To obtain information on migraine and its impact on the family, particularly on the healthcare utilization and productivity of family members without migraine. 

Retrospective study using linked medical and pharmacy claims data that allowed identification of families and individuals with migraine. 

Families with at least 1 migraineur were matched with up to 3 nonmigraineur families on employer, age of migraineur, number of family members, sex, and index date quarter. For a subset of employees, data on short-and long-term disability as well as absenteeism also were evaluated to determine the impact of migraine on these indirect costs. 

The total healthcare costs of a family with a migraineur were 70% higher than those of the nonmigraine family, with most of the difference concentrated in outpatient costs. The impact of having a migraineur in the family persisted even after controlling for other disorders and demographics in a multivariate model. Total healthcare costs per family depended on which family member was the migraineur, and were about $600 higher when the sole migraineur was a child versus a parent and almost $2500 higher when both a parent and child were affected (compared with families in which the sole migraineur was a child). Work absence days, short-term disability, and workmanâ€™s compensation days all were higher among migraine families than among families without a migraineur. 

Migraine families incur far higher direct and indirect healthcare costs than nonmigraine families, with variation depending on which family member is the clinically detected migraineur. 

Migraine is a common disorder, affecting about 18% of women and 6% of men in the United States.

It is characterized by recurrent episodes of disability that impact quality of life,

and add substantially to direct medical costs.

Migraine prevalence is highest between the ages of 25 and 55 years, corresponding with the most productive years for individuals, and places a substantial burden on employers in terms of lost or diminished productivity and increased healthcare costs. 

The impact of migraine and its treatment on productivity has been reviewed in detail elsewhere,4 and although we can speculate that decreases in productivity at work also would be reflected at home, that research has not been undertaken to date. Although there is an extensive body of literature on pain and the family, it contains few data on migraine and its impact on the family. Most of the available research on the impact of migraine on the family focuses on the likelihood of time lost by migraineurs from family activities.

In one study, 73.6% of migraineurs claimed that their headache adversely affects their lifestyle through interference with family relations, work attendance, and work efficiency.

Among a national sample of migraine sufferers, 61% believed that their families were significantly affected by their migraines, including delay in accomplishing household duties and activities with children (62%) and spouses (69%); a substantial proportion reported cancellation of pleasure trips, celebrations, and social activities.

Most striking was that 5% reported that their migraines caused their divorce or separation. Most stated that their families were understanding, but a considerable number reported that family members took a negative attitude. This was particularly so with young children deprived of parental care during their parent's migraine attacks. Spousal relationships also suffered, and frequency and quality of sexual relationships were adversely affected by migraine. 

If one looks at the family as a production unit, it is possible to determine the impact of the child's headaches on his or her own production as well as that of the parents. Furthermore, the impact of the parent's headache on the "production" of the child may be determined. Statistically and clinically, we can determine what impact, if any, migraine has on the rest of the family. We undertook this analysis using a data source that allowed us to link families, identify migraineurs, and analyze medical claims (healthcare utilization) to address the question of the impact of migraine on the family, particularly on the healthcare utilization and productivity of the nonmigraine family members. Additional data were used for a subset of employees for whom we also had short- and long-term disability and absenteeism data. 

The analytic files were derived from 2 MarketScan research databases: the Commercial Claims and Encounters Database and the Health and Productivity Database. The Commercial Claims and Encounters Database, the largest of the MarketScan research databases, reflects the combined healthcare service use of more than 3.5 million privately insured individuals (in the 1999 database), covered under a variety of fee-for-service (FFS), fully capitated, and partially capitated health plans. The Health and Productivity Database was used to assess indirect costs. This database includes information on 1997-1999 medical care and pharmacy utilization for 6 large US employers, with data on workplace absence, short-term disability, and workman's compensation for a subset of the employers. The database contains person-level information on more than 800 000 individuals, including approximately 350 000 employees. Data on absenteeism, short-term disability, and workers' compensation varies by year. For example, absence data is available for approximately 15 000 to 16 000 workers in 1997-1998, and 112 000 persons in 1999. Data on short-term disability and workman's compensation is more complete than absence data. Short-term disability data are available for approximately 227 000 to 275 000 workers, and workman's' compensation data are available for 227 000 to 242 000 workers. These differences stem from the fact that absenteeism records tend to be kept in noncentralized locations at regional offices of large employers. 

The dollar amounts reported are payments inclusive of cost-sharing (copayments, deductibles, etc). Because the individuals captured in the Health and Productivity Database database are a proper subset of those in the full MarketScan files, we linked the absenteeism, shortterm disability, and workers' compensation data to the corresponding subset of episodes in the MarketScan files. This enabled us to investigate the covariation of Health and Productivity Database metrics with traditional measures of healthcare utilization captured in standard claims databases. 

The analytic file contains a mix of patients with FFS health plans and those with partially or fully capitated plans. In the economic analyses, we examined healthcare utilization using information on service utilization (eg, outpatient visits, emergency room use, hospitalizations) that is available from both the FFS and encounter records. For the capitated group, the value of patients' service utilization was priced using average payments for corresponding procedure codes from the FFS sample. We imputed the costs in the encounter plans using costs from the FFS plans. Average payments by Current Procedural Terminology (CPT) procedure code from the FFS plans were calculated and assigned, by CPT code, to the services in the encounter data; this was undertaken by region and year. This analysis takes the perspective of the employer/payer. 

International Classification of Diseases, Ninth Revision (ICD-9-CM)

A migraine case was defined as any subject having an 

 diagnosis code for migraine (346.xx), or a pharmaceutical claim for an ergot, triptan, or isometheptene (Midrin). We evaluated the possibility of using other headache codes (307.81 and 784.0), but thought that they would lack sensitivity (even though code 784 includes vascular headache, according to some 

 coding manuals). The remaining medication classes were not specific enough to migraine sufferers. 

Nonmigraineurs were matched to migraineurs in up to a 3:1 ratio on employer, age (5 year bands), number of family members, sex, and index date quarter. Enrollees were randomly selected within each stratum. Because of the sampling with no replacement requirement, it was not always possible to achieve the target 3:1 ratio for each stratum, but this ratio was achieved in most instances. The exceptions were higher ratios for pediatric enrollees and lower ratios at the tail end of the age distribution. The control group does not contain migraine patients, but other headache patients who may have not met our criteria for migraine were eligible for inclusion. 

International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)

To construct the analytic file, all inpatient, outpatient, and pharmaceutical claims for the first observed 

 code and/or prescription medication indicating migraine during the period January 1, 1997, through December 31, 1999, were searched. The index date is the first date on which a patient qualifies as a migraine case by the definition above. For those who may have a medication from the list first and a diagnosis sometime later, the index date was that of the first prescription. 

