Association between provider specialty and healthcare costs and glycemic control for patients with diabetes

Abstract

Aims: To analyze the association between provider, healthcare costs, and glycemic control for patients with diabetes mellitus (DM).

Materials and methods: This cross-sectional study identified adults with type 1 or 2 DM (T1D, T2D) in the Optum database. The main independent variable was provider (endocrinologist or primary care). Regression analysis compared total medical and pharmacy costs, adjusting for health status and other patient differences, by provider.

Results: For all patients, HbA1C improvement was greater, and medical costs significantly lower with an endocrinologist rather than a primary care provider. The largest HbA1C improvement (4%) occurred for insulin-dependent patients seen by endocrinologists. Significant medical savings with endocrinologist management occurred within the Medicare Advantage population in every sub-group of patients, with 14% lower costs ($4,767) for patients with T1D, 11% lower costs ($3,160) for patients with macro- and microvascular complications, and 10% lower costs ($2,237) for insulin-dependent patients. Within the commercial insurance population, medical costs were reduced by ≥9% in every sub-group of patients, with a 20% reduction ($8,450) for patients with micro- and macrovascular complications. Overall total costs (medical and pharmacy) were 8% ($1,541) higher for patients receiving endocrinologist rather than primary care, although endocrinologist care resulted in a 9% reduction (–$3,710) in costs for Medicare Advantage patients with T1D. Total medical costs (excluding pharmacy costs) may be a more accurate indicator of costs associated with patients in various stages of DM.

Limitations: There was insufficient data to develop risk-adjustment payments for pharmacy costs based on disease severity. The cross-sectional design identifies associations and not cause–effect relationships.

Conclusion: DM management by an endocrinologist was associated with greater HbA1C improvement and significantly lower medical costs. Total costs were higher with an endocrinologist, but for patients with T1D lower costs were seen, ranging from 2–9% regardless of insurance type.

Keywords:

• Diabetes mellitus
• medical costs
• endocrinologist
• glycemic control
• healthcare utilization

Introduction

Healthcare in the US is rapidly transitioning from a volume-based to a value-based model. Value-based diabetes care seeks to improve the quality of care (clinical outcomes), as well as its cost efficiency (economic outcomes). Value-based care considers both clinical and economic outcomes, unlike volume-based care, which preferentially rewards clinical encounters and procedures. The shift to value-based care is especially germane for chronic diseases such as diabetes mellitus (DM), which have long-term clinical and economic consequences for patients and society. An estimated 23.1 million Americans, 7.2% of the population, had diagnosed DM in 2015¹. In 2012, the total direct and indirect estimated cost of treating diagnosed diabetes in the US was $245 billion². These numbers underscore the importance of determining how to improve clinical outcomes in a cost-efficient manner.

Determining the “value” of healthcare is challenging, given the varying interests of multiple constituencies, such as patients, providers, payers, employers and society, and the complexity of the healthcare delivery system. Nuckols et al.³ proposed a conceptual framework that assesses the complex interactions of external factors, structure, process, outcomes, and cost. For example, there may be increased costs associated with optimal disease management, but these costs may be offset by savings associated with fewer complications. Provider characteristics and practice structure are aspects of that framework which may affect both the quality and cost of care.

Our retrospective cross-sectional study was designed to assess the association between provider specialty and healthcare costs and glycemic control for patients with DM by analyzing a Medicare Advantage and commercial insurance database of patients with type 1 or type 2 diabetes (T1D or T2D).

Materials and methods

Database

We used a sub-set of the Optum Clinformatics^® Data Mart database (OptumInsight, Eden Prairie, MN) containing more than 6 million commercial insurance and Medicare Advantage members ≥18 years old with T1D or T2D in the US. Patient sub-sets included those with uncontrolled DM, disease-associated complications, and insulin dependence. The database contains de-identified, person-specific health data including medical, inpatient, outpatient, and pharmacy claims. We searched records dating from January 2013 through December 2014, spanning coverage years (CY) 2013 and 2014 and including 2 million patients with DM. Institutional review board oversight was not necessary because the database was de-identified in compliance with HIPAA rules.

Subject selection

Individuals included in this study were identified as having T1D or T2D anytime during CY2013 (the baseline year) and CY2014 (the analyzed year), and being ≥ 18 years old as of CY2013. Having DM was defined as having at least one emergency department visit with a diagnosis of diabetes (ICD-9-CM codes 250.xx), OR at least one inpatient admission with a diagnosis of DM, OR at least two outpatient visits with a diagnosis of DM at least 30 days apart. During CY2013 and CY2014, if an individual had claims indicating gestational diabetes (ICD-9-CM codes 648.0x, 648.8x), non-clinical diabetes (ICD-9-CM code 790.29), or secondary diabetes (ICD-9-CM codes 249.x), then that individual was excluded. Subjects had to be enrolled continuously during CY2013 and CY2014 with medical and pharmacy benefits, and in the same insurance/health plan during this time frame, defined as the same insurer (commercial or Medicare Advantage) and same indicator of consumer-driven health plan (no CDH, Health Savings Account, or Health Reimbursement Account). Selected subjects also had pharmacy claims for either insulin or oral anti-diabetic medications during CY2013 and CY2014.

