Economic Outcomes Associated with Microvascular Complications of Type 2 Diabetes Mellitus

Economic Outcomes Associated withMicrovascular Complications of Type 2Diabetes MellitusResults from a US Claims Data Analysis

Elise M. Pelletier,1 Boyung Shim,1 Rami Ben-Joseph2 and J. Jaime Caro3

1 Health Economics and Outcomes Research, IMS Health Incorporated, Watertown, Massachusetts, USA

2 Health Outcomes, sanofi-aventis, Bridgewater, New Jersey, USA

3 United BioSource Corp, Concord, Massachusetts, USA

Abstract Background: Patients with diabetes mellitus have been shown to be at high

risk for both macrovascular and microvascular complications (MVC). Re-

cent studies have focused on MVC and their effect on the healthcare system,

but limited published data exist on long-term costs associated with MVC in

patients with type 2 diabetes mellitus (T2DM).

Objective: The objective of this study was to compare resource utilization and

medical costs over a 12-month period among patients diagnosed with T2DM

with versus without MVC in a managed-care population.

Methods: Patients aged ‡18 years, diagnosed with T2DM between 1 January

2003 and 31 December 2004 were identified in an administrative claims data-

base of approximately 55 million beneficiaries in private and public health

plans. The date of the first T2DM diagnosis during this period was the ‘index

date’ for each patient. All patients had to have a minimum of 12 months of

continuous enrolment both prior to and following the index date. MVC was

identified during the 12 months prior to the first T2DM diagnosis and these

patients were matched (1 : 2) by age, sex and ten co-morbid conditions to

those with no evidence of MVC during the entire study period.

Results: Among the 15 326 MVC patients included in the study, 61% had a

history of peripheral neuropathy, 28% diabetic retinopathy and 19% nephro-

pathy. Compared with 30 652 patients without MVC, the MVC patients were

more likely to use oral antidiabetics and insulin and had a higher co-

morbidity score. Over 12 months, patients with MVC had more (mean 0.3 vs

0.2; p < 0.001) and longer (mean length of stay 1.79 days vs 0.85 days;

p < 0.001) hospital stays; physician office visits (19.7 vs 13.7; p < 0.001); andprescriptions for oral antidiabetic (6.3 vs 5.6 scripts; p < 0.001) and insulin

(0.7 vs 0.2 scripts; p< 0.001) use. Average total costs per patient over 12 months

were $US14 414 with MVC versus $US8669 without MVC (p < 0.001).

ORIGINAL RESEARCH ARTICLEPharmacoeconomics 2009; 27 (6): 479-490

1170-7690/09/0006-0479/$49.95/0

ª 2009 Adis Data Information BV. All rights reserved.

Conclusion: This study indicates that in patients with T2DM, MVC is asso-

ciated with significant consumption of healthcare resources. Mean total

costs with MVC were almost double those of patients without MVC over a

12-month period.

Background

Patients with type 2 diabetes mellitus (T2DM)have been shown to be at high risk of bothmacrovascular and microvascular complications(MVC), including retinopathy, nephropathy andneuropathy.[1-3] Diabetic retinopathy leads to12 000–24 000 cases of blindness in the US eachyear while diabetic nephropathy affects approxi-mately 40% of patients with diabetes; about60–70% of patients with diabetes have neuro-pathy, with >60% of non-traumatic lower-limbamputations attributable to the disease.[1,4] TheseMVC lead to unnecessary clinical and economicburdens for both patients and payers. Prior studieshave examined the effect of macrovascular com-plications on the healthcare system.[5-13] Gandraet al.[5] recently published an analysis that focusedsolely on macrovascular disease such as acutemyocardial infarction, congestive heart failureand peripheral vascular disease. The objective ofthis study was to estimate the resource utilizationand costs associated with patients with T2DMwith chronic MVC compared with those with noevidence of MVC to provide managed-care deci-sion makers with a more complete understandingof the total economic impact of T2DM.

Methods

Data Source

Data were obtained from the PharMetricsPatient-Centric Database, which comprises fullyadjudicated medical and pharmacy claims for >50million unique patients in 88 health plans acrossthe US; adjudication ensures that claims have allrequired and relevant information that results inthe release of payment to the provider of thehealthcare services and, therefore, removes anyclaims that were denied. The database includes in-patient and outpatient diagnoses (in International

Classification of Diseases [9th Edition]-ClinicalModification [ICD-9-CM] format) and procedures(in CPT [Current Procedural Terminology]-4 andHCPCS [Healthcare Common Procedure CodingSystem] formats), as well as both retail andmail-order prescription records, demographicvariables, product and payer type, provider speci-alty, allowed, paid and charged amounts, and startand stop dates for plan enrolment.

Patient Selection

Data for this study spanned the period from1 January 2002 through 31 December 2005(complete claims for the year 2006 were unavail-able at the time of study initiation). Patients aged‡18 years with a diagnosis of T2DM recordedfrom January 2003 through December 2004 wereidentified. The date of the first T2DM diagnosisduring this period was the ‘index date’ for eachpatient. Patients were defined as having T2DM ifthey had at least one diagnosis code of 250.x0and/or 250.x2 on a medical, facility or surgicalclaim and met at least one additional criterionduring the study period: (i) at least one othermedical, facility or surgical claim with the T2DMcodes at least 30 days apart; (ii) at least onepharmacy claim for any oral antidiabetic (OAD)medication; or (iii) at least two procedure codesfor a glycosylated haemoglobin HbA1c test (CPT-4 code 83036). All patients were required to haveat least 12 months of continuous enrolment bothprior to and following the index date.

The prior 12 months were used to identifypatients with chronic MVC, defined as having atleast one medical, facility or surgical claim with adiagnosis or procedure code for nephropathy,peripheral neuropathy or diabetic retinopathy,or a diagnosis, procedure or revenue code fordialysis (table I); patients without MVC werethose with no evidence of MVC during the entire

480 Pelletier et al.

ª 2009 Adis Data Information BV. All rights reserved. Pharmacoeconomics 2009; 27 (6)

24-month period. Patients were excluded if theyhad evidence of type 1 diabetes at any time duringthe study period, defined as any diagnosis of250.x1 or 250.x3 (5.5% of patients) or evidence ofinsulin prescriptions only with no evidence oforal antidiabetic therapy (0.2% of patients); pa-tients were also excluded if they had no evidenceof MVC during the prior 12 months but hadMVC during the 12-month follow-up (to removenew-onset MVC and focus the study on chronicMVC), or they were aged ‡65 years and not en-rolled in Medicare Advantage. The latter wereexcluded because their complete claim historiesmay not be available as a result of a lack of co-ordination of benefits with traditional Medicareor other payer.

Each diabetic patient with MVC was matchedrandomly to two diabetic patients withoutMVC by age, sex and ten co-morbid conditions(asthma, arthritis, depressive disorders, chronicobstructive pulmonary disease, dyslipidaemia,migraine, osteoporosis, congestive heart failure,ischaemic heart disease or stroke) because thesecould influence treatment regimens and were feltto be the factors most likely to differ betweengroups. A co-morbidity was present if at least onemedical, facility or surgical claim with the corre-sponding ICD-9-CM diagnosis code was foundduring the pre-index period. Each non-MVCcontrol patient had to be an exact match to eachMVC case patient; if more than two non-MVCpatients were found for each MVC patient, thefirst two patients were chosen. All unmatchedMVC patients (n = 846) were excluded.

Measures

Outcomes included resource utilization anddirect medical costs. Demographic (age, sex,geographic region, health plan and payer type)and clinical characteristics (history of MVC,history of T2DM, prior and concomitantantidiabetic use, history of the ten selected co-morbidities) were evaluated based on data ob-tained on the index date or during the pre-indexperiod. Co-morbidity burden was estimatedusing the Dartmouth-Manitoba adaptation ofthe Charlson co-morbidity index.[14]

Resource utilization and direct medical costs,including medications (OAD, insulin, all otherpharmacy claims), outpatient care (ED visits,physician office visits, laboratory diagnostictests, all other outpatient services) and inpatientcare (hospitalizations), during the 12-month fol-low-up were evaluated; services were identifiedusing standard episode treatment grouper pro-cessing and included all medical and pharmacyservices. Costs were defined from the perspectiveof the health plan, and paid claims were used as aproxy for costs; the paid variable does not includeco-payments or deductibles. Costs were ex-pressed in $US, year 2006 values, adjusted asnecessary using the medical care component ofthe US Consumer Price Index.

Statistical Analyses

Descriptive statistics were used to describedifferences in patient characteristics for MVCversus without MVC prior to and following

Table I. Codes for microvascular complications

Complication ICD-9-CM code CPT-4 code Revenue code

Diabetic retinopathy 250.5, 362, 362.0x

Peripheral neuropathy 250.6, 337.1, 351.x–355.x,

357.x–358.1, 536.3, 707.1x,

707.8–707.9, 782.0,

885.x–887.x, 895.x–897.x,

V49.6x–V49.7x

20802–20838, 24900–24940,

25900–25931, 26910–26952,

27590–27598, 27880–27899,

28800–28825

Nephropathy 250.4, 580.xx–588.xx, 593.xx

Dialysis V56.x 90935–90940, 90945–90999,

G0320–G0327, S9339

800–809, 820–859, 880–889

Other diabetic complications 250.7, 250.9

CPT-4 =Current Procedural Terminology; ICD-9-CM = International Classification of Diseases (9th Edition)-Clinical Modification.

Costs of Microvascular Complications in Type 2 Diabetes Mellitus 481


matching. Summary statistics were presented aspercentages for categorical variables and asmean – standard deviation (SD) for continuousvariables; statistical tests included the chi-squaretest and Wilcoxon rank-sum test.

Resource use and costs were compared for withversus without MVC using both parametric (chi-square tests) and nonparametric (Wilcoxon rank-sum) analyses, where appropriate. A p-value <0.05was considered statistically significant. A multi-variate analysis on total direct costs was also per-formed using a generalized linear model (GLM)with a g distribution and log-link function; candi-date variables in the base model included patientdemographics, the Charlson co-morbidity index,history of T2DM and previous antidiabetic use,with collinearity among variables evaluated duringmodel development.

Sensitivity analyses were conducted and includedthe log transformation of medical costs and anevaluation of the study results including patientswithout a history of MVC. All data managementand analyses were conducted using StatisticalAnalysis Software (SAS�), versions 8.2 and 9.1(SAS Institute Inc., Cary, NC, USA).

Results

Of the 903 697 patients who were initiallyidentified with at least one diagnosis code forT2DM, 761 677 (84.3%) met the more stringentT2DM criteria. After applying additional entrycriteria, 128 983 T2DM patients remained in thesample (16 172 MVC and 112 811 non-MVC pa-tients). Evidence of type 1 diabetes and con-tinuous eligibility were the main reasons forexclusion. Of this sample, 15 326 patients withMVC (95%) were matched to 30 652 withoutMVC (figure 1).

The average age for patients with and withoutMVCwas 56 years. In theMVCpopulation, 60.5%had peripheral neuropathy, 28.1% diabetic retino-pathy, 18.9% nephropathy and 1.0% peripheralcirculatory disorders or an unspecified complica-tion; only 1.4% were on dialysis (table II). Patientswith MVC were more likely to use antidiabetictherapy: 55% had at least one prescription forOAD medication compared with 46% without

MVC; and 6% took insulin versus 2% withoutMVC (both p< 0.001). The Charlson co-morbidityindex was significantly higher among patients withMVC (mean 1.73 vs 1.17). The majority were pa-tients with evidence of T2DM prior to the indexT2DM diagnosis, with 70% of MVC and 60% ofnon-MVC having a history of the disease.

Utilization and Cost of Healthcare Services12 Months’ Post-Index Date

Significant differences in resource use betweenthose with and without MVC were found for allcategories (table III). MVC patients continued tohave a higher use of OAD, averaging 6.31 scriptsversus 5.62 scripts (p < 0.001) in the non-MVCpatients; few ED visits or hospitalizations wereseen in either group. The most common reasonsfor hospitalization for both MVC and non-MVCpatients included other forms of chronic ischae-mic heart disease (7% and 9%, respectively), re-spiratory and other chest symptoms (5% and 6%,respectively), osteoarthritis (3% and 4%, respec-tively), acute myocardial infarction (3% and 4%,respectively) and heart failure (3% and 2%, re-spectively). The mean length of stay for theseconditions did not differ substantially; forchronic ischaemic heart disease, the averagehospital stay was 3.9 days for MVC and 3.7 daysfor non-MVC patients, while that for respiratorysymptoms, osteoarthritis, myocardial infarctionand heart failure averaged 3.0 days, 5.3 days, 8.7days and 8.4 days, respectively, for patients withMVC and 2.6 days, 5.0 days, 4.6 days and 5.2days, respectively, for those without MVC.

Direct medical costs over the 12-month follow-up period for patients with MVC was almostdouble that for patients without MVC (table IV).This difference in total costs was driven by higherhospitalization costs and outpatient servicesamongMVC patients. Average prescription costswere also greater, with a 294% difference in in-sulin costs and a 129% difference in OAD medi-cations. Mean costs for all other prescriptionswere approximately 141% higher with MVC(p < 0.001). Costs associated with end-stage renaldisease were higher among MVC patients (mean$US14 114 vs $US4846); however, few patients in



either group had evidence of the disease (2.7%with MVC and 0.3% without MVC). As is com-mon with costs, the medians were much lowerthroughout, implying outliers with more sig-nificant resource utilization, perhaps due to moresevere disease. Approximately 80% of patientswith MVC had a total of 12-months’ costs below$US16 000; in patients without MVC, almost80% were below $US10 000; the costliest 10% of

patients with MVC were above $US30 000 andup to $US2.1 million over 12 months, whereasin patients without MVC, these ranged from$US19 000 to $US426 000. Although the meannumber of hospitalizations was quite low, ad-missions resulted in 12-month costs of more than$US1 million for some patients.

