Association between glycemic control and short-term healthcare costs among commercially insured diabetes patients in the United States

Abstract

Objectives:

Glycemic control, measured by HbA_1c, is well known to be a risk marker for long-term costly diabetes-related complications. The relationship between HbA_1c and short-term costs is unclear. This study investigates how HbA_1c is correlated to short-term diabetes-related medical expenses.

Methods:

Patients with diabetes with an HbA_1c reading ≥6% between April and September 2007 were identified from a large US managed-care organization. Healthcare utilization data was obtained during the subsequent 12-month period. Multivariate analyses were performed to estimate the correlation between HbA_1c and diabetes-related healthcare costs.

Results:

In all, 34,469 and 1,837 patients with type 2 and type 1 diabetes, respectively, were identified with an HbA_1c reading ≥6% (mean HbA_1c: 7.4% and 7.9%). The majority of patients with type 1 diabetes were treated with insulin, while most patients with type 2 diabetes were treated with metformin. The multivariate analysis showed that several characteristics, including HbA_1c, significantly correlate with diabetes-related medical costs for both patients with type 1 and type 2 diabetes. A 1-percentage-point increase in HbA_1c will, on average, lead to a 6.0% and 4.4% increase in diabetes-related medical costs for type 1 and type 2 diabetes, respectively. This corresponds to an annual cost increase of $445 and $250 for patients with type 1 and type 2 diabetes, respectively.

Limitations:

Retrospective data analyses inherently associated with selection bias which can only partly be adjusted by statistical techniques. Furthermore, the study population is not necessarily representative of the general population and there can be isolated coding or data errors in the dataset.

Conclusions:

These results suggest that tighter glycemic control is associated with short-term cost benefits for patients with diabetes. This supplements conventional wisdom that HbA_1c affects risk of long-term complications and long-term costs.

Key words::

• Economic benefits
• Glycemic control
• Healthcare costs
• Retrospective database analysis

Introduction

Diabetes is associated with a heavy economic burden to managed-care organizations in the US and due to the continually rising prevalence of diabetes, the cost burden for the US healthcare system is likely to continue to rise1–4. Indeed, in 2007 an estimated 17.9 million individuals had a diagnosis of diabetes in the US1. In comparison, in 2002 an estimated 12.1 million individuals were diagnosed with diabetes. Thus, the number of individuals with diabetes appears to increase by ∼1 million/year2, and by 2050 as many as 48.3 million individuals may be diagnosed with diabetes3. In 2007, the total diabetes-related expenditure in the US was estimated at $174 billion2 – an increase compared with the estimated $132 billion in 20024. Furthermore, diabetes-related complications account for much of the economic burden of diabetes on individuals and society through premature morbidity and mortality4. Indeed, on average, the costs of complications in the US over 30 years were estimated to be more than $47,000 per patient with type 2 diabetes5. Additionally, annual diabetes-related medical cost increase with diabetes duration, most likely due to the risk of complications that develop over time6. In line with the above, delaying of preventing long-term complications through glycemic control would be cost effective.

Deterioration of glycemic control is one of the hallmark features of the progression of diabetes. The American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists (AACE) recommend specific target levels of HbA_1c (ADA: <7.0%; AACE: ≤6.5%)7,8 and achieving these HbA_1c targets is a key objective in the management of type 2 diabetes. Unfortunately, the National Health and Nutrition Examination Survey (NHANES) demonstrated that 43% of US adults with type 2 diabetes still do not achieve HbA_1c targets (<7.0%)9. However, improving glycemic control is an important part of diabetes, as elevated HbA_1c is linked to the development of long-term complications, such as cardiovascular disease, retinopathy and lower-limb amputation. Of note, even though the ADA recommends an HbA_1c target of <7.0%, it is increasingly becoming recognized that less stringent goals may be appropriate in certain patients, such as those with severe hypoglycemia, limited life expectancy or co-morbidities10. For example, the Action to Control CardiOvascular Risk in Diabetes (ACCORD) trial, which compared intensive glucose control with standard control, identified a possible harm of intensive glucose-lowering in high-risk patients with type 2 diabetes – near normal glucose control was associated with increased all-cause mortality and CV mortality11. On the other hand, the Action in Diabetes and Vascular Disease (ADVANCE) trial showed no increased mortality in intensively controlled high-risk patients achieving HbA_1c values of 6.5%12. The results of these trials may have some impact on future HbA_1c targets.

