Cost-effectiveness of Diabetes Prevention Interventions Targeting High-risk Individuals and Whole Populations: A Systematic Review

Diabetes is a major global health issue. In 2019, there were an estimated 463 million adults aged 20–79 years with diabetes globally (∼9.3% of the population in this age-group), a figure that is projected to increase to 700 million by 2045 (1). Health care expenditures attributable to diabetes were estimated at $1.3 trillion in 2015 (2). Fortunately, type 2 diabetes (T2D), which accounts for 90–95% of the disease burden (3), can be prevented or delayed through nutrition and lifestyle changes as well as through pharmacologic interventions (4).

Approaches to prevent T2D fall under two categories: targeting individuals at high risk for developing T2D (high-risk approaches) and targeting the whole population regardless of the level of risk (population-based approaches). In general, high-risk individuals are those who have prediabetes (a health condition with a blood glucose level that is higher than normal but does not reach the level of diagnosed T2D) or who have risk factors for developing T2D, such as having a family history of T2D, being overweight or obese, being physically inactive, being 45 years old or older, or being a woman with a history of gestational diabetes mellitus (5). Interventions targeting high-risk individuals include screening for T2D in clinics and communities and providing lifestyle or pharmacologic interventions. On the other hand, population-based approaches aim to impact public health through policy implementation, campaigns, and other environmental change strategies. For example, imposing taxes on sugar-sweetened beverages (SSBs) has been proposed as a population-based approach to combat T2D and cardiovascular disease by the World Health Organization (6). Epidemiological evidence on the association between added sugars and T2D incidence and implementation experiences from Mexico and selected cities in the U.S. (Berkeley, for example) have led decision makers to explore the feasibility and effectiveness of scaling up such policies (7,8). Some experts suggest that the goal of reducing the number of new cases of T2D in the U.S. and worldwide is likely best achieved through approaches that combine both high-risk and population-based approaches (9,10).

T2D prevention approaches, whether high-risk or population-based, vary in terms of intervention effectiveness and cost. However, their cost-effectiveness (CE) has not been evaluated comprehensively or systematically. Most literature reviews to date have assessed the efficacy of T2D prevention approaches only without considering their CE or with focus on a single strategy (11–15). For example, one review and meta-analysis focused on nutrition education and examined the cost and CE of using diet modification as a T2D preventive intervention (16). Another systematic review measured the CE of T2D high-risk prevention approaches but focused on lifestyle interventions only (17). A recent study reviewed the CE of both lifestyle and metformin for T2D prevention among high-risk individuals but did not include population-based approaches (18). Another review evaluated both high-risk and population-based approaches (19); however, it did not examine key features of lifestyle interventions—such as intervention delivery mode and format—that might affect the CE outcome, and it only included fiscal policies among the population-based approaches. In addition, many new studies on the CE of T2D prevention interventions that have been published in recent years need to be evaluated in a review.

Here, we systematically review the CE of both high-risk and population-based approaches for T2D prevention. The goal is twofold: 1) to update evidence on high-risk approaches implemented in real-world settings, including whether to screen, whom to screen, and which formats are best for delivering lifestyle interventions (in-person vs. virtual, one-on-one vs. group, etc.) and 2) to synthesize evidence on population-based prevention strategies.

We searched the Cumulative Index to Nursing and Allied Health Literature (CINAHL), Cochrane databases, Excerpta Medica (EMBASE), Medical Literature Analysis and Retrieval System Online (MedlinePlus), PsycINFO, Scopus, and Sociological Abstracts (Soc Abs) to identify original economic evaluations of approaches to prevent T2D published in English from January 2008 to July 2017. Search keywords included 1) diabetes, impaired glucose tolerance, and insulin resistance; 2) expenditure, health care cost, and cost of illness; 3) quality-adjusted life year (QALY), disability-adjusted life year (DALY), and incidence of diabetes; and 4) cost-effectiveness analysis (CEA), cost-utility analysis, cost-benefit analysis, and economic evaluation (20). In addition to searching the seven databases above, we manually screened the reference lists of all included studies as well as the table of contents of major diabetes journals (Diabetes Care, The Lancet Diabetes & Endocrinology, Diabetologia, and Diabetes Research and Clinical Practice) during the search period.

