Abstract
OBJECTIVE
Despite improvements in screening, Hispanic/Latino individuals bear a disproportionate burden of undiagnosed diabetes in the U.S. Identifying who is at risk within this large and diverse population is important for targeting interventions. In this study, we sought to characterize risk factors for undiagnosed diabetes among Hispanic/Latino adults. We also investigated determinants among insured adults to explore barriers for those with access to care.
RESEARCH DESIGN AND METHODS
We used data from 1,883 Hispanic/Latino adults aged ≥20 years from the National Health and Nutrition Examination Surveys (2005–2018). Sequential multivariable logistic regression models were used to examine a range of social, health care, and individual-level determinants of undiagnosed diabetes (defined as having elevated fasting plasma glucose ≥126 mg/dL or HbA1c ≥6.5% [48 mmol/mol] in participants self-reporting as not having diabetes) in the overall sample and among those with health insurance (n = 1,401).
RESULTS
Younger age (20–44 years), male sex, and having immigrated (compared with being U.S. born), but not socioeconomic factors, were significantly associated with a higher odds of undiagnosed diabetes compared with being diagnosed. These estimates were attenuated after adjusting for health care utilization variables. In fully adjusted models, having no health care visits in the past year, reporting no family history of diabetes, and having better self-reported health were the predominant risk factors for undiagnosed diabetes in the overall sample and among insured Hispanic/Latino adults.
CONCLUSIONS
Our findings highlight the importance of reaching younger, male, and immigrant Hispanic/Latino adults and addressing barriers to health care utilization, even among insured adults, to improve diabetes awareness.
Graphical Abstract
Introduction
Despite improvements over time in diabetes screening and diagnosis in the U.S., Hispanic/Latino individuals continue to bear a disproportionate burden of undiagnosed disease compared with non-Hispanic White individuals (1–7). From 1999 to 2014, there was improved screening and detection of diabetes primarily among non-Hispanic White individuals, adults aged ≥65 years, and those of the highest income (6). In contrast, Mexican American adults experienced an increase in undiagnosed diabetes from 3.7% to 6% over the same period (6). A more recently published National Health and Nutrition Examination Survey (NHANES) analysis using data from 1988 to March 2020 demonstrated a high prevalence of undiagnosed diabetes among adults with low income and who were uninsured, racial and ethnic minorities, and aged <45 years, a pattern that has persisted throughout the past 30 years (7).
Undiagnosed diabetes has been linked to more advanced disease at diagnosis, worse glycemic control, and a greater burden of complications (8,9). Data from the Centers for Disease Control and Prevention in 2018 demonstrated that Hispanic/Latino individuals were 1.3 times more likely to die of diabetes than non-Hispanic/Latino White individuals (10). Furthermore, the cost of diabetes and its associated morbidity and mortality was estimated to be $327 billion in 2017, representing ∼14% of the health care costs in the U.S. (11). According to 2020 census data, Hispanic/Latino individuals comprise 19% of the U.S. population and one of the fastest growing racial/ethnic groups (12). Thus, addressing gaps in undiagnosed diabetes represents a significant opportunity to improve overall population health in the U.S., especially for this fast-growing segment of Americans.
A major limitation in this research has arisen from the fact that Hispanic/Latino individuals are often characterized as a single, high-risk ethnic group. However, just under one-half of all Hispanic/Latino individuals in the U.S. are foreign-born, and there is considerable diversity with respect to socioeconomic status, culture, language use, and country of origin (13). Several of these factors have been shown to contribute to variation in health outcomes. Identifying who is at highest risk of undiagnosed diabetes within this large and diverse population is an important step toward targeting interventions and identifying the roots of disparities. Thus, one aim of this study was to investigate the risk factors for undiagnosed diabetes among Hispanic/Latino individuals in the U.S.
A second aim of this study was to examine the risk factors for undiagnosed diabetes among Hispanic/Latino adults with health insurance. Lack of health insurance and poor access to care have been identified as among the most important modifiable contributors to disparities in undiagnosed diabetes (1,14–16). Data from the U.S. Census Bureau indicated that 18.3% of Hispanic/Latino adults living in the U.S. were uninsured in 2020, the highest proportion of any racial/ethnic group (17). Uninsured individuals are less likely to receive needed routine care to screen for diabetes (1,14–16). However, it is unclear whether barriers to diabetes awareness remain among Hispanic/Latino individuals with health insurance. Such barriers may include social determinants of health that are increasingly appreciated as major risk factors for poor health, even for people with access to care (18–21). For Hispanic/Latino individuals with health insurance, social factors such as limited English proficiency and cultural barriers, financial hardship, and poor health care utilization may interfere with the ability to receive a timely diagnosis. While expansion of health insurance coverage remains key to improving diabetes screening and detection for Hispanic/Latino individuals, it is important to understand the risk factors for undiagnosed disease, even for individuals with health insurance.
Research Design and Methods
We used cross-sectional, complex survey data from NHANES, a series of nationally representative surveys continuously conducted by the National Center for Health Statistics. NHANES uses a complex, multistage sample design and is intended to be nationally representative of the U.S. noninstitutionalized population (22). There were two phases of data collection. In the first phase, researchers collected information from household interviews on demographics, socioeconomic indicators, past medical history, and health behaviors. Participants were administered a physical examination in a mobile examination center in the second phase. NHANES protocols were approved by the National Center for Health Statistics institutional review board, and all participants provided written informed consent (22).
We pooled data from six 2-year survey cycles from 2005 to 2018 to generate an adequate sample size. We initially included 1,897 Mexican American and other Hispanic nonpregnant adults aged ≥20 years with complete data on diabetes status, including laboratory measures. We excluded observations with missing data on selected covariates, yielding a sample size of 1,883.