Improving glycemic control reduces the rate of micro- and macrovascular complications, as shown by landmark trials such as the Diabetes Complications and Control Trial (DCCT) in type 1 diabetes patients and the UK Prospective Diabetes Study (UKPDS) in type 2 diabetes patients13,14. UKPDS, for example, showed that reducing HbA_1c by 1% in type 2 diabetes significantly decreases the risk of microvascular complications (37%, p < 0.0001), diabetes-related death (21%, p < 0.0001), and myocardial infarction (14%, p < 0.0001)15.

Several studies have suggested that improving glycemic control will be beneficial for the longer-term economic burden of diabetes16–18. DCCT, for example, showed that intensive therapy led to improvements in glycemic control, which helped to delay the onset and the progression of diabetes-related complications. This in turn led to increased quality of life and life-expectancy. The associated incremental cost per year gained was estimated to be $28,661, which is well within the range of long-term cost effectiveness16. However, the short-term relationship between glycemic control and cost remains unclear. This study, therefore, investigates how HbA_1c is correlated to diabetes-related medical expenses over 12 months in a managed-care plan.

Patients and methods

This was a retrospective analysis of medical and pharmacy administrative claims data and laboratory data for patients with type 1 and type 2 diabetes who were enrolled in a large US managed-care organization (data collected and made anonymous by i3 Innovus, a division of Ingenix Pharmaceutical Services, Inc; ∼15 million covered lives/year). Patient information was made anonymous to protect patient privacy in accordance with the Health Insurance Portability and Accountability Act.

Patient population

Eligible patients for this retrospective study were adults (age ≥18 years) diagnosed with type 1 or type 2 diabetes that were continuously enrolled in a health plan for at least 1 year. Included patients were required to have had at least one HbA_1c reading between April and September 2007 inclusive. Furthermore, patients’ average HbA_1c value had to be ≥6% to reflect that this analysis is relevant for patients where glycemic management is an issue. Medical diagnosis of diabetes was coded using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 250.xx. Patients were classified as having type 1 diabetes if all or the majority of medical claims were coded as ICD-9-CM 250.x1 or 250.x3. Patients were classified as having type 2 diabetes if they had at least one prescription refill for an oral antidiabetic drug (OAD); or if all or most of the medical claims were coded as ICD-9-CM 250.x0 or 250.x2.

Patient characteristics

Patient characteristics were identified from administrative enrollment and claims data. The patient characteristics analyzed included gender, age, and type of healthcare plan. Pharmacy claims were also examined to identify diabetes therapy and medical claims were used to identify baseline co-morbidities.

Healthcare resource utilization

Data on overall and diabetes-related healthcare resource utilization were collected during a 12-month period from October 2007 to September 2008. The number of overall and diabetes-related medical visits and prescription refills and the associated healthcare costs were identified. Overall medical and diabetes-related costs were defined as the sum of office, inpatient, outpatient, ER and other costs, such as costs associated with laboratory tests, rehabilitation facilities and nursing homes. Diabetes-related costs included prescription refills and medical claims coded with ICD-9-CM 250.xx.

Statistical methods

Statistical analyses were carried out using SAS version 9.1.3 (SAS Institute Inc., Cary, NC, USA). Descriptive analyses were carried out to identify patient characteristics. Multivariate analyses were used to estimate the correlation between patient characteristics and diabetes-related healthcare costs using the following baseline co-morbidities as fixed effects: high blood cholesterol, hypertension, cardiovascular disease, kidney disease, non-alcoholic fatty liver disease and obesity. The following variables were included in the multivariate analysis: healthcare costs, glycemic control, age, gender, healthcare plan type and diabetes treatment (for coding of these variables, please refer to Table 3 in the results section). These variables do not follow a normal distribution, as established by normality tests. Therefore, the multivariate analyses were performed using a generalized linear model (GLM) regression technique with a gamma variance and log-link, a model which was proven to be suitable for this analysis (scaled deviance over degree of freedom: 1.2518 for the type 2 diabetes patients only model and 1.2049 for the type 1 diabetes patients only model). Positive GLM estimates indicate a higher cost, while negative GLM estimates indicate a lower cost. GLM estimates were used to calculate the percentage increase in costs, using the following formula: 100 × [exp(GLM estimate) − 1]. The percentage decrease in costs can be calculated as follows: 100 × [exp(−GLM estimate) − 1].

