Diabetes mellitus is a group of metabolic diseases characterized by hyperglycemia resulting from defects in insulin secretion, insulin action, or both, that affects 29 million (9.3%) people in the United States as of 2014.1 In 2010, diabetes was the seventh leading cause of death in the United States, along with being the leading cause of kidney failure, nontraumatic lower limb amputations, retinopathy, heart disease, and cardiac events, such as stroke.1,2 The standard procedure for diagnosing diabetes in the United States has historically been fasting plasma glucose (FPG) measures, oral glucose tolerance test (OGTT), and random plasma level measures in hyperglycemic individuals.3 In 2006, the American Diabetes Association (ADA) introduced the use of the glycated hemoglobin (HbA1c) test for the accurate measurement of blood glucose levels in diagnosing diabetes mellitus.3
Simultaneously, globally acclaimed agencies, such as the World Health Organization (WHO) and the International Diabetes Federation, recognized the application of HbA1c testing for the measurement and monitoring of long-term glucose levels. In 2009, the ADA International Expert Committee recommended the use of the HbA1c test as a measure in the screening or diagnosis of diabetes mellitus and impaired fasting glucose.4 At that time, the ADA announced that the HbA1c assays were highly standardized, and that the results from these tests could be uniformly applied temporally and across populations.4
The HbA1c measure has several advantages over other tests. It measures an average HbA1c level, which can be converted into estimated average glucose levels from the past 2 to 3 months, enabling physicians to use HbA1c levels as primary treatment targets.4 Furthermore, it is a convenient test, because fasting is not required, which enables greater frequency of testing use among patients; it is standardized and precise; and it has demonstrated evidence to suggest greater preanalytical stability and less day-to-day perturbations during periods of stress and illness.4
In addition to its use in diagnosing diabetes, the HbA1c measure is also widely used as a glycemic marker for the long-term monitoring of diabetes. HbA1c levels provide valuable information about the degree of glucose control during the previous 8 to 12 weeks, and it measures the risk for complications from diabetes.5,6 HbA1c levels can be used to segregate patients into different categories, such as low-risk prediabetes, high-risk prediabetes, uncontrolled diabetes, and controlled diabetes. Thus, HbA1c levels can be used as direct treatment targets, making the test a powerful monitoring tool.5 Thanks to its ease of use, standardization, precision, close association with diabetes complications, and its ability to be translated into estimated average glucose values, the HbA1c test is an effective prediabetes or diabetes monitoring and diagnostic tool.6
Diabetes-related expenditures impose a huge financial burden on the US economy. The total diabetes-related cost in the United States was $245 billion in 2012, including $176 billion in direct medical costs and $69 billion in reduced productivity, which was almost doubled from a decade earlier.7,8 Diabetes screening and monitoring procedures are relevant components of the total diabetes expenditures that need to be evaluated. After recommendations by the ADA and WHO in 2009 to use HbA1c as a glycemic marker, the test emerged as a principal diagnostic and monitoring tool given its convenience and ease of use.9,10 However, data citing the magnitude of its adoption or how quickly it was adopted by clinicians are not available.
Although HbA1c testing is more convenient, it is also more costly than the older procedures used in clinical practice. Therefore, the advent of a newer, more costly screening and monitoring tool (every 2-3 months) could have impacted the healthcare expenditures burden for patients with diabetes. Several studies have examined diabetes-related healthcare expenditure predictors, such as the utilization of healthcare services and demographic factors.7,8,11 However, the impact of a new diagnostic and monitoring tool, specifically after the introduction of HbA1c testing, on the total diabetes-related healthcare expenditures has never been evaluated. This assessment could afford a better understanding for clinicians and payers of the impact of a change in the diagnostic tools on the total healthcare spending for the diabetic population.
Thus, our primary objective is to determine if there is an association between the HbA1c tests and the total healthcare expenditures among patients with diabetes. This study is based on an assumption that in 2009, diabetic patients were newly diagnosed using the older criteria, whereas in 2011, patients were newly diagnosed with the HbA1c test. Therefore, new diabetes diagnoses in 2009 and in 2011 are considered proxies for pre– and post–HbA1c testing implementation, respectively. As a corollary to this objective, this study also determined the factors that are associated with the total healthcare expenditures among diabetic patients before and after the change in diagnostic measure with the implementation of HbA1c testing in 2009 and in 2011, respectively.
It is important to note that the HbA1c diagnostic measure was introduced as a concept in 2006. It was not standardized by the ADA in 2006, and it is, therefore, safe to assume that it was not often used in 2006 and until 2009. Only after its standardization in 2009 did the use of HbA1c testing enter into common practice. Moreover, we are assuming that this newly supported diagnostic tool would increase the number of patients who were diagnosed by this mechanism in 2011 versus 2009.
