Journal of Diabetes and Its Complications 29 (2015) 529–533

Contents lists available at ScienceDirect

Journal of Diabetes and Its Complications journal homepage: WWW.JDCJOURNAL.COM

Health care resource utilization and costs during episodes of care for type 2 diabetes mellitus-related comorbidities☆ S.D. Candrilli a,⁎, J.L. Meyers a, K. Boye b, J.P. Bae b a b

RTI Health Solutions, 300 Park Offices Drive, Research Triangle Park, NC 27709, USA Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, USA

a r t i c l e

i n f o

Article history: Received 3 September 2014 Received in revised form 11 December 2014 Accepted 19 December 2014 Available online 24 December 2014 Keywords: Diabetes T2DM Comorbidities Episodes of care Economic burden Costs

a b s t r a c t Aims: To obtain costs of episodes of care for type 2 diabetes mellitus (T2DM)-related comorbidities. Methods: Data from the MarketScan Commercial Claims and Encounters Database were analyzed with the Medical Episode Grouper software, which uses proprietary algorithms to identify episodes of care. Episodes relevant to the T2DM population were examined, including: coronary artery disease with acute myocardial infarction, ventricular fibrillation, shock, and/or cardiac arrest (CAD episodes); cerebrovascular disease with stroke (CVD episodes); hypoglycemia; T2DM with complications (complication episodes); and renal failure. Results: 45,350 CAD; 85,287 CVD; 29,886 hypoglycemia; 40,339 complication; and 211,673 renal failure episodes were identified. Mean (SD) episode durations were 15.2 (39.1), 25.5 (55.0), 5.9 (24.0), 21.2 (54.6), and 364.0 (0.0) days, respectively. Inpatient visits were the largest component of unadjusted costs for CAD, CVD, and complication episodes (93.4%, 78.3%, and 91.9%, respectively). Other ancillary care represented the largest component of unadjusted costs for hypoglycemia (53.3%) and renal failure (80.5%) episodes. Mean adjusted total costs were $16,435; $4558; $445; $5675; and $8765 for CAD, CVD, hypoglycemia, complication, and renal failure episodes, respectively. Conclusions: This study adds important information to the literature regarding costs of episodes of care for patients with T2DM in the US. © 2015 Elsevier Inc. All rights reserved.

1. Introduction The health care delivery system in the United States is typically fragmented with services administered across different care settings by various health care providers. The corresponding fee-for-service health care payment system has also evolved in a complex and similarly fragmented structure, with each provider being paid individually for the services administered. In an effort to improve the quality of care, coordinate care across settings, and minimize unnecessary health care expenditures, a payment system based on a beneficiary’s episode of care has been proposed (Mechanic & Altman, 2009). Additionally, episodebased payment models have been proposed as part of the Affordable Care Act as a key initiative to reduce health care expenditures (Centers for Medicare and Medicaid Services, 2012). Conflict of interest: This study was supported by research funding from Eli Lilly and Company. JB and KB are employees of Eli Lilly and Company. SC and JM are employees of RTI Health Solutions; RTI Health Solutions received funding from Eli Lilly and Company to perform this research. ☆ Source of support: This study was funded by Eli Lilly and Company. ⁎ Corresponding author at: RTI Health Solutions, 300 Park Offices Drive, Research Triangle Park, NC 27709, USA. E-mail addresses: [email protected] (SD. Candrilli), [email protected] (JL. Meyers), [email protected] (K. Boye), [email protected] (JP. Bae). http://dx.doi.org/10.1016/j.jdiacomp.2014.12.009 1056-8727/© 2015 Elsevier Inc. All rights reserved.