Following the Cochrane Collaboration’s protocol for systematic reviews (21), two people independently reviewed each study for inclusion/exclusion in our review, quality assessment, and data abstraction. We focused on three types of economic evaluations of high-risk approaches to T2D prevention: CE, cost-utility, and cost-benefit analyses. We included studies that reported quantitative measures for the CE outcomes. The outcome was the incremental cost-effectiveness ratio (ICER), which is in the form of cost-per-additional QALY gained or cost-per-additional DALY averted.

To assess the quality of included studies, we used a tool based on The BMJ authors’ guide for economic studies (22), which was used previously (20). In brief, the tool assesses each study based on 13 attributes: sources of cost data, sources of benefit data, categories of cost data, categories of benefit data, analytical time horizons, study perspectives, model descriptions, structure diagrams, currency and year of the costs, discounting factor for costs, discounting factor for benefits, ICERs, and sensitivity analyses. Each attribute was given one point—an equal weight—if the study clearly stated it. We included studies with a quality score of seven and above (20).

We abstracted the following information from each study: publication information, study objective, prevention approach, comparison, target population, delivery method, provider, analytical time horizon, study method, perspective of the evaluation, and results. We adjusted ICERs and costs to 2017 U.S. dollars using the Consumer Price Index (23). For studies conducted in countries other than the U.S., we used the annual exchange rate from the Federal Reserve Bank to convert the foreign currencies into U.S. dollars before adjusting them for inflation (24). In rare cases where the study did not report the specific year of currency used to calculate costs, we assumed the costs were calculated 1 year before the publication date. Studies were considered cost-effective if the ICER was below the $50,000/QALY threshold (25).

We grouped high-risk approaches into four categories based on their study objectives: 1) articles focused on deciding whether to screen for prediabetes, 2) articles determining the target population for screening that would generate the optimal CE outcomes, 3) articles evaluating the CE of specific T2D prevention interventions, and 4) articles evaluating the CE of managing gestational diabetes mellitus.

To better understand what features contribute to the CE outcomes of prevention interventions, we examined interventions from the third category above (those evaluating the CE of specific T2D prevention interventions) and summarized the median and range of ICERs for interventions sharing similar features in terms of how the intervention is delivered (i.e., whether delivered one-on-one or in a group and whether conducted in-person or via virtual media, such as internet or mobile applications) and by whom (i.e., whether taught by health care providers or lay health workers, such as trained community health workers or diabetes educators). The high-risk approaches included lifestyle interventions (translational Diabetes Prevention Program [DPP] and translational non-DPP) and pharmacologic interventions (metformin). Translational DPPs refer to nutrition and physical activity interventions that follow the DPP curriculum that translated to the real world, such as those provided in the community or primary care setting. In contrast, translational non-DPPs are lifestyle interventions that do not strictly follow the DPP curriculum.

We modified our study protocol to accommodate the methods and results reported in studies on population-based approaches because many of them did not use the standard framework for assessing CE due to a lack of data. For study screening, we included population-based interventions if they reported ICERs or if they compared costs given a certain level of benefits if benefits were measured as T2D cases prevented or QALY due to reduction in diabetes. Consequently, the result of cost-saving (CS) for population-based interventions should be interpreted with caution, as it could refer to a reduction in health costs only rather than savings as measured by ICER, which is a negative incremental cost.

Quality assessment for population-based approaches was less restrictive and reduced to nine scoring attributes (the other four pertained to formal CEA and did not apply in these cases): sources of cost data, sources of benefit data, categories of cost data, categories of benefit data, analytical time horizons, study perspectives, model descriptions, currency and year of the costs, and results. Again, we included studies with a quality score of seven and above.

