The Effect of Cost Sharing on Employees With Diabetes

December 15, 2006
Sean Nicholson, PhD
Sean Nicholson, PhD

Volume 12, Issue 12 SP

Most employers are responding to rapidly growinghealthcare costs by asking workers to pay moreof the premium and by increasing prices (eg,deductibles and copayments) workers face at the pointof care. In 1999, for example, 84% of people covered byemployer-sponsored health insurance had a physiciancopayment of $10 or less, whereas 72% had physiciancopayments of $15 or more in 2005.1 Likewise, copaymentsfor prescription drugs on the second (preferredbranded) and third (nonpreferred branded) tiers havealmost doubled between 2000 and 2005, rising from $13to $25 and from $17 to $33, respectively. The objectiveof these changes is to encourage employees and theirdependents to compare the benefits of medical carewith the cost, to discourage the use of low-value medicalservices, and thereby to slow down the growth rate ofhealth insurance premiums.

It is clear that increasing employee cost sharingachieves its primary objective of reducing costs.2,3 Forpeople with certain health conditions, however, increasingemployee cost sharing may have unintended consequences.Several recent studies show that patients withhigh blood pressure, elevated cholesterol levels, or diabetesmellitus (DM) are less likely to adhere to medicaltherapy when drug copayments rise.4-9 If the health ofnonadherent patients worsens such that they requireadditional medical services, then reducing drug copaymentsmay actually increase total medical expenditures.Furthermore, if one defines health costs broadly toinclude health-related absenteeism and reduced on-the-jobproductivity (often referred to as presenteeism) dueto a health condition, direct medical costs may actuallyrepresent a small proportion of total health-relatedcosts.10,11

At least 20 employers, including Pitney Bowes Incand the University of Michigan, are adopting a contrarianapproach.12-14 Rather than increasing copayments forall drug classes, these employers are selectively reducingcopayments for conditions such as diabetes, asthma,and hypertension in which the clinical evidence suggeststhere is a strong connection between adherenceto drug therapy and health. Unfortunately, there is nosingle study that performs a complete analysis of thefinancial effect of changing DM drug copayments onadherence, medical expenditures, absenteeism, and on-the-job productivity from the perspective of an employer,to my knowledge. The objective of this article is toperform such an analysis for DM by constructing afinancial model that links together results from severaldifferent published studies. The Diabetes OutcomesAnalyzer model, which is a disease-specific applicationof a general model proposed recently by Nicholson etal,15,16 focuses on DM because there are more studiesexamining the key links required for the financialmodel than for other diseases and because it is a fairlycommon and expensive condition.

Although this financial model can help employersdesign health benefits, one must be careful when interpretingthe results. The model uses mean values (eg, thepercentage change in pharmaceutical expendituresassociated with a $1 change in the drug copayment)across several different studies and then links these valuestogether to estimate the financial effect of drugcopayments on total health-related costs. For pragmaticreasons, I assume that the employee populations andthe examined interventions are similar to one anotherand are representative of employers nationally.


The baseline situation for an employer with 5000workers is given in Table 1.2,7,10,11,14,17-24 The analysisfocuses on workers only. Dependents and retirees couldbe incorporated, although the employer would not benefitfrom any changes in absenteeism or presenteeism.Each data element is described in column 1, the meanvalue based on the literature is given in column 2, andthe sources or assumptions are reported in column 3.Using the National Health and Nutrition ExaminationSurvey, Cowie et al17 reported the percentages of USemployees between 1999 and 2002 by sex and by agecategory (20-39 years vs 40-64 years), as well as the percentagesof employees (men and women combined) inthese age categories who were diagnosed as having DM.Based on these data, an estimated 6.6% of the employedpopulation has DM, or 332 workers for a typical employerwith 5000 workers.

The patient populations, interventions, and resultsfrom several key studies are summarized in Table2.1,7,14,18,25 Two studies examined the medication adherencerates for patients with DM. Among a population of137 300 nonelderly employees and dependents insuredby a manufacturing firm, Sokol et al18 found that 55% ofthe patients with DM adhered to their medication regimen.A patient was considered adherent if he or shereceived a supply of drugs for at least 80% of the year.Dor and Encinosa7 reported that 58% of patients withDM adhered to their physicians' recommended drugtreatment among a population of 27 100 workers andretirees who received employer-based health insurancefrom 9 large firms. In the latter study, a patient was consideredadherent if he or she had a full 3-month supplyof drugs following the expiration of his or her first DMprescription, among patients who filled at least 1 DMprescription between June 1999 and September 2000. Iaveraged these 2 results and assumed that, before anydrug copayments were changed, 56.5% of employeeswith DM adhered to their physicians' recommendeddrug therapy. For an employed population of 5000, thisimplies that 144 workers are not adherent among 332workers with DM.

