Effects of statin treatment on primary and hospital care use: a microsimulation model
1Health Economics Research Centre, Nuffield Department of Population Health, University of Oxford, Oxford, UK
2Health Economics and Policy Research Unit, Wolfson Institute of Population Health, Queen Mary University of London, London, UK
3Centre for Primary Care, Wolfson Institute of Population Health, Queen Mary University of London, London, UK
4North East London Integrated Care Board, London, UK
*Corresponding to: Dr. Junwen Zhou, Health Economics Research Centre, Nuffield Department of Population Health, Old Road Campus, Headington, Oxford OX3 7LF, UK. email: junwen.zhou@ndph.ox.ac.uk.Abstract
Background
Statin treatment’s efficacy, safety and cost-effectiveness are well accepted, but the long-term impact on healthcare use is not fully understood. We assessed statin treatment effects on primary and hospital care use over time.
Methods
The UK Biobank population cohort with linked hospital admissions (N=501,807) and primary healthcare data (N=192,983) informed models of hospital admissions, hospital inpatient days, primary care services (consultations, diagnostic and monitoring tests, and medication prescription items) associated with individual characteristics and occurrences of myocardial infarction, stroke, coronary revascularization and vascular death. These models were integrated into a validated cardiovascular disease (CVD) microsimulation policy model to assess statin treatment effects on healthcare use in population categories by age (40-60 years and 60-70 years) and prior CVD history.
Results
Statin treatment was associated with improved survival and lower rates of hospital admissions, hospital inpatient days and prescription items per person-year over lifetime. Compared to no treatment, healthcare use with statin treatment was lower in earlier years after initiation, these reductions diminished over time and transitioned into higher healthcare use. The number of years to net neutral effect (95%CI) ranged from 9 (7-12) to 17 (9-25) for consultations/tests, from 22 (17-28) to 38 (28-48) for prescription items/hospital admissions, and from 40 (30-47) to 51 (43-62) for hospital inpatient days. Earlier transitions were observed in older people and people with prior CVD history.
Conclusions
Statin treatment reduces individual rates of hospital inpatient services and medication prescriptions but increases overall healthcare use driven by increased longevity and ageing.
Funding
NIHR-HTA, NIHR-Barts-BRC.
Article notes
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
Support from the UK NIHR Health Technology Assessment (HTA) Programme (17/140/02) and the National Institute for Health Research Barts Biomedical Research Centre (NIHR203330) is acknowledged. The study was designed and analysed independently of all funders and the views expressed are those of the author(s) and not necessarily those of the NIHR, the Department of Health and Social Care or any other funder.
Introduction
Cardiovascular diseases (CVDs) are the leading causes of death worldwide, contributing to substantial health and economic burden globally1. In the United Kingdom (UK), 7.6 million people are living with CVD, which is associated with 0.17 million deaths and £12 billion healthcare costs annually2. Statin therapy is widely recommended to reduce CVD events in people at increased risk3-6. Evidence on the efficacy4 and safety7 of statins has informed clinician-patient discussions of statin treatment3 and, alongside evidence on statins’ effects on quality-adjusted life years, healthcare costs and cost-effectiveness5,6, has supported widening guideline recommendations for statin use3.
However, impact of statin treatment on healthcare use has received less attention. This is important for both healthcare planning and resource allocation as well as for patients themselves. While preventive interventions lower the risk of CVD, they increase survival and may increase overall healthcare use due to healthcare need during extended life years and increased morbidity with ageing. The need for healthcare due to ill health also substantially affects patients’ lives consuming time, energy and financial resources and causing personal discomfort8,9.
In this study, we assessed the impact of CVD events on hospital admissions and primary care services in the UK and evaluated the impact of statin therapy on healthcare use using a CVD microsimulation model.
Methods
Study population and data
The UK Biobank (UKB) study is a prospective cohort study of over 500,000 adults, aged 40-70 years at recruitment from 2006 to 2010 across England, Scotland, and Wales10. All participants with established linkage to primary (40% of the cohort) or hospital care records, except a small number at end-stage kidney disease at recruitment, contributed to our current study (Table S1). Participants’ data from enrolment into UKB until the earliest of 31 March 2016, death or loss to follow-up, contributed to the present analysis. During the follow-up period, we evaluated the number of hospital admissions and inpatient days for any cause from the linked hospital care records, and the number of primary care consultations, primary care diagnostic and monitoring tests, and medication prescription items by primary care clinicians from the linked primary care records. We focused on the impact on healthcare use of the first occurrence of four cardiovascular events post-enrolment: myocardial infarction (MI), stroke, coronary revascularization (CRV), and vascular death (VD). The identification of baseline characteristics, disease events and healthcare use and the approach to missing data imputation were previously reported11,12.
