McGarry and colleagues did an excellent job in providing an economic perspective on the effectiveness of thromboprophylaxis with a low-molecular-weight heparin versus unfractionated heparin in acutely ill medical inpatients.

Low molecular weight heparins (LMWHs) have gained wider acceptance as the drugs of choice in prophylaxis of venous thromboembolism (VTE) due to their superior efficacy and cost effectiveness to unfractionated heparin, and the reduced risk of heparin-induced thrombocytopenia and osteoporosis.

However, new developments in the realm of VTE prophylaxis in medically ill patients have provided us some valuable insights. The ARTEMIS trial demonstrated the efficacy and safety of fondaparinux in medically ill patients versus placebo.

Although there are no comparative trials between LMWHs and fondaparinux in medically ill patients, there have been a few head-to-head pharmacoeconomic studies in other patient types that have shown fondaparinux to be cost effective.

McGarry's study notwithstanding, we believe the current question to be addressed is how to choose between LMWHs and fondaparinux and which agent is a more cost-effective option in medically ill patients. 

1. McGarry LJ, Thompson D, Weinstein MC, Goldhaber SZ. 

Cost effectiveness of thromboprophylaxis with a low-molecular-weight heparin versus unfractionated heparin in acutely ill medical inpatients. 

2. Cohen AT, Gallus AS, Lassen MR, et al. 

Fondaparinux vs placebo for the prevention of venous thromboembolism in acutely ill medical patients (ARTEMIS). Program and abstracts of the XIX Congress of the International Society on Thrombosis and Haemostasis; Birmingham, UK, July 12-18, 2003. P2406. Available online at http://www.blackwellpublishing.com/isth2003/abstract.asp?id=10228. Accessed November 29, 2004. 

3. Dranitsaris G, Kahn SR, Stumpo C, et al on behalf of The Fondaparinux

Canadian Health Economic Study Investigators. 

Pharmacoeconomic analysis of fondaparinux versus enoxaparin for the prevention of thromboembolic events in orthopedic surgery patients. 

4. Sullivan SD, Davidson BL, Kahn SR, Muntz JE, Oster G, Raskob G. 

A cost-effectiveness analysis of fondaparinux sodium compared with enoxaparin sodium as prophylaxis against venous thromboembolism: use in patients undergoing major orthopaedic surgery. 

Articles

LETTERS TO THE EDITORS

Correspondence regarding, "Cost Effectiveness of Thromboprophylaxis With a Low-Molecular-Weight Heparin Versus Unfractionated Heparin in Acutely Ill Medical Inpatients"

have done a laudable job addressing some of the concerns around measurement of return on investment of disease management programs, including seasonality and regression to the mean. However, their evaluation of the success of their program to diminish asthma expense in the Colorado Medicaid population has introduced substantial potential biases that diminish the strength of their conclusions. 

Tinkelman and Wilson have chosen to have a "convenience" control group, as opposed to using a randomized control group. This control group is on average 4 years older than the intervention group, raising the question of overrepresentation of the disabled in the control group compared to the intervention group. There is substantial evidence that disabled and Temporary Assistance for Needy Families (TANF) Medicaid patients are very different populations,

and they should each be analyzed separately when reviewing efficacy of medical management programs. The need to eliminate 2.8% of the control group for "high use of nonasthma claims"also suggests that there were systematic differences between the intervention and control groups. 

The control group was chosen because the researchers were unable to reach the provided home phone number. It is very likely that there is a systematic difference between those who researchers can reach and those who they cannot. The authors surmise that "those who were not at home were healthy and working, while those who were at home were sicker and not working."It is just as likely that those for whom the researchers did not have workable phone numbers were actually more impoverished, and therefore nonequivalent to the intervention group. 

The authors also removed noncompliant patients from the intervention group, while they could not remove such patients from the control group. Clearly, those who had "lost telephone service"might have been at higher risk of adverse outcomes of their asthma. It would have been more appropriate to evaluate the intervention group on an "intention to treat"basis, since there was no similar elimination of members of the control group. 

The authors acknowledge that their evaluation fails to consider the incremental cost of higher use of disease modifying inhalation agents in the intervention group. In fact, this additional cost might offset a considerable portion of the stated savings from this program. 

Tinkelman and Wilson have done us a service to identify substantial regression to the mean (30.7%) in the control group of their population. More rigorous studies will be required to ascertain whether there was really substantially more decrease in cost in a comparable intervention group. 

Asthma disease management: regression to the mean or better? 

2. KCMU (Kaiser Commission on Medicaid and the Uninsured). 

Medicaid program at a glance. Fact sheet. Washington: KCMU. January, 2004. Available at: http://www.kff.org/medicaid/loader.cfm?url=/commonspot/security/getfile.cfm&PageID=30463. Accessed December 31, 2004. 

The foundation of Tinkelman & Wilson's conclusion

 on the effectiveness of an asthma disease management (DM) program is based on the assumption that the reference group was equivalent to the population that received disease management services. Unfortunately, the authors do not present sufficient credible evidence to support that essential assumption. 

The authors do assert that answering the phone on the first call for both the participants and non-participants was a "random variable"and that the populations selected for DM and the comparison were "statistically and clinically well-matched as cases and controls."Yet, no supporting evidence from the literature is presented to back-up such a contention. Indeed, evidence available in the paper itself reveals that the 2 populations were quite obviously different. 

First, the investigators state that "individuals with extreme non-asthma use"were present in the reference group and were removed from the analysis, but no such outliers were cited as being in the intervention group. This fact is an obvious contradiction to the idea that the "random variable"caused, by itself, a group of "well-matched"controls. Second, the evidence presented in Table 1 to show comparability on age, in fact shows just the opposite. A simple statistical test using the values presented on age (mean and standard deviation) calculates to a 

statistic of 3.511 and a statistically significant difference on this important baseline characteristic (

This study would have been far more persuasive if the authors had 1) presented credible evidence from the scientific literature suggesting that their selection method (answering the telephone on the first attempt) has been shown to produce equivalent comparison groups elsewhere; and 2) if they increased the number of baseline metrics available in claims data (eg, comorbid conditions, inpatient and other utilization statistics, etc) in both the reference and intervention groups and offered appropriate statistical tests of their differences. 

The "equivalence"principle is perhaps the most essential component of any study.

Like many before us, we recommend that investigators studying defined population interventions include a self-critical discussion of the "strength of evidence"supporting their conclusions,

especially the degree to which selection bias was avoided.

The extent to which a study achieved "internal validity"(and the extent to which it did not) should be supported by evidence from the literature and/or data produced by the study itself. 

Framework for assessing causality in disease management programs: principles. 

The environment and disease: association or causation? 

A user's guide to the disease management literature: recommendations for reporting and assessing program outcomes. 

The correspondents have brought up some important points regarding the limitations of our study. So as to not "throw the baby out with the bath water,"however, we would like to address some of these concerns. While acknowledging that both the methods and the data may not be perfect, at the same time we submit that some useful information can be garnered from these results. 

We would argue that both Levin-Scherz and Wilson and Linden have perhaps overstated the likelihood of systematic differences between the treatment and control groups. While it is correct that the control group was not formally randomized and thus, could reasonably be construed as one of "convenience,"at the same time, the manifest similarity of the 2 groups suggests that they likely were not systematically dissimilar in several important ways. This methodology is consistent with that described previously by Haley.

We certainly agree, in accordance with that principle, "since the members of a convenience sample were not selected by a true random sampling method, one will not have the same degree of assurance that the results can be generalized to the target universe as one would have with a random sample. However, many convenience samples are free enough of obvious biases to provide accurate estimates of their target universes."

Another issue raised in both letters relates to the age of the populations. While the ages were statistically different in the 2 groups (with the given group sizes and standard deviations, a difference of 2.1 years would have been statistically significant), the mean and range were very similar. This was also true for the other measured variables (sex, asthma utilization, and non-asthma utilization). We acknowledge it is possible that there were more disabled individuals in the control group or that the control group was more impoverished. However, it is difficult to conceive how these differences might not affect costs in the baseline year, but then go on to affect costs in the treatment year. Thus, we would argue that this supposition is possible but not likely. 

Dr. Levin-Scherz makes an excellent point about using an intention to treat analysis, at least on the individuals who were excluded from the treatment group for reasons other than loss of eligibility for insurance. We would absolutely remedy this oversight in future studies. We also agree with his conclusion that replication of these findings is important step to further substantiating the benefits of asthma disease management. In addition, we agree with all of the valuable suggestions from Wilson and Linden, which indeed would have strengthened this paper significantly. We were constrained by space limitations by the journal and had to limit our presentation of these points in the Discussion. We thank both Drs. Levin-Scherz and Wilson for sharing their valued insights. 

National Jewish Medical and Research Center

Techniques of Patient-Oriented Research. 

Designing Clinical Research. In: Pak CYC, Adams PM, eds. 

Correspondence regarding, "Asthma Disease Management: Regression to the Mean or Better?"

To assess variation in the content of Medicare's localmedical review policies.

Six case studies to compare differences in coveragepolicies by diagnosis codes, procedure codes, and indicationsfor use.

All carrier policies from 48 carrier contracts (n = 5213)posted to the Centers for Medicare & Medicaid Services Web sitewere downloaded on May 31, 2001. All policies in the data setwere coded based on a typology: new technology (NT), extensionsof new technology (TE), and utilization management (UM) of widelyused procedures. We identified policies addressing the sameprocedure or technology. We required at least 20 separate policiesin each case study to allow meaningful comparisons. We randomlyselected 1 case study of a diagnostic and 1 for a treatmentmodality from each policy type (NT, TE, and UM).

Given previous research on local carriers, we expectedto find variations among policies in each case study. We foundsubstantial similarity, however, among policies covering the NTand TE types. We found significantly more variation among ourUM-type case studies.

Medicare legislation has called for greater coveragepolicy consistency in Medicare. This analysis on variation inpolicy content, part of a larger study on variation in Medicare'slocal coverage process, provides data on policy content differences.Policy reform should reflect the nature of and reasons forpolicy variation as suggested by the findings of this research.

Local and regional variations in healthcare havebeen well documented and raise important issuesabout quality of care in the system.

A payer'scoverage policies are intended to define whether a particularservice or technology will be offered and, if so,set forth the conditions of use. There is growing interestamong payers, providers, and policy makers about theuse of scientific and economic evidence in healthcarecoverage decision making to encourage more appropriateuse and reduce variation.

Managed care organizations have a strong interest inunderstanding practice variations and encouragingappropriate use to improve outcomes and reduce costs.They have an additional interest in regional variationwithin the Medicare program because the calculation ofthe Medicare Advantage premium is tied to fee-for-serviceadjusted average per capita costs. Regional variationstrongly influences the amount of a Medicare Advantagepremium. Because Medicare Advantage plans must followthe coverage policy in the county where a beneficiaryresides, plans offering services over a largegeographic region or nationally face operational challengeswhen Medicare coverage policies differ.Moreover, recently enacted Medicare provisions aremoving the program toward new regional models, raisingadditional concerns about the implications of policyvariations and practice patterns on participating healthplans.

has recently described efforts toenhance Medicare's national coverage process, mostMedicare coverage decisions are decentralized.Medicare relies on a patchwork of nearly 50 local contractingorganizations (called 

for Part A) that develop thousandsof local medical review policies (LMRPs) applicable onlywithin their own jurisdictions. The 1965 Medicarestatute authorized local contractors to process claimsas a buffer between providers and government.

Contractorssubsequently acquired the coverage policymaking function through administrative action in the1990s.

The Medicare Payment Advisory Commission andthe US General Accounting Office have recommendedelimination of Medicare's local coverage process,

andthe 2003 Medicare Prescription Drug Improvement andModernization Act directs the Centers for Medicare &Medicaid Services to work to achieve greater coveragepolicy consistency. However, little empirical work hasbeen done to examine the extent of, reasons for, or consequencesof variation in local coverage.

Our research team has been involved in a multiyearstudy of Medicare's local coverage processes. The variationsin the size, resources, and stability of the diversecontracting organizations have been reported elsewhere.

Significant variations in contractor productivity(number of policies each contractor posted), use ofevidence (based on evidence cites in LMRPs), and effectivedates of policies among local contractors have alsobeen found.

These findings document variation acrossa wide range of variables.

There has been no systematic analysis of variation inpolicy content, to our knowledge. For example, whilewe know when local policies covering deep brain stimulation(DBS) were issued, how those policies differ inspecific detail one from another has not been analyzed.What can the analysis tell us about variation that wouldinform payers, providers, and policy makers? To answerthese questions, we developed 6 case studies that revealpatterns of policy variation with important implicationsfor Medicare.

Medicare carriers are required to post LMRPs on theCenters for Medicare & Medicaid Services Web site withprescribed data fields.

We downloaded all LMRPs onMay 31, 2001. We observed that all LMRPs are not thesame. We identified 3 types of LMRPs: "new technology"(NT), "technology extension"(TE), and "utilizationmanagement"(UM) of widely diffused interventions,defined as follows: NT policies provide guidance for, andlimitations on, the use of new clinical interventions; TEpolicies focus on new uses of procedures or technologiesalready covered for other uses; and UM policies focus onwidely diffused procedures to avoid misuse or overuse.Two physician consultants coded all policies in ourdatabase. More than 85% of LMRPs focus on widely diffusedtechnologies (UM); the rest (< 15%) evaluate newtechnologies (NT) or extensions of technologies (TE). Acomplete discussion of the coding methods has beenpublished previously.

For our case studies, we established the following criteria:(1) We chose only policies from among 48 carrierfiles (n = 5213) because carriers are more active policymakers than fiscal intermediaries and their regions aremore specifically drawn. A complete discussion of thecomplex structure of Medicare contracting organizationshas been previously published.

(2) We selected onlycases in which our physician consultants agreed on thecoding by type. (3) To enable meaningful comparisons,we selected only case studies with at least 20 postedpolicies.

Using these criteria, we selected 6 case studies froma pool of 80 cases that met our criteria. We selected 2case studies from each of our 3 policy types, 1 diagnosticand 1 treatment or procedure in each type. The 6case studies represent a detailed analysis of 195 separatecoverage policies. Although additional case studiesmight confirm or challenge our findings, the clear patternswe found suggest that our results will be replicatedwith further research. The case descriptions aredrawn from posted policies on the Centers for Medicare& Medicaid Services Web site.

Deep brainstimulation is a neurosurgical procedure that uses electricalstimulation of subcortical structures (the thalamusor the basal ganglia) to control tremors.

is a gram-negative rod that has beencausally linked to chronic gastritis, peptic ulcers, gastriccancer, and gastric lymphoma. 

breath tests are noninvasive diagnostic procedures todetermine the presence of active infection.

Male urethral obstructionsmay result from infections, an enlargedprostate, prostate cancer, prostatitis with fibrosis, andother constrictions. Elimination of the obstructionincludes medical and surgical options. Urethral stentingis an extension of the growing use of stents to hold openoccluded vessels.

Transesophageal Echocardiography (TE, Diagnostic).

Transesophageal echocardiography (TEE) appliesan ultrasound generator to the exterior chest wall or inthe esophagus to obtain additional cardiovascular information.The instrumentation is invasive with potentialfor serious complications.

Toenail debridementinvolves the reduction of a thickened dystrophicnail resulting from mycosis or a severe systemiccondition using specialized equipment, such as forcepsor a rotary drill.

Cardiovascular Stress Test (UM, Diagnostic). 

Cardiovascularstress testing uses cardiac physiologicalmonitoring during and after stress, with or without subsequentcardiac imaging. A diseased heart respondsabnormally, allowing a diagnostic determination.

