Introduction. David Epstein 1,2 Leticia García-Mochón - PDF

Eur J Health Econ DOI /s ORIGINAL PAPER Modeling the costs and long-term health benefits of screening the general population for risks of cardiovascular disease: a review of methods

Please download to get full document.

View again

of 12
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.


Publish on:

Views: 12 | Pages: 12

Extension: PDF | Download: 0

Eur J Health Econ DOI /s ORIGINAL PAPER Modeling the costs and long-term health benefits of screening the general population for risks of cardiovascular disease: a review of methods used in the literature David Epstein 1,2 Leticia García-Mochón 2 Stephen Kaptoge 3 Simon G. Thompson 3 Received: 5 March 2015 / Accepted: 25 November 2015 The Author(s) This article is published with open access at Abstract Background Strategies for screening and intervening to reduce the risk of cardiovascular disease (CVD) in primary care settings need to be assessed in terms of both their costs and long-term health effects. We undertook a literature review to investigate the methodologies used. Methods In a framework of developing a new healtheconomic model for evaluating different screening strategies for primary prevention of CVD in Europe (EPIC-CVD project), we identified seven key modeling issues and reviewed papers published between 2000 and 2013 to assess how they were addressed. Results We found 13 relevant health-economic modeling studies of screening to prevent CVD in primary care. The models varied in their degree of complexity, with between two and 33 health states. Programmes that screen the whole population by a fixed cut-off (e.g., predicted 10-year CVD risk [20 %) identify predominantly elderly people, who may not be those most likely to benefit from long-term treatment. Uncertainty and model validation were generally poorly addressed. Few studies considered the disutility of taking drugs in otherwise healthy individuals or the budget impact of the programme. Electronic supplementary material The online version of this article (doi: /s ) contains supplementary material, which is available to authorized users. & David Epstein Department of Applied Economics, University of Granada, Campus de la Cartuja, Granada, Spain Escuela Andaluza de Salud Pública, Granada, Spain Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK Conclusions Model validation, incorporation of parameter uncertainty, and sensitivity analyses for assumptions made are all important components of model building and reporting, and deserve more attention. Complex models may not necessarily give more accurate predictions. Availability of a large enough source dataset to reliably estimate all relevant input parameters is crucial for achieving credible results. Decision criteria should consider budget impact and the medicalization of the population as well as cost-effectiveness thresholds. Keywords Cost-effectiveness analysis Screening Cardiovascular disease Primary prevention Statins Literature review JEL Classification Introduction I180 H510 Cardiovascular disease (CVD) is a major public health problem with a huge impact on health service budgets in European countries [1]. Current guidelines for primary prevention of CVD generally involve a combination of advice for lifestyle change and/or pharmacological intervention (e.g., statins or anti-hypertensives) in those assessed to be at sufficiently high-risk [2 5]. The parameters of such programmes vary greatly between countries. Most countries use opportunistic case finding, although the UK has recently launched a national screening programme [6]. National guidelines recommend initiating statin therapy when the 10-year risk of CVD exceeds 7.5 % in the USA [2], 10 % in the UK [7], and 20 % in other countries [8]. An explicit comparison of the costs and benefits of CVD risk assessment and treatment informs some guidelines [7], D. Epstein et al. but not others [5]. Cost-effectiveness of a screening strategy might be optimized by appropriate choice of the risk algorithm, employing the most efficient threshold for initiating treatment [9], or using stepwise or targeted screening strategies [10]. There are also concerns about the long-term side effects of statins and medicalizing a large proportion of the general population [11]. In this paper, we report a literature review conducted to help develop a new health-economic model for evaluating different screening strategies and interventions to prevent CVD in European countries ( We identify a series of questions that an economic analysis in this area ought to address, and describe and comment on the approaches used. These questions are based on the authors experience and discussions while preparing the paper. Several published reviews of the health-economic evidence for primary prevention of CVD already exist [12 17]. Each offers useful insights, but none considers all of the following methodological questions that we believe need to be addressed together: 1. What are the criteria used for cost-effectiveness? 2. What is the structure of the economic model? 