Secondary Logo

Journal Logo

Original Articles

A treatment-based algorithm for identification of diabetes type in the National Health and Nutrition Examination Survey

Mosslemi, Mitraa,,b; Park, Hannah L.b; McLaren, Christine E.c; Wong, Nathan D.a,,b

Author Information
Cardiovascular Endocrinology & Obesity: March 2020 - Volume 9 - Issue 1 - p 9-16
doi: 10.1097/XCE.0000000000000189
  • Free

Abstract

Introduction

The American Diabetes Association classifies diabetes into four general categories: type 1 diabetes (T1D) (due to autoimmune β-cell destruction), type 2 diabetes (T2D) (due to a progressive loss of β-cell insulin secretion, frequently on the background of insulin resistance, non-autoimmune), gestational diabetes mellitus (diabetes diagnosed in the second or third trimester of pregnancy that was not clearly overt diabetes before gestation), and specific types of diabetes (due to other causes, e.g. monogenic forms of diabetes, diseases of the exocrine pancreas, and drug- or chemical- induced diabetes) [1,2]. Gestational diabetes mellitus is only present during pregnancy, and other specific types represent a small fraction of patients with diabetes (<5%) [1]. Therefore, diabetes is mainly classified as T1D, comprising approximately 5–10% of cases, and T2D, comprising 90–95% [1,2].

The National Health and Nutrition Examination Survey (NHANES) is a population-based program of studies with complex, multistage sampling of the civilian, non-institutionalized U.S. population to represent the population of all ages [3]. NHANES is unique in that it combines interviews with medical examination data, including laboratory tests and physiological measurements [3]. Thus, NHANES has been used to monitor, among other conditions, the prevalence of diabetes in the U.S. and to study risk factors, comorbidities, and complications associated with diabetes. However, like many other health surveys and registries, NHANES does not record the diabetes subtypes (T1D or T2D). Because of the differences in etiology, associated risk factors, and pathophysiology of complications between T1D and T2D, studies focusing on a specific diabetes type or performing comparisons between them should be able to differentiate between participants with T1D and T2D. However, there is currently no standard method to distinguish between T1D and T2D in NHANES, as summarized in Table 1, and various strategies have been used in different studies.

Table 1
Table 1:
Diabetes type identification methods used in the National Health and Nutrition Examination Survey

This article describes a method for identifying the NHANES participants with self-reported diagnosed diabetes according to whether they have T1D or T2D. Our treatment-based algorithm is independent of the patients’ age at diagnosis and instead relies on information about their reported diabetes treatment(s), in conjunction with differences between the T1D and T2D treatment recommendations from the American Diabetes Association Standard of Care [4].

Methods

We used NHANES data from 1999 to 2016 and created corresponding weighted variables to appropriately represent the non-institutionalized U.S. civilian resident population. We used all the available NHANES cycles, from 1999–2000 to 2015–2016, to have access to a large enough sample size of T1D patients [3]. The NHANES questionnaire asks about non-gestational diabetes and gestational diabetes in two separate sections, namely, the Diabetes Questionnaire (DIQ) and the Reproductive Health Questionnaire, respectively [3]. Although the DIQ includes ~27 questions regarding the diagnosis, treatment, and screening for complications, our treatment-based algorithm performed T1D vs. T2D classification using a subset of these questions:

DIQ.010: Other than during pregnancy, have you ever been told by a doctor or other health professional that you have diabetes or sugar diabetes?

DIQ.040: How old were you when a doctor or other health professional first told you that you had diabetes or sugar diabetes?

DIQ.050: Are you now taking insulin?

DIQ.060: For how long have you been taking insulin?

DIQ.070: Are you now taking diabetic pills to lower your blood sugar? These are sometimes called oral agents or oral hypoglycemic agents.

Our treatment-based algorithm was derived from the standard treatment differences between T1D and T2D. According to treatment recommendations from the American Diabetes Association from 1998 to 2018, all T1D patients should have been treated with insulin after diagnosis, and none of the oral hypoglycemic agents were FDA approved for T1D [4–6]. Recommendations for the pharmacological therapy for T2D were to initiate treatment with monotherapy oral hypoglycemic agents, though in cases of severe HbA1c level, recommendations called for dual therapy (oral hypoglycemic agents plus insulin) instead [4,5,7]. Over time, some T2D patients may become insulin deficient, and thus need insulin therapy, as well [8–10]. Some of these patients may discontinue oral medication for relief from adverse side effects of the oral agents and use insulin monotherapy [8–10]. Overall, the treatment guidelines for T1D and T2D have been reasonably distinct, which enables the identification of diabetes type among the NHANES participants with diagnosed diabetes through their treatment information.