For selected studies, we abstracted data on publication information, objective, prevention strategy, comparison, target population, analytic time horizon, study method, the perspective of the evaluation, and results. We then grouped population-based approaches into four categories and summarized CE of each one: 1) implementing fiscal policy, 2) implementing a regulation, 3) promoting health by education and information, and 4) changing the built or food environment.

Figure 1 shows the 39 studies that met our inclusion criteria: 28 articles on high-risk approaches and 11 articles on population-based approaches.

Table 1A shows studies arranged chronologically and then alphabetically by the last name of the first author within each category (26–53). Among these studies, the analytic time horizon ranged from 1 year to a lifetime. Studies were evaluated from either a societal perspective or a health care perspective. Most studies discounted costs and benefits at 3%. While most of the studies were based on simulation modeling, eight studies assessed prevention strategies using randomized controlled trials. Results indicate that screening for prediabetes and providing interventions, either lifestyle or pharmacologic interventions, is either cost-effective or CS among individuals with a high risk of T2D. The conclusion held for both the societal and health care perspectives and for shorter or longer time horizons.

Seventeen studies (or study arms) evaluated the CE of specific interventions compared with no intervention (status quo or placebo) from a health care system perspective (Table 2) (35–37,39,40,42–44,46,49–51). Results indicate that all interventions were cost-effective, but the magnitude of the ICERs differed by intervention features. Lifestyle interventions were more cost-effective than metformin interventions, regardless of analytical time horizon, delivery method, media, mode, and personnel type. Among lifestyle interventions, translational DPP was more cost-effective than translational non-DPP prevention approaches. The median ICER for translational non-DPP was twice as high as that for translational DPP. Analytical time horizon also affects CE outcomes; studies evaluated over a longer time horizon have a lower ICER. Among lifestyle interventions, in-person interventions had slightly better CE outcomes than virtual interventions. The median ICERs for interventions delivered in groups and for interventions provided by a combination of health professionals and trained lay health workers were less than half of those for the one-on-one interventions or interventions provided by health professionals alone.

Table 1B describes the 11 studies that evaluated 28 population-based approaches to preventing T2D (54–64). Some studies appear in more than one category because they evaluated multiple interventions that were applied to different categories. All studies were evaluated at 10 years or longer. More than half of these studies (or study arms) assessed the CE of two fiscal policies—SSB taxation and fruit and vegetable subsidies. Among the nine studies (or study arms) that evaluated the CE of SSB taxation, the most common taxation rate was 20% of the total amount paid. All nine studies used computer-simulation models and used effectiveness outcomes from published articles. Two studies used a governmental perspective while the other seven used a health care perspective. All nine studies found the SSB tax to be CS. The included studies also evaluated a sugar tax, a fruit and vegetable subsidy, and a combination of taxing unhealthy foods and subsidizing healthy foods and found large variations in CE outcomes. For nonfiscal policy interventions, such as a walking group in the community, opening supermarkets to increase food access, and increasing healthy food options in the workplace, most of the interventions were cost-effective or CS from the health care system perspective. However, the CE results were inconsistent from the societal perspective. In addition, many of these interventions were only evaluated by one study, such that we were unable to make a definite conclusion on the CE of these interventions.

Table 3 summarizes the CE of population-based approaches. The SSB tax was found to be CS from the health care system and governmental perspectives. The four studies (or study arms) that evaluated the CE of subsidies for fruits and vegetables found mixed results, from more costly with no net health outcomes benefits to CS. Similarly, the five studies that evaluated community-wide interventions also found them to have various CE outcomes from the health care system and societal perspective. Interventions of incentive programs and environmental change were cost-effective from the health care system perspective.

Our systematic review assessed the CE of approaches for preventing T2D from 39 studies. Three key findings emerged. First, the ICERs of most of the high-risk approaches were well below the range that is generally considered to be cost-effective. Importantly, differences between delivery methods were small, and the group-delivered translational DPP provided by a combination of health professionals and trained lay health workers seemed most cost-effective. Our findings reinforced the fact that interventions to prevent T2D among high-risk individuals are highly cost-effective and practical in any given setting. Second, implementing a population-wide SSB tax was CS and has the potential to benefit a large population. SSB taxation can be considered as an important population-based policy approach to prevent T2D globally. Third, although there were many proposed population-based interventions (including subsidies for fruits and vegetables, health promotion approaches, and environmental changes), the CE of these interventions needs further investigation with real-world data in order to draw a conclusion.