The case studies include 36 DBS policies, 39 

breath test policies, 33 urethral stent policies, 33TEE policies, 23 toenail debridement policies, and31 stress test policies. There were 41 foot care policies,but only 23 of them addressed toenail debridementand other toenail problems exclusively,while 18 embedded debridement with several otherfoot care procedures.Similarly, we identified43 policies coveringtransesophagealechocardiography; ofthose, 33 focusedspecifically on TEE,and 10 included TEEwithin a wider range ofechocardiography. Weexcluded the broaderpolicies from theanalysis because wecould not clearly identifywhich componentsin the policy applied tothe specific procedureand which referred tothe broader conditions.Coding all datapoints would artificiallyinflate the diversityof the policies.

Our policy contentcomparisons focus on 3areas: diagnosis codes,procedure codes, andindications.

Classification of Diseases, Ninth Revision,

One key measure of variation among policies is thearray of diagnosis codes in each policy. The 

is the principal system for codingpatient diagnoses and conditions. Local medical reviewpolicies specify which diagnosis codes are required forcoverage. We found substantial variation in the waysthat carriers cited diagnosis codes. Some policies listedcodes as ranges (ie, 427.0-427.9); others listed all codesseparately. We coded all ranges as including the mostcomprehensive range; we coded single diagnosis codesseparately.

illustratedegrees of uniformity or concentration by policy type.Each point on the x-axis reflects 1 carrier; the y-axisshows the frequency of specific codes cited. Figure 1demonstrates the concentration in our 2 NT case studies,DBS and the 

breath test. Of the 36 DBSpolicies, all contain 2 codes, idiopathic Parkinson diseaseand essential tremor. Only 2 carriers include 3additional codes, all of which refer to more generalizedspasmodic conditions. In other words, 34 of the 36 carriershave identical diagnosis codes; only 2 are marginallymore inclusive.

There is also considerable consistency among NTpolicies applicable to the 

breath test. There are24 different codes cited across all 39 policies. Fourappear in all policies (gastric ulcer, duodenal ulcer, gastrojejunalulcer, and gastric ulcer), 32 policies includethe diagnosis code for peptic ulcer, 29 cover 

infection and general abdominal pain, followed by a fewother abdominal disorders.

Figure 2 shows significant concentration in our 2 TEcases, urethral stent and TEE. Of the 33 LMRPs for urethralstent, there are only 7 cited codes or code ranges.All the policies include a code range that includes differenturethral strictures (eg, unspecified infection ortrauma). Twenty-three policies cover stents for benignprostatic hypertrophy, and only 2 cover strictures dueto shistosomiasis, syphilis, or gonococcal infection.Therefore, for the wide variety of strictures, the policiesare similar. However, in some jurisdictions, other specificstrictures can be treated with stenting.

Transesophageal echocardiography includes 130 differentdiagnosis codes or code ranges. The mean numberof codes or code ranges per policy is 39.9. Figure 2shows that there is high concentration for many codes,but there is variation among policies for a wide range ofother cardiac disorders.

Finally, Figure 3 demonstrates significant variationamong UM policies covering toenail debridement andstress testing. These are widely used procedures in theconventional medical arsenal. The UM policies tend toemerge when carriersidentify overuse or misusein their claims dataand want to set clearerlimits. Because there issubstantial variation inpractice patterns acrossregions, it is not surprisingto find thesecorresponding variationsin UM policies.

The HealthcareCommon ProceduralCoding System specifieswhich provider activitieswill be reimbursedas part of the coveredservice. There are 3levels of codes: Level I

HealthcareCommon ProceduralCoding Systemnational codes for servicesand supplies notincluded in the 

and local codesif no national code isyet available. Typically,policies listed discreteprocedure codes; occasionally,related codeswere listed in ranges.The 

demonstratesthat patterns ofconcentration amongprocedures (Healthcare Common Procedural CodingSystem) vary from those related to diagnosis codes. Theconcentration is measured by the percentage of policiesthat included the 5 most commonly cited codes. Whilethere are several ways to measure concentration (eg,the Herfindahl index), we used the percentage of policiesciting the most commonly cited codes because it isthe most straightforward method and is not confoundedwith the number of procedures cited. A high percentageof policies citing the 5 most common procedure codesindicates a high degree of homogeneity among policies.

We found highest levels of concentration in TE, withsignificant but somewhat lower levels of concentrationin NT. On closer examination,we found thatNT variation was relatedto new procedures. Forexample, the variationin the DBS policies maybe explained by thepresence of temporary(E) codes in some policiesand by variation inthe listing of ancillaryprocedures, such assubcortical mapping orpreliminary monitoring.These variations do notappear to affect beneficiaryaccess to the coreprocedure but mightaffect total providerreimbursement.

Some of this incidentalvariation presumablywill ameliorate asNT procedures proliferate,temporary codesbecome permanentcodes, and policies arerevised to reflect evolvingpatterns of use. Theconcentration in procedurecodes for TE policiesprobably reflects agreater understandingof a technology beingextended to a new use.

POLICY

Variation in Medicare"s Local Coverage Policies: Content Analysis of Local Medical Review Policies

Reimbursement options for pharmaceuticals reimbursed underMedicare Part B (physician-dispensed drugs) are changing and thenew comprehensive Part D Medicare outpatient drug benefit bringsfurther changes. The Medicare Prescription Drug, Improvementand Modernization Act of 2003 (MMA) replaces traditional policy,of reimbursing Part B drugs at 95% of average wholesale price(AWP, a list price), with a percentage markup over the manufacturer'saverage selling price; in 2005 an indirect competitive procurementoption will be introduced. In our view, althoughAWP-based reimbursement has been fraught with problems in thepast, these could be fixed by constraining growth in AWP and periodicallyadjusting the discount off AWP. With these revisions, anAWP-based rule would preserve incentives for competitive discountingand deliver savings to Medicare. By contrast, basingMedicare reimbursement on a manufacturer's average selling priceundermines incentives for discounting and, like any cost-basedreimbursement rule, may result in higher prices to both public andprivate purchasers. Indirect competitive procurement for drugsalone, using specialty pharmacies, pharmacy benefit managers, orprescription drug plans, is unlikely to constrain costs to acceptablelevels unless contractors retain flexibility to use standard benefitmanagement tools. Folding Part B and Part D into comprehensivecontracting with health plans for full health services is likely tooffer the most efficient approach to managing the drug benefit.

The Medicare Prescription Drug, Improvementand Modernization Act of 2003 (MMA) radicallychanged coverage and reimbursement for outpatientpharmaceuticals. The MMA established MedicarePart D, which provides comprehensive outpatient prescriptiondrug coverage for seniors, starting in 2006.This benefit is to be administered by stand-alone privateprescription drug plans (PDPs), whose role is to negotiatedrug discounts and process claims on behalf of thegovernment. More immediately, MMA changed the rulesfor reimbursement of physician-dispensed drugs, whichare already covered under Medicare Part B. Specifically,starting in January 1, 2004, reimbursement for mostPart B drugs was reduced from 95% of their currentaverage wholesale price (AWP) to 85% of their AWP onApril 1, 2003. Beginning in January 1, 2005, reimbursementfor single-source (mostly on-patent originator)drugs is 106% of their average selling price (ASP) orwholesale acquisition cost (WAC), whichever is lower.Reimbursement could be lower if the widely availablemarket price is at least 5% less than the ASP. For multiple-source drugs (generics and off-patent brands),reimbursement is based on the ASP subject to a maximumallowable charge (MAC) limit. (The MAC is basedon the ASP of a low-priced product in the molecule.)The ASP is intended to represent the volume-weighted,average manufacturer selling price net of rebates anddiscounts, to all purchasers excluding Medicaid andfederal purchasers.

Beginning in 2006, Centers for Medicare & MedicaidServices (CMS) will begin a competitive acquisition programas an option for dispensing physicians. The organizationwill select, by competitive bid, at least 2contractors in each area to be responsible for negotiatingdrug prices with manufacturers and delivering drugsto physicians. Physicians may opt to continue their owndrug purchasing, with the ASP/WAC reimbursement inplace for 2005, or they may obtain Part B drugs from 1of the contractors in their area. Under this option,physicians will no longer acquire Part B drugs frommanufacturers or seek reimbursement from CMS; thesefunctions will be assumed by contractors, along withany associated profit or loss. Many details of the competitiveacquisition program remain to be determined,including defining the geographic areas and setting standardsfor quality, access, financial stability, and solvency.Under the legislation, CMS is to make recommendationsto Congress by January 2005 on whether Part B drugsshould become part of the Part D benefit, but newlegislation would be required in order for that changeto occur.

The efficient design of Medicare payment policies fordrugs is a critical issue, given the fiscal pressure impliedby this benefit and its potential effect on private sectorpayers, the pharmaceutical industry, and ultimatelyconsumers. The elderly, who currently represent about12% of the US population, consume about 40% of pharmaceuticalssold in the United States; this share isexpected to increase as the US population ages. Althoughmore than 78% of seniors already have some drug coverage,

the Medicare drug benefit expands and consolidatesthis coverage, making Medicare the dominantpayer for pharmaceuticals in the United States. ThusMedicare's reimbursement strategies will influence pharmaceuticalprices and profitability, and ultimately incentivesfor research and development (R&D) and theavailability of new drugs for consumers worldwide.

Here we review Medicare's payment options fordrugs, comparing the effects of alternative administeredpricing rules—specifically, rules based on list versustransactions price—and alternative competitive strategies,in the context of the economic realities of pharmaceuticalpricing. We evaluated the effects of thesevarious policies for Part B drugs, because reimbursementrules are already defined and Medicare is alreadythe dominant payer. We draw conclusions from the PartB experience for the larger Part D benefit and for stateMedicaid and private payers.

Several recent reviews of Medicare's Part B drug policyhave been excellent.

Here we provide a brief historicalperspective, as background to the discussion offuture reimbursement options.

Medicare Part B traditionally covers outpatient drugsthat are related to a physician's service or a coveredmedical device and some immunosuppressive drugs.About 75% of current reimbursement is paid to physiciansfor drugs that they acquire, primarily chemotherapy,other cancer treatments, and other infusion drugs.In 2002, Medicare spent about $8.4 billion on thesedrugs, up from $6.5 billion in 2001 and $4 billion in1999 (Bill London, personal communication, February2, 2005). The almost 3-fold increase between 1998 and2002 implies an average annual growth of 27%, farexceeding the annual growth rate of private-sector outpatientdrug spending.

Prior to 2004, Medicare's payment for Part B drugswas set at the AWP minus 5%. The AWP as a concept forreimbursement originated in the 1960s in response to awide variety of "cost"prices used by wholesalers,direct-selling manufacturers, specialty wholesalers, andpharmacies. Led by Medi-Cal (California's Medicaidprogram), AWP was adopted as a "standard cost definitionthat would reflect what wholesalers normallycharged for the drug."Starting in 1969, Medi-Cal paidpharmacies AWP plus a flat dispensing fee to cover thepharmacist's professional service. The initial problem ofnot having a listing of such prices was resolved whenfirst the Red Book and then the Blue Book starting publishingAWPs as reported by drug wholesalers.

In 1991, the Health Care Financing Administration(HCFA, the predecessor agency to CMS) announced thatoutpatient drugs would be reimbursed at the lower ofAWP or the estimated acquisition cost, to be determinedby surveys of actual prices paid by physicians and otherproviders who dispensed these drugs. However, HCFAsubsequently decided that this strategy was unworkablebecause of the large number of providers and the widevariations in their use of drugs and the prices paid. Inthe 1997 Balanced Budget Act, reference to estimatedacquisition cost was deleted and AWP minus 5% becamethe sole basis for reimbursement.

The AWP has become a list price that is set by thedrug's manufacturer and published by pricing servicessuch as the Red Book. With AWP not defined in eitherstatute or regulation and no requirements or conventionsthat AWP reflect actual market prices, theGeneral Accounting Office (GAO) reflected consensusopinion when it concluded that AWP "may be neitheraverage nor what wholesalers charge."

Some eventranslate AWP as "ain't what's paid."Studies by theGAO and the Department of Health and HumanServices Office of the Inspector General have consistentlyconcluded that this standard results in paymentby Medicare that significantly exceeds providers'acquisitioncost for most drugs and biologics, and excessiveco-payments by beneficiaries, who pay 20% of thecharge.

The physicians who administer Part B-covereddrugs and are the chief beneficiaries of the spreadbetween AWP minus 5% and their actual acquisitioncosts argue that overpayment is needed to compensatefor inadequate Medicare payments for drug administration.

Whatever its empirical validity, this claim thatoverpayment for drugs makes up for inadequate physicianfees kept Congress from reducing Part B drug paymentsuntil the MMA, which raised payments forselected professional oncology services.

Before dismissing an AWP-based approach, it is worthnoting that many private sector payers, who are presumablymotivated to minimize costs and who lack thestatutory constraints faced by Medicare, also traditionallybase their drug reimbursement to pharmacies on discountedAWP. The critical difference is that privatepayers apply a larger percentage discount and revise thepercentage periodically, to capture most of the spreadbetween AWP and the pharmacy's actual cost of acquiringdrugs. Thus, the real questions are (1) could an AWP-basedapproach for Medicare be revised to eliminate itsprior problems; (2) how does a revised AWP approachcompare to using ASP or some other transactions priceas the basis for reimbursement; and (3) will competitivebidding strategies provide a superior alternative?

Sound payment policy for any medical serviceshould provide reasonable cost control while creatingincentives for efficient utilization. In the case of pharmaceuticals,sound payment policy must also considereffects on incentives for R&D. Given the cost structureof research-based pharmaceuticals, with high fixedcosts of R&D and low marginal costs, the concern is thata large payer such as Medicare might use its monopsony(single buyer) power to force prices down to marginalcost, which would erode incentives for R&D. Becausepharmaceutical R&D takes 8 to 15 years, the effect ofinadequate prices on the flow of new drugs would not beevident until many years later, in contrast to inadequateprices for other medical services in which providers'withdrawal of services acts as a prompt corrective toinadequate prices.

So far, both Medicare and Medicaid have adoptedreimbursement practices similar to those used by privatepayers, but with important differences. Both publicand private payers overwhelmingly pay for drugs indirectly,reimbursing the dispensing pharmacist or physicianfor their cost of acquiring the drugs plus adispensing fee, rather than paying manufacturersdirectly. A fundamental challenge in designing thesereimbursement rules is that payers do not observe theacquisition costs paid by dispensing pharmacists andphysicians, and these acquisition costs may vary,reflecting manufacturer discounts geared to volume,prompt payment, incentives to increase market share,and bundling. Any spread between the payer's reimbursementfor the drug and its acquisition cost accruesas income to the dispensing pharmacy or physician,regardless of whether the payer's benchmark for reimbursementis AWP or some proxy of acquisition cost.Because each provider's actual acquisition cost is unobservable,the payer can at best hope to roughly approximateaverage acquisition cost, while having a neutraleffect on prescribing and dispensing choices (unless thepayer specifically wishes to influence usage, such asencouraging generic substitution).

The fact that providers capture some spread betweenreimbursement and acquisition cost of drugs is oftenviewed as inappropriate. In fact, such margins betweenreimbursement and cost may occur for all medical services.Although the payer may appear to pay too much inthe short run, in the longer run the potential forproviders to capture such margins reinforces theirincentives to try to reduce their costs. Indeed, makingproviders cost-conscious, by giving them both upsideand downside risk for costs, was a key objective in theswitch from cost-based reimbursement to prospectivepayment for hospitals and physicians. In the case ofdrugs, permitting dispensers to capture some spreadbetween reimbursement and acquisition cost createsincentives for them to be price-sensitive, which in turnleads drug manufacturers to compete by cutting prices.

Such competition can yield savings to payers, providedthat manufacturers cannot simply increase thespread by raising the AWP and that payers reduce theirpayment to reflect lower acquisition prices. Many privatepayers take advantage of this system, reimbursingretail pharmacies based on AWP minus a percentage— now roughly 15% to 18%—that has increased over timeas the spread has increased. Many Medicaid programsuse a similar approach as at least 1 option for pharmacyreimbursement. Similarly, the key drivers of genericprice competition in the United States are that pharmaciescan substitute between approved generics and cancapture the spread between the payer's reimbursement,usually a MAC and their acquisition cost. By makingpharmacists price-sensitive, this system creates incentivesfor generic manufacturers to compete on the pricesthat they offer pharmacies, to increase their share of thepharmacy's business. Thus although the reimbursementspread may appear inappropriate, it serves the importantpotential function of encouraging price competitionby manufacturers, which can yield savings to payers ifreimbursement is adjusted down in line with costs.

An alternative approach is for the payer to define thepermissible manufacturer price for the drug, adding amarkup to cover wholesalers and dispensing costs.Countries that regulate drug prices generally apply thisapproach to regulating the manufacturer's price atlaunch. Price regulation uses 1 of 3 benchmarks: (1) theprice of competitor products in the same class as thenew drug (internal referencing); (2) the price of thesame drug in other countries (external referencing); or(3) some estimate of manufacturer costs. None of theseapproaches meets the criteria for sound payment policyto manufacturers. Internal referencing raises issues ofappropriate markups for innovation or superior safetyor efficacy, and for on-patent versus generic status;external referencing simply imports foreign price controlsand undermines appropriate cross-national pricedifferentials; and cost-based reimbursement for pharmaceuticalsis doomed by problems of measuring andallocating joint cost.

The hope of both the administered pricing rules andcompetitive contracting options proposed for Medicareis that these can constrain prices to reasonable levelswhile avoiding the opposing risks of forcing prices downto marginal cost or, alternatively, paying inappropriatelyhigh prices.

We focus here on options for paying for Part B drugsas outlined in MMA; options for Part D would be similar.These options fall into 2 broad categories (

). First,Medicare could define administered pricing rules, basedon either a list price or an estimated transactions price,with discounts or markups. Any administered pricingrule could use internal reference pricing, wherebyMedicare sets a single reimbursement level for all drugsin designated groups. Second, Medicare could use competitivebidding strategies, either contracting with manufacturersdirectly or indirectly, through pharmacybenefit managers (PBMs), PDPs, or health plans.Indirect contracting could encompass a range of possibleservices, from Part B drugs only to comprehensivehealth benefits.

A discounted list price approach, such as AWP minus5% or 15%, has the obvious advantage that list prices areeasily obtained from pricing services. A less obvious butkey economic advantage is that this approach encouragesprice competition below the prevailing reimbursementlevel, as manufacturers seek to increase thespread to providers, as discussed above.

Two modifications of Medicare's traditional AWP-basedstrategy are necessary to make it attractive. First,Medicare should have the flexibilityto adjust the discountover time to keep reimbursementroughly in line withacquisition costs, as determinedby audit of markettransactions, and enableMedicare to share in the savingsfrom price competition.Second, the rate of increase inAWP should be constrained,for example, to the rate ofincrease in the consumer priceindex or producer price indexfrom a predetermined date, oran excess-inflation penaltycould be imposed. These modificationswould reduce thepotential for manufacturers toincrease the spread by raisingthe list price rather than cuttingthe acquisition price.

An obvious disadvantage ofany list price, including AWP,is that it may deviate substantiallyfrom actual transactionprice. However, if transactionsprices are monitored quarterlyand the discount percentageadjusted to capture the spread for the payer, as occursin the private sector, the resulting payment level wouldapproximate transactions prices with a lag, and theaudit burden would be less than under a transactionsprice approach. Of course, even if margins are appropriateon average, both positive and negative deviationsare possible for individual products, which could distortprescribing choices. However, this is true for everyadministered pricing rule, whether based on list ortransactions prices, and for diagnosis-related groupreimbursement for hospitals and resource-based relative-value scales for physicians. As long as the rate ofincrease of list prices is constrained, differences in thespread across drugs should reflect differences in manufacturerprice cuts, not in manipulation of the list price.

An alternative list price that appears to be moregrounded in acquisition costs is the WAC. Some statesalready use WAC plus a markup as an option to reimbursepharmacies under Medicaid. The wholesale acquisitioncost is a manufacturer catalogue price towholesalers and is also published by pricing servicessuch as First Data Bank. However, if Medicare were toadopt WAC as a basis for reimbursement, exactly thesame incentives that have distorted AWP would applyto WAC: manufacturers would have incentives to raisethe listed WAC and then to offer discounts below WACto increase the provider's spread. Thus, any manufacturer-determined list price will face the same potentialfor distortion as AWP, unless constrained by an excess-inflationpenalty.

Under a discounted AWP approach with consumerprice index or producer price index constraints, regulationsshould clarify that selling below the list priceshould not be grounds for charges of fraud and abuse.Given the 2 provisions proposed here, that the increasein AWP is constrained and the payer captures thespread by adjusting the discount factor, then any marginbetween the AWP and the actual selling price canbe presumed to reflect competitive discounting and isto be encouraged.

In general, reimbursement to physicians or pharmaciesfor dispensing costs is better done through a fixeddispensing fee, rather than by paying a percentage of thedrug price. Dispensing and overhead costs usually do notvary with the price of the drug. Paying a percentage ofthe drug price encourages use of higher-priced drugs.

Setting reimbursement based on a transaction priceplus a markup rather than a list price is intended toground reimbursement more firmly on the provider'sactual acquisition cost. With this objective in mind,starting on January 1, 2005, Medicare will pay for PartB drugs based on ASP plus 6% or a widely available marketprice. A drug's ASP is defined as the average salesprice to all purchasers, excluding certain public purchasers,and is to be reported quarterly by manufacturersto CMS, with significant penalties for misreporting.

The fundamental concern with this approach is thatit undermines manufacturers'incentives to compete onprice. At the limit, if a reduction in price is perfectlymatched by a reduction in reimbursement, the physicianhas no incentive to use the lower-priced drug andthe manufacturer has no incentive to reduce price;rather, the manufacturer may have an incentive toraise price.

This tendency for ASP-based reimbursement, likeany cost-based reimbursement, to lead to price increasesmay be attenuated by 2 factors. First, because reimbursementto all physicians is based on average salesprice, not the physician-specific sales price, the manufacturermay still have an incentive to discount to highlyprice-sensitive physicians. However, this incentive ismuted because ASP is a volume-weighted average price.Second, if private payers continue to reimburse physiciansbased on discounted AWP, then the incentive toraise prices to maximize reimbursement from Medicarepatients may be constrained. However, given Medicare'sdominant market share for most Part B drugs, the likelyeffect of ASP-based reimbursement is to reduceincentives for price competition, leading to higherprices to both private payers and Medicare than wouldoccur under our proposed constrained-AWP-minus-X%,where X% is revised based on actual margins.

The manufacturer's incentives to discount would beslightly less under our proposed flexible discount offAWP rule, compared to the current fixed 5% discount offAWP, to the extent that manufacturers anticipate that adiscount granted in period 

may lead to a larger reimbursementdiscount off AWP in 

+ 1. However, this negativeeffect on incentives to discount is weak becausethe reimbursement discount off AWP is an average overall products. By contrast, the negative effect on incentivesto discount is much greater under the ASP rule,because each drug's ASP reimbursement in 

+ 1depends on drug-specific discounts in period 

A similar reduction in manufacturer discounts to privatepayers occurred after Medicaid's adoption of theOmnibus Budget Reconciliation Act of 1990, whichrequired manufacturers to give a rebate to Medicaidequal to the greater of 15.1% of average manufacturerprice (AMP) or AMP-"Best Price"to private buyers.

Administered pricing rules based on either list ortransaction prices could be combined with internal referencing, whereby Medicare would set a single reimbursementfor all drugs in the designated group.

Maximum allowable charge reimbursement for multisourcedrugs, as already used by Medicaid and manyprivate payers, exemplifies generic referencing. Weendorse the MMA's adoption of MAC reimbursementfor multisource drugs under Parts B and D. Genericreferencing offers significant cost savings to payerswithout significant health risk to patients or disincentivefor R&D.

More controversial is therapeutic referencing, whichapplies the same reimbursement to different compoundsin a therapeutic class—for example, all &#946;-blockers—with the reimbursement level usually set at thelowest price in the class. Manufacturers may charge aprice above the reference price, but the patient mustpay the difference. The recent attempt by CMS to reimbursesimilar products (such as Procrit and Aranesp) atthe same rate based on "functional equivalence"is therapeuticreferencing in all but name. If applied to drugsthat are equivalent in all relevant dimensions, thentherapeutic referencing is consistent with cost-effectivenessprinciples; however, if applied without rigorousevidence, then therapeutic referencing may deny paymentfor real improvements in treatments, for at leastsome patients, and undermine incentives for researchin such improvements. Of course, the differencebetween real improvements and minor variations is afine line. However, such judgments should be based onempirical evidence, given their potential effects onincentives for innovation. The MMA prohibits the use offunctional equivalence unless it was in place prior to thelegislation's passage.

Competitive Contracting for Medicare Part B Drugs

Competitive contracting approaches seek to avoidthe need for Medicare to define administered pricingrules. Under a direct purchase approach, Medicarewould procure drugs directly from manufacturers, asdoes the Department of Veteran's Affairs (VA).However, the VA direct procurement model is impracticalfor Medicare. The VA delivers services through itsown, compact provider system. By contrast, Medicareuses the private sector delivery system, includingthousands of private physicians and pharmacies whomust be reimbursed for the drugs that they acquireand dispense.

Under indirect contracting for Part B drugs, asplanned starting in 2006, Medicare will contract withspecialty pharmacies, distributors, or PBMs to contractwith manufacturers, negotiate prices and deliver drugsto physicians. Specialty pharmacies currently handlemost Part B drugs that require special handling, such aspatient-specific dosing and just-in-time delivery tophysician offices; however, chain pharmacies and PBMsare increasingly entering this market. The MMA provisionsfor Part B drugs will thus provide a test—albeit aspecialized one—of the indirect competitive contractingmodel for Medicare drug benefits.

Under the MMA indirect contracting approach, specialtypharmacy contractors for Part B drugs will bereimbursed for a drug at the lowest bid price of anycontractor in the area, plus an administrative fee fordrug handling and delivery. Contractors on Part Bdrugs are not authorized to engage in active formularymanagement and will not be at risk for total drugexpenditures. Whether this indirect contractingapproach, with contractors as passive administrators,can yield acceptable control over drug prices is far fromcertain. The leverage of private sector PBMs in negotiatingdiscounts with manufacturers depends criticallyon their use of tiered formularies to shift market shareto preferred drugs. If Part B drug contractors must offera totally open formulary, undifferentiated by tieredcopayments, manufacturers would have little incentiveto offer price discounts except to the extent that thebeneficiaries'25% copayment creates some price-sensitivityof final demand. If a manufacturer offered discountsto a particular contractor, this might increasethe likelihood that contractor would win the Medicarecontract but would have no effect on patient or physicianbehavior and hence no effect on the manufacturers'market share.

Thus an indirect contracting model with specialtypharmacies/PBMs that lack discretion to design restrictiveformularies is likely to be ineffective at stimulatingprice competition between manufacturers. Suchmanacled indirect contracting is likely to generate lesscompetitive pressure on drug prices than a constrained-AWP-minus-X% administered pricing rule, which preservesmanufacturers'incentives to cut prices toprescribing physicians while capturing much of the savingsfor Medicare, provided that the discount percentageis revised periodically.

In fact, if Medicare adopts indirect contracting withspecialty pharmacies/PBMs but without formularies tocontrol prices, Medicare would probably still have todefine an administered pricing rule in order to determinereasonable reimbursement to the PBM contractorsfor the drugs they procure, if passive pass-through ofdrug prices is to be avoided. Similarly, competitiveselection of intermediaries and carriers to administerhospital and physician services under Parts A and B hasnot relieved Medicare of the need to define administeredpricing rules that these intermediaries apply inpaying providers. An alternative reimbursement rule fordrugs would be to pay the lesser of the lowest indirectcontractor price or constrained-AWP-minus-X%. Underthis approach, Medicare could experiment with competitivecontracting, but would be no worse off thanunder the revised AWP rule.

Competitive Contracting for a Comprehensive

The MMA gives the PDPs some flexibility in formularydesign for Part D drugs and places them at partial riskfor Part D drug spending. Thus if administration of PartB is merged into Part D, the Part D formulary flexibilityand associated incentives and ability to constrain pricesmay be extended to Part B. However, it remains uncertainwhether PDPs will be willing to participate as riskbearingcontractors if they are required to offer openenrollment despite substantial adverse selection risk,have limited ability to constrain formularies and otheraspects of benefit design, and are limited in their abilityto raise beneficiary co-payments or premiums.However, if PDPs are constrained in benefit design orare only minimally at risk for part of the administrativefee, with drug costs largely passed through toMedicare, then Medicare may again resort to administeredpricing rules in order to limit its expenditures toacceptable levels.

Competitive Contracting for Comprehensive

Alternative Strategies for Medicare Payment of Outpatient Prescription Drugs"Part B and Beyond

Neonatal intensive care unit admission rates arean important birth outcome indicator for Medicaid managed careorganizations.

To reduce neonatal intensive care unit admissionrates by at least 15% and to maintain that reduction through implementationof a quality improvement program.

The organization performed a longitudinalpopulation-based review of its birth outcomes from 1997 through2003, focusing on neonatal intensive care unit admission rates.The return-on-investment evaluation reflected attributable incrementalprogram costs and resultant savings.

Interventions included enhanced identification andstratification of high-risk women with the use of a health riskassessment form; outreach through nursing care coordination offeringhome visits, transportation, support services, social work services,and connection with other community-based organizations;and implementation of a strong informatics structure.

Neonatal intensive care unit admission rates decreasedfrom 107.6 per 1000 births in 1998 to 56.7 per 1000 births in2003. The return on investment from the incremental programenhancements was just over $2 per $1 expended.

A program that identifies its high-risk pregnantenrollees in a timely fashion, provides outreach using a strongnursing care coordination and social work emphasis, and has anenhanced informatics structure can significantly affect birth outcomesfor a Medicaid managed care population.

Pregnancy is one of the primary events that lead toeligibility for Medicaid, and deliveries account foralmost 50% of Medicaid inpatient discharges.

Women from lower socioeconomic groups experiencepoorer birth outcomes than those from higher socioeconomicgroups.

Consequently, delivery claims andhigh-cost neonatal intensive care unit (NICU) expensesconsume a large portion of Medicaid managed care medicalexpense budgets, despite advances in perinataltechnology in the United States. With the shift of mostMedicaid-eligible individuals to Medicaid managed care,enhancing birth outcomes becomes a major challengefor any Medicaid managed care plan.

Monroe Plan for Medical Care (MP), an independentpractice association with more than 3000 providers inthe Rochester region, partners with Excellus BlueCrossBlueShield, Rochester, to serve as Excellus BlueCrossBlueShield's delivery system for publicly financed programstargeting underserved populations. It providescare for nearly 48 000 Blue Choice Option (Medicaidmanaged care) enrollees in a program that covers thecategories of individuals that include women and children(Temporary Assistance to Needy Families), adultswho are unable to work (Safety Net), and a segment ofthe disabled populations (Supplemental SecurityIncome). Monroe Plan for Medical Care is the dominant(70% of the market share) provider for Medicaid managedcare in this region. It is the exclusive communityprovider for 14 000 enrollees in Family Health Plus, anexpansion of the Medicaid managed care program in NewYork State for working-poor individuals. Finally, MP is theexclusive community provider of care for 11 000 childrenenrolled in Child Health Plus (New York State'schildren's insurance program).

In its relationship with Excellus BlueCross BlueShield,MP is financially responsible for the provision of all coveredmedical care services for enrollees in these publiclyfinanced programs. It also is accountable for care management,disease management, and quality improvementactivities that target the enrollees in these programs.

Before 1998, Medicaid managed care was an optionalprogram for Medicaid recipients. During 1998, themost populous counties in MP's service area requiredmandatory enrollment of individuals covered byMedicaid in the Temporary Assistance to NeedyFamilies and Safety Net categories into Medicaid managedcare programs. Consequently, MP's overall enrollmentincreased from about 28 000 enrollees in 1998 tomore than 73 000 in 2003, with a resultant increase inthe number of births annually from about 600 in 1998to more than 1300 in 2003.

During the late 1990s, the NICU admission rates forMedicaid, including MP, consistently were considerablygreater than 100 per 1000 births. In 1997, there was arudimentary prenatal and perinatal case managementprogram that consisted of a part-time nurse case manager,ad hoc use of community support services (includinghome-based prenatal education), and aninconsistent approach to case finding. It made sense toimplement a more comprehensive approach to prenatalcare, because a reduction in NICU admission rateswould favorably reduce overall costs for deliveries. Thiswould provide an opportune area in which to achieveimprovements in terms of patient care and outcomesand medical expenses.

The medical literature reports that there are manyrisk factors that significantly affect birth outcomes forlow-income and working-poor women, including medicalcomorbidities, mental health and substance abuseissues, smoking, previous preterm birth, and social-relatedproblems such as social isolation, spousal abuse,and homelessness.

The birth outcomes database ofMP (obtained through quarterly facility medical recordreviews) identified social issues as correlating stronglywith poor birth outcomes in comparison with other riskfactors. Because comprehensive care through outreach,coordination, and education of patients seems to havedemonstrated improvements in outcomes,

MPbelieved that early identification of risk factors, withsubsequent coordinated interventions, would hold thegreatest promise in mitigating the effects of risk factorsin terms of birth outcomes.

Although MP's enhanced quality improvement effortsbegan in late 1997, the Center for Health CareStrategies invited MP in 2000 to participate in its BestClinical and Administrative Practices "Toward ImprovingBirth Outcomes"program. This was a nationwidework group of 11 Medicaid managed care entitiescommitted to developing and pilot-testing best-practicemodels. Best Clinical and Administrative Practices providedMP a framework in which to implement qualityinitiatives, using its typology of identification, stratification,outreach, and intervention. Furthermore, the collaborationwith Best Clinical and AdministrativePractices allowed MP to network and to identify bestpractices (and failed efforts) from other Medicaid managedcare entities.

During the literature review and ongoing interactionswith other Medicaid managed care organizations, it wasevident that there was no one magic bullet for improvingbirth outcomes. Consequently, MP decided that sustainedimprovement requires change in the care deliverysystem

to assist practitioners in doing the right thing atthe right time. Monroe Plan for Medical Care thus adopteda quality improvement approach for its prenatal careimprovement activities that calls for the use of learningcycles to plan and test changes in systems and processes.

Such cycles have been referred to as Plan-Do-Study-Act cycles, which will guide improvement teamsthrough a systematic analysis and improvement process.As part of this process, MP solicited the professionalinput of its Obstetrics/Gynecology Advisory Committeein the development of its prenatal care program"Healthy Beginnings."The objectives of the programwere to reduce NICU admission rates from baseline in1998 by at least 15% during the subsequent 3 years andto maintain that reduction in the following years.

This program is part of MP's quality improvementprogram to enhance healthcare outcomes for all of itsenrollees. The focus for any such quality improvementprogram is to institute organizational system changes toensure adherence to appropriate practice guidelinesthrough the coordination of care. This approach emphasizesorganizational and care delivery improvementsusing existing standards of care. Consequently, therewas no randomization of enrollees into intervention orcontrol groups, and the services provided were availableto all eligible enrollees, who at all times were able torefuse or terminate any services offered. Because theseactivities are essentially organizational system changes,institutional review board approval and participants'informed consent were not sought or required.

Before 1997, practitioners rarely notified MP aboutthe pregnancy of enrollees (notification would occur foralmost 3% of pregnant women). In late 1997, MP developedits prenatal registration form (PRF) to serve as ameans for practitioners to notify the program when anenrollee was pregnant and to provide the program witha health risk assessment for each woman. The PRFassesses risk categories of social risk factors, maternalmedical history, psychoneurological history, maternalobstetrical history, and previous infant findings. Duringthe rollout for the PRF, MP began to reimburse practices$30 for the submission of the PRF; alternatively, if thePRF was not submitted, the practitioner could potentiallylose the prenatal care reimbursement. This actionresulted in PRF submission rates of 85% in 1998 andsubsequent annual rates in the 88% to 98% range.

Although the submission of PRFs reached 90% andhigher, challenges remained regarding the timeliness oftheir submission. Often, practitioners would submitPRFs during the late third trimester, when the ability tomitigate any significant risks would be at a minimum. InApril 2001, Healthy Beginnings implemented a tieredpayment system for the submission of the PRF in whichthe program would pay practitioners $50 for submissionin the first trimester, $30 in the second trimester, and$20 in the third trimester. In addition, program staff visitedpractitioner offices to educate personnel regardingthe submission of the PRF and its importance in assistingthe practitioners in managing high-risk pregnantwomen. This intervention resulted in submission ratesof the PRF within the first trimester that were consistentlyin the 60% and higher range.

The PRF serves as an invaluable tool to stratify therisk for pregnant enrollees. Staff members input the PRFinto the care management database, which then scoresthe reported findings to reflect the risk for each patientand to engage members in needed medical, behavioralhealth, and social and support services as identified. Acommittee of obstetrical practitioners and the local perinatalnetwork developed the scoring system to stratifypatients at risk throughout the community. This scoringsystem serves as an adjunct for care coordination decisionsand is not used as admission criteria for the NICU.

During 2000, staff discovered that there was no consistentreview process in place for addressing behavioralhealth issues identified through the PRF. In November2000, the program instituted an integrated reviewprocess between the behavioral health and clinical medicalmanagement staff that now ensures that behavioralhealth staff address all mental health and substanceabuse issues reported on the PRF and work to engagepatients in necessary care.

Through the quality improvement process, HealthyBeginnings has evolved its approach to outreach, fromusing generalized community outreach services (thelocal county community healthcare worker programand contracted home health agencies), to instituting atrial of using its own prenatal outreach workers, tofinally engaging outreach services through the localBabyLove Program in early 2002. Whenever theHealthy Beginnings perinatal nurse coordinator identifiesmembers at moderate-to-high risk through thePRF, the coordinator manages these individualsthrough communications with the practitioners, outreachprograms (such as the county's MedicaidCommunity Health Worker program), and referral toMP's internal social work program as needed. Individualswith medical complications of pregnancyreceive complex case management, home care services,or skilled nursing services as required.

The perinatal nurse coordinator refers all pregnantenrollees identified as high risk because of psychosocialproblems to the BabyLove Program. This community-basedprogram has a strong history of working effectivelywith high-risk pregnant women, with the addedfeature of social work supervision that is necessary toeffectively provide outreach. The BabyLove Programoffers home visits, arranges transportation, provideslinks to support services and social work services, andconnects high-risk pregnant women with other criticallyneeded services.

In early 2003, MP engaged the services of an additionalBabyLove Program outreach worker to addressthe needs of depression in pregnancy. Later in 2003, MPadded its own social worker to support the HealthyBeginnings clinical staff in addressing social problems.

With enhanced outreach, MP has been able to moreeffectively connect its pregnant women with medical,mental health, chemical dependency, community-based,governmental, and social services. Before 2001,MP stored PRF data in an internally developed MicrosoftAccess (Redmond, Wash) database; however, thisapproach to information systems did not support caremanagement activities. In 2001, MP installed a commerciallyavailable care management software system(CaseTrakker; IMA Technologies, Sacramento, Calif) tosupport prenatal and perinatal care activities. The systemidentifies risk factors and scores the PRF, providesmember demographics, identifies related practitioners,provides progress notes, creates reminders and ticklersfor care management activities, stores birth outcomedata, creates reports, and provides an interface forcomorbidity issues. The system supports care managersand social workers by linking care management activities,risk factors, and outcomes associated with patients.As now implemented, it is not accessible to providersand does not provide a general electronic medicalrecord; rather, it focuses on structuring the contacts thatthe care managers have with patients, practitioners, andcommunity agencies, based on identified risk factors.

The measurement for program effectiveness is theNICU admission rate for all pregnant women in MP.Because any significant programmatic changes firstoccurred in late 1997 and early 1998, and given thatpregnancy is usually a 9-month phenomenon, the baselineyear for NICU admissions is 1998. In terms of programmatic costs, 1997 serves as the baseline year, withincremental new program costs reflected in the followingyears.

demonstrates, the NICU admissionrates have progressively decreased relative to the 1998baseline rate of 107.6 per 1000 births as MP implementedand improved the prenatal care program.Concomitantly, the NICU admission rates for Medicaidrecipients in upstate New York have remained essentiallythe same during the same period (M. Whitbeck,Finger Lakes Health Systems Agency, Rochester,unpublished data, 2004). The NICU admissions andtheir associated costs in this analysis also include thoseinfants weighing less than 1200 g. The costs for theseinfants were shifted out ("carved out") of Medicaid managedcare to New York State Medicaid fee for servicebeginning in 1999; however, these children and theircosts were included in the results and analysis. Theseprogram results have exceeded the original projectobjective of achieving a 15% reduction in the NICUadmission rate.

Based on the 1998 NICU admission rate of 10.8%, arate of 9.1% would have to be obtained to achieve thecorporate objective of a 15% reduction in rates.However, to obtain a statistically credible reduction ofat least 15% requires rates to fall below 6.8% to accountfor the potential effect of random variation. This is therate calculated to show at least a 15% reduction statisticallysignificant at 

<.05 (&#945;= .05, 80% power) for abinomial test. Although there were reductions in NICUadmission rates during 1999 and 2000, the rates inthose years exceeded 9.1%. From 2001 through 2003,the rates were 8.8%, 8.9%, and 5.7%, respectively. A1-sided exact binomial test was used to verify whetherNICU admission rates were 15% belowthe 1998 rate. The rate during 2003was statistically significant (

< .01).Arguably, this is conservative becausethis requires a one-third reduction inNICU admission rates. Using theweaker criterion of simply showing astatistically significant reductioncompared with 1998, the rates from2001 through 2003 were statisticallysignificant at 

In support of the observeddecrease in NICU admissions, 2 otherbirth outcome measures (gestationalage < 32 weeks and birthweight <1900 g) demonstrated trends in thesame direction. The rate of birthswith a gestational age younger than32 weeks decreased from 2.9% in2001 to 0.9% in 2003, and the rate of births with abirthweight less than 1900 g decreased from 6.1% in2001 to 1.6% in 2003. Birthweights less than 2500 galso decreased but less dramatically, from 10.2% to7.6%. The mean birthweight and gestational ageremained essentially flat during this period; however, areduction in the tails of these distributions wasobserved. Although the mean birthweight and gestationalage did not change, the reduction in extremebirthweights and gestational ages is consistent withreduced NICU admissions.

From 1998 through 2003, criteria for NICU admissionremained unchanged. Therefore, within the communitythere was no reduction in the number ofindications for NICU admissions, nor was there anyredirection of children who would be candidates forNICU admission to other units. Analysis of New York'sStatewide Planning and Research Cooperative Systemdatabase, which captures all payer hospital data,demonstrated no concurrent changes in NICU admissionrates in upstate New York for Medicaid patients (feefor service or managed care) through 2002 (M.Whitbeck, Finger Lakes Health Systems Agency,Rochester, unpublished data, 2004). Therefore, itappears that there were no external effects on NICUadmission rates to explain the observed change in ratesfor MP.

Implementation of a comprehensive care managementprogram for prenatal care is not a trivial financialmatter, particularly given that there has not been anyone approach documented in the medical literature thatis deemed the magic bullet. A health plan must demonstrate that whatever care or disease management programit implements will improve health outcomes andyield an economic return. Resources are scarce for mostMedicaid managed care programs, and it is imperativefor them to measure the effectiveness of their effortsrather than to simply continue them because they seemright. This is where measuring and analyzing the returnon investment for a program are important.

For the purposes of this analysis, MP used the return-on-investment methods developed by AmericanHealthways Inc and Johns Hopkins University.

Thisapproach examines the ratio of the realized savings tothe incremental program (prgm) costs, as shown in thefollowing formula:

In this case, medical costs (Med Costs) are the NICUcosts, with the MP NICU costs in 1998 serving as thebaseline for expected future years'NICU costs. The differencebetween actual costs and the projected costsbased on the baseline year provides the value for thenumerator of the ratio. Program costs for the denominatorinclude reimbursement costs for submissionof the PRF, Medicaid Obstetrical and Maternal Servicesprogram and Community Health Worker programexpenses (primarily home-based prenatal educationservices), care management software expenses, externaloutreach services, internal personnel costs (full-timeemployees), and allocated organizational overhead.

Using the aggregate NICU cost savings (1999 through2003) of $1 875 463 (

) and the incremental prenatalprogram costs (1998 through 2003) of $924 300(

) in the return-on-investment ratio yields aratio of 2.03, where for each $1 spent the enhancedprenatal care program provided a cost savings to theorganization of just over $2. The ongoing rise in administrativecosts per birth for the Healthy Beginnings programreflects the growth in the pregnant population, whichrequired an increased investment in program infrastructureand the shift to a "high-touch"approach to outreachand to interventions that would require the addition ofmore personnel (a high-cost expense) to the program.

Despite the rising administrative costs for the program,the substantial savings realized by the program(with a demonstrated savings-to-cost ratio of 2.03)makes a strong business case for the Healthy Beginningsprogram. One could hypothesize that fewer NICUadmissions potentially indicate overall healthier infantsat birth for this high-risk population, resulting in animproved quality of life for these children and their parentsand an overall societal benefit. The observedreduction in the rates of infants with gestational agesyounger than 32 weeks and infants with birthweightsless than 1900 g would support this. More in-depth testingof this hypothesis could be the subject of anotherstudy looking at child developmental metrics.

Monroe Plan for Medical Care contends that theimplementation of its enhanced prenatal care program,Healthy Beginnings, in late 1997 and early 1998 hasresulted in a marked decrease in its NICU admissionrates. Other than this program, there were no knownexternal forces that would have caused a drop in theNICU admission rates, such as a change in NICU admissioncriteria or coding changes or evidence that theoverall population experienced a drop in rates. In 1998,the larger counties that MP serves (with most of itsenrollees) converted to mandatory enrollment intoMedicaid managed care for their Medicaid beneficiaries.This would mitigate any argument that the pregnantwomen enrolled within Medicaid managed care wouldbe healthier than those in fee-for-service Medicaid.

There does not appearto be a consistent changein case-mix indicatorsthat would cause theobserved changes inNICU admission rates.Use of a diagnostic costgroup case-mix adjuster(DxCG, Inc, Boston,Mass) showed essentiallyno change in the relativerisk for women eachyear, and the mother'smean age at deliveryremained the same atabout 25 years. Trendsfor other risk factorswere mixed, with incrementaldecreases innumbers of individualswith social risk factors,while there wereincreases of similarmagnitude in the percentageof women withprevious low birthweightpregnancies orpreterm labor.Interestingly, the rate ofwomen entering prenatalcare during theirthird trimester decreased from 13.0% in 2001 to 7.7% in2003. Late entrance into prenatal care is a risk factorthat the program has sought to minimize throughprovider incentives and patient education.

March 2005

Correspondence regarding, "Cost Effectiveness of Thromboprophylaxis With a Low-Molecular-Weight Heparin Versus Unfractionated Heparin in Acutely Ill Medical Inpatients"

Correspondence regarding, "Asthma Disease Management: Regression to the Mean or Better?"

Variation in Medicare"s Local Coverage Policies: Content Analysis of Local Medical Review Policies

Alternative Strategies for Medicare Payment of Outpatient Prescription Drugs"Part B and Beyond

Reduction in Neonatal Intensive Care Unit Admission Rates in a Medicaid Managed Care Program

Physician Evaluations of Care Management Practices in Medicaid Programs

Achieving Value in Healthcare

Can an Algorithm for Appropriate Prescribing Predict Adverse Drug Events?

The Impact of Medicare Modernization on Coverage Policy: Recommendations for Reform