Prior research has demonstrated

Transcription

1 Costs Of Severely Ill Members And Specialty Medication Use In A Commercially Insured Population Understanding cost distribution and burden among plan members is key to devising equitable benefit designs. by Vincent J. Willey, Michael F. Pollack, Wayne M. Lednar, William N. Yang, Charles Kennedy, and Grant Lawless ABSTRACT: This study examines the overall profile and costs associated with severely ill commercially insured people. We found severely ill members to have the highest costs, from both the insurer and member perspective. Even for the most costly members where specialty medication use was highest, biologics represented less than one-third of the pharmacy spending and 6.6 percent of overall spending. Out-of-pocket spending rose dramatically when medications were paid for under the pharmacy benefit rather than the medical benefit. The advantages of paying for specialty medications under the pharmacy benefit should be evaluated in conjunction with the potential consequences of increased out-ofpocket burden. [ Affairs 27, no. 3 (28): ; 1377/hlthaff ] Prior research has demonstrated that the treatment of chronic illnesses is associated with increased expenses compared with what the general population spends for health care. These increased expenses are borne by both the institutional payer (health insurance plan, employer, or government) and the individual patient. Much of this research has focused on chronic diseases such as diabetes, asthma, and heart disease. 1 However, some studies have begun to look at more severe diseases such as cancer, rheumatoid arthritis, and multiple sclerosis. 2 These diseases differ from the previously described diseases in that there are specialty medications used in their treatment that can modify or even arrest the course of disease. In addition, some of these severe chronic conditions (such as rheumatoid arthritis and multiple sclerosis) may severely reduce both individual and employer productivity, because they may affect patients at a younger age than other chronic conditions, last longer, and produce functional decline at the age of maximal Vincent Willey (vwilley@healthcore.com) is vice president, Research Development and Operations, at Core in Wilmington, Delaware. At the time of this writing, Michael Pollack was with Core. He is now associate director, Economics and Outcomes Research, at AstraZeneca LP, in Wilmington. Wayne Lednar was with Eastman Kodak at the time of this writing. He is now global chief medical officer at E.I. dupont de Nemours Company, also in Wilmington. William Yang is a health management physician at the Coca-Cola Company/Emory Clinic in Atlanta, Georgia. Charles Kennedy is vice president, Information Technology, at WellPoint in Camarillo, California. Grant Lawless is director, Corporate Accounts, at Amgen in Thousand Oaks, California. Core is an independent subsidiary of WellPoint Inc. 824 May/June 28 DOI 1377/hlthaff Project HOPE The People-to-People Foundation, Inc. Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

2 working productivity. Specialty medications include injectable biologic-derived agents that target specific immune processes and proteins. They are traditionally used in patients with severe and chronic diseases for which alternative treatments either are nonexistent or have failed to be continuously effective. The cost per unit for biologics is much higher than that for traditional pharmaceuticals. Further, biologics are used in a small percentage of plan members. The combination of high cost per unit and relatively focused use may attract health plans attention for cost management strategies. For the patients who require these treatments, depending upon the cost-sharing arrangement they have with the payer and the out-of-pocket burden they face, the economic burden may pressure them to forgo some essential medical care. Prior research has demonstrated that chronically ill people with medical bill problems are four to five times more likely than those without medical bill problems to forgo or delay medical care because of the economic burden. 3 This behavior can produce negative unintended consequences, both clinical and economic. Theprimaryobjectiveofthisstudywasto characterize and quantify the medical services andmedicationspendingforahealthplanand its members with severe and chronic conditions. Although previous work has found that a majority of health care costs are borne by a small proportion of the population, we believe that better characterization of these populations and quantification of costs will be useful from a cost, quality, and safety perspective. 4 We further examined the characteristics of members in the top 1 percent of the plan population in health plan costs, with a focus on the effect of what types of services were driving plans and members spending and the use of biologics on health plans costs. Study Data And Methods For this study, we used a geographically diverse managed care database representing approximately 26.8 million members Three study cohorts the severely ill, those with a chronic disease, and representative plan population were identified between 1 July 2 and 31 August 24 and evaluated through 31 August 25. The disease states that were assigned to each of these cohorts were selected after consultation with employer group and health plan personnel for economic comparisons and not clinical comparisons. They were selected because it was felt that they were of greatest clinical and economic concern and that management interventions might be possible. Severe illness diagnoses were defined as diseases that are progressive in nature and that commonly use biologics in their treatment. The chronic disease diagnoses we studied were chosen to represent more common conditions that have often been the target of various disease management and wellness programs by health plans and employers. Cohort identification. Outcomes were evaluated from the date of initial plan enrollment to the first disenrollment from the health plan or the end of available data. Therefore, the index date was identified as the date of the member s first enrollment segment in the managed care database during the study period. Members who were enrolled prior to the study period were excluded, because their previous experience with the plan might influence subsequent behavior and cost outcomes. Members were further required to have at least ninety days of continuous health plan enrollmentandeligibilityandtobeages18 64at thetimeofenrollment.thatagerangewasselected so that a traditional commercial population could be studied and so that the issue of incomplete cost capture associated with Medicare supplemental insurance products could be eliminated. Continuous eligibility was defined as having no more than a thirty-day gap between health plan enrollment segments. Eligibility gaps ranging from zero to sixty days were also investigated and found to have no significant effect on cohort size and subsequent results. Severely ill plan members were identified as those having two or more medical claims with an International Classification of Disease, Ninth Revision (ICD-9), diagnosis code for at least one HEALTH AFFAIRS ~ Volume 27, Number Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

3 of the following conditions: breast cancer, gastrointestinal cancer, lung cancer, non- Hodgkin s lymphoma (NHL), chronic kidney disease (CKD), rheumatoid arthritis (RA), multiple sclerosis (MS), and hepatitis C. Since reimbursement is shared between private and government payers as CKD patients continue on dialysis, CKD patients who had a history of dialysis procedures prior to their diagnosis were excluded, to avoid incomplete capture of costs. A total of 225,285 members with severe conditions were initially identified in the claims database for the study period. After members not ages 18 64, not meeting continuous eligibility criteria, and with initial enrollment segments starting after the beginning of the study period were excluded, the final severely ill cohort totaled 45,77 members. The largest drop was the result of members having eligibility that started prior to the study period. Members in the chronic disease cohort were identified as having two or more medical claims with an ICD-9 diagnosis code for at least one of the following conditions: diabetes mellitus, coronary heart disease, chronic obstructive pulmonary disease (COPD), cerebrovascular accident (CVA), and asthma. A total of 22,583 members were identified as having at least one of these conditions and meeting all of the exclusion criteria as described for the severely ill cohort. Because this cohort was much larger than the severely ill cohort and a random sample would need to be selected for the representative plan population, it was decided to randomly select a sample for the chronic disease cohort that was approximate in size to the severely ill cohort. Therefore, the chronic disease cohort used in the analysis totaled 4,59 members. Lastly, a representative sample of the managed care database was created by randomly identifying and selecting members who met the exclusion criteria outlined above. After the study criteria were applied, the representative plan cohort consisted of 39,825 members. Cohorts were identified and constructed so that each cohort is exclusive of one another. The representative plan population does includememberswhohavesevereorchronic conditions. However, the specific members with these conditions were not retained as part of those two cohorts, removing the possibility that costs are counted twice for the same member. Since the primary outcome of interesttothisstudywasthetotalmedicalspending by plan and members, we decided that if a plan member was identified as having both a severe illness and a chronic condition during the study period, the member would be placed within the severely ill cohort. Analysis of outcomes. Univariate analyses of frequencies and means were performed to describe the study cohorts. Once plan members were placed into their respective cohorts, their total health care use and costs were evaluated from the claims database. care costs represented all claims for these members. Cost outcomes are delineated and reported according to amounts paid by the health plan andamountspaidoutofpocketbymembers. Out-of-pocket values include coinsurance, deductibles, copayments, and other associated fees incurred by members and reported through the managed care claims database. Insurance premiums were not captured in the database, so they were not included in the calculation of out-of-pocket costs. Costs were further defined as either medical services costs or medication costs. The former were stratified into hospitalization and ambulatory costs. Hospitalization costs were defined as inpatient hospitalizations and emergency department (ED) visits. Ambulatory costs were defined as physician visits, outpatient hospital care, and other laboratory/ambulatory services. It is important to note that medication costs were captured from both the medical and the pharmacy claims and reported out separately. care cost values within a calendar year were adjusted to 25 U.S. dollar figures using the medical services and the medical care Consumer Price Index (CPI), where appropriate. 5 The minimum observation period for evaluating health care costs among the cohorts was three months, with a maximum ob- 826 May/June 28 Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

4 servation period of more than forty-eight months. The decision to use a relatively short minimum observation period was based on the premisethatsomememberswithsevereillnesses may disenroll from the health plan more quickly than those in the other cohorts as a result of death. Therefore, we decided that it was appropriate to include those members since their costs during that time period are representative of that cohort. However, as Exhibit 1 shows, the mean follow-up time for the cohorts was 25.5 to 3.4 months. All health care cost values were annualized by dividing each member s observed total cost values by the member s enrollment length in years. Members who fell within the top 1 percent and 2.5 percent of health plan spending were analyzed to better understand the magnitude of the financial impact of these conditions and to explore the cost drivers for those members with the greatest expenses. The cost thresholds were determined by using the 9 ($5,968) and 97.5 ($16,31) percentiles for total plan-paid costs for the representative plan population. EXHIBIT 1 Characteristics Of Plan Members, By Study Cohort Severely ill Number (N = 45,77) Percent Chronic disease Number (N = 4,59) Percent Representative plan population Number (N = 39,825) Percent Sex Male Female 16,266 29, ,188 2, ,95 22, Age group at enrollment ,59 21,83 21, ,591 18,547 16, ,82 19,775 6, Member Primary beneficiary Dependent/spouse 31,272 14, ,232 12, ,217 11, Follow time (months) Geographic region West Midwest South 32,637 2,855 1, ,838 3,31 8, ,746 2,776 8, Diseases of severely ill Breast cancer Lung cancer GI cancer NHL Rheumatoid arthritis Multiple sclerosis Hepatitis C Kidney disease 11,184 3,581 5,622 2,354 8,96 3,85 6,243 9, Chronic diseases Diabetes mellitus Coronary artery disease Cerebrovascular accident COPD Asthma 6,194 3,848 1,253 3,74 3, ,612 6,689 1,584 4,128 12, , , NOTES: Percentages were rounded to the first decimal place. GI is gastrointestinal. NHL is non-hodgkin s lymphoma. COPD is chronic obstructive pulmonary disease. a Not applicable. HEALTH AFFAIRS ~ Volume 27, Number Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

5 Study Results A total of 45,77 members were identified with one of the defining severe illnesses. This represented.8 percent of the overall populationthatwouldhavebeeneligibleforcohort inclusion based on the exclusion criteria (age, continuous eligibility, and time of initial enrollment during the study period). The severely ill group had a greater proportion of females compared to the other groups and was older. Gender differences were most likely due to the selection of disease states more common in women in the severely ill group, such as breast cancer. Regarding age, it is worthwhile to note the magnitude of people in the younger age categories, as many of these diseases are typically considered to occur in the elderly. Also, almost 7 percent of members with a severe illness were the primary beneficiary/ employee. The four oncology diagnoses accounted for almost 5 percent of the severely ill group, while diabetes was the most common of the chronic diseases (Exhibit 1). plan and member costs. Exhibit 2 displays the breakdown of the health plan and member costs for each of the study groups. For the severely ill group, the mean total annualized costs ($29,273) were more than three times greater than those observed in the chronic disease group ($8,225) and more than nine times greater than the overall plan population ($3,75). In all of the study groups, medical services costs were the major cost driver: they contributed three to five times the cost of medications. On average, hospitalization costs were approximately 5 percent of the medical services cost. Exhibits 3 and 4 display the total annualized costs for each disease state categorized by overall, medical services, and medication costs. We found that members contributed much more when medications were covered under the pharmacy benefit as opposed to the medical benefit. In the case of the severely ill members, even when the health plan paid more for medications under the medical benefit than under the pharmacy benefit ($3,421 versus $2,639), members out-of-pocket spending was still greater when the medications were covered under the pharmacy benefit ($655), compared with $123 when they were covered under the medical benefit. For the severely ill cohort, members incurred 19 percent of the EXHIBIT 2 Plan And Member Costs In 25 U.S. Dollars, By Study Cohort Severely ill Chronic disease Representative plan population Mean Median Mean Median Mean Median Total annualized costs $29,273 $9,3 $8,225 $3,321 $3,75 $989 Total plan-paid costs Medical services costs Hospitalization costs Ambulatory costs Medication costs Pharmacy benefit Medical benefit Total member-paid costs Medical services costs Hospitalization costs Ambulatory costs Medication costs Pharmacy benefit Medical benefit 27,196 21,136 11,627 9,59 7,64 4, ,195 7,31 5,581 3,683 1,898 2,47 1, ,55 2,1 1, ,6 2,639 3,421 1, ,45 1, ,77 1, , , May/June 28 Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

6 EXHIBIT 3 Total Annualized Spending, Medical Services Spending, And Medication Spending For The Severely Ill Cohort, In 25 U.S. Dollars, By Disease Diagnosis Thousands of dollars Total spending Medical serv. spending Medication spending Lung cancer GI cancer NHL CKD Breast cancer MS Hepatitis C RA NOTES: GI is gastrointestinal. NHL is non-hodgkin s lymphoma. CKD is chronic kidney disease. MS is multiple sclerosis. RA is rheumatoid arthritis. cost burden under the pharmacy benefit, while incurring only 3 percent of the medication spending within the medical benefit. Top 1 percent of health plan costs. Exhibit 5 shows the percentage of members who fall within the top 1 percent and 2.5 percent of health plan spending. All of the severely ill conditions were associated with at least a 5 percent chance of being in the top 1 percent of plan spending. Exhibit 6 shows the health plan and member costs associated with the severely ill members in the top 1 percent, broken down by quartile within the top 1 percent, as well as the utilization and cost data for biologics. Consistent with the overall groups, the largest component of health plan and total annualized costs were from medical services. The amount of variability in member-paid costs within the top 1 percent is large, with the costs ranging from $1,344 in the bottom quartile to $3,92 in the highest quartile. Also consistent with the overall groups, the member costs associated with medications were much greater when the EXHIBIT 4 Total Annualized Spending, Medical Services Spending, And Medication Spending For The Chronic Disease Cohort, In 25 U.S. Dollars, By Disease Diagnosis Thousands of dollars 2 Total spending Medical serv. spending Medication spending Cerebrovascular accident Coronary artery disease COPD Diabetes mellitus Asthma NOTE: COPD is chronic obstructive pulmonary disease. HEALTH AFFAIRS ~ Volume 27, Number Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

7 EXHIBIT 5 Percentage Of Plan Members Who Fall Within The Top 1 Percent And 2.5 Percent Of Plan Spending, By Disease State, 25 Total number of patients Patients within top 1% Patients within top 2.5% Number Percent Number Percent Diseases of severely ill Breast cancer Lung cancer GI cancer NHL Rheumatoid arthritis Multiple sclerosis Hepatitis C Kidney disease Chronic diseases Diabetes mellitus Coronary artery disease Cerebrovascular accident COPD Asthma 11,184 3,581 5,622 2,354 8,96 3,85 6,243 9,136 7,73 3,178 4,64 1,727 4,3 2,277 3,43 5, ,819 2,667 3,543 1,187 1,799 1,55 1,416 3, ,612 6,689 1,584 4,128 12,59 5,87 3, ,731 3, ,766 1, NOTES: Percentages were rounded to the first decimal place. GI is gastrointestinal. NHL is non-hodgkin s lymphoma. COPD is chronic obstructive pulmonary disease. drug was paid for using the pharmacy benefit versus the medical benefit. Use of biologic medications ranged from 1. percent in the bottom quartile of the top 1 percent cohort to 45.2 percent in the highest quartile. In the highest quartile, where biologic use was most frequent, biologics accounted for only 6.6 percent of the overall plan spending. Moreover, biologics accounted for approximately 31.7 percent of the total plan spending on all inpatient and ambulatory medications among members in the highest quartile. Discussion Understanding how health care dollars are spent, and their specific distribution among all members of a health plan, is vitally important if one is going to attempt to put programs in place to maximize the appropriateness of the dollars spent. Our study used administrative claims data to examine the health plan and member costs associated with the management of the severely ill and the use of biologic medications. We also determined the cost patterns associated with some more common chronic diseases and the general population to serve as a comparison. Employers stake in study population. In defining the seriously ill population, our study showed that these diseases are not reserved for just the retired elderly. All of the members we studied were under age sixty-five, and more than half of our severely ill members were under age fifty. More than two-thirds of our population was the primary beneficiary, meaning that they were employed for at least some of the time during their illness. This is extremely important information for employers that traditionally offer health insurance as a benefit to their employees. One of the reasons for sponsoring this benefit is that employees can then receive health care that allows them to work and be as productive as possible, in addition to recruitment and retention of a valued workforce. Therefore, employers have a large stake in how well these diseases are treated and should consider these demographic trends when determining the structure of their benefits and the availability of wellness, screening, and disease management programs for employees that may prevent or decrease the risk for these costly conditions. 83 May/June 28 Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

8 EXHIBIT 6 Plan And Member Costs For The Severely Ill Who Are In The Top 1 Percent Of Plan Spending, By Quartile, In 25 U.S. Dollars Bottom quartile (1% to 7.5%) Lower quartile (7.5% to 5.%) Upper quartile (5.% to 2.5%) Highest quartile (Top 2.5%) Mean Median Mean Median Mean Median Mean Median Total annualized costs $6,698 $6,685 $8,686 $8,633 $12,793 $12,683 $71,634 $3,38 Total plan-paid costs Medical services costs Hospitalization costs Ambulatory costs 5,353 3,519 1,9 2,429 5,42 3, ,288 7,112 4,73 1,59 3,14 7,156 4, ,11 1,874 6,877 2,61 4,276 1,783 7, ,672 67,714 53,688 3,718 22,97 36,52 27,393 8,9 13,97 Medication costs Pharmacy benefit Medical benefit 1,834 1, ,465 1,236 2,382 2, ,796 1, ,997 3, ,834 1, ,26 4,771 9,255 8,28 1,852 1,342 Total member-paid costs Medical services costs Hospitalization costs Ambulatory costs 1, , , , ,919 1, , ,92 2,648 1,22 1,626 2,759 1, Medication costs Pharmacy benefit Medical benefit , Cost of biologics Plan-paid costs Member-paid costs , Percent Percent Percent Percent Member spending on medications Medical benefit Pharmacy benefit Use of biologics Percent of total plan spending on biologics Percent of total plan medication spending on biologics Primary drivers of costs. In the severely ill population, medical services costs accounted for more than 75 percent of the health plan costs. Hospitalization costs were almost double the entire cost of medications. These data show that medication costs do not seem to be the drivers of health care costs for these members. However, it should be noted that some of the medical services costs are incurred during the infusion of medications. 6 In terms of the spending on medications, the severely ill were unique compared to those with other chronic diseases and health plan members in general, in that more than half of the costs were paid from the medical benefit. This is important since traditionally the cost burden for the member receiving medications is less when paidforunderthemedicalbenefitasopposed to the pharmacy benefit. Our study found that as well, with the severely ill having out-ofpocket costs five times greater for medications paidforunderthepharmacybenefitasopposed to the medical benefit. Out-of-pocket burden. We also explored the out-of-pocket burden that members faced in the form of copays, coinsurance, and deductibles during this time frame. Members with severe illnesses faced an average out- HEALTH AFFAIRS ~ Volume 27, Number Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

9 of-pocket burden of more than $2, per year almost four times that of the general health plan population. This increased to just under$4,peryearforseverelyillmembers in the top 2.5 percent. The magnitude of this out-of-pocket burden may lead to detrimental clinical and economic outcomes and should be examined more closely as out-of-pocket burden continues to increase. John Hsu and colleagues studied the clinical and economic outcomes associated with caps on Medicare drug benefits, comparing outcomes between a group of beneficiaries whose annual drug benefits were capped at $1, and a group that had unlimited benefits. Although there were significantly less pharmacy costs associated with beneficiaries with an upper-limitof-coverage cap, total medical spending was equivalent between the groups. Beneficiaries with coverage caps had poorer clinical outcomes, manifested by increased rates of ED visits, nonelective hospitalizations, anddeath,comparedtothosewithnocap.in patients with dyslipidemia and diabetes, beneficiaries with a cap had poorer adherence to medication and poorer control of lipid and glucose levels. 7 Dana Goldman and colleagues examined varying copays based on clinical status of patients on cholesterol-lowering therapy and found that the increased compliance with therapybythehighest-riskpatientsresulted in tremendous cost savings. 8 Benefit design. Inthetimeframeofour study, members paid much less for medications covered under the medical benefit than under the pharmacy benefit. As health plans develop strategies to decrease the annual rise in medical costs, some have focused on shifting the coverage of biologic medications from the medical benefit to the pharmacy benefit. Although moving coverage of biologics under the pharmacy benefit does have advantages (better data and transparency via use of NDC versus J-codes, improved clinical management, and effective utilization controls), this switch The cost of biologics was not insignificant from an absolutedollar perspective, but biologics were clearly not the driver of costs. may lead to a great increase in the out-ofpocket burden for members using these products if the historical cost-sharing dynamics continue as observed in our study. As benefit design changes are implemented, studies should be performed using a combination of administrative claims and medical record data to assess their impact on clinical and economic outcomes. Goldman and colleagues studied the effect of benefit design on specialty drug use for severely ill populations with limited treatment alternatives. 9 They found that increased cost sharing would not reduce the use of these drugs, since specialty drug use was largely insensitive to cost sharing at current outof-pocket levels in which these medications are paid for primarily under the medical benefit, especially compared to traditional pharmaceuticals. They concluded that health plans should focus on making sure that the appropriate members receive these medications. Once these members are identified, limiting coverage or placing barriers to access based on excessive copay contributions makes little economic sense. However, numerous studies have found a relationship between increasing copays and decreased patient compliance. 1 If benefit design changes are implemented and out-of-pocket burden for biologic medications rises as steeply as we observed when comparing medications paid for under the medical versus the pharmacy benefit to a point where compliance is compromised, members may not get the full benefit of therapy. This may result in the worst scenario for payers, where they bear much of the cost for biologic therapies but have a compromised effect as a result of suboptimal compliance. For other therapies paid for through the pharmacy benefit, a relationship between high levels of medication compliance has been associated with lower disease-related medical costs. 11 Severely versus chronically ill. In the health plan populations represented in our 832 May/June 28 Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

10 study, the severely ill cost on average three times more than those with other chronic diseases and nine times more than the general plan population. However, the severely ill made up only a small percentage of the population (.8 percent). So while health care costs for the severely ill are high, they can be spread over the much larger and relatively healthier plan population, to achieve the basic principles of insurance and population-based risk management. Since biologic medications are part of the treatmentapproachfortheseseverelyilldiagnoses, and they are generally expensive on a unit basis, we explored their contribution to the health plan spending for these members. Even in members in the top 2.5 percent of overall costs to the health plan, fewer than half used biologics. Biologics represented only 6.6 percent of the total costs paid by the plan and less than one-third of the overall medication costs. At an average cost of $4,449 per severely illmemberinthetop2.5percentofhealth plan paid costs, the cost of biologics was not insignificant from an absolute-dollar perspective, but biologics were clearly not the driver of costs in these members. Study limitations. Although this study provides valuable quantitative information for policymakers regarding this population, several limitations warrant consideration when interpreting the results. The study was performed within a commercially insured population, which may limit the generalizability of the results to other populations because of differing health care benefit designs or age (such as Medicaid, Medicare, or the uninsured). In addition, we used a relatively short minimum continuous health plan enrollment period (ninety days) after diagnosis for inclusion into the study. Although the mean follow-up time was between two and two and a half years in each of the study cohorts (with a maximum of forty-eight months), the abbreviated followup time may have influenced the results. However, the shortest follow-up was observed in the representative plan population, and all costs were annualized to account for varying follow-up time. Finally, this analysis was retrospective in design and therefore does not allow for the determination of causality in the findings. Our study attempts to provide payers and employers with greater detail around the costs associated with the treatment of the severely ill. Our results showed that a large percentage of the severely ill, commercially insured population are covered employees and that they are relatively young. The main cost driver for these members is medical services costs specifically, costs associated with hospitalization. Biologic medications were not a major driver of costs for these members, even those with the greatest use of biologics and the highest health plan costs. Out-of-pocket burden for members increased dramatically as health plan costs rose. plans and employers should explore the overall clinical and economic effects of increasing out-of-pocket burden for members to obtain therapies such as biologic medications, to make sure that policies aimed at controlling medical spending in one specific area do not have unintended negative consequences in another. These negative consequences could ultimately lead to suboptimal outcomes and even greater costs overall. The authors thank Marcus Wilson for his helpful comments on the manuscript. Portions of these data were presented at the Sixth Annual Forum on Patient Adherence, Compliance, and Persistency in Philadelphia, Pennsylvania, 23 April 27. This research was supported by funding from Amgen Inc. HEALTH AFFAIRS ~ Volume 27, Number Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.