3. What are the population and strategies of interest? 4. How are primary CVD outcomes defined and assessed? 5. How are individuals at high risk of CVD identified and treated? 6. How are resources, costs and quality of life measured? 7. How is the model implemented and validated? The structure of the paper is as follows. First, we describe the literature search. Second, we discuss the health-economic approaches used to address each question in the selected articles. We compare and critique these approaches as we go. Lastly, we discuss some general themes raised by the review and tentatively propose some recommendations. The recommendations reflect our opinion, but are intended to summarize the advantages and drawbacks of each approach in different decision contexts. Literature search We conducted a literature review to identify studies describing health-economic models of cost-effectiveness of screening strategies for primary prevention of CVD in the general population. The web appendix (etable 1) provides details of the bibliographic terms used and the search results obtained from PubMed and Web of Science databases. Studies were included in the final review if they were published between January 2000 and September 2013, concerned CVD screening strategies or general health checks that could be implemented in a primary care setting with current technology, were full economic evaluations (i.e., include both costs and benefits), targeted the adult general population without previous history of CVD, and were based on models with a time horizon [1 year. Studies were excluded if they assessed tests or technology not commonly available in primary care settings in western Europe, did not include CVD screening as the initial step (e.g., economic evaluation of statin treatments), or were targeted at subgroups of the general population (e.g., people already identified as intermediate risk, or patients with diabetes mellitus). As this paper is a review of methodological approaches, rather than a quality assessment of the articles themselves, we also excluded articles that replicated broadly similar methods to another included study. The literature search initially identified 459 articles, of which 47 were selected for full text retrieval based on relevance of title and abstract (Fig. 1). After reading the full text, 13 articles met the inclusion criteria specified above. The main reasons for excluding the remaining 34 articles were that they did not evaluate screening strategies (n = 14), did not involve full economic evaluations (n = 7), did not evaluate screening strategies and did not involve full economic evaluation (n = 1), were not based in the adult general population (n = 8), or had a time horizon \1 year (n = 4). Table 1 and etable 2 summarize the main characteristics of the included studies [18 30]. Critique of the health-economic approaches used by the included studies Question 1: What are the criteria for costeffectiveness? The quality-adjusted life-year () was the most commonly used health outcome, measured over the patient s lifetime or restricted to 10 years. The is a composite measure calculated as the product of survival and health-related quality of life, and is therefore appropriate for a condition such as CVD which impacts on both dimensions of health. Use of alternative metrics such as the number of CVD events prevented or CVD-free life-years gained does not take account of the patient experience after the CVD event. While the captures both morbidity and mortality, it has been criticized for excluding other considerations that might be important to decision-makers, for example, the effect of the programme on health-related inequalities or vulnerable groups [31], the impact on labor productivity [19], moral hazard (e.g., statins may give a false sense of health security to treated individuals, counteracting the Modeling the costs and long-term health benefits of screening the general population for Fig. 1 Flow chart for the selection of economic evaluation studies Identified articles (n=459) Articles duplicated (n=52) Articles excluded by Title and Abstract (n=360) Full text retrieval (n=47) Dropped after critical appraisal (n=34) - Not full economic evaluation (n=7) - Not screening strategies (n=14) - Not general population (n=8) - Not long term time horizon (n=4) - Not screening strategies & no full economic evaluation (n=1) Final articles included (n=13) incentive to adopt lifestyle changes), and medicalizing a generally healthy population [32]. Any health gained by implementing a new programme has an opportunity cost of health (and other goods) foregone elsewhere. Some studies used the threshold approach, comparing the incremental cost-effectiveness ratio (ICER) of the intervention with the national threshold set by relevant health-care authorities (Table 1). A fixed ICER threshold may not be appropriate for making decisions about large-scale public health programmes such as national screening if financing these gross changes would successively cut into more essential and productive health