A multivariate analysis of total direct medicalcare costs over 12 months in patients with versus

Poor data quality(n = 461)

Evidence of MVC following but notprior to the index T2DM diagnosis

(n = 110 696) Not continuously eligible for entire

24-month study period(n = 429 996)

<18 years of age, or ≥65 years of ageand not in a Medicare Advantage plan

(n = 47 854)

With MVC(n = 16 172)

Without MVC(n = 112 811)

Patients with T2DM identified from Jan 2003 to Dec 2004:(1) at least two ICD-9-CM codes of 250.x0 and/or 250.x2 at least 30 days apart; or(2) at least one ICD-9-CM code of 250.x0 and/or 250.x2 and at least one pharmacy claim for any OAD medication; or(3) at least one ICD-9-CM code of 250.x0 and/or 250.x2 and at least two procedure codes for an HbA1c test (CPT-4 83036) (n = 761 677)

Total patients with T2DM(n = 128 983; 17%)

With MVC(n = 15 326)

Without MVC(n = 30 652)

Evidence of T1DM(n = 43 687)

846 excluded 1 : 2 match

Fig. 1. Sample selection. CPT-4 =Current Procedural Terminology-4; ICD-9-CM = International Classification of Diseases (9th Edition)-Clinical Modification; HbA1c =glycosylated haemoglobin; MVC =microvascular complications; OAD =oral antidiabetic; T1DM = type 1 dia-betes mellitus; T2DM = type 2 diabetes mellitus.



Table II. Characteristics of patients with type 2 diabetes mellitus (T2DM) with and without microvascular complications (MVC)

Characteristics Patients with T2DM [n (%)]a p-Valueb

with MVC (n =15 326) without MVC (n =30 652)

Age (y) 1.000

18–34 255 (1.7) 510 (1.7)

35–64 13 370 (87.2) 26 740 (87.2)

‡65 1701 (11.1) 3 402 (11.1)

Sex (male) 7 276 (47.5) 14 552 (47.5) 1.000

Geographical region (in US) <0.001Northeast 4 361 (28.5) 7 974 (26.0)

Midwest 4 485 (29.3) 9 443 (30.8)

South 2 535 (16.5) 6 093 (19.9)

West 3 945 (25.7) 7 142 (23.3)

Plan type <0.001health maintenance organization 7 097 (46.3) 14 374 (46.9)

preferred provider organization 5 176 (33.8) 11 005 (35.9)

point of service 2 194 (14.3) 3 678 (12.0)

indemnity 634 (4.1) 1 147 (3.7)

other/unknown 225 (1.5) 448 (1.5)

Payer type <0.001commercial insurance 11 389 (74.3) 23 493 (76.6)

Medicare Advantage 461 (3.0) 612 (2.0)

Medicaid 1 806 (11.8) 3 577 (11.7)

self-insured 1 449 (9.5) 2 571 (8.4)

other/unknown 221 (1.4) 399 (1.3)

History of T2DM in the 12 months pre-index period 10 752 (70.2) 18 286 (59.7) <0.001Microvascular complicationsc

peripheral neuropathy 9 792 (60.5) NA

diabetic retinopathy 4 549 (28.1) NA

nephropathy 3 057 (18.9) NA

dialysis 159 (1.0) NA

other diabetic complicationsd 233 (1.4) NA

Antidiabetic therapy

oral antidiabetics 8 361 (54.6) 13 995 (45.7) <0.001insulins 946 (6.2) 659 (2.1) <0.001

Charlson co-morbidity index [mean (SD)] 1.73 (1.50) 1.17 (1.19) <0.001Co-morbidities of interest

asthma 881 (5.7) 1 762 (5.7) 1.000

arthritis 3 634 (23.7) 7 268 (23.7) 1.000

depressive disorders 1 201 (7.8) 2 402 (7.8) 1.000

chronic obstructive pulmonary disease 643 (4.2) 1 286 (4.2) 1.000

dyslipidaemia 7 192 (46.9) 14 384 (46.9) 1.000

migraine 303 (2.0) 606 (2.0) 1.000

osteoporosis 393 (2.6) 786 (2.6) 1.000

congestive heart failure 749 (4.9) 1 498 (4.9) 1.000

ischaemic heart disease 2110 (13.8) 4 220 (13.8) 1.000

stroke 808 (5.3) 1 616 (5.3) 1.000

a Unless otherwise indicated.

b Chi-square test for categorical variables; Wilcoxon rank-sum test for continuous variables.

c Totals do not sum to 100% because patients may have more than one event in the pre-index period and therefore be counted in multiple

categories.

d Diabetes with peripheral circulatory disorders or unspecified complications.

NA= not applicable.



without MVC showed that MVC patients in-curred 66% higher total costs than patientswithout MVC (p < 0.001), controlling for demo-graphic and clinical factors (table V). Being fe-male, enrolled in preferred provider organizationor point-of-service plans versus health main-tenance organization plans, and having multipleco-morbid conditions also resulted in higher costs(p < 0.001), with increasing costs associated with

an increase in the number of co-morbidities.Patients with a history of the disease in the pre-index period had lower overall costs (p < 0.001)than patients with newly diagnosed T2DM.

Sensitivity Analyses

Differences in 12-month costs were evaluatedfollowing log transformation of all costs, to account

Table III. Utilization of healthcare services over 12 months in patients with type 2 diabetes mellitus (T2DM)

Medications and services With MVC

(n =15 326)Without MVC

(n= 30652)p-Valuea

Pharmacy claims per person

Oral antidiabetics

mean (SD) 6.31 (7.33) 5.62 (6.71) <0.001

median 4.00 4.00

Insulin prescriptions per person

mean (SD) 0.65 (2.78) 0.24 (1.67) <0.001

median 0.00 0.00

All other pharmacy servicesb

mean (SD) 35.07 (32.91) 28.86 (28.11) <0.001

median 27.00 22.00

Outpatient claims

ED visits

mean (SD) 0.39 (1.05) 0.28 (0.84) <0.001

median 0.00 0.00

Physician office visits

mean (SD) 19.70 (26.82) 13.67 (19.18) <0.001

median 12.00 9.00

Laboratory diagnostic tests

mean (SD) 20.15 (27.23) 15.74 (18.14) <0.001

median 14.00 12.00

All other outpatient servicesc

mean (SD) 22.80 (29.65) 15.89 (19.62) <0.001

median 15.00 10.00

Inpatient claims

No. of hospitalizations per person

mean (SD) 0.27 (0.78) 0.16 (0.52) <0.001

median 0.00 0.00

Days in the hospital

mean (SD) 1.79 (9.44) 0.85 (4.25) <0.001

median 0.00 0.00

a Wilcoxon rank-sum test used to compare utilization of healthcare services.

b All pharmacy claims except oral antidiabetics and insulins.

c All outpatient services except emergency department visits, physician office visits, and laboratory diagnostic tests.

MVC =microvascular complications.



for non-Normal distribution related to potentialoutliers. Following the log transformation, MVCpatients continued to show significantly highertotal costs than non-MVC patients (p < 0.001),and higher costs for all pharmacy services (in-cluding OADs, insulins and all other therapies),all outpatient services (including ED and physi-cian office visits, laboratory diagnostic testing

and other outpatient services) and all inpatientadmissions (p< 0.001 for all comparisons).

To address the inclusion of only patients witha history of MVC, we re-ran the analysis includ-ing all patients with MVC in the post-index per-iod, regardless of pre-index evidence. Followingmatching, this increased our sample to 87 210patients (29 070 with MVC, 58 140 without

Table IV. Per-patient payments ($US, year 2006 values) for healthcare services over 12 months in patients with type 2 diabetes mellitus

(T2DM)

Costs With MVC (n= 15326) Without MVC (n =30 652) p-Valuea

Total overall costs

mean (SD) 14 414 (37 522) 8 669 (16 403) <0.001

median 5 985 3 940

Total pharmacy costs

mean (SD) 3 042 (6 366) 2 172 (4 368) <0.001

median 1 753 1 310

Oral antidiabetics

mean (SD) 453 (767) 350 (695) <0.001

median 123 81

Insulins

mean (SD) 53 (270) 18 (152) <0.001

median 0 0

All other pharmacy services

mean (SD) 2 537 (6 236) 1 804 (4 234) <0.001

median 1 257 988

Total outpatient costs

mean (SD) 6 177 (13 650) 3 770 (6 833)

median 2 938 1 817

ED visits

mean (SD) 290 (1 075) 215 (880) <0.001

median 0 0

Physician office visits

mean (SD) 1 607 (6 288) 921 (1 974) <0.001

median 784 551

Laboratory diagnostic tests

mean (SD) 433 (931) 325 (647) <0.001

median 233 200

All other outpatient services

mean (SD) 3 846 (10 295) 2 309 (5 445) <0.001

median 1 385 717

Total inpatient costs

mean (SD) 5 195 (30 455) 2 727 (12 040) <0.001

median 0 0

a Wilcoxon rank-sum test used to compare per-patient payments.

MVC =microvascular complications.



MVC). No changes in results were noted. Total12-month costs continued to be significantlyhigher among patients with MVC ($US15 188)versus those without MVC ($US8352; p < 0.001).The difference in total costs was again driven byhigher average hospitalization costs ($US6245 vs$US2624) and outpatient costs ($US6060 vs$US3633) among MVC patients. Costs for in-sulin were 253% higher and those for OADmedications were 128% higher in the MVC co-hort, while total pharmacy costs were 138%greater among MVC patients (p< 0.001). Logtransformation of the costs did not affect thesesignificant differences, while multivariate ana-lyses controlling for demographic and clinicalcharacteristics found that MVC patients contin-ued to incur 40% higher total costs than patientswithout MVC (p < 0.001).

Discussion

A retrospective studywas conducted to evaluatethe economic outcomes associated with MVC.Over half of the patients with MVC had a historyof peripheral neuropathy, while almost one-thirdexperienced diabetic retinopathy. The results of theanalyses suggest that healthcare costs of patientswith MVC are dramatically higher than those forpatients without MVC; greater utilization of in-patient and outpatient services, and to a smallerdegree OAD and insulin use, contributed to thisdifference.

Between one-third and one-half of patientswith T2DM had evidence of tissue or organ da-mage, while an epidemiologic study reportedthat almost one-fifth of patients with T2DMhave two or more complications.[12,13,15] A 2004

Table V. Multivariate analysis (generalized linear model [GLM])a of total direct medical care costs over 12 months

Variable Parameter estimate Standard error 95% CI Chi-square p-Value

With vs without MVC 0.507 0.01 0.49, 0.52 4026.36 <0.001

Female vs male 0.042 0.01 0.03, 0.06 31.83 <0.001

Age

18–34 y vs 35–44 y 0.074 0.03 0.02, 0.13 6.66 0.010

45–54 y vs 35–44 y 0.152 0.01 0.13, 0.18 130.19 <0.001

55–64 y vs 35–44 y 0.243 0.01 0.22, 0.27 357.41 <0.001

65+ y vs 35–44 y 0.158 0.02 0.12, 0.19 82.35 <0.001

Region (in US)

Northeast vs West -0.104 0.01 -0.13, -0.08 88.28 <0.001

Midwest vs West 0.057 0.01 0.04, 0.08 26.89 <0.001

South vs West -0.103 0.01 -0.13, -0.08 73.37 <0.001

Plan type

indemnity vs HMO -0.112 0.02 -0.15, -0.07 30.59 <0.001

PPO vs HMO 0.071 0.01 0.05, 0.09 57.59 <0.001

POS vs HMO 0.079 0.01 0.05, 0.10 41.42 <0.001

other/unknown vs HMO -0.074 0.03 -0.14, -0.01 5.24 0.022

Charlson co-morbidity index

1 vs 0 0.329 0.01 0.30, 0.36 590.11 <0.001

2 vs 0 0.646 0.02 0.62, 0.68 1807.13 <0.001

3 vs 0 0.961 0.02 0.92, 1.00 2456.44 <0.001

4 vs 0 1.317 0.02 1.28, 1.36 3780.43 <0.001

History of T2DM: yes vs no -0.449 0.01 -0.47, -0.43 1375.51 <0.001

History of insulin use: yes vs no 0.107 0.02 0.07, 0.15 26.34 <0.001a GLM performed with a g distribution and log-link function.

HMO =health maintenance organization; MVC =microvascular complications; POS = point of service; PPO =preferred provider organization;

T2DM = type 2 diabetes mellitus.



cross-sectional analysis of the prevalence ofMVC in types 1 and 2 diabetes found that 55%had retinopathy, 67% nephropathy and 68.5%neuropathy,[15,16] in accordance with otherdata.[1-4] The prevalence of MVC in our study,particularly nephropathy, was slightly lower.This could be due to the exclusion of type 1 dia-betic patients, who are also at risk of developingMVC, from our analysis as well as the relativelyshort (1-year) follow-up period that we used tocapture MVC claims. Additionally, we did notrequire diabetic patients to have had the diseasefor any particular length of time. Our patientsample also tended to be younger because of theclaims data utilized; therefore, we may be ex-cluding older patients or those with a longerduration of diabetes who might be at greater riskof MVC. Nevertheless, the high prevalence ofMVC among our patients suggests the need toreduce the development of these conditions,which have substantial economic implications.

The total annual cost of diabetes in the USwas estimated to be $US132 billion in 2002, orapproximately 11% of the US healthcare ex-penditure. Direct medical expenditures totalled$US92 billion, an increase of 109% from 1997.[2]

Diabetes care comprised $US23.2 billion of thesedirect costs, while $US24.6 billion went forchronic complications.[2] A report issued by theAmerican Association of Clinical Endocrino-logists, The State of Diabetes Complications inAmerica, found that poorly managed T2DMcosts the US health system an extra $US22.9 bil-lion per year (2006 values) in direct medical coststo treat serious health problems associated withthe disease; annual per capita costs totalled$US1785 for eye damage, $US4687 for foot pro-blems and over $US9000 for chronic kidney dis-ease.[17] A recent Markov model found that in-dividuals with cardiometabolic risk factors willpose a large economic burden on the healthcaresystem, even in the short term.[18]