Our treatment-based algorithm for identifying diabetes type is depicted in Fig. 1. It defines T2Ds as patients whose treatment was one of the following: no medication, oral hypoglycemic agents alone, or oral hypoglycemic agents plus insulin. Our treatment-based algorithm identifies T1Ds as patients who used only insulin therapy and started taking insulin within one year of diagnosis. The remaining group contains patients who took only insulin but started insulin therapy over one year after diagnosis. This group should include T2D patients who became insulin-requiring years after diagnosis. However, T2D patients who are also positive for at least one of the autoimmune β-cell destruction autoantibodies (known as type 1.5 or latent autoimmune diabetes in adults) [11,12] and T1D patients who initially were misdiagnosed, so their insulin therapy was delayed could also be in this category. Therefore, we kept this group separate, and our treatment-based algorithm classifies this group as ‘possible-T2D’.

Fig. 1
Fig. 1:
Treatment-based algorithm for identification of diabetes type.

The target population for our treatment-based algorithm is participants with self-reported diagnosed diabetes, that is, who replied ‘yes’ to having been told by a doctor or other health professional that they have diabetes (DIQ.010 = 1) (n = 5645). For all these participants, we calculated the time since diagnosis as age at participation minus the age at diagnosis (DIQ.040). Our treatment-based algorithm requires five steps to identify the NHANES participants with self-reported diagnosed diabetes as having T1D or T2D (Fig. 1) as follows:

Step 1

Participants who did not have a reply to treatment questions (DIQ.050, DIQ.060, DIQ.070) – either ‘don’t know’ or ‘refused’ response – and those whose insulin treatment period (DIQ.060) appeared longer than their time since diagnosis were excluded (n = 188). The rest were moved forward (n = 5457), comprising our analytic cohort.

Step 2

Participants who replied ‘no’ to DIQ.050 and ‘no’ to DIQ.070, meaning that they were neither currently taking insulin nor an oral hypoglycemic agent, were assigned as T2D (n = 860). The rest were moved forward (n = 4597).

Step 3

Participants who replied ‘no’ to DIQ.050 and therefore were not currently taking insulin were assigned as T2D (n = 3141). The rest were moved forward (n = 1456).

Step 4

Participants who replied ‘yes’ to DIQ.050 and ‘yes’ to DIQ.070, meaning they were taking both insulin and an oral hypoglycemic agent, were assigned as T2D (n = 731). The rest were moved forward (n = 725).

Step 5

Participants whose time since diagnosis minus their answer to DIQ.060, insulin treatment period, was one year or less, meaning they started taking insulin within a year after diagnosis, were assigned as T1D (n = 280). The rest were labeled as ‘possible-T2D’ (n = 445).

Using this treatment-based algorithm, we classified all participants who self-reported as having been diagnosed with diabetes and replied to the treatment questions in NHANES 1999–2016 as T1D, T2D, and possible-T2D.

Results

Of the 5457 (representing 18.1 million in the US) participants with self-reported diagnosed diabetes and valid medication information, our treatment-based algorithm identified 280 (1.1 million, 6.1%) T1D, 4732 (15.6 million, 86.1%) T2D, and 445 (1.42 million, 7.8%) possible-T2D cases.

Treatment information of the T1D, T2D, and possible-T2D participants identified by the treatment-based algorithm, which reflect the variables used to make the assignments, is presented in Table 2. As defined by the algorithm criteria, all T1D respondents were taking insulin and started insulin treatment within one year of their diabetes diagnosis, and no T1D participant was taking oral hypoglycemic medication. The timing of the start of insulin treatment after diagnosis was used to differentiate between diabetes type at the end of the treatment-based algorithm, and to be classified as a T1D case, the participant had to have started taking insulin within one year of diagnosis. Based on the American Diabetes Association Standard of Care, a patient with T1D must start insulin treatment right after the diagnosis, but in practice, the confirmation of the diagnosis and the initiation of treatment takes time for many cases [4]. We kept one year as the cutoff for initiation of insulin therapy after diagnosis for T1D identification, since it has been previously used by Menke et al. [13] in their estimate of the prevalence of T1D using NHANES 1999–2010.

Table 2
Table 2:
Treatment information of participants with diagnosed diabetes classified by our treatment-based algorithm as T1D, T2D, and possible-T2D

Among cases identified as T2D by our treatment-based algorithm, 19% were taking no medications, 66% were taking only oral hypoglycemic medication, and 15% were taking both oral hypoglycemic medication and insulin. About 11% of T2D patients (1.7 million) identified by our treatment-based algorithm had been diagnosed when they were younger than 30. On the other hand, among participants identified as T1D by our treatment-based algorithm, 43% (481 237) had been diagnosed when they were 30 years or older.

We also examined the demographics and clinical characteristics of the T1D and T2D cases assigned by our treatment-based algorithm for validation (Table 3). As expected, the T1D group had a younger average diagnosis age than T2D (28 vs. 49 years) and longer time since diagnosis (17 vs. 10 years). Moreover, the T1D group had higher average HbA1c (8.1% vs. 7.2%) and lower average BMI (30.0 vs. 32.8 kg/m2), and the gender and race distributions for T1D were more male (58% vs. 49%) and considerably more from the White race (75% vs. 60%).

Table 3
Table 3:
Demographics and clinical characteristics of participants with diagnosed diabetes classified by our treatment-based algorithm as T1D, T2D, and possible-T2D

The demographics and BMI of the possible-T2D group were similar to the T2D group. Only their time since diagnosis and HbA1c appeared closer to the T1D group, suggesting that the group is comprised mainly of T2D cases with longer time since diagnosis.

Discussion

We introduced a treatment-based algorithm to identify cases of T1D and T2D across all ages among participants with diagnosed diabetes in the cross-sectional population of NHANES 1999–2016. Our treatment-based algorithm-defined T1Ds represented 6.1% (1.1 of 18.1 million) of all diagnosed diabetes cases, with 57% (0.6 million) of T1D cases diagnosed when patients were aged younger than 30 years, and 43% (0.5 million) diagnosed aged 30–80 years or older. Our treatment-based algorithm-defined T2Ds and possible-T2Ds together comprised 93.9% (17 of 18.1 million) of all diagnosed diabetes cases, with 9.5% (1.6 million) of all T2D cases diagnosed under the age 30, and 90.5% (15.4 million) diagnosed at age 30 or above. To our knowledge, our treatment-based algorithm is the first algorithm to consider contemporary data on age distributions of T1D and T2D and consequently exclude the age at diagnosis criterion for diabetes type identification in health registries [13–21].

Age at diagnosis was applied in all the methods for diabetes type identification in previous NHANES studies as the initial criterion for type identification because T1D is more common than T2D in younger ages [13,16–21]. Although it had been known that T1D could occur at any age which gradually led to the eradication of the use of the term ‘juvenile diabetes’ for T1D, the exact distribution of the disease across all ages was not determined [22]. Thomas et al. [23], using a novel validated T1D genetic risk score, demonstrated that, in their UK population-based cohort with a broad age range (0–60 years), 42% of genetically defined T1D cases were diagnosed after age 30. Therefore, using the age of diagnosis as a classification criterion could misclassify a substantial proportion of T1D cases as T2D. Additionally, the age at diagnosis would misclassify all T2D cases diagnosed younger than the threshold (commonly 30 years) as T1D. Since T2D among children and adolescents in the US have considerable prevalence (24:100 000 in 2009) and increasing incidence (4.8% annual increase between 2002 and 2012), such misclassification is not negligible [24–26].

Applying age at diagnosis as the only criterion (Table 1, method #2) would have resulted in 12.4% of our analytic cohort being identified as T1D, which, as previously mentioned, is an overestimation since estimates of the prevalence of T1D among US adults were closer to 6% [16,27–29]. In contrast, our treatment-based algorithm identified 6.1% of all self-reported diabetes cases as T1D, which is compatible with the current estimates [28,29].

Criteria besides the age at diagnosis that have been used in the previous NHANES studies (Table 1) improve the accuracy of classification compared to using the age at diagnosis alone; however, they impose some specific differential misclassifications. For example, it is well-established that patients with young-onset T2D have a higher-risk of cardiovascular complications compared with patients with older age of onset [30–32]. By using the age at diagnosis as one of the criteria for the classification, as has been the case in all the NHANES diabetes type identification methods, these higher risk participants or a portion of them would be misclassified as T1D cases. Such misclassification would underestimate the cardiovascular disease risk in T2D and overestimate it in the T1D cohort.

