Managed Health Plan Effects on the Specialty Referral Process

Results from the Ambulatory Sentinel Practice Network Referral Study

CHRISTOPHER B. FORREST, MD, PHD,* PAUL NUTTING, MD, MSPH,‡ JAMES J. WERNER, MS,§

BARBARA STARFIELD, MD, MPH,* SARAH VON SCHRADER, MA,* AND CHARLES ROHDE, PHD†

OBJECTIVES. The specialty referral process isone of the chief targets of managed care con-straints on ambulatory medical decision-making. This study examines the influence ofgatekeeping arrangements and capitated pri-mary care physician (PCP) payment on the spe-cialty referral process in primary care settings.

RESEARCH DESIGN. Primary care practice-basedstudy of referred and nonreferred office visits.

SUBJECTS. The study comprised 14,709 visitsmade by privately insured, nonelderly patientswho were seen by 139 primary care physiciansin 80 practices located in 31 states.

MEASURES. Visits were grouped by healthplan type: gatekeeping with capitated PCPpayment; gatekeeping with fee-for-servicePCP payment; no gatekeeping. Dependentmeasures included the proportion of visitsreferred, characteristics of referrals, and physi-cian coordination activities.

RESULTS. The percentages of office visits re-sulting in a referral were similar between thetwo gatekeeping groups and higher than theno gatekeeping group. Patients in plans withcapitated PCP payment were more likely to bereferred for discretionary indications thanthose in nongatekeeping plans (15.5% v 9.9%,P <0.05). The frequency of referring physician

coordination activities did not vary by healthplan type. The proportion of patients in gate-keeping health plans within a practice wasdirectly related to employing staff as referralcoordinators, allowing nurses to refer withoutphysician consultation, and permitting pa-tients to request referrals by leaving recordedtelephone messages.

CONCLUSION. The specialty referral processfor privately insured nonelderly patients en-rolled in managed health plans is generallysimilar, regardless of the presence of gatekeep-ing arrangements and capitated PCP payment.An increase in the number of discretionaryreferrals among patients in plans with capi-tated PCP payment provides support for ex-ploring strategies that encourage PCPs to man-age in their entirety conditions that straddlethe boundaries between primary and specialtycare. In response to increasing numbers ofpatients enrolled in managed health planswith gatekeeping arrangements, physiciansappear to modify the structure of their prac-tices to facilitate access to and coordination ofreferrals.

Key words: Referral-consultation; primarycare; managed care; gatekeeping; capitation;coordination. (Med Care 2003;41:242–253)

From the *Department of Health Policy and Manage-ment and the †Department of Biostatistics, Johns Hop-kins Bloomberg School of Public Health, Baltimore,Maryland.

‡From the Department of Family Medicine, Univer-sity of Colorado and Center for Research Strategies,Denver, Colorado.

§From the Department of Health and BehavioralSciences, University of Colorado at Denver, Denver,Colorado.

Supported by grant no. R01 HS09377 from theAgency for Healthcare Research and Quality.

Address correspondence and reprint requests to:Christopher B. Forrest, MD, PhD, Johns Hopkins Med-ical Institutions, 624 N. Broadway, Room 689, Baltimore,MD 21205. E-mail: cforrest@jhsph.edu

Received December 19, 2001; initial review March 8,2002; accepted July 25, 2002.

MEDICAL CAREVolume 41, Number 2, pp 242–253©2003 Lippincott Williams & Wilkins, Inc.

242

The specialty referral process is a commontarget for managed care utilization controls. Two ofthe strongest constraints on physician decision-making are gatekeeping arrangements and physi-cian payment.1 Gatekeeping has a clinical functionin which patients use a specific physician as anentry point into medical care. The physician gate-keeper identifies patients’ health care needs andthe services required to address those needs. Inaddition to this traditional first-contact role ofPCPs,2 managed health plans have added anadministrative function in which physician gate-keepers authorize patients’ specialty referrals.