In an episode-based system, payment would be bundled for all services required during a set period of care for a given condition. It is expected that in an episode-based payment system, health care providers would have a financial incentive to coordinate care and improve efficiency while also allowing physicians to have the flexibility to choose the treatment pathway best suited for the individual. Currently, the Centers for Medicare and Medicaid Services (CMS) are testing an episode-based payment system in practice for patients receiving acute care. In this program, hospitals receive bundled payment for hospital and physician services for a select set of inpatient care episodes (i.e., cardiac and orthopedic procedures) (Centers for Medicare and Medicaid Services, 2009). Additionally, initial studies are being conducted by managed care organizations to assess the value of bundling payment for patient care in areas such as oncology (Newcomer, Gould, Page, Donelan, & Perkins, 2014). As episode-based payment systems are gaining increasing attention, it is worth acknowledging that one of the most noted limitations of the episode-based payment system is how best to define an episode of care (Mechanic, 2011). Retrospective insurance claims databases allow for reporting of cost per service through International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM) diagnostic and procedure codes, Current Procedural Terminology-4 codes, National Drug Codes, and standard place of service codes.

530

S.D. Candrilli et al. / Journal of Diabetes and Its Complications 29 (2015) 529–533

However, the cost per service alone may not offer an accurate accounting of the total health care-related cost of a clinical event. Historically, when assessing the costs of diseases through the analysis of retrospective health insurance claims databases, the baseline approach has been to identify a sentinel event (e.g., first observed diagnosis), then look forward over some predefined period (e.g., 30 days, 90 days, 6 months, 1 year) and aggregate costs in various service categories (e.g., Bach, Mirkin, & Luke, 2011; Mehta, Suzuki, Glick, & Schulman, 1999; Min et al., 2008; Zaman et al., 2000). However, a one-size-fits-all approach for defining episodes of care may not be ideal, especially given the enormous potential for variation in clinical characteristics of different conditions and differences in severity within conditions themselves. Recently, software applications have been developed that contain clinically driven approaches to identifying episodes of care using administrative databases. One such software application is the Medical Episode Grouper (MEG) by Truven Health Analytics (Truven). This software is capable of identifying episodes for 555 disease categories. The objective of this analysis was to generate episode costs for patients with diabetes using the MEG software. 2. Materials and methods Data from Truven’s MarketScan Commercial Claims and Encounters Database (CCAE) were used in this analysis (Truven Health, 2014). The CCAE database primarily consists of employer-sourced and health plan-sourced data containing medical and drug utilization information. Between 40 and 50 million unique individuals were included in the database during the study period (i.e., 2009 to 2011), encompassing employees, their spouses, and their dependents covered by employer-sponsored private health insurance. In total, more than 100 large employers and 12 unique health plans throughout the US are represented in the database. Medical claims in the CCAE database include complete payment and charge information, dates and place of service (e.g., inpatient, outpatient, emergency), diagnoses, procedures, and detailed information on hospitalizations, including admission and discharge dates. Pharmacy claims in the CCAE database include complete outpatient prescription drug information, which consists of patient co-payments, mail order drugs, injectable medications, drugs from specialty pharmacies, and all standardized prescription-level fields collected on a typical pharmacy claim (e.g., date of fill/refill, drug name and class, strength, quantity, days’ supply). RTI International’s institutional review board (IRB) determined that this study met all criteria for exemption from IRB review. To define episodes of care in retrospective setting, we utilized MEG, a commercially available medical grouper software package for use with administrative claims data. This software uses a unique algorithm to chronologically sift through claims and organize data into unique episodes of care based on 555 disease staging and severity categories (Rattray, 2008). Specifically, MEG relies on ICD-9-CM codes for the basis of episode classification, and patients are followed for specific periods to identify episode start and end dates. Episode end dates are defined based on a designated clean window unique to each disease category during which no claims for a specific condition are present. Episodes include both acute (e.g., abdominal aneurysm) and chronic (e.g., renal failure) conditions, with episodes lasting up to a maximum of 1 year. Patients may have more than one episode type at a time, and not all claims are necessarily assigned to an episode (e.g., routine check-ups). Pharmacy claims are included in episode groups based on mappings of National Drug Codes to specific disease episode groups. Patients with at least two medical claims for outpatient services with a primary or nonprimary diagnosis on or after January 1, 2009, through December 31, 2011, for type 2 diabetes mellitus (T2DM) (ICD-9-CM diagnosis code 250.x0 or 250.x2) were selected for initial

