This national projection of diabetes prevalence for the US is the first to model the projection on the number of individuals at high risk of developing diabetes using a multivariable risk assessment. Projections suggest a rising and substantial diabetes burden for the population. Hispanic adults will be most affected, with estimates suggesting that by 2031 more than 20% of the adult Hispanic community will have diabetes. These results are particularly worrisome for this community in light of recent evidence that the gap in healthcare quality between Hispanic and non-Hispanic White individuals has continued to widen [23
Many previous diabetes projections have been limited to estimates of diagnosed diabetes and thus have lower estimates of projected diabetes burden, and have not incorporated an evaluation of the population at high risk of diabetes, with clinical indicators, into their models [5
]. Our estimates will be less likely to be affected by changes in screening strategies. Additionally, they incorporate potential changes in the level of risk for diabetes in the US population, a change which is likely given national trends in obesity [3
]. Moreover, recent data have suggested that individuals with undiagnosed diabetes are similar to those with diagnosed diabetes with regard to the development of complications; thus our estimates are more robust in describing the burden of disease in the population [24
Comparing our projections with those from other studies, we note that an estimate, published in 2006, for diagnosed diabetes in the US among individuals aged 20 to 64 years in 2030 is 16.8 million [25
]. Our estimates are based both on diagnosed and undiagnosed diabetes, and our projection of total diabetes among that age group for 2031 is higher, namely 19 million. It is possible that estimates based solely on diagnosed diabetes could become more consistent with our estimates, if greater vigilance were shown for screening for undiagnosed diabetes. However, not accounting for the at-risk population in the estimates is likely to lead to inaccurate estimates. A comparison of our estimates of total diabetes with those of another study [7
], which projected total diabetes but did not account for the population at high risk of developing diabetes, reveals that the latter’s projections are most probably underestimates. Using data from 1993, the investigators projected a population prevalence estimate of total diabetes in the US among individuals aged 20 years and older for the year 2000 to be 7.6%, while the NHANES 1999–2002 yielded a prevalence of 8.8%. For 2025 the same team [7
] projected a prevalence of 8.9% versus 13.5% for 2021 in our study.
The results have several implications for the delivery of healthcare and healthcare financing.
First, we estimated our models under several assumptions for the number of individuals at high risk of diabetes in the population. Regardless of these assumptions, the US will have a substantial number of individuals at high risk of diabetes in 2011, 2021 and 2031. Interventions to modify lifestyle are critical to decrease the number of individuals at high risk, and consequently to lower the expected increase in diabetes in the future. Although some of the diabetes estimates suggest seemingly small decreases in future prevalence, based on decreases in the population at risk, the actual numbers are substantial. For example, a one-percentage point drop in the US population estimate of diabetes among individuals aged 20 and older in 2031 is quite substantial and would account for a decrease in prevalence of diabetes equivalent to 2,600,000 people.
Second, the projection that a substantial proportion of the population will have diabetes indicates greater spending will be necessary to manage the disease. This will include spending on drugs, ongoing monitoring, and treating of complications including nephropathy, retinopathy, and cardiovascular disease.
Third, the disproportionate impact of diabetes on minorities, particularly Hispanics, demands new intervention strategies to decrease the number of individuals at high risk and to deliver care to individuals who have historically had poor access to care. Additionally, with the projected increase in diabetes prevalence among 30 to 39-year-olds, a population not currently targeted for screening, a re-examination of current public health policy and screening strategies may be warranted [26
There are several strengths to the design of this study. One is that the study utilised multiple NHANES data sets, which have the advantage of allowing for nationally representative population estimates. Thus, the initial data used to fit the model as well as to make mortality estimates of diabetes, both diagnosed and undiagnosed, are nationally representative. Another strength is that this study is the first to make a nationally representative assessment of the at-risk population for development of diabetes and then use that assessment to model the future prevalence of diabetes. The assessment of risk used, moreover, is based on the ARIC diabetes risk score [18
], a multivariable risk score that used clinical indicators.
When interpreting our results, however, several limitations need to be considered. Thus, although this is the first study to use a validated diabetes risk score to assess the high-risk population for the development of diabetes for the entire US population, potential limitations exist with regard to the diabetes risk score. The ARIC diabetes risk score [18
] was based on a cohort of individuals aged 45 to 64 years at baseline and may therefore be limited when estimating diabetes development among individuals aged 20 years and older. However, we estimated diabetes prevalence in 10-year age increments. Moreover, the risk score’s moderate sensitivity and specificity may cause the model to under- or potentially overestimate future prevalence projections. Another possible limitation is that estimates of future disease burden are based on assumptions about the number at risk of disease and about mortality within the population. We have attempted to address this limitation by presenting the results of a sensitivity analysis, which includes variations in the proportion of the population at risk and in mortality. The third limitation is the diagnosis of diabetes in the NHANES data on the basis of a single FPG value. This strategy, although common in epidemiological studies, could potentially underestimate the prevalence of diabetes associated with isolated post-challenge hyperglycaemia, which occurs more commonly in women, the elderly, and in lean populations. It could also overestimate diabetes prevalence, because a clinical diagnosis of diabetes in asymptomatic patients requires two abnormal fasting glucose levels.
In summary, a continued focus on effective interventions for lifestyle modifications to decrease diabetes risk, as well as vigilant ascertainment of diabetes, appears crucial if the future prevalence and burden of diabetes in the US population are to be adequately addressed. This is especially important for minority populations, particularly the Hispanic community, which is projected to have an overwhelming future diabetes burden. Considering that minorities have historically had limited access to healthcare, these findings emphasise the importance of interventions targeting these populations.