Methods
The Medical Expenditure Panel Survey-Household Component (MEPS-HC) 2009 and 2011 databases were used for this study. The MEPS-HC collects detailed annual data on demographic characteristics, health conditions, the utilization of healthcare services, charges and payments, expenditures at personal and household levels, access to care, and satisfaction with care for the US civilian noninstitutionalized population.12
We combined the self-reported patients with diabetes from the MEPS-HC component from 2009 and 2011 to form our study cohort. The outcome variable was comprised of total healthcare expenditure (ie, the self-reported amount spent by the participants in 2009 and in 2011). The MEPS-HC respondents who reported that they were diagnosed with diabetes (excluding gestational diabetes) were sampled as patients with diabetes. To best represent the HbA1c-associated diagnosis, we segregated only newly diagnosed patients in the years 2009 and 2011. Patients with diabetes were categorized as newly diagnosed in that year if the difference between the patient’s age and the age of diagnosis was 0 years or 1 year. This step ensured that the study cohort excluded patients who were diagnosed before 2009, who might have been diagnosed with older diagnostic tests such as FPG or OGTT; therefore, this maximized the possibility of including only patients diagnosed with HbA1c testing.
We created a binary variable representing new diabetes diagnoses in 2009 and in 2011, which also acted as a proxy for before and after the introduction of the HbA1c diagnostic test (1, HbA1c not used in 2009; 0, HbA1c used in 2011). This proxy variable formed the main independent variable. Other independent variables were comprised of those that influenced the diabetes-related expenditures in previous studies, including categorical sociodemographic factors such as age, sex, race, and age at diagnosis; other comorbidities; and continuous variables, such as the self-reported utilization of healthcare service variables, including total office-based visits, office-based plus outpatient visits, hospital outpatient visits, emergency department visits, inpatient hospital stays, prescription medicines, dental visits, home healthcare, and other medical events for the years 2009 and 2011 as self-reported by the MEPS-HC respondents.
Statistical Analysis
Our descriptive analysis included age-adjusted diabetes prevalence estimates; frequency distributions, including nationally weighted means (for continuous variables); and contingency tables (for categorical variables) of the independent variables across diabetic patients in 2009 and in 2011. Because the outcome variable of the total healthcare expenditure was observed to be skewed to the left, it was log transformed and included in the regression analysis. A general linear regression model tested the association of total healthcare expenditures and the HbA1c variable, while keeping all other covariates constant. All the variables were entered into the model, and nonsignificant variables were eliminated with backward selection (P <.05) to arrive at the final model.
All estimates were weighted to produce nationally representative estimates and to account for complex stratified sampling, oversampling, and nonresponse. A sensitivity analysis was carried out using linear regression, in which each variable was added one at a time to determine the impact of each parameter based on the relative magnitude of the regression coefficient. The statistical analysis was performed using SAS version 9.3 (SAS Institute Inc; Cary, NC). A sensitivity analysis was not performed, because there were too many variables to perform such an analysis. (Sensitivity analysis is usually done on a select few variables that may be subject to change, which we could not do in this case.)
Results
In 2009, the weighted frequency of diabetes mellitus was 19.8 million, with a prevalence of 8.53%. Among the overall diabetes sample in the MEPS-HC databases, 11.55% of the total patients were newly diagnosed in 2009. In 2011, the prevalence of diabetes was 9.55% (weighted frequency, 22.66 million), of which 10.44% of the patients were newly diagnosed. Overall, the weighted prevalences of newly diagnosed diabetes in 2009 and in 2011 were 0.71% and 0.79%, respectively.
The demographic distribution among the patients in 2009 and in 2011 was similar. Patients in the 45- to 64-year age-group, white race, and patients enrolled in private insurance formed the majority of the patient cohort in 2009 and in 2011 (Table 1). The mean age among the newly diagnosed patients with diabetes was approximately 55 years in both cohorts. Among the comorbidities, the prevalence of obesity was the highest in both cohorts, followed by hypertension, high cholesterol, and arthritis. The prevalence of obesity and hypertension increased in 2011, whereas high cholesterol and arthritis decreased. The overall percentage of patients with diabetes and 3 or more comorbidities decreased from 2009 to 2011 (Table 1).
Among the utilization variables, patients had an average of approximately 31 prescription refills (95% confidence interval [CI], 26-36) in 2009, which decreased slightly to approximately 30 in 2011 (95% CI, 26-34). The majority of healthcare services use in 2009 and in 2011 was through provider visits, physician visits, home health provider days, and nonphysician visits, with an increase in the utilization of these services in 2011. Broadly, the expenditures in these major healthcare service categories increased in 2011, whereas those in other categories decreased in 2011 (Table 2).