The effects of gatekeeping arrangements onreferrals are becoming clearer. Patients enrolled ingatekeeping plans are more likely than counter-parts to be referred during office visits.3–5 Whetherthis positive effect of gatekeeping on the volumeof referrals made from physicians offices is a resultof cooccurring payment systems that encouragereferrals (eg, PCP capitation) has not been rigor-ously explored. When patients’ utilization patternsare examined over time, those who switch intogatekeeping plans have had reduced rates of spe-cialist use,6–8 whereas those who switch out ofgatekeeping plans have had no change in special-ist use.9 Gatekeeping’s effect on increasing thechances that a physician decides to refer a patientcoupled with minimal impact on specialist usesuggest that the main impact of gatekeeping onreferral rates is to limit patient self-referral. Pa-tients in health maintenance organizations aresubstantially less likely to see a specialist via theself-referral pathway than those with less re-stricted access.3

Gatekeeping should enhance coordination ofcare to the extent that referral approvals increasePCP awareness of the specialty care their patientsreceive. Although a study of physician attitudessuggested poorer communication among man-aged care versus indemnity patients,10 a prospec-tive study of pediatricians’ referral practices foundno difference in coordination between patients inhealth plans with and without gatekeepingarrangements.5

A review of the effects of alternative paymentsystems on PCP decision-making concluded thatthere was scant empirical information on whethercapitation influenced the specialty referral pro-cess.11 In Denmark12 and Norway13 capitated PCPpayment has been associated with increased spe-cialty referrals. In recognition of capitation’s incen-tives to promote referrals, some medical groups in

California have developed blended payment sys-tems that combine capitated PCP payment withfee-for-service for procedures that straddle thescope-of-practice boundaries between primaryand specialty care physicians.14 The influence ofcapitation on overall rates of referral and its po-tential effect on discretionary referrals merits fur-ther attention.

In this manuscript, we assess how managedhealth plans influence the specialty referral pro-cess among nonelderly, privately insured patients.The research was conducted in the offices of 139primary care physicians who practiced in 31 states.Our aims were to examine the influence of gate-keeping arrangements and capitated PCP pay-ment on the frequency of referrals, characteristicsof referral decisions, and coordination of referralcare.

Materials and Methods

Physician Recruitment

The study was coordinated by the staff of theAmbulatory Sentinel Practice Network (ASPN).Physicians were recruited in successive waves be-tween March 1997 and May 1998. Recruitmentwas directed to physician members of ASPN,physicians affiliated with Medical Group Manage-ment Association, other practice-based researchnetworks (Minnesota Academy of Family Physi-cians Research Network, the Wisconsin ResearchNetwork, and the Dartmouth COOP), and thelarger community of physicians. The study waspublicized via direct mailings to physicians, articlesin research network newsletters, notices in jour-nals, and presentations at conferences. Telephonecontact was made with physicians expressinginterest.

Physicians were included in the study if theypracticed in the United States and were not intraining. Of the 342 expressing interest in thestudy, 155 completed all aspects of data collection.The final sample had 139 physicians (42% weremembers of ASPN); 11 were excluded because ofinadequate plan information and five because theyhad no visits with privately insured patients. Nophysician worked in a group/staff-model HMO.

Survey Procedures

Study protocols and materials were approved bythe Johns Hopkins School of Public Health’s Com-

Vol. 41, No. 2 MANAGED CARE AND SPECIALTY REFERRALS

243

mittee on Human Research and the ColoradoMulti-Institutional Review Board. Procedures andquestionnaires were adapted from a study weconducted with pediatricians.5,15,16 Refinements ofmethods and questionnaires were made after apilot test done in five practices.17

Data were collected from August 1997 to April1999. Within a practice, this study was done in twophases. In Phase 1, physicians completed a back-ground questionnaire. Each practice selected acoordinator who communicated with ASPN,learned study protocols, trained office staff, andmonitored data quality. Practice coordinators com-pleted a questionnaire about the organizationaland financial components of their practices. Coor-dinators described all health plans their practiceaccepted. Specific items included the plan’s use ofspecialty referral authorization (gatekeeping),capitated PCP payment, and payer-type. Researchstaff provided practices with a numeric identifierfor each plan.

During Phase 2, coordinators kept logs of visitsand referrals and physicians completed question-naires about referrals they made. Phase 2 lasted for15 consecutive practice days, defined as either ahalf or full workday during which the physicianheld routine office hours. Each patient’s date ofbirth, gender, health plan (using the unique iden-tifier), principal diagnosis, and up to two addi-tional diagnoses were recorded in the visit log. Amedical record abstractor converted the diagnosesinto ICD-9-CM codes.

In the referral log, coordinators recorded allreferrals made by physicians and other staff. Phy-sician referrals involved verbal communication be-tween the physician and parent/patient during avisit or telephone encounter. Because the focus ofthis study was on new referrals made by primarycare physicians to specialist clinicians in ambula-tory settings, we excluded referrals made to labo-ratories, radiologic facilities, emergency depart-ments, hospitals for inpatient admission, andcurb-side consultations in which the referringphysician obtains verbal advice from a specialist.

An iterative feedback process was used to op-timize validity of data collected by practice coor-dinators. Before each phase of data collection,small group conference calls (3 to 5 coordinatorsand a member of the research staff) were con-ducted to explain and review study protocols.Once a sample of data was collected, research staff

reviewed it for completeness and accuracy. Find-ings from this review were communicated to prac-tice coordinators.

Managed Health Plan Characteristics

The evolution of managed care and resultantheterogeneity of plans within organizational mod-els has led to the recognition that traditional labels(eg, HMO, PPO, and Indemnity) have limitedresearch utility.18 Analysts have suggested thatplans be characterized by their specific at-tributes.1,18 We divided plans into those with andwithout gatekeeping arrangements. The formergroup was divided based on type of PCP pay-ment–capitation or fee-for-service. All plans usingcapitated PCP payment also had gatekeeping.Work by Grembowski et al1 has shown that phy-sician payment and gatekeeping had the largesteffects among five variables included in a “mana-gedness index,” designed to measure the restric-tions on medical decision-making. Plan data weremissing for 1% of privately insured visits in thesample.

Diagnostic Data

ICD-9-CM diagnosis codes were grouped intoclinically similar categories called expanded diag-nosis clusters (EDCs). A sub-set of diagnoseswithin clusters is categorized as a chronic condi-tion using the criterion of expected continuousduration of 12 months or more. To account for thenumber and severity of conditions with whichpatients presented, we assigned a previously de-veloped “treated morbidity index”to each visit.19,20

The index is predictive of the chances of referralduring an office visit.20 It is derived from Aggre-gated Diagnostic Groups (ADGs). Each ADG con-tains diagnostic codes that have similar clinicalcriteria and expected health care needs. Largerscores suggest a greater complexity level for thepatient visit. Further information on EDCs andADGs available at the following Web site:http://www.acg.jhsph.edu.

Referral Decision Variables

Characteristics of physicians’ referral decisionswere obtained from questionnaires physicianscompleted after making a referral. Items with a

FORREST ET AL MEDICAL CARE

244

yes/no response format included: referred healthproblem was newly presenting to medical care;referral was made to a specialist within the phy-sician’s practice; patient’s reason for encounterwas to get a referral; patient’s health plan re-stricted the physician from referring to the mostappropriate specialist; physician or staff made thespecialty appointment; physician provided patientwith the name of a specific specialist; and physi-cian sent the specialist relevant patient informa-tion. Additional items included: locus of referral(office visit or telephone encounter); type of spe-cialist referred to; certainty ratings (range 1–5) ofthe diagnosis and treatment of the referred healthproblem; and, urgency of the referral (dichoto-mized as patient needs to see specialist within ormore than a week). Physicians indicated theirreasons for making the referral using a previouslydeveloped taxonomy.15

Discretionary medical decision-making hasmore variability across practitioners than decision-making associated with nondiscretionary clinicalproblems,21 and, as a result, may be more sensitiveto gatekeeping and capitation effects.22 Referralsfor uncommon diseases are usually nondiscretion-ary, requiring little physician judgment regardingnecessity of referral. Most clinical situations arenot so clear-cut and involve some degree of phy-sician discretion concerning the referral decision.To identify “discretionary referrals,” we created avariable that dichotomized referrals according towhether they meet all four of the following crite-ria: common condition—highest third for fre-quency of presentation to physicians in the sam-ple; high certainty—certainty ratings of 4 or 5 (ona 5-point scale) for the diagnosis and treatment ofreferred health problem; low urgency—patientcould wait at least a week to see the specialist; andcognitive assistance only requested—main reasonfor referral was for advice, to direct medical man-agement, because of time constraints or patient/third party request.