Diabetes Definitions
Diagnosed diabetes was defined as a physician diagnosis of diabetes (other than during pregnancy) that was self-reported by the participant. Undiagnosed diabetes was defined as meeting American Diabetes Association (ADA) diagnostic criteria, including elevated fasting plasma glucose ≥126 mg/dL (among those fasting ≥8 h) or HbA1c ≥6.5% (48 mmol/mol), in adults who self-reported not having diabetes (23). A phlebotomist collected blood samples from participants according to standardized protocols at mobile examination centers. Changes to the laboratory methods, such as instrumentation and calibration procedures, occurred during the waves of the NHANES that we evaluated; however, these methodological changes had negligible effects on the measures, and adjustments in calibrations were not advised across these NHANES waves (22). Total diabetes refers to the combination of participants with undiagnosed and diagnosed diabetes. Participants also self-reported whether they had been diagnosed with prediabetes. This information was used in the sensitivity analyses described below.
Risk Factors
A growing body of research has explored the utility of measuring and addressing social determinants of health to improve health outcomes (18–21,24). We considered the following domains among the essential targets that impact individual health and well-being and contribute to a wide spectrum of health disparities and inequities (24).
Sociodemographic and Immigrant-Related Factors
We included the relevant variables available in NHANES that fit within this domain, including education, household poverty level, food security, home ownership (yes/no), and employment status (employed/not employed) (24). We also assessed variables that are relevant to U.S. Hispanic/Latino populations, including nativity and time living in the U.S., language of interview, citizenship status, and whether Mexican or non-Mexican). Education was categorized as less than high school or high school or more. We also evaluated education as a four-category variable (less than high school, high school, some college, college or more), but the sample size at the upper end was small, and results were similar regardless of the way education was operationalized. The poverty income ratio (PIR) was calculated by dividing total family income by the federal poverty threshold based on family size and categorized as <1.00, 1.00–2.99, and ≥3.00 (25). Food security was measured using the 18-item U.S. Household Food Security Survey Module. Participants with three or more affirmative responses were defined as being food insecure (26). For nativity and time living in the U.S., we combined country of birth and time in the U.S. as follows: US-born, foreign-born living in the U.S. 0–19 years and foreign-born living in the U.S. ≥20 years. The language of the interview (Spanish vs. English) was used to characterize participants with limited English proficiency (27). Finally, citizenship status was self-reported as yes or no to the question, “Are you a citizen of the United States?”
Health Care–Related Factors
We evaluated the role of health care access and utilization in undiagnosed diabetes by assessing the following variables: 1) health insurance coverage and type, 2) having a routine place for health care, and 3) the number of health care visits in the past year. Health insurance coverage was self-reported as yes or no in response to the question, “Are you covered by health insurance or some other kind of health care plan?” Although NHANES captures various insurance types, we categorized this variable as private, Medicare/Medicaid, and other due to small sample sizes for other insurance types. We also assessed whether participants had a routine place for health care (yes/no), and we categorized responses to the question, “During the past 12 months, how many times have you seen a doctor or other health care professional about your health at a doctor’s office, a clinic, or some other place,” as 0, 1, or ≥2.
Other Risk Factors
We also assessed the role of BMI, self-rated health (excellent/very good, good, and fair/poor), and family history of diabetes (yes/no). BMI was derived from weight (in kg) divided by height (in m2), which were clinically measured at the mobile examination center. BMI was categorized as <30 kg/m2 and ≥30 kg/m2.
Statistical Analysis
We completed our statistical analyses using Stata 16.1 software. All analyses were weighted using the recommended sample weights to adjust for sampling probability and nonresponse according to guidelines established by NHANES. Descriptive statistics were generated on the distribution of the variables for total, undiagnosed, and diagnosed diabetes. Significance tests comparing adults with undiagnosed and diagnosed diabetes were conducted using χ2 statistics for categorical variables. Multivariable logistic regressions were used to estimate associations between the range of risk factors and undiagnosed diabetes. Risk factors that were significantly associated with undiagnosed diabetes in bivariate analyses were formally tested in regression models (Table 1). We used a sequential modeling approach to first identify the sociodemographic groups most burdened by undiagnosed diabetes among Hispanic/Latino adults. Specifically, we first evaluated associations with age (20–44, 45–64, ≥65 years [reference]), sex (male vs. female [reference]), and nativity/length of time living in the U.S. (0–19 years, ≥20 years, U.S.-born [reference]) (model 1). We initially tested but later dropped variables such as citizenship status and language of interview from the regression models due to high collinearity with nativity/length of time living in the U.S. and because the latter variable appeared to be more strongly related to undiagnosed diabetes. We also did not include education, PIR, food security, and home ownership in our models as these were not associated with undiagnosed diabetes and did not improve model fit. We then added employment status, followed by the health care–related variables, followed by the other individual-level risk factors to evaluate whether these attenuated associations for the sociodemographic variables. All models also included adjustment for survey year. A two-sided P < 0.05 was considered statistically significant for all analyses.
Table 1.
Characteristics of Hispanic/Latino adults aged ≥20 years by diabetes awareness status, NHANES 2005–2018
| Total | Undiagnosed diabetes† | Diagnosed diabetes± | P * | |
|---|---|---|---|---|
| Unweighted, n (%) | 1,883 | 469 (29.3) | 1,414 (70.7) | |
| Age-groups (years) | ||||
| 20–44 | 253 (24.4) | 93 (34.1) | 160 (20.4) | <0.001 |
| 45–64 | 960 (51.5) | 243 (48.5) | 717 (52.7) | |
| ≥65 | 670 (24.1) | 133 (17.4) | 537 (26.9) | |
| Sex | ||||
| Female | 964 (49.4) | 233 (44.3) | 731 (51.6) | 0.042 |
| Male | 919 (50.6) | 236 (55.7) | 683 (48.5) | |
| Nativity/length of time in the U.S. (years) | ||||
| 0–19 | 318 (21.1) | 109 (28.4) | 209 (18.0) | 0.002 |
| ≥20 | 839 (39.0) | 211 (38.8) | 628 (39.2) | |
| Born in U.S. | 652 (36.2) | 130 (28.9) | 522 (39.3) | |
| Missing | 74 (3.7) | 19 (4.0) | 55 (3.6) | |
| Mexican origin | 0.400 | |||
| Mexican | 1,202 (62.5) | 294 (64.3) | 908 (61.7) | |
| Non-Mexican | 681 (37.5) | 175 (35.7) | 506 (38.3) | |
| U.S. citizenship status | ||||
| Citizen | 1,276 (63.5) | 280 (54.7) | 996 (67.1) | 0.003 |
| Noncitizen | 595 (35.8) | 186 (44.5) | 409 (32.2) | |
| Missing | 12 (0.7) | 3 (0.8) | 9 (0.7) | |
| Language of interview | ||||
| English | 841 (47.0) | 180 (39.7) | 661 (50.1) | 0.002 |
| Spanish | 1,042 (53.0) | 289 (60.3) | 753 (49.9) | |
| Education | ||||
| Less than high school | 1,089(52.3) | 264 (51.0) | 825 (52.9) | 0.592 |
| High school or more | 794 (47.7) | 205 (49.0) | 589 (47.1) | |
| PIR | 0.864 | |||
| <1.00 | 556 (27.8) | 134 (26.1) | 422 (28.5) | |
| 1.00–2.99 | 774 (42.3) | 196 (43.2) | 578 (41.9) | |
| ≥3.00 | 309 (19.0) | 79 (19.5) | 230 (18.8) | |
| Missing | 244 (10.9) | 60 (11.3) | 184 (10.7) | |
| Own home | 0.688 | |||
| Yes | 1,107 (56.6) | 269 (55.6) | 838 (57.1) | |
| No | 776 (43.4) | 200 (44.4) | 576 (42.9) | |
| Employment status | 0.0003 | |||
| Employed | 712 (47.6) | 219 (56.0) | 493 (44.2) | |
| Not employed | 1,171 (52.4) | 250 (44.0) | 921 (55.8) | |
| Food security** | ||||
| Secure | 995 (52.2) | 244 (52.7) | 751 (52.0) | 0.360 |
| Insecure | 838 (45.6) | 217 (46.1) | 621 (45.3) | |
| Missing | 50 (2.3) | 8 (1.3) | 42 (2.7) | |
| Health insurance coverage | ||||
| Yes | 1,401 (67.2) | 308 (61.5) | 1,093 (73.7) | 0.001 |
| No | 482 (32.8) | 161 (38.5) | 321 (26.3) | |
| Health insurance type (n = 1,401) | ||||
| Private | 463 (41.6) | 135 (48.7) | 328 (39.2) | 0.033 |
| Medicare/Medicaid | 632 (37.9) | 117 (31.4) | 515 (40.1) | |
| Other/unknown | 306 (20.5) | 56 (19.9) | 250 (20.7) | |
| Routine place for care | <0.001 | |||
| Yes | 1,642 (84.0) | 341 (70.8) | 1,301 (89.5) | |
| No | 241 (16.0) | 128 (29.3) | 113 (10.5) | |
| Health care visits in the past year | ||||
| 0 | 217 (15.0) | 137 (32.1) | 80 (7.9) | <0.001 |
| 1 | 206 (12.9) | 88 (19.7) | 118 (10.1) | |
| ≥2 | 1,460 (72.1) | 244 (48.2) | 1,216 (82.0) | |
| BMI (kg/m2) | <0.001 | |||
| <30 | 754 (38.0) | 164 (28.9) | 590 (41.8) | |
| ≥30 | 1,035 (59.6) | 297 (69.3) | 738 (55.7) | |
| Missing | 94 (2.4) | 8 (1.8) | 86 (2.6) | |
| Family history of diabetes | ||||
| Yes | 1,221 (66.5) | 249 (54.7) | 972 (71.4) | <0.001 |
| No | 627 (32.3) | 213 (44.5) | 414 (27.2) | |
| Missing | 35 (1.2) | 7 (0.8) | 28 (1.4) | |
| Self-rated health | ||||
| Excellent/very good | 196 (11.1) | 82 (18.9) | 114 (7.8) | <0.001 |
| Good | 516 (30.1) | 167 (36.1) | 349 (27.7) | |
| Fair/poor | 1,171 (58.8) | 220 (45.0) | 951 (64.5) |
Proportions are weighted.
Undiagnosed diabetes defined as fasting plasma glucose ≥126 mg/dL (48 mmol/mol) or HbA1c ≥6.5% in adults who self-reported not having diabetes.
Diagnosed diabetes defined as physician’s diagnosis (other than during pregnancy) as self-reported by participant.
P values compare estimates between those with undiagnosed and diagnosed diabetes.
Food security measured using 18-item U.S. Household Food Security Survey Module. Participants with three or more affirmative responses were defined as food insecure.
We also performed a secondary analysis among only participants with health insurance (n = 1,401). We used the same approach as in the main analyses to select the risk factors to assess in the regression models. However, we also included insurance type as part of the health care–related variables in models among insured participants. In sensitivity analyses, we reran all models excluding individuals with undiagnosed diabetes who reported having ever been told that they had prediabetes.