In 2006, Gandra et al.[5] estimated that total12-month costs for T2DM-patients with macro-vascular complications averaged $US10 450 com-pared with $US3385 among those without thesecomplications, a 3-fold difference; inpatient costswere a large proportion of the total. Our study

found that average total direct costs for T2DMpatients with MVC were $US14414 comparedwith $US8669 for those without MVC, almost a2-fold difference, and higher than the total costs re-ported for macrovascular disease inGandra et al.[5]

Examining these data collectively provides health-care policymakers and managed-care plans witha more complete picture of the economic burdenof diabetic patients. The results from these stu-dies can be used in economic modelling to ex-amine effects of early prevention on long-termoutcomes. Programmes and treatments to pre-vent diabetes and its complications could resultin tremendous savings to the healthcare system,even in the short term. However, these data mostlikely underestimate the total cost burden, giventhat only direct medical costs were taken intoaccount in both studies; indirect costs such aswork loss and family care were not evaluated butcan have significant impact on patients and theirfamilies.

Several models show that appropriate treat-ment of diabetic patients results in long-termsavings. A model using 1992–2002 data fromover 734 000 patients with type 1 and 2 diabetespredicted that the introduction of a diabetesmanagement programme to reduce risk factors,including raised HbA1c and cholesterol levels,decreased macro- and microvascular complica-tions, resulted in a life-time incremental cost perQALY of $US5527 if the programme effectslasted 1 year and $US4448 per QALY if theycontinued for 10 years (payer perspective).[19,20]

Another diabetes model found that effective tar-geting of financial resources toward loweringHbA1c levels in line with guidelines could lead tosavings in the range of $US35–$US72 billion indirect healthcare costs over a decade.[21]

Limitations

Patient selection relied on ICD-9-CM codesfor detecting both T2DM and MVC. This mayhave included patients without T2DM or ex-cluded patients with T2DM, or may have mis-categorized patients with and without MVC. Inparticular, we found a sample of patients with-out MVC but with evidence of macrovascular



complications; because of the relationship be-tween these two, there may be patients withoutMVC who actually do have these complicationsbut are not receiving treatment for them. If thesenon-MVC patients were appropriately classifiedas MVC, this could reduce the difference in coststhat we found. Additionally, clinicians may notaccurately record a co-morbid diagnosis at thetime of treatment and, therefore, the proportionswith the co-morbid conditions of interest may beunder- or overestimated. Although there were nodifferences in the proportions with any of the tenconcomitant conditions used in matching, theCharlson co-morbidity index, which includes ad-ditional co-morbidities, was significantly higheramong patients with MVC; thus, co-morbid con-ditions not used in matching, including macro-vascular complications, may produce a moreserious burden of disease among those with MVC.

The amount of claims data available over timedoes not allow for an evaluation of how longpatients may have had T2DM. Our study lookedback only 12 months; therefore, it does not ac-count for duration of illness among patients in-cluded in the study. This could impact the resultsif patients in one cohort were more likely to havehad T2DM for a much longer period of time thatultimately resulted in different treatment regi-mens or complicating factors. However, control-ling for age, sex and co-morbidities helped toaccount for potential differences among the co-horts because of duration and severity of disease.

Identifying patients with MVC from claimsdata focused the study on patients with activedisease who might be more likely to undergotreatment and, in turn, potentially be more cost-ly. By limiting our study to those with prevalentdisease (with a history of MVC), we may havereduced the impact of initial treatment costs inparticular.

The database does not provide information onsystemic factors that could affect care, includingplan limits on medication use or complex medicalprocedures; however, because of the large anddiverse nature of the plans in the database, thesefactors should not impact study results. Ad-ditionally, we used managed-care reimbursementrates as a proxy for cost. While other measures

might have been considered, this has the advantageof reflecting costs to providers covered by one keydecision maker – managed-care payers.

Conclusion

This study indicated that T2DM patients withMVC consumed significantly more healthcarethan those without MVC, at almost twice thecost. This provides further evidence of the sub-stantial economic impact to health plans andpolicymakers of MVC among T2DM patients.

Acknowledgements

This study was funded by sanofi-aventis; Rami Ben-Joseph is a former employee of sanofi-aventis and holds stockin the company. Elise Pelletier, Boyung Shim and J. Jaime Carohave all received funding for research work from sanofi-aventis. All authors confirm that this paper is an accuraterepresentation of the study results; Rami Ben-Joseph, thestudy sponsor, was involved in the study design, writing ofthe report and decision to submit the paper for publication.The authors would like to thank Robert Rosofsky, AndrewHostetter and Sergey Sian for their assistance with the statis-tical programming and analysis of this study. We also wish tothank the PharmacoEconomics reviewers of our study for theirthoughtful and insightful comments.

References1. WHO. Diabetes: fact sheet no. 312 [online]. Available from

URL: http://www.who.int/mediacentre/factsheets/fs312/en [Accessed 2007 Mar 6]

2. Avogaro A, Giorda C, Maggini M, et al. Incidence of cor-onary heart disease in type 2 diabetic men and women:impact of microvascular complications, treatment andgeographic location. Diabetes Care 2007; 30 (5): 1241-7

3. Klein R, Klein BEK, Moss SE, et al. The Wisconsin epide-miologic study of diabetic retinopathy: III. Prevalence andrisk of diabetic retinopathy when age at diagnosis is 30 ormore years. Arch Ophthalmol 1984; 102: 527-32

4. National Institute of Diabetes and Digestive and KidneyDiseases (NIDDK), National Institutes of Health (NIH),National database statistics, 2007 [online]. Available fromURL: http://diabetes.niddk.nih.gov/dm/pubs/statistics [Ac-cessed 2007 Mar 6]

5. Gandra SR, Lawrence LW, Parasuraman BM, et al. Totaland component health care costs in a non-Medicare HMOpopulation of patients with and without type 2 diabetesand with or without macrovascular disease. J Manag CarePharm 2006; 12 (7): 546-52

6. Gilmer TP, O’Connor PJ, Manning WG, et al. The cost tohealth plans of poor glycemic control. Diabetes Care 1997;20 (12): 1847-53



7. Currie CJ, Morgan CL, Peters JR. Patterns and costs ofhospital care for coronary heart disease related and notrelated to diabetes. Heart 1997; 78 (6): 544-9

8. Glauber H, Brown J. Impact of cardiovascular disease onhealth care utilization in a defined diabetic population.J Clin Epidemiol 1994; 47 (10): 1133-42

9. Nichols GA, Brown JB. The impact of cardiovascular dis-ease onmedical care costs in subjects with and without type2 diabetes. Diabetes Care 2002; 25 (3): 482-6

10. O’Brien JA, Patrick AR, Caro J. Estimates of direct medicalcosts for microvascular and macrovascular complicationsresulting from type 2 diabetes mellitus in the United Statesin 2000. Clin Ther 2003; 25 (3): 1017-38

11. O’Brien JA, Shomphe LA, Kavanagh PL, et al. Directmedical costs of complications resulting from type 2 dia-betes in the US. Diabetes Care 1998; 21 (7): 1122-8

12. UK Prospective Study Group. UK Prospective DiabetesStudy (UKPDS) VIII: study design, progress and perfor-mance. Diabetologia 1991; 34: 877-90

13. Morgan CL, Currie CJ, Stott NC, et al. The prevalence ofmultiple diabetes-related complications. Diabet Med 2000;17 (2): 146-51

14. Romano PS, Roos LL, Jollis JG. Adapting a clinical co-morbidity index for use with ICD-9-CM administrativedata: differing perspectives. J Clin Epidemiol 1993 Oct; 46(10): 1075-9, 1081-90

15. Girach A, Manner D, Porta M. Diabetic microvascularcomplications: can patients at risk be identified? A review.Int J Clin Pract 2006 Nov; 60 (11): 1471-83

16. Rahman S, Zia I. Prevalence of microvascular complicationsamong diabetic patients. Pakistan JMed Res 2004; 43: 163-5

17. American Association of Clinical Endocrinologists. Thestate of diabetes complications in America. The AmericanAssociation of Clinical Endocrinologists (AACE) 16thAnnual Meeting and Clinical Congress; 2007 Apr 11-15;Seattle (WA)

18. Caro JJ, O’Brien JA, Hollenbeak CS, et al. Economic bur-den and risk of cardiovascular disease and diabetes inpatients with different cardiometabolic risk profiles.Value Health 2007; 10: S12-S20

19. O’Reilly DJ, Clarke P, Hopkins R, et al. Development andapplication of a diabetes economic model [abstract]. 28thAnnual Meeting of the Society for Medical DecisionMaking; 2006 Oct 17; Boston (MA)

20. O’Reilly D, Hopkins R, Blackhouse G, et al. Developmentof an Ontario Diabetes Economic Model (ODEM) andapplication to a multidisciplinary primary care diabetesmanagement program: report prepared for the OntarioMinistry of Health and Long-Term Care. 2006 Nov 2[online]. Available from URL: http://www.path-hta.ca/diabetes.pdf [Accessed 2007 Mar 6]

21. Minshall ME, Roze S, Palmer AJ, et al. Treating diabetesto accepted standards of care: a 10-year projection of theestimated economic and health impact in patients with type1 and type 2 diabetes mellitus in the United States.Clin Ther 2005 Jun; 27 (6): 940-50

Correspondence: Elise M. Pelletier, MS, IMS Health Inc.,311 Arsenal Street, The Arsenal on the Charles, Watertown,MA 02472, USA.E-mail: [email protected]



Documents

Economic Outcomes Associated with Microvascular Complications of Type 2 Diabetes Mellitus