The model assumes that an employer's pharmaceuticaland nonpharmaceutical expenses per employee withDM are $3922 and $10 776, respectively. These are themean amounts reported by Sokol et al18 and by Garrettand Bluml.14 Garrett and Bluml studied a population of256 patients with DM in Asheville, NC, who were coveredby employer-sponsored health insurance (Table2). The spending amounts reported in these 2 studieswere adjusted from the period examined (1999 forSokol et al18 and 2004 for Garrett and Bluml14) to 2006values by using the overall US growth rate of pharmaceuticaland nonpharmaceutical expenses. Specifically,I applied a 12.5% annual growth rate for pharmaceuticalexpenses between 1999 (the date of the study by Sokolet al18) and 2004; I applied a 9.1% annual growth rate fornonpharmaceutical expenses during the same period.The growth rates from 2004 (the date of the study byGarrett and Bluml14) to 2006 (8.2% for pharmaceuticalexpenses and 7.9% for nonpharmaceutical expenses) arethe actual growth rates in the United States between2003 and 2004.

Eight studies examined how often workers with DMwere absent because of their health condition. Althoughthe questions differed somewhat across the studies, thedata were self-reported by surveyed employees, andemployees were usually asked to distinguish health-relatedabsences from overall absences. The numbers ofannual self-reported absence days ranged from 1.2 daysto 12.6 days across the 8 studies, with a mean of 4.4days. Goetzel et al10,11 summarized 4 of these studies: theMedstat MarketScan Health and Productivity Managementdatabase that contained absenteeism informationon 375 000 employees between 1997 and 1999 at 6 largecorporations, the Employers Health Coalition, Inc,9 thatsurveyed 10 000 employees at 8 large employers in 1998and 1999, the Work Productivity Short Inventory(Ozminkowski et al19) that evaluated 619 employees at alarge telecommunications company, and the Bank OneWorker Productivity Index (Burton et al20) that collecteddata from 1000 Bank One customer service operators inIllinois in 1995. To arrive at the overall mean of 4.4 daysper year, the data from these 4 studies were combinedwith the data from 4 other studies on how frequentlyemployees with DM were absent: Collins et al21 surveyed6000 employees at Dow Chemical Company in 2004,Cranor et al22 and Cranor and Christiansen23 examined164 patients with DM covered by employer-sponsoredhealth insurance in Asheville, Wang et al24 examined anemployer-based population, and Egede et al26 examinedpatients with DM in the general population.

To estimate the annual cost associated with DM-relatedabsences, I multiplied the mean daily wage of anemployee in the United States by 4.4 days. This is a reasonableassumption if workers are paid according to themean value they provide a firm and if an absence resultsin 1 day's worth of output not being produced. Pauly etal27 showed that the minimum cost of an absence is anemployee's daily wage if labor markets are competitive,so this assumption is conservative. Pauly et al argue thatif it is difficult to substitute for an absent worker if theworker operates as part of a team or if the worker's outputcannot be postponed without some penalty (eg, lostsales), the true cost of an absence will exceed the worker'sdaily wage. With a mean annual salary of $41 414,fringe benefits that represent 28.6% of the salary, and240 work days per year, the estimated annual cost to anemployer of 4.4 absences is $980. The mean salary dataare from the Bureau of Labor Statistics, and the fringebenefit data as a percentage of salary are from the USDepartment of Labor. In the short run, an employercould capture the financial benefits of improvements ina worker's health-related productivity. However,employees who are absent less frequently due toimproved health should capture the benefits in the formof higher wages if other potential employers observeimproved productivity.27

Although some employers consider the financialeffect of reduced absenteeism when deciding whether toinvest in the health of their workers, most employersare skeptical about whether to include presenteeismimprovements in such calculations.28 In this model, Ireport the benefits associated with improved on-the-jobproductivity separately so that an employer can includeor omit them as he or she sees fit. Most of the researchto date on impaired presenteeism is based on informationcollected from workers regarding their perceptionof how often health impairs their productivity and of themagnitude of the impairment on their usual productivity.21,29,30 Three studies summarized by Goetzel et al,10,11which were already described in the absenteeism analysis,also examined presenteeism among workers whohave DM. Workers reported that their productivity duringthe 4 weeks before the survey was 11.4% lower, onaverage, than their usual level because of their healthcondition. If this productivity decrement occurred persistentlythroughout the year, the productivity losswould amount to $6138 based on the mean annualsalary (with benefits) in the United States.