Table 3 shows the expenditures for the various categories of healthcare services among newly diagnosed diabetic patients in 2009 and in 2011. The mean total healthcare expenditure among newly diagnosed diabetic patients in 2009 was $9476 (95% CI, $5529-$13,422), which decreased to $8417 (95% CI, $6149-$10,684) in 2011. In 2009, the expenditures among the healthcare services were comprised mostly of the total inpatient expenses ($2492; 95% CI, $1010-$3975), the total outpatient and inpatient facility expenses, and prescription medications ($3232; 95% CI, $3040-$3424). In 2011, the majority of the healthcare expenditures were comprised of the total inpatient expenses, inpatient facility expenses, and prescription medications.
The linear regression model did not reveal any significant association (P = .12) between the total healthcare expenditures and the HbA1c diagnosis in 2009 and in 2011 (Table 4). The utilization of a home health agency (P = .02), emergency department care (P <.0001), dental care (P <.0001), ambulatory optometrist visits (P <.0001), outpatient provider visits (P <.0001), physician and nonphysician services (P <.0001), prescription medication refills (P <.0001), hospital stays (P <.0001), number of hospital discharges (P <.0001), total number of comorbidities (P <.0001), and private insurance (P <.0001) were drivers of total healthcare expenditures. Similarly, comorbidities such as cancer (P <.0001), high cholesterol (P <.0001), obesity (P <.0001), asthma (P <.0001), arthritis (P <.0001), and cardiac conditions, such as heart attack (P <.0001) and stroke (P = .001), were positively associated with healthcare spending.
Discussion
In 2009 and onward, the HbA1c test was recommended by the ADA as a primary diagnostic tool for diabetes in addition to its functionality as a blood glucose monitor. Our study tested the supposition that this change in diagnostic criteria could have led to an impact on total healthcare expenditures among patients with diabetes. According to our model, the HbA1c factor did not have a marked impact on healthcare expenditures. However, before making accurate inferences, it is important to consider a few aspects of our study methodology.
First, the HbA1c factor was used as a proxy and could not be accounted for directly in the model. This study is based on the assumption that many of the patients in 2009 were diagnosed with an older diagnosis method, whereas in 2011 patients were more likely to be diagnosed using the HbA1c test. Because the use of HbA1c is not mandatory and is a decision left to the discretion of the medical provider, it may not be certain that all patients in 2011 were diagnosed using the HbA1c test. However, it is likely that the change introduced with the diagnostic test recommendation might have influenced more clinicians to use this test.13 Therefore, the comparison of expenditures from data that clearly document the use of HbA1c testing may lead to more accurate conclusions.
We found that the mean total healthcare expenditures among newly diagnosed diabetic patients decreased in 2011 compared with 2009. Part of this decrease can be related to a reduction in patient spending toward the use of healthcare services in 2011. This decrease suggests promising developments in cost-containment, and thereby better disease and resource management.
Spending on major healthcare services (eg, on prescription medications, dental care, home health agency, equipment and supplies, total outpatient services, outpatient physician visits, and facility expenditures to the doctor and nondoctor) also decreased in 2011 as is seen in Figure 1.
Figure 2 shows that the utilization of services such as prescription medications did not decline greatly in 2011. This is a clear indication of cost-containment and optimal resource management. Another explanation could be that there were efforts to reduce reimbursement in 2011, which could have brought down the total expenditures in 2011.
One of the other contributing factors to this decrease includes the economic slowdown from 2007 to 2009 and its subsequent struggle to recover; decreases in some prescription drug costs, which were caused by the expansion of generic medications (eg, atorvastatin and clopidogrel); and the structural changes to the payer and healthcare sector cost-sharing dynamics.14 Likewise, 2009 saw the launch of new antiglycemic drugs, such as saxagliptin. Consequently, the dipeptidyl peptidase-4 inhibitors were used more frequently.14 Moreover, the use of insulin has been increasing over time, as more physicians prescribe it.14
Part of the expenditure containment may also be linked to the reduced prevalence of known high-cost comorbidities, such as high cholesterol, arthritis, and cardiovascular events (eg, emphysema, angina, and other attacks), among patients with newly diagnosed diabetes in 2011.13 A combination of these factors might have resulted in better disease management and might have contributed to lower healthcare costs among patients with diabetes. Similarly, because the HbA1c test is easier to use, it may help to detect diabetes earlier, which would prevent the cost of other unnecessary tests or costs related to a misdiagnosed or an unknown disorder.
