NHS Health Checks: an observational study of equity and outcomes 2009–2017

Background The NHS Health Check cardiovascular prevention programme is now 10 years old. Aim To describe NHS Heath Check attendance, new diagnoses, and treatment in relation to equity indicators. Design and setting A nationally representative database derived from 1500 general practices from 2009–2017. Method The authors compared NHS Health Check attendance and new diagnoses and treatments by age, sex, ethnic group, and deprivation. Results In 2013–2017, 590 218 (16.9%) eligible people aged 40–74 years attended an NHS Health Check and 2 902 598 (83.1%) did not attend. South Asian ethnic groups were most likely to attend compared to others, and females more than males. New diagnoses were more likely in attendees than non-attendees: hypertension 25/1000 in attendees versus 9/1000 in non-attendees; type 2 diabetes 8/1000 versus 3/1000; and chronic kidney disease (CKD) 7/1000 versus 4/1000. In people aged ≥65 years, atrial fibrillation was newly diagnosed in 5/1000 attendees and 3/1000 non-attendees, and for dementia 2/1000 versus 1/1000, respectively. Type 2 diabetes, hypertension, and CKD were more likely in more deprived groups, and in South Asian, Black African, and Black Caribbean ethnic groups. Attendees were more likely to be prescribed statins (26/1000) than non-attendees (8/1000), and antihypertensive medicines (25/1000 versus 13/1000 non-attendees). However, of the 117 963 people with ≥10% CVD risk who were eligible for statins, only 9785 (8.3%) were prescribed them. Conclusion Uptake of NHS Health Checks remains low. Attendees were more likely than non-attendees to be diagnosed with type 2 diabetes, hypertension, and CKD, and to receive treatment with statins and antihypertensives. Most attendees received neither treatment nor referral. Of those eligible for statins, <10% were treated. Policy reviews should consider a targeted approach prioritising those at highest CVD risk for face-to-face contact and consider other options for those at lower CVD risk.


INTRODUCTION
Cardiovascular disease (CVD) remains the largest cause of premature mortality. CVD reduction has slowed in all Western countries, with widening socioeconomic gradients in the UK and, since 2018, the first increases in CVD mortality for 50 years. 1,2 In 2009, the NHS Health Check 5-yearly programme started in England, with the aim of reducing heart attack and stroke, and identifying dementia in people aged 40-74 years by assessing major risk factors and providing individual behavioural support and treatment. 3 In 2020, the programme completed its first decade, with coverage averaging 1 million annually. The effectiveness of the programme has been challenged by some researchers and clinicians. 4,5 Uptake has been variable and low at <25%, contrasting with 50%-75% uptake in cancer screening. Half of the population have a low 10-year risk of CVD (<10%) for whom trials of health checks showed no evidence of reduced CVD, though new disease was identified. 5,6 Whole population trials of behaviour change interventions for dietary and physical activity are difficult to implement and also lack evidence of CVD benefit. 7,8 Evidence of whole population behavioural change thus relies on observational, epidemiological, and modelling studies or trials in high-risk subgroups. [9][10][11] In contrast, pharmacological treatments or dietary substitution for hypertension or statins are supported by robust trial evidence. [12][13][14] Public Health England has highlighted the need to address equity of provision and inequalities, 15 and the aim of this current study was to compare NHS Health Checks from 2009-2013 with the period 2013-2017 and assess changes in attendance by age, sex, and ethnic and socioeconomic group, and new diagnosis of type 2 diabetes, hypertension, chronic kidney disease (CKD), dementia, and atrial fibrillation (AF), as well as new statin and antihypertensive treatment. 16

