Using General Practice Patient Survey data to explore prevalence and patient uncertainty about Long Covid
1School of Primary Care, Population Sciences and Medical Education, Faculty of Medicine, University of Southampton, Southampton, UK
2School of Geography and Environment, Faculty of Environmental and Life Sciences, The University of Southampton, UK
3University Hospital Southampton NHS Foundation Trust, Southampton, UK
4NIHR Applied Research Collaboration Wessex, UK
* Corresponding author; email: n.a.alwan@soton.ac.ukAbstract
Background
The high global burden of Long Covid (LC) has significant implications for population wellbeing, healthcare, social care and national economies.
Aim
To explore associations between patient sociodemographic and health characteristics with two outcomes: reporting having LC and expressing uncertainty about having LC, as described by general practice (GP) survey respondents.
Design and setting
Analysis of GP Patient Survey (England), a random sample of 759,149 patients aged 16yrs+ registered with a GP in England (2023).
Method
Multivariable logistic regression modelling comparing those with and without LC, and those who were unsure in relation to patient characteristics.
Results
4.8% of respondents reported having LC, and 9.1% were unsure. Significant adjusted associations indicating higher risk of LC included age (highest odds 35-54yrs), sex (females), ethnicity (White Gypsy/Irish Traveller, mixed/multiple ethnic groups), sexual orientation (gay/lesbian or bisexual), living in a deprived area, being a carer or a parent and having a long-term condition (LTC). Those aged ≤25yrs, males, non-binary, heterosexual, not parents or carers, from other white, Indian, Bangladeshi, Chinese, Black or Arab backgrounds, former and current smokers, and with no defined LTC were more likely to be unsure about having LC compared to answering ‘yes’.
Conclusion
There is an unequal distribution of LC in England, with the condition being more prevalent in minoritised and disadvantaged groups. There are also high levels of uncertainty about having LC. Improved awareness is needed amongst the general population and healthcare professionals to ensure those most vulnerable in society are identified and provided with care and support.
Article notes
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
This work forms part of MW's PhD which is funded by a University of Southampton Presidential Scholarship
Introduction
In the UK, the Office for National Statistics estimated in March 2024 that 2 million adults and children experience self-reported Long Covid (LC) (3.3% of the population), with 69% and 41% experiencing it for at least 1 and 2 years respectively and 19% reporting that LC limits their ability to undertake day to day activities ‘a lot’1. There is also evidence of inequalities in LC prevalence, with higher prevalence associated with being female2–9, of older or middle age2–4, 6, 7, having a higher body mass index (BMI)2, 3, 5, 7, 8, smoking3, 5, and belonging to an ethnic minority group5, 8. Higher prevalence is also associated with greater deprivation3–5, 9. There is also unequal impact of the condition on people’s lives10, 11 12 and inequitable support13, 14 15 for the condition. More research is needed to understand the extent of this inequality and the action needed to address it.
As a relatively novel condition, knowledge about LC is still in its infancy amongst researchers and healthcare professionals but there is even more lack of awareness, exacerbated by barriers and stigma, in the general population leaving unwell people unsure if they have LC16, 17. This high and unequal burden of illness from LC has significant implications for society in terms of the infrastructure of accessible support and treatment needed for people with LC.
The General Practice Patient Survey (GPPS) is an annual survey of people aged 16yrs+ registered with a GP in England, administered by Ipsos on behalf of NHS England18. Established in 2007, it asks respondents about their experience of their local GP, other local NHS services and respondents’ general health. In 2022 a new question was added that asked whether respondents had Long Covid.
The aim of this study was to (1) explore prevalence of LC, (2) examine its potential sociodemographic and health characteristics and (3) explore factors associated with being unsure about having LC.
Methods
Sampling
Each year in January the GPPS is issued by post to a random sample of patients registered with a GP in England. It can be completed on paper (returned using a FREEPOST envelope), online or using a dedicated telephone helpline. Invitations and reminders are also issued by text message and the online survey is available in 14 different languages and British Sign Language. Paper copies can be requested in Braille and large print. Data is grouped by geographical and health system factors including general practice, Primary Care Network, Integrated Care System, Local Authority and NHS Region. Indices of deprivation are also available according to patient postcode and further technical detail is available online19.
The survey is issued to approximately 2.5 million registered patients and aimed to ensure a minimum number of responses per practice and Primary Care Network (PCN) and a set confidence interval for each practice. A proportionately stratified (by age, gender and postcode) unclustered sample is drawn for each practice.19 Patient identifiable data is not collected. It is a standalone self-reported survey that is not linked to GP patient records. The questionnaire was last significantly revised in 2021 and is available online20. More detail about the survey is available at https://gp-patient.co.uk/FAQ
Design
This is an analysis of cross-sectional survey data using the GPPS dataset for 2023. Permission to access and use the individual patient level survey data from 2023 was requested and granted by agreement between NHS England and the University of Southampton. Ethical approval for this analysis was granted by the University of Southampton Faculty of Medicine Ethics Committee (No.82369).
Variables
The outcome was the response to the question ‘Would you describe yourself as having “Long Covid”, that is, you are still experiencing symptoms more than 12 weeks after you first had COVID-19, that are not explained by something else?’. Respondents could select ‘yes’, ‘no’, ‘not sure’ or ‘prefer not to say’.
The exposure variables used in the analysis are listed in Tables 1 and 2. All were categorical and used as collected with the exception of smoking status. Smoking status was recorded as never, former, occasional and regular smoker; and was condensed to never, former and current smoker (regular and occasional) for analysis.
Analysis
Data was analysed using Stata v1721. Descriptive statistics, frequencies and proportions were calculated followed by univariable and multivariable logistic regression. The primary analysis compared those answering ‘yes’ and ‘no’ to the LC question, excluding those answering unsure or ‘prefer not to say’. The same analysis comparing those answering ‘not sure’ and ‘no’ was also undertaken, as well as comparing ‘not sure’ and ‘yes’. We conducted the latter analysis to understand the factors associated with uncertainty about having the health condition which may act as a barrier to seeking care and support.
We conducted multivariable analysis to assess the independent effects of variables given they can potentially confound each other. The multivariable analysis included age, sex, sexuality, parental/guardian status, carer status, ethnicity, patient’s area of residence index of multiple deprivation (IMD) quintile (a combined index of relative deprivation based on data in seven domains)22, religion, smoking status and having at least one defined LTC (other than LC). A new variable was created to identify respondents with at least one defined LTC. The question that asked if respondents had any LTCs could not be included as respondents with LC could have counted their LC as a LTC (the LTC variable could include the LC outcome). Work status was included as an adjustment in the analysis comparing ‘yes’ and ‘not sure’ but not included when comparing ‘yes’ and ‘no’ (work status could have been impacted by having LC).
Results
The survey was completed by 759,149 people (28.6% response rate).
Figure 1 shows the percentage of respondents self-reporting having LC and being unsure if they have LC.
Respondents describing themselves as having LC (compared with those who answered ‘No’)
The percentage of respondents self-describing as having LC was 4.8% (n=35,445) (Table 1). The highest prevalence of LC was in the North West region (5.5%) and lowest in the South West (3.9%) (Table 1). 9% of people who described themselves as permanently sick or disabled had LC (Table 1). Compared to being in full-time work, people who were students, retired, looking after the family or home, or ‘doing something else’ were less likely to report having LC (Table 2). People who were unemployed or permanently sick or disabled were more likely to report having LC (unadjusted 1.12, 95%CI 1.06-1.19; 1.92, 95%CI 1.84-2.00 respectively).
Of those who stated they had LC, 59.9% (n=20,973) were female. Males were less likely to report having LC compared to females (adjusted OR 0.88, 95%CI 0.86-0.90) (Table 2). 60.7% of those with LC were aged between 35 and 64yrs. Compared to those aged 16-24yrs, having LC was significantly less common in those ≥65yrs and more common in those aged 25-64yrs even after adjustment for other characteristics (Table 2). The prevalence of LC significantly increased as area deprivation increased, with living in the most deprived area quintile associated with 47% higher odds compared to the least (adjusted OR 1.47, 95%CI 1.42-1.53) (Table 2).