Data Analysis

To compare physicians’ perceptions about pastreferral experiences, we created three groups ac-cording to the proportion of visits made during thestudy period by patients with gatekeeping healthplans. The percentage of patients with gatekeep-ing arrangements was determined for all partici-pating physicians within a practice. To conduct

statistical analyses contrasting physician percep-tions regarding the referral process, three groupswere formed–lowest quartile, middle quartile, andhighest two quartiles. The group with the lowestproportion of gatekeeping patients (the referent)was compared separately with each of the othertwo groups using t tests for means and the �2

statistic for proportions.Three managed health plan groups comprised

the main independent variable. We divided plansinto those with and without gatekeeping arrange-ments, and the former group was subdividedbased on type of PCP payment: capitation orfee-for-service. The gatekeeping with capitationgroup was the referent and was compared witheach of the other 2 groups separately in analyseson characteristics of patients, referral rates, char-acteristics of referrals, and coordination activities.Patient and referral characteristics were statisticallyevaluated across the groups using �2 tests forproportions and t tests for continuous variables.

Rates of referral were calculated as the percent-age of office visits leading to a referral, the per-centage of referrals made by physicians duringtelephone encounters with patients, and the per-centage of all referrals made from the physician’spractice that were a result of decisions made bynonphysician staff. Because of the known effects ofage, sex, and treated morbidities on primary carephysician referral rates,20 we adjusted referral rateproportions for each of these characteristics. Theseregression analyses entered the gatekeeping withcapitation group as the referent. We accounted forthe unequal distribution of patient visits by phy-sician and the clustering of visits within physiciansusing mixed effects regression models. Each phy-sician was entered as a random effect, and patientcharacteristics were entered as fixed effects. TheSAS GLIMMIX module was used for these anal-yses.24 For analyses of the percentage of referralsmade by nonphysician staff and referral character-istics, we adjusted proportions for physician ran-dom effects only.

Results

The physician sample was composed of 139PCPs (127 family physicians, 11 internists, and onepediatrician) in 80 practices located in 31 states.Mean age was 45.3 years, 18.7% were female,10.4% were located in rural practices, and 33.3%worked �28 hours per week. On average, each

Vol. 41, No. 2 MANAGED CARE AND SPECIALTY REFERRALS

245

physician participated in 10.8 private health insur-ance plans (range 1–31).

Before beginning office data collection, physi-cians’ reported on their prior referral experiences(Table 1). Compared with the low managed caregroup, the high group reported more problemswith patient self-referral and greater financial riskfor making referrals, but fewer plan restrictions onobtaining specialty care. The proportion of pa-tients in gatekeeping plans seen in a practice wasdirectly related to use of referral coordinators,nurses being allowed to refer without physicianconsultation, and allowing patients to requestreferrals by leaving a recorded message on thepractice’s voicemail system.

The sample of office visits, which was com-posed of privately insured patients only, consti-tuted 56.1% of all visits made to participatingphysicians. Table 2 shows the characteristics of thevisit sample.

The percentages of office visits resulting in aspecialty referral were similar between the twogatekeeping groups and higher than the no gate-keeping group (Table 3). About half the referralsmade during telephone conversations with pa-tients were a result of patients’ requesting a refer-ral. When these patient-requested referrals wereremoved from the sample, more referrals in thecapitation group were made during telephoneconversations than the other two groups.

Referrals among the capitation group weremore likely to be for chronic conditions than theother groups (Table 4). Compared with the nogatekeeping group, patients in health plans withcapitated PCP payment were more likely to bereferred for discretionary indications. Across thethree groups, referrals made within the practiceand prior curbside consultation were equallylikely.

Although patients in the gatekeeping healthplan groups were more likely to request a referralthan the nongatekeeping group, there were nosignificant differences in 14 other reasons forreferral (data not shown). Just 1% of referrals weremade because of insurance plan requirements forreferral reauthorization; this proportion did notdiffer across groups.