Provider designation and definition of visits

Endocrinologists and primary care physicians were identified by provider category codes associated with their practices. Visits were counted once per date and defined as face-to-face appointments in an outpatient setting. Visits in an inpatient setting or ambulatory surgical center were excluded because we were interested in regular visits to monitor therapy, rather than interventions to stabilize acute episodes, as would occur with inpatient care. Furthermore, primary care physicians and endocrinologists who see patients on an outpatient basis typically do not provide inpatient care. Patients were categorized into two mutually exclusive groups based on provider seen as: (1) those having at least two visits to an endocrinologist during CY2014, and (2) those not seeing an endocrinologists but seeing a primary care physician twice during the same time period.

Key measures

Costs were pulled for DM (T1D or T2D) and insulin dependence, as well as for outcomes (any complication, uncontrolled DM [ICD-9-CM 250.xx and fifth digit of 2 or 3], and micro- and macrovascular complications). “Insulin dependent” was defined as using any insulin therapy (basal insulin, intermediate-acting insulin, or insulin plus an oral antidiabetic medication). “Uncontrolled DM” was identified as ICD-9-CM 250.xx and a fifth digit of 2 or 3. The ICD-9-CM codes associated with DM-related complications came from Menzin et al.⁴. Microvascular complications included retinopathy, nephropathy, and neuropathy (sensory, history of foot lesions; autonomic, including sexual dysfunction and gastroparesis). Macrovascular complications included coronary heart disease, cerebrovascular disease, and peripheral arterial disease.

HbA1C results, when available, were retrieved for patients with DM who met selection criteria. These results were categorized using the same characteristics (T1D, T2D, insulin dependent, any complication, uncontrolled DM, and micro- and macrovascular complications) and provider types (endocrinologist or primary care) as in the cost analyses.

Costs

Medical costs included care delivered in all locations besides pharmacies. Total costs included both medical and pharmacy costs. Costs were based on “standard price”, defined for the database as “an estimate of the allowed amount (the insurance paid amount plus the amount the patient is responsible for)”. This method treats the total dollar amount as if it came from a single payer with a consistent policy rather than from multiple sources. Differences in price amount to differences in utilization. $$\text{Total costs }=\text{ Total medical costs }+\text{ Pharmacy costs}$$

Analyses

Ordinary least squares regression was used to make an adjusted comparison between the type of provider, and medical and total medical and pharmacy costs of DM care categorized by various sub-groups of patients and type of insurance. Medical costs included inpatient and outpatient costs. Total costs included medical and pharmacy costs. Factors in the model included demographics, type of insurance, type of DM, CY2014 common complications of DM, and the CY2013 hierarchical condition categories (HCCs). HCCs are used in risk-adjusted payment models by the Centers for Medicaid and Medicare Services (CMS; www.cms.gov/Research-Statistics-Data-and-Systems/.). Analyses were conducted for sub-groups of T1D, T2D, insulin dependent, any complication, uncontrolled DM, and those with both microvascular and macrovascular complications. An additional set of regression analyses analyzed the interaction between type of insurance and provider type, plus all other factors mentioned previously, to see if the effect of provider type was statistically different (p ≤ .05) by insurance type. Given our large sample size and our interest in associations only, normality of the coefficient estimates should be met⁵. However, as a sensitivity analysis, we also performed a generalized linear model to address concerns with using linear regression, by conducting a modified park test⁶ to determine the family with a log link. Based on this test, a Poisson regression was conducted to see how results compared to linear regression.

Results

Figure 1 presents study group selection criteria and numbers of patients in each group. Ninety per cent of study patients with DM were managed in the primary care setting. Endocrinologists saw their patients more frequently than primary care physicians did (3.9 vs 2.11 annual visits, respectively), and endocrinologists saw a larger percentage of patients with T1D (19% vs 4%, respectively). Patients seen in primary care, compared to those seen by endocrinologists, were older (68.5 years vs 61 years, respectively), more likely to be covered by Medicare (71% vs 44%, respectively), and less likely to use blood glucose monitoring (58% vs 82%, respectively), have uncontrolled DM (40% vs 77%, respectively), or have DM-associated complications (46% vs 58%, respectively). The Charlson Comorbidity Index was 3.92 for primary care and 3.88 for specialty care patients. In the primary care setting, <30% of patients with DM were on any insulin therapy, whereas 67% of patients treated by endocrinologists were using advanced insulin therapies.