Based on this set of assumptions, the total health-relatedcost per worker with DM is $21 749 per year(Table 1). About two thirds of these costs are directmedical expenditures and one third are indirect productivity-related costs. For a firm with 5000 workers, thetotal health-related costs for all workers with DM areestimated to be $7.2 million per year.


Are health-related costs likely to increase or decreaseif an employer cuts the copayment for DM drugs? Mostemployers have instituted a 3-tier drug copayment system.As summarized in Table 3, the mean copaymentfor generic drugs in employer-sponsored health planswas $10 in 2005, while the mean copayments for preferredand nonpreferred branded drugs were $22 and$35, respectively.1 Nationally, 53% of all prescriptionswere for generic drugs in 2004.31 No information wasavailable regarding the percentage of prescriptions thatwere for preferred versus nonpreferred branded drugs.For purposes of calculating a mean baseline drug copaymentper prescription, I assume that 40% of prescriptionsare for preferred branded drugs and that 7% are fornonpreferred branded drugs. This implies that employeespaid a mean of $16.55 per prescription at the pointof care in 2005.

In this article, I simulate the effect if an employerwere to put all DM drugs on the first tier. That is,employees would pay the $10 generic copayment for allDM drugs, even for branded drugs. Therefore, for anemployee with DM, the mean copayment would bereduced by $6.55, from $16.55 to $10. The model isflexible enough so that an employer could simulate anypossible copayment structure, such as eliminatingcopayments for DM drugs or raising all copayments tothe nonpreferred branded amount.

The first outcome of interest is whether medicationadherence rates would improve as a result of the $6.55reduction in the mean drug copayment. As summarizedin Table 2, Dor and Encinosa7 found that a $4 increasein drug copayments is predicted to reduce drug adherencerates by 3.4 percentage points, or 9%, amongemployees who have DM. As in other investigations, Iassume that the measured relationship between changesin copayments and adherence, medical spending, absenteeism,and presenteeism are symmetric. For example,if copayments were reduced by $4 per prescriptioninstead of increased by $4, I assume that adherencerates in the sample described by Dor and Encinosawould have increased by 9%. Furthermore, the measuredrelationships are assumed to be linear. In themodel, the relationship between copayments and adherencein the study by Dor and Encinosa is calibrated tothe magnitude of the simulated copayment change.Therefore, a $6.55 reduction in the mean copaymentwould be expected to improve adherence by 14.7%([-$6.55/$4] [-9%] = 14.7%). The model predicts that209 of 332 workers with DM would adhere to their recommendeddrug treatment under the lower copaymentdesign, an increase of 21 workers.

I adopt a similar approach to predict the effect oflower copayments on pharmaceutical and nonpharmaceuticalspending. Three of the studies summarized inTable 2 examined the relationship between copaymentsand pharmaceutical spending for patients with DM.Goldman et al25 reported that doubling the copayment,which represents an increase of $9.39 in the meancopayment in their sample, was associated with a 23%reduction in spending on DM drugs; Garrett and Bluml14found that a $14.40 reduction in copayments was associatedwith a 27.3% increase in pharmaceutical spendingfor employees with DM; and Dor and Encinosa7 predictedthat a $4 increase in copayments would reduce pharmaceuticalspending among patients with DM by 4.9%.Garrett and Bluml14 also examined a situation in whichemployers in Asheville entirely eliminated the copaymentsfor DM drugs. Because the study did not indicatehow large the copayments were before the change, Iassume for purposes of calibrating the model that thecopayments were equal to $14.40, the national mean in2004.1 Based on these 3 studies, the simulated $6.55decrease in the mean drug copayment is predicted toincrease pharmaceutical spending for employees withDM by 12.2%, on average, or $477 per worker per year.The spending before and after the simulated reductionin drug copayments is summarized in Table 3.Dor and Encinosa7 and Garrett and Bluml14 alsoexamined the relationship between copayments andnonprescription drug spending, as summarized inTable 2. They found evidence of cost offsets: reducingthe price of pharmaceuticals via the copayment wasassociated with reductions in nonpharmaceuticalexpenses among employees who have DM, and increasingdrug copayments had the opposite effect on nonpharmaceuticalcosts. Specifically, Dor and Encinosafound that a $4 increase in copayments was associatedwith a 6.4% increase in nonpharmaceutical spendingamong persons with DM, and Garrett and Bluml foundthat a $14.40 reduction in copayments was associatedwith a 14.4% decrease in nonpharmaceutical spending.