Results

Patient population

A total of 529,771 patients with diabetes were identified who were continuously enrolled in a healthcare plan for at least 1 year. Of these patients, 45,779 had at least one HbA_1c reading between April and September 2007. The majority (n = 36,306; 80%) had an average HbA_1c value ≥6%. Among these patients, 1,837 patients were diagnosed with type 1 diabetes, while 34,469 patients were diagnosed with type 2 diabetes.

Patient characteristics

Patient characteristics are summarized in Table 1. The mean age was 41 years and 56 years for patients with type 1 and type 2 diabetes, respectively, and the majority of patients were male (51% and 55%, respectively). Mean (SD) HbA_1c values, among the included patients with a HbA_1c value above 6%, were 7.9 (1.5)% and 7.4 (1.6)% for patients with type 1 and type 2 diabetes, respectively. The study population is geographically diverse, with the majority of patients situated in the South (71.5%), West (11.4%) and Midwest (10.3%). However, the study population does not accurately reflect the geographical distribution of the country as it is overrepresented in the South. Most patients were enrolled in a point-of-service healthcare plan (POS): 65% and 60% of patients with type 1 and type 2 diabetes, respectively. The most commonly used diabetes treatment for patients with type 1 diabetes was, not surprisingly, insulin. Among patients with type 2 diabetes, metformin was the most commonly used oral antidiabetic drug (OAD) (51%), followed by sulfonylureas (29%) and thiazolidinediones (20%). At baseline, 38% and 74% of patients with type 1 and type 2 diabetes, respectively, suffered from hypertension, while 53% and 80%, respectively, had high blood cholesterol. Also, 1% and 4–5% of patients with type 1 and type 2 diabetes suffered from cardiovascular and kidney disease, respectively.

Table 1.  Patient characteristics.

Characteristics	Type 1 diabetes (n = 1,837)	Type 2 diabetes (n = 34,469)
Mean age (years ± SD)	41 ± 16	56 ± 9
Female/male (%)	49/51	45/55
Healthcare plan type, n (%)
EPO	330 (18)	6,453 (19)
HMO	132 (7)	3,928 (11)
POS	1,185 (65)	20,527 (60)
PPO	184 (10)	3,441 (10)
Other	6 (0)	120 (0)
Diabetes treatment, n (%):
DPP-4	NA	3,308 (10)
Metformin	NA	17,449 (51)
OAD combination	NA	6,354 (18)
Sulfonylurea	NA	10,086 (29)
Thiazolidinedione	NA	6,985 (20)
Exenatide	NA	2,601 (8)
Other OAD	NA	1,084 (3)
Basal insulin	879 (48)	5,273 (15)
Bolus insulin	1,567 (85)	3,010 (9)
Premix insulin	88 (5)	1,516 (4)
Co-morbidities, n (%):
Hypertension	702 (38)	25,551 (74)
High blood cholesterol	969 (53)	27,646 (80)
Obesity	63 (3)	4,550 (13)
Non-alcoholic fatty liver disease	11 (1)	915 (3)
Kidney disease	91 (5)	1,491 (4)
Cardiovascular disease	13 (1)	371 (1)
HbA1c, (%)
>6–≤7	583 (31.7)	19,249 (55.8)
>7–≤7.5	316 (17.2)	4,389 (12.7)
>7.5–≤8	285 (15.5)	2,802 (8.1)
>−≤8.5	207 (11.3)	1,835 (5.3)
>8.5–≤9	133 (7.2)	1,444 (4.2)
>9	313 (17.0)	4,750 (13.8)

EPO. exclusive provider organization; HMO, health maintenance organization; POS, point-of-service plan; PPO, preferred provider organization; NA, not applicable; DPP-4, dipeptidyl peptidase–4 inhibitor; OAD, oral antidiabetic drug; OAD combination, combination products of more than one OAD.

Co-morbidities are identified by ICD–9 diagnosis codes.