study inclusion. From this group of patients, those who were at least 18 years of age on January 1, 2009, and who were continuously enrolled in their health plan from January 1, 2009, through December 31, 2011, were selected. Medical and pharmacy claims for the selected patients were input into the MEG software so that episodes of care could be identified for both acute and chronic conditions. Based on the episodes identified by MEG, patients were placed into non-mutually exclusive episode group cohorts if they had an episode in that group, and study measures were assessed separately for each of the episode group cohorts. This study focused on episodes relevant to the diabetes population. This included episodes for coronary artery disease with acute myocardial infarction, ventricular fibrillation, shock, and/or cardiac arrest (i.e., CAD episodes, which were considered acute); cerebrovascular disease with stroke (i.e., CVD episodes, which were considered acute); hypoglycemia (which were considered acute); diabetes mellitus with complications1 (i.e., complication episodes, which were considered acute), and renal failure episodes (which were considered chronic). Finally, all episodes that were truncated at the end of the data period (i.e., December 2011) were removed from the analysis to eliminate partial data bias. Background characteristics for patients experiencing episodes included in this analysis were reported. Background characteristics that were measured at the start of the observation period (i.e., January 1, 2009) included demographics (i.e., age, sex, geographic location) and the Charlson Comorbidity Index (CCI) score (Charlson et al., 2008). Additionally, the mean length per episode was also reported. For each episode type identified, episode-related health care costs were aggregated for the entire episode. Costs were reported by the major service sector (i.e., inpatient, emergency department, physician office visit, other ancillary care visits, pharmacy, and total costs) in which they occurred based on the place of service, billing, and other relevant codes associated with each medical claim observed as well as the total costs incurred across all service sectors. All episode-related health care claims were identified based on the output provided by MEG. Similarly, the percentage of patients with a claim during each episode was reported, overall and by care setting. All study measures were analyzed descriptively through the tabular display of mean values, medians, ranges, and standard deviations of continuous variables of interest and frequency distributions for categorical variables, and all analyses were conducted using SAS version 9.3 (Cary, NC: SAS Institute, Inc.; 2011). Prior to assessing the cost data, claims with a negative or missing cost value were removed from the analytic data set; no imputation was conducted. Finally, all cost data were adjusted to 2013 dollars and represent actual reimbursements paid by health plans for each medical or prescription encounter. Unadjusted costs were generated (and presented) for all episodes by episode type, as well as for just those episodes with care in each particular setting (e.g., inpatient costs across all CVD episodes and inpatient costs among those CVD episodes that included inpatient care). Multivariable models were also estimated to formally generate covariate-adjusted estimates of episode costs. Specifically, costs associated with an episode were reported for each care setting, as well as overall. Adjusted costs for each care setting and overall were estimated using generalized linear models (GLM) with a log-link function and a gamma distribution for the error term to resolve the issue of skewed cost distribution that is common in claims data (Wedderburn, 1974). Specifically, GLMs of the following general form were estimated: Y ¼ f ðβ0 þ βi Xi þ εÞ; 1 Diabetes mellitus complications include cellulitis, pyelonephritis, gangrenous infection, osteomyelitis, hyperosmolar state, ketoacidosis, acute myocardial infarction, acute cerebrovascular accident, sepsis, coma, hyperosmolar coma, and shock.

S.D. Candrilli et al. / Journal of Diabetes and Its Complications 29 (2015) 529–533