We found no statistically significant differencesin the frequency of several referring physiciancoordination activities by type of health plan.Overall, physicians provided patients with thename of a specific specialist for 85.7% of referrals;they or their office staff made the specialty ap-

pointment for 68.2%; and they sent relevant infor-mation to the specialist for 83.7%. Physiciansreported that they had referred to the specialistpreviously for 92.8% of referrals for which physi-cians gave patients a specific name of a specialist;no significant differences in this proportion werefound by plan category.

Discussion

This study examined how managed healthplans, as characterized by gatekeeping arrange-ments and capitated primary care physician (PCP)payment, influence the specialty referral process.Despite managed health plans’ widespread use ofrestricted provider networks, choice of specialistdid not differ by type of health plan. Physicians feltrestricted in their ability to refer to the mostappropriate specialist in just 5% of cases. Theseactual referral experiences contrast with PCPs’perceptions of the restrictions managed careplaces on their referral decision-making. In somestudies, physicians have reported more barriers toobtaining necessary referrals for their managedcare patients compared with those in indemnityplans.25–27 However, among physicians in oursample, an increasing proportion of managed carepatients in their practice was associated with lessdissatisfaction with their ability to obtain neces-sary referrals. Similarly, Kerr et al28 found that anincreasing proportion of revenue from capitatedhealth plans was associated with greater physiciansatisfaction with the referral process.

In response to increasing numbers of patientsenrolled in managed health plans with gatekeep-ing arrangements, physicians appear to modify thestructure of their practices to facilitate access toand coordination of referrals. In our study, thelevel of managed care in a practice was associatedwith permitting staff with or without physicianinput to make referrals and with employing refer-ral coordinators. Gatekeeping and capitated PCPpayment independently shifted some referraldecision-making from office visits to telephoneencounters. These changes may have resultedfrom higher administrative burden associated withmanaged care referrals. They may also serve toincrease the efficiency of the referral process,resulting in more physician satisfaction.

Consistent with other research,3–5,9 this studysuggests that primary care gatekeepers are not

FORREST ET AL MEDICAL CARE

246

restricting patients’access to specialists. Gatekeep-ing patients were more likely than nongatekeep-ing patients to be referred during an office visit.On the other hand, the ability of patients innongatekeeping plans to directly access specialistsmay lead to higher overall specialist utilizationrates compared with those in gatekeeping plans.One would expect the amount of patient requestsfor referral to increase among gatekeeping pa-tients, if limited direct access to specialists was amajor reason for the “gatekeeping effect” on pri-

mary care referral decisions. However, even afterremoving all referrals that were requested by pa-tients, we found substantially higher referral ratesamong patients in gatekeeping plans comparedwith those in nongatekeeping plans.

Capitated physician payment did not influencerates of referral. European studies have foundincreased referrals among physicians whoswitched from a primarily fee-for-service paymentsystem to capitated payment.12,13 A key differencebetween our study and the European work is that

TABLE 1. Prior Referral Experiences and Practice Characteristics of 139 Primary Care Physicians byLevel of Managed Care

“Thinking about referrals you make to medical andsurgical specialists, how much of a problem is eachof the following?”

Proportion of Visits Made to PCP by Privately InsuredPatients in Gatekeeping Health Plans

Lowest Quartile(N � 34) Mean

Item Score*

Middle Quartile(N � 34) Mean

Item Score*

Highest Quartiles(N � 71) Mean

Item Score*

Patients refer themselves without first discussing thereferral with me.

2.15 2.29 2.46†

My patients are cross-referred (from one consultantto another) without consulting me.

2.24 2.53 2.59

Insurance plans restrict referrals for necessaryspecialty care.

2.56 2.18 2.11†

Insurance plans restrict me from referring a patientto the most appropriate specialist.

2.42 2.38 1.99†

Peer pressure to control costs. 1.53 1.41 1.89†

Financial risk for referring. 1.29 1.44 1.68†

The number and complexity of forms you must fillout for a referral.

2.44 2.45 2.30

For children: pressure to refer to adult specialists. 1.21 1.26 1.42Patients request referrals when they are not

clinically indicated.2.35 2.41 2.51

Practice Characteristics Related to Referral Process

Proportion of Visits Made to Participating PCPs in thePractice by Privately Insured Patients in Gatekeeping

Health Plans

Lowest Quartile% (N � 20)

Middle Quartile% (N � 20)

Highest Quartiles% (N � 40)

Practice has a referral coordinator(s) 45.0 55.0 73.7†

Nurses permitted to refer patientsAfter consultation with physicians 89.5 90.0 82.0Without consultation with physicians 0.0 15.8 21.0†

Administrative staff permitted to refer patientsAfter consultation with physicians 52.6 85.0† 67.5Without consultation with physicians 0 5.0 10.5

Patients can request a referral by leaving a recordedmessage on practice voice mail

5.0 35.0† 20.0

*The response categories for the items ranged from a low of 1 (no problem) to 4 (big problem).†The lowest managed care group was the reference category. The other two groups were separately compared

with the referent. P � 0.05 using t tests for means and �2 statistics for proportions.

Vol. 41, No. 2 MANAGED CARE AND SPECIALTY REFERRALS

247

physicians in our sample did not work exclusivelyin a capitated environment. On average, just 29%of privately insured patients were in capitatedhealth plans. Concerns that PCP capitation willlead to “dumping”of patients into specialty care22

appear to be unjustified, at least among physiciansin community-based practices that participate inmultiple health plans.

Prior research has found no managed care effecton the types of referrals made for pediatric pa-tients,5 patients with chest pain,29 or those re-ferred for appendicitis.30,31 In this study, gatekeep-ing plans were associated with more patient-requested referrals, but other reasons for referralwere similar to nongatekeeping plans. Comparedwith patients in nongatekeeping plans, those inplans with capitated PCP payment had morereferrals for discretionary indications. Becausethere are no commonly accepted empirical defini-tions of what constitutes a discretionary referral,we created a new variable that was based on thepresence of all four of the following characteristics:referred condition is commonly managed in pri-mary care, little clinical uncertainty, low urgency,and the need for cognitive expertise only from thespecialist. Common conditions are usually re-

tained in primary care,3 and when they are re-ferred, it is to obtain access to a specialized skill,such as refraction for myopia. One would expectPCPs to have the knowledge and cognitive skillsnecessary for diagnosing and treating commonconditions, particularly those with little clinicaluncertainty. Capitated payment may act at themargins of PCPs’ scope-of-practice, increasing thelikelihood that health problems which could bemanaged in either a primary care or specialtysetting are referred.

Managed health plans use a variety of tech-niques to manage demand and to limit supply ofhealth care services. An important limitation ofthis study is our focus on just one organizationalfeature (gatekeeping arrangements) and one fi-nancial incentive (capitated PCP payment). Otherfeatures such as utilization management, providerprofiling, specialty withhold payments, composi-tion of the specialist network, and patient costsharing were not examined. As shown in a studyon mental health referrals, specialty withhold pay-ments may decrease the chances of referral.32

Exclusion of this variable from our study may haveinflated true differences between the managedand nonmanaged care health plan groups.

TABLE 2. Characteristics of Privately Insured Nonelderly Patients Who Made Office Visits With 139Primary Care Physicians’ by Type of Health Plan

Characteristic Total Sample

Type of Health Plan*

Gatekeeping WithCapitated PCP

Payment

Gatekeeping WithFee-For-ServicePCP Payment

No GatekeepingArrangements

No. visits (% of total) 14,709 (100) 4231 (28.8) 5688 (38.6) 4790 (32.6)Age, mean (SD) 33.7 (18.1) 33.7 (17.9) 33.7 (17.7) 33.6 (18.7)% 17 years or younger 22.8 21.6 22.4 24.3‡

% Female 59.8 60.9 59.7 59.0Treated morbidity index, mean

(SD)0.45 (0.59) 0.44 (0.59) 0.47 (0.59)† 0.46 (0.60)

% With a chronic medicalcondition

25.7 22.2 26.6§ 27.8§

Type of primary diagnosis, % § §

Sign/symptom 14.6 15.6 14.3 14.0Medical 28.1 25.0 28.4 30.6Surgical 31.3 29.6 31.3 32.8Ob/Gyn 6.5 7.9 6.1 5.6Mental health 3.8 3.1 4.5 3.6Administrative/preventive care 15.7 18.8 15.4 13.4

*P values for categorical variables were derived from �2 tests, and for continuous variables from t tests. Thegatekeeping with capitated PCP payment group was the reference category for statistical comparisons.

†0.01 � P � 0.05; ‡0.001 � P � 0.01; §P � 0.001.

FORREST ET AL MEDICAL CARE

248

One possible explanation for the higher referralrate among the gatekeeping compared with thenongatekeeping health plans is selection bias. Recentstudies suggest slightly favorable selection amongmanaged health plans.33 In this study, we found thatpatients in the capitated health plans were less likely tohave a chronic condition than patients in the other twohealth plan groups. The favorable selection among thecapitated health plan visits could have the effect of

narrowing true differences in referral rates. Althoughour analyses controlled for morbidity burden, un-measured differences in health status could explainsome of this effect.

Physicians in the study sample were volunteers.The generalizability of this sample to the largerpopulation of PCPs has been examined previous-ly.17 Rates of referral (overall and stratified by age,sex, and selected diseases) among study physi-

TABLE 3. Frequency of Specialty Referral Among Privately Insured Nonelderly Patients by Type ofHealth Plan

Referral Rates and RatiosTotal

Sample

Type of Health Plan§

GatekeepingWith CapitatedPCP Payment

Gatekeeping WithFee-for-ServicePCP Payment

NoGatekeeping

Arrangements

Specialty referrals made during office visitsNo. office visits 14,709 4231 5688 4790Adjusted overall % of office visits

referred¶5.5 6.5 6.2 3.9‡

Adjusted overall % of office visitsreferred after removing referrals inwhich patient’s reason for the officevisit was to get referral¶

4.5 5.1 5.1 3.4†

Adjusted % of office visits made forchronic conditions referred�

6.0 8.8 6.8 3.2‡

Adjusted % of office visits made formental health disorders referred�

6.8 10.6 7.2 3.4†

Specialty referrals made during telephone conversations with patientsTotal number of referrals made by

physicians for privately insuredpatients

981 362 397 222

Adjusted % of referrals PCPs madeduring telephone conversations withpatients�

22.7 28.1 20.8 17.6

Adjusted % referrals PCPs madeduring telephone conversations withpatients after removing referrals inwhich patient’s reason for thetelephone call was to get a referral�

11.2 15.4 10.1* 9.1*

Specialty referrals made by staff other than physiciansTotal number of referrals made by

either physicians or staff1188 401 530 257

Adjusted % of all referrals made fromPCP’s office that were made by staff�

16.6 19.3 16.7 13.8

Adjusted % of all referrals made fromPCP’s office that were made by thePCP during an office visit�

64.3 59.1 65.1 70.1*

*0.01 � P � 0.05; †0.001 � P � 0.01; ‡P � 0.001.§Statistical comparisons use the gatekeeping with capitated PCP payment as the reference group.¶Proportions were adjusted for patient age, sex, and the treated morbidity index using a mixed effects model.

Each physician was entered into the model as a random effect.�Proportions were adjusted for referring physician random effects only.

Vol. 41, No. 2 MANAGED CARE AND SPECIALTY REFERRALS

249

cians were comparable to those found amongphysicians participating in the National Ambula-tory Medical Care Survey, which is representativeof US physicians’ practice patterns. Family physi-cians in the study sample had a similar number ofvisits per day as the typical family physician in theUnited States. It should also be noted that al-though this study included physicians whoworked in a variety of practice arrangements,those in group/staff model HMOs were excluded.Our managed care findings, therefore, should begeneralized to patients of physicians who contractwith network-model HMOs.

Conclusion

The specialty referral process in primary caresettings is generally similar for privately insurednonelderly patients enrolled in managed healthplans, regardless of the presence of gatekeepingarrangements and capitated PCP payment. De-spite prior research to the contrary, capitatedPCP payment had no effect on the proportion ofoffice visits referred, although it did appear toshift the balance toward more referrals made fordiscretionary indications. The absence of a cap-itation effect on overall referral rates is reassur-

TABLE 4. Characteristics of Specialty Referrals Among Privately Insured Nonelderly Patients by Type ofHealth Plan

Referral CharacteristicTotal Sample% (N � 981)

Type of Health Plan

Gatekeeping WithCapitated PCP

Payment% (N � 362)

Gatekeeping WithFee-for-ServicePCP Payment% (N � 397)

No GatekeepingArrangements% (N � 222)

Adjusted Proportions‡

Type of referred health problemNewly presenting to medical care 57.3 56.4 56.3 60.6Chronic medical condition 20.3 26.3 20.1* 14.8†

Mental health disorder 5.9 9.1 3.9* 3.5Sign or symptom 23.0 20.6 25.0 22.6Surgical condition 39.2 37.9 39.5 41.0

Type of specialist referred toMedical Specialist 31.3 31.6 30.3 32.5Surgical Specialist 44.8 39.5 50.4 43.2Non-Physician Clinician 12.8 14.7 10.1 15.1Mental Health Practitioner 4.7 6.9 3.3 1.8*

Referral made for discretionaryindications§

12.4 15.5 11.1 9.9*

Referral made to specialist in PCP’spractice

17.5 18.1 14.2 18.8

Curbside consultation obtained beforereferral made

7.0 6.2 6.8 8.2

Reason for visit/telephone encounterwas to get a referral

33.9 39.1 33.9 24.5†

Health plan restricted physician fromreferring to most appropriatespecialist

5.0 5.3 5.9 4.4

*0.01 � P � 0.05; †0.001 � P � 0.01.‡Statistical comparisons used the gatekeeping with capitated PCP payment as the reference group. Proportions

are adjusted for referring physician random effects only.§Discretionary referrals were defined by the presence of four characteristics: commonly occurring condition, a

high level of certainty regarding the diagnosis and treatment of the referred health problem, low urgency (i.e.,patient does not need to see specialist within a week), and the main reason for referral was for advice, to directmedical management, time constraints, and/or because of patient/third party request.

FORREST ET AL MEDICAL CARE

250

ing that it is unlikely to lead to patient “dump-ing” into specialty care. An increase in thenumber of discretionary referrals among pa-tients in plans with capitated PCP paymentprovides support for exploring strategies, suchas blended payment systems, that encouragePCPs to manage in their entirety conditions thatstraddle the boundaries between primary andspecialty care.

Acknowledgments

Laurie Vorel provided technical assistance with datacollection and project implementation. The authorsthank the physicians and office coordinators whodonated their time and provided their energy to makethis study possible. Their involvement was invaluable.These physicians are listed by the state where theirpractice is located: Arizona: Scott Ekdahl, DO; Arkan-sas: Batesville Family Practice Clinic—John Scott, MD;California: Springhill Medical Group—Andrew Ness,MD; Colorado: Howard Corren, MD; ComprehensiveFamily Care Center—Steven Milligan, MD, LouiseSchottsteadt, MD, Laura Stein, MD, Lynn Strange,MD; Example Orchard Family Practice—Frank Reed,MD, Nell Davis, MD, Charles Kay, MD, Lynne Spicer,MD; Family Physicians of Western Colorado—DanSullivan, MD; Plan de Salud del Valle—Tillman Farley,MD, Audrey Farley, MD; Southern Ute Health Cen-ter—Joan MacEachen, MD, George Maxted, MD; St.Anthony Family Medicine Residency—Timothy Dud-ley, MD, John Miller, MD, Kathy Miller, MD; Georgia:Memorial Family Practice—Roslyn Taylor, MD, KeithEllis, MD, Craig Fabel, MD; Promina Gwinnett Physi-cians - Buford—Bruno Denis, MD, Kelly Erola, MD,David Najjar, MD, Linda Casteel, MD, Richard Liotta,DO; Promina Gwinnett Physicians - Duluth—RandyCronic, MD, Mark Majoch, MD; Southeast GeorgiaRegional Medical Center—Russell Leubbert, DO,James Snow, DO; Illinois: Mt. Morris Health Center—Anna Meenan, MD, Loyd Wollstadt, MD, EduardoScholcoff, MD, Steven Lidvall, MD; Indiana: Clay CityCenter for Family Medicine—Steven Phillipson, MD,Paul Daluga, MD; Iowