Impact of changes in diabetes screening guidelines on testing eligibility and potential yield among adults without diagnosed diabetes in the United States

Abstract

Aims:

Recent USPSTF and ADA guidelines expanded criteria of whom to test to identify prediabetes and diabetes. We described which Americans are eligible and report receiving glucose testing by USPSTF 2015 and 2021 as well as ADA 2003 and 2022 recommendations, and performance of each guideline.

Methods:

We analyzed cross-sectional data from 6,007 non-pregnant U.S. adults without diagnosed diabetes in the 2013–2018 National Health and Nutrition Examination Surveys. We reported proportions of adults who met each guideline’s criteria for glucose testing and reported receiving glucose testing in the past three years, overall and by key population subgroups,. Defining prediabetes (FPG 100–125 mg/dL and/or HbA1c 5.7–6.4 %) or previously undiagnosed diabetes (FPG ≥ 126 mg/dL and/or HbA1c ≥ 6.5 %), we assessed sensitivity and specificity.

Results:

During 2013–2018, 76.7 million, 90.4 million, 157.7 million, and 169.5 million US adults met eligibility for glucose testing by USPSTF 2015, 2021, and ADA 2003 and 2022 guidelines, respectively. On average, 52 % of adults reported receiving glucose testing within the past 3 years. Likelihood of receiving glucose testing was lower among younger adults, men, Hispanic adults, those with less than high school completion, those living in poverty, and those without health insurance or a usual place of care than their respective counterparts. ADA recommendations were most sensitive (range: 91.0 % to 100.0 %) and least specific (range: 18.3 % to 35.3 %); USPSTF recommendations exhibited lower sensitivity (51.9 % to 66.6 %), but higher specificity (56.6 % to 74.5 %).

Conclusions:

An additional 12–14 million US adults are eligible for diabetes screening. USPSTF 2021 criteria provide balanced sensitivity and specificity while ADA 2022 criteria maximize sensitivity. Glucose testing does not align with guidelines and disparities remain.

1. Introduction

Diabetes is a common, chronic, and burdensome metabolic disorder that is associated with extensive morbidity, health care costs, [1] and both higher diabetes-specific mortality and death from other causes (e. g., infections) [2]. An estimated 37.3 million Americans have diabetes, of which 28.7 million (76.9 %) are diagnosed [3]. An additional 96 million have prediabetes, a metabolic state characterized by elevated fasting or 2-hour glucose or glycated hemoglobin levels, yet do not meet diagnostic thresholds for diabetes. Adults with prediabetes are 4–12 times more likely than the general public to progress to type 2 diabetes over the ensuing 5–25 years [4,5] and are at higher risk for cardiovascular disease, chronic kidney disease, and mortality [6,7].

Many randomized trials have demonstrated that engaging in preventive health behaviors [8,9] and management of risk factors for diabetes complications (such as glycemia, blood pressure, lipids, and avoiding tobacco) [10] can lower the incidence of type 2 diabetes, as well as morbidity and mortality associated with diabetes, [11] respectively. Identifying people with prediabetes or type 2 diabetes is a critical step in linking them with preventive and care services; national data have shown that those who are aware of their diagnoses are more likely to engage in these services [12,13]. To maximize efficiency and feasibility, targeted rather than universal screening is endorsed in the context of US healthcare [14]. Given the continued growth in numbers of US adults affected by diabetes, over the past two decades, expert groups have updated guidelines regarding whom to test and at what frequency. Of note, the American Diabetes Association (ADA) recommendations for adults have included a wider array of diabetes risk factors since 2003 [15] and recently lowered the age eligibility criteria to 35 years for all adults [16]. The US Preventive Services Task Force (USPSTF) guidance has been updated in 2008,[17] 2015, [18] and 2021.[19] The multitude of guidelines creates confusion in clinical practice,[20] and thus may limit the extent to which screening is implemented. Previous analyses show that 23.7 % and 73.0 % of US adults without diagnosed diabetes met USPSTF 2008 and ADA 2003 recommended eligibility for glucose testing, and 51 % reported receiving tests.[21] Furthermore, these same analyses highlighted inequities in receipt of glucose testing despite the fact that lower socioeconomic and underrepresented minority subgroups of the population tend to be at higher risk of developing diabetes.

Since the ADA and USPSTF are the two most commonly-used recommendations for glucose testing and management in clinical care, we used nationally-representative data to assess and compare who would be eligible for targeted screening if clinicians use the ADA 2022 versus 2003 or the USPSTF 2021 versus 2015 guidelines (Supplementary Appendix 1). We estimated proportions of adults who reported receiving or not receiving glucose testing, whether these screening practices aligned with the guidelines, and which subgroups of the population were most or least likely to receive testing. We also evaluated, among those without known diabetes, how well the screening guidelines perform with regard to sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). These data offer insights into current diabetes screening practices nationally and likely yields associated with different guidelines.

2. Methods

2.1. Data source

We analyzed data from the 2013–2018 National Health and Nutrition Examination Surveys (NHANES), conducted by the Centers for Disease Control and Prevention’s National Center for Health Statistics (NCHS). NHANES uses a complex, multistage probability sampling design such that weighted estimates are nationally-representative of the non-institutionalized US civilian population.[22,23] NHANES survey instruments and protocols have been described extensively.[22,23] Briefly, NHANES collects survey data, anthropometric measurements, and biospecimens from participants and releases data in 2-year cycles. Over the 2013–2018 survey period, the 2-year examination response rates were 68.5 %, 58.7 %, and 48.8 %.

NHANES received approval for human subjects research from the Centers for Disease Control and Prevention NCHS Ethics Review Board.

2.2. Study participants

This analysis included adult participants (aged ≥ 18 years) from the morning examination session who had fasted 8-<24 h and reported no previous diagnosis of diabetes. We excluded pregnant people (n = 57), and those missing data regarding self-reported receipt of glucose testing (n = 192) or glycemic measures (n = 4). In total, our analysis included 6,007 participants.

2.3. Exposure definitions

For all guidelines, we identified characteristics of adults who were eligible for glucose testing (Supplementary Appendix 2).

Per USPSTF guidelines, those eligible for glucose testing must have body mass index (BMI [weight in kilograms divided by height in meters squared]) ≥ 25 kg/m² (or ≥ 23 kg/m² for Asian Americans) and be 35–70 years (2021)[19] or 40–70 years of age (2015).[18].

The ADA guidelines[15] recommended glucose testing at age ≥ 45 years (2003) which has been revised to ≥ 35 years (2022) or at age ≥ 18 years with a BMI ≥ 25 kg/m² (or ≥ 23 kg/m² for Asian Americans) and another risk factor for diabetes. Diabetes risk factors and their definitions in NHANES included: physical inactivity (no work-, transport-, or recreation-related physical activity in a typical week); family history of diabetes in a first-degree relative; minority race/ethnicity as those that are not non-Hispanic White (i.e. non-Hispanic Black, Hispanic, non-Hispanic Asian, American Indian or Alaska Native); history of gestational diabetes or delivering a macrosomic (>9lb) baby; hypertension (blood pressure ≥ 140/90 mmHg or use of antihypertensive medication); dyslipidemia (HDL-cholesterol < 35 mg/dL, triglycerides > 250 mg/dL, or lipid-lowering medication use); self-reported diagnosis of prediabetes (impaired fasting glucose, impaired glucose tolerance, or borderline diabetes); and self-reported history of vascular disease (myocardial infarction, coronary heart disease, or stroke). NHANES did not collect data on polycystic ovary syndrome.

For each of the guidelines, we identified individuals who met criteria for glucose testing. If an individual did not meet guideline-specific criteria, then that individual was considered ineligible for glucose testing based on that specific guideline. We also identified individuals who did not meet any guideline criteria for glucose testing.

2.4. Outcome definitions

We defined self-reported receipt of glucose testing as an affirmative answer to the question, “Have you had a blood test for high blood sugar or diabetes within the past three years?”

We estimated performance of each guideline among those without known prediabetes and diagnosed diabetes. To estimate sensitivity (the proportion with prediabetes or diabetes who would be eligible according to each set of screening criteria), specificity (the proportion of those free of prediabetes and diabetes who are classified as ineligible by screening criteria), PPV (proportion of those classified as eligible for glucose testing that actually have prediabetes or diabetes), NPV (the proportion of those classified as ineligible for glucose testing that are actually free of prediabetes and diabetes), and area under the curve (AUC; based on the Mann-Whitney test),[24] we defined prediabetes as fasting plasma glucose (FPG) 100–125 mg/dL[25] or hemoglobin A1c (A1C) 5.7–6.4 %; in additional analyses, we used a confirmatory definition where both FPG and A1C thresholds had to be met.[26] We defined previously undiagnosed diabetes as FPG ≥ 126 mg/dL or A1C ≥ 6.5 %; here again, we also examined a confirmed definition of FPG ≥ 126 mg/dL and A1C ≥ 6.5 %. To ensure consistency of FPG across survey years, we applied regression equations recommended by NCHS.[27-29].

2.5. Statistical analysis

We used SAS, version 9.4 (SAS Institute, Cary, North Carolina) and SAS-callable SUDAAN, version 11 (Research Triangle Park, NC) for statistical analyses, accounting for NHANES’ complex survey design, including specific weights for the fasted subsample. We determined the proportion of adults reporting receipt of glucose testing in the past three years by eligibility based on each guideline using cross-tabulations (eligible and tested, eligible and not tested, ineligible and tested, ineligible and not tested). We described the characteristics of mutually exclusive groups according to their eligibility based on each guideline and self-reported glucose testing status. We tested for significant differences in proportions by eligibility status using chi-square or Wald F tests. Finally, we assessed the performance (sensitivity, specificity, PPV, and NPV) of screening guidelines to identify prediabetes and undiagnosed diabetes, including a sensitivity analysis that excluded adults with self-reported prediabetes.

2.6. Role of the funding source

This study received no external funding.

3. Results

Our analysis included 6,007 participants, representing 206.7 million US adults without diagnosed diabetes. During 2013–2018, using the ADA 2003 and 2022 recommendations, 157.7 million and 169.5 million US adults without diagnosed diabetes, respectively, would be eligible for glucose testing (Fig. 1; Supplementary Appendix 2). Using the USPSTF 2015 recommendations, 76.7 million adults would be eligible for glucose testing; 75.2 million of these individuals would also be eligible by ADA 2003 guidelines. With the introduction of the USPSTF 2021 guideline, 90.4 million would be eligible for glucose testing; 87.1 million of these were eligible by ADA 2003 guidelines, and 3.4 million were eligible by USPSTF 2021 guidelines alone. Based on ADA 2022 guidelines, an additional 8.4 million adults who would not be eligible based on ADA 2003 guidelines are now eligible for glucose testing.

Fig. 1.

Populations eligible for diabetes or prediabetes testing by recommendation among US adults aged ≥ 18 years without diagnosed diabetes (N = 206.7 Million), 2013–2018 The estimated numbers of adults eligible for diabetes or prediabetes testing by guideline were as follows: (A) ADA 2003, orange circle: 157.7 million; (B) USPSTF 2015, green dashed circle: 76.7 million; (C) USPSTF 2021, green circle: 90.4 million; (D) ADA 2022, blue circle: 169.5 million. Population sizes may not sum to total due to rounding.

Compared to participants who were pan-eligible for glucose testing based on ADA 2003 and 2022 and USPSTF 2015 and 2021 guidelines (Table 1), those who were ineligible or eligible based on ADA 2022 recommendations alone were younger (mean age 25.0 [ineligible] and 39.0 [ADA 2022 alone] vs 51.9 years [pan-eligible]), less diverse (28.0 % [ineligible] and 26.1 % [ADA 2022] vs 38.7 % not non-Hispanic White [pan-eligible]), had lower mean BMI (23.7 [ineligible] and 22.4 [ADA 2022] vs 31.9 kg/m² [pan-eligible]), lower waist circumference (83.9 [ineligible] and 83.4 [ADA 2022] vs 106.4 cm [pan-eligible]) and lower mean systolic blood pressures (112 [ineligible] and 111 [ADA 2022] vs 125 mmHg [pan-eligible]). Those ineligible and eligible based on ADA 2022 alone, compared to those who were eligible by all guidelines, were also more likely to be uninsured (19.9 % [ineligible] and 26.6 % [ADA 2022] vs 14.1 % [pan-eligible]), more likely to have no usual place of care (31.2 % [ineligible] and 24.7 % [ADA 2022] vs 14.5 % [pan-eligible]), and had no visits to a doctor in the past year (23.0 % [ineligible] and 24.2 % [ADA 2022] vs 15.9 % [pan-eligible]).

Table 1.

Characteristics of US adults aged ≥ 18 years without diagnosed diabetes by eligibility status ^a, 2013–2018.

	Total		Eligible based on ADA and USPSTF guidelines b		Eligible based on ADA 2022 alone, but not USPSTF guidelines		Ineligible based on all guidelines c
Unweighted n	6007	SE	2580	SE	240	SE	886	SE
Population size, millions	206.7		87.1		8.4		37.2
Age, % (SE) d
18–44 years	50.3	1.0	27.8	1.4	100.0		100.0
45–64 years	33.1	0.9	58.2	1.4	0.0		0.0
≥65 years	16.6	0.8	14.0	1.2	0.0		0.0
Age (years), mean (SE) d	45.3	0.4	51.9	0.4	39.0	0.2	25.0	0.2
Male, % (SE) d	48.4	0.6	50.5	1.0	40.9	3.7	53.4	2.3
Race-ethnicity, % (SE) d
Hispanic	15.6	1.5	17.3	1.8	13.9	2.4	11.9	1.6
Non-Hispanic Asian	5.7	0.6	5.5	0.7	9.9 e	1.7	4.7	0.7
Non-Hispanic Black	11.3	1.1	12.3	1.2	8.6 e	1.8	8.8	1.3
Non-Hispanic White	63.6	2.0	61.3	2.3	63.9	3.9	72.0	2.6
Education, % (SE)
< High School	14.5	1.0	14.4	1.1	14.7	2.9	12.7	1.8
High School	24.2	1.2	23.4	1.3	19.3	3.8	26.2	2.1
> High School	61.4	1.8	62.1	1.9	65.9	4.9	61.0	2.8
At or below poverty, % (SE)	14.2	0.9	12.9	1.0	13.6	2.6	16.7	2.0
Uninsured, % (SE) d	16.7	1.1	14.1	1.0	26.6	4.0	19.9	1.9
No usual place of care, % (SE) d	19.8	0.8	14.5	1.0	24.7	3.6	31.2	1.6
No doctor visit in past year, % (SE) d	18.2	0.7	15.9	1.1	24.2	3.2	23.0	2.2
BMI, kg/m2 mean (SE) d	28.9	0.2	31.9	0.2	22.4	0.1	23.7	0.3
Waist circumference, cm mean (SE) d	98.6	0.4	106.4	0.4	83.4	0.6	83.9	0.7
SBP, mmHg - mean (SE) d	121.2	0.3	124.7	0.5	111.3	0.9	112.1	0.5
DBP, mmHg - mean (SE) d	70.1	0.3	73.3	0.3	69.6	0.7	65.6	0.4
FPG, mg/dL - mean (SE) d	97.8	0.3	101.4	0.4	93.0	0.8	91.0	0.4
A1c, % mean (SE) d	5.44	0.01	5.58	0.01	5.25	0.01	5.13	0.01
Self-reported prediabetes, % (SE) d	11.2	0.5	16.0	1.0	23.4	3.6	14.0	1.6
Lab-defined Prediabetes, % (SE) d	40.7	1.2	53.9	1.4	6.2 e	2.1	1.9	0.5
Lab-defined Diabetes, % (SE) d	3.4	0.2	5.2	0.4	– e	– e	– e	– e

Abbreviations: SE, standard error; BMI, body mass index; cm, centimeters; SBP, systolic blood pressure; DBP, diastolic blood pressure; FPG, fasting plasma glucose; A1c, glycated hemoglobin.

Lab-defined prediabetes defined as fasting plasma glucose 100–125 mg/dL and/or hemoglobin A1c [HbA1c] 5.7–6.4 %.

Lab-defined diabetes defined as fasting plasma glucose ≥ 126 mg/dL and/or hemoglobin A1c ≥ 6.5 %).

^aData not shown for 49 participants who were eligible for screening based on ADA 2022 and USPSTF 2021 but not ADA 2003, representing about 3.4 million US adults.

^bEligible based on ADA 2003 or 2022 and USPSTF 2015 or 2021 guidelines.

^cIneligible based on ADA 2003, ADA 2022, USPSTF 2015, and USPSTF 2021 guidelines.

^dSignificant difference by eligibility status based on chi-square or Wald F test; p < 0.05.

^eUnreliable data (defined by relative standard error > 30 % of estimate) not shown.

During 2013–2018, when asked about receiving glucose testing in the past three years, 52.0 % responded affirmatively. Using the ADA 2003 and 2022 guidelines, 45.3 % to 47.4 % of US adults who were eligible and 4.7 to 6.7 % who were ineligible reported receiving a glucose test (Fig. 2). Applying the USPSTF 2015 and 2021 guidelines, 24.2 % to 27.6 % of US adults without diagnosed diabetes were eligible and tested while 24.4 % to 27.9 % were ineligible and tested.

Fig. 2.

Glucose testing by eligibility according to screening guidelines.

For adults who were considered eligible by each of the guidelines (Table 2), compared to their respective comparative subgroups, the prevalence of receiving a glucose test was lowest among younger adults (aged 18–44 years), men, Hispanic adults, the least educated (less than high school completion), those living in poverty, and those without health insurance or a usual place of care. These patterns were similar among adults who were ineligible by the guidelines and still reported receiving a glucose test in the past 3 years. The only exception was that Hispanic adults who were ineligible based on the guidelines had a higher prevalence (33.4 % compared to 24.4 % to 30.6 % of other race/ethnic groups) of reporting glucose tests in the past 3 years.

Table 2.

Prevalence of self-reported receipt of glucose testing in past 3 years among US adults without diagnosed diabetes, according to eligibility, overall and by selected subgroups.

	ADA 2003		ADA 2022		USPSTF 2015		USPSTF 2021		Ineligible based on all guidelines
	%	SE	%	SE	%	SE	%	SE	%	SE
Overall	59.4	1.1	57.8	1.1	65.11	1.8	63.1	1.6	28.0	1.9
Age (years), %(SE) a,b,c,d,e
18-44 years	48.1	1.7	46.0	1.7	53.9	2.8	53.1	1.9	28.0	1.9
45-64 years	62.3	1.8	62.3	1.8	65.1	2.1	65.1	2.1	--
≥65 years	71.6	2.2	71.6	2.2	77.6	3.4	77.6	3.4	--
Sex, % (SE) a,b,d,e
Male	55.6	1.5	53.9	1.5	63.8	2.1	61.0	1.9	21.4	3.0
Female	62.8	1.2	61.2	1.2	66.3	2.1	65.3	1.8	35.0	2.1
Race-ethnicity, % (SE) a,b,d
NH White	62.1	1.6	59.8	1.6	66.6	2.4	65.2	2.1	27.0	2.3
NH Black	58.2	1.8	57.5	1.8	64.8	2.4	62.8	2.2	24.4	4.1
Hispanic	51.0	1.9	50.8	1.8	58.2	2.1	54.3	2.2	33.4	3.3
NH Asian	54.9	2.5	53.4	2.5	63.0	3.6	63.6	3.4	30.6	3.8
Education <HS, % (SE) a,b,c,d,e
<HS	48.4	1.8	47.0	1.9	48.9	3.4	47.6	2.9	18.7	3.0
HS	54.5	1.9	53.1	1.9	59.9	2.9	57.8	2.8	22.1	3.0
>HS	64.0	1.3	62.2	1.3	70.4	2.1	68.6	1.9	32.4	2.7
At or below poverty, % (SE) a,b,e
No	61.1	1.2	59.3	1.2	65.8	2.1	64.2	1.8	30.2	2.2
Yes	50.2	2.6	48.9	2.6	58.0	3.8	56.4	3.2	15.1	2.0
Insured, % (SE) a,b,c,d,e
No	39.8	2.0	38.5	1.8	45.5	4.2	45.1	3.6	19.4	2.4
Yes	63.0	1.2	61.5	1.2	68.1	1.6	66.1	1.4	30.4	2.2
Usual place of care, % (SE) a,b,c,d,e
No	33.6	2.6	32.6	2.4	41.9	4.8	40.9	4.3	18.3	3.0
Yes	64.7	1.2	63.0	1.2	68.6	1.7	66.9	1.6	32.2	2.2

Unweighted sample size is 6007, representing 206.7 million adults without diagnosed diabetes.

P < 0.05 for subgroup differences for those eligible by a) ADA 2003, b) ADA 2022, c) USPSTF 2015, d) USPSTF 2021 or e) those ineligible by ADA and USPSTF guidelines.

Abbreviations: SE, standard error; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; FPG, fasting plasma glucose; A1c, glycated hemoglobin; HS, high school; NH, non-Hispanic.

Across all glycemic definitions of prediabetes or diabetes based on FPG and HbA1c, the ADA’s screening recommendations were the most sensitive (range: 91.0 % to 100.0 %) and least specific (range: 18.3 % to 35.3 %); sensitivity was higher and specificity lower for the ADA 2022 compared to 2003 recommendations (Table 3). Both sets of USPSTF recommendations provided lower sensitivity and higher specificity. The 2021 USPSTF guidelines provided greater sensitivity (range: 57.5 % to 66.6 %) and lower specificity (range: 56.6 % to 67.1 %) than the 2015 recommendations (sensitivity ranged from: 51.9 % to 59.4 % and specificity ranged from 63.2 % to 74.5 %). The probability of having prediabetes or diabetes –i.e., the yield or PPV among those recommended for screening by the guidelines– was highest for the USPSTF 2015 and lowest for the ADA 2022 guidelines. The PPV of all guidelines was especially low for undiagnosed diabetes and when using confirmatory blood glucose thresholds (both elevated HbA1c and FPG) for diagnosis. However, of those that the guidelines identified as ineligible for glucose testing –the NPV– the proportion without prediabetes or undiagnosed diabetes neared 100 %. AUC values ranged from 57.2 to 62.6 for all guidelines and all definitions of prediabetes and diabetes, and were not meaningfully different. Estimates of sensitivity, specificity, PPV, NPV, and AUC were similar to those reported above after excluding adults with self-reported prediabetes (Supplementary Appendix 3).

Table 3.

Performance ^a of screening criteria in identifying prediabetes or diabetes among US adults without diagnosed diabetes.

Definition	Guideline	Sensitivity b	Specificity c	PPV d	NPV e	AUC f
Prediabetes or diabetes
FPG ≥ 100 mg/dL or A1C ≥ 5.7 % g	USPSTF 2015	51.9 (49.4–54.3)	74.5 (72.4–76.4)	61.5 (58.6–64.3)	66.3 (63.8–68.7)	62.6 (61.4–63.8)
	USPSTF 2021	57.5 (55.0–59.9)	67.1 (64.6–69.4)	57.8 (54.9–60.7)	66.8 (64.3–69.1)	61.6 (60.4–62.8)
	ADA 2003	91.0 (89.1–92.6)	35.3 (32.6–38.1)	52.5 (50.1–54.9)	83.3 (79.9–86.2)	62.6 (61.6–63.5)
	ADA 2022	94.2 (92.6–95.4)	27.6 (25.1–30.2)	50.5 (48.2–52.9)	85.7 (82.3–88.6)	60.2 (59.4–61.0)
FPG ≥ 100 mg/dL and A1C ≥ 5.7 % g	USPSTF 2015	57.7 (53.8–61.5)	66.3 (64.4–68.2)	22.3 (20.2–24.6)	90.3 (89.2–91.3)	62.5 (60.9–64.1)
	USPSTF 2021	63.4 (59.8–66.9)	59.6 (57.2–61.9)	20.9 (18.9–22.9)	90.6 (89.4–91.8)	61.7 (60.1–63.3)
	ADA 2003	97.0 (94.7–98.4)	27.2 (25.0–29.5)	18.3 (17.0–19.7)	98.2 (96.8–99.0)	60.8 (60.1–61.5)
	ADA 2022	98.2 (96.3–99.2)	20.7 (18.8–22.8)	17.2 (16.0–18.5)	98.6 (97.1–99.3)	58.3 (57.7–58.9)
Diabetes
FPG ≥ 126 mg/dL or A1C ≥ 6.5 % g	USPSTF 2015	59.4 (51.3–67.1)	63.7 (61.9–65.4)	5.4 (4.5–6.5)	97.8 (97.2–99.3)	61.3 (58.4–64.1)
	USPSTF 2021	64.6 (55.9–72.5)	57.0 (54.9–59.0)	5.0 (4.2–5.9)	97.9 (97.2–98.4)	60.2 (57.4–63.0)
	ADA 2003	97.9 (95.9–98.9)	24.5 (22.5–26.6)	4.3 (3.8–5.0)	99.7 (99.4–99.8)	59.4 (58.5–60.3)
	ADA 2022	98.9 (98.1–99.3)	18.6 (16.8–20.5)	4.1 (3.5–4.7)	99.8 (99.7–99.9)	57.2 (56.5–58.0)
FPG ≥ 126 mg/dL and A1C ≥ 6.5 % g	USPSTF 2015	58.9 (46.6–70.1)	63.2 (61.4–64.9)	2.3 (1.7–3.1)	99.0 (98.6–99.4)	61.4 (57.1–65.7)
	USPSTF 2021	66.6 (53.5–77.5)	56.6 (54.6–58.6)	2.2 (1.7–2.9)	99.1 (98.6–99.5)	62.1 (58.0–66.2)
	ADA 2003	99.1 (93.4–99.9)	24.1 (22.1–26.1)	1.9 (1.5–2.4)	99.9 (99.6–100.0)	59.6 (58.7–60.5)
	ADA 2022	100.0	18.3 (16.5–20.2)	1.8 (1.4–2.3)	100.0	57.5 (57.1–58.0)

Abbreviations: PPV, positive predictive value; NPV, negative predictive value; USPSTF, United States Preventive Services Task Force; ADA, American Diabetes Association; DM, diabetes mellitus; FPG, fasting plasma glucose; A1c, glycated hemoglobin.

^aData are presented as weighted percentages and 95% confidence intervals based on data from 6,007 adults without diagnosed diabetes, representing 207 million.

^bProbability of being eligible for screening among those with prediabetes or diabetes (true positives).

^cProbability of being ineligible for screening among those without prediabetes or diabetes (true negatives).

^dProbability of having prediabetes or diabetes among those eligible for screening.

^eProbability of not having prediabetes or diabetes among those ineligible for screening.

^fProbability that being eligible for screening produces a higher probability of having prediabetes or diabetes than being ineligible.

^gA1c values 5.7 % and 6.5 % correspond to A1c values 39 mmol/mol and 48 mmol/mol.

4. Discussion

Based on the 2021 update of the USPSTF’s recommendations, 90.4 million asymptomatic adult Americans without diagnosed diabetes are eligible to be screened for prediabetes and type 2 diabetes, 13.7 million more than would be eligible according to the 2015 USPSTF guideline. Similarly, an additional 11.8 million adults would be eligible based on the 2022 (169.5 million) compared to the 2003 ADA recommendations (157.7 million). Those still ineligible for glucose testing in the US tend to be younger, predominantly non-Hispanic White, lower body weight individuals with lower likelihood of being insured or having access to a usual care provider. Still, on average, just over a half of all adults reported receiving a glucose test in the past three years, 28 % of whom were ineligible based on all three guidelines. Furthermore, though the PPV of the USPSTF 2021 and both ADA guidelines are now similar, only 50–60 % of those considered eligible by these guidelines would actually have prediabetes or diabetes, even when using the broadest definitions (FPG ≥ 100 mg/dL or A1C ≥ 5.7 %). The use of two-step confirmatory diagnostic thresholds is associated with substantially lower yields.

Tradeoffs in the clinical performance characteristics of screening criteria have implications for their use in practice. Ideal screening criteria will exhibit both high sensitivity and high specificity; as noted here, current guidelines do not offer both. While prediabetes and diabetes rarely requires urgent treatment, the long-term benefits of early lifestyle change and pharmacotherapy are well established.[8-11] Further, the small proportion of patients found to have uncontrolled diabetes based on a screening test result could experience more substantial short-term benefit, offering an opportunity to prevent acute hyperglycemic complications. Relative to other screening tests such as mammography, the cost of recommended glycemic tests and the potential harms of false positive results are low.

There are a number of health care and preventive services implications of expanding sensitivity and lowering specificity of current guidelines. In the near term, at the local level, the newer guidelines may influence clinicians to offer glucose testing to lower-risk individuals and potentially, patients to request glucose testing. Anecdotal data suggest that clinicians already screen younger adults who have risk factors for diabetes, likely because they ascribe to the ADA 2003 criteria. Our data confirm this as 28 % of adults who would be ineligible based on guidelines reported receiving glucose testing in the past 3 years. So, one perspective is that the new USPSTF criteria are “catching up” with current practice. That said, the value in the endorsement is that USPSTF recommendations tend to be covered fully by payers,[30] lowering the cost barrier to obtaining these services.

Currently, nationally, 23 % of 37.3 million Americans with diabetes and an estimated 81 % of the 96 million with prediabetes remain unaware of their metabolic risk.[3] These gaps in awareness could be related to a variety of system- (e.g., coverage for glucose testing), clinician- (e.g., lack of familiarity with guidelines), or patient-level factors (e.g., reduced care-seeking).[31] Our data further point to the influence of social determinants of health[32] wherein greater proportions of adults who have lower income, lower educational attainment, and Americans lacking health insurance or a usual health care provider reported not receiving a glucose test; many of these social determinants are interconnected and also possibly influence lower health literacy and understanding of what transpired in the clinical setting.[33,34] Unfortunately, the updated guidelines, in and of themselves, do not directly address access and other barriers to glucose testing, as well poor communication of results by clinicians, that are routinely faced by vulnerable groups.[35].

Alignment of the age criterion (i.e. 35 years) of the ADA and USPSTF guidelines is an important step in reducing confusion experienced by clinicians in ordering, interpreting, and communicating glucose testing results. As such, the USPSTF’s recent expansion of criteria is welcomed, though the criteria still exclude normal-weight people and younger adults. In certain race/ethnic groups –for example, people of Asian, Hispanic, or sub-Saharan African descent– this is problematic as these populations appear to develop diabetes at lower BMI and younger age. [36] Furthermore, the expansion of the ADA 2022 guidelines –though aligned with USPSTF 2021 guidelines on the age threshold– further augments glucose testing recommendations in favor of greater sensitivity and lower specificity.

In addition to appropriate glucose testing, operationally linking high-risk individuals to an adequate, physically-accessible, and affordable supply of type 2 diabetes prevention and care services –that include education and lifestyle modification programs, as well as accessible medications– appears to be a major need nationally.[37] In the vast market-based arena of US health services, there is no single stakeholder responsible for offering services, but rather opportunities, incentives via coverage (e.g., Medicare, some Medicaid, and some private payers authorizing coverage for diabetes preventive services),[38] and sometimes seed funds to initiate delivery of preventive and care services. Current infrastructure and capacity are insufficient to facilitate lifestyle change for 96 million individuals with prediabetes[39] and provide clinical care for 37.3 million people with diabetes. This is one of the factors associated with gaps in the continua of prevention[13] and care, [12] from testing, awareness of risk, communication and referral by clinicians, and attendance and continued adherence to intervention programs and medications. Financial coverage for services and medications, or the lack thereof, contributes to gaps and the variability across insurers and benefit types may deepen disparities in diabetes and related outcomes. Lastly, demand for preventive and care services itself can be variable and multifactorial; awareness, motivation, and competing priorities are major concerns, highlighting the need for enhancing communication about diabetes and related risks in a way that motivates engagement.

The strengths of this analysis include using nationally-representative data to assess the expansion of eligibility and performance of the USPSTF 2021 and ADA 2022 diabetes screening recommendations compared to the previous USPSTF 2015 and ADA 2003 guidelines. We examined a variety of indicators of guideline performance (sensitivity, specificity, PPV, and NPV) and different definitions of prediabetes and diabetes.

Our study has some limitations. Our estimates are based on cross-sectional data which limits causal inferences regarding the ultimate effect of the implementation of the guidelines on diagnoses and health outcomes. Furthermore, receipt of glucose testing in the past three years was self-reported which may be subject to misclassification associated with recall bias. With regard to glucose testing, our findings could not account for routine biochemical assessments by clinicians that patients might have been unaware of; still, the very low prevalence of self-reported testing and prevalent diabetes or prediabetes suggest that even passive testing and results are not well communicated to individuals when they seek preventive care. For undiagnosed prediabetes or diabetes, our estimates were based on a single measure of elevated FPG or A1C values only. However, to examine the sensitivity, specificity, PPV, and NPV of guidelines, we used a range of less and more stringent definitions for prediabetes and diabetes.[26] The application of these eligibility criteria only to those without diagnosed diabetes likely estimates a lower overall performance; though, the focus of our analysis aligns with the goals of screening guidelines, which is to identify disease in those currently undiagnosed. Some high-risk groups like PCOS were not included in the analysis, though they account for a very small fraction of those at risk for developing diabetes nationally. We did not define diabetes based on oral glucose tolerance test (OGTT), which could miss those who solely meet the postprandial criteria for prediabetes or diabetes, but OGTTs are rarely used in contemporary clinical practice.

5. Conclusions

US national data show that recent guideline changes on whom to screen for prediabetes and diabetes would result in additional 12–14 million asymptomatic adults becoming eligible for glucose testing. Notably, the USPSTF 2021 criteria provide balanced sensitivity and specificity and higher yield than ADA 2022 criteria which maximize sensitivity. The data also point to low alignment of practice with guidelines as large proportions of the population who are ineligible report receiving glucose testing. The expanded sensitivity of both the recent USPSTF and ADA guidelines also do not directly address the social determinants of health which remain persistent barriers to receiving testing; other policy and programmatic efforts may be needed to close testing gaps in those segments of the population which also happen to be the most at-risk for diabetes.