It is essential to point out that since only around 5% of diabetes cases are T1D, when the target population is T2D, considering all diabetes cases as T2D would have merely 5% misclassification. Therefore, performing the study without any classification could be a more coherent choice than using an imprecise classification criterion like age at diagnosis, which would lead to specific biases in the results. However, a precise algorithm is needed for identifying T1D cases, either for an estimate of the prevalence of T1D or comparison studies between T1Ds and T2Ds. The current trend of using inconsistent and less precise diabetes type identification methods not only decreases the comparability of the results across the NHANES studies but also distorts the understanding of the differences between diabetes type, as observed in a study by Zhang et al. [21] where they considered age at participation for diabetes type identification (Table 1, method #5).

Since no validated method exists to distinguish between diabetes types in surveys, the exact national prevalence of T1D in the US is not known. Menke et al. [13] previously provided an estimate for the prevalence of T1D in the US, using NHANES 1999–2010 (Table 1, method #4). They formulated their classification method by adding the initiation time of insulin therapy after the diabetes diagnosis to the age of diagnosis criterion [13]. They identified participants to have T1D if they started insulin within one year of diabetes diagnosis, were currently using insulin and were diagnosed with diabetes under age 30 (definition 1) or age 40 (definition 2) [13]. They reported the prevalence of T1D in the US population as 2.6/1000 based on definition 1 and 3.4/1000 based on definition 2. The limitations of the Menke et al. approach was that their estimates would have missed T1D cases diagnosed after the age of 30 (or 40) and included some younger people with T2D cases who required insulin within a year of diagnosis. Our algorithm improved these limitations by allowing the inclusion of T1D cases diagnosed at any age and restricting the inclusion of T2D cases who take insulin, by checking the oral medication status. Based on our treatment-based algorithm, the estimated prevalence of T1D, for NHANES 1999–2010, was 3.8/1000 (1,110,011 people), which is higher than both Menke et al. [13] estimates (2.6/1000 and 3.4/1000).

In addition, the ‘SEARCH for Diabetes in Youth’ (SEARCH), which was a multi-center observational study conducting population-based ascertainment of physician-diagnosed diabetes in youth, identified a T1D prevalence of 1.9/1000 among youth under the age of 20 years in 2009 [24–26]. As Menke et al. mentioned in their T1D prevalence report, their approach made a higher estimate (2.4/1000) than the one reported by the SEARCH [13,24]. When we used our treatment-based algorithm for the participants under 20 years of age, the prevalence for T1D became 1.7/1000, which is closer to the reported prevalence of T1D under the age of 20 years by SEARCH (1.9/1000) [24].

Moreover, since 2016, the National Health Interview Survey (NHIS) asked participants who had ever received a diagnosis of diabetes to report the subtypes (type 1 or type 2, other types, or unknown). Using the NHIS supplemented question, Bullard et al. reported 5.8% as the proportion of T1Ds among US adults (age ≥ 18 years) with diagnosed diabetes in 2016. Our treatment-based algorithm identified 5.6% of adults (age ≥ 18 years) with diagnosed diabetes in 1999–2016 NHANES as T1D, which is compatible with the NHIS report [29].

Additionally, the demographic and clinical characteristics of our treatment-based algorithm-defined T1D and T2D groups, as presented in Table 3, appear reasonable in comparison with contemporary epidemiologic reports. Overall, our treatment-based algorithm-defined T1D group had a younger average diagnosis age, longer diabetes duration, higher average HbA1c, and lower average BMI than the T2D group, as expected [33–35]. The sex and race distribution for the T1D group defined by our treatment-based algorithm were more male (1.4 male/female ratio) and more from the White race (the proportion of T1D cases among all diagnosed diabetes cases were 6.1% for all races but 7.5% among the non-Hispanic white population). Many reports have indicated an excess of T1D incidence in males after the pubertal years in populations of European descent, with the incidence rate ratio of male-female between 1.3- and 2-fold [33,36,37]. Besides, the SEARCH study, which provided specific data on race/ethnicity among children and adolescents with diabetes within the US, indicated that non-Hispanic whites had the highest rate of new cases of T1Ds compared to members of other US racial groups [38].