Results
Overall, 29.3% of Hispanic/Latino adults with diabetes were undiagnosed. Table 1 compares characteristics of the study population by diabetes awareness status across NHANES survey years 2005–2018. There were statistically significant differences in diabetes awareness by age, sex, and immigrant- and health care–related factors but not by most of the socioeconomic factors, except for employment status. Specifically, individuals with undiagnosed diabetes tended to be younger (20–44 years), male, and employed (vs. not employed) compared with those with diagnosed diabetes. Furthermore, Hispanic/Latino adults with undiagnosed diabetes were more likely to be recent immigrants (living in the U.S. <20 years), noncitizens, and Spanish-speaking than those with diagnosed diabetes. In terms of health care factors, participants with undiagnosed diabetes were also more likely to have no health insurance, no routine place for care, and zero health care visits in the past year. Notably among participants with health insurance, those with private insurance were more likely to be undiagnosed, whereas those with Medicare/Medicaid were more likely to be diagnosed. Finally, participants reporting no family history of diabetes and higher levels of self-rated health but with BMI levels above the threshold for obesity were more likely to be undiagnosed than diagnosed. There were no statistically significant differences in diabetes awareness by Mexican or non-Mexican origin, education, PIR, food security status, or home ownership.
Table 2 presents the results of multivariable regression analyses. In a model adjusted for only sociodemographic variables (model 1), younger age (20–44 years) (odds ratio [OR] 2.41 [95% CI 1.55–2.72]), male sex (OR 1.32 [1.00–1.76]), and fewer years living in the U.S. among the foreign-born participants (0–19 years: OR 1.79 [1.20–2.68]; ≥20 years: OR 1.42 [1.07–1.91]) compared with U.S.-born participants were associated with a higher odds of undiagnosed diabetes relative to having diagnosed diabetes. There was some attenuation of estimates after adjusting for employment status, which was not significantly associated with undiagnosed diabetes, though the estimate was large (OR 1.28 [0.95–1.73]) (model 2). There was even further attenuation of estimates for age and for nativity/time living in the U.S. with adjustment for health insurance status (model 3), but younger age and immigrant status remained associated with undiagnosed diabetes. However, after accounting for health care utilization (a routine place for care and number of health care visits in the past year) (model 4), estimates for age and nativity/time living in the U.S. decreased even further, and only younger age (20–44 years) remained significantly associated with undiagnosed diabetes. Moreover, having zero or only one health care visit in the past year was associated with a 6.14- and 3.06-higher odds of undiagnosed diabetes, respectively, than having two or more visits. Having no routine place for health care was also associated with a 50% higher odds of undiagnosed diabetes, but the estimate was not statistically significant. Although having health insurance was not significantly associated with awareness of diabetes status in model 3, not only did the estimate become statistically significant in model 4 but also the relationship was reversed (not having insurance was associated with a lower odds of undiagnosed diabetes). To better understand this finding, we conducted a sensitivity analysis that revealed that ∼45% of Hispanic/Latino adults who did not have insurance still had two or more health care visits in the past year (data not shown). In model 5, having no self-reported family history of diabetes (OR 1.99 [1.48–2.68]), higher self-rated health (excellent/very good [OR 3.97 (2.37–6.65)] or good [OR 1.77 (1.24–2.54)] vs. fair/poor), and having a BMI classified as obese (OR 1.94 [1.42–2.64] vs. normal/overweight) were associated with a higher odds of undiagnosed diabetes. Moreover, accounting for these variables, particularly family history of diabetes and self-rated health, led to increases in model estimates for younger individuals (20–44 years old) and for all immigrant groups compared with the U.S.-born group.
Table 2.
ORs of undiagnosed diabetes among Hispanic/Latino adults in the U.S. (n = 1,883), NHANES 2005–2018
| OR (95% CI) | |||||
|---|---|---|---|---|---|
| Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | |
| Age (years) | |||||
| 20–44 | 2.41 (1.55–3.72) | 2.11 (1.30–3.43) | 1.92 (1.23–2.99) | 1.68 (1.03–2.72) | 1.72 (1.10–2.69) |
| 45–64 | 1.37 (0.99–1.89) | 1.24 (0.89–1.74) | 1.16 (0.84–1.60) | 1.12 (0.80–1.57) | 1.11 (0.79–1.56) |
| ≥65 | Ref | Ref | Ref | Ref | Ref |
| Sex | |||||
| Male | 1.32 (1.00–1.76) | 1.26 (0.96–1.67) | 1.26 (0.95–1.67) | 1.04 (0.76–1.41) | 0.95 (0.68–1.33) |
| Female | Ref | Ref | Ref | Ref | Ref |
| Nativity/time living in U.S. (years) | |||||
| 0–19 | 1.79 (1.20–2.68) | 1.77 (1.18–2.66) | 1.63 (1.07–2.48) | 1.46 (0.89–2.38) | 1.70 (0.99–2.93) |
| ≥20 | 1.42 (1.07–1.91) | 1.40 (1.05–1.87) | 1.36 (1.02–1.82) | 1.27 (0.93–1.73) | 1.45 (1.02–2.06) |
| Missing | 1.59 (0.75–3.39) | 1.55 (0.74–3.27) | 1.36 (0.63–2.96) | 1.09 (0.53–2.22) | 1.55 (0.72–3.32) |
| U.S.-born | Ref | Ref | Ref | Ref | Ref |
| Employment status | |||||
| Employed | 1.28 (0.95–1.73) | 1.30 (0.96–1.75) | 1.07 (0.78–1.47) | 0.99 (0.73–1.34) | |
| Not employed | Ref | Ref | Ref | Ref | |
| Health insurance | |||||
| Yes | Ref | Ref | Ref | ||
| No | 1.30 (0.92–1.83) | 0.63 (0.41–0.96) | 0.68 (0.44–1.06) | ||
| Routine place for health care | |||||
| No | 1.50 (0.91–2.49) | 1.54 (0.91–2.60) | |||
| Yes | Ref | Ref | |||
| Number of health care visits in past year | |||||
| 0 | 6.14 (3.81–9.88) | 6.16 (3.88–9.78) | |||
| 1 | 3.06 (2.15–4.35) | 2.83 (1.97–4.06) | |||
| ≥2 | Ref | Ref | |||
| Family history of diabetes | |||||
| Yes | Ref | ||||
| No | 1.99 (1.48–2.68) | ||||
| Missing | 0.55 (0.52–3.38) | ||||
| Self-rated health | |||||
| Excellent/very good | 3.97 (2.37–6.65) | ||||
| Good | 1.77 (1.24–2.54) | ||||
| Fair/poor | Ref | ||||
| BMI (kg/m2) | |||||
| <30 | Ref | ||||
| ≥30 | 1.94 (1.42–2.64) | ||||
| Missing | 1.32 (0.52–3.38) | ||||
Model 1 includes sociodemographic variables (age, sex, and nativity/time living in U.S.) and NHANES survey year. Model 2 includes model 1 plus employment status. Model 3 includes model 2 plus health insurance. Model 4 includes model 3 plus health care utilization variables. Model 5 includes model 4 plus family history of diabetes, self-rated health, and BMI. Ref, reference.