Dor and Encinosa did not directly examine the effect ofdrug copayments on nonpharmaceutical expenses.Instead, they applied their results on the relationshipbetween copayments and adherence to the findings byWagner et al32 regarding the relationship between adherence,glycemic control, and medical costs. The estimatednet spending increase associated with highercopayments in the study by Dor and Encinosa,7 combiningpharmaceutical and nonpharmaceutical spending,was $38 (or 3.7%) per patient with DM per year. Garrettand Bluml14 found that net spending fell by $72 (or0.9%) per person in the year following the reduction incopayments, whereas costs had been expected to rise by14.6% during this period. Based on these 2 studies, thesimulated $6.55 decrease in the mean drug copaymentis predicted to reduce nonpharmaceutical spending foremployees with DM by 8.5% on average, or $917 perworker per year. As summarized in Table 3, the totalmedical spending per employee with DM is predicted tofall by $439 per worker per year, or 3%.

The intervention in Asheville analyzed by Garrett andBluml14 had 2 components: (1) setting DM drug copaymentsto zero and (2) instituting a pharmaceutical careservices program in which patients could meet with apharmacist at no cost, learn about DM treatment and howto measure glucose levels from home, and obtain informationon how and why to adhere to treatment. Therefore,the results cited likely overstate the effects of changing onlycopayments on pharmaceutical and nonpharmaceuticalspending. Furthermore, because employees volunteeredfor the program, it is possible that employees who werehighly motivated to improve their health participated,such that the results might be different if an employerrequired all employees to participate.

Employers who believe that access to medical servicesaffects workers' productivity may be interested inwhether and how copayments affect absenteeism and onthe-job productivity. Although we cited 8 published studies2,10,19-24 documenting how often employees with DM areabsent, there is only a single study (to my knowledge)that examines whether absenteeism decreases when DMdrug copayments are changed. Cranor et al22 and Cranorand Christiansen,23 who examined the same interventionas Garrett and Bluml,14 tracked absenteeism of 164employees during a 5-year period. They found that a$14.40 reduction in the mean copayment for DM drugsreduced absenteeism in the first year by 48% (from 12.6days to 6.0 days). Absenteeism during the second throughthe fifth years essentially remained at this new lowerlevel. I assume that a similar effect would occur in themodel. Specifically, absenteeism would decrease by21.7% following a $6.55 reduction in the mean DM drugcopayment. Based on the mean daily wage, this translatesinto a $212 savings (Table 3) for the employer orthe employee, depending on how wages are subsequentlyestablished.

To the best of my knowledge, there are no studiesthat directly examine how a change in the copayment ofDM drugs (or any other kind of drug) affects the productivityof workers with DM when they are present forwork. Therefore, I assume that the direction and magnitudeof the on-the-job productivity effect are the sameas for absenteeism. That is, a $6.55 reduction in themean copayment is assumed to lead to a 21.7% improvementin the on-the-job productivity of workers with DM.There is some support for this assumption in the studyby Goetzel et al.10,11 The correlation between absenteeismand presenteeism across 10 health conditionswas 0.44. Workers with diseases that were associatedwith frequent absenteeism also had low productivitywhile present for work. If lower drug copayments reducepresenteeism costs by 21.7%, this would translate into asavings of $1315 per worker per year.

Because no published study (to my knowledge)exists on the relationship between copayments and onthe-job productivity and because many employers areskeptical about presenteeism in general, I first summarizethe results without presenteeism savings.Decreasing copayments on DM drugs is predicted toreduce medical and absenteeism costs by 4.2%, from$15 678 to $15 027 per worker per year. For an employerwith 5000 employees and 332 workers with DM, thispolicy is predicted to generate total savings of $216 300per year. About two thirds of the savings are due toreduced medical services and the remaining one thirdto reduced absenteeism. If estimated presenteeism savingsare included in the model, reduced copaymentsare predicted to reduce total health-related costs by9.0%. Twenty-two percent of the $1965 reduction incosts per worker with DM are due to reduced use ofmedical services. Reductions in absenteeism costs($212) represent 10.8% of the total savings, andimprovements in on-the-job productivity account forabout two thirds of the savings.