where Y is the episode cost, βs are coefficients to be estimated, f represents the gamma distribution, X1 is a vector of background characteristics (i.e., age, sex, CCI score, and geographic region), and ε is the error term. Following estimation of each model, mean predicted adjusted costs were presented for each episode type, overall and by each care setting (e.g., for each observation, the regression equation was solved using the regression parameter estimates and each observation’s actual data). 3. Results A total of 1,846,287 patients with T2DM aged at least 18 years on January 1, 2009, and continuously enrolled in their health plan from January 1, 2009, through December 31, 2011, were selected from the CCAE database and had their claims input into MEG. From these patients, a total of 34,786,690 unique episodes of care were created by MEG. Of these, 45,350 CAD episodes, 85,287 CVD episodes, 29,886 hypoglycemia episodes, 40,339 complication episodes, and 211,673 renal failure episodes were identified. Among patients with CAD episodes, the mean (SD) age was 66.0 (12.5) years, the mean (SD) CCI score was 5.2 (3.2), 33.7% were female, and patients were located primarily in the Midwest (31.1%), South (27.1%), and Northeast (25.0%) geographic regions. Similarly, among patients with CVD episodes, the mean (SD) age was 68.0 (13.0) years, the mean (SD) CCI score was 5.2 (3.1), 48.9% were female, and patients were located primarily in the Midwest (33.3%), South (26.9%), and Northeast (22.4%) geographic regions. Patients with hypoglycemia episodes were slightly younger (mean [SD] 60.1 [16.34] years) and had a slightly lower CCI score (mean [SD] 4.4 [3.2]), approximately 56% were female, and patients were located primarily in the Midwest (32.7%), South (26.3%), and Northeast (24.1%) geographic regions. Among patients with complication episodes, the mean (SD) age was 55.6 (15.5) years, the mean (SD) CCI score was 4.4 (3.1), 46.2% were female, and patients were located primarily in the Midwest (34.6%), South (22.0%), and Northeast (23.9%) geographic regions. Among patients with renal failure, the mean (SD) age was 66.9 (12.7) years, the mean (SD) CCI score was 6.3 (2.9), 43.6% were female, and patients were located primarily in the Midwest (35.7%), South (26.8%), and Northeast (19.4%) geographic regions (Table 1). CAD episodes lasted a mean (SD) of 15.2 (39.1) days and had mean (SD) unadjusted costs of $15,389 ($34,297). The largest component of unadjusted costs during the episode was inpatient visits (93.4% of total costs), followed by other ancillary care (6.0% of total costs). Over half of patients (57.2%) had an inpatient visit during the episode, 5.4% had an emergency department visit, and 54.8% had other ancillary care during the episode.

531

CVD episodes lasted a mean (SD) of 25.5 (55.0) days and had mean (SD) unadjusted costs of $4263 ($21,245). The largest component of unadjusted costs during the episode was inpatient visits (78.3% of total costs), followed by other ancillary care (18.1% of total costs). Approximately 22.1% of patients had an inpatient visit during the episode, 6.0% had an emergency department visit, 52.0% had an office visit, and 56.7% had other ancillary care during the episode. Hypoglycemia episodes lasted a mean (SD) of 5.9 (24.0) days and had mean (SD) unadjusted costs of $444 ($1855). The largest component of unadjusted costs during the episode was other ancillary care (53.3% of total costs), followed by inpatient visits (14.2% of total costs), emergency department visits (12.0% of total costs), and office visits (11.0% of total costs). Approximately 1% of patients had an inpatient visit, 15.4% had an emergency department visit, 49.1% had an office visit, 51.2% received other ancillary care, and 16.3% had a pharmacy claim during the episode. Complication episodes lasted a mean (SD) of 21.2 (54.6) days and had mean (SD) unadjusted costs of $5559 ($14,812). The largest component of unadjusted costs during the episode was inpatient visits (91.9% of total costs) followed by other ancillary care (5.9% of total costs). Approximately 37.3% of patients had an inpatient visit, 4.4% had an emergency department visit, 45.2% had an office visit, and 39.2% received other ancillary care during the episode. Finally, renal failure episodes were considered chronic episodes and lasted for a mean (SD) of 364 (0) days for all patients and had mean (SD) unadjusted costs of $8285 ($38,716). The largest component of unadjusted costs during the episode was other ancillary care (80.5% of total costs) followed by inpatient visits (9.1% of total costs). Approximately 7.3% of patients had an inpatient visit, 5.9% had an emergency department visit, 78.9% had an office visit, and 60.8% received other ancillary care during the episode (Table 2). Adjusted total costs were similar to unadjusted values (presented in Table 1) for each episode type examined. For example, mean adjusted total episode costs were $16,435 for CAD episodes, $4558 for CVD episodes, $445 for hypoglycemia episodes, $5675 for complication episodes, and $8765 for renal failure episodes. Similar relationships were observed across care settings for each of the episodes assessed (e.g., mean unadjusted inpatient cost for CAD episodes, among episodes with inpatient care, was $25,139 versus mean adjusted inpatient cost of $24,305) (Table 3). 4. Discussion This study used a commercially available episode grouper software package to estimate health care costs associated with conditions of relevance to a diabetes population. Specifically, this study examined the duration and corresponding costs for episodes of care for CAD,