In a secondary analysis among only individuals with health insurance (n = 1,401), bivariate analyses indicated that younger age (20–44 years), employed versus not employed, private insurance, no routine place for health care, no health care visits in the past year, obesity, no family history of diabetes, and higher levels of self-rated health were significant risk factors for undiagnosed diabetes (Supplementary Table 1). Results also indicated that Spanish-speaking participants and those owning a home had slightly higher proportions with undiagnosed than diagnosed diabetes. In multivariable regression models (Table 3), younger age was associated with a higher odds of undiagnosed diabetes (20–44 years: OR = 1.96 [95% CI 1.23–3.11]; 45–64 years: OR 1.47 [1.07–2.01]) (model 1). However, with the addition of employment status (model 2), health insurance type (model 3), and the health care utilization variables (model 4), these estimates were considerably attenuated. Moreover, the odds of undiagnosed diabetes for employed versus not employed participants and among those with private insurance versus Medicare/Medicaid were greatly attenuated with the addition of the health care utilization variables (comparing model 4 with model 3). Insured patients with zero health care visits in the past year had a more than five times higher odds of being undiagnosed with diabetes than those with two or more health encounters. As in the main models, having no family history of diabetes and reporting high levels of self-rated health were also associated with significantly higher odds of undiagnosed disease, and adjustment for these factors led to an increase in estimates among younger aged adults and Spanish speakers. In sensitivity analyses, results from all models were similar when we excluded individuals with undiagnosed diabetes who reported ever having been told that they had prediabetes (data not shown).
Table 3.
ORs of undiagnosed diabetes among Hispanic/Latino adults in the U.S. among individuals with health insurance (n = 1,401), NHANES 2005–2018
| OR (95% CI) | |||||
|---|---|---|---|---|---|
| Model 1 | Model 2 | Model 3 | Model 4 | Model 5 | |
| Age (years) | |||||
| 20–44 | 1.96 (1.23–3.11) | 1.64 (0.96–2.80) | 1.52 (0.90–2.58) | 1.32 (0.75–2.32) | 1.54 (0.90–2.63) |
| 45–64 | 1.47 (1.07–2.01) | 1.26 (0.90–1.76) | 1.16 (0.78–1.72) | 1.18 (0.80, 1.76) | 1.23 (0.79–1.89) |
| ≥65 | Ref | Ref | Ref | Ref | Ref |
| Sex | |||||
| Male | 1.16 (0.82–1.65) | 1.11 (0.78–1.58) | 1.11 (0.78–1.57) | 1.02 (0.70–1.48) | 0.89 (0.59–1.35) |
| Female | Ref | Ref | Ref | Ref | Ref |
| Language of interview | |||||
| English | Ref | Ref | Ref | Ref | Ref |
| Spanish | 1.32 (0.99–1.75) | 1.33 (1.00–1.77) | 1.39 (1.03–1.86) | 1.22 (0.89–1.69) | 1.59 (1.13–2.23) |
| Employment status | |||||
| Employed | 1.43 (0.98–2.11) | 1.30 (0.84–2.01) | 1.18 (0.75–1.85) | 1.02 (0.67–1.55) | |
| Not employed | Ref | Ref | Ref | Ref | |
| Insurance type | |||||
| Medicare/Medicaid | Ref | Ref | Ref | ||
| Private | 1.32 (0.85–2.05) | 1.13 (0.71–1.82) | 1.04 (0.64–1.70) | ||
| Other/unknown | 1.20 (0.77–1.88) | 1.13 (0.73–1.75) | 1.02 (0.62–1.66) | ||
| Routine place for health care | |||||
| No | 1.29 (0.61–2.71) | 1.43 (0.66–3.12) | |||
| Yes | Ref | Ref | |||
| Number of health care visits in past year | |||||
| 0 | 5.03 (2.67–9.47) | 4.68 (2.52–8.69) | |||
| 1 | 3.06 (1.90–4.95) | 2.82 (1.77–4.49) | |||
| ≥2 | Ref | Ref | |||
| Family history of diabetes | |||||
| Yes | Ref | ||||
| No | 1.73 (1.21–2.47) | ||||
| Missing | 0.52 (0.16–1.67) | ||||
| Self-rated health | |||||
| Excellent/very good | 4.17 (2.30–7.55) | ||||
| Good | 1.71 (1.15–2.56) | ||||
| Fair/poor | Ref | ||||
| BMI (kg/m2) | |||||
| <30 | Ref | ||||
| ≥30 | 1.91 (1.32–2.76) | ||||
| Missing | 0.38 (0.09–1.51) | ||||
Model 1 includes sociodemographic variables (age, sex, and language of interview) and NHANES survey year. Model 2 includes model 1 plus employment status. Model 3 includes model 2 plus insurance type. Model 4 includes model 3 plus health care utilization variables. Model 5 includes model 4 plus family history of diabetes, self-rated health, and BMI. Ref, reference.