Healthcare resource utilization

Overall and diabetes-related healthcare resource utilization for type 1 and type 2 diabetes is summarized in Table 2. The annual medical costs were $14,912 and $17,492 for patients with type 1 and type 2 diabetes, respectively. The majority of the medical utilization were accounted for by office visits (11.2 and 11.2 visits for type 1 and type 2 diabetes, respectively), and these corresponded to an annual cost of $3,195 and $3,716. On the other hand, the number of inpatient visits was 1.3 and 2.0 for patients with type 1 and type 2 diabetes, and these visits were associated with an annual cost of $3,838 and $6,486, respectively. A major part of the overall medical expenditures can be attributed to other costs, such as laboratory tests and rehabilitation ($4,139 and $2,380 for type 1 and type 2 diabetes, respectively). Total diabetes-related medical costs were $7,421 and $5,680 related to type 1 and type 2 diabetes, respectively. Similarly to overall medical utilization, most of the diabetes-related medical visits for type 1 and type 2 diabetes were accounted for by office visits (4.3 and 3.8 visits, respectively), and the associated costs were $877 and $658, respectively. Diabetes-related inpatient visits for type 1 and type 2 diabetes (0.5 visits for both) were associated with an annual cost of $2,305 and $2,723, respectively.

Table 2.  Annual overall and diabetes-related healthcare resource utilization for type 1 and type 2 diabetes.

Healthcare resource	Type 1 diabetes		Type 2 diabetes
	Number*	Costs ($)	Number*	Costs ($)
Overall healthcare resources
Medical visits	21.6 (22.4)	14,912 (30,807)	21.4 (25.3)	17,492 (48,567)
Office visits	11.2 (11.1)	3,195 (7,084)	11.2 (11.3)	3,716 (12,495)
Outpatient visits	2.5 (5.0)	3,631 (12,961)	2.8 (5.7)	4,816 (17,260)
ER visits	0.4 (2.2)	107 (900)	0.5 (2.6)	95 (795)
Inpatient visits	1.3 (6.9)	3,838 (19,361)	2.0 (11.3)	6,486 (35,476)
Other†	6.2 (9.5)	4,139 (6,735)	4.8 (6.8)	2,380 (6,257)
Prescription refills	33.5 (25.7)	5,010 (4,512)	40.6 (30.4)	4,352 (4,690)
Diabetes-related healthcare resources
Medical visits	9.8 (9.5)	7,421 (16,227)	6.9 (7.0)	5,680 (18,627)
Office visits	4.3 (3.3)	877 (1,423)	3.8 (3.3)	658 (1,143)
Outpatient visits	0.7 (1.8)	1,399 (5,386)	0.5 (1.9)	1,536 (7,509)
ER visits	0.1 (0.5)	24 (377)	0.1 (0.5)	17 (452)
Inpatient visits	0.5 (2.5)	2,305 (13,296)	0.5 (3.1)	2,723 (15,385)
Other†	4.2 (6.7)	2,816 (3,912)	2.1 (2.3)	746 (1,692)
Prescription refills	7.7 (5.6)	1,811 (1,240)	9.8 (8.9)	1,342 (1,583)

Data are mean (SD).

*Number of visits or refills.

†Laboratory test, rehabilitation facilities, and nursing homes.

The number of overall prescription refills for patients with type 1 diabetes and type 2 diabetes was 33.5 and 40.6 refills, respectively. These prescription refills were associated with a total annual cost of $5,010 and $4,352 for type 1 and type 2 diabetes, respectively. Of this total number of prescription refills, 7.7 and 9.8 refills were related to type 1 and type 2 diabetes, respectively, and the associated annual costs were $1,811 and $1,342, respectively. The low number of diabetes-related prescription refills is most likely due to the fact that no refills were recorded during the study period for 8.5% and 17.3% of patients with type 1 and type 2 diabetes, respectively.

Correlation between patient characteristics and diabetes-related healthcare costs

The multivariate analysis showed that several patient characteristics, such as age and HbA_1c, significantly correlate with diabetes-related healthcare costs for both type 1 and type 2 diabetes (Table 3). It was calculated that a 1-percentage-point increase in HbA_1c is, on average, associated with a significant increase in diabetes-related costs of 6.0% (100 × [exp(0.0583) − 1]) and 4.4% (100 × [exp(0.0431) − 1]) for type 1 (p = 0.0061) and type 2 (p < 0.0001) diabetes, respectively. This corresponds to an increase in annual diabetes-related medical costs of $445 and $250 for type 1 and type 2 diabetes, respectively (calculated using healthcare resource utilization data shown in Table 2). Additionally, a 1-percentage-point increase in HbA_1c is associated with an increase in annual diabetes-related pharmacy costs of $109 and $59 for type 1 and type 2 diabetes, respectively. Similarly, a 1-percentage-point decrease in HbA_1c is, on average, associated with a decrease in diabetes-related costs of 5.7% (100 × [exp(−0.0583) − 1]) for type 1 diabetes and 4.2% (100 × [exp(−0.0431) − 1]) for type 2 diabetes. This correlates to a reduction in annual diabetes-related medical costs of $423 and $239 for type 1 and type 2 diabetes, respectively (calculated using healthcare resource utilization data shown in Table 2). Furthermore, a significant correlation was observed between age and diabetes-related costs for type 2 diabetes (p < 0.0001) (Table 3). Certain healthcare plan types and diabetes treatments were also shown to be significantly correlated with diabetes-related costs (p < 0.0001 to p < 0.05) (Table 3).

Table 3.  Multivariate analysis of diabetes-related medical costs.

Characteristics	Type 1 diabetes		Type 2 diabetes
	Estimate	p-value	Estimate	p-value
Intercept	7.4559 (0.2485)	<0.0001	6.2698 (0.0758)	<0.0001
Age	−0.0016 (0.0024)	NS	0.0084 (0.0009)	<0.0001
HbA1c	0.0583 (0.0213)	0.0061	0.0431 (0.0056	<0.0001
Gender, female	0.1168 (0.0608)	NS	0.1469 (0.0169)	<0.0001
Healthcare plan type*
EPO	0.1968 (0.1194)	NS	0.1297 (0.0322)	<0.0001
HMO	0.0421 (0.1475)	NS	0.1667 (0.0355)	<0.0001
POS	−0.0340 (0.1026)	NS	0.0673 (0.0281)	0.0168
Other	1.3261 (0.5387)	0.0138	−0.6120 (0.1418)	<0.0001
Diabetes treatment
DPP-4	NA	NA	0.0153 (0.0282)	NS
Metformin	NA	NA	0.2236 (0.0175)	<0.0001
OAD combination	NA	NA	0.1443 (0.0223)	<0.0001
Sulfonylurea	NA	NA	0.2227 (0.0188)	<0.0001
Thiazolidinedione	NA	NA	0.0516 (0.0207)	0.0127
Exenatide	NA	NA	0.0451 (0.0314)	NS
Other OAD	NA	NA	−0.0153 (0.0473)	NS
Basal insulin	−0.3588 (0.0611)	<0.0001	0.2218 (0.0266)	<0.0001
Premix insulin	0.9742 (0.1581)	<0.0001	0.5834 (0.0410)	<0.0001
Bolus insulin	0.8031 (0.0958)	<0.0001	0.9404 (0.0339)	<0.0001
Co-morbidities
Hypertension	0.4136 (0.0756)	<0.0001	0.7054 (0.0200)	<0.0001
High blood cholesterol	0.1416 (0.1416)	NS	0.1861 (0.0218)	<0.0001
Obesity	0.0462 (0.1679)	NS	0.6748 (0.0245)	<0.0001
Non-alcoholic fatty liver disease	1.3699 (0.3926)	0.0005	0.5934 (0.0511)	<0.0001
Kidney disease	0.9288 (0.1442)	<0.0001	0.8995 (0.0414)	<0.0001
Cardiovascular disease	0.8544 (0.3581)	0.017	0.9052 (0.0795)	<0.0001

Dummy variables are shown in italic.

Data are estimate (SD).

Variables were coded as follows: age – continuous; gender – male and female; healthcare plan type – EPO, HMO, POS, PPO and Other; diabetes treatment use – 1 (yes) and 0 (no); co-morbidities – 1 (yes) and 0 (no).

*PPO healthcare plan was used as benchmark. EPO, exclusive provider organization; HMO, health maintenance organization; POS, point-of-service plan; PPO, preferred provider organization; NS, non-significant; NA, not applicable; OAD, oral antidiabetic drug.

Discussion

The results presented above showed that for patients with type 1 diabetes, approximately half of overall medical expenditures are diabetes-related, while for patients with type 2 diabetes only ∼30% of total medical cost were associated with diabetes treatment. Inpatient visits are a major contributor to diabetes-related medical costs. Indeed, inpatient visits were associated with an annual cost per patient of $2,305 and $2,723 for type 1 and type 2 diabetes, respectively. This analysis has also shown that poorer glycemic control, as measured by HbA_1c, is significantly correlated with higher diabetes-related medical and pharmacy costs in the short-term (p < 0.0001 to p < 0.01).

It is fairly well established that improving glycemic control is cost effective over the long term16–18. Beyond the clinical incentive of improving HbA_1c to avoid long-term complications and costs, there may also be shorter-term economic advantages to adequate glycemic control in patients with diabetes. Some studies have suggested that better glycemic control may lead to cost savings in the short term. For example, an analysis of data from a large health maintenance organization estimated that every 1% reduction in HbA_1c is associated with healthcare cost savings of $400 to $4,000 per patient over the following 3 years19. Furthermore, sustained reductions in HbA_1c have been associated with cost savings within 1–2 years of improvement20, and also maintaining HbA_1c levels below 7.0% over a 1-year period was associated with lower diabetes-related costs than HbA_1c levels continuously above 7.0%21. In line with previous studies, the present analysis has also demonstrated a significant correlation between HbA_1c and diabetes-related costs. However, the current analysis has established this correlation over a 12-month period, a shorter timeframe than has previously been described in many studies. These results therefore suggest that improving glycemic control may have short-term economic benefits. This may have implications for future cost-effectiveness analyses of new therapies with increased potential for short-term glycemic control. Of note, the purpose of this study was to investigate the correlation between diabetes-related costs and HbA_1c for patients that are in need of improving or monitoring their glycemic control, and was therefore designed to only include patients with type 1 or type 2 diabetes with an average HbA_1c ≥ 6%. Besides any short- or long-term cost savings, the benefits to patients’ long-term health outcomes are well established. For example, a US-specific model analysis of type 2 diabetes estimated that a reduction in HbA_1c from 9.5% to 8% improved mean life expectancy by 1.11 years and discounted (3% per year) quality-adjusted life expectancy by 0.58 QALYs (quality-adjusted life-years). When HbA_1c was reduced from 8% to 7% the results showed increase in mean life expectancy of 0.72 years and increase in QALYs of 0.5822.

The current study results must be interpreted within the limitations of the data and the study design. Firstly, the observational nature of the study does not allow causal inference of the results. An inherent limitation of retrospective database research is selection bias, which can only be partly adjusted for by using statistical techniques like regression analyses as is done in the present study. Furthermore, the study population was a sample of managed-care members, and it therefore may not be possible to generalize the results to a non-managed-care population. Claims data may also be subject to coding errors, and hence the number of type 1 and type 2 diabetes patients may not be entirely accurate. However, claims data are an important source of data, as they provide data on health care utilization and associated expenditures in a real-world setting. Finally, there are limitations regarding the specifications of the regression model. Costs data distributions are generally right-skewed and are not easily fit when conducting multivariate analyses. However, the statistical analysis undertaken in this study is the best possible approach.

Conclusion

In this managed-care population of type 1 and type 2 diabetes patients, a significant correlation was observed between HbA_1c and short-term diabetes-related costs. It is well known that improving glycemic control is associated with a decrease in the risk of long-term complications and associated long-term costs. The results of this analysis suggest that improved glycemic control may also confer short-term cost benefits for patients with diabetes. This needs to be taken into account in future cost-effectiveness analyses of new therapies for type 2 diabetes.

Transparency

Declaration of funding

The preparation of this article was supported by Novo Nordisk Inc., Princeton, NJ, USA.

Financial support

The preparation of this article was supported by Novo Nordisk Inc., Princeton, USA

Declaration of financial/other relationships

Both authors are employed by Novo Nordisk Inc.

Acknowledgment

The assistance of Dr Elien Moës, Watermeadow Medical plc, UK, in preparing this article is gratefully acknowledged.