Our algorithm may still underestimate the prevalence of T1D, despite reporting a higher prevalence than the Menke et al. report, by excluding T1D patients who took oral hypoglycemic medications. The use of oral hypoglycemic medications as adjunct to insulin is not the standard treatment for T1D patients. Still, a relatively small number of patients with T1D take oral hypoglycemic medications. According to a report from the T1D Exchange Clinic Network Registry, which covers patients with T1D at all ages across the lifespan, and is currently active in 33 US states, less than 4% of patients with T1D take oral glucose-lowering medications [35]. Based on this estimate, our algorithm may misclassify around 4% of T1D patients who took oral hypoglycemic medication, which for NHANES 1999–2016 would represent 46,410 patients.

Since the T1D group is considerably smaller than the T2D group, it is more sensitive to any misclassification. Accordingly, we compromised for the abovementioned misclassification to protect the precision of our T1D classification. The ADA guidelines recommend metformin as the first-line treatment for T2D patients, which implies that early use of insulin in T2D patients would be as an adjunct to metformin. Therefore, we applied taking oral hypoglycemic medications as a determinant for identifying T2D patients who may have started insulin therapy within a year after diagnosis.

Furthermore, using the status of taking oral medications as an identifier for T2D could prevent the misclassification of T2D patients who use diabetes medications that are neither tablets nor insulin (like GLP-1 agonists) but have mistakenly reported these as insulin. Based on a report from the Medical Expenditure Panel Survey, which is a nationally representative US survey, in 2015, 4.4% of T2D patients take GLP-1. Still, none of them takes it as the first-line treatment/monotherapy, compatible with the ADA treatment guidelines for T2D patients [39]. Therefore, the patients who may have mistakenly reported injection-type diabetes medication as insulin would have been identified as T2D by our algorithm, based on their concurrent use of oral hypoglycemic medication.

Although our method is still prone to some misclassification due to inaccuracies in self-reporting or potential deviations from the Standard of Care in clinical practice, it does not induce additional misclassification biases. Nonetheless, further development and validation of our treatment-based algorithm is needed. We did not have access to any health registry or clinical database that contains diabetes type information. Consequently, the chief limitation in this work is the lack of validation of the algorithm against such data for deriving the sensitivity and specificity measures of our treatment-based algorithm classification.

We conclude that the use of diabetes treatment information is a pragmatic and valid approach to identifying diabetes type in health registries like NHANES, without any specific contradiction with the existing body of knowledge on diabetes type. Therefore, the application of our treatment-based algorithm for diabetes-related studies in NHANES would improve the performance of diabetes type identification and enable comparability across studies.

Acknowledgements

We would like to acknowledge the staff and participants of the National Health and Nutrition Examination Survey program.

This work was previously presented, in part, at the 52nd Annual Meeting of the Society for Epidemiologic Research (SER), held 18–21 June 2019, in Minneapolis, Minnesota, USA.

Data sets supporting the results of this article are available in the NHANES Data Portal repository, https://wwwn.cdc.gov/nchs/nhanes/. No patient consent form is required as we have used publicly available data.

Conflicts of interest

N.D.W. receives research support through his institution not related to this study from Boehringer-Ingelheim and Novo Nordisk and is a consultant for Astra-Zeneca. For the remaining authors, there are no conflicts of interest.