Similarly, a study that assessed the impact of the Affordable Care Act on overall healthcare expenditures reported that the overall per capita healthcare expenditures increased from 2000 to 2010.15 However, despite its progressive increase, the percent growth has dramatically decreased in recent years, especially from 2008 to 2012.15 This pattern is reflected in our findings, which showed a similar pattern of percentage decrease in the mean expenditures from 2009 to 2011.
In the regression model, the comorbidities (ie, cancer, cholesterol, obesity, arthritis, asthma, and cardiac conditions, specifically stroke and heart attack) added significantly to the total healthcare expenditures. These are common comorbidities among diabetic patients, and some have been cited among the top 5 most costly conditions to treat in a report by the Agency for Healthcare Research and Quality.16 Most patients with diabetes may be concomitantly diagnosed with 3 or more comorbidities, thus requiring a polymedication regimen. Prescription of a multidrug regimen can multiply the costs, depending on the number of costly comorbidities; the type and brand of the additional prescribed medications; and the nature of patients’ health insurance coverage.16 Our findings also concur with previous reports of prescription medication refills, home healthcare, and outpatient and inpatient care contributing incrementally to the overall expenditures.17-19 The pronounced contribution of prescription medication refills in 2009 and 2011 may be substantiated by a 2008 ADA report that associated 17% of the overall retail prescription costs to patients with diabetes (attributed to diabetes and its accompanying complications).11
Private insurance payments contributed to expenditures in the final model. This finding underlines the escalation in private insurance deductibles, copays, and premiums.20 As cost-effective measures increase, several employers shift cost-sharing to their employees by way of lower insurance premiums, while keeping deductibles and copays high.21,22
Our model also found that increasing age contributed to rising expenditures. Because of the prevalence of concomitant conditions among geriatric patients, healthcare expenditures per patient (especially with chronic diseases such as diabetes) increases with age, primarily as a result of the increased use of hospital and nursing facility resources, physician visits, concomitant medications (other than insulin and oral agents), and home care.11,13 It is well-documented that the aging baby boom population and geriatric patients are adding appreciably to the total healthcare spending.15,16,23
Limitations
Some limitations need to be considered before making definite conclusions. MEPS is a cross-sectional database; therefore, no temporal causation can be established. The data used here were self-reported by patients, which could lead to systematic bias and overrepresentation or underrepresentation of the actual data or to skewed results.
Similarly, the diabetes-related costs were self-reported by patients. The HbA1c factor could not be accounted for directly in the models; it was included in the model as a proxy. Thus, it may not be completely accurate to conclude that the HbA1c tool directly impacted expenditures.
The change in expenditures could be embedded in several factors, some of which may not be accounted for in our study. One such consideration was that the landscape regarding treatment patterns (eg, insulin use) among patients with diabetes and the uptake of new treatments was also changing during this time between 2009 and 2011. These dynamics would be expected to have a significant impact on healthcare expenditures, as well as the diagnostic test. Other miscellaneous extraneous factors, such as inflation, market monopolies, fraudulent billing, and unnecessary services recommended by medical providers, could influence expenditures and could not be accounted for in the study.
Finally, as mentioned earlier, the premise of this study is based on an assumption that patients in 2009 were diagnosed with an older diagnostic factor, such as OGTT or FPG, whereas in 2011 patients were diagnosed using the HbA1c test. However, the possibility that in 2011 diabetic patients could still be diagnosed with older criteria cannot be ruled out. The comparison of expenditures clearly documenting diagnosis with and without HbA1c testing needs to be carried out to reach even more appropriate conclusions.
Conclusion
From the data available for this analysis it cannot be affirmed that the use of HbA1c testing as a diagnostic or monitoring tool impacts the total expenditures of patients with diabetes. Therefore, the current level of evidence may be insufficient to clearly establish a connection between HbA1c testing and the total healthcare expenditures. Further investigations on the total costs attributed to HbA1c, FPG, and OGTT testing, as well as a comparison of expenditures for people with and without a diabetes diagnosis using HbA1c testing, would help draw more accurate conclusions.
Author Disclosure Statement
Ms Bhounsule and Dr Peterson reported no conflicts of interest.
Ms Bhounsule is a PhD candidate, and Dr Peterson is John Wyeth Dean, Mayes College of Healthcare Business and Policy, University of the Sciences in Philadelphia, PA.
References
1. Centers for Disease Control and Prevention. National diabetes statistics report: estimates of diabetes and its burden in the United States, 2014. 2014. www.cdc.gov/diabetes/pubs/statsreport14/national-diabetes-report-web.pdf. Accessed October 11, 2014.