Table 1 Patient demographics and episode duration, by episode type. Characteristic

Episode type CAD

N Age (mean [SD]) CCI Score (mean [SD]) Female (N, %) Geographic region (n, %) Northeast Midwest South West Unknown Episode duration (mean [SD])

a

CVDb

Hypoglycemia

Complication

13.04 3.09 48.92

29,886 60.47 4.35 16,698

16.25 3.22 55.87

40,339 55.60 4.41 18,634

15.46 3.09 46.19

211,673 66.93 6.28 92,301

12.67 2.93 43.61

22.40 33.33 26.86 16.71 0.70 54.97

7191 9779 7845 4788 283 5.91

24.06 32.72 26.25 16.02 0.95 23.98

9647 13,970 8856 7544 322 21.19

23.91 34.63 21.95 18.70 0.80 54.62

41,053 75,551 56,696 36,976 1397 364.00

19.39 35.69 26.78 17.47 0.66 0.00

45,350 65.96 5.16 15,276

12.46 3.15 33.68

85,287 68.03 5.24 41,719

11,346 14,110 12,285 7284 325 15.24

25.02 31.11 27.09 16.06 0.72 39.11

19,101 28,427 22,907 14,251 601 25.52

CCI = Charlson Comorbidity Index; SD = standard deviation. a CAD episodes include coronary artery disease with acute myocardial infarction, ventricular fibrillation, shock, and/or cardiac arrest. b CVD episodes include cerebrovascular disease with stroke.

Renal failure

532

S.D. Candrilli et al. / Journal of Diabetes and Its Complications 29 (2015) 529–533

Table 2 Unadjusted health care resource utilization and costs, by episode type. Care setting

Episode type

N Inpatient Patients with a visit (N, column %) Costs (mean [SD]), among all Costs (mean [SD]), among those with inpatient care ED Patients with a visit (N, column %) Costs (mean [SD]), among all Costs (mean [SD]), among those with ≥1 ED visits Office visits Patients with a visit (N, column %) Costs (mean [SD]), among all Costs (mean [SD]), among those with ≥ 1 office visits Other ancillary care Patients with a claim (N, column %) Costs (mean [SD]), among all Costs (mean [SD]), among those with ≥ 1 ancillary care encounters Pharmacy Patients with a claim (N, column %) Costs (mean [SD]), among all Costs (mean [SD]), among those with ≥ 1 filled prescriptions Total Patients with a visit/claim (N, column %) Costs (mean [SD]), among allc