Statistical Models of healthcare use
The healthcare use outcomes were established over annual periods from recruitment in UKB by summing up the number of respective resource uses incurred by each participant during each year of follow-up in the study. Annual numbers of each type of healthcare use were modelled using two-part models with the first part modelling the probability of any healthcare use using logistic regression, and the second part modelling the number of non-zero healthcare use, using generalised linear model with Gamma distribution and identity link13. This model was chosen based on the comparison of model performance across different models (see Supplementary section 1).
Each model included the following pre-specified participant characteristics at entry: sex, ethnicity, quintile of Townsend deprivation index, smoking status, physical activity, diet quality, body mass index, low- and high-density lipoprotein cholesterol, serum creatinine, systolic and diastolic blood pressure, use of antihypertensive treatment and histories of diabetes mellitus, severe mental illness or cardiovascular disease. Each model included annually updated temporal histories of the first occurrences of MI, stroke, and CRV during follow-up, each of which had the following 4 categories: no event; same year (event in the annual period); 1 year ago (event in the previous annual period); and ≥2 years ago (event in an annual period two or more years ago). The annually updated participant characteristics also included current age, temporal histories of incident diabetes or cancer, and vascular and non-vascular death. We included the interactions between same year vascular death and each of the other vascular events (MI, stroke, CRV), and between the temporal history of MI and the temporal history of CRV. Cluster robust standard errors were estimated acknowledging the lack of independence between annual periods for the same participant.
Model performance was checked in deciles of predicted annual use overall and by age group and prior CVD history (Figure S1). The mean absolute and relative excess annual healthcare use associated with CVD events in the UKB was estimated using the estimated models and recycled prediction. This was calculated by subtracting (for absolute excess) or dividing (for relative excess) the predicted annual healthcare use in years with a CVD event by the predicted use in years without a CVD event.
CVD microsimulation model integrated with healthcare use models
Statin effects on healthcare use were assessed by propagating the statin effects on CVD adverse events to associated healthcare use. The previously reported UK CVD microsimulation policy model11 was used to perform this assessment. Briefly, this decision-analytic model projects annually the first occurrence of four CVD events: MI, stroke, CRV and VD; and three key non-vascular events: incident diabetes, incident cancer and non-vascular death with the occurrence of any of these non-fatal events impacting the risks of subsequent events. The model was developed using the individual participant data of 16 large statin versus control randomised clinical trials14, calibrated using the UKB data10, and validated across categories of UKB10 and Whitehall II cohort15 participants, and against national mortality and cancer incidence rates and other published data. The model was previously used to project the event risks and survival over individuals’ remaining lifetimes (i.e. death or reaching 110 years of age) without and with statin therapy11, and to assess the cost-effectiveness of different statin therapies in categories of individuals5,6. This microsimulation model was adapted by integrating the healthcare use models (Figure S2), enabling the projection of healthcare use without and with statin therapy over time.
Effects of statin therapy on healthcare use of UK Biobank participants
We assessed the lifetime effects of atorvastatin 40 mg/day on healthcare use using UKB participants’ data. We randomly sampled 10,000 UKB participants from each of the UKB subpopulations by age (40-60 and 60-70 years) and prior CVD history (without and with) at recruitment (Table S2) to use in the simulation of statin’s effects in the adapted CVD policy model. As in previous work5, effects of statin treatment on CVD events4,16 and adverse effects of statin17-19 were informed from meta-analyses of trials and cohort studies (Table S3). We summarised the effects of statin therapy on cumulative healthcare use per person-year and per person over time for each of the four subpopulations by age and prior CVD history. The mean estimate was derived from base-case analysis running 500 microsimulations per individual. Uncertainty around the estimates were derived from 500 Monte-Carlo simulations capturing the uncertainty in effects of statin therapy, event risk equations and healthcare use equations related to participant characteristics and events5,11. (Figure S3-S4)
Results
Study population
A total of 501,807 UKB participants were included in the study, with 501,807 and 192,983 participants, contributing to hospital and primary care use analyses, respectively (Figure S5). Compared to participants aged 40-60 years, participants aged 60-70 years were more likely to have prior CVD history. Participants aged 60-70 and those with prior CVD history were more likely to be male, former or current smokers, with higher BMI, higher creatinine levels, with history of hypertension, diabetes, cancer, or severe mental illness (Table 1).