The medical costs for patients in the HMO/capitated plans were imputed by using the FFS data from MarketScan. We calculated average payments by CPT procedure, year, and region, and then assigned the imputed costs to each individual service in the encounter portion of the project data, again by CPT procedure, year, and region. The costs were annualized for reporting. For migraine patients, only postindex costs are reported and were considered migraine related if a visit was associated with a primary diagnosis code of 346.xx; only prescriptions for an ergot, triptan, or isometheptene were considered migraine related. For the control group (matched sample), costs were reported for enrollment during the 1997-1999 study period. We used total healthcare costs as the basis of comparison with matched nonmigraine families, because total costs show the incremental differences in healthcare costs that can be due to migraine itself and to its influence on comorbidities. Multivariate analysis (ordinary least squares of logged total expenditures) was used to test the effect of other family demographic and healthcare expenditures on the impact of migraine on total healthcare expenditures. Logged expenditures were used in the model to reduce the skewness in the expenditure distribution. 

A family-level file containing the inpatient, outpatient, and pharmaceutical claims for all individuals in families containing at least 1 migraineur was created. A sample of nonmigraine families (ratio of 3:1) that were matched on number of family members, plan, and geography was used as a comparator group. Members of the same family were identified by eligibility codes that also distinguish employee from dependent. Family position was determined from eligibility files that included both demographic information (age, sex) and relationship to the policy beneficiary. 

Major Diagnosis Categories, a Centers for Medicare & Medicaid Services classification system that groups diagnosis codes into 24 major categories, were used as proxies for extent of medical and mental health comorbidities. Similarly, Charlson Comorbidity Index scores were generated.

For the family cost measures, we identified the earliest migraine index date within the family. All services after that date (up through December 31, 1999) were reported for all family members. If there was another migraineur in the family with a later index date, that index date was ignored for the purposes of the family analysis. Simple descriptive statistics were used to characterize the cohorts under study. 

We identified 73 094 families with at least 1 migraineur per family (

). The vast majority of the families (96%) had only 1 migraineur identified and the majority of families had more than 3 members (57%). About 20% were single-person households. The likelihood of triptan use increased with the number of migraineurs in the family, going from 36% in families with 1 migraineur to 74% in those families with 3 or more migraineurs. This pattern was repeated for the other drug classes as well. Similarly, both the mean number of Major Diagnosis Categories and the Charlson Comorbidity Index increased with an increasing number of migraineurs in a family. 

The total healthcare costs of a family with a migraineur were 70% higher than those of the nonmigraine family (

), with most of the difference concentrated in outpatient and pharmacy costs. Pharmacy costs showed the greatest discrepancy; they were more than twofold higher in migraine families than in nonmigraine families. However, pharmacy costs accounted for 20% of total healthcare costs in the migraine families versus 15% in the nonmigraine families. Outpatient costs also were 80% higher for migraine families than for nonmigraine families, but accounted for just over 50% of the total costs in each group. 

Total healthcare costs per family depended on which family member was the migraineur (

). Total healthcare costs for the family were about $600 higher when the sole migraineur was a child versus a parent and almost $2500 higher when both a parent and child were affected (2.5% of migraine families) (

). We also examined the impact of migraine on other family members who apparently did not have migraine (Table 3). As is shown above, the total healthcare costs per family were 70% higher in the migraine families than in the matched nonmigraine families (Figure 1). The differential was even higher (90%) among families with both parent and child migraineurs. Homes with just a child migraineur incurred more total healthcare costs than homes with just an adult migraineur, but the difference was not striking. Adult family members of migraineurs had more than twice the total healthcare costs than the sibling or child of a migraineur ($3500 vs $1400), while the migraineurs themselves incurred more than $6400 in total healthcare costs (

). All of these estimates were far higher than the mean total healthcare costs (weighted average) of the matched individuals in the original cohort ($2691). 

Among migraineurs, the proportion of total healthcare costs that could be attributed to migraine was 9%, while between 5% and 7% of the total healthcare costs in migraine families were attributable to migraine. The impact of the migraineur on the rest of the family's total healthcare costs also could be approximated, as the spouse of a migraineur had 24% higher total healthcare costs and the child of a migraineur had 11% higher total healthcare costs compared with their nonmigraine family counterparts. Both the nonmigraine parent of a migraineur (26% increase) and specifically the mother of a child migraineur (25% increase) also were substantially affected when compared with parents in nonmigraine families. 

 shows that absence days, short-term disability, and workman's compensation days were higher among migraine families than among families without a migraineur (see also 

). These data were available for 7307 migraine families and 19 981 matched families. Migraine appeared to increase the number of annual work loss days caused by absence or short-term disability by 54% or more; however, it had less impact on workman's compensation. Short-term disability appeared to be particularly influenced by migraine status in the adult family members; the number of short-term disability days was 2.3 times higher than the number in the matched nonmigraine families. A 62% increase in short-term disability days (59% higher absenteeism) was evident when only a parent was the migraineur, but that increase was more than twofold higher when a parent and child were both migraineurs. Even in those families with just a child migraineur, there were 11% more short-term disability days and absenteeism was 35% higher than in nonmigraine families with children. The number of absence days in families with just a child migraineur also was higher, but was approximately the same as the number in families with both a parent and a child migraineur compared with nonmigraineur families. However, the employed parent in migraineur families lost, on average, 4 additional sick days per year, 5 additional short-term disability days, and 3 additional workman's compensation days compared with his/her nonmigraine counterpart. 

 reports the results of the multivariate analysis of the persistence of the effect of migraine on total healthcare expenditures after controlling for other family demographic and health status variables. The results indicate that, after controlling for family size, regional location, and medical comorbidities measured at the individual and family level, total healthcare expenditures were significantly higher for families containing a member suffering from migraine than in matched families without a migraineur. Collectively, these variables accounted for more than 46% of the variation in logged total healthcare expenditures. 

This is the first extensive study of migraine and its impact on family healthcare costs. Total healthcare costs are a valuable method of assessing the true impact of an illness, because migraine, like many disorders, is comorbid with a number of other illnesses that make it difficult to determine to what extent costs are driven by which illness. Ideally, effective treatment of 1 illness should offset the costs of comorbid or consequent illness; however, it would be difficult to determine whether this is the case from these data. 

This analysis provides new and interesting information regarding migraine and its impact on the family. The results suggest that migraine, even under treatment, is associated with a substantial increase in total healthcare costs for a family. The impact of having a migraineur in the family persisted even after controlling for other disorders and demographics in a multivariate model. These costs are disproportionately concentrated in outpatient and pharmacy costs. Costs were substantially higher when both parent and child were affected; however, adult migraineurs had a higher incremental cost (versus adult nonmigraineurs) than their pediatric counterparts. Similarly, first-degree relatives of a migraineur appeared to incur about 25% higher total healthcare costs. These differences could be due to the increased stress of the illness on other family members. These data are of more interest when one considers that these estimates may be conservative. There are a substantial number of subjects in the "nonmigraine" category (more than 100 000; data not shown) who suffer from severe headaches that did not meet our criteria for migraine. Our analysis compared migraineurs with those who did not meet our conservative criteria for migraine: an analysis comparing migraineurs to the total true nonheadache population is likely to show even more striking differences. 

A few limitations of the current study must be borne in mind. First and foremost is that we are examining the impact of clinically detected and treated migraine. Studies have shown that a substantial proportion of migraineurs do not seek care.

This would suggest that there were "undetected" migraine subjects in our comparison group who likely generated additional costs that we did not detect. The net effect would be to bias our results toward a conservative estimate of differences between groups. We used medical claims data as our source of medical care and charge information. Although intended to reflect actual practice, our identification of subjects and assignment of migraine status were strongly dependent on coding. It is possible that these data are distorted because of incentives for reimbursement; however, this has not been shown to be the case in other migraine studies. The proper diagnosis of migraine is part of the art and practice of medicine, and despite formal guidelines provided by the International Headache Society,

we were unable to validate the extent to which the subjects in this study met formal criteria. Research has shown a high positive predictive value of a migraine diagnosis in the clinical record and medical claims.

Our analysis of "migraine-related" costs (unpublished date, 2003) shows that migraine itself accounted for a very small proportion of all healthcare costs for the migraine subject, which suggests that migraine may be a marker for more systemic or comorbid illnesses. Our comorbidity analysis did show that comorbidity rates among migraineurs were higher than those among nonmigraineurs; however, in the multivariate model adjusting for comorbidities, the effect of having a migraineur in the family persisted, which provides compelling evidence that a migraineur in the family is a strong and independent predictor of higher family costs. Finally, our costing methodology annualized costs of care and may have distorted the results; it is unclear whether this would overestimate or underestimate costs. 

Our analysis of absenteeism, short-term disability, and worker's compensation days lost was in a subset of subjects for whom these data were captured. The strength of our findings lie in the objective capture of these data and the lack of reliance on self-report or attribution of impact to migraine. Although the results are striking, it is possible that the subset is in some way not representative of the general experience of migraineurs in an employer setting. There was a modest effect of migraine in other family members on absenteeism and short-term disability, as being the parent of a migraineur increased absenteeism days 28% and short-term disability days 11%. The magnitude of absenteeism in our current study was higher than that reported in previous studies from a managed care organization14 and was broadly consistent with results from more general US population samples.

We found far fewer work days lost than those reported in a clinical trial sample,

 due at least in part to the severity of migraine among the patients enrolled in the trial and to the theory that the employed population from which our data were derived are generally thought to be in better health than those not employed (the "healthy worker effect").

Given that these data probably reflect healthier migraineurs, these findings are that much more important, as past research has shown migraine to be associated with significant impact on workforce participation,

 which in turn could have significant effects on the family. 

Migraine is a common and costly illness that appears to have financial impact not only on the sufferers, but also on others in their families. This is evident for both direct and indirect costs. Migraine families incurred far higher direct and indirect costs than nonmigraine families, with the variation depending on which family member was the clinically detected migraineur. Further analyses need to be undertaken to continue to better understand the relationship between migraine, comorbidities, and their relative impact on families. 

From Galt Associates, Inc, Blue Bell, Pa, and the Department of Epidemiology, University of North Carolina School of Public Health, Chapel Hill, NC (PES); The MEDSTAT Group, Cambridge, Mass (WHC, RB); MLG Consulting Services, Inc, Chadds Ford, Pa (MLC); and AstraZeneca Inc, Wilmington, Del (RW). 

This study was funded by AstraZeneca Pharmaceuticals LP. 

Portions of this manuscript were presented at the 44th Annual Scientific Meeting of the American Headache Society, Seattle, Wash, June 21-23, 2002. 

Address correspondence to: Paul E. Stang, PhD, Galt Associates, Inc, 640 Sentry Parkway, Suite 305, Blue Bell, PA 19422. E-mail: 

1. Stewart WF, Lipton RB, Celentano DD, Reed ML.

 Prevalence of migraine headache in the United States. 

2. Lipton RB, Stewart WF, Diamond S, Diamond ML, Reed M.

 Prevalence and burden of migraine in the United States: data from the American Migraine Study II. 

. Evolution of the measurement of quality of life in migraine. 

4. Stang P, Cady R, Batenhorst A, Hoffman L.

 Workplace productivity: a review of the impact of migraine and its treatment. 

 Reduced labor force participation among primary care patients with headache. 

6. Fishman P, Von Korff M, Lozano P, Hecht J. 

Articles

CLINICAL

The Family Impact and Costs of Migraine

To examine actual use of formal benefit/cost and benefit/risk results in health system decision making by public and private healthcare organizations. 

A direct survey with questions about healthcare decisions made by the respondent or the respondentâ€™s organization. The scope of this survey precluded meaningful quantitative analysis, thus descriptive and qualitative analyses were performed. 

An initial questionnaire was tested in 2001 with 15 respondents in 4 countries. In 2002, a revised questionnaire was sent to a convenience sample of 116 individuals representing information users (providers, payers, and regulators) and information producers (technology firms and academics) in France, Sweden, the United Kingdom, and the United States. Responses were received from 104 people (89.7%). 

Every information user employed benefit/risk analyses to accept or reject new interventions and delete existing technologies. In addition, 42.1% of information users also used formal benefit/ cost results (cost effectiveness, cost benefit, and/or cost utility). Seven providers/payers in the United States, 1 in France, and 1 in the United Kingdom required such analyses, as did 1 UK regulator. Most did not produce their own analyses but relied on those of public organizations (eg, Food and Drug Administration, National Institute of Clinical Effectiveness), academics, and pharmaceutical firms. 

A surprisingly high percent of information users (42.1%) employed any formal economic cost-effectiveness, costbenefit, or cost utility analysis, CEA, CBA, or CUA evaluations in deciding whether to accept, pay for, or reject new interventions or to delete old interventions. Still, this figure was substantially higher than expected given the results of previous studies, nearly all of which found low use of formal benefit/risk and benefit/cost analyses. 

The few economic evaluations of health services prior to the mid-20th century focused on public health issues rather than on medical care services, the focus of most current evaluations.

Increasingly sophisticated questions, methods, and analyses aided impressive growth of formal economic evaluations (eg, cost effectiveness, cost benefit, cost utility) over the past 2 decades.

Such evaluations commonly occur with or after clinical and quality-of-life evaluations of health and medical technologies and population-based health improvement programs. However, it remains uncertain whether increased interest and increased numbers of formal benefit/cost analyses are accompanied by increased use of such research results in actual healthcare decision making. 

The limited evidence available suggests slow adoption of economic evaluations for use in decision making. The main reasons seem to be a disconnect in the vernaculars of producers and users and concerns about reliability of results, especially related to issues of study transparency, timeliness, and study relevance to individual organizations. The recent report by Drummond and colleagues clearly delineates the issues; it offers guidance and recommendations to help decision makers incorporate results of economic analyses in their policy-making processes.

One important concern is that results of the many completed benefit/cost evaluations are seldom used to help medical services delivery or health system decision making, and, when they are used, the application is often incorrect.

Many have lamented infrequent use of these results and, with rare exception, their negligible impact on health system and medical care decision making at national and local levels.

Multiple reasons have been suggested for low use of benefit/cost results, reasons similar to those given for the low use of clinical evidence and technology assessments.

Most of the above cited literature, though, is based primarily on opinions and perceptions of how people and organizations make healthcare decisions, from implementing new technology to paying for services to controlling expenditure growth.

However, there are few empirical data directly supporting or rejecting these opinions and quantifying directly the actual role, use, and direct effects of economic evaluations in healthcare decision making, either at national, individualsector, or health system level. The goal of this study was therefore to examine the extent to which economic evaluations are actually being used as an aid to healthcare decision making by public and private healthcare organizations. 

A questionnaire was developed that asked about use of specific benefit/risk and benefit/cost techniques, results, and decision outcomes, plus other benefit, risk, and economic data, in the acceptance or rejection of new or the deletion of existing interventions. Questions elicited specific examples in which formal economic evaluations had been used during 2001 and 2002 to decide whether to accept, pay for, or reject new interventions or to delete old diagnostic and treatment interventions previously approved and reimbursed. The instrument was tested with 15 experts in economic evaluation and policy. These experts were located in France, Sweden, the United Kingdom, and the United States, and all were known by the investigator. Their recommended changes to the questionnaire were incorporated, including additional and more detailed questions and the inclusion of benefit/risk questions. At the investigator's request, these 15 people recommended 116 potential respondents. The only criterion was that participants had to be decision makers knowledgeable about use of benefit/risk and benefit/cost results, or lack thereof, in their organizations. 

Two categories of organizations were surveyed: information users (providers, payers, and regulators) and information producers (technology firms and academics). With the widest range of participants from multiple health care sectors, this was the largest direct survey undertaken to date on the use of benefit/cost evaluations in healthcare decision making. The working hypothesis was that this survey would confirm results of previous studies of fewer participants. 

The questionnaire had 4 parts. First were questions about actual decisions made by the respondent to accept, reject, pay for, or delete new or existing technologies. A second set of questions asked about actual decisions made by the 

 to accept, reject, pay for, or delete new or existing interventions. The third part of the questionnaire requested examples of both types of decisions made during 2001 and 2002. The final section inquired about reasons for the decisions to accept, reject, pay for, or delete technologies; about changing indications for use of previously accepted technologies. 

Between May 1 and December 31, 2002, the revised survey was administered in person (26.9%), over the telephone (72.1%), and in 2 cases by e-mail (1.9%) to the population of 116 people in France, Sweden, the United Kingdom, and the United States. Responses were received from 104 participants (89.7%). Before this study, the investigator knew fewer than half the respondents personally. Only 3 organizations contributed more than 1 participant: 2 each from a European payer, a US regulator, and a US provider/payer. All who declined to participate did so solely because their organizations forbade response. 

The sample size and its breadth, wide range of responses, types of organizations recruited, and responsibilities of participants precluded meaningful quantitative analyses across organizations and countries. Consequently, analyses were mainly descriptive and qualitative. 

Respondents came from private and public sectors, for-profit and not-for-profit organizations (

). They were between 35 and 61 years old, and 17.3% (n = 18) were female. All were senior decision makers in their organizations. Information users who responded had substantial responsibility for accepting, rejecting, or deleting new or existing technology and interventions. In addition, 31.7% (n = 33) of total respondents (n = 104) were also responsible for payment decisions. Four (11.4%) of the 35 participating US providers/private payers were from private health insurance firms, 25 (71.4%) worked for combined insurance and service providers (eg, managed care organizations or MCOs), and the remaining 6 (17.1%) represented public insurers. Among French providers/payers, 1 was from a private insurance firm and 6 from public insurers/payers. Among public organizations such as the Food and Drug Administration (FDA), payment and acceptance decisions generally were made by groups with the organization at different times, and payment decisions were made after benefit/risk acceptance decisions.

All respondents confirmed that their organizations used benefit/risk results in essentially all decisions to accept, reject, pay for, or delete technology. However, a majority (86.5%) of all 104 respondents never produced their own formal analyses but relied on those of public (eg, the FDA) or private organizations (eg, pharmaceutical firms) and on published research results. Everyone acknowledged that their organizations used relative rather than absolute benefit/risk ratios (ie, comparisons among interventions with similar uses) in nearly every decision to accept or reject new technology, or to delete previously accepted interventions. 

Nearly all respondents (93.3%) found surgical operations the most difficult to evaluate, mainly because of lack of control trials establishing benefit and risk. Discontinuing existing operations found to be less effective or more dangerous than a new surgical procedure was equally difficult. Respondents were concerned particularly that new, often high-risk, and usually costly operations were introduced with little evaluation of benefits and risks or the economic effects on patients, health systems, and payers. 

With the exception of participating academic organizations, 2 US MCOs, and 1 Swedish regulatory agency, every organization that responded had written benefit/ risk criteria for adoption and rejection. None had specific benefit/risk thresholds for either acceptance or rejection; rather, acceptable ratios varied by severity of illness, treatment risk, and alternatives available. At the same time, nearly all providers/payers noted that once the national regulatory agency accepted a new technology for use, their organizations also accepted it, eventually if not immediately. Sildenifil was the most frequently cited example of a new technology accepted on the basis of benefit/risk evaluation but the use of which was delayed because of reluctance to pay. 

Every respondent noted that in most cases, although not that of pharmaceuticals, the information available to evaluate benefits and risks of new technologies varied greatly in quality and quantity. The main complaint was that data, primarily for medical devices and surgical operations, were drawn mostly from uncontrolled observational or technical studies, such as of hardware function; thus they provided no confident rates of quantifiable absolute or relative adverse outcomes or benefits. Nearly all providers and payers also noted that if a hardware technology (eg, a new type of noninvasive imaging device) appeared unlikely to cause harm and did what it was purported to do, then it was accepted. Payment, however, was not necessarily approved automatically. 

Examples of interventions rejected on the basis of benefit/risk results included many complementary and alternative medicine treatments, polymerase chain reaction screening of blood supply for hepatitis C, and electrical stimulation of hand and arm for post-stroke patients. The combination weight-loss treatment of fenfluramine and phentermine was the most frequently cited example of an accepted treatment subsequently delisted because of new benefit/risk results. 

In nearly all provider and payer organizations, decision makers had little or no direct responsibility for diffusion and actual use of an intervention in clinical practice after its acceptance. Ongoing evaluation was rare also, occurring only after severe adverse outcomes or the availability of new alternatives. 

Medications clearly were the easiest to evaluate for benefit/risk as pharmaceutical firms are responsible in study countries for such evaluations. All responding providers/payers were confident that they could nearly always accept benefit/risk results for prescription medications on behalf of their organizations. This is not to understate respondents' concerns that randomized-control efficacy trials are rarely large enough to define all important adverse effects. They also expressed concern that, given its widely recognized problems, post-marketing surveillance is a poor substitute for trials and systematic follow-up in defining long-term benefit and risk. 

Even with high benefit/risk profiles, the perception of low total benefits can influence decisions. For example, respondents from 2 US MCOs noted that gabapentin for peripheral neuropathy was not accepted onto their formularies because of insufficient benefit, even though the FDA had approved it for use. 

All respondents were concerned that new technology generally added to, not substituted for, existing diagnostic tools and treatments. They were also troubled by the uncertain health benefits but cost-increasing consequences of multiple interventions for the same diagnosis. Providers, payers, and regulators noted that deleting previously accepted technology posed difficulties different from those encountered when accepting new interventions. One difficulty was the disruptive personal effect of discontinuing the use of a technology or treatment that both physician and patient perceived as beneficial, especially when effective alternatives were unavailable. A recent example was the FDA decision to overturn their previous approval of alosetron for irritable bowel syndrome because of risk of intestinal blockage. Patient complaints, some US respondents claimed, forced the FDA to re-approve alosetron for use. 

A large majority (94.4%) of all respondents and their organizations used some form of economic evaluations in decision making (

). Restricting the definition to formal benefit/cost analyses brought the number of providers, payers, and regulators (n = 78) that regularly used formal benefit/cost analyses (most commonly costeffectiveness analyses) to 42.1% (

). Nearly all remaining members of these groups used such analyses occasionally. 

UK respondents noted that all new technology was required to have formal benefit/cost analyses prior to decisions on acceptance and payment. Formal benefit/ cost evaluation was mandatory for 18 private payers (including 7 MCOs) in the United States and 1 private insurer in France. Interestingly, pharmaceutical and device manufacturers (ie, technology firms) reported that they did not use their firms' benefit/cost analyses in pricing products, although they did use these results in sales and marketing efforts. 

Surprisingly, generalizability or transferability of results was not identified by any information user as an important issue. However, timeliness was clearly an important factor in benefit/ cost use; economic outcomes were rarely available when needed for decision making. As Wallace and colleagues found,

service providers and payers did not use benefit/cost results in developing clinical practice guidelines. No responding provider or payer organization had a predefined benefit/cost ratio by which it would accept, reject, pay for, or delete interventions. Only 1 US MCO based its decisions to accept, reject, or pay for a new technology mainly on formal benefit/cost (value) calculations. However, 92.3% of European respondents predicted that the European Union would soon require all member nations to provide formal benefit/cost evaluations, whether for pharmaceuticals alone or for all new healthcare technology. 

Respondents from the United Kingdom and elsewhere commented that the National Institute of Clinical Excellence (NICE) is the only public body in the 4 study countries currently responsible for both benefit/risk and benefit/cost evaluations. However, 2 British respondents pointed out that benefits still predominate, and expected benefit/cost ratios are used mainly to confirm benefit/risk decisions. Two respondents noted that Sweden would independently require formal benefit/cost (likely cost/quality-adjusted life-year [QALY]) analyses, after acceptance based on benefit/risk, for all pharmaceuticals, probably within the next 2 years. The United Kingdom, Ontario, and Australia already have these requirements. One respondent also noted that in deciding to accept and pay for computerized axial tomography scans in 1975, Sweden became the first country to make a public decision in healthcare based on formal economic evaluation. 

Those responsible for payment may still reject a new intervention even if others in their organization accept it based on benefit/risk outcomes. Examples of technology accepted but not paid for, at least initially, included orlistat for weight reduction, electrical stimulation for nonunion of bone fractures, islet cell and pancreas transplantation for diabetes mellitus, electron-beam computed tomography to diagnose coronary artery calcification, continuous glucose-monitoring devices, chemosensitivity assays for oncology chemotherapy, and vertical axial decompression for back pain. Less common examples provided by French insurers and some US MCOs included screening densitometry to diagnose osteoporosis or laser keratotomy to improve vision acuity. At the other extreme, many US MCOs and insurers would not pay for the nicotine patch unless accompanied by antismoking group therapy, a clear example of willingness to pay more for increased benefit. 

The most commonly cited examples of interventions initially denied payment for benefit/cost reasons but subsequently accepted were positive emission tomography scans and chiropractic treatments for back and neck pain. US providers/payers were the most aggressive in not paying for specific interventions based on benefit/cost criteria. Definitions of medical necessity, the experimental nature of some new interventions, and contractual restrictions also had important roles in decisions to deny acceptance or refuse payment. 

Two examples were raised of policy changes that bear on deletion of interventions previously accepted and paid. Both have important implications for expanding the role of sound medical evidence in benefit/cost analysis as the basis for health system resource allocation and payment decisions. If such examples become common, best medical evidence and outcomes of appropriate formal economic evaluations will be used increasingly in helping to make such decisions. Their importance is not from whence the evidence came but rather that the evidence is applicable and obtained by appropriately designed study. First, Sweden is implementing new policies that require pharmaceutical firms to re-evaluate benefits of currently marketed prescription medications. Substantial delisting is expected. Second, a decision was made in France to reduce or stop reimbursement for pharmaceuticals of no demonstrated benefit. It is expected that 30% of such medications will be removed from payment lists in the first round, with complementary and alternative medications disproportionately affected. 

Broad use of benefit/risk results in decision making was not surprising. Unexpectedly, however, a high percentage of respondents (42.1%) of information users reported use of benefit/cost results in decision making. Current literature paints a picture different from this study's results.

For example, a 4-country study on actual use found large numbers of economic evaluations submitted to public decision makers but uncertainty about whether such information influenced decisions.

Weatherly and colleagues, in their survey of 102 health authorities in England, found that cost-effectiveness results were well understood but not often used.

Hoffmann and associates found that UK health authorities recognized the usefulness of published economic evaluations but were concerned about narrow research questions, poor generalizability of results, and uncertain methodological rigor.

A case study by Pausjenssen et al directly measured actual use of economic analyses by the Drug Quality and Therapeutics Committee of the Ontario (Canada) Drug Benefit Program.

Surveys of 9 members after each of 12 committee meetings concluded that economic analyses affected decisions but mainly on highcost innovative medications; otherwise, clinical outcomes predominated. 

Bethan and colleagues indirectly measured use of economic evaluations in Australia. They found that benefit/ cost standards required for licensing and payment of pharmaceuticals were closely related to decisions to accept and pay for new pharmaceuticals.

The Pharmaceutical Benefits Advisory Committee was unlikely (2/26 submissions) to recommend a medication for acceptance and reimbursement if incremental cost per life-year saved was greater than AU $76 000, and unlikely (1/26 submissions) to reject one if incremental cost per life-year saved was less than AU $42 000. 

Raftery's evaluation of early implementation of guidelines found that with only 1 exception NICE recommended for approval and payment every new technology with a cost/QALY less than Â£31 000.

Neither Raftery nor Bethan et al found clear and conscious use of specific or preselected costs per outcome, but both studies had results similar to the widely accepted 2-decade standard of US $50 000 per life-year saved, a standard that is itself being questioned.

Thus, these studies indicate the use of economic evaluations by multiple decision makers.

There was no disagreement among respondents that all organizations involved in producing benefit/cost evaluations, including pharmaceutical and other technology firms and private consultants, can do appropriate and methodologically sound benefit/risk, benefit/cost, and clinical-outcome studies. However, respondents expressed concerns about the outcomes of economic evaluations by these firms because such studies nearly always yield findings positive to funding firms. As 1 French respondent asked, "Why are there so infrequently studies with negative results, and why do such studies nearly always show I am going to save money?" 

Two conflicting themes emerged from responses of information users and academics. On the one hand, most respondents (65.4%) agreed that a large percentage of benefit/cost studies were generally methodologically sound enough to help decision makers, even though nearly all agreed that such studies could be even better methodologically. At the same time, 89.2% of providers, payers, regulators, and academics felt there was insufficient transparency, of methods and analyses mainly, to judge quality and results of benefit/cost studies. Additionally, economic outcomes were rarely available when needed, and most decision makers were not knowledgeable enough to evaluate the quality of methods and data used. Narrowly focused decision analytic models were the most troublesome because research questions, assumptions, logic, technical details, and variable interpretations were often difficult to understand and relate to results. The comments on decision modeling are different from those of a recently published survey on the use of pharmacoeconomic models in decision making.

No respondents noted the conflict between general acceptance of benefit/risk results produced by firms and widespread suspicion of benefit/cost studies. It may be that financial implications from benefit/risk studies are less powerful incentives to make results appear more beneficial and less risky than they are, and less variable in interpretation, while direct economic effects from benefit/cost results are clear for providers/payers (expected increased budgets) and pharmaceutical and device firms (increased sales and profits). 

A first step toward resolving the issue of underuse of formal economic evaluations in healthcare decision making is for decision making groups (mainly payers) to require its use. A second option is for payers to require adherence to accepted standards for benefit/cost analysis and reporting, improve applicability to user needs, and address concerns about existing benefit/cost evaluation.

A desirable side effect may be increased use of all—evidence-based care.

Next, independent expert reviewers should examine benefit/cost studies for methodological soundness and appropriateness of results, as government regulators do now with benefit/risk studies. Then these benefit/cost evaluations can be confidently used as a basis for acceptance (in conjunction with benefit/risk results) and payment. Such policies would increase the likelihood of better clinical and economic decision making and resource allocation, thereby improving value for private and public money spent on health and medical services.

The main limitation of this study was the sample selected, small groups of selected persons in 4 countries— certainly no random probability sample. Although the results may not be completely generalizable, it is likely that the main conclusion is valid. 

Economic considerations are important factors in decisions to accept and pay for health and medical services. Though they considered the current methodologies and variability of information less than ideal, respondents indicated widespread and growing use of multiple formal economic data and evaluations in healthcare decision making. 

From the Departments of Medicine and Health Care Systems and the Leonard Davis Institute of Health Economics, University of Pennsylvania, Philadelphia. 

Address correspondence to: Bernard S. Bloom, PhD, University of Pennsylvania, 3615 Chestnut Street, Philadelphia, PA 19104-2676. E-mail: 

The Economic Writings of Sir William Petty.

 Hull CH, ed. Cambridge: Cambridge University Press; 1948.

Report of the Sanitary Commission of Massachusetts. 1850.

 Cambridge, Mass: Harvard University Press; 1948.

3. Elixhauser A, Halpern M, Schmier J, Luce BR.

 Health care CBA and CEA from 1991 to 1996: an updated bibliography. 

4. Drummond M, Brown R, Fendrick AM, Fullerton P, Neumann P, Taylor R, Barbieri M.

 Use of pharmacoeconomic informationâ€“report of the ISPOR task force on use of pharmacoeconomic/health economic information in health-care decision making. 

5. Drummond M, Dubois D, Garattini L, et al.

 Current trends in the use of pharmacoeconomics and outcomes research in Europe. 

MANAGERIAL

Use of Formal Benefit/Cost Evaluations in Health System Decision Making

Sean R. Tunis, MD, MSc, is chief clinical officer at the Centers for Medicare and Medicaid Services(CMS) and director of the Office of Clinical Standards and Quality. Dr. Tunis is responsible for makingevidence-based national coverage policies; setting quality standards for Medicare and Medicaid providers;and leading quality measurement and improvement activities for the CMS. Over his career,he has devoted considerable attention to the role of cost effectiveness in guiding decision making.Given his stature as a thought leader on these issues, we felt he was the ideal person to comment onthe topics raised in this issue of the 

The potential use of economic analysis in healthcaredecision making has strong natural appealas it becomes clear how an individual's healthcaredecisions can affect, through insurance coverage,the ability of others to obtain the care they need.Healthcare spending in the United States continues torise faster than inflation and worker salaries. Financialpressures force difficult decisions about whether toincrease premiums, reduce benefits, or reallocateresources. Using analytic methods that are designed tomaximize the value of the health services purchased fora given level of spending makes intuitive sense. The failureto adopt such a framework given the current economicrealities in healthcare seems, at best, unwise.The two articles by Neumann 

 that appearelsewhere in this issue illustrate this point quite well.

The application of economic analysis in healthcaredecision making, however, particularly with regard tomedical necessity, has proven to be highly controversial.Evidence varies regarding the extent to which decisionmakers actually use economic analysis, a point thatBernard Bloom discusses in this issue of the 

The reluctance to explicitly address financial issues inthis context may well reflect a "deep-seated abhorrenceof limits and of the corporate or government officialswho impose them," and Peter Neumann's article cogentlypresents this case.

 In addition, the discussion of theuse of cost-effectiveness analysis (CEA) can benefitfrom a broader understanding of other importantsources of tension within the policy process throughwhich decisions are made about coverage of specificitems and services. Significant disagreement exists overthe appropriate locus for medical decision making, theintegrity of the process used by payers to make thesedecisions, and the proper role of scientific evidence.

One important source of tension regarding decisionsabout medical necessity is the degree of deferenceaccorded to judgments and preferences of patients andtheir physicians. The idealized view is that clinicians andpatients arrive at optimal choices by carefully consideringthe pros and cons of the existing alternatives, free ofall factors other than clinical risks and benefits and individualpatient preferences. In this model, cliniciansattempt to identify the clinical intervention that bestmeets the patient's need, separating this clinical decisionfrom any consideration of what the indirect impact of thisdecision might be. Only at the level of these individualdecisions, it is argued, can all the subtle variations in clinicalpresentation and social context be thoroughly understoodand factored into determinations about what is thebest medically necessary care for the patient.

Unquestionably, clinical decisions are best made byinformed patients consulting with their clinicians.However, many outside factors that have little to dowith clinical benefits and preferences also influencehealthcare decisions. Patients are exposed to anincreasing volume of direct-to-consumer advertisingand to healthcare information from the Internet andother sources that are of variable objectivity and quality.Clinicians also receive a large volume of informationfrom commercial entities and the media that may notreflect a carefully balanced perspective on the value ofmedical alternatives. Several decades of health servicesresearch, documenting widespread underuse, overuse,and misuse of healthcare technologies and services,demonstrate that reliance on judgments at the level ofindividual patient encounters does not ensure that medicallynecessary care will be consistently provided.

The persistent discomfort with payers as a locus formaking decisions about medical necessity is manifestedin a range of social and legal developments. The backlashagainst managed care in the 1990s was primarilydriven by the powerful concern of patients and policymakers over the idea that those who pay for healthcareservices could override the decisions of clinicians andpatients regarding diagnosis, treatment, and specialistreferral. In legal proceedings, challenges to noncoveragedecisions are generally received with considerable sympathyby judges and juries who see sick patients withlimited options for getting well having their doctorsoverruled by organizations whose primary functions areadministrative, not clinical. States continue to pass lawsmandating that specific services be covered in order toprevent insurers from limiting access to certain technologiesand services, and requiring independentappeals mechanisms outside the control of the payer formedical necessity decisions. In the past few years, theUS Congress passed federal laws that allow beneficiariesto appeal local and national medical necessity policies,and that require Medicare to issue draft coverage decisionsfor public comment. These changes reflect steadypressure from industry advocates and a demand foradditional accountability when making policy thatblocks payment for medical services deemed to be medicallyappropriate by patients and their physicians.

Process for Medical Necessity Decision Making

A second important source of tension over coveragedecisions made by public and private payers is theprocess for making these decisions, particularly withrespect to explicitness, consistency, transparency, andopportunity for outside input. With a few exceptions,payers have been rightfully criticized on each of theseelements of the process. Historically, the process bywhich payers made these decisions was not describedin detail; few payers provided any clear description intheir contracts or other documents about how thesedecisions were made, by whom, with what information,and in what time frame. Prior to 1999, the Medicareprogram communicated coverage decisions throughbrief statements in the policy manual describing whatwould or would not be covered. Furthermore, no formalprocess was in place for making these decisions, andoutside stakeholders had limited opportunity to participatein or even track the development of specific policydecisions.

Many public and private payers have taken a very differentapproach over the last several years, investingconsiderable resources to make the medical policyprocess more scientific and transparent. The Blue Crossand Blue Shield Association has conducted rigorous scientificreviews of medical technologies for nearly 20years and applies well-defined criteria in evaluatingexisting evidence. Initially available only to subscribinghealth plans, the technology assessment reports havebeen made available to the public since June 2003. Morerecently, the Medicare program has made significantchanges to improve the process for making medicalnecessity decisions.

 A description of the process nowappears in the Federal Register and can be found on theMedicare coverage Web site (www.cms.gov/coverage).Clinical studies are evaluated through the well-describedframework of evidence-based medicine, and decisionmemos are now posted on the coverage Web site, providingdetailed explanations of what evidence wasreviewed and how the coverage policy was derived fromthe evidence. Substantial opportunity for outside inputexists, with numerous calls and meetings with stakeholderstaking place for each coverage decision. An independentexpert Medicare Coverage Advisory Committeewas also established to provide a public forum in whichparticularly complex or controversial decisions can bediscussed. While these process improvements have noteliminated significant differences of opinion regardingspecific decisions, the aggregate effect has been to substantiallyreduce the frequency and intensity of conflictswith stakeholders, and to allow for an orderly andinformed exchange of views on a better-defined andmore tractable set of issues.

While considerable progress has taken place, thereremains a high level of discomfort with many aspects ofthe process for making coverage decisions. Significantvariations in medical necessity policy still exist betweenhealth plans that cannot be easily explained. Manyhealth plans continue to develop these policies throughmechanisms that are neither reliable nor transparent.

Medicare has not yet been successful in publishing regulationsor guidance documents providing a definitionof medical necessity, or articulating the criteria that areapplied to medical technologies in order to determinewhether they are "reasonable and necessary."

The Role of Evidence-based Medicine inMedical Necessity Decisions

While an increasing number of organizations haveadopted evidence-based medicine as the analyticframework for making medical necessity decisions, considerablediscussion continues regarding the appropriateapplication of this framework. First and foremost isthe question of how to determine when the overallquality of evidence is adequate to demonstrate medicalnecessity for individual technologies and services.While the framework provides a well-defined set ofrules to guide the appraisal of evidence from clinicalstudies, the translation of this assessment into categoricaldecisions of medical necessity may require subtlejudgments about study methods.

A second critical and challenging issue is how differentsources of information about risks and benefitsshould be weighed when making medical necessitydecisions. Specifically, different views exist about therelative importance of local expert opinion as reflectedin community standards of care, national expert clinicalconsensus as articulated in clinical practice guidelinesand consensus documents, and conclusions aboutclinical effectiveness derived directly from high-quality,peer-reviewed clinical research. Many stakeholdersmaintain that the state of the art in clinical practicemay sometimes be ahead of the state of the science inthe published literature, and that the clinical communitydetermines the value of clinical interventionsthrough experience long before confirmation is providedin published studies. Others raise questions aboutthe reliability of professional consensus by pointing tothe large geographic variations in care,

 high rates ofinappropriate care, and underuse of services of proveneffectiveness.

Beyond these fundamental issues, numerous secondaryissues must also be addressed. Should evidencerequirements be applied differently for different categoriesof technologies (eg, drugs versus devices)? Shoulddiagnostic technologies be evaluated based on diagnosticaccuracy alone, or should there be evidence to supportthe impact of the technology on patient managementand health outcomes? What is the optimal point in timeto assess new technologies? Should technologies in widespreaduse be reviewed to determine whether they aresupported by good clinical studies, and should they beheld to a different standard of evidence than new technologies?Who is responsible for generating evidenceregarding the effectiveness of existing and new technologies?These and many other issues related to theuse of evidence-based medicine in healthcare policymaking are the subject of highly spirited debate, againhighlighting the fact that the challenges and resistanceattributed to the use of cost-effectiveness analysis mustbe viewed as part of a broader set of challenges.

In addition to the 3 categories of issues discussedabove, several critical challenges must be confrontedregarding the use of CEA in medical necessity decisions.These major challenges include concerns aboutvalidity and transparency of CEA methods, uncertaintiesabout motivations for use of CEA, and the difficultiesinherent in applying results to individual patients.

Neumann correctly notes that tremendous progresshas been made in developing and standardizing themethods of economic analysis in healthcare over thepast 25 years. The number of such studies published infirst-tier peer-reviewed journals provides strong validationof the credibility and usefulness of thesereports. Whether or not the methods are sufficientlyrobust or transparent to be used in medical necessitydeterminations is a separate question. Economic modelsare inherently more complex and less transparentthan analyses of the clinical studies that comprise theeffectiveness component of the models. Many importantassumptions are embedded in decisions about themodel structure, and it is challenging to ensure thatevery important decision is fully disclosed andexplained. This complexity poses particular problemsbecause CEA sponsors' motivations may be suspectedof guiding the analysis in ways that may be difficult todetect. Given the critical role of transparency andcomprehensibility in the legitimacy of policy decisionsabout medical technology, CEA raises special challengesover and above those faced when weighing clinicalrisks and benefits.

Even under the assumption of robust and generallyaccepted CEA methods, the application of these techniquesto payers' decisions remains problematic. Somediscomfort with these techniques may be explained bya general public resistance to recognizing that healthcareresources are limited, but the discomfort isheightened when decisions to limit care are made byorganizations for whom reduced spending may belinked to improved financial performance. Use of economicanalysis under the premise that resources arelimited raises questions about who is actually benefitingfrom any savings, particularly when applied by forprofitentities.

 Greater acceptance of economicanalysis for decision making in the United Kingdom(and other countries with a publicly funded nationalhealth service) may reflect in part a greater culturalemphasis on equity, but may also emerge from thestructure and financing of the British healthcare system.Because virtually all healthcare in the UnitedKingdom is paid for through a single national program inwhich all individuals participate, the reality of a fixedbudget is more self-evident, and the logic of applyingCEA to use resources efficiently becomes less suspect.While fundamental differences in history and valuesmay explain why the United States and United Kingdomhave such different approaches to financing and delivering healthcare, the substantial differences that nowexist between these systems could themselves explainobserved differences in receptivity to the use of CEA indecision making.

Finally, and possibly most significant, the applicationof economic analysis in making decisions aboutmedical necessity becomes highly problematic wheneverit is translated to decisions about the care of individualpatients. The determination that a particulartechnology or service is not cost effective means thatspecific individuals, someone's parent, spouse, child, orfriend, may be unable to obtain that potentially beneficialintervention because it has been deemed not to beworth the price. Confronting such dilemmas on theindividual level invariably raises virtually impossibleobjections to the use of CEA for individual yes/no decisionsabout what will and won't be covered. When adecision not to pay for a particular service is made byweighing the clinical risks against the benefits, theunderlying logic is to avoid exposing the individual topossible harm. Using CEA for such decisions can betranslated to mean that a clinical benefit must be foregonebecause the price is too high.

It has become a widely held view that the inability tocomplete a federal rule-making process to define theterm "reasonable and necessary" results primarily fromdiscomfort with the proposed inclusion of cost effectivenessas one of the criteria. Neumann's articleexplains this in the context of our unwillingness as asociety to accept that healthcare resources are limited.I would agree that the public desire for healthcare servicesexceeds the allocated resources. However, successfulcompletion of rule-making related to coveragedecisions, with or without inclusion of a cost-effectivenesscriterion, would also require the development ofsome consensus regarding the locus of decision making,transparency of the process, and the appropriateuse of evidence in decision making.

Despite the challenges facing the use of economicanalysis in decisions about technology and procedures,it does have a critical role to play in healthcare. Whileit may not be a perfected tool, it provides a robustmethodology to support efforts to purchase the greatestamount of health for any given level of healthcarespending. Because of the special challenges raised byusing economic methods to make categorical decisions,it may be preferable to further develop policyapproaches that include cost outside the realm of medicalnecessity. Medicare has recently taken some stepsin this direction with respect to 2 very high cost technologieswith marginal, operator-dependent benefits:lung volume reduction surgery for end-stage lung diseaseand left ventricular assist devices for destinationtherapy for end-stage heart failure. In each case,Medicare limited coverage of these procedures to centerswith evidence of unique expertise, both to increasethe likelihood that patients would have improved outcomesand to promote appropriate use of these high-risk,costly interventions. The pricing of a newimmunoassay used to screen for colorectal cancer wasderived in part from an economic model comparingthis test to alternative colorectal cancer screening technologies.In this case, the statutory language authorizingMedicare to add coverage for the new colorectalcancer screening test explicitly directed the Secretaryto include costs as one of the factors to be considered.

Obtaining good value from healthcare spending is anurgent policy priority, made more so by the acceleratingpace of medical discovery. Perhaps the most importantbenefit of expanding the use of economic analysisin health policy decision making would be to supportmore informed public dialogue about how best toachieve a healthcare system that produces high-qualityand safe healthcare, allows patient freedom of choice inhealthcare decisions, maintains healthy innovation,and expands access to care while remaining affordable.A decision-making framework that explicitly includeseconomic analyses would enable us to adopt explicitand consistent reimbursement guidelines that linkhealthcare benefits to the amount paid.

 Without sucha framework, linkage between quality and cost is considerablymore difficult. In an environment of increasingpressure on healthcare resources, importantopportunities to improve public health may be missed.

From the Centers for Medicare & Medicaid Services, Baltimore, MD.

This editorial reflects the views of the author, and does not necessarily represent theviews of the Centers for Medicare & Medicaid Services or the Department of Health andHuman Services.

Address correspondence to: Sean R. Tunis, MD, MSc, Centers for Medicare &Medicaid Services, 7500 Security Boulevard, Baltimore, MD 21244.

Crossing the Quality Chasm: A New Health Systemfor the 21st Century 

No Margin, No Mission: Heathcare Organizations and theQuest for Ethical Excellence.

 Why donâ€™t Americans use cost-effectiveness analysis? 

 Use of formal benefit/cost evaluations in health system decision making.

. Washington, DC: National Academy Press; 2001.

 Improvements in Medicare coverage of new technology.

 Medical necessity, coverage decisions andmedical policy. In: 

 New York: Oxford University Press; 2003:67-95.

 The mammography dilemma: a crisis for evidence-based medicine?

7 Fisher ES, Wennberg DE, Stukel TA, Gottlieb DJ, Lucas FL, Pinder EL.

 Theimplications of regional variations in Medicare spending. Part 1: the content, quality,and accessibility of care. 

 The quality of health care delivered toadults in the United States. 

 NewYork: Oxford University Press; 2002:1-12.

 Benefit language: criteria that will improve quality while reducingcosts. 

EDITORIAL

Economic Analysis in Healthcare Decisions

To develop and evaluate a set of questionnaire-based screening tools to identify risk for 1-year adverse outcomes in adults with moderate to severe asthma. 

Prospective cohort study in 16 managed care organizations in the United States. 

Patients (n = 4888) with moderate-to-severe asthma completed baseline and 1-year questionnaires (response rate, 79%). Adverse outcomes included hospitalization in the past year; emergency department (ED) visit in the past year; days of lost activity in the past month; a composite measure combining hospitalization, ED use, and lost days; and severe symptoms. Risk models were constructed for each of these 5 outcomes. Candidate predictors included baseline demographic characteristics, prior asthma healthcare use, access to care, symptoms, and treatment. Outcome variables were dichotomized, and logistic regression analysis was used to estimate the probability of 1-year outcomes. 

The patients' mean age was 45 years; 69% were female, and 83% were white. At 1-year follow-up, 9% had been hospitalized in the past year, 35% had used the ED, and 36% had reduced activity in the past month; 54% reported at least 1 of these, and 53% reported severe symptoms. Twenty-one items were retained for the 5 final risk models. Overall, the strongest predictors were comorbid illnesses and prior ED use. Model discrimination using receiver operating characteristic area ranged from 0.67 to 0.78 for predicting hospitalization, ED use, lost days, any one of these outcomes, and symptoms. 

The questionnaire-based risk models identified with good discrimination asthmatics at increased risk for a range of adverse outcomes. Risk models based on patient-reported data could be used to target individuals for intervention. 

Asthma, a common chronic condition, has substantial effects on individuals, the healthcare system, and society, accounting annually for 2 million emergency department (ED) visits, more than 400 000 hospitalizations, 4500 deaths, 2 million lost work days, and a cost of $6 billion.

The enormous burden of asthma on patients and society in part reflects the under treatment of patients at risk of poor outcomes.

Therefore, there is a need for screening tools to identify patients needing closer scrutiny of their therapeutic regimen.

Screening populations to identify higher-risk individuals for disease management programs can improve efficiency of resource use and lead to cost savings.

In managed care, there are opportunities to improve asthma care, particularly with medication use.

May 2004

The Family Impact and Costs of Migraine

Use of Formal Benefit/Cost Evaluations in Health System Decision Making

Economic Analysis in Healthcare Decisions

Predicting Patient-Reported Asthma Outcomes for Adults in Managed Care

Why Don"t Americans Use Cost-Effectiveness Analysis?

An Illness in the Family: Accounting for the Complex Effects of Illness on Other Family Members

Assessing Symptoms, Disease Severity, and Quality of Life in the Clinical Context: A Theoretical Framework