References

1. American Diabetes Association. 2. Classification and diagnosis of diabetes: standards of medical care in diabetes—2018. Diabetes Care. 2018; 41Suppl. 1S13–S27
2. Genuth SM, Palmer JP, Nathan DM. Cowie CC, Casagrande SS, Menke A. Classification and diagnosis of diabetes. Diabetes in America. 20183rd edNational Institutes of Health, NIH Pub No. 17-14681.1–1.39. https://www.niddk.nih.gov/about-niddk/strategic-plans-reports/diabetes-in-america-3rd-edition#suggest
3. Centers for Disease Control and Prevention (CDC). National Center for Health Statistics (NCHS). National Health and Nutrition Examination Survey. 2018, Hyattsville, MD US: Centers Dis Control Prev (CDC) Natl Cent Heal Stat (NCHS), Dep Heal Hum Serv Centers Dis Control Prev. http://www.cdc.gov/ncidod/dvbid/westnile/index.htm. [Accessed 1 August, 2018]
4. American Diabetes Association. 8. Pharmacologic approaches to glycemic treatment: standards of medical care in diabetes—2018. Diabetes Care. 2018; 41Suppl. 1S73–S85
5. American Diabetes Association. Standards of medical care for patients with diabetes mellitus. Diabetes Care. 2003; 26:S33–S50
6. Skyler JS. Medical Management of Type 1 Diabetes. 19983rd ed, Alexandria, VA: American Diabetes Association
7. Zimmerman BR. Medical Management of Type 2 Diabetes. 19984th ed, Alexandria, VA: American Diabetes Association
8. Abdi H, Azizi F, Amouzegar A. Insulin monotherapy versus insulin combined with other glucose-lowering agents in type 2 diabetes: a narrative review. Int J Endocrinol Metab. 2018; 16:e65600
9. Vos RC, van Avendonk MJ, Jansen H, Goudswaard AN, van den Donk M, Gorter K, et al. Insulin monotherapy compared with the addition of oral glucose-lowering agents to insulin for people with type 2 diabetes already on insulin therapy and inadequate glycaemic control. Cochrane Database Syst Rev. 2016; 9:CD006992
10. Bianca H, Lundby CL, Jørn W, Vaag A, Gluud C, Lund SS, et al. Comparison of metformin and insulin versus insulin alone for type 2 diabetes: systematic review of randomised clinical trials with meta-analyses and trial sequential analyses. BMJ. 2012; 344:e1771
11. Palmer JP, Hampe CS, Chiu H, Goel A, Brooks-Worrell BM. Is latent autoimmune diabetes in adults distinct from type 1 diabetes or just type 1 diabetes at an older age?. Diabetes. 2005; 54Suppl 2S62–S67
12. Pieralice S, Pozzilli P. Latent autoimmune diabetes in adults: a review on clinical implications and management. Diabetes Metab J. 2018; 42:451–464
13. Menke A, Orchard TJ, Imperatore G, Bullard KM, Mayer-Davis E, Cowie CC. The prevalence of type 1 diabetes in the united states. Epidemiology. 2013; 24:773–774
14. Eastwood SV, Mathur R, Atkinson M, Brophy S, Sudlow C, Flaig R, et al. Algorithms for the capture and adjudication of prevalent and incident diabetes in UK biobank. PLoS One. 2016; 11:e0162388
15. Ng E, Dasgupta K, Johnson JA. An algorithm to differentiate diabetic respondents in the Canadian Community Health Survey. Health Rep. 2008; 19:71–79
16. Wong ND, Glovaci D, Wong K, Malik S, Franklin SS, Wygant G, Iloeje U. Global cardiovascular disease risk assessment in United States adults with diabetes. Diab Vasc Dis Res. 2012; 9:146–152
17. Koopman RJ, Mainous AG III, Diaz VA, Geesey ME. Changes in age at diagnosis of type 2 diabetes mellitus in United States, 1988 to 2000. Ann Fam Med. 2005; 3:60–63
18. Wong K, Glovaci D, Malik S, Franklin SS, Wygant G, Iloeje U, et al. Comparison of demographic factors and cardiovascular risk factor control among U.S. adults with type 2 diabetes by insulin treatment classification. J Diabetes Complications. 2012; 26:169–174
19. Wong ND, Patao C, Wong K, Malik S, Franklin SS, Iloeje U. Trends in control of cardiovascular risk factors among US adults with type 2 diabetes from 1999 to 2010: comparison by prevalent cardiovascular disease status. Diabetes Vasc Dis Res. 2013; 10:505–513
20. Wang Y, Lopez JMS, Bolge SC, Zhu VJ, Stang PE. Depression among people with type 2 diabetes mellitus, US National Health and Nutrition Examination Survey (NHANES), 2005–2012. BMC Psychiatry. 2016; 16:88