2. Centers for Disease Control and Prevention. Diabetes Report Card 2012. 2012. www.cdc.gov/diabetes/pubs/pdf/diabetesreportcard.pdf. Accessed June 5, 2014.
3. American Diabetes Association. Standards of medical care in diabetes—2006. Diabetes Care. 2006;29(suppl 1):S4-S42. Erratum in: Diabetes Care. 2006;29:1192.
4. American Diabetes Association. Standards of medical care in diabetes—2010. Diabetes Care. 2010;33(suppl 1):S11-S61. Erratum in: Diabetes Care. 2010;33:692.
5. National Diabetes Information Clearinghouse; National Institute of Diabetes and Digestive and Kidney Diseases. The A1C test and diabetes. NIH Publication No 14-7816. March 2014. http://diabetes.niddk.nih.gov/dm/pubs/A1CTest/. Accessed April 2, 2015.
6. Gupta MK. Glycemic biomarkers as tools for diagnosis and monitoring of diabetes. MLO Med Lab Obs. 2013;45:8-12, vi; quiz 14.
7. Hogan P, Dall T, Nikolov P; for the American Diabetes Association. Economic costs of diabetes in the U.S. in 2002. Diabetes Care. 2003;26:917-932.
8. American Diabetes Association. Economic costs of diabetes in the U.S. in 2012. Diabetes Care. 2013;36:1033-1046. Erratum in: Diabetes Care. 2013;36:1797.
9. Dairman T. HbA1c test to be used for diagnosis. Diabetes Self-Management Blog. February 20, 2009. www.diabetesselfmanagement.com/blog/hba1c-test-to-be-used-for-diagnosis/. Accessed September 21, 2014.
10. McDonald TJ, Warren R. Diagnostic confusion? Repeat HbA1c for the diagnosis of diabetes. Diabetes Care. 2014;37:e135-e136.
11. American Diabetes Association. Economic costs of diabetes in the U.S. in 2007. Diabetes Care. 2008;31:596-615. Erratum in: Diabetes Care. 2008;31:1271.
12. Agency for Healthcare Research and Quality; Medical Expenditure Panel Survey. MEPS HC-147: 2011 full year consolidated data file. September 2013. http://meps.ahrq.gov/mepsweb/data_stats/download_data/pufs/h147/h147doc.pdf. Accessed June 25, 2014.
13. Bonora E, Tuomilehto J. The pros and cons of diagnosing diabetes with A1C. Diabetes Care. 2011;34(suppl 2):S184-S190.
14. Waugh N, Cummins E, Royle P, et al. Newer agents for blood glucose control in type 2 diabetes: systematic review and economic evaluation. Health Technol Assess. 2010;14:1-248.
15. Farley R. ACA impact on per capita cost of health care. FactCheck.org: The Wire. February 14, 2014. www.factcheck.org/2014/02/aca-impact-on-per-capita-cost-of-health-care/. Accessed July 23, 2014.
16. Cohen SB. The concentration of health care expenditures and related expenses for costly medical conditions, 2012. Statistical brief #455. October 2014. Agency for Healthcare Research and Quality, Rockville, MD. http://meps.ahrq.gov/data_files/publications/st455/stat455.pdf. Accessed November 5, 2014.
17. Klonoff DC. Spending for diabetes drugs is increasing in the United States. J Diabetes Sci Technol. 2008;2:543-545.
18. CostHelper, Inc. Diabetes medication cost: how much does diabetes medication cost? http://health.costhelper.com/diabetes-medication.html. Accessed August 3, 2014.
19. Rathmann W, Haastert B, Icks A, Giani G. Trends in outpatient prescription drug costs in diabetic patients in Germany, 1994-2004. Diabetes Care. 2007;30:848-853.
20. Hartman MB, Kornfeld RJ, Catlin AC. A reconciliation of health care expenditures in the national health expenditures accounts and in gross domestic product. Surv Curr Bus. 2010;90:42-52.
21. Volsky I. The remarkable slowdown in health care costs since the passage of Obamacare. ThinkProgress. August 20, 2013. http://thinkprogress.org/health/2013/08/20/2498391/growth-in-health-care-costs-continues-to-decrease-since-passage-of-obamacare/. Accessed August 3, 2014.
22. Ku L, Broaddus M. Public and private health insurance: stacking up the costs. Health Aff (Millwood). 2008;27:w318-w327.
23. National Institute on Aging. Aging in the United States—past, present, and future. 2000. www.census.gov/population/international/files/97agewc.pdf. Accessed July 18, 2014.