The presumed strength of this study is also its weakness.Taking a broad perspective on how changes in drugcopayments affect medical costs, absenteeism, and presenteeismcan provide employers with a more completeunderstanding of the financial effect of their actionsthan now exists in the literature, but it also requiresseveral simplifying assumptions that should make readerscareful when interpreting the results. For example,I assume that the measured relationships in the literatureare symmetric and linear. It is possible that the truerelationships between copayments and adherence, medicalspending, and productivity are nonlinear. For purposesof the model, I assume that the measured effectsoccur in the year following a change in copayments andare permanent, whereas in practice the effects maydiminish or accelerate over time. Therefore, the modelshould be updated as additional studies create morerefined estimates of the key variables.


Although most employers are increasing drug copaymentsin an effort to slow down the growth rate of healthinsurance premiums, at least 20 employers are adoptingthe opposite strategy of selectively reducing drug copayments for certain expensive health conditions.12,13 Tomy knowledge, there is no study that has performed acomprehensive financial analysis of the implications ofreducing drug copayments. This article constructs amodel based on different investigations that collectivelyexamined the relationships between copayments andadherence to drug therapy, pharmaceutical spending,nonpharmaceutical spending, absenteeism, and on-the-jobproductivity. For DM, the model predicts that reducingdrug copayments from their current levels wouldsave money by decreasing nonpharmaceutical, absenteeism,and presenteeism costs, which are more thanthe resulting increase in pharmaceutical costs.

The model is flexible enough to examine the effect ofany copayment structure an employer is considering.Furthermore, an employer can decide whether to includeall health-related costs, including absenteeismand presenteeism, or whether to focus only on directmedical spending.

If reducing drug copayments more than pays foritself, why are more employers not adopting this strategy?One explanation is that DM is an exception; formany other health conditions, reducing copaymentsmay increase pharmaceutical spending without conveyingimprovements in health and productivity. A secondexplanation is that employers or health benefit consultantsdo not want to treat (or are not used to treating)certain health conditions such as DM, asthma, andhypertension differently. Finally, two thirds of the savingsin the DM model come from improved on-the-jobproductivity, and there is little published researchdemonstrating a connection between benefit design andtangible increases in employee productivity.

From the Department of Policy Analysis and Management, Cornell University, Ithaca,NY, and the National Bureau of Economic Research, Cambridge, Mass.

This project was supported by Pfizer Inc.

Address correspondence to: Sean Nicholson, PhD, Department of Policy Analysis andManagement, Cornell University, Ithaca, NY 14853. E-mail:


Health Benefits: Annual Survey 2005.

1. Kaiser Family Foundation and Health Research and Educational Trust. Menlo, Calif: Kaiser Family Foundation andHealth Research and Educational Trust; 2005.

Am Econ Rev.

2. Manning WG, Newhouse JP, Duan N, Keeler EB, Leibowitz A, Marquis MS.Health insurance and the demand for medical care: evidence from a randomizedexperiment. 1987;77:251-277.


3. Joyce GF, Escarce JJ, Solomon MD, Goldman DP. Employer drug benefit plansand spending on prescription drugs. 2002;288:1733-1739.

J Gen Intern Med.

4. Ellis JJ, SR Erickson, JG Stevenson, SJ Bernstein, RA Stiles, Fendrick M.Suboptimal statin adherence and discontinuation in primary and secondary preventionpopulations. 2004;19:638-645.

Am J Manag Care.

5. Taira DA, Wong KS, Frech-Tamas F, Chung RS. Copayment level and compliancewith antihypertensive medication: analysis and policy implications for managedcare. 2006;12:678-683.

Am J Manag Care.

6. Goldman DP, Joyce GF, Karaca-Mandic P. Varying pharmacy benefits with clinicalstatus: the case of cholesterol-lowering therapy. 2006;12:21-28.

The Case of

Prescription Drugs.

7. Dor A, Encinosa W. Does Cost Sharing Affect Compliance? Cambridge, Mass: National Bureau of Economic Research;2004:1-30. NBER Working Paper 10738.

Am J Manag Care.

8. Gibson TB, Mark TL, McGuigan KA, Axelsen K, Wang S. The effects of prescriptiondrug copayments on statin adherence. 2006;12:509-517.