CONCLUSIONS
Using combined NHANES data from 2005 to 2018, we observed that among Hispanic/Latino adults, younger individuals, males, and immigrants to the U.S. had a higher odds of undiagnosed diabetes. These disparities seemed partly attributable to poor health care utilization and not necessarily due to a lack of health insurance. Having no known family history of diabetes and self-reporting better health was also associated with a higher odds of undiagnosed disease. Among insured Hispanic/Latino adults, poor health care utilization and perceived better health also appeared to be prominent drivers of undiagnosed diabetes.
Our findings align with prior studies that have demonstrated a higher prevalence of undiagnosed diabetes among adults aged <45 years compared with older adults in the U.S. across all racial/ethnic groups (1,7). Accordingly, in 2021–2022, the U.S. Preventive Services Task Force and the ADA lowered the age threshold for prediabetes/diabetes screening from 45 to 35 years (23,28). Since Hispanic/Latino ethnicity is also part of the criteria for screening, this lower age threshold is expected to identify more individuals among this high-risk population (23). However, the success of any changes to screening will depend on whether individuals engage with the health care system.
More recent Hispanic/Latino immigrants to the U.S. also had a significantly higher odds of undiagnosed diabetes than U.S.-born Hispanic/Latino individuals. These differences in diabetes awareness appeared to be partly attributable to poor health care utilization. Research has highlighted the barriers that immigrants face in accessing health care services (13,29). For example, restrictive U.S. laws and policies have historically excluded undocumented immigrants from qualifying for health insurance and other social services. This reduces access to preventive health care services, thereby decreasing the likelihood of being screened for diabetes. Research has also documented disparities in health care access, demonstrating that recent immigrants were the least likely to have a primary care provider or recent health care visit (13). Other social and systemic factors that may prevent immigrants from seeking and receiving optimal care include cultural and language barriers, health literacy, financial hardship, and lack of familiarity with the U.S. health care system (13,30–32).
Although expanding health insurance coverage remains a critical intervention strategy to improve access to screening and treatment, much more work is needed to improve engagement with the health care system to address disparities in diabetes awareness. Hispanic/Latino individuals are the racial/ethnic group least likely to visit a physician (30). A combination of structural and individual-level factors is likely at play. From a health education standpoint, emphasizing the importance of annual primary care checkups for all Hispanic/Latino adults and the reasons for them, even in the context of perceived good health, may improve attendance at preventive health screenings. In addition, expanding the availability of culturally and linguistically competent care will be important from a broader structural standpoint. For Hispanic/Latino individuals, not having a language-concordant provider or access to interpreter services has been identified as a barrier to quality medical care (13,20,21,30). Moreover, an overall mistrust of the Western approach to health care has also contributed to increased apprehension about engaging with providers, even among Hispanic/Latino individuals with health insurance (33).
Aside from these factors, there may also be a need to reduce constraints associated with traditional in-person medical visits. In our analyses, being employed was associated with undiagnosed diabetes in bivariate analyses and partly attenuated estimates of undiagnosed diabetes for younger aged individuals. The higher OR estimates for the employed (vs. unemployed) also approached the null after incorporating the health care utilization variables into the model. Factors such as unpaid leave from work and limited time in general may create barriers to in-person attendance at physicans’ offices. Moreover, limited transportation and childcare responsibilities have often been cited as reasons for missing medical appointments (34–36). Expanding telehealth may be one way to address some of these challenges, though other creative approaches will likely be necessary to improve health care utilization for Hispanic/Latino individuals.
Although a lack of health insurance has been associated with undiagnosed diabetes in other studies (14,37), null findings have also been reported, especially after fully adjusting for other variables, including health care utilization (15). However, in our study, being uninsured was paradoxically associated with a lower odds of undiagnosed diabetes after adjustment for health care utilization variables. The reasons for this are unclear. Health insurance status is a widely documented determinant of health care access, but access to quality care and utilization of health care services involves more than just having health insurance (31,38). Even for uninsured individuals, diabetes screening may be accessible via federally qualified health centers, other community clinics, health fairs, and less ideally, hospital emergency departments (39). In our study, 45% of Hispanic/Latino adults without health insurance nevertheless had two or more health care encounters in the past year. If uninsured Hispanic/Latino adults in our sample were more likely to be screened once in care, this could potentially account for the patterns we observed. Nevertheless, these findings do not discount the importance of having health insurance for improving access to care more generally and for improving management outcomes after receiving positive test results upon screening. However, they do highlight the importance of addressing barriers to engagement with the health care system, outside of just having health insurance, to improve diabetes awareness for Hispanic/Latino adults.
This study has some limitations. First, we cannot conclude that any of the risk factors we identified are a cause of undiagnosed diabetes. For some of the risk factors, reverse causation is a possibility. For example, individuals with diagnosed diabetes may have had more health care visits than individuals with undiagnosed diabetes because of the need to manage their illness. They are also likely to rate their health as worse than individuals who are unaware of their disease status. Future research will need to move beyond cross-sectional designs to better identify causal risk factors. Second, ADA screening guidelines recommend repeating abnormal HbA1c or glucose measurements for diagnostic confirmation; however, NHANES only collected these measurements once. Third, although the time frame of our study covered the period during which the Affordable Care Act (ACA) was enacted, it was underpowered to detect whether risk factors for undiagnosed diabetes differed before and after the implementation of the ACA. Continued monitoring of undiagnosed diabetes and associated risk factors in the years since the ACA, especially for a high-risk group like Hispanic/Latino individuals, will be important. Fourth, NHANES does not distinguish between type 1 and type 2 diabetes, preventing us from evaluating whether the risk factors for awareness differed for these two disease processes. Nevertheless, the study had several strengths. We used nationally representative data to assess risk factors for undiagnosed diabetes within the racial/ethnic group with the highest burden of undiagnosed disease in the U.S. The practice of characterizing Hispanic/Latino individuals as a single, high-risk racial/ethnic group masks enormous heterogeneity and limits our ability to appropriately tailor interventions accordingly. We also used sequential regression models to understand the explanatory power of different variable sets rather than adjusting for all covariates at once. This contributed to our understanding of why some segments of the population were more likely to be undiagnosed. Future studies with more rigorous study designs will be needed to confirm these findings. Finally, our secondary analysis among Hispanic/Latino adults with health insurance adds to our understanding of why undiagnosed diabetes remains a problem, even for those with health care coverage.
Poor diabetes awareness contributes to worse glycemic control and disease-related complications due to prolonged undetected periods (1,8,9). Improving diabetes detection among Hispanic/Latino individuals is vital to reducing disparities in diabetes-related morbidity, mortality, and associated health care costs. Overall, our study shows that younger age, male sex, and being an immigrant were predictors of undiagnosed diabetes. Lack of health care utilization, no family history of diabetes, and better perceived health were factors that attenuated estimates for these sociodemographic groups, most of which remained key predictors of undiagnosed diabetes among Hispanic/Latino adults in fully adjusted models. Younger age was also a major contributor of undiagnosed diabetes among individuals with health insurance. Our findings highlight the importance of creating an accessible health care system, which includes, but is not limited to, expanding health insurance coverage. Future work should explore ways to increase access to screening for diabetes, especially for younger and immigrant adults, while also addressing the barriers that prevent Hispanic/Latino adults from utilizing health care.
This article contains supplementary material online at https://doi.org/10.2337/figshare.25888720.
Article Information
Funding. This research received support from National Institute of Diabetes and Digestive and Kidney Diseases grant K01DK107791 and the Margaret E. Mahoney Fellowship in Health Policy through the New York Academy of Medicine.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. J.N.d.R. and S.S.A. designed the study concept and acquired data. J.N.d.R. and S.S.A. performed data analysis and interpreted the data. J.N.d.R. drafted the initial manuscript. S.S.A. revised and finalized the manuscript for publication. J.N.d.R. and S.S.A. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Handling Editors. The journal editors responsible for overseeing the review of the manuscript were John B. Buse and Casey M. Rebholz.
Funding Statement
This research received support from National Institute of Diabetes and Digestive and Kidney Diseases grant K01DK107791 and the Margaret E. Mahoney Fellowship in Health Policy through the New York Academy of Medicine.
References
- 1. Cowie CC. Diabetes diagnosis and control: missed opportunities to improve health: the 2018 Kelly West Award Lecture. Diabetes Care 2019;42:994–1004 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Cheng YJ, Kanaya AM, Araneta MRG, et al. Prevalence of diabetes by race and ethnicity in the United States, 2011–2016. JAMA 2019;322:2389–2398 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Fisher-Hoch SP, Vatcheva KP, Rahbar MH, McCormick JB. Undiagnosed diabetes and pre-diabetes in health disparities. PLoS One 2015;10:e0133135. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Russell E, Oh KM, Zhao X. Undiagnosed diabetes among Hispanic and White adults with elevated haemoglobin A(1c) levels. Diabetes Metab Res Rev 2019;35:e3153. [DOI] [PubMed] [Google Scholar]
- 5. Britton LE, Hussey JM, Crandell JL, Berry DC, Brooks JL, Bryant AG. Racial/ethnic disparities in diabetes diagnosis and glycemic control among women of reproductive age. J Womens Health (Larchmt) 2018;27:1271–1277 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Geiss LS, Bullard KM, Brinks R, Hoyer A, Gregg EW. Trends in type 2 diabetes detection among adults in the USA, 1999–2014. BMJ Open Diabetes Res Care 2018;6:e000487. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Fang M, Wang D, Coresh J, Selvin E. Undiagnosed diabetes in U.S. adults: prevalence and trends. Diabetes Care 2022;45:1994–2002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Saydah S, Cowie C, Eberhardt MS, De Rekeneire N, Narayan KM. Race and ethnic differences in glycemic control among adults with diagnosed diabetes in the United States. Ethn Dis 2007;17:529–535 [PubMed] [Google Scholar]
- 9. Chiou T, Tsugawa Y, Goldman D, Myerson R, Kahn M, Romley JA. Trends in racial and ethnic disparities in diabetes-related complications, 1997–2017. J Gen Intern Med 2020;35:950–951 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Murphy SL, Xu J, Kochanek KD, Arias E, Tejada-Vera B. Deaths: final data for 2018. Natl Vital Stat Rep 2021;69:1–83 [PubMed] [Google Scholar]