Impact of CVD events on healthcare use
Study participants were followed for a mean duration of 7 years from recruitment. Compared to those without prior CVD, participants with CVD history were more likely to experience cardiovascular events during follow-up: MI (3.9% vs. 1.0%), stroke (3.3% vs. 0.9%), CRV (5.4% vs. 1.3%) and VD (2.5% vs. 0.4%). They also had larger annual healthcare use per person: hospital admissions (0.7 vs. 0.3), inpatient days (1.9 vs. 0.7), primary care consultations (8.3 vs. 5.1), primary care diagnostic and monitoring tests (5.8 vs. 3.0), and primary care medication prescription items (48.3 vs. 17.9) (Table S4). In the two-part regression models (Table S5-S6), MI, stroke and CRV events were associated with higher healthcare use. These excesses decreased over the two years following the events (except for medication prescription items, where the effect remained high), but longer-term higher use remained. Compared to years without an event, the relative ratio (95% CI) of hospital admissions more than two years following a CVD event ranged from 1.29 (1.22-1.36) to 1.56 (1.43-1.70), inpatient days 1.18 (1.10-1.27) to 2.95 (2.46-3.43), primary care consultations from 1.18 (1.13-1.22) to 1.31 (1.25-1.36), primary care diagnostic and monitoring tests from 1.36 (1.29-1.42) to 1.61 (1.51-1.71), and primary care prescription items from 2.10 (1.99-2.20) to 2.42 (2.18-2.66). VD was associated with higher use of hospital services but lower use of primary care services in the year of death. (Figure 1, Table S7)
Burden of illness in the absence of statin treatment
In the absence of statin treatment, the predicted 10-year risk of new major vascular event (MVE, defined as MI, stroke, CRV, or VD) ranged from 3.3% to 6.9% (for individuals aged 40-60 to 60-70 respectively) in those without CVD history, and from 13.8% to 21.4% in those with prior CVD history. Over a lifetime, these risks increased to 35.6%-38.2% and 52.7%-53.5%, respectively. The predicted remaining life expectancy was 35.2 and 26.3 years for participants age 40-60 years and 60-70 years without prior CVD history, decreasing to 27.4 and 20.8 years respectively in those with prior CVD history (Table S8). Estimated rate of healthcare use increased over time and was consistently higher among older individuals and those with CVD history. Per person-year, over the first 10 years, mean number of hospital admissions ranged from 0.29 (aged 40-60 without CVD history) to 0.85 (aged 60-70 with CVD history), rising to 0.52-1.06 over lifetime. Hospital inpatient days increased from 0.65-2.49 to 1.59-3.57, primary care consultations from 4.66-8.90 to 6.04-9.75, primary care diagnostic and monitoring tests from 2.76-6.51 to 3.94-7.32, and primary care medication prescription items from 14.7-52.1 to 24.2-59.7. (Figure 2, Table S9)
Impact of statin treatment on hospital admissions and primary care use
Over longer time horizons, statin treatment was associated with larger reductions in the proportion of people with new MVE, ranging from 1.2%-2.5% and 4.8%-7.1% in people without and with CVD history respectively over 10 years to 7.5%-8.9% and 10.2%-11.5%, respectively, over lifetime. These effects were paralleled by increasing gains in life expectancy ranging from 0.005-0.015 and 0.039-0.062 years per person over 10 years among people without and with prior CVD history, respectively, to 0.64-0.71 and 0.80-0.94 over lifetime. (Figure 3, Table S9)
The joint effects of statin treatment in reducing the risk of CVD events and increasing life expectancy led to large reductions in healthcare use in earlier years after statin initiation followed by an increased longer-term use. The estimated time from statin initiation to neutral net effect on cumulative healthcare use occurred earliest for primary care tests and consultations [ranging from 9 (7-12) to 17 (9-25) years] respectively, followed by primary care medication prescription items and hospital admissions [22 (17-28) to 38 (28-48) years] respectively, and hospital inpatient days [40 (30-47) to 51 (43-62) years]. These transitions occurred earlier in older individuals and those with prior CVD history. Over lifetime, statin treatment led to an estimated per-person increase of 0.33 (0.17, 0.48) to 0.60 (0.33, 0.88) hospital admissions, 0.29 (-0.35, 0.93) to 0.89 (-0.14, 1.92) hospital admission days, 4.93 (3.48, 6.38) to 8.35 (5.94, 10.76) consultations, 3.70 (2.57, 4.82) to 6.39 (4.51, 8.27) tests, and 15.4 (6.2. 24.5) to 33.6 (17.8, 49.3) prescription items. (Figure 4, Table S10) Unlike healthcare use per person, healthcare use per person-year remained lower over lifetime for hospital admissions [0.0000002 (-0.002, 0.002) to -0.014 (-0.019, -0.009)] and hospital inpatient days [-0.016 (-0.030, - 0.002) to -0.115 (-0.150, -0.081)], and primary care prescription items [-0.002 (-0.196, 0.191) to – 0.927 (-1.252, -0.602)]. (Figure 4, Table S9)
Discussion
This study provides novel insights into the impact of statin therapy on healthcare use over time. We report that MI, stroke, and CRV events were associated with marked increases in healthcare use in the first two years post events which declined but remained elevated thereafter, with persistently high demand for medications. Statin treatment resulted in increased survival and lower rates of hospital admissions, inpatient days and primary care prescriptions per person-year over lifetime, with larger reductions for older adults and those with a prior CVD history. Over lifetime, however, per person net healthcare use was greater among statin users. Transitions to greater net healthcare use occurred sooner for primary care consultations and tests (10–20 years after statin initiation) and later for hospital inpatient days (40–50 years after statin initiation), with earlier transitions observed in older individuals and those with a prior history of CVD.