CADa

CVDb

Hypoglycemia

Complication

Renal failure

45,350

85,287

29,886

40,339

211,673

25,921 $14,369 $25,139

57.16 $33,826 $41,607

18,822 $3337 $15,122

22.07 $20,240 $40,965

281 $63 $6714

0.94 $1367 $12,437

15,055 $5110 $13,692

37.32 $14,440 $21,004

15,376 $751 $10,338

7.26 $6583 $22,304

2452 $45 $833

5.41 $848 $3555

5150 $27 $450

6.04 $347 $1344

4605 $53 $345

15.41 $289 $664

1761 $35 $804

4.37 $319 $1309

12,539 $34 $580

5.92 $478 $1881

13,589 $57 $190

29.96 $241 $411

44,305 $103 $199

51.95 $475 $645

14,678 $49 $99

49.11 $213 $296

18,251 $66 $146

45.24 $270 $386

16,7052 $324 $410

78.92 $1662 $1861

24,867 $916 $1670

54.83 $4116 $5444

48,390 $770 $1357

56.74 $3593 $4686

15,314 $237 $462

51.24 $1008 $1371

15,821 $327 $833

39.22 $2425 $3817

128,654 $6667 $10,970

60.78 $37,452 $47,545

472 $3 $287

1.04 $66 $581

1577 $26 $1398

1.85 $313 $1841

4860 $42 $258

16.26 $412 $994

904 $21 $946

2.24 $286 $1663

98671 $509 $1091

46.61 $1681 $2330

45,350 $15,389

100.00 $34,297

85,287 $4263

100.00 $21,245

29,886 $444

100.00 $1855

40,339 $5559

100.00 $14,812

211,673 $8285

100.00 $38,716

ED = emergency department; SD = standard deviation. a CAD episodes include coronary artery disease with acute myocardial infarction, ventricular fibrillation, shock, and/or cardiac arrest. b CVD episodes include cerebrovascular disease with stroke. c By definition, all episodes had a positive total cost.

CVD, hypoglycemia, complications associated with diabetes, and renal failure among commercially insured patients with T2DM. Total adjusted episode costs ranged from an average of $445 per hypoglycemia episode (mean episode duration of 6 days) to $16,435 per CAD episode (mean episode duration of 15 days). The analysis suggests that mean episode duration varies significantly from one condition to another, ranging from less than a week to a year. But also important is the recognition that episode duration varies from patient to patient, as evidenced by a coefficient of variation ranging from 2.15 (CVD) to 4.06 (hypoglycemia). This emphasizes the need

for using episode-based costing approaches rather than aggregating costs across a simple fixed duration (such as 30 days). Limited real-world data exist on episode costs for patients with T2DM. O’Byrne and colleagues examined T2DM patients with CAD using Episode Treatment Groups software and found that total CAD episode payments were $65,661 over 4 years (averaging $16,415 per year) (O’Byrne et al., 2013). This is similar to our (unadjusted) estimate of $15,389 per CAD episode found using the MEG software. Quilliam and colleagues used MarketScan data to assess costs associated with hypoglycemia among patients with T2DM (Quilliam,

Table 3 Adjusted health care costs per episode for each care setting among patients with utilization in the care setting, by episode type. Care setting

Episode type CADa

Inpatient costs Patients with a visit (N, column %) Adjusted costs (mean [SD]) ED costs Patients with a visit (N, column %) Adjusted costs (mean [SD]) Office visits costs Patients with a visit (N, column %) Adjusted costs (mean [SD]) Other ancillary care costs Patients with a claim (N, column %) Adjusted costs (mean [SD]) Pharmacy costs Patients with a claim (N, column %) Adjusted costs (mean [SD]) Total costs Patients with a visit/claim (N, column %) Adjusted costs (mean [SD])

25,921 $24,305 2452 $886

CVDb 57.16

5.41

18,822 $16,493 5150 $424

Hypoglycemia

Complication

22.07

281 $7562

0.94

15,055 $13,363

6.04

4605 $356

15.41

1761 $693

Renal failure 37.32

15,376 $10,495

7.26

4.37

12,539 $550

5.92

13,589 $185

29.96

44,305 $184

51.95

14,678 $90

49.11

18,251 $146

45.24

167,052 $400

78.92

24,867 $1792

54.83

48,390 $1490

56.74

15,314 $502

51.24

15,821 $841

39.22

128,654 $10,571

60.78

1.04

1577 $1383

1.85

4860 $263

16.26

904 $918

2.24

98,671 $1085

46.61

100.00

85,287 $4558

100.00

211,673 $8765

100.00

472 $318 45,350 $16,435

29,886 $445

100.00

40,339 $5675

ED = emergency department; SD = standard deviation. a CAD episodes include coronary artery disease with acute myocardial infarction, ventricular fibrillation, shock, and/or cardiac arrest. b CVD episodes include cerebrovascular disease with stroke.

100.00

S.D. Candrilli et al. / Journal of Diabetes and Its Complications 29 (2015) 529–533

Simeone, Ozby, & Kogut, 2011). They found that the mean cost for an outpatient hypoglycemia visit was $394, which is similar to the (unadjusted) episode cost reported in our analysis of $444 with slightly less than 1% of patients having an inpatient visit and 15.4% of patients having an emergency department visit. However Quilliam and colleagues did not account for care received during the entire episode, which likely explains their lower cost estimate. Hussey and colleagues examined episodes of care for patients with acute myocardial infarction, diabetes, and other conditions in a Medicare population using the Episode Treatment Groups software (Hussey, Sorbero, Mehrotra, Liu, & Damberg, 2009). Consistent with our analysis, they found that patients often had numerous comorbid conditions; among patients with acute myocardial infarction, 63% also had hypertension, 54% also had congestive heart failure, and 35% also had diabetes. Similarly, our analysis found CCI scores that ranged from a mean (SD) of 4.4 (3.2) for T2DM patients with a hypoglycemia episode to 5.2 (3.2) for patients with a CAD episode and 6.3 (2.9) for patients with a renal failure episode. Furthermore, the study by Hussey and colleagues found that during an episode of care, patients often sought care with numerous different providers, including both primary care physicians and specialists. The fact that patients often have multiple comorbid conditions and often seek treatment from multiple providers highlights the many issues to be resolved when trying to implement a payment scheme for an episode of care and how a one-size-fits-all approach to episode costing may not be feasible in real-world setting where many patients have multiple comorbidities. This study was conducted using an administrative claims database and has limitations common to retrospective database analyses. No clinical data or electronic medical records were available to confirm diagnoses of clinical events. Episode grouping was based on recorded diagnosis codes, which may not be 100% accurate. Cost data provided in the database were based on health plan paid amounts, which may only reflect an average cost incurred to payers. Furthermore, patients in both capitated and fee-for-service health plans were included in this analysis. Historically, patients with capitated plans may not include physician payments for some services provided, as such cost estimates may be conservative by virtue of missing cost data for capitated enrollees. However, guidelines suggest that managed care plans are beginning to enhance encounter records with fee-forservice equivalent financials for patients in capitated health plans. Patients were required to have 2 years of continuous health plan enrollment; therefore, patients who may have died during study follow-up were not included in this analysis. Further, this study examined patients in a commercial claims population. While the study data are generally representative of the US commercially insured population, results may be different for patients in fee-forservice Medicare or Medicaid programs. The regions and local areas reflected in the study may not span the entire US since the health plans that contribute data to the database may not cover every region or local area in the US. Finally, chronic episodes were defined to be 364 days long; however, it is almost certainly the case that chronic conditions may have episodes lasting longer than 364 days, suggesting that cost estimates for such conditions may be conservative. The limitations previously denoted notwithstanding, the results presented in this article add important information to the body of literature regarding costs of episodes of care for patients with diabetes in the US, particularly given the paucity of published data in this area. Findings from this study may address important literature and methodological gaps related to US health care costing as well as inform economic modeling. Role of the funding source This study was supported by research funding from Eli Lilly and Company. The publication of study results is not contingent upon the

533

sponsors’ approval or censorship of the manuscript. Aside from the participation of JB and KB, the study sponsor had no role in study design, analysis and interpretation of data, writing the manuscript, or the decision to submit the manuscript for publication. Submission declaration and verification The material in this manuscript has not been published and is not under consideration for publication elsewhere. If accepted for publication in Journal of Diabetes and Its Complications, the manuscript will not be published in the same form, in English or in any other language, including electronically, without the written consent of the copyright holder. Acknowledgements The authors would like to thank Joseph A. Johnston and Dara Schuster of Eli Lilly for contributions to elements of the original study design. Further, the authors would like to thank Shweta Madhwani and Daniel Siepert of RTI-HS for assistance with quality review and editorial preparation of the manuscript, respectively. References Bach, P. B., Mirkin, J. N., & Luke, J. J. (2011). Episode-based payment for cancer care: A proposed pilot for Medicare. Health affairs (Project Hope), 30(3), 500–509, http:// dx.doi.org/10.1377/hlthaff.2010.0752. Centers for Medicare and Medicaid Services (2009). Roadmap for implementing value driven healthcare in the traditional Medicare fee-for-service program. Available at: http://www.cms.gov/Medicare/Quality-Initiatives-Patient-AssessmentInstruments/QualityInitiativesGenInfo/downloads/vbproadmap_oea_1-16_508. pdf (Last accessed November 24, 2014) Centers for Medicare and Medicaid Services (2012). The Affordable Care Act: Lowering Medicare costs by improving care efforts will save over $200 billion for taxpayers through 2016, nearly $60 billion for beneficiaries in traditional Medicare. Available at: http://www.cms.gov/apps/files/aca-savings-report-2012.pdf (Last accessed November 24, 2014) Charlson, M. E., Charlson, R. E., Peterson, J. C., Marinopoulos, S. S., Briggs, W. M., & Hollenberg, J. P. (2008). The Charlson comorbidity index is adapted to predict costs of chronic disease in primary care patients. Journal of Clinical Epidemiology, 61(12), 1234–1240, http://dx.doi.org/10.1016/j.jclinepi.2008.01.006. Hussey, P. S., Sorbero, M. E., Mehrotra, A., Liu, H., & Damberg, C. L. (2009). Episode-based performance measurement and payment: Making it a reality. Health affairs (Project Hope), l28(5), 1406–1417, http://dx.doi.org/10.1377/hlthaff.28.5.1406. Mechanic, R. E. (2011). Opportunities and challenges for episode-based payment. New England Journal of Medicine, 365(9), 777–779, http://dx.doi.org/10.1056/ NEJMp1105963. Mechanic, R. E., & Altman, S. A. (2009). Payment reform options: Episode payment is a good place to start. Health affairs (Project Hope), 28(2), w262–w271, http://dx.doi. org/10.1377/hlthaff.28.2.w262. Mehta, S. S., Suzuki, S., Glick, H. A., & Schulman, K. A. (1999). Determining an episode of care using claims data. Diabetic foot ulcer. Diabetes Care, 22(7), 1110–1115. Min, J. K., Robinson, M., Shaw, L. J., Lin, F., Legorreta, A. P., & Gilmore, A. (2008). Differences in episode-based care costs for multidetector computed tomographic coronary angiography versus myocardial perfusion imaging for the diagnosis of coronary artery disease. Journal of Medical Economics, 11(2), 327–340, http://dx. doi.org/10.3111/13696990802134291. Newcomer, L. N., Gould, B., Page, R. D., Donelan, S. A., & Perkins, M. (2014). Changing physician incentives for affordable quality cancer care: Results of an episode payment model. Journal of Oncology Practice, http://dx.doi.org/10.1200/JOP.2014. 001488. O’Byrne, T. J., Shah, N. D., Wood, D., Nesse, R. E., Killinger, P. J., Litchy, W. J., et al. (2013). Episode-based payment: Evaluating the impact on chronic conditions. Medicare & Medicaid Research Review, 3(3), http://dx.doi.org/10.5600/mmrr.003.03.a07. Quilliam, B. J., Simeone, J. C., Ozby, A. B., & Kogut, S. J. (2011). The incidence and costs of hypoglycemia in type 2 diabetes. The American Journal of Managed Care, 17(10), 673–680. Rattray, M. C. (2008). Measuring healthcare resources using episodes of care. http:// carevariance.com/images/Measuring_Healthcare_Resources.pdf (Last accessed March 1, 2014) Truven Health (2014). Marketscan Commercial Claims and Encounters database. Available at https://marketscan.truvenhealth.com/marketscanportal/portal.aspx (Accessed October 2014) Wedderburn, R. W. M. (1974). Quasi-likelihood functions, generalized linear models, and the Gauss–Newton method. Biometrika, 61, 439–447. Zaman, A., Goldberg, R. J., Pettit, K. G., Kaniecki, D. J., Benner, K., Zacker, C., et al. (2000). Cost of treating an episode of variceal bleeding in a VA setting. American Journal of Gastroenterology, 95(5), 1323–1330.

Health care resource utilization and costs during episodes of care for type 2 diabetes mellitus-related comorbidities.

To obtain costs of episodes of care for type 2 diabetes mellitus (T2DM)-related comorbidities...
206KB Sizes 0 Downloads 5 Views