12 months post-COVID-19 infection, even in non-hospitalized individuals. The findings highlight the considerable long-term neurological burden of COVID-19 and underscore the need for improved healthcare systems planning and public health policies to address these challenges. Further research should focus on the underlying mechanisms, preventative strategies, and effective treatments for these long-term neurological complications.

The study's primary limitations stem from its cohort composition, which was predominantly White and male, potentially limiting the generalizability of findings to other populations. While extensive adjustments were made to minimize bias, the possibility of residual confounding or misclassification bias cannot be entirely excluded. The contemporary control group may have included some individuals with undetected COVID-19 infections, potentially leading to underestimation of the true risk. The analysis was conducted before widespread vaccine availability, so the results might not fully reflect the impact of vaccination on long-term neurological outcomes.