METHOD
The study conforms to the STROBE recommendations. 17 The authors used QResearch, a nationally representative database including 35 million people registered with 1500 UK general practices using the same Egton Medical Information System (EMIS). The primary study period included the 4 years from 1 April 2013 to 31 March 2017 and the earlier study period 1 April 2009 to 31 March 2013. Adults aged 40-74 years registered for at least 1 year who were eligible for an NHS Health Check were included. Those excluded as ineligible had pre-existing hypertension, ischaemic heart disease, stroke or transient ischaemic attack, AF, heart failure, peripheral arterial disease, CKD, familial hypercholesterolaemia, or diabetes, were already on statins, or had had an NHS Health Check within 5 years before the study entry date. 18 Data were extracted on 31 May 2018 using Read codes. 18 Outcome data were obtained on or within 12 months of an NHS Health Check, or an index date in non-attendees of 1 April in the year of cohort entry. NHS Health Check attendance was based on specific codes and not imputed as it has been in some studies. 19 Sociodemographic data and risk factors were obtained on the closest date before or on the NHS Health Check or index date. Attendance was defined as attendees as a proportion of the eligible population in that year or period. The authors included sex, age group in years, and self-reported ethnic group using Office of National Statistics categories: White (British, Irish, and other White ethnic groups); South Asian (Bangladeshi, Indian, and Pakistani); Black African; Black Caribbean; Chinese; other, including mixed ethnic groups; and not recorded. 20 Deprivation assessed by the Townsend score, derived at small area level from Census data on housing, car ownership, and unemployment, 21 was grouped into fifths, with quintile 1 the least deprived and quintile 5 the most deprived.
Risk factors included smoking status, alcohol units per day, blood pressure, blood glucose, serum cholesterol, body mass index (BMI), and QRisk2 10-year cardiovascular risk, including a family history of ischaemic heart disease coded positive in first-degree relatives, with angina or heart attack <60 years. GP referrals for obesity, smoking, or alcohol reduction were recorded. Delays in diagnosis and treatment may occur due to repeat or detailed testing or referral. 22 The authors therefore used a 12-month period after the NHS Health Check to allow completion of diagnostic and treatment processes. New diagnoses were recorded on or within 12 months of the NHS Health Check/index date, and included hypertension, CVD (ischaemic heart disease, or stroke or transient ischaemic attack), CKD, type 2 diabetes, AF, familial hypercholesterolaemia, and dementia. New diagnoses for hypertension and type 2 diabetes were based on practitionerrecorded disease register codes and not imputed from measurements. Diagnosis of CKD was based on national standards for estimated glomerular filtration (eGFR) values <60 ml/min/1.73 m 2 (categories 3-5), and non-diabetic hyperglycaemia from HbA1c 42-47 mmol/mol. New medication within 12 months was at least two statin prescriptions or two prescriptions out of the three main classes of antihypertensive medications of thiazide, calcium channel blocker, and angiotensin converting enzyme inhibitors or receptor blockers. 'Not stated or recorded' described missing values for ethnicity and deprivation.
A medical statistician used Stata MP (version 16) with Cox proportional hazards models to describe associations between outcomes and sociodemographic variables, using Royston-Parmar proportional hazards models when proportional hazard assumptions were not met. Models were adjusted for clustering by general practice, with calculated unadjusted and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) using a two-tailed test of significance of 0.01. HRs were adjusted for sociodemographic variables (see Supplementary Tables S1-S11).  Figure S1). The coverage of the 5-year rolling NHS Health Check programme, assuming attendance of onefifth of the eligible population each year, is described in Table 1, increasing from

How this fits in
The uptake of NHS Health Check has remained low. Half of the attendees were <50 years or people at low CVD risk who received neither treatment nor referral. The NHS Health Check identified important new diagnoses of hypertension, type 2 diabetes, and chronic kidney disease, and this study reports increased new diagnoses of atrial fibrillation and dementia in people aged ≥65 years. Black African, Black Caribbean, and South Asian ethnic groups were more likely to be identified with disease compared to other ethnic groups. Treatment with statins was three times more likely in attendees than non-attendees, and treatment with antihypertensives was also increased. However, of those eligible for statins, only 8.3% were prescribed them. More targeted approaches should be considered to improve efficiency and cost effectiveness.  Supplementary Table S13).
Conversely, one in eight females and almost one in three males were at ≥10% CVD risk (see Supplementary Table S13 Tables S1 and S12). Table 2 shows CVD risk by attendance. Missing data for some risk factors were more likely in non-attendees. Major risk factor recording for attendees and nonattendees is described in Supplementary  Table S4).

Risk factors by attendance status and by period
New diagnosis of CKD was higher in Black Caribbean and Black African ethnic groups and in more deprived quintiles (see Supplementary Table S5). New CVD diagnosis was similar in attendees versus non-attendees (HR 1.06, 95% CI = 0.99 to 1.14). In attendees, Pakistani and Bangladeshi ethnic groups had higher risks of new CVD diagnosis and Black Africans lower risks, with an increasing gradient of CVD diagnosis with deprivation (see Supplementary Table S6). Familial hypercholesterolaemia is described in Supplementary Table S7. Atrial fibrillation and dementia diagnosis was more likely in males and at older ages 70-74 years; HRs are described in Supplementary Tables S8  and S9.  Table 3 shows statin treatment by period and attendance. In 2013-2017, new treatment with statins was more frequent among attendees (26/1000) than nonattendees (8/1000; HR 2.98, 95% CI = 2.84 to 3.13). Similarly, antihypertensive medicines were more likely to be prescribed in attendees (25/1000) than non-attendees (13/1000; HR 1.65, 95% CI = 1.59 to 1.72) (see Supplementary Tables S10 and S11).   Prescribing of statins and antihypertensives was higher in the earlier period because attendees were older and at higher risk than in the later period. In 2009-2013, statins were prescribed to 37/1000 and 12/1000 of attendees and non-attendees, respectively, and antihypertensives to 28/1000 and 16/1000, respectively. Supplementary Table S15 shows the gradient of increased statin prescriptions by category of CVD risk and attendance in the two study periods. Statin prescription was higher in the recorded risk categories in the later period, with the exception of those with <5% CVD risk.
Of those attendees eligible for statins with 10%-19% CVD risk, only 6.0% (5754/96 095) were treated and, of those at ≥20% CVD risk, 18.4% (4031/21 868) were treated. In 2013-2017, in people with a QRisk2 score ≥10%, statins were prescribed to 82.9/1000 and in 2009-2013 to 75.5/1000, with the later increase most pronounced in the group with risks of 10%-19%, at 38.9/1000 to 59.9/1000, respectively. South Asians were more likely and Black African, Black Caribbean, and Chinese ethnic groups were less likely to be prescribed statins than White ethnic groups (see Supplementary Table S10). Table 5 shows statin prescription by CVD risk.

DISCUSSION Summary
This study provides new insights on equity of provision of NHS Health Checks, and new diagnosis and treatment by age, sex, and ethnic and socioeconomic group. Coverage increased initially in 2009-2013, but thereafter remained persistently low, averaging 24% over the years 2013-2017. Equity of attendance was variable, more likely in South Asian and less likely in Black African, Black Caribbean, and Chinese ethnic groups than White ethnic groups, and lower in males compared to females. In 2013-2017, people aged <50 years accounted for almost half of the attendances; 57.8% of females and 32.9% of males who attended were at low CVD risk (<5%). Conversely, one in eight females and almost one in three males were at ≥10% CVD risk.
Of those attendees eligible for statins with 10%-19% CVD risk, only 6.0% (5754/96 095) were treated and, of those at ≥20% CVD risk, 18.4% (4031/21 868) were treated. New statin treatment was almost three times more likely among attendees than nonattendees, and antihypertensive prescription was more likely in attendees. In attendees with CVD risk of 10%-19%, statin prescribing increased after revised 2014 National Institute for Health and Care Excellence guidance. 23 South Asian ethnic groups (Indian, Pakistani, and Bangladeshi) were more likely and Black and Chinese ethnic groups were less likely to be prescribed statins than White ethnic groups. One new case of diagnosed hypertension was detected for every 40 people attending an NHS Health Check, one new case of type 2 diabetes for every 130 attendances; one new case of CKD for every 138 attendances; and one new familial hypercholesterolaemia case for every 1118 attenders. In attendees aged 65-74 years, one new case of AF was detected for every 209 patients and every 578 patients for dementia. Type 2 diabetes diagnosis was between three and six times more likely in South Asian than White ethnic groups, and CKD and hypertension were more likely in Black Caribbean and Black African ethnic groups. Increased diagnosis of these conditions was more likely in more socially deprived attendees, most pronounced with type 2 diabetes. At a public health level, increased diagnosis and treatment is likely to contribute equitably  to the health needs of socially diverse populations. Earlier detection of hypertension, type 2 diabetes, CKD, AF, and dementia was more likely in attendees. Attendees were almost three times more likely to be treated with statins and more likely to receive antihypertensive medication than nonattendees. For those at higher CVD risk, diagnosis and treatment confer important health benefits, but for those at low CVD risk, attendance lacks evidence of benefit.

