Reach and Use of Diabetes Prevention Services in the United States, 2016-2017

Key Points Question What number and proportion of US adults are referred to and use diabetes preventive services? Findings In this nationally representative cross-sectional analysis of 50 912 survey respondents, 73.5% of those with diagnosed prediabetes and 50.6% of those with risk factors reported receiving any advice or referrals to reduce risk; of those advised, 35.0% to 75.8% of those with diagnosed prediabetes and 33.5% to 75.2% with risk factors reported engaging in various risk-reducing activities or programs in the past year. Those with a formal prediabetes diagnosis were more likely to receive advice or referrals by health care professionals, and those advised were more likely to engage in risk-reducing activities and/or programs. Meaning To improve reach and use of diabetes preventive services, expansion of some combination of programs (supply), awareness (demand), and access and referrals by health care professionals (linkage) appears to be needed.


Introduction
A number of national organizations in the United States are currently coordinating efforts to deliver type 2 diabetes (hereinafter referred to as diabetes) prevention services to attempt to curb the economic and disabling physical and psychosocial effects of the disease. 1,2 This effort is supported by robust evidence from the Diabetes Prevention Program (DPP) study and others, [3][4][5] which showed that lifestyle modification (LSM) and/or insulin-sensitizing medications among people at high risk for diabetes can delay progression to diabetes onset. Behavioral counseling for LSM (to eat fiber-rich foods, reduce calories, be physically active, and manage weight) offers the most sustainable, 6,7 most cost-effective, 8 and broadest benefits in terms of lowering cardiovascular risks, 9 disability, obstructive sleep apneas, retinopathy, and urinary incontinence. 10 Accordingly, the US Preventive Services Task Force and American Diabetes Association (ADA) recommend screening and multiplevisit lifestyle counseling to achieve weight loss and reduce diabetes risk in high-risk adults. 11,12 To facilitate scaling LSM programs, several interventions applying principles from the DPP trial have been tested in community, workplace, health care, and online settings and have been associated with at least modest benefits. [13][14][15] In addition, commercial and public payers in the United States, including Medicare, now pay for LSM programs delivered to people at high risk for diabetes.
Coverage for diabetes prevention services is based on data and aspirations that these programs will be high-value investments by lowering long-term costs of medications, use of health care services, and lost productivity associated with diabetes, 16,17 which currently total $327 billion annually. 1 Notably, models suggest that the value from diabetes-prevention LSM programs is sensitive to enrollment by those at risk of diabetes, risk level, attendance at program sessions, and achievement of lifestyle goals. 5,13,18 Also, awareness and internalization of one's risk are crucial motivators to engage in and maintain behavioral changes. Previous data 2 have shown that 88.4% of people with biochemically confirmed prediabetes (ie, those with elevated blood glucose levels not yet in the diabetes range) were unaware of their prediabetes.
Organized efforts to increase the availability of evidence-based programs online and in communities, health care settings, and workplaces have been aided by support and referral from clinical settings. Despite the initiation and momentum of these programs, few population-based data exist on their reach, implementation, and adoption. We herein assessed national progress in diabetes prevention and present findings using a prevention continuum diagram that facilitates identification of the most prominent gaps.

Data Source
The National Center for Health Statistics of the Centers for Disease Control and Prevention conducts the annual National Health Interview Survey (NHIS), the largest nationally representative crosssectional survey of noninstitutionalized US civilians. The National Center for Health Statistics Research Ethics Review Board approved data collection in the NHIS. All participants provided informed oral consent before participation, and all data were deidentified before analyses. This analysis was exempted from review by the National Center for Health Statistics because we used only publicly available, deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.

Data Collection
Details of the methods, survey instruments, data collection procedures, and analysis protocols of the NHIS are available online. 19 Trained interviewers administered questionnaires to obtain social, demographic, economic, and behavioral characteristics of respondents. Respondents also selfreported their medical history, including information about medications, height, and weight.

Study Population
We estimated proportions of respondents considered to be at high risk for diabetes using several definitions. Among those without self-reported diabetes, we classified diagnosed prediabetes as a positive response to the question, "Other than during pregnancy, have you ever been told by a physician or other health professional that you have borderline diabetes or prediabetes?" For those at high risk without a known prediabetes diagnosis (ie, those with a negative response to the previous question), we used the ADA composite score 22  hypertension, and lack of physical activity) (eTable 1 in the Supplement), which is highly sensitive for diabetes. 23 A score of 5 or higher (possible range, 0-11) was considered high risk for diabetes. We also reported the prevalence and numbers with a high ADA risk score with diagnosed prediabetes.

Statistical Analysis
We used SAS, version 9.3 (SAS Institute Inc) and SUDAAN, version 11.0 (Research Triangle Institute) software to account for the complex survey design. For all analyses, we calculated weighted percentages and standard errors or 95% CIs. To estimate the population size of those affected, we applied the weighted percentages to the July 1, 2017, US resident civilian and noninstitutionalized population estimates from the US Census Bureau.
We described the sociodemographic and clinical characteristics of each group. To assess the association of a prediabetes diagnosis with the actions of a health care professional and the likelihood of engagement in diabetes prevention activities, we stratified all analyses a priori by prediabetes diagnosis or elevated ADA risk score without a formal prediabetes diagnosis.
Because weight loss is 1 of the key LSM goals in diabetes prevention programs, we restricted analyses to overweight individuals (BMIՆ23.0 for Asian American respondents or Ն25.0 for all others), which also aligns with eligibility criteria for the National Diabetes Prevention Program. 24 We conducted sensitivity analyses to assess whether the addition of people with normal weight at high risk of diabetes changed our results.
We generated national diabetes prevention continuum diagrams-a tool used in health services research 25 -to assist with identification of gaps in delivery and uptake. To track how commonly health care professionals advised adults at high risk for diabetes to increase physical activity, reduce dietary fat or total calorie intake, or participate in weight loss programs or referred them to diabetes prevention programs in the past year, we calculated frequencies based on participant responses to survey questions (eTable 2 in the Supplement). Because variation often occurs in practice of what health care professionals choose or remember to emphasize, we also estimated what proportion of respondents reported receiving advice from their health care professionals for any of these activities. To assess engagement, we estimated proportions of respondents reporting that they reduced fat or total caloric intake, increased physical activity, participated in a weight loss program, or participated in a DPP-like intervention in the past year. Recognizing user variation in preferences and competing priorities, we conducted analyses to estimate proportions of respondents who engaged in any of the activities or programs in the past year, but we also examined each activity and program separately in sensitivity analyses.
We also examined whether engagement in any behavior to reduce diabetes risk, stratified by whether the participant was advised and/or referred by a health care professional, varied by year, age, sex, race/ethnicity, educational attainment, insurance status, history of GDM, hypertension, or BMI. To assess whether differences between groups were statistically significant, we used multivariate logistic regression models and presented adjusted prevalence estimates, using Wald F tests to assess whether associations were statistically significant. P < .05 was considered statistically significant for 2-sided tests. Unless otherwise indicated, data are expressed as percentages (SEs).
Among civilian, noninstitutionalized adults reporting no physician diagnosis of diabetes (Table 1), an estimated 80.0 million, or 36.0%, were at high risk of diabetes based on a physician diagnosis of prediabetes or an elevated ADA risk score. Of these, 17.9 million (22.2% of those at high risk or 8.0% overall) had diagnosed prediabetes, 73.3 million (91.9% of those at high risk or 33.3% overall) had an elevated ADA risk score, and 11.3 million (14.1% of those at high risk or 5.1% overall) had both. Among 145.5 million Americans with elevated BMI (overweight or obese) who would be eligible for diabetes prevention services, 68.1 million were considered at high risk based on reported prediabetes diagnoses (14.6 million) or an elevated ADA risk score (53.5 million).
Compared with US adults with elevated BMI who are not at high risk, adults with diagnosed prediabetes and those with ADA scores of at least 5 without a prediabetes diagnosis, were older (mean, Adults with elevated BMI who were at high risk for diabetes (by ADA risk score or diagnosed prediabetes) were 1.5 to 2.0 times as likely as those with elevated BMI but not at high risk to undergo

JAMA Network Open | Diabetes and Endocrinology
Reach  In sensitivity analyses that included all adults without diabetes (eTable 5 in the Supplement)-17.9 million with prediabetes and 62.1 million without prediabetes but with elevated ADA risk scores-

JAMA Network Open | Diabetes and Endocrinology
Reach In sensitivity analyses to estimate proportions of respondents who engaged in each specific activity or program separately and variation across sociodemographic and clinical characteristics (eTable 6 in the Supplement), we noted greater engagement among those with greater educational attainment, those who were middle aged, and those with insurance, a history of GDM, hypertension, or obesity. In addition, greater engagement appeared to occur among women and minority races/ ethnicities.

Discussion
These data provide, to our knowledge, the most comprehensive assessment of the degree to which US adults who are likely to benefit from diabetes prevention services are undergoing testing, counseling, and actual engagement in risk-reduction activities and/or programs. One-fifth of those with an elevated ADA risk score in 2017 reported a formal prediabetes diagnosis. Individuals with prediabetes diagnoses were more likely than those without to receive diabetes risk-reduction advice and/or referrals by health care professionals. Advice and/or referral by a health care professional was associated with a higher likelihood of participation. Overall, engagement in LSM programs designated for diabetes prevention was exceedingly low among all high-risk US adults.
The backdrop of health reforms 26 and large-scale efforts and investments directed toward diabetes prevention in the United States, 27 the United Kingdom, 28 and other countries accentuate the magnitude and importance of this issue. Our findings suggest that efforts to expand the supply of diabetes prevention LSM programs are, at least now, insufficient, not being matched by uptake, or both. Our population prevention continuum identified 3 specific gaps that are potentially modifiable.
First, 1 in 5 adults at high risk for diabetes reported a prediabetes diagnosis. Of the remaining 53.5 million who were overweight or obese with elevated ADA risk scores, 85% to 90% underwent a blood glucose test in the past 3 years. Therefore, people at high risk may not meet biochemical prediabetes thresholds, 23 or factors involving the test (eg, inaccuracy of laboratory findings), health care professional (eg, poor recognition, poor communication, or inaction), respondent (eg, recall bias), or some combination of these may be at play. This is important because our data suggest that prediabetes diagnoses may lead to a higher likelihood of counseling or referral by health care professionals.
Second, health care professionals were 2 to 3 times more likely to give general physical activity or dietary advice to patients than refer them to formal programs. Even among respondents with elevated BMI and diagnosed prediabetes, general advice about diabetes risk reduction (approximately 60%) was far more common than referral to a weight loss (approximately 20%) or a diabetes prevention (approximately 5%) program. Health care professionals may not believe in the effectiveness and cost-benefit of diabetes prevention programs, may be less aware of these programs, or may believe their patients would have less accessibility to programs. 29,30 Indeed, prediabetes constitutes a large heterogeneous group with varied risk levels. Because the motivating results from diabetes prevention trials were mostly observed in individuals with impaired glucose tolerance, health care professionals may be less inclined to refer people with lower-risk prediabetes subtypes. Furthermore, the proportions were even lower for adults without a prediabetes diagnosis, which may mean that health care professionals are more concerned about the long-term costs 31,32 and comorbidities 33 among people with biochemical evidence of prediabetes than among those without.
Third, among those advised or referred, sizeable gaps occurred between the 66% to 76% (or 40%-60% overall) of US adults at high risk of diabetes reporting some LSM in the past year compared with the 33% to 40% of those advised (or less than 10% overall) attending a formal program. Our study cannot assertively explain low levels of program engagement. Insurance coverage may influence engagement in diabetes risk-reducing activities, but insufficient power was available to examine whether coverage was associated with program engagement specifically.
Competing priorities may also be barriers to engagement, 34 especially among young adults and especially if participants do not perceive a benefit that outweighs the costs and time to participate in diabetes prevention programs. Comorbidities and higher self-perceived risk-notably, in the case of those with a history of GDM, hypertension, or obesity-appeared to motivate participation in riskreducing activities, especially among those not advised or referred.
Given that barriers associated with these gaps are likely multifactorial, a variety of concurrent implementation processes and supportive policies may increase the supply and coverage of programs, awareness (among those at risk and health care professionals), initial engagement, and Evidence from natural experiment studies show small associations between employer and health plan policies to facilitate diabetes prevention and engagement of adults at high risk for diabetes. 41 With regard to financial incentives, studies show associations between incentives directly given to respondents and better initial weight loss, but no improvement in maintenance of weight loss. 42,43 Studying what psychological, economic, and time preferences motivate people at high risk to engage remains an important area of study 37,38 and of high value to payers and organizations delivering LSM programs.
Despite data regarding the efficacy of metformin in reducing diabetes incidence, 3 14.5% of those with prediabetes reported using medications. Low use of metformin 44 and other medications to prevent diabetes may be associated with patient or health care professional preferences, concerns about adverse effects, and possibly consideration that medications to prevent diabetes likely will have to be taken in perpetuity because the effect wears off after discontinuation. 6 Previous national reports showing low achievement of diabetes care goals 45,46 were followed by subsequent improvements over time. 33,47 In addition, those with diagnosed diabetes tend to experience more aggressive treatment and achievement of care goals. 33 These findings highlight the value of this report as a benchmark from which to monitor future program availability and coverage, identification of prediabetes, referral, and retention as these programs mature. Furthermore, quality

JAMA Network Open | Diabetes and Endocrinology
Reach

Limitations
We used cross-sectional data that represented respondents' activities and recall at the time of the survey. If someone was diagnosed with prediabetes several years ago and made lifestyle changes then, these would not get enumerated in the surveys. Self-reported data, especially for height and weight, can be subject to recall or social desirability biases, which may have influenced estimates of those with elevated ADA risk scores. The ADA risk score itself does not have perfect associations with who will develop diabetes. Our definitions of high risk for diabetes may be overly sensitive, although our estimates of the total numbers of adults at high risk align with national data based on biochemical testing to confirm high risk for diabetes. 2,49 Respondents may have misinterpreted what was meant regarding diabetes programs in the questions posed. Our estimates of engagement in LSM programs for diabetes prevention do not reflect whether respondents attended a single session or completed the program because they were only asked about participation.

Conclusions
Our analyses used 2 recent waves of nationally representative surveys with large sample sizes offering a contemporary perspective on diabetes prevention in the United States. Our study offers a first assessment of participant-reported advice and/or referral by their health care professionals and engagement by adults at high risk using a national diabetes prevention continuum, demonstrating where the gaps occur, providing insights into possible policy and program actions, 26,50 and providing a benchmark for future population-level monitoring.