People who were gay/lesbian, bisexual or ‘other’ sexual orientations had significantly higher prevalence compared to heterosexuals (Table 2). Parents were more likely to have LC compared to non-parents, as were carers in all caring hours categories compared to non-carers (Table 2). Compared to white British ethnicity, significantly higher adjusted odds of having LC were seen in White Gypsy or Irish Traveller groups, any other White background, White and Asian and ‘other’ mixed or multiple ethnic groups, and ‘any other ethnic group’ (Table 1). All three Black as well as Chinese ethnic groups were less likely to report having LC (Table 2). Compared to having no religion, adjusted ORs of having LC were higher for Buddhist, Christian, Muslim, Sikh and ‘other’ religions (Table 2).
The prevalence of LC amongst current regular smokers was 4.8% (Table 1) and the proportion of people with LC who were regular or occasional smokers was 12.2% compared to 11.2% smokers among those with no LC. Compared to people who had never smoked, the adjusted OR of having LC in occasional/regular smokers was 0.96 (95%CI 0.93-1.00) and for former smokers was 1.19 (95%CI 1.16-1.22) (Table 2).
There was particularly high prevalence of LC amongst people who were deaf and using sign language (14.6%), and those who had Alzheimer’s/dementia (9.2%), autism (8.8%), a breathing condition (9.0%), a mental health condition (9.2%) or another LTC not listed (7.8%) (Table 1). People with at least one defined LTC (which was not LC) had higher odds of having LC (adjusted OR 1.97, 95%CI 1.92-2.01).
Respondents who were ‘not sure’ if they would describe themselves as having LC (compared with those who answered ‘Yes’)
Those aged 25-74yrs were significantly less likely to say they are unsure about having LC (compared to those who answered ‘yes’ to having LC) than those aged 16-24yrs (Table 3). Males and non-binary people were significantly more likely to be unsure compared to females. Gay/lesbian and ‘Other’ sexual orientations were significantly less likely to be unsure compared to heterosexuals. Parents were less likely to be unsure compared to non-parents (adjusted OR 0.93, 95% CI 0.89-0.96), and carers less likely to be unsure compared to non-carers (Table 3).
White Gypsy or Irish Traveller were significantly less likely to be unsure that they had LC (compared to those who answered ‘yes’ to having LC) than White British ethnicity. ‘Any other White background’, Indian, Bangladeshi, Chinese, ‘Any other Asian background’, Black Caribbean, Black African, Arab and ‘Any other ethnic group’ were significantly more likely to be unsure compared to White British. There was no significant difference by level of area deprivation (IMD quintile) between those who were unsure and those who said they have LC or between those with a religion and those without.
Unemployment, working part-time, being retired, looking after the family or home, or ‘doing something else’ were positively linked to not being sure as opposed to saying ‘yes’ to have LC, compared with people working full-time. People who were permanently sick or disabled were less likely to be unsure (Table 3).
Compared with those who answered ‘yes’ to having LC, former or current smokers were significantly more likely to be unsure than people who never smoked. People with at least one defined LTC were less likely to be unsure compared to people without (Table 3).
Results comparing responses of those answering ‘not sure’ and ‘no’ are included in Table 4.
Discussion
Our analysis of this national general practice survey data found that 4.8% of respondents self-described as having LC while 9.1% were unsure whether they have it or not. There is an unequal distribution of LC in England, with the condition being more prevalent in minoritised and disadvantaged groups. There are also high levels of uncertainty about LC, with certain groups that are already disadvantaged being more likely to be uncertain if they have the condition.
Many of this study’s findings support existing research on risk factors for LC, for example lower odds of having LC in males compared to females2–9, higher odds in those of working age2–4, 6, 7, the strong trend of increasing odds of LC with increasing deprivation1, 3–5, 9 and higher odds for people who are unemployed1, and who have existing health conditions2, 5, 7. UK studies have also found higher prevalence of LC in particular ethnic minority groups1, 3, 5, 8 but findings have not been consistent. Lower odds were also found for people of Black ethnicity, something that has been found in other UK studies1, 7. Higher prevalence in people from sexual orientation minorities has also been identified in one US study23.
In this study, there were higher odds of having LC for people who are parents and carers. This could be explained by disproportionate exposure to SARS-CoV-2 through schools and caring roles, or greater awareness of LC in these groups. The latter is supported by the reduced odds found in this study of being unsure about having LC in those groups. The impact of LC symptoms could be more noticeable due to caring roles, or the likelihood of being able to rest during the acute COVID-19 phase may have been limited (a potential factor in some LC cases10). There were also higher odds for people of particular faiths, even after adjusting for ethnicity, deprivation, age, gender and other potential confounding factors. However, this still may be explained by residual confounding. Odds of having LC were higher in former smokers but lower for current smokers, which is contrary to studies where smoking has been identified as a risk factor3, 5. However, reviews of this area recognise the variation in findings about the role of smoking in LC24.
This study also adds new evidence about the lack of certainty people feel about whether or not they have LC. Groups who were less confident about whether they had LC included males and non-binary people, people from ‘Any other white background’, Indian, Bangladeshi, Chinese, ‘Any other Asian background’, Black Caribbean, Black African, Arab and ‘Any other ethnic group’, those who were working part-time, unemployed, retired, or staying at home and smokers. Those who were less likely to be unsure about having LC included those aged 24-74yrs, those who were not heterosexual, parents/guardians and carers, White Gypsy/Irish Traveller, people who were sick or disabled and people with at least one defined LTC. Those under the age of 25yrs were less confident about having LC. This could explain why most research about LC includes older people as most research uses self-identification with LC as study inclusion strategy. It could also be the case that there is wider associated stigma with having LC in younger people17.
Given the high burden of LC in the global population and its impact on everyday lives of individuals and the functioning of society, it is important to understand the drivers for uncertainty around LC to enable people to get the right support. Levels of uncertainty may indicate a lack of awareness of LC symptoms or confusion, or there may be an overlap in symptoms with other conditions. A lack of uncertainty in particular groups may be because a diagnosis has already been secured, indicating that some groups may be more likely than others to be successfully diagnosed, or it could be related to levels of health literacy and self-advocacy. Stigma and self-doubt can shape self-identification with LC16, 17, 25, 26 which could be discouraging people from coming forward for diagnosis and support and there is emerging evidence that intersectional identity may compound LC stigma leading to worse outcomes for marginalised groups16.
Strengths and Limitations
A key strength of this study is its large sample size and a national survey sample. Limitations of this study include the lack of data about severity or symptoms of LC, the timing of its onset and the COVID-19 vaccination status of respondents. In addition, interpreting survey responses about LTCs with respect to LC is problematic. LC is a long-term condition but 21.9% of people who described themselves as having LC said they did not have any LTCs. To take account of this the multivariable models in this study included a variable constructed from named LTCs as reported by respondents rather than the answer to the question ‘any other LTC’ to eliminate potentially including the LC outcome. A further limitation is that we cannot role reverse causality for some of the variables such as work status, hence this variable was not included in the multivariable analysis.
Conclusions
This study provides further evidence that particular population groups in England, often those minoritised and disadvantaged, are more likely to have LC. It adds new evidence that there are more people who are unsure whether they have LC than those who are confident they have it, again some from more vulnerable groups. This clearly demonstrable health inequality lends weight to the argument for diagnosis, treatment and support to be better distributed according to need27, to avoid LC becoming another factor contributing to the widening health gap between different groups in society. Improved awareness about the condition amongst the general population and training about barriers and stigma among healthcare professionals are required to ensure people with LC can be identified and provided with the right treatment and support.
Data Availability
GP Patient Survey data is available at GP practice-level via https://gp-patient.co.uk/analysistool. Requests for GPPS patient-level datasets can be made directly to the survey team whose details can be found here https://gp-patient.co.uk/confidentiality