Health Plans' Coverage Determinations for Technology-based Interventions: The Case of Electrical Bone Growth Stimulation

December 1, 2004
Alison J. Huang, MD, MPhil
Alison J. Huang, MD, MPhil

,
Marcel P. Gemperli, MD, MPH
Marcel P. Gemperli, MD, MPH

,
Linda Bergthold, PhD
Linda Bergthold, PhD

,
Sara S. Singer, MBA
Sara S. Singer, MBA

,
Alan Garber, MD, PhD
Alan Garber, MD, PhD

Volume 10, Issue 12

Objectives: To determine (1) whether commercial health plans'coverage criteria for a costly technology-based medical interventionare consistent with recent clinical effectiveness evidence, (2)whether medical directors adhere to planwide coverage criteriawhen making coverage determinations for individual patients, and(3) if any organizational characteristics are associated with havingmore stringent coverage criteria or making more frequent coveragedenials.

Study Design: Case-based survey of medical directors of UScommercial health plans.

Methods: A close-ended survey was mailed to 346 medicaldirectors meeting eligibility criteria, asking about the criteria specifiedin their plans' coverage policies for electrical bone growth stimulation(EBGS) and whether they would cover this intervention for ahypothetical patient with abnormal union of long-bone fracture.

Results: Responses from 228 (66%) of the 346 directors indicatedthat approximately 72% of plans have a formal coverage policyfor EBGS for long-bone fractures. More than 30% of plansspecify that longer than 4 months must elapse before EBGS isattempted, although clinical studies do not support absolute waitingtimes. Directors of approximately 61% of plans with policiesrequiring extended waiting periods would nevertheless authorizeEBGS for patients who did not meet this criterion.

Conclusions: Health plans apply varied criteria in coveragepolicies for technology-based treatments such as EBGS, but do notalways adhere to stated criteria when determining coverage forindividual patients. For-profit status, accreditation status, geographiclocation, and size of plan are not associated with beingmore or less likely to authorize EBGS.

(Am J Manag Care. 2004;10:957-962)

Since managed care emerged as the dominantmode of healthcare delivery in the United States,there has been widespread concern about the wayhealth plans make coverage determinations about technologicallyadvanced medical interventions. One strategyused by plans to regulate use of these often costlyinterventions is to issue coverage policies containingspecific criteria for determining whether interventionsare medically necessary for patients with particularindications.

Although some see coverage policies primarily as atool for restricting access to care, others argue that theycan be used to promote more consistent and evidence-basedpractice across plans.1 At least in theory, theseutilization review tools can provide the means by whichindividual physicians'clinical judgments are subjectedto more systemwide standards.

To date, there are limited data to indicate whethercoverage criteria promote consistency or the applicationof evidence-based standards. Previous work suggeststhat medical directors rely to varying extents oncoverage policies, published scientific evidence, expertopinion, community standards, personal experiences,and patient preferences when making individual coveragedeterminations.2

To investigate this issue, we presented medical directorsof commercial health plans with a hypotheticalrequest for authorization for a costly technology-basedintervention. We sought to determine (1) whetherplans'coverage criteria for this intervention are consistentwith the most recent clinical effectiveness evidenceand (2) whether plans apply criteria consistently whenfaced with the ambiguities and exigencies of individualcases. We also examined whether organizational characteristicssuch as plan size, tax status, and accreditationstatus are predictive of plan behavior in this area.

Our research group, after consultation with experts,selected electrical bone growth stimulation (EBGS) asthe focus of our study. We considered EBGS to be anideal technology for a study of coverage decision makingbecause it raises many of the issues that can lead tovariable coverage policy. Despite its use for more than 2decades to promote healing in fractures that fail conservativetherapy, EBGS has long been in an "evidencelimbo,"as the medical community did not reach consensusabout appropriate selection of fractures for thistreatment until recently.

Although early research was conducted in patients inwhom conservative measures had been attempted for 9months or longer,3 investigations after 1990 began to relyon serial radiographs to confirm nonunion rather thanabsolute waiting times.4,5 In 1999, the Health CareFinancing Administration revised Medicare criteria tostipulate that EBGS was appropriate for patients in whomserial x-ray films confirm that fracture healing has ceasedfor 3 months or longer.6 Therefore, EBGS has had proponentsand opponents for many years, with potential forthe introduction of new evidence about clinical effectivenessto lead to agreement on covered indications.

In addition, like many other technologicallyadvanced medical interventions, EBGS is costly. Retailprices for external, noninvasive devices range from$3500 to more than $4000; worldwide sales of thesedevices reached nearly $200 million in 1998.7 The costof EBGS is likely to be at least an implicit considerationamong payers.

Finally, EBGS has not figured widely in publicdebates about medical cost cutting and managed care.Many other technologies for which there is limited evidencehave been the subject of controversy, whichmight inhibit medical directors from candidly expressingtheir views.

We expected different health plans to develop differentcoverage criteria for EBGS. We anticipated that fewplans would include criteria that were no longer consistentwith recent evidence on the appropriate indicationsfor EBGS. Plans that were affiliated with nationalor multistate organizations were expected to specify criteriathat were more consistent with recent evidence.Tax status, accreditation status, and plan size were notexpected to be predictors of having more evidence-basedcriteria.

We expected that variability in the clinical criterialisted in plans'coverage policies would translate intovariability in their approach to day-to-day coveragedeterminations. That is, we expected medical directorsto adhere to written coverage criteria when decidingwhether to cover EBGS for the patient in our studycase, with 80% adherence being our threshold for consideringmedical directors to be consistently adherent.

METHODS

Sample Selection

We identified a set of characteristics that defined a"commercial managed healthcare organization"for thepurposes of studying coverage decision making. To beeligible, organizations needed to (1) provide and financea comprehensive range of general healthcare services,(2) offer and include enrollment in at least 1 managedcare product line (ie, health maintenance organization[HMO], preferred provider organization, or point of serviceplan), (3) have been continuously marketing for atleast a year (ie, since January 1, 2000), (4) assume financialrisk for the medical services they provide, and (5)have at least 10% commercially sponsored beneficiaries.

Preliminary review of published directories of managedcare organizations suggested that many multistateorganizations could be subdivided into several "planunits"that made independent coverage decisions forbeneficiaries in different states or regions.8-10 In suchcases, we considered each plan unit (hereafter referredto as a "plan") to be a separate decision-making entityfor the purposes of our analysis. We relied on organizations'self-report to identify all of their affiliated plans; ifan organization indicated that decision making wasmade centrally for all regions in which it operated, weconsidered the entire organization to be a single plan.We identified plans that were affiliated with national ormultistate organizations to assess whether their behaviordiffered from that of nonaffiliated plans.

Within each plan, we identified a single individual toanswer questions about decision making for all productsoffered for that plan. We generally sought the most senior-level medical director who was responsible for decisionmaking for the largest product in the largest statewithin that plan and who had the authority to evaluatemedical interventions for coverage on a daily basis. Weasked medical directors to indicate whether theiranswers to our questions would have differed if theywere responding on behalf of a different product or statein the plan. Medical directors were encouraged to consultwith other staff members when necessary to provideaccurate information.

Names and addresses of the medical directors wereverified by telephone before mailing of the survey inmid January 2001, and ineligible plans were removedduring the telephoning process. A reminder postcardwas sent 3 weeks after the initial mailing, and 3 subsequentrounds of reminder telephone calls were madeuntil early May 2001.

Questionnaire Development

Survey questions were developed in consultationwith 20 members of the managed care industry,including several former medical directors of healthplans. The first part of the survey elicited informationabout general organizational characteristics of plans,including geographic area of operation, number of coveredlives, range of products, physician payment methods,tax status, accreditation status, affiliation with anational or multistate organization, and use of qualityand utilization management strategies.

Medical directors were also presented with a hypotheticalcase involving a request for EBGS for a patientwith a long-bone fracture that had not healed with conservativemanagement:

A 55-year-old man suffered a partially displacedfracture of the right tibia after falling down the stairs.Shortly afterward, he was seen by an orthopedic surgeonin a local emergency room. The surgeon feltthat open reduction and internal fixation was notmedically indicated and applied a cast to thepatient's right leg. The patient initially appeared torecover well, but follow-up x-rays over the next severalweeks revealed very little callus formation at thesite of the fracture. After 3 months, the surgeondetermined that healing was delayed and not progressingat a satisfactory rate. She recommended anexternal (noninvasive) electrical bone growth stimulatorbecause she felt it was clear that satisfactoryunion would not be achieved in 6 months based onthe clinical course of the first 3 months.

Participants were asked to indicate whether theirplans would consider the intervention to be medicallynecessary for the patient described.* Participants couldindicate whether the decision would be made by a medicaldirector or whether the treating physician would beallowed to make this decision as a matter of policy.We avoided incorporating any details that would forcemedical directors to reject the request for obvious inconsistencieswith widely established standards of care forlong-bone fractures (eg, conservative measures notattempted or no radiographic evaluation of healing performed).At the same time, the case was designed toreflect some of the ambiguities that might characterize areal-life case (eg, the treating physician makes estimatesabout future healing based on a shorter trial period).

Through a review of 8 different EBGS coverage policiesposted on health plan Web sites, we identified 5 clinicalcriteria that are commonly used to guide coveragedeterminations for EBGS. We asked medical directorswhether their plans had any coverage policies for EBGSand, if so, whether these policies required any of the following:(1) minimum period to have elapsed after fracture,(2) serial radiographic evidence of delayed healing,(3) fracture gap within a defined range, (4) presence ofclinical signs or symptoms of nonunion at the fracturesite, or (5) absence of infection at the fracture site.

Statistical Analysis

P

We created variables to describe plans'organizationalcharacteristics; continuous variables such as enrollmentwere translated into categorical variables, inwhich plans were separated into mutually exclusiveand collectively exhaustive groups. Univariate analysiswas used to uncover any simple associations betweenorganizational characteristics of the plans and theirresponses to our questions about EBGS decision making,using the &#967;2 test to assess statistical significance.We then developed multivariate logistic regressionmodels to evaluate the independent effect of each organizationalvariable as a predictor of plan behavior andused a stepwise elimination approach to simplify models.We only report results for variables for which < .05in the final regression model.

RESULTS

Of 479 health plans initially identified, 133 (28%)did not qualify for inclusion based on informationobtained from telephone screening and further investigation.Forty-four of these plans had closed or convertedto rehabilitation care, 29 had merged withanother plan, 26 were equivalent to another planunder a different name, 19 had greater than 90% oftheir enrollment covered by Medicaid, 13 were innon-risk-bearing networks only, and 2 did not haveactive enrollment in any managed care products. Ofthe remaining 346 plans, we obtained complete or partialresponses from 228 (66%). Survey responses representedcovered lives in all 50 states and the Districtof Columbia. The total number of covered lives representedwas approximately 119 million, or 77% of thecovered lives represented by our universe of eligibleplans (estimated at 155 million).

Organizational Characteristics ofParticipating Organizations

Table 1 gives the characteristics of the respondingplans. More than half of participating directors representedplans with more than 100 000 covered lives, andalmost a quarter of directors represented plans withfewer than 50 000 lives. A third of plans reported theirgreatest enrollment in independent practice associationor network HMO products. Responding plans were fairlyevenly distributed across census regions, with proportionatelydecreased representation of regions withlow HMO penetration.8-10 Most medical directorsbelieved that their responses also applied to other productsand states in their plan; only 8% thought that theresults did not apply universally, and in our analysis weapplied their answers to onlythose covered lives in the relevantstate and product.

0

More than half of theresponding plans reimbursetheir primary care physiciansprimarily on a fee-for-servicebasis, with slightly fewer than aquarter paying mainly by capitation.Slightly more than half ofthe plans are for profit, one thirdare not accredited by any organization,and one third are affiliatedwith a national system ofplans operating in more than 5states. More than 90% of plansuse preauthorization requirementsto manage use.

We compared several organizationalcharacteristics ofrespondents and nonrespondentsusing previously publishedsources about health plans.8-10Our review suggested that planswith a larger number of coveredlives are slightly over-representedamong our respondents; 56% ofrespondents represent plans withmore than 100 000 covered lives,as opposed to 46% of nonrespondents.However, respondents andnonrespondents do not differ significantlyin terms of geographicregion, product type, accreditationstatus, or HMO penetrationof the primary state.

Health Plans'Responsesto the Case

Only 14% of medical directors,representing 8% of covered lives,indicated that EBGS would notbe considered medically necessaryfor the patient described inthe case. Approximately 77% ofmedical directors indicated thattheir plans would approve therequest. The remaining 9% ofmedical directors indicated thatthey did not know how theirplans would respond to the case.Of note, plans'decisions to coveror not to cover EBGS in this casewere not associated with any particular healthplan characteristics.

Criteria Specified in Coverage Policies

P

P

Slightly more than 72% of plans, representingslightly more than 82% of covered lives, have aformal coverage policy for EBGS for nonunionof long-bone fractures. On multivariate regressionanalysis, we found that the 2 health plancharacteristics associated with being morelikely to have a coverage policy are affiliationwith a multistate managed care system (<.01) and the use of preauthorization for certaininterventions (< .05). Tax status, accreditation, sizeof plan, and geographic region do not appear to be significantlyassociated with having an EBGS policy.

Of those plans with coverage policies for EBGS,approximately 23% of plans, representing 37% of coveredlives in this category, do not require serial radiographicconfirmation of nonunion (Table 2), althoughthe clinical literature and Medicare policy support theuse of this method of confirming nonunion over moresubjective measures.6 Slightly fewer than 84% of planswith coverage policies, representing 80% of coveredlives in this category, require that a specific amount oftime must elapse after initial fracture. No measuredorganizational characteristics predict having either ofthese requirements.

Actual time frames specified in coverage policiesvary among our data. Of those plans that require thata specific amount of time must elapse after initialfracture, 69% (90% of covered lives in this category)specify a period shorter than 4 months, as is consistentwith Medicare coverage policies for EBGS, and31% (10% of covered lives in this category) specify aperiod of 4 months or longer.* A few (3%) medicaldirectors say that their coverage policies require thatpatients must wait longer than 9 months before EBGScan be considered as an intervention. No health plancharacteristics are predictive of having more restrictivetime frames.

*As noted earlier, Medicare's coverage policy manual emphasizes the time elapsedbetween serial radiographs documenting failed healing, rather than total time elapsed sinceinitial fracture. However, this implies a minimum waiting time after initial fracture, becauseit would not be possible for patients to obtain serial radiographs performed 90 days apart if90 days had not yet elapsed since initial fracture.

In addition, a few plans include coverage criteria thatare not included in the 1999 Medicare EBGS coveragepolicy revisions. Slightly fewer than 45% of plans withcoverage policies require that the fracture site not beinfected, slightly fewer than 17% require that specificclinical signs or symptoms of nonunion be present (eg,pain or motion at the fracture site), and slightly morethan 36% set a maximum limit on the size of the fracturegap.

Comparison Between Coverage Criteriaand Case Responses

We compared medical directors'responses to thecase with the information they provided us about theirhealth plans'coverage policies for EBGS. Of those medicaldirectors representing plans with waiting periods of4 months or longer in their coverage policies, slightlymore than 60% (52% of covered lives in this category)indicated that they would consider EBGS to be medicallynecessary for the patient described in the case,despite the fact that only 3 months had elapsed. In addition,of those medical directors representing plans thatspecify limits on fracture gap size, 90% (98% of coveredlives in this category) reported that they would considerEBGS to be medically necessary, despite the fact thatno fracture gap measurement was provided.

Validation of Medical Directors'Responses

As a rough assessment of the validity of medicaldirectors'responses to our questions, we compared theinformation we obtained about coverage policiesthrough our review of plans'Web sites with the informationprovided to us by medical directors of thosesame plans in their answers to our survey questions.The overall consistency of medical directors'responsesto our questions about EBGS coverage policies (comparedwith posted health plan coverage policies) was87%, with consistency rates for individual questionsvarying from 80% to 100%.

DISCUSSION

To our knowledge, this is the first study to collectnational data about the way health plans apply coveragecriteria to day-to-day decisions about technology-basedinterventions. Our results indicate that there is significant variation in plans'coverage criteria, and these criteriaare not always consistent with the most recentclinical effectiveness evidence. If medical directors wereto adhere rigidly to planwide criteria, almost a thirdwould compel patients with fractures to wait for extendedperiods before authorizing a treatment that clinicalstudies have shown to be effective when applied earlier.

At the same time, this variation in plans'coveragecriteria does not appear to translate into variation incoverage determinations about individual patients.More than 60% of medical directors from plans withextended waiting times in their policies indicated thatthey would authorize EBGS in a case in which this criterionwas not met. Therefore, regardless of whethercriteria are consistent with available clinical effectivenessevidence, they do not appear to have a major effecton decisions about individual patients.

The closed format of the survey did not permit participantsto explain why they might approve EBGS in acase in which their plans'written coverage criteria werenot met. However, our findings were robust enough thatit is unlikely that this discrepancy between planwidecriteria and individual decision making is due to randomerror. Given the close juxtaposition of questionsabout coverage criteria and questions about the case,we believe that directors were fully cognizant of theirplans'coverage criteria when answering questionsabout our hypothetical patient.

Previous studies have suggested that different healthplan characteristics may be associated with differentapproaches to coverage decision making. Research oncoverage of laser-based technologies, for example,indicated that staff-model HMOs were likely to covermore technologies than independent practice association-model HMOs.11 A 2001 study of plans participatingin Medicaid concluded that for-profit plans are morelikely to use aggressive utilization review techniquesand deny preauthorization than not-for-profit plans.12

Our results do not support the argument that organizationalcharacteristics such as plan size, for-profit status,or accreditation status predict whether plans imposemore restrictive coverage policies or use preauthorizationmore aggressively. Although we found that plansthat are affiliated with a national or multistate systemare more likely to have a coverage policy for EBGS, noneof the characteristics that we measured were predictiveof plans'responses to our case or their coverage criteria.

To the extent that we could verify them, such as byexamining their consistency with written plan criteria,medical director responses to our survey appearedaccurate. There will always be uncertainty, however,about the validity of self-responses. Furthermore, thefindings for EBGS do not necessarily generalize to otherinterventions. Interpretation of criteria might be differentin the emotion-laden context of a potentially lifesavingprocedure for a serious illness, for example, orfor an inexpensive preventive intervention.

This study is of interest in part because it representsdecision making regarding a large fraction of commerciallyinsured Americans. The survey incorporated adiverse, national cross section of plans, with representationfrom all geographic regions and plan sizes. Byconsidering the complex organizational structure ofmanaged care organizations, we were able to capturedifferences in the way different plan units affiliated witha single multistate organization might approach coveragedecisions. A key advantage of incorporating a specificcase containing detailed clinical information,despite limits on its generalizability, is that it enabled usto explore subtle differences in plans'coverage.

In this case, existing coverage policies did not uniformlypromote consistency, nor did they all limit coveragecriteria to evidence standards. Not only do manyplans'policies contain non-evidence-based criteria, butmost medical directors appear to ignore planwide policieswhen faced with complex patients.

From the Center for Health Policy and Center for Primary Care and OutcomesResearch, Stanford University (AJH, MPG, LB, SSS, and AG), and the VA Palo Alto HealthCare System (AG), Stanford, Calif.

This study was supported by the Robert Wood Johnson Foundation, Princeton, NJ.

Address correspondence to: Alison J. Huang, MD, MPhil, Department of InternalMedicine, University of California, San Francisco, 3930 24th Street, #15, San Francisco, CA94114. E-mail: ajhuan@itsa.ucsf.edu.

Health Aff (Millwood).

1. Miller RH, Luft HS. Does managed care lead to better or worse quality of care?1997;16(5):7-25.

Health Aff (Millwood).

2. Singer SJ, Bergthold LA. Prospects for improved decision making about medicalnecessity. 2001;20(1):200-206.

Clin Orthop.

3. Borsalino G, Bagnacani M, Bettati E, et al. Electrical stimulation of humanfemoral intertrochanteric osteotomies. 1988;237:256-263.

J Bone Joint Surg Br.

4. Sharrard WJ. A double-blind trial of pulse electromagnetic fields for delayedunion of tibial fractures. 1990;72:347-355.

Clin Orthop.

5. Brighton CT, Shaman P, Heppenstall RB, Esterhai JL Jr, Pollack SR, FriedenbergZB. Tibial nonunion treated with direct current, capacitive coupling, or bone graft.1995;321:223-234.

6. National Coverage Analysis (NCA): Electrical Stimulation for FractureHealing. Baltimore, Md: Centers for Medicare & Medicaid Services; 1999. PublicationCAG-00043N.

Clin Orthop.

7. Ryaby JT. Clinical effects of electromagnetic and electric fields on fracture healing.1998;355(suppl):205-215.

8. The Competitive Edge HMO Directory. St Paul, Minn: InterStudy Publications;2000.

9. Directory of Health Plans. Philadelphia, Pa: American Association of HealthPlans and Dorland Healthcare; 2000.

10. 2000 Report and Directory of Preferred Provider Organizations. Washington,DC: SMG Marketing Group and the American Association of Preferred ProviderOrganizations; 2000.

Am J Manag Care.

11. Steiner CA, Powe NR, Anderson GF. Coverage decisions for medical technologyby managed care: relationship to organizational and physician payment characteristics.1996;2:1321-1331.

Health Aff (Millwood).

12. Landon BE, Epstein AM. For-profit and not-for-profit health plans participatingin Medicaid. 2001;20(3):162-171.