CoChat
Sign Up
Cited by 2,189Long-Term Effects of COVID-19COVID-19 Clinical Research StudiesCOVID-19 and Mental Health

6-month neurological and psychiatric outcomes in 236 379 survivors of COVID-19: a retrospective cohort study using electronic health records

Open access
Taquet, Maxime, Geddes, John R, Husain, Masud, Luciano, Sierra, Harrison, Paul J·2021·The Lancet Psychiatry
DOI 10.1016/s2215-0366(21)00084-5PMC PMC8023694

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

BACKGROUND: Neurological and psychiatric sequelae of COVID-19 have been reported, but more data are needed to adequately assess the effects of COVID-19 on brain health. We aimed to provide robust estimates of incidence rates and relative risks of neurological and psychiatric diagnoses in patients in the 6 months following a COVID-19 diagnosis. METHODS: For this retrospective cohort study and time-to-event analysis, we used data obtained from the TriNetX electronic health records network (with over 81 million patients). Our primary cohort comprised patients who had a COVID-19 diagnosis; one matched control cohort included patients diagnosed with influenza, and the other matched control cohort included patients diagnosed with any respiratory tract infection including influenza in the same period. Patients with a diagnosis of COVID-19 or a positive test for SARS-CoV-2 were excluded from the control cohorts. All cohorts included patients older than 10 years who had an index event on or after Jan 20, 2020, and who were still alive on Dec 13, 2020. We estimated the incidence of 14 neurological and psychiatric outcomes in the 6 months after a confirmed diagnosis of COVID-19: intracranial haemorrhage; ischaemic stroke; parkinsonism; Guillain-Barré syndrome; nerve, nerve root, and plexus disorders; myoneural junction and muscle disease; encephalitis; dementia; psychotic, mood, and anxiety disorders (grouped and separately); substance use disorder; and insomnia. Using a Cox model, we compared incidences with those in propensity score-matched cohorts of patients with influenza or other respiratory tract infections. We investigated how these estimates were affected by COVID-19 severity, as proxied by hospitalisation, intensive therapy unit (ITU) admission, and encephalopathy (delirium and related disorders). We assessed the robustness of the differences in outcomes between cohorts by repeating the analysis in different scenarios. To provide benchmarking for the incidence and risk of neurological and psychiatric sequelae, we compared our primary cohort with four cohorts of patients diagnosed in the same period with additional index events: skin infection, urolithiasis, fracture of a large bone, and pulmonary embolism. FINDINGS: Among 236 379 patients diagnosed with COVID-19, the estimated incidence of a neurological or psychiatric diagnosis in the following 6 months was 33·62% (95% CI 33·17-34·07), with 12·84% (12·36-13·33) receiving their first such diagnosis. For patients who had been admitted to an ITU, the estimated incidence of a diagnosis was 46·42% (44·78-48·09) and for a first diagnosis was 25·79% (23·50-28·25). Regarding individual diagnoses of the study outcomes, the whole COVID-19 cohort had estimated incidences of 0·56% (0·50-0·63) for intracranial haemorrhage, 2·10% (1·97-2·23) for ischaemic stroke, 0·11% (0·08-0·14) for parkinsonism, 0·67% (0·59-0·75) for dementia, 17·39% (17·04-17·74) for anxiety disorder, and 1·40% (1·30-1·51) for psychotic disorder, among others. In the group with ITU admission, estimated incidences were 2·66% (2·24-3·16) for intracranial haemorrhage, 6·92% (6·17-7·76) for ischaemic stroke, 0·26% (0·15-0·45) for parkinsonism, 1·74% (1·31-2·30) for dementia, 19·15% (17·90-20·48) for anxiety disorder, and 2·77% (2·31-3·33) for psychotic disorder. Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza (hazard ratio [HR] 1·44, 95% CI 1·40-1·47, for any diagnosis; 1·78, 1·68-1·89, for any first diagnosis) and those who had other respiratory tract infections (1·16, 1·14-1·17, for any diagnosis; 1·32, 1·27-1·36, for any first diagnosis). As with incidences, HRs were higher in patients who had more severe COVID-19 (eg, those admitted to ITU compared with those who were not: 1·58, 1·50-1·67, for any diagnosis; 2·87, 2·45-3·35, for any first diagnosis). Results were robust to various sensitivity analyses and benchmarking against the four additional index health events. INTERPRETATION: Our study provides evidence for substantial neurological and psychiatric morbidity in the 6 months after COVID-19 infection. Risks were greatest in, but not limited to, patients who had severe COVID-19. This information could help in service planning and identification of research priorities. Complementary study designs, including prospective cohorts, are needed to corroborate and explain these findings. FUNDING: National Institute for Health Research (NIHR) Oxford Health Biomedical Research Centre.

AI snapshot

Written from the paper’s full text.

TL;DR

In a retrospective cohort of 236,379 COVID-19 survivors, 33.62% received a neurological or psychiatric diagnosis within 6 months, and 12.84% received a first such diagnosis. Risks were higher than after influenza or other respiratory tract infections and were greatest in patients with more severe COVID-19, including ITU admission or encephalopathy.

Key findings

  • Among 236,379 patients with COVID-19, the estimated 6-month incidence of any neurological or psychiatric diagnosis was 33.62% (95% CI 33.17–34.07), and of any first diagnosis was 12.84% (12.36–13.33). For ITU-admitted patients, these were 46.42% (44.78–48.09) and 25.79% (23.50–28.25); for patients with encephalopathy, 62.34% (60.14–64.55) and 31.13% (27.29–35.36).
  • Individual 6-month incidences in the whole COVID-19 cohort included anxiety disorder 17.39% (17.04–17.74), mood disorder 13.66% (13.35–13.99), psychotic disorder 1.40% (1.30–1.51), ischaemic stroke 2.10% (1.97–2.23), intracranial haemorrhage 0.56% (0.50–0.63), dementia 0.67% (0.59–0.75), and parkinsonism 0.11% (0.08–0.14).
  • Compared with influenza, COVID-19 was associated with higher risk of any diagnosis (HR 1.44, 95% CI 1.40–1.47) and any first diagnosis (1.78, 1.68–1.89); compared with other respiratory tract infections, HRs were 1.16 (1.14–1.17) and 1.32 (1.27–1.36), respectively. ITU admission versus no ITU admission was associated with HR 1.58 (1.50–1.67) for any diagnosis and 2.87 (2.45–3.35) for any first diagnosis.
  • Most diagnostic categories were more common after COVID-19 than after influenza, except parkinsonism and Guillain-Barré syndrome, for which HRs were not significantly greater than 1 versus influenza.

Methods

  • Retrospective cohort and time-to-event analysis using the TriNetX electronic health records network (over 81 million patients, 62 health-care organisations, primarily USA). The primary cohort had confirmed COVID-19 (ICD-10 U07.1); control cohorts had influenza or any respiratory tract infection including influenza, with COVID-19/SARS-CoV-2-positive patients excluded. All cohorts included patients older than 10 years with an index event on or after Jan 20, 2020, alive on Dec 13, 2020.
  • Fourteen neurological and psychiatric outcomes were assessed 1–180 days after the index event, with first versus any diagnoses estimated depending on whether outcomes were chronic or recurrent. Cohorts were propensity-score matched 1:1 on 55 variables; incidence was estimated with Kaplan-Meier, comparisons with log-rank tests, and hazard ratios with Cox proportional hazards models.
  • Secondary analyses examined COVID-19 severity via hospitalisation, ITU admission, and encephalopathy, with matched comparisons; sensitivity analyses included patients who died, test-confirmed COVID-19, and pre-pandemic influenza comparisons, plus benchmarking against four additional index events.

Limitations

  • The paper states limitations inherent to electronic health records: unknown completeness of records, no validation of diagnoses, and sparse information on socioeconomic and lifestyle factors. It also notes that the choice of comparator cohort influenced hazard ratio magnitude, that encephalopathy was likely under-recorded and results should not be generalised to all patients with delirium, that timing differences and undiagnosed COVID-19 in controls likely made HRs underestimates, and that the study can only show associations, not causality.
  • Evident from the text: despite propensity-score matching, residual confounding may remain; diagnoses rely on ICD-10 codes; and follow-up is limited to 6 months, so longer-term trajectories and duration of risk are not established.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/s2215-0366(21)00084-5

Cites

31 of 31 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 10 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Among patients with COVID-19 who did not require hospitalisation, compared with patients who had other respiratory tract infections, the hazard ratio for any first diagnosis was 1.28 (95% CI 1.23-1.33).

    “For a first diagnosis, the HRs were 1·83 (1·71–1·96) versus patients who had influenza and 1·28 (1·23–1·33) versus those who had other respiratory tract infections.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
    • Rests on an assumption: The propensity score matching adequately controls for confounding by measured covariates.
    • Rests on an assumption: The comparator cohorts (influenza and other respiratory tract infections) are appropriate to estimate the excess risk attributable to COVID-19.
  • Findingown data

    In matched cohorts of patients with COVID-19, the overall hazard ratio for any first diagnosis was 2.87 (95% CI 2.45-3.35) for those requiring ITU admission versus those not requiring ITU admission.

    “with a HR of 1·58 (1·50–1·67) for any diagnosis and 2·87 (2·45–3·35) for any first diagnosis (table 5, appendix p 22).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
    • Rests on an assumption: The propensity score matching adequately controls for confounding by measured covariates.
  • Findingown data

    Among 236 379 patients diagnosed with COVID-19, the estimated incidence of a neurological or psychiatric diagnosis in the following 6 months was 33.62% (95% CI 33.17-34.07).

    “Among 236 379 patients diagnosed with COVID-19, the estimated incidence of a neurological or psychiatric diagnosis in the following 6 months was 33·62% (95% CI 33·17–34·07), with 12·84% (12·36–13·33) receiving their first such diagnosis.”
    • Also stated in: § Results · plain
      “In the whole cohort, 33·62% (95% CI 33·17–34·07) of patients received a diagnosis (table 2).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
  • Findingown data

    In the ITU admission group, the estimated incidence of ischaemic stroke in the 6 months after COVID-19 diagnosis was 6.92% (95% CI 6.17-7.76).

    “In the group with ITU admission, estimated incidences were 2·66% (2·24–3·16) for intracranial haemorrhage, 6·92% (6·17–7·76) for ischaemic stroke, 0·26% (0·15–0·45) for parkinsonism, 1·74% (1·31–2·30) for dementia, 19·15% (17·90–20·48) for anxiety disorder, and 2·77% (2·31–3·33) for psychotic disorder.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
  • Findingown data

    In matched cohorts of patients with COVID-19, the overall hazard ratio for any first diagnosis was 1.70 (95% CI 1.56-1.86) for those requiring hospitalisation versus those not requiring hospitalisation.

    “This comparison showed greater hazard rates for all outcomes in the hospitalised group than in the non-hospitalised group, except for nerve, nerve root, or plexus disorders (table 5, figure 2), with an overall HR of 1·33 (1·29–1·37) for any diagnosis and 1·70 (1·56–1·86) for any first diagnosis.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
    • Rests on an assumption: The propensity score matching adequately controls for confounding by measured covariates.
  • May combine two claims. 24
  • Classified as probably not a substantial claim. 10

  • May combine two claims.

    Findingown data

    Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza, with HR 1.78 (95% CI 1.68-1.89) for any first diagnosis.

    “Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza (hazard ratio [HR] 1·44, 95% CI 1·40–1·47, for any diagnosis; 1·78, 1·68–1·89, for any first diagnosis) and those who had other respiratory tract infections (1·16, 1·14–1·17, for any diagnosis; 1·32, 1·27–1·36, for any first diagnosis).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
    • Rests on an assumption: Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

  • May combine two claims.

    Findingown data

    HRs were significantly greater than 1 for all diagnoses for patients who had COVID-19 compared with those who had influenza, except for parkinsonism and Guillain-Barré syndrome.

    “HRs were significantly greater than 1 for all diagnoses for patients who had COVID-19 compared with those who had influenza, except for parkinsonism and Guillain-Barré syndrome, and significantly greater than 1 for all diagnoses compared with patients who had respiratory tract infections (table 3).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
    • Rests on an assumption: Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

  • May combine two claims.

    Findingown data

    Most diagnostic categories were more common in patients who had COVID-19 than in those who had other respiratory tract infections, with HR 1.16 (95% CI 1.14-1.17) for any diagnosis.

    “Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza (hazard ratio [HR] 1·44, 95% CI 1·40–1·47, for any diagnosis; 1·78, 1·68–1·89, for any first diagnosis) and those who had other respiratory tract infections (1·16, 1·14–1·17, for any diagnosis; 1·32, 1·27–1·36, for any first diagnosis).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
    • Rests on an assumption: Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The COVID-19 cohort included 46 302 patients who were hospitalised.

    “The COVID-19 cohort was divided into subgroups of patients who were not hospitalised (190 077 patients), those who were hospitalised (46 302 patients), those who required ITU admission (8945 patients), and those who received a diagnosis of encephalopathy (6229 patients).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The COVID-19 cohort included 8945 patients who required ITU admission.

    “The COVID-19 cohort was divided into subgroups of patients who were not hospitalised (190 077 patients), those who were hospitalised (46 302 patients), those who required ITU admission (8945 patients), and those who received a diagnosis of encephalopathy (6229 patients).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Most diagnostic categories were more common in patients who had COVID-19 than in those who had other respiratory tract infections, with HR 1.32 (95% CI 1.27-1.36) for any first diagnosis.

    “Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza (hazard ratio [HR] 1·44, 95% CI 1·40–1·47, for any diagnosis; 1·78, 1·68–1·89, for any first diagnosis) and those who had other respiratory tract infections (1·16, 1·14–1·17, for any diagnosis; 1·32, 1·27–1·36, for any first diagnosis).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
    • Rests on an assumption: Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The COVID-19 cohort included 6229 patients who received a diagnosis of encephalopathy.

    “The COVID-19 cohort was divided into subgroups of patients who were not hospitalised (190 077 patients), those who were hospitalised (46 302 patients), those who required ITU admission (8945 patients), and those who received a diagnosis of encephalopathy (6229 patients).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza, with HR 1.44 (95% CI 1.40-1.47) for any diagnosis.

    “Most diagnostic categories were more common in patients who had COVID-19 than in those who had influenza (hazard ratio [HR] 1·44, 95% CI 1·40–1·47, for any diagnosis; 1·78, 1·68–1·89, for any first diagnosis) and those who had other respiratory tract infections (1·16, 1·14–1·17, for any diagnosis; 1·32, 1·27–1·36, for any first diagnosis).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
    • Rests on an assumption: Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

  • May combine two claims.

    Findingown data

    Similar results were observed when patients who had COVID-19 were compared with those who had one of the four other index events, except when an outcome had a predicted relationship with the comparator condition.

    “Similar results were observed when patients who had COVID-19 were compared with those who had one of the four other index events (appendix pp 11–14, 34), except when an outcome had a predicted relationship with the comparator condition (eg, intracranial haemorrhage was more common in association with fracture of a large bone).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.
    • Rests on an assumption: The ICD-10 code U07.1 reliably identifies patients with COVID-19.
    • Rests on an assumption: Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The COVID-19 cohort included 190 077 patients who were not hospitalised.

    “The COVID-19 cohort was divided into subgroups of patients who were not hospitalised (190 077 patients), those who were hospitalised (46 302 patients), those who required ITU admission (8945 patients), and those who received a diagnosis of encephalopathy (6229 patients).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    For the overall psychiatric disorder category (ICD-10 F20-48), the hazard ratio varied with time, declining but remaining significantly higher than 1, indicating that the risk was attenuated but maintained 6 months after COVID-19 diagnosis.

    “For the overall psychiatric disorder category (ICD-10 F20–48), the HR did vary with time, declining but remaining significantly higher than 1, indicating that the risk was attenuated but maintained 6 months after COVID-19 diagnosis (appendix p 9).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    The timing of index events was such that most infections with influenza and many of the other respiratory tract infections occurred earlier on during the pandemic, whereas the incidence of COVID-19 diagnoses increased over time.

    “The timing of index events was such that most infections with influenza and many of the other respiratory tract infections occurred earlier on during the pandemic, whereas the incidence of COVID-19 diagnoses increased over time (appendix p 24).”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Findingown data

    The data provide some support for the possibility of post-COVID-19 parkinsonian syndromes, although the incidence was low and not all HRs were significant.

    “Our data provide some support for this possibility, although the incidence was low and not all HRs were significant.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Findingown data

    The data on Guillain-Barré syndrome after COVID-19 were equivocal, with HRs increased compared with other respiratory tract infections but not with influenza, and increased compared with three of the four other index health events.

    “Whether COVID-19 is associated with Guillain-Barré syndrome remains unclear;26 our data were also equivocal, with HRs increased with COVID-19 compared with other respiratory tract infections but not with influenza (table 3), and increased compared with three of the four other index health events (appendix p 34).”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [26] Keddie, 2020, Epidemiological and cohort study finds no association between COVID-19 and Guillain-Barré syndrome, Brain

      Not checked in the free run.

    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
    • Rests on an assumption: The comparator cohorts (influenza and other respiratory tract infections) are appropriate to estimate the excess risk attributable to COVID-19.

  • May combine two claims.

    Findingcitation

    The risk of cerebrovascular events (ischaemic stroke and intracranial haemorrhage) was elevated after COVID-19, with the incidence of ischaemic stroke rising to almost one in ten (or three in 100 for a first stroke) in patients with encephalopathy.

    “Consistent with several other reports,23, 24 the risk of cerebrovascular events (ischaemic stroke and intracranial haemorrhage) was elevated after COVID-19, with the incidence of ischaemic stroke rising to almost one in ten (or three in 100 for a first stroke) in patients with encephalopathy.”
    Quote in the paperSubstantial claim0 of 2 sources checked
    • cites [23] Siow, 2021, Stroke as a neurological complication of COVID-19: a systematic review and meta-analysis of incidence, outcomes and predictors, J Stroke Cerebrovasc Dis source

      Not checked in the free run.

    • cites [24] Hernandez-Fernandez, 2020, Cerebrovascular disease in patients with COVID-19: neuroimaging, histological and clinical description, Brain source

      Not checked in the free run.

    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Findingcitation

    Even among patients who were hospitalised, only about 11% received a diagnosis of encephalopathy, whereas much higher rates would be expected.

    “Even among patients who were hospitalised, only about 11% received this diagnosis, whereas much higher rates would be expected.18, 30”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [18] Oldham, 2020, Characterising neuropsychiatric disorders in patients with COVID-19, Lancet Psychiatry source

      Not checked in the free run.

    • cites [30] Docherty, 2020, Features of 20 133 UK patients in hospital with covid-19 using the ISARIC WHO Clinical Characterisation Protocol: prospective observational cohort study, BMJ

      Not checked in the free run.

    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Findingown data

    2.66% of patients older than 65 years and 4.72% of patients who had encephalopathy received a first diagnosis of dementia within 6 months of having COVID-19.

    “Although the estimated incidence was modest in the whole COVID-19 cohort (table 2), 2·66% of patients older than 65 years (appendix p 28) and 4·72% who had encephalopathy (table 2), received a first diagnosis of dementia within 6 months of having COVID-19.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Findingreasoning

    HRs for most neurological outcomes were constant, and hence the risks associated with COVID-19 persisted up to the 6-month timepoint.

    “HRs for most neurological outcomes were constant, and hence the risks associated with COVID-19 persisted up to the 6-month timepoint.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Findingown data

    Substance use disorders and insomnia were more common in COVID-19 survivors than in those who had influenza or other respiratory tract infections, except for the incidence of a first diagnosis of substance use disorder after COVID-19 compared with other respiratory tract infections.

    “Substance use disorders and insomnia were also more common in COVID-19 survivors than in those who had influenza or other respiratory tract infections (except for the incidence of a first diagnosis of substance use disorder after COVID-19 compared with other respiratory tract infections).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
    • Rests on an assumption: The comparator cohorts (influenza and other respiratory tract infections) are appropriate to estimate the excess risk attributable to COVID-19.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The proportional hazard assumption was tested with the generalised Schoenfeld approach.

    “The proportional hazard assumption was tested with the generalised Schoenfeld approach.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The study was reported according to the Reporting of studies Conducted using Observational Routinely-collected health Data (RECORD).

    “Our study was reported according to the Reporting of studies Conducted using Observational Routinely-collected health Data (RECORD, appendix pp 55–60).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Premisecitation

    Survivors of COVID-19 are at increased risk of mood and anxiety disorders in the 3 months after infection.

    “Similar concerns have been raised regarding psychiatric sequelae of COVID-19,12, 13 with evidence showing that survivors are indeed at increased risk of mood and anxiety disorders in the 3 months after infection.14”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [14] Taquet, 2021, Bidirectional associations between COVID-19 and psychiatric disorder: retrospective cohort studies of 62 354 COVID-19 cases in the USA, Lancet Psychiatry source

      Not checked in the free run.

  • May combine two claims.

    Study design and conductown data

    The proportional hazards assumption was violated for intracranial haemorrhage and ischaemic stroke in patients who had COVID-19 compared with patients who had other respiratory tract infections (p=0.012 for intracranial haemorrhage and p=0.032 for ischaemic stroke).

    “The only exception was for intracranial haemorrhage and ischaemic stroke in patients who had COVID-19 when compared with patients who had other respiratory tract infections (p=0·012 for intracranial haemorrhage and p=0·032 for ischaemic stroke).”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationnone given

    Some patients in the comparison cohorts are likely to have had undiagnosed COVID-19; this would also tend to make our HRs an underestimate.

    “Some patients in the comparison cohorts are likely to have had undiagnosed COVID-19; this would also tend to make our HRs an underestimate.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Study design and conductown data

    The increased rates of neurological and psychiatric sequelae were robust in all three sensitivity analyses: including patients who had died, confirming COVID-19 diagnosis by RNA or antigen test, and comparing with influenza in 2019 or 2018.

    “The increased rates of neurological and psychiatric sequelae were robust in all three sensitivity analyses: when patients who had died by the time of the analysis were included (appendix p 51), when the COVID-19 diagnosis was confirmed by use of an RNA or antigen test (appendix p 52), and when the sequelae were compared with those observed in patients who had influenza in 2019 or 2018 (appendix pp”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Interpretationown data

    Overall, COVID-19 was associated with increased risk of neurological and psychiatric outcomes, but the incidences and HRs were greater in patients who had required hospitalisation, and markedly so in those who had required ITU admission or had developed encephalopathy, even after extensive propensity score matching.

    “Overall, COVID-19 was associated with increased risk of neurological and psychiatric outcomes, but the incidences and HRs of these were greater in patients who had required hospitalisation, and markedly so in those who had required ITU admission or had developed encephalopathy, even after extensive propensity score matching for other factors (eg, age or previous cerebrovascular disease).”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
    • Rests on an assumption: The propensity score matching adequately controls for confounding by measured covariates.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Study design and conductown data

    Our findings are robust given the sample size, the propensity score matching, and the results of the sensitivity and secondary analyses.

    “Our findings are robust given the sample size, the propensity score matching, and the results of the sensitivity and secondary analyses.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: The propensity score matching adequately controls for confounding by measured covariates.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

    Interpretationnone given

    A study of this kind can only show associations; efforts to identify mechanisms and assess causality will require prospective cohort studies and additional study designs.

    “Finally, a study of this kind can only show associations; efforts to identify mechanisms and assess causality will require prospective cohort studies and additional study designs.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    The psychiatric sequelae of COVID-19 appear widespread and to persist up to, and probably beyond, 6 months.

    “Therefore, as with the neurological outcomes, the psychiatric sequelae of COVID-19 appear widespread and to persist up to, and probably beyond, 6 months.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Study design and conductcitation

    The study has weaknesses inherent to an electronic health records study, such as unknown completeness of records, no validation of diagnoses, and sparse information on socioeconomic and lifestyle factors.

    “Nevertheless, they have weaknesses inherent to an electronic health records study,29 such as the unknown completeness of records, no validation of diagnoses, and sparse information on socioeconomic and lifestyle factors.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [29] Casey, 2016, Using electronic health records for population health research: a review of methods and applications, Annu Rev Public Health source

      Not checked in the free run.

  • May combine two claims.

    Interpretationreasoning

    The effect of these timing differences on observed rates of sequelae is unclear but, if anything, they are likely to make the HRs an underestimate because COVID-19 cases were diagnosed at a time when all other diagnoses were made at a lower rate in the population.

    “The effect of these timing differences on observed rates of sequelae is unclear but, if anything, they are likely to make the HRs an underestimate because COVID-19 cases were diagnosed at a time when all other diagnoses were made at a lower rate in the population (appendix p 24).”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

  • May combine two claims.

    Interpretationreasoning

    Under-recording of delirium during acute illness probably means that the diagnosed cases had prominent or sustained features; as such, results for this group should not be generalised to all patients with COVID-19 who experience delirium.

    “Under-recording of delirium during acute illness is well known and probably means that the diagnosed cases had prominent or sustained features; as such, results for this group should not be generalised to all patients with COVID-19 who experience delirium.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.
32 of 32

  • The paper relies on this without support.

    Diagnoses recorded in the TriNetX electronic health records accurately reflect the neurological and psychiatric outcomes under study.

    “Available data include demographics, diagnoses (using codes from ICD-10), medications, procedures, and measurements (eg, blood pressure and body-mass index).”

  • The paper relies on this without support.

    The ICD-10 code U07.1 reliably identifies patients with COVID-19.

    “The primary cohort was defined as all patients who had a confirmed diagnosis of COVID-19 (ICD-10 code U07.1).”

  • The paper relies on this without support.

    Propensity score matching on the measured covariates adequately controls for confounding by unmeasured factors.

    “We used propensity score matching19 to create cohorts with matched baseline characteristics, done within the TriNetX network.”

  • The authors state this as an assumption.

    Hospitalisation, ITU admission, and encephalopathy are valid proxies for COVID-19 severity.

    “We investigated how these estimates were affected by COVID-19 severity, as proxied by hospitalisation, intensive therapy unit (ITU) admission, and encephalopathy (delirium and related disorders).”

  • The paper relies on this without support.

    Diagnoses recorded in the electronic health records accurately reflect patients' true clinical diagnoses.

    “Nevertheless, they have weaknesses inherent to an electronic health records study,29 such as the unknown completeness of records, no validation of diagnoses, and sparse information on socioeconomic and lifestyle factors.”

  • The paper relies on this without support.

    The propensity score matching adequately controls for confounding by measured covariates.

    “even after extensive propensity score matching for other factors (eg, age or previous cerebrovascular disease)”

  • The paper relies on this without support.

    The comparator cohorts (influenza and other respiratory tract infections) are appropriate to estimate the excess risk attributable to COVID-19.

    “the choice of cohorts against which to compare COVID-19 outcomes influenced the magnitude of the HRs (table 3, appendix p 34).”

  • The paper relies on this without support.

    The 6-month follow-up period is sufficient to capture the main neurological and psychiatric outcomes of interest.

    “Longer-term studies are needed to ascertain the duration of risk and the trajectory for individual diagnoses.”
8 of 8

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Conflicting matched hospitalisation cohort sizes

    Minor

    The results text gives 44 927 matched patients for the hospitalisation comparison, whereas the corresponding Table 5 heading gives N=45 167.

  • First substance use diagnoses contradict the claimed increase versus respiratory infections

    Minor

    The text says hazard ratios were significantly greater than 1 for all diagnoses versus respiratory tract infections. Table 3 instead reports a significantly lower hazard ratio of 0·92 for a first substance use disorder diagnosis.

2 of 2

Cites

Flagged first, then by citations

4 of the 35 references couldn’t be matched to a known paper.

Cited by

Sorted by citations

Related

Authors
Taquet, M., Geddes, J. R., Husain, M., Luciano, S., & Harrison, P. J.
Affiliations
  • Oxford Health NHS Foundation Trust (GB)
  • TriNetX (United States) (US)
Corresponding author
Harrison, Paul J
ORCID
Venue
The Lancet Psychiatry
Publisher
Elsevier BV
Published
2021-05
Volume
8
Issue
5
Pages
416-427
ISSN
2215-0366 (ISSN-L)
DOI
10.1016/s2215-0366(21)00084-5
PMC
PMC8023694
OpenAlex
W3148333766
Semantic Scholar
81d0f9614f3454de7682cc062b48e04d82329866
Cited by
2,189
References
35
Citations by year
2026: 137 · 2025: 186 · 2024: 346 · 2023: 467 · 2022: 640 · 2021: 408 · 2020: 2
FWCI
222.72
Type
journal article
Language
en
Open access
Licence
Locations
  • The Lancet Psychiatry (publishedVersion, cc-by)
  • medRxiv (acceptedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
NIHR
Topics
  • Health Sciences › Medicine › Neurology › Long-Term Effects of COVID-19
  • Health Sciences › Medicine › Infectious Diseases › COVID-19 Clinical Research Studies
  • Social Sciences › Psychology › Clinical Psychology › COVID-19 and Mental Health
Keywords
COVID-19 pandemic, neurological sequelae, psychiatric sequelae, long COVID, incidence, respiratory infections, stroke, dementia, psychosis, anxiety disorders, substance use disorders, insomnia, Guillain-Barré syndrome, parkinsonism, encephalitis, delirium, COVID-19 severity, hospitalization, intensive care unit admission
MeSH terms
COVID-19, SARS-CoV-2, Mental Disorders, Female, Humans, Influenza, Human, Male, Middle Aged, Nervous System Diseases, Outcome and Process Assessment, Health Care, Research Design, Respiratory Tract Infections, Risk Factors, Severity of Illness Index, Global Health, Cohort Studies, Incidence, Proportional Hazards Models, Risk Assessment, Electronic Health Records

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-07.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-07.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review