Our findings for high-risk approaches are consistent with previous literature in that lifestyle programs utilizing the translational DPP curriculum are somewhat more cost-effective than lifestyle interventions that do not follow the DPP curriculum (17). The translational DPP lifestyle program is widely used in the Centers for Disease Control and Prevention (CDC)-led National Diabetes Prevention Program (National DPP)—a U.S. translational program providing a framework and infrastructure for targeting high-risk individuals, and this program is covered by several commercial and public insurers (65). For example, the Centers for Medicare & Medicaid Services began covering the CDC-recognized DPP lifestyle change programs in 2018 for Medicare beneficiaries (66).

A noteworthy change in the high-risk approach category is the adoption of virtual media for intervention delivery. In recent years, virtual media interventions have become available via online counseling calls, emails, and text messages (29,30,39,49). One benefit of virtual media is that it reaches individuals who have barriers to in-person interventions, such as the elderly and people who live in rural areas. People may take advantage of virtual media interventions to save time and travel expenses (67). Virtual media interventions also allow participants to access the program any time and with a greater frequency (68). Our review found that few studies evaluated the CE of interventions delivered virtually. The results from this limited evidence show that virtually delivered programs were cost-effective but not as cost-effective as the in-person lifestyle program as measured by cost per QALY. Additionally, more rigorous studies are needed to assess the CE of virtually delivered programs.

The results of our review also demonstrate great potential for population-based interventions to prevent T2D (69). Among fiscal policies, taxing SSBs may be a better approach than subsidies for healthy foods for two main reasons: 1) tax policies generated better CE outcomes and 2) evidence supporting tax policies was stronger as multiple studies collectively reached a consistent conclusion. From a health care system perspective, SSB taxes would be CS. The SSB tax would reduce SSB consumption at zero or little health intervention costs and would also reduce health care spending. The nine studies in our review showed how much health care costs would be saved from SSB taxation. In addition, these studies showed that such an intervention would be CS or cost-effective from the governmental perspective. On the other hand, the ICERs of interventions to promote the consumption of fruits and vegetables ranged widely. These studies differed in features that would change the results, such as the targeted population (general population vs. participants in the Supplemental Nutrition Assistance Program [SNAP]), analytical time horizon (10 years vs. lifetime), and study perspective (governmental, health care system, or societal). Although evidence indicates that an SSB tax could be a CS intervention to prevent T2D, there are political and other considerations that impact its implementation in the real world (70). The uptake of that strategy is dependent on state and local decision-making (70–72).

Our study is one of the first to include articles evaluating the CE of population-based approaches to prevent T2D in a systematic review. The adoption of population-based approaches could have great potential for improving population health. A recent analysis found that only 3.1% of U.S. adults without T2D (regardless of prediabetes status) met T2D risk reduction lifestyle goals in 2007–2012 (73), suggesting the need for broader public health efforts to reach the majority of the U.S. population for reducing their risk of T2D. Individuals at high risk for T2D could benefit from population-based prevention efforts in conjunction with targeted, high-risk approaches. For those who have not been screened for T2D, population-based interventions may also slow their progression to T2D and provide other health benefits from better nutrition and more physical activity (9).

Our findings on the CE of both high-risk approaches and population-based approaches indicate that investing in T2D prevention is an efficient use of limited health care and societal resources. Since the development of T2D is a result of a combination of multiple risk factors including genetics, environment, and behaviors, a combined strategy of both high-risk and population-based approaches may be the best one to achieve optimal outcomes of T2D prevention (9,10). Interventions targeting high-risk individuals are effective and cost-effective among individuals at risk for T2D. However, the low uptake and resource-intensive nature of high-risk approaches limit their application. In contrast, while population-based approaches use “upstream” approaches that reach a broader population, their impact at the individual level is weaker, and the evidence of their effectiveness is more limited.

Based on our review, we suggest two avenues for the future economic evaluation of T2D prevention approaches. The first is to conduct rigorous CEAs using real-world data on population-based interventions. The studies in this review generated considerable variation in CE, indicating uncertainty about the CE of these interventions. Many studies are based on simulation modeling. Although high-quality simulation models can generate reliable results, they rely on strong assumptions that may or may not be reflected in reality. In contrast, data from empirical studies—natural experiments, for example—are directly observed and reflect the “true” behavioral change of the population to interventions. Although such studies usually last for a couple of years, they are often the foundation for modeling studies. Additional research that evaluates the impact of taxes, subsidies, food labeling, and other approaches that are already implemented (“natural experiments”) are needed to obtain stronger data. Second, effective and cost-effective population-based approaches are needed for both developing and developed countries. Although we found a disproportionate imbalance in the number of studies published involving high-income countries, population-based approaches are strategies to reach a large scale of the population to address the dramatic increase in diabetes prevalence worldwide.

Conclusions from this review need to be interpreted with caution. First, most of the evaluations, especially population-based approaches, utilized simulation modeling, which can be heavily influenced by assumptions. Unlike data from clinical trials, which are directly observed, model data are usually from published articles. Even though many models used data from clinical trials for the initial years of interventions, they must make assumptions on the persistence of costs and effectiveness beyond the trial study period to simulate a longer time horizon. Because of these contraints, in our review, we tried to rely on evidence if it was consistent across multiple modeling studies. Second, in order to include as many studies on population-based interventions as possible, we used somewhat “looser” quality criteria for these studies. Many of the population-based approaches did not conduct formal CEA. As a result, the CS results from these studies needs to be better understood, as they were a simple comparison of costs given a certain level of health benefit. Also, many of the CE results were estimated from governmental or health care system perspectives rather than a societal perspective. Third, the societal perspective defined in population-based approaches was not as inclusive as it was for high-risk approaches. Some cost categories were not included in the societal perspective, such as productivity loss or time cost. Fourth, we compared the CE of interventions based on the median ICERs without explicitly considering other study information, such as the evaluation method and rigorousness of data. This comparison follows previous literature (17) but may not reflect a real difference in CE. Fifth, our results provide information for decision makers to choose among interventions based on CE criteria only. Many other issues such as health equality, acceptability, and feasibility should also be considered in real-world decision-making.

Evidence from our review indicates that investing in T2D prevention, using either high-risk approaches or population-based approaches, is an efficient use of health care and societal resources. Given the enormous cost associated with T2D, if health care resources are limited, then prevention is a highly efficient use of such resources. Interventions targeting high-risk individuals with group-delivered translational DPP lifestyle intervention, provided by a combination of health professionals and trained lay health workers, was more cost-effective compared with one-on-one interventions provided by health professionals solely; however, all interventions targeting high-risk individuals were cost-effective. Among population-based approaches, the SSB taxation saves costs and resources of the health care system and government. Therefore, expansion of insurance-covered, professional, and lay-delivered group DPPs with a simultaneous institution of SSB taxation can be considered as a priority to stem the rising tide of T2D. A combined approach that targets both high-risk individuals and the whole population could be a policy choice for preventing T2D in the U.S. and probably in other high-income countries.

Acknowledgments. This work is a collaboration between the Centers for Disease Control and Prevention and the American Diabetes Association. The authors thank the external and internal reviewers for their valuable comments during the review process. The authors thank Rui Li (CDC) for generously sharing materials from her previous review and providing guidance, William Thomas (CDC) for his timely help with the literature search, and Clarice G. Conley (CDC) for her editorial assistance.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. X.Zho. and P.Z. designed the research. X.Zho. analyzed data, interpreted results, and drafted the manuscript. K.R.S. made a critical revision of the manuscript. X.Zho., K.R.S., B.P.N., S.J., and K.K.P. screened studies and abstracted data. B.P.N., X.Zha., A.L.A., and P.Z. provided important intellectual content to the manuscript.