11 NOTES 1. P. Fishman et al., Chronic Care Costs in Managed Care, Affairs 16, no. 3 (1997): ; R.I. Garis and K.C. Farmer, Examining Costs of Chronic Conditions in a Medicaid Population, Managed Care 11,no.8(22):43 5;C.Hoffman, D. Rice, and H.Y. Sung, Persons with Chronic Conditions: Their Prevalence and Costs, Journal of the American Medical Association 276, no. 18 (1996): ; and W. Yu et al., The Relationships among Age, Chronic Conditions, and care Costs, American Journal of Managed Care 1, no. 12 (24): Fishman et al., Chronic Care Costs in Managed Care ; and Yu et al., The Relationships. 3. H.T. Tu, Rising Costs, Medical Debt, and Chronic Conditions, Issue Brief no. 88 (Washington: Center for Studying System Change, 24). 4. Henry J. Kaiser Family Foundation, Illustrating the Potential Impacts of Adverse Selection on Insurance Costs in Consumer Choice Models, November 26, insurance/snapshot/chcm1116oth2.cfm (accessed 3 January 28). 5. Bureau of Labor Statistics, Measuring Price Change for Medical Care in the CPI, 1 March 27, (accessed 13 September 27). 6. The medical services costs associated with infusion of medications are typically a small percentage of the acquisition costs for biologics. For example, Wu and colleagues found that the annual drug administration cost for infliximab, a biologic medication infused in the outpatient setting for the treatment of rheumatoid arthritis, was $725 and represented 4 percent of the annual cost of the drug. E. Wu et al., Cost of Care for Patients with Rheumatoid Arthritis Receiving TNF-Antagonist Therapy Using Claims Data, Current Medical Research and Opinion 23, no. 8 (27): In addition, changes in payment methodologies to physicians for drugs administered in their offices have greatly decreased reimbursement for the drug product. This may affect patient care if physicians decide not to administer drugs in their offices and refer patients to outpatient infusion centers for treatment; research on the influence of these reimbursement changes on the quality of care is needed. D. Stern and D. Reissman, Specialty Pharmacy Cost Management Strategies of Private Care Payers, Journal of Managed Care Pharmacy 12, no. 9 (26): J. Hsu et al., Unintended Consequences of Caps on Medicare Drug Benefits, New England Journal of Medicine 354, no. 22 (26): D.P. Goldman, G.F. Joyce, and P. Karaca-Mandic, Varying Pharmacy Benefits with Clinical Status: TheCaseofCholesterol-LoweringTherapy, American Journal of Managed Care 12, no. 1 (26): D.P. Goldman et al., Benefit Design and Specialty Drug Use, Affairs 25, no. 5 (26): Goldman et al., Varying Pharmacy Benefits ; D.P. Goldman et al., Pharmacy Benefits and the Use of Drugs by the Chronically Ill, Journal of the American Medical Association 291, no. 19 (24): ; and C.W. Tseng et al., Cost-Lowering Strategies Used by Medicare Beneficiaries Who Exceed Drug Benefit Caps and Have a Gap in Drug Coverage, Journal of the American Medical Association 292, no. 8 (24): M.C. Sokol et al., Impact of Medication Adherence on Hospitalization Risk and care Cost, Medical Care 43, no. 6 (25): May/June 28 Downloaded from Affairs.org on April 18, 219. Copyright Project HOPE The People-to-People Foundation, Inc. For personal use only. All rights reserved. Reuse permissions at Affairs.org.