21. Zhang N, Yang X, Zhu X, Zhao B, Huang T, Ji Q. Type 2 diabetes mellitus unawareness, prevalence, trends and risk factors: National Health and Nutrition Examination Survey (NHANES) 1999–2010. J Int Med Res. 2017; 45:594–609
22. Daneman D. Type 1 diabetes. Lancet. 2006; 367:847–858
23. Thomas NJ, Jones SE, Weedon MN, Shields BM, Oram RA, Hattersley AT. Frequency and phenotype of type 1 diabetes in the first six decades of life: a cross-sectional, genetically stratified survival analysis from UK biobank. Lancet Diabetes Endocrinol. 2018; 6:122–129
24. Pettitt DJ, Talton J, Dabelea D, Divers J, Imperatore G, Lawrence JM, et al. Prevalence of diabetes in U.S. Youth in 2009: The SEARCH for diabetes in youth study. Diabetes Care. 2014; 37:402–408
25. Hamman RF, Bell RA, Dabelea D, D’Agostino RB Jr, Dolan L, Imperatore G, et al; SEARCH for Diabetes in Youth Study Group. The SEARCH for diabetes in youth study: rationale, findings, and future directions. Diabetes Care. 2014; 37:3336–3344
26. Mayer-Davis EJ, Lawrence JM, Dabelea D, Divers J, Isom S, Dolan L, et al. Incidence trends of type 1 and type 2 diabetes among youths, 2002–2012. N Engl J Med. 2017; 376:1419–1429
27. Centers for Disease Control and Prevention. National Diabetes Statistics Report, 2017. 2017, Atlanta, GA: Centers for Disease Control and Prevention, U.S. Dept of Health and Human Services. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf. [Accessed 1 April 2019]
28. Fitch K, Weisman T, Engel T, Turpcu A, Blumen H, Rajput Y, Dave P. Longitudinal commercial claims-based cost analysis of diabetic retinopathy screening patterns. Am Health Drug Benefits. 2015; 8:300–308
29. Bullard KM, Cowie CC, Lessem SE, Saydah SH, Menke A, Geiss LS, et al. Prevalence of diagnosed diabetes in adults by diabetes type — United States, 2016. MMWR Morb Mortal Wkly Rep. 2018; 67:359–361
30. Wannamethee SG, Shaper AG, Whincup PH, Lennon L, Sattar N. Impact of diabetes on cardiovascular disease risk and all-cause mortality in older men: influence of age at onset, diabetes duration, and established and novel risk factors. Arch Intern Med. 2011; 171:404–410
31. Huo L, Magliano DJ, Ranciere F, Harding JL, Nanayakkara N, Shaw JE, et al. Impact of age at diagnosis and duration of type 2 diabetes on mortality in Australia 1997-2011. Diabetologia. 2018; 61:1055–1063
32. Sattar N, Rawshani A, Franzén S, Rawshani A, Svensson AM, Rosengren A, et al. Age at diagnosis of type 2 diabetes mellitus and associations with cardiovascular and mortality risks. Circulation. 2019; 139:2228–2237
33. Maahs DM, West NA, Lawrence JM, Mayer-Davis EJ. Epidemiology of type 1 diabetes. Endocrinol Metab Clin North Am. 2010; 39:481–497
34. Gudbjörnsdottir S, Cederholm J, Nilsson PM, Eliasson B; Steering Committee of the Swedish National Diabetes Register. The national diabetes register in Sweden: an implementation of the st. Vincent declaration for quality improvement in diabetes care. Diabetes Care. 2003; 26:1270–1276
35. Miller KM, Foster NC, Beck RW, Bergenstal RM, DuBose SN, DiMeglio LA, et al; T1D Exchange Clinic Network. Current state of type 1 diabetes treatment in the U.S.: updated data from the T1D exchange clinic registry. Diabetes Care. 2015; 38:971–978
36. Rogers MAM, Kim C, Banerjee T, Lee JM. Fluctuations in the incidence of type 1 diabetes in the United States from 2001 to 2015: a longitudinal study. BMC Med. 2017; 15:199
37. Wändell PE, Carlsson AC. Time trends and gender differences in incidence and prevalence of type 1 diabetes in Sweden. Curr Diabetes Rev. 2013; 9:342–349
38. Mayer-Davis EJ, Bell RA, Dabelea D, D’Agostino R Jr, Imperatore G, Lawrence JM, et al. The many faces of diabetes in American youth: type 1 and type 2 diabetes in five race and ethnic populations: the SEARCH for Diabetes in Youth Study. Diabetes Care. 2009; 32Suppl 2S99–101
39. Raval AD, Vyas A. National trends in diabetes medication use in the United States: 2008 to 2015. J Pharm Pract. 20180897190018815048
Keywords:

diabetes epidemiology; diabetes type identification; endocrinology; National Health and Nutrition Examination Survey; type 1 diabetes; type 2 diabetes

© 2020Wolters Kluwer Health Lippincott Williams Wilkins