services elsewhere. An alternative way to estimate the opportunity cost of introducing a new screening programme is to use the fixed-budget method, in which the additional number of individuals treated is fixed up front (e.g., top quartile of the population at greatest CVD risk) and then the strategy that maximizes total health given the fixed budget is considered as the optimal screening strategy [23]. One study [20] calculated an efficiency frontier [33]. This allows dominated options (those at higher cost but no more effective) to be identified and excluded, but unless the decision-maker is willing to specify a cost-effectiveness threshold, does not offer any guidance about choosing between options on the frontier. Question 2: What is the structure of the economic model? The structure of a model represents the important events or states whose occurrence or state-occupancy are to be predicted. As CVD is a chronic condition, the model should predict events over the full lifetime of the cohort of patients. Decision models can facilitate extrapolation (prediction of events beyond the time horizon of the primary studies), synthesis (bringing together evidence from different and diverse sources), and sensitivity analysis (prediction or simulation under alternative assumptions or data). The models reviewed were implemented with varying degrees of complexity with between 2 and 33 states (see Table 2 and etable 3 for a description of the health states in each model). Simpler structures included states such as no CVD, non-fatal CVD event, and dead. Other models distinguished between types of non-fatal CVD events (e.g., stroke, myocardial infarction (MI)), causes of death (e.g., CVD-related, other causes), or included adverse events of treatment as separate health states. More complex models included successive non-fatal CVD events (e.g., stroke followed by MI) or time-dependency (e.g., a tunnel state in a state- model to incorporate a D. Epstein et al. Table 1 Summary of the characteristics of the included studies Articles Year Region/country Target population/result by subgroup Blake et al. [18] Johannesson [19] Marshall and Rouse [20] 2003 US People aged 35 to 85 years without hyperlipidemia (LDL cholesterol \149 mg/ dl)/age and sex 2001 Sweden People aged C35 years/sex and age 2002 UK People aged 30 to 74 years/no analysis by subgroup Pletcher et al. [21] Kok et al. [22] 2009 US People aged 35 to 85 years/age, sex and risk level 2009 The Netherlands People aged [30 years/ age and sex Rapsomaniki et al. [23] Wald et al. [24] Choudhry et al. [25] Lovibond et al. [26] 2011 North America, Western Europe, and Japan People aged C40 years/ no analysis by subgroup 2011 UK People aged 0 to 89 years/age and CVD risk cut-off 2011 US Men aged C50 years and women C60 years with LDL cholesterol \130 mg/dl/no analysis by subgroup 2011 UK People aged C40 years/ age and sex Strategies of screening compared (S1, S2, S3, etc.) Treatment Model Time horizon Perspective Outcome S1: no screening and no treatment (usual care); S2: C-reactive protein screening and treatment; S3: no screening and treat all Statin State 10 years Health care Cost per Screening at different risk thresholds Statin State S1: clinical risk assessment for all patients at age 30; S2: pre-selection of patients for assessment using a prior estimate of their CVD risk that include age, sex, diabetes status, and default values for other risk factors S1: Adult Treatment Panel III guidelines; S2: range of risk- and age-based alternative strategies S1: old guideline S2: new guideline (SCORE) S1: gender, region, age and year of birth; S2: additionally includes three established CVD risk factors: SBP, total cholesterol, and smoking status Aspirin, statin and antihypertensives 5-year probability of CVD Statin State Statin and antihypertensives State Statin Partitioned survival curve Lifetime Social Cost per 5 years Primary health care Cost per CVD event prevented 30 years Health care Cost per 20 years Health care Cost per LY; cost per 10 years Health care Cost per CVD-free year of life S1: testing hs-crp and rosuvastatin for patients with hs-crp C2 mg/l; S2: no screening and no treatment (usual care); Individual patient simulation Statin State Lifetime Health care Cost per CVD-free year of life Lifetime Social Cost per S1: BP monitoring in the clinic (measurements at monthly intervals over 3 months); S2: BP monitoring in the home (measurements over a week); S3: ambulatory monitoring (measurements over 24 h) S1: age alone; S2: FRS Statin and antihypertensives Antihypertensives treatment State Lifetime Health care Cost per Modeling the costs and long-term health benefits of screening the general population for Table 1 continued Articles Year Region/country Target population/result by subgroup Strategies of screening compared (S1, S2, S3, etc.) Treatment Model Time horizon Perspective Outcome Cobiac et al. [27] Shiffman et al. [28] den Ruitjer et al. [29] 2012 Australia People aged C35 years/ absolute risk and sex 2012 US People aged 45 to 79 years/sex 2013 US People aged years/sex