Figure 1. Study group characteristics.

Table 1 presents the economic analyses for all study patients and also separately for those sub-groups covered by commercial or Medicare Advantage insurance. Medical costs for all patients were significantly lower (p < .05) for every patient sub-group treated by endocrinologists compared with primary care physicians. The greatest medical cost savings came from specialist management of micro- and macrovascular complications ($4,273, –14%, p < .05) among all patients and in the commercial insurance population ($8,450, –20%, p < .05), and treatment of T1D ($4,767, –14%, p < .05) in the Medicare Advantage population. The greatest total medical and pharmacy cost savings came from specialist management of T1D ($2,345, –7%) among all study patients and Medicare Advantage patients ($3,710, –9%), and of micro- and macrovascular complications in the commercial insurance population ($5,088, –10%). Our Poisson regression with log link also showed similar results in terms of the direction of effect and statistical significance. The largest HbA1C improvement (4%) occurred for insulin-dependent patients seen by endocrinologists (Table 2).

Table 1. Comparative medical and total (medical and pharmacy) costs categorized by patient characteristics and healthcare providers.

	Endocrinologists’ patients (n) (Endo)	Primary care patients (n) (PCP)	Regression adjusted medical costs* Endo vs PCP	Regression adjusted total costs* Endo vs PCP
All patients	34,070	310,227	–$1,006** (–7%)	$1,541** (8%)
Type 1	6,469	12,622	–$3,444** (–13%)	–$2,345** (–7%)
Type 2	27,601	297,605	–$361** (–3%)	$2,406** (14%)
Insulin dependent	22,973	90,239	–$2,455** (–12%)	–$555 (–2%)
Any complication	19,774	142,818	–$1,794** (–10%)	$712** (3%)
Uncontrolled DM	26,309	124,577	–$1,311** (–8%)	$1,286** (6%)
Micro- and macrovascular complications	8,096	69,950	–$4,273** (–14%)	–$1,853** (–5%)
Medicare advantage sub-group	14,860	221,063	–$893** (–6%)	$1,227** (6%)
Type 1	2,028	8,894	–$4,767** (–14%)	–$3,710** (–9%)
Type 2	12,832	212,169	–$178** (–1%)	$2,077** (12%)
Insulin dependent	9,833	65,574	–$2,237** (–10%)	–$613 (–2%)
Any complication	10,384	117,574	–$1,247** (–7%)	$983** (4%)
Uncontrolled DM	11,664	89,833	–$1,137** (–6%)	$1,057** (4%)
Micro- and macrovascular complications	5,587	62,143	–$3,160** (–11%)	–$1,091 (–3%)
Commercial insurance sub-group	19,210	89,164	–$1,214** (–10%)	$1,634** (10%)
Type 1	4,441	3,728	–$1,573 (–9%)	–$425 (–2%)
Type 2	14,769	85,436	–$986** (–9%)	$2,131** (13%)
Insulin dependent	13,140	24,665	–$2,454** (–15%)	–$228 (–1%)
Any complication	9,390	25,244	–$2,638** (–14%)	$279 (1%)
Uncontrolled DM	14,645	34,744	–$1,700** (–12%)	$1,265** (6%)
Micro- and macrovascular complications	2,509	7,807	–$8,450** (–20%)	–$5,088** (–10%)

*Medical costs include inpatient and outpatient costs. Total costs include medical costs and pharmacy costs in US dollars.

** p < .05.

Abbreviation. DM, diabetes mellitus.

Table 2. Change in HbA1C categorized by patient characteristics and healthcare provider.

	Patients (n)		HbA1C value
	Endo	PCP	Endo	PCP	Endo vs PCP
All patients	13,650	120,722	7.39	7.45	–0.06* (–0.8%)
Type 1	2,143	4,095	7.80	8.04	–0.24* (–3%)
Type 2	11,507	116,627	7.39	7.42	–0.03 (–0.4%)
Insulin dependent	8,687	32,633	8.00	8.35	–0.35* (–4%)
Any complication	8,022	60,201	7.50	7.53	–0.03 (–0.4%)
Uncontrolled DM	10,960	53,816	7.81	7.86	–0.06* (–0.7%)
Micro- and macrovascular complications	3,233	28,241	7.50	7.47	0.03 (0.4%)

*Statistically significant (p < .05) but not clinically significant. The American Diabetes Association recommends an HbA1C goal of <7.5% in healthy, non-pregnant adults with few co-existing chronic illnesses, and intact cognitive and functional status⁷.

Abbreviation. DM, diabetes mellitus.

Discussion

Given the individual and societal burden of diabetes, it is important to find ways of delivering care in a cost-efficient manner while achieving better quality of care. One such way is through steering patients to certain providers or care teams with expertise in managing chronic conditions. Most patients with DM are managed in primary care settings, yet this study suggests that sub-groups of these patients would realize greater clinical and economic benefits under specialist care. For all patients, HbA1C improvement was greater, and medical costs significantly lower, with DM management by an endocrinologist. This improvement was statistically significant for patients with T1D, insulin dependence, and uncontrolled DM. The improvement was not clinically significant given the cross-sectional study design, which included newly diagnosed patients as well as those on long-term maintenance therapy. Change in HbA1C is greatest when therapy first begins. Standards of Medical Care in Diabetes–2018 recommends an HbA1C goal of <7.5% in healthy, non-pregnant adults with few co-existing chronic illnesses, and intact cognitive and functional status⁷. HbA1C targets are individualized, based on numerous factors (age, life expectancy, comorbidities, duration of DM, risk of hypoglycemia, patient motivation, and adherence), and may change over time⁸. In this study, considering all patients (Table 2), patients with the most severe disease (T1D, insulin-dependent, and uncontrolled) were farther from the target HbA1C.

Significant medical savings with endocrinologist management occurred within the Medicare Advantage population in every sub-group of patients, with a 14% reduction ($4,767) for patients with T1D, an 11% reduction ($3,160) for patients with macro- and microvascular complications, and a 10% reduction ($2,237) for insulin-dependent patients. Within the commercial insurance population, medical costs were reduced by ≥9% in every sub-group of patients, with a 20% reduction ($8,450) for patients with micro- and macrovascular complications. Compared to primary care physicians, endocrinologists achieved savings even though they saw their patients more often and managed a larger percentage of patients with any complication (46% vs 58%, respectively) and uncontrolled DM (40% vs 77%, respectively).

Overall total costs, including medical and pharmacy claims, were 8% ($1,541) higher for patients receiving specialist rather than primary care. In the Medicare Advantage population, specialist care costs were 6% ($1,227) higher than for primary care, although endocrinologist care resulted in a 9% reduction (–$3,710) in costs for patients with T1D. Higher pharmacy costs would be expected for the patients seen by endocrinologists, as their disease is typically more severe. Endocrinologists also had a much higher percentage of patients receiving insulin than in primary care (67% vs <30%, respectively). CMS does not receive enough Medicare Part D data to develop risk-adjustment payments for pharmacy costs. Thus, there is no adjustment of pharmacy costs based on disease severity, as there is for medical costs based on HCC categories of disease. Consequently, savings in medical costs may be more meaningful and relevant to CMS, as new programs and payment models are developed to address the needs of Medicare beneficiaries with DM.

In the context of the example of glycemic control and the conceptual framework for quality of care and costs³, our finding of significant savings associated with management of micro- and macrovascular complications by an endocrinologist is important. The average per-patient medical care costs for DM-related complications have been calculated as $68,037 over 30 years (2009 USD)⁹. For all patients in our study, lowering the annual per-patient medical costs for any complication by 10% and for micro- and macrovascular complications by 14% represents a short- and potentially long-term economic advantage for specialist care, but also suggests that patients are experiencing fewer or less severe complications and, thus, better quality-of-life.

Several limitations apply to our analysis. It was a retrospective claims analysis, which assumed that data entry and coding were accurate. Accurate and timely HCC coding is often a challenge for physicians and healthcare systems¹⁰. Claims data do not capture adherence to therapy, which may affect clinical outcomes, or quality-of-life, which may influence patient and physician decision-making about therapies. Furthermore, the cross-sectional study design can only identify associations, not cause–effect relationships.

This study found that endocrinologists saw their patients more often and treated them more aggressively than primary care physicians. Why this approach resulted in fewer complications and significantly lower total costs could be the subject for future research.

Conclusions

For patients in this study, DM management by an endocrinologist rather than a primary care physician resulted in greater HbA1C improvement and significantly lower medical costs. Although total medical and pharmacy costs were higher for patients treated by an endocrinologist, specialist care resulted in savings ranging from 2–9% for patients with T1D, regardless of the type of insurance coverage.

Transparency

Declaration of funding

Funding provided by Medtronic – Diabetes, Northridge, CA.

Declaration of financial/other relationships

MG, HC, MS, and CZ are employees of Medtronic – Diabetes, Northridge, CA. HL and FC have no relationships to disclose. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.

Acknowledgments

Sarah Staples, MA, ELS, assisted with manuscript preparation and was paid by Medtronic – Diabetes.