- 11. American Diabetes Association. Economic costs of diabetes in the U.S. in 2017. Diabetes Care 2018;41:917–928 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Funk C, Lopez MH. A brief statistical portrait of US Hispanics. Pew Research Center, 14 June 2022. Accessed 1 May 2024. Available from https://www.pewresearch.org/science/2022/06/14/a-brief-statistical-portrait-of-u-s-hispanics/ [Google Scholar]
- 13. Velasco-Mondragon E, Jimenez A, Palladino-Davis AG, Davis D, Escamilla-Cejudo JA. Hispanic health in the USA: a scoping review of the literature. Public Health Rev 2016;37:31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Zhang X, Geiss LS, Cheng YJ, Beckles GL, Gregg EW, Kahn HS. The missed patient with diabetes: how access to health care affects the detection of diabetes. Diabetes Care 2008;31:1748–1753 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Menke A, Casagrande S, Avilés-Santa ML, Cowie CC. Factors associated with being unaware of having diabetes. Diabetes Care 2017;40:e55–e56 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. McWilliams JM, Meara E, Zaslavsky AM, Ayanian JZ. Differences in control of cardiovascular disease and diabetes by race, ethnicity, and education: U.S. trends from 1999 to 2006 and effects of Medicare coverage. Ann Intern Med 2009;150:505–515 [DOI] [PubMed] [Google Scholar]
- 17. Keisler-Starkey K, Bunch LN.; U.S. Census Bureau. Current Population Reports, P60-274, Health Insurance Coverage in the United States: 2020, Washington, DC, U.S. Government Publishing Office, September 2021 [Google Scholar]
- 18. Hill J, Nielsen M, Fox MH. Understanding the social factors that contribute to diabetes: a means to informing health care and social policies for the chronically ill. Perm J 2013;17:67–72 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Walker RJ, Smalls BL, Campbell JA, Strom Williams JL, Egede LE. Impact of social determinants of health on outcomes for type 2 diabetes: a systematic review. Endocrine 2014;47:29–48 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Walker RJ, Strom Williams J, Egede LE. Influence of race, ethnicity and social determinants of health on diabetes outcomes. Am J Med Sci 2016;351:366–373 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Morales J, Glantz N, Larez A, et al. Understanding the impact of five major determinants of health (genetics, biology, behavior, psychology, society/environment) on type 2 diabetes in U.S. Hispanic/Latino families: Mil Familias - a cohort study. BMC Endocr Disord 2020;20:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Johnson CL, Paulose-Ram R, Ogden CL, et al. National Health and Nutrition Examination Survey: analytic guidelines, 1999–2010. Vital Health Stat 2 2013:1–24 [PubMed] [Google Scholar]
- 23. American Diabetes Association . Standards of Medical Care in Diabetes–2022 abridged for primary care providers. Clin Diabetes 2022;40:10–38 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. U.S. Department of Health and Human Services . Social determinants of health. Healthy People 2030. Accessed 11 March 2024. Available from https://health.gov/healthypeople/priority-areas/social-determinants-health
- 25. U.S. Census Bureau . How the Census Bureau measures poverty. Accessed 25 April 2024. Available from https://www.census.gov/topics/income-poverty/poverty/guidance/poverty-measures.html
- 26. Bickel GW. Guide to measuring household food security. Washington, DC, U.S. Department of Agriculture, Food and Nutrition Service, Office of Analysis, Nutrition, and Evaluation, 2000 [Google Scholar]
- 27. Holman H, Müller F, Bhangu N, Kottutt J, Alshaarawy O. Impact of limited English proficiency on the diagnosis and awareness of diabetes: the National Health and Nutrition Examination Survey, 2003–2018. Diabetes Care 2022;45:e124–e125 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Jonas DE, Crotty K, Yun JDY, et al. Screening for prediabetes and type 2 diabetes: updated evidence report and systematic review for the US Preventive Services Task Force. JAMA 2021;326:744–760 [DOI] [PubMed] [Google Scholar]
- 29. Fernández A, Rodriguez RA. Undocumented immigrants and access to health care. JAMA Intern Med 2017;177:536–537 [DOI] [PubMed] [Google Scholar]
- 30. Funk C, Hugo Lopez M. Hispanic Americans’ experiences with health care. Pew Research Center. June 14, 2022. Accessed 11 March 2024. Available from https://www.pewresearch.org/science/2022/06/14/hispanic-americans-experiences-with-health-care/ [Google Scholar]
- 31. Escarce JJ, Kapur K. Access to and quality of health care. In Hispanics and the Future of America. National Research Council (US) Panel on Hispanics in the United States; Tienda M, Mitchell F, Eds. Washington, DC, National Academies Press, 2006, pp. 410–446 [PubMed] [Google Scholar]
- 32. Zakaria NI, Tehranifar P, Laferrère B, Albrecht SS. Racial and ethnic disparities in glycemic control among insured US adults. JAMA Netw Open 2023;6:e2336307. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Juckett G. Caring for Latino patients. Am Fam Physician 2013;87:48–54 [PubMed] [Google Scholar]
- 34. Kaplan-Lewis E, Percac-Lima S. No-show to primary care appointments: why patients do not come. J Prim Care Community Health 2013;4:251–255 [DOI] [PubMed] [Google Scholar]
- 35. Alkomos MF, Mendez D, Mazzei-Pifano D, et al. Patients’ reasons for missing scheduled clinic appointments and their solutions at a major urban-based academic medical center. J Community Hosp Intern Med Perspect 2020;10:426–430 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Rust G, Ye J, Baltrus P, Daniels E, Adesunloye B, Fryer GE. Practical barriers to timely primary care access: impact on adult use of emergency department services. Arch Intern Med 2008;168:1705–1710 [DOI] [PubMed] [Google Scholar]
- 37. Casagrande SS, Menke A, Aviles-Santa L, et al. Factors associated with undiagnosed diabetes among adults with diabetes: results from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL). Diabetes Res Clin Pract 2018;146:258–266 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Strand de Oliveira J. Access to care: beyond health insurance. J Am Acad Physician Assist 2013;26:40–45 [DOI] [PubMed] [Google Scholar]
- 39. Tolbert J, Drake P, Damico A. Key facts about the uninsured population. Kaiser Family Foundation, 18 December 2023. Accessed 11 March 2024. Available from https://www.kff.org/uninsured/issue-brief/key-facts-about-the-uninsured-population/ [Google Scholar]