Lee et al. [30] 2010 US People aged C40 years/ age, sex, and absolute risk S1: usual care; S2: single risk factor-based guidelines; S3: absolute risk-based guidelines Statin and antihypertensives State S1: FRS; S2: FRS? lipoprotein(a) Aspirin State S1: FRS; S2. FRS? carotid intima-media thickness S1: Adult Treatment Panel III guidelines; S2: hs-crp in those without an indication for statin followed by targeted statin for patients with elevated hs-crp levels; S3: statin therapy at specified predicted risk thresholds without hs-crp Statin, antihypertensives and platelet aggregation inhibitor State Statin State Lifetime Health care Cost per 10 years Health care Cost per CVD event prevented; cost per 10, 20, and 30 years Health care Cost per Lifetime Health care perspective Cost per CVD cardiovascular disease, FRS Framingham Risk Score, hs-crp high-sensitivity C-reactive protein, LY life-year, quality-adjusted life-year, S screening strategy, SBP systolic blood pressure D. Epstein et al. Table 2 Health states included in the different models Number of health states Number of studies References Non-fatal health states Causes of death 2 1 Marshall et al. [20] Alive without CHD; Alive after CHD No fatal state 3 1 Johannesson [19] Alive without CHD; Alive after CHD Death 4 2 Rapsomaniki et al. [23] and Wald et al. [24] Alive without CVD; Alive after CVD CVD; OCM 6 2 Shiffman et al. [28], Alive without CVD; Alive after MI; Alive after stroke MI; Stroke; OCM Lee et al. [30] 6 1 Cobiac et al. [27] Alive without CHD; Alive after CHD; Alive after stroke Stroke; CHD; OCM 8 1 Kok et al. [22] Alive without CVD; Alive after MI; Alive after stroke; Alive MI; Stroke; CHD; OCM after other CHD 8 1 Blake et al. [18] Alive without CVD; Alive after MI; Alive after stroke; Alive after MI after stroke; Alive after stroke after MI MI; Stroke; OCM 11 1 Den Ruitjer et al. [29] Alive without CVD; Alive after first MI; Alive after second MI; Alive after stroke; Alive after hemorrhagic stroke; Alive after gastrointestinal bleeding 11 1 Pletcher et al. [21] Alive without CVD; Alive after MI; Alive after stroke; Alive after SA; Alive after MI after SA; Alive after stroke after MI; Alive after revascularization after SA 12 1 Lovibond et al. [26] Alive without CVD; Alive after MI; Alive after stroke; Alive after UA; Alive after SA; Alive after TIA 33 1 Choudhry et al. [25] States are combination of CVD events and complications, diabetes onset, myopathy, and VTE MI; Stroke; Hemorrhagic stroke; Gastrointestinal bleeding; OCM MI; Stroke; SA; OCM MI; UA; SA; Stroke; TIA; OCM MI; UA; Stroke; VTE; OCM CHD coronary heart disease, CVD cardiovascular disease, MI myocardial infarction, OCM other cause mortality, SA stable angina, TIA transient ischemic attack, UA unstable angina, VTE venous thromboembolism higher rate of death in the first year after a non-fatal CVD event, compared to subsequent years after the event). The authors of each study rarely justified why they chose the given model structure and neither did they acknowledge that alternative structures could be implemented. While additional states may allow greater accuracy to predict outcomes, it may be difficult to reliably estimate all the necessary parameters in a complex model. This gives rise to a trade-off between desirable model structure and reliable parameter estimation [34]. Even large epidemiological datasets may not have sufficient observations to give precise estimates of all the s in a complex model. Such modeling may produce unreliable results, and so validation is an essential part of the model-building process [35]. Question 3: What are the population and strategies of interest? A summary of the population and strategies evaluated in each article is shown in etable 2. Age is a risk factor for both CVD and competing non-vascular causes of death. Of the 13 studies, seven stratified the population by age [18, 19, 21, 22, 24, 26, 30] while the others estimated an average result across all age groups. A concern arises when comparing screening strategies based on risk scoring systems that include age as a risk factor for CVD, that age by itself is a strong non-modifiable risk factor, and therefore a strategy that treats patients above a fixed threshold of absolute risk will predominantly select older people. Risk scores such as the Framingham risk score (FRS) may assign the same absolute 10-year CVD risk to a young person with, say, multiple modifiable risk factors such as high cholesterol and hypertension, as an otherwise healthy older person with no modifiable risk factors [36]. Also, the absolute risk of CVD predicted from scores with age as a risk factor can be misleading as they do not take into account competing risks (i.e., the 10-year CVD risk is calculated as if other causes of death do not occur) and are therefore likely to over-estimate the true cumulative probability of CVD especially for older people. Stratifying the population into age groups, and evaluating the model separately for each of
Related Search
Similar documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks