Employment outcomes of people with Long Covid symptoms: community-based cohort study

Abstract Background Evidence on the long-term employment consequences of SARS-CoV-2 infection is lacking. We used data from a large, community-based sample in the UK to estimate associations between Long Covid and employment outcomes. Methods This was an observational, longitudinal study using a pre–post design. We included survey participants from 3 February 2021 to 30 September 2022 when they were aged 16–64 years and not in education. Using conditional logit modelling, we explored the time-varying relationship between Long Covid status ≥12 weeks after a first test-confirmed SARS-CoV-2 infection (reference: pre-infection) and labour market inactivity (neither working nor looking for work) or workplace absence lasting ≥4 weeks. Results Of 206 299 participants (mean age 45 years, 54% female, 92% white), 15% were ever labour market inactive and 10% were ever long-term absent during follow-up. Compared with pre-infection, inactivity was higher in participants reporting Long Covid 30 to <40 weeks [adjusted odds ratio (aOR): 1.45; 95% CI: 1.17–1.81] or 40 to <52 weeks (aOR: 1.34; 95% CI: 1.05–1.72) post-infection. Combining with official statistics on Long Covid prevalence, and assuming a correct statistical model, our estimates translate to 27 000 (95% CI: 6000–47 000) working-age adults in the UK being inactive because of Long Covid in July 2022. Conclusions Long Covid is likely to have contributed to reduced participation in the UK labour market, though it is unlikely to be the sole driver. Further research is required to quantify the contribution of other factors, such as indirect health effects of the pandemic.


A
proportion of people infected with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) continue to experience symptoms months or years later, known as Long Covid, post coronavirus disease 2019 (COVID-19) syndrome or post COVID-19 condition.
4][5] A substantial proportion of people with Long Covid also report psychiatric symptoms associated with depression and anxiety, 6 although these may not be a direct result of past SARS-CoV-2 infection. 7,8In March 2023, an estimated 1.9 million people in the UK (2.9% of the population) self-reported Long Covid, with prevalence being highest among working-age people. 6he growing number of SARS-CoV-2 infections in the UK has coincided with rising labour market inactivity, defined as neither being in work nor actively seeking it.By the end of 2022, the number of working-age people who were inactive had increased by over 350 000 since pre-pandemic, with inactivity due to long-term sickness growing most quickly. 9One study suggests that nearly half of UK firms had employees with Long Covid in 2022, and a quarter reported it as a substantial cause of long-term absence. 10Long Covid may have resulted in long-term absence for 110 000 UK workers and the loss of 4.4 million working hours per week. 11This may have implications for the livelihood of individuals but also for the health of the macroeconomy, including income and earnings, labour market supply, productivity, tax receipts, benefit payments and consumer demand.
However, despite the potential impact of Long Covid on global labour markets, there is limited evidence on its relationship with employment.In one study, 45% of participants with Long Covid required a reduced work schedule and 22% were not working due to their illness seven months post-infection. 12In another analysis, 19% of participants with Long Covid were unable to work, 10% reported reduced working hours, and 37% said that their income had been affected. 13Among participants with SARS-CoV-2, Long Covid has been associated with 44% higher odds of not working and 27% lower odds of working full-time. 14A study of 50 000 participants with SARS-CoV-2 found an average reduction in self-reported working capacity of 10.7% after a mean follow-up of 8.5 months post-infection. 15hile providing valuable insights, these studies are largely crosssectional and descriptive in nature, lack robust control groups for comparison, and are based on convenience samples that may not be representative.Inferential studies on Long Covid and employment outcomes are scarce, and limited sample sizes and follow-up time have precluded detailed analysis of population subgroups.Therefore, in this study, we used longitudinal data from a large, community-based sample to estimate associations between Long Covid and employment outcomes.

Study design and data
We performed an observational, longitudinal, pre-post analysis of participants from the UK COVID-19 Infection Survey (CIS, ISRCTN21086382), 16 a study of people aged 2 years from randomly sampled households (excluding hospitals, care homes, halls of residence and prisons).The CIS received ethical approval from the South Central Berkshire B Research Ethics Committee (20/SC/ 0195).Enrolment took place from 26 April 2020 to 31 January 2022, accruing over 530 000 participants from over 260 000 households.Supplementary table S1 reports household enrolment rates, which were as high as 51% at the start of the survey (comprising previous respondents to government surveys who had consented to participate in future research) but dropped to 12% by the end of recruitment (when sampling was conducted randomly from national address lists; the majority of participating households were recruited during this phase); >97% of enrolled participants consented to monthly assessments for 1 year.
Participants provided a self-swab for polymerase chain reaction (PCR) testing at each assessment, while a sub-sample also provided blood samples for antibody testing.Participants reported whether they had tested positive for SARS-CoV-2 or antibodies outside of the study and their current Long Covid status ('Would you describe yourself as having 'Long Covid', that is, you are still experiencing symptoms more than 4 weeks after you first had COVID-19 that are not explained by something else?').Data collection was conducted via face-to-face interviews until June 2022 before switching to remote (online or telephone) collection.

Inclusion and exclusion criteria
We included monthly assessments from 3 February 2021 (when Long Covid data were first collected) to 30 September 2022 when participants responded to the Long Covid question, were aged 16-64 years, and were not in full-time education.
We excluded participants with a first positive swab for SARS-CoV-2 (either a PCR test via study assessments or any selfreported test outside of the study) at enrolment, as the timing of infection could not be determined for these participants.We also excluded participants with a positive spike-antibody test (excluding any tests after COVID-19 vaccination) or who suspected they had COVID-19 14 days before their first positive swab, as the latter may represent a reinfection.To ensure we could fully observe participants' self-reported Long Covid experience, we excluded participants first testing positive before 11 November 2020 (12 weeks before Long Covid data started being collected).
When analysing long-term absence, we excluded assessments when participants were not in employment, as well as those before 1 October 2021 (when the UK Coronavirus Job Retention Scheme was in operation).

Exposures
The time-varying exposure was past SARS-CoV-

Outcomes
The outcomes were labour market inactivity (excluding retirement, i.e. neither working nor looking for work, and not retired); and workplace absence for 4 weeks for any reason whilst in employment.The CIS question used to derive these outcomes can be found in Supplementary Appendix S1.

Covariates
We considered a range of socio-demographic variables hypothesized to be related to both Long Covid and employment status: age, sex, white or non-white ethnicity, country/region, area deprivation quintile group, and self-reported health/disability status.We also examined labour market variables: employment status, employment sector, occupation group and whether self-employed.All variables were measured at CIS enrolment and thus before any SARS-CoV-2 infection.

Statistical methods
We descriptively compared confounders between participants who ever or never reported Long Covid during follow-up using means and proportions for continuous and categorical variables, respectively.Large differences between groups were indicated by absolute standardized differences >10%, a threshold widely used in previous research. 17e fitted conditional logit (fixed effects) models controlling for background labour market conditions over the study period by adjusting for the calendar day of each study assessment (modelled as a restricted cubic spline with boundary knots at the 10th and 90th percentiles and an internal knot at the median of the time distribution) interacted with current age (restricted cubic spline, as for calendar day), sex and health/disability status at CIS enrolment.We performed several sensitivity analyses for the primary outcome (labour market inactivity), detailed in Supplementary Appendix S2.A range of heterogeneity tests for both study outcomes can be found in Supplementary Appendix S3.
Models were fitted using the 'clogit' function in R's 'survival' package. 18We reported results as adjusted odds ratios (aORs) and 95% confidence intervals (CIs), with pre-infection being the reference group.All statistical analyses were performed using R version 4.0.

Population attributable risk
By combining our aORs with official statistics on the population prevalence of Long Covid by inactivity status and time since infection, 19 we estimated the number of working-age adults in the UK who were inactive in July 2022 because of Long Covid (assuming a correct statistical model); that is, those reporting Long Covid who would have been working had they not been infected with SARS-CoV-2.A detailed description of these calculations can be found in Supplementary Appendix S4.

Characteristics of study participants
The analysed population comprised 206 299 participants (Supplementary figure S1) contributing a total of 2 547 618 monthly follow-up assessments (mean 12.3 assessments per participant).147 895 participants (contributing 1 155 207 assessments) were in employment from 1 October 2021 and were therefore included in the analysis of long-term absence.97 751 participants (47.4% of the total analysed population of 206 299) tested positive for SARS-CoV-2 during follow-up, while the remaining 108 548 (52.6% of the total) never tested positive.Of the 97 751 participants who tested positive, 8440 (8.6%) reported Long Covid 12 weeks post-infection.
Compared with participants infected with SARS-CoV-2 without reporting Long Covid, those who reported Long Covid were on average older (46.3 vs. 44.3years) at enrolment, and were more likely to be female (63.2% vs. 55.1%),living in the most deprived quintile group (14.9% vs. 10.8%),living with a long-term health condition/disability (24.2% vs. 16.2%), and inactive in the labour market (9.7% vs. 6.6%)(table 1).Among infected participants who were employed at enrolment, those who reported Long Covid during follow-up were more likely to be working in the teaching and education sector (16.0% vs. 12.8%) and in caring, leisure and other service occupations (10.3% vs. 6.5%).A comparison of study participants who reported Long Covid versus those not infected with SARS-CoV-2 can be found in Supplementary table S2.

Labour market inactivity
A total of 31 248 study participants (15.1%) were ever inactive during follow-up.Irrespective of timing, 17.7% of participants who ever reported Long Covid were ever inactive, compared with 13.4% of those infected with SARS-CoV-2 without reporting Long Covid.
Applying these aORs to published official statistics on the population prevalence of Long Covid by inactivity status and time since infection, an estimated 27 000 (95% CI: 6000-47 000) working-age non-students were inactive because of Long Covid in July 2022, assuming a correct statistical model.
In sensitivity analysis, similar results were obtained when restricting the population to participants testing positive for SARS-CoV-2, excluding assessments when participants were retired, and increasing the number of internal knots in the splines for calendar time and age (Supplementary Appendix S2).Supplementary figure S2 demonstrates the benefit of adjusting for calendar time compared with unadjusted estimates.

Long-term absence
Of 147 895 participants in employment from 1 October 2021, 14 493 (9.8%) ever reported long-term absence.Irrespective of timing, longterm absence was experienced by 13.1% of participants who reported Long Covid, compared with 9.8% of those infected with SARS-CoV-2 without reporting Long Covid.

Findings in context
The main finding of our study is that reporting Long Covid after SARS-CoV-2 infection is associated with increased odds of labour market inactivity and long-term absence compared with preinfection, with the period of greatest excess risk being 30 to <40 weeks post-infection for inactivity and 18 to <30 weeks postinfection for absence.Our results add to those from a limited number of longitudinal studies examining the employment outcomes of COVID-19, all with smaller sample sizes and shorter follow-up, from which the findings to date are mixed.Among 17 000 UK participants recruited via social and traditional media, SARS-CoV-2 was associated with a five-fold increase in the odds of sickness absence beyond the acute phase of infection, but there was no evidence of a relationship between COVID-19 and inadequate household income in the long-term. 20Among 36 000 UK Household Longitudinal Study (UKHLS) respondents, Long Covid was associated with reduced working hours but not being out of employment, with the former appearing to dissipate after six months. 11ur findings are also coherent with a small body of research internationally.In the US, Long Covid has been shown to be associated with a greater likelihood of unemployment and a reduced likelihood of working full-time. 14Participants in a German study experienced reduced working capacity 6-12 months after SARS-CoV-2 infection. 15Mean work ability scores were lower in a Swiss cohort with Long Covid 1-year post-infection compared with those without Long Covid. 21ur findings are coherent with the broader labour market landscape in the UK, which has been characterized by rising levels of inactivity throughout the pandemic. 22However, it seems unlikely that Long Covid is the sole, or even main, driver of this trend.The number of working-age adults who were inactive due to ill-health had been gradually rising since early 2019, nearly a year before the emergence of COVID-19. 23Furthermore, persistently increasing levels of inactivity during the pandemic has not been commonplace internationally, despite Long Covid having a global burden. 4The UK is among only 9 of 38 Organisation for Economic Co-operation and Development (OECD) member states for which the inactivity rate among people aged 15-64 years was higher in the third quarter of 2022 than three years earlier, and one of only six for which the rate continued to rise over the latest four quarters. 24o contextualize our estimated attributable risk of 27 000 people inactive because of Long Covid, this represents 0.5% of total inactivity (excluding retirement) among working-age non-students in the UK in July 2022, 22 and 13% of those reporting Long Covid. 19This suggests the majority of inactive people with Long Covid may have been absent from the labour market even if they had not been infected with SARS-CoV-2, for example due to comorbidities.

Employment outcomes of people with Long Covid symptoms 491
Table 1 Characteristics at enrolment of study participants ever infected with SARS-CoV-2 during follow-up, stratified by whether participants ever subsequently reported Long Covid Other studies suggest that 80 000 25 -96 000 26 people might have left employment directly because of Long Covid by March 2022.These numbers are higher than our estimate as they represent cumulative exits from employment rather than point-in-time measures.
We observed an inverted U-shaped relationship between Long Covid and inactivity over time, with the odds ratios peaking at 30 to <40 weeks post-infection before subsiding thereafter.This observation, coupled with lack of evidence of a relationship between time since infection and inactivity for individuals who previously had Long Covid symptoms, suggests that some people who left employment while experiencing Long Covid may have later returned to work (either in remission or with residual symptoms).This may have been facilitated by there being 1.2 million vacancies in the UK during the fourth quarter of 2022, 40% more than pre-pandemic. 27The rate of consumer price inflation was also high in the UK during 2022, peaking at 11.1% in October 2022 compared with 12 months earlier, 28 thus financial pressures may have encouraged some people with Long Covid to return to work.Our finding that the risk of absence levels off after 30 weeks postinfection may partly reflect people returning to work (including those with persistent illness) upon completion of their 28-week period of statutory sick pay.Continuing to work while sick, so called 'presenteeism', has been linked to reduced productivity 29 and increased risk of future absence. 30However, we cannot rule out a survivorship effect: study participants remaining at risk of long-term absence beyond 30 weeks post-infection were those who had not yet left the workforce, including due to ill-health.

Strengths and limitations
To the authors' knowledge, this is the largest longitudinal study to examine the relationship between Long Covid and employment outcomes, with over 200 000 participants included in the analysis.Selection bias was minimized by selecting households at random from national address lists, while prospective data collection meant that responses were not affected by recall effects.The CIS enrolment rate had dropped to 12% by the end of recruitment, so possible selection effects mean that our findings are not necessarily generalizable to the broader population.However, the attrition rate was low, so our analysis is unlikely to have been substantially affected by loss-to-follow-up bias.
The conditional logit modelling approach implicitly controlled for all observed and unobserved time-invariant confounders, whereby each participant acted as their own control.However, we cannot rule out residual confounding by unmeasured time-varying factors (which may be evident from the observed negative association between SARS-CoV-2 infection and labour market inactivity <18 weeks post-infection), and reverse causality may have affected the estimated relationships between Long Covid and employment status to some extent.This means our estimate of labour market inactivity attributable to Long Covid, which relies on the assumption that the underlying statistical model is correct, should be interpreted with caution.
We did not have data on working hours, so we were unable to investigate the relationship between Long Covid and reduced working time.It was also not possible to distinguish long-term absence due to sickness from that due to other reasons, for example maternity/paternity leave.However, the latter are unlikely to be influenced by SARS-CoV-2, so we expect the impact on our findings to be small.Our estimates of inactivity due to Long Covid do not consider indirect impacts, such as family members of people with Long Covid reducing their working hours or leaving employment to take on caring responsibilities.Furthermore, our estimates do not include the effects of early retirement, transitioning to unemployment (not working but looking for work and available to start), or students choosing to remain in education instead of finding work.However, we expect these effects to be small because: there were 27 000 fewer working-age retirees in the UK in the final quarter of 2022 compared with 2019, before the pandemic; 22 the unemployment rate in the UK dropped to 3.5% during 2022, a historic low since comparable records began in 1992; 22 and the population prevalence of Long Covid is considerably lower among people aged <25 years (a group comprising the majority of students) than in older people. 6uture studies with longer follow-up could provide insights on a broader range of socio-economic outcomes following SARS-CoV-2 infection, such as income and earnings, social security benefit claims, and socio-economic position.
We excluded participants first testing positive for SARS-CoV-2 before 11 November 2020 and those with confirmed SARS-CoV-2 > 14 days before their first positive swab.This means that our findings are not necessarily generalizable to people who were infected in the 'first wave' of the COVID-19 pandemic (irrespective of whether they were later re-infected).
Long Covid status was self-reported by study participants, so exposure misclassification is possible (e.g.symptoms being caused by a medical condition unrelated to COVID-19).However, there is currently no biological test for Long Covid, so case ascertainment based on recorded clinical diagnoses is likely to lack sensitivity due to under-presentation, under-diagnosing and under-coding. 31

Conclusion
Long Covid is associated with labour market inactivity among working-age people, with the greatest risk occurring 30 to <40 weeks after SARS-CoV-2 infection.Long Covid is also associated with long-term absence 18 to <30 weeks post-infection.This reduced working capacity, and its downstream effects on income and living conditions, may be detrimental to the physical and mental health and wellbeing of the individuals affected. 32At a macroeconomic level, it is likely that Long Covid has contributed to reduced participation in the UK labour market to some extent, but it is unlikely to have been the sole driver, with fewer than 30 000 working-age adults estimated to be inactive because of Long Covid in July 2022.The contribution of factors besides Long Covid to reduced labour market participation, such as indirect health effects of the pandemic and extended healthcare waiting lists, remains unknown, and further research is required.

3 )
Notes: SD, standard deviation; SOC, standard occupational classification.Area deprivation was based on the English Indices of Deprivation 2019, the Welsh Index of Multiple Deprivation 2019, the Scottish Index of Multiple Deprivation 2020 and the Northern Ireland Multiple Deprivation Measure 2017.

Figure 1 Figure 2
Figure 1 Adjusted odds ratios for inactivity (excluding retirement) compared with the pre-infection period, by time since SARS-CoV-2 infection and Long Covid status.Notes: Estimates are from a conditional logit model adjusted for calendar day of study assessment, current age, and interactions between calendar day and each of current age, sex, and self-reported health/disability status at survey enrolment Figure 2 Adjusted odds ratios for long-term (4 weeks) absence compared with the pre-infection period, by time since SARS-CoV-2 infection and Long Covid status.Notes: Estimates are from a conditional logit model adjusted for calendar day of study assessment, current age, and interactions between calendar day and each of current age, sex, and self-reported health/disability status at survey enrolment.The model was fitted to study assessments from 1 October 2021 when participants were in employment