Long-term neurologic outcomes of COVID-19

Long COVID, the umbrella term describing postacute sequelae following SARS-CoV-2 infection, can involve extrapulmonary organ dysfunction including neurologic abnormalities. Prior studies of postacute clinical neurologic disorders largely focused on hospitalized patients, had follow-up durations under 6 months and evaluated a narrow set of outcomes. A comprehensive 12-month evaluation across the care-setting spectrum (nonhospitalized, hospitalized and ICU) had not been undertaken. Addressing this gap is important for guiding postacute care and healthcare system capacity planning. The authors leveraged US Department of Veterans Affairs national healthcare databases to assemble a cohort of 154,068 individuals who survived the first 30 days of COVID-19, alongside large contemporary (n=5,638,795) and historical (n=5,859,621) control groups, to estimate 12-month risks and burdens of prespecified neurologic outcomes overall and by acute care setting, using inverse probability weighting to balance cohorts.

Previous literature established neurologic and psychiatric sequelae within 6 months after COVID-19 but was limited by shorter follow-up, focus on hospitalized populations, and restricted outcome sets. Studies documenting neurologic abnormalities and Long COVID features suggested potential risks but lacked comprehensive 12-month, care-setting–stratified analyses. This study builds on those reports by extending follow-up to 12 months, expanding the neurologic outcome spectrum and including both nonhospitalized and critically ill populations, with comparisons to both contemporary and prepandemic cohorts.

Design and setting: Longitudinal observational cohort study using the US Department of Veterans Affairs (VHA) national electronic healthcare databases (Corporate Data Warehouse). The VHA operates 1,255 healthcare facilities nationwide. Ethics approval was obtained with waiver of informed consent (protocol 1606333). Cohorts: COVID-19 cohort included VHA users in 2019 with a first positive SARS-CoV-2 test between March 1, 2020 and January 15, 2021 (n=169,476); those alive 30 days after the positive test were included (n=154,068). T0 was date of first positive test; follow-up ended December 31, 2021. Contemporary control cohort comprised 2019 VHA users alive on March 1, 2020 not in the COVID-19 group (n=5,809,137); T0 was randomly assigned to match the COVID-19 cohort’s T0 distribution; those alive 30 days after T0 formed the final cohort (n=5,638,795). Historical control cohort comprised 2017 VHA users alive on March 1, 2018 (n=6,009,794 not in COVID-19 group); T0 assigned using the COVID-19 T0 distribution shifted by −730 days; those alive 30 days after T0 formed the final historical cohort (n=5,859,621); follow-up ended December 31, 2019. Data sources: Demographics from CDW Patient domain; outpatient and inpatient encounters; pharmacy dispensing; barcode medication administration; laboratory results; COVID-19 Shared Data Resource; and Area Deprivation Index (ADI) as a contextual socioeconomic measure. Outcomes: Prespecified incident neurologic outcomes defined using ICD-10 codes, assessed from 30 days after T0 to end of follow-up among participants without the outcome in the prior year. Individual outcomes were grouped into composites: cerebrovascular disorders (ischemic stroke, TIA, hemorrhagic stroke, cerebral venous thrombosis), cognition and memory (memory problems, Alzheimer’s disease), peripheral nerve disorders (peripheral neuropathy, paresthesia, dysautonomia, Bell’s palsy), episodic disorders (migraine, epilepsy/seizures, headache), extrapyramidal and movement disorders (abnormal involuntary movements, tremor, Parkinson-like disease, dystonia, myoclonus), mental health disorders (major depressive, stress/adjustment, anxiety, psychotic disorders), musculoskeletal disorders (joint pain, myalgia, myopathy), sensory disorders (hearing/tinnitus, vision abnormalities, loss of smell, loss of taste), and other neurologic/related disorders (dizziness, somnolence, Guillain-Barré syndrome, encephalitis/encephalopathy, transverse myelitis). Exposure stratification: Analyses conducted overall and by acute infection care setting: nonhospitalized (n=131,915), hospitalized (n=16,764), ICU (n=5,389). Covariates: Dual approach including predefined covariates (age; race [White, Black, other]; sex; ADI; BMI; smoking [current, former, never]; healthcare utilization; comorbidities including cancer, CKD, chronic lung disease, diabetes, hypertension; eGFR; systolic/diastolic BP; continuous covariates modeled with restricted cubic splines) and algorithmically selected high-dimensional covariates from diagnoses (540 groups), medications (543 types), and laboratory abnormalities (62), selecting variables present in ≥100 participants per exposure group; the top 100 variables by univariate relative risk were retained per outcome-specific cohort. Weighting and balance: Multinomial logistic regression estimated probabilities of group membership conditional on covariates; stabilized inverse probability weights P(group)/P(group|L) were calculated, truncated at 30 (<0.001% truncated). Covariate balance assessed using standardized mean differences (good balance achieved). Statistical analysis: Cause-specific hazard models accounting for death as a competing risk estimated hazard ratios (HRs) comparing COVID-19 with contemporary and historical controls after weighting. Burdens per 1,000 persons at 12 months were derived from survival probabilities within groups; excess burden was the difference between COVID-19 and control burdens. Subgroup analyses by age, race, sex, obesity, smoking, ADI, diabetes, CKD, hyperlipidemia, hypertension, immune dysfunction; age modeled as a spline with interaction tests for effect modification. Care setting analyses estimated HRs and burdens within nonhospitalized, hospitalized, and ICU groups vs controls. Sensitivity analyses: (1) Weighting using only predefined covariates (no high-dimensional variables). (2) Doubly robust models applying both weighting and covariate adjustment. Controls: Positive outcome control (fatigue) and negative-outcome controls (three outcomes with no expected association) and negative-exposure controls (influenza vaccination on odd vs even calendar days). Variance estimation used robust sandwich estimators. Statistical significance defined by 95% CI excluding 1. Analyses conducted in SAS EG v8.2 with visualization in R v4.0.4.