Our findings align with McConnachie et al.20, which showed that allocation to statin over five years in the WOSCOPS study significantly reduced cardiovascular-related hospital admissions and inpatient days, with a modest increase in admissions for other causes over 15 years. We extend this evidence by assessing estimated statin effects with longer duration of statin treatment, over a longer timeframe, across a broader population (including women and individuals with prior CVD), and on additional healthcare resources such as primary care services. In contrast, Cooke et al. found no association between statin use and reductions in re-hospitalizations or physician visits.21 However, their analysis spanned only about two years — likely too short to detect differences. In addition, they focused on statin use in 1997-2001, a period dominated by less potent statin formulations.
Despite robust evidence for statins’ efficacy and safety, millions of statin eligible people in the UK remain untreated,22 and adherence is often poor even among individuals with prior CVD23. Programs aimed at improving adherence to statin therapy, including pharmacist interventions24, mobile phone text messaging25, and financial incentives26, have limited evidence for effectiveness and cost-effectiveness. Our study provides new evidence that statin treatment reduces average rate of healthcare use, particularly hospital admissions, broadening their value beyond clinical benefits. Communicating the benefits of preventive interventions such as statin therapy on reducing hospital admission in addition to reducing risk of CVD events — could improve public awareness of the broader consequences of CVD and benefits of prevention and further incentivise heathier lifestyles and adherence to preventive measures3,27.
We also show how statins impact long term healthcare use. Among statin treated individuals, healthcare use is progressively lower after statin initiation, levelling off after the first decade and transitioning into increased use after the second decade — earlier for primary care services and in older individuals or those with prior CVD history. The increased survival with statin treatment contributed to increased healthcare use in later years as a result of ageing and morbidity such as cancer and diabetes (Figure S6) among survivors. These insights can inform healthcare resource planning. While our analysis is UK-based, the findings are likely relevant to other settings, including resource-limited setting.8
The key strengths of the study include the use of the large-scale, population-based UKB cohort, which enabled a comprehensive assessment of healthcare use linked to CVD events across a diverse, real-world population. The integration of the detailed healthcare use data into a validated CVD microsimulation model allowed for a robust estimation of statin impact on healthcare use over time— addressing a gap in the literature on healthcare use implications of statin treatment and illustrating broader impact of preventive therapies. A study limitation is the lack of data availability for hospital outpatient, laboratory services, emergency department care, and social care use; further research to study statin effects on the broader range of healthcare and social care services will be helpful. Finally, the UK Biobank cohort is not representative of the general UK population, with reduced social diversity and a healthy volunteer effect. However, we present findings by two key factors, age and prior CVD history, which we hope improves generalizability.
In summary, our study quantifies the increased healthcare use following CVD events, and statins’ longer term effects on healthcare use. While statins reduce healthcare use in earlier years, extended life expectancy and ageing lead to increased healthcare use in survivors in later years, especially for primary care services and among older people or people with a prior CVD history at statin initiation. These dynamics underscore the need for long-term planning in healthcare resource allocation, with a shift from acute CVD care toward sustained preventive and chronic care services.
Declaration statement
Data availability statement
Data may be obtained from a third party and are not publicly available. The datasets used in the current study are available from UK Biobank (https://www.ukbiobank.ac.uk/). Researchers can apply to use the UK Biobank resource.
Supporting information
Data Availability
Data may be obtained from a third party and are not publicly available. The datasets used in the current study are available from UK Biobank (https://www.ukbiobank.ac.uk/). Researchers can apply to use the UK Biobank resource.
Acknowledgements
This research has been conducted using data from UK Biobank, a major biomedical database www.ukbiobank.ac.uk. We thank all the participants, staff and other contributors to the resource.
Funding
Support from the UK NIHR Health Technology Assessment (HTA) Programme (17/140/02) and the National Institute for Health Research Barts Biomedical Research Centre (NIHR203330) is acknowledged. The study was designed and analysed independently of all funders and the views expressed are those of the author(s) and not necessarily those of the NIHR, the Department of
Health and Social Care or any other funder.
Conflict of interest statement
The authors declare that they have no competing interests.