1999 Survey Data.

9. Employers Health Coalition, Inc. Tampa, Fla: EHC; 2000.

J Occup Environ Med.

10. Goetzel RZ, Long SR, Ozminkowski RJ, Hawkins K, Wang S, Lynch W. Health,absence, disability, and presenteeism cost estimates of certain physical and mentalhealth conditions affecting U.S. employers. 2004;46:398-412.

J Occup Environ Med.

11. Goetzel RZ, Hawkins K, Ozminkowski RJ, Wang S. The health and productivitycost burden of the "top 10" physical and mental health conditions affecting sixlarge U.S. employers in 1999. 2003;45:5-14.

Am J Manag Care.

12. Fendrick AM, Chernew ME. Value-based insurance design: a "clinically sensitive"approach to preserve quality of care and contain costs. 2006;12:18-20.

Am J Manag Care.

13. Fendrick AM, Smith DG, Chernew ME, Shah SN. A benefit-based copay forprescription drugs: patient contribution based on total benefits, not drug acquisitioncost. 2001;7:861-867.

J Am Pharm Assoc (Wash).

14. Garrett DG, Bluml BM. Patient self-management program for diabetes: first-yearclinical, humanistic, and economic outcomes. 2005;45:130-137.

Appl Health Econ Health Pol.

15. Nicholson S, Pauly MV, Polsky D, et al. How to present the business case forhealthcare quality to employers. 2005;4:209-218.

Health Econ.

16. Nicholson S, Pauly MV, Polsky D, Sharda C, Szrek H, Berger ML. Measuringthe effects of workloss on productivity with team production. 2006;15:111-123

Diabetes Care.

17. Cowie CC, Rust KF, Byrd-Holt DD, et al. Prevalence of diabetes and impairedfasting glucose in adults in the U.S. population: National Health and NutritionExamination Survey. 2006;29:1263-1268.

Med Care.

18. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medicationadherence on hospitalization risk and healthcare cost. 2005;43:521-530.

J Occup Environ Med.

19. Ozminkowski RJ, Goetzel RZ, Chang S, Long S. The application of two healthand productivity instruments at a large employer. 2004;46:635-648.

J Occup Environ Med.

20. Burton WN, Conti DJ, Chen CY, Schultz AB, Edington DW. The role of healthrisk factors and disease on worker productivity. 1999;41:863-877.

J Occup Environ Med.

21. Collins JJ, Baase CM, Sharda CE , et al. The assessment of chronic health conditionson work performance, absence, and total economic impact for employers.2005;47:547-557.


Am Pharm Assoc (Wash).

22. Cranor CW, Bunting BA, Christensen DB. The Asheville project: long-term clinicaland economic outcomes of a community pharmacy diabetes care program. 2003;43:173-184.

J Am Pharm Assoc (Wash).

23. Cranor CW, Christiansen DB. The Asheville project: short-term outcomes of acommunity pharmacy diabetes care program. 2003;43:149-159.

J Occup Environ Med.

24. Wang PS, Beck A, Berglund P, et al. Chronic medical conditions and workperformance in the Health and Work Performance Questionnaire calibration surveys.2003;45:1303-1311.


25. Goldman DP, Joyce GF, Escarce JJ, et al. Pharmacy benefits and the use ofdrugs by the chronically ill. 2004;291:2344-2350.

Diabetes Care.

26. Egede, LE. Effect of depression on work loss and disability bed days in individualswith diabetes. 2004;27:1751-1753.

Health Econ.

27. Pauly MV, Nicholson S, Xu J, et al. A new general model of the impact ofabsenteeism on employers and employees. 2002;11:221-231.

On the Brink of Change: How CFOs View Investments in Health and


28. Parry T. San Francisco, Calif: Integrated Benefits Institute; 2002. Available at: Accessed November 24, 2006.

J Occup Environ Med.

29. Burton WN, Chen CY, Conti DJ, Schultz AB, Pransky G, Edington DW. Theassociation of health risks with on-the-job productivity. 2005;47:769-777.


30. Stewart WF, Ricci JA, Chee E, Hahn SR, Morganstein D. Cost of lost productivework time among US workers with depression. 2003;289:3135-3144.

Intelligence Applied: 2005 Annual Report.

31. IMS Health Incorporated. Fairfield,Conn: IMS; 2005.


32. Wagner EH, Sandhu N, Newton KM, McCulloch DK, Ramsey SD, Gothaus LC.Effect of improved glycemic control on health care costs and utilization. 2001;285:182-189.