• Overall cohort sizes and follow-up: COVID-19 (n=154,068), contemporary controls (n=5,638,795), historical controls (n=5,859,621); median follow-up ~408–409 days; total 14,064,985 person-years. - Any neurologic outcome: HR 1.42 (95% CI 1.38, 1.47); excess burden 70.69 (63.54, 78.01) per 1,000 persons at 12 months. - Cerebrovascular disorders: Composite HR 1.56 (1.48, 1.64); burden 4.92 (4.26, 5.62) per 1,000. • Ischemic stroke HR 1.50 (1.41, 1.61); burden 3.40 (2.75, 4.09). • TIA HR 1.62 (1.50, 1.75); burden 2.03 (1.64, 2.46). • Hemorrhagic stroke HR 2.19 (1.63, 2.95); burden 0.21 (0.11, 0.35). • Cerebral venous thrombosis HR 2.69 (1.29, 5.62); burden 0.05 (0.01, 0.14). - Cognition and memory disorders: Composite HR 1.80 (1.71, 1.88); burden 10.35 (9.27, 11.47). • Memory problems HR 1.77 (1.68, 1.85); burden 10.07 (9.00, 11.20). • Alzheimer’s disease HR 2.03 (1.79, 2.31); burden 1.65 (1.27, 2.10). - Disorders of peripheral nerves: Composite HR 1.34 (1.29, 1.39); burden 8.64 (7.44, 9.87). • Peripheral neuropathy HR 1.34 (1.28, 1.40); burden 5.64 (4.67, 6.65). • Paresthesia HR 1.32 (1.25, 1.39); burden 2.89 (2.27, 3.55). • Dysautonomia HR 1.30 (1.21, 1.40); burden 1.60 (1.12, 2.12). • Bell’s palsy HR 1.48 (1.24, 1.77); burden 0.32 (0.16, 0.51). - Episodic disorders: Composite HR 1.32 (1.26, 1.39); burden 4.75 (3.79, 5.76). • Migraine HR 1.21 (1.14, 1.28); burden 2.04 (1.36, 2.76). • Epilepsy and seizures HR 1.80 (1.61, 2.01); burden 2.01 (1.47, 2.63). • Headache disorders HR 1.35 (1.25, 1.45); burden 1.46 (1.06, 1.89). - Extrapyramidal and movement disorders: Composite HR 1.42 (1.34, 1.50); burden 3.98 (3.24, 4.77). • Abnormal involuntary movements HR 1.41 (1.32, 1.50); burden 2.85 (2.24, 3.49). • Tremor HR 1.37 (1.25, 1.51); burden 1.10 (0.73, 1.51). • Parkinson-like disease HR 1.50 (1.28, 1.75); burden 0.89 (0.50, 1.34). • Dystonia HR 1.57 (1.29, 1.90); burden 0.40 (0.21, 0.63). • Myoclonus HR 1.42 (1.13, 1.79); burden 0.14 (0.04, 0.26). - Mental health disorders: Composite HR 1.43 (1.38, 1.47); burden 25.00 (22.40, 27.69). • Major depressive disorders HR 1.44 (1.39, 1.48); burden 17.28 (15.43, 19.18). • Stress/adjustment disorders HR 1.39 (1.34, 1.44); burden 14.34 (12.66, 16.07). • Anxiety disorders HR 1.38 (1.33, 1.42); burden 12.44 (10.93, 13.99). • Psychotic disorders HR 1.51 (1.33, 1.71); burden 1.02 (0.66, 1.43). - Musculoskeletal disorders: Composite HR 1.45 (1.42, 1.48); burden 40.09 (37.22, 43.01). • Joint pain HR 1.34 (1.31, 1.38); burden 27.65 (25.01, 30.35). • Myalgia HR 1.83 (1.77, 1.90); burden 15.97 (14.75, 17.23). • Myopathy HR 2.76 (2.30, 3.32); burden 0.71 (0.52, 0.93). - Sensory disorders: Composite HR 1.25 (1.22, 1.28); burden 17.03 (14.85, 19.26). • Hearing abnormalities or tinnitus HR 1.22 (1.18, 1.25); burden 11.87 (10.05, 13.75). • Vision abnormalities HR 1.30 (1.24, 1.36); burden 5.59 (4.55, 6.68). • Loss of smell HR 4.05 (3.45, 4.75); burden 1.07 (0.86, 1.32). • Loss of taste HR 2.26 (1.54, 3.32); burden 0.11 (0.05, 0.21). - Other neurologic or related disorders: Composite HR 1.46 (1.40, 1.52); burden 7.37 (6.41, 8.38). • Dizziness HR 1.44 (1.38, 1.50); burden 6.65 (5.72, 7.61). • Somnolence HR 1.67 (1.31, 2.12); burden 0.56 (0.26, 0.94). • Guillain-Barré syndrome HR 2.16 (1.40, 3.35); burden 0.11 (0.04, 0.22). • Encephalitis or encephalopathy HR 1.82 (1.16, 2.84); burden 0.07 (0.01, 0.16). • Transverse myelitis HR 1.49 (1.11, 2.00); burden 0.03 (0.00, 0.11). - Care-setting gradient: Elevated risks and burdens were observed even among nonhospitalized individuals, with progressively higher risks from nonhospitalized to hospitalized to ICU patients. - Subgroups and age interactions: Increased risks evident across age, race, sex, obesity, smoking, ADI, diabetes, CKD, hyperlipidemia, hypertension, and immune dysfunction subgroups. Age interactions indicated increasing risks with age for episodic, mental health, musculoskeletal disorders, and any neurologic disorder, and decreasing risks with age for cognition/memory, sensory, and other neurologic disorders. - Robustness: Findings consistent when using historical controls; across sensitivity analyses (predefined covariates only; doubly robust models); and supported by positive outcome control (fatigue). No significant associations in negative-outcome and negative-exposure controls.

The study demonstrates that beyond the acute phase, COVID-19 is associated with significantly increased 12-month risks and absolute burdens of a wide range of neurologic disorders, including cerebrovascular events, cognitive and memory problems, peripheral nerve and movement disorders, episodic conditions, mental health disorders, musculoskeletal and sensory disorders, and rare conditions such as Guillain-Barré syndrome and encephalitis/encephalopathy. Increased risks were present even among nonhospitalized patients and scaled with acute disease severity, indicating that long-term neurologic sequelae are not confined to those with severe initial illness. The associations were consistent across demographic and clinical subgroups and robust across multiple analytic strategies and control groups. These findings underscore the potential for substantial population-level burden and highlight implications for healthcare capacity planning and postacute care pathways. Potential biological mechanisms include persistent viral components and immune activation, autoimmunity, dysbiosis, endothelial injury, complement activation with coagulopathy and microangiopathy, and neuroinflammatory processes affecting neural cells, synaptic signaling, neurogenesis, and myelination. Age interactions suggest differing vulnerabilities: stronger effects of cognitive/memory, sensory, and other neurologic disorders in younger adults, and stronger effects on mental health, musculoskeletal, and episodic disorders in older adults.

This comprehensive 12-month evaluation shows that COVID-19 confers increased risk and burden of diverse neurologic disorders across the care-setting spectrum and subgroups, including nonhospitalized individuals. The results call for integrating long-term neurologic sequelae into public health policy and healthcare planning, with emphasis on prevention of infection and development of care pathways for survivors. Future research should clarify underlying mechanisms, identify at-risk populations, evaluate preventive and therapeutic interventions to mitigate long-term neurologic complications, and monitor how evolving variants, vaccination, and treatments influence the epidemiology of postacute neurologic outcomes.

Generalizability may be limited due to the predominantly White male veteran population. Although extensive covariate adjustment and validated outcome definitions were used, residual confounding and misclassification cannot be excluded. Some contemporary controls may have had undiagnosed SARS-CoV-2 infection, potentially biasing estimates toward the null and underestimating risks. Inverse probability weighting results can be sensitive to model specifications, though multiple sensitivity analyses yielded consistent findings. Because enrollment occurred before widespread vaccination in the US (<1% vaccinated before T0), results may differ in vaccinated populations and with new variants or therapeutics. Subgroup analyses estimate risks within groups but are not necessarily comparable across groups. The pandemic’s dynamic nature may alter long-term sequelae over time.