Strengths and limitations
GP payment for NHS Health Checks was based on specific codes, resulting in substantial coding completeness. 18 NHS Health Checks provided by pharmacists or local authorities may not be recorded in GP records, but represent a small proportion of attendances.
Preventive programmes consistently report a healthy attendee effect, with substantial residual confounding in comparisons with non-attendees. Some studies propensity-matched attendees and non-attendees to reduce confounding. 19 However, despite adjustment for all known confounders, the Danish Inter99 CVD prevention study showed non-attendees had substantially higher rates of accidental death and other unrelated causes than attendees. 41,42 These biases reduce cardiovascular risks and events in attendees because they are likely to be healthier than non-attendees. The finding of higher rates of new diagnosis and treatment after NHS Health Checks runs counter to that bias and strengthens these findings.

Comparison with existing literature
Other studies of NHS Health Checks have raised concerns about effectiveness, equity of delivery, and lack of benefit in people at low cardiovascular risk. 5,24,25 A literature review of NHS Health Checks, largely before 2014, 24 identified low coverage and poor lifestyle modification as issues of concern, a finding echoed by patients. 26 The current study would indicate that little has changed. More than 75% of those eligible to attend do not do so. Though a referral may be recorded by GPs, there is no information on whether the patient attended. More than 80% of attendees did not receive a referral or treatment and added value was based on brief and often superficial advice from a healthcare assistant during a single visit taken up largely with recording. 25,27 This study uptake, based on the registered GP population, was lower than national reporting, which used mid-year population estimates and invitation response as denominators. 28 NHS Digital data on NHS Health Checks 2012-2018 did not include new diagnoses and treatments, and statin prescribing was almost twice as high as in this and previous studies indicating major methodological differences. 29 No socioeconomic difference in attendance was reported. 29 Like other earlier studies of 2009-2013, 16,[30][31][32][33][34][35] the present study of 2013-2017 identified more diagnoses of new hypertension, diabetes, and CKD, and higher prescription of statins among NHS Health Check attendees. 36 This study is the first to report increased diagnoses of AF and dementia resulting from an NHS Health Check. Three previous studies considered AF, and none reported dementia, using populations comprised largely of people <65 years in which both conditions are rare and, hence, unlikely to identify changes in diagnosis. 36,37 Chang et al found increased AF diagnosis in attendees, which was not significant after matching. 30 The present study used a denominator of age 65-74 years followed for 12 months, and the authors observed significantly more new diagnoses of AF in attendees (5/1000) than non-attendees (3/1000), and new diagnoses of dementia were recorded in 2/1000 attendees versus 1/1000 non-attendees. Approximately 2% of people aged 65-74 years in community settings are estimated to have dementia. 38 Lower attendance by Chinese and Black African patients was confirmed in the current study, which showed highest attendance in South Asians similar to reports in previous studies. 39,40 Implications for practice Modelling of NHS Health Check effectiveness is a more appropriate method to estimate the CVD benefits of additional diagnoses and treatment, with estimates of 300 fewer premature deaths and 1000 more people living free of CVD. 43,44 There is a policy decision to be made about whether it is more effective to improve persistently low attendance or target those at increased CVD risk. 45 There is little evidence that attendees at lower CVD risk benefit from brief advice at NHS Health Checks and targeted approaches are more efficient. 46,47 The COVID-19 pandemic halted the NHS Health Check programme in 2020, highlighting opportunities for online engagement for more than half of the population who have a CVD risk of <10% (Table 5). This would free up resources to improve targeting of people at higher CVD risk. 48

Provenance
Freely submitted; externally peer reviewed.

Ethical approval
The Trent Research Ethics committee has approved use of the QResearch database for anonymised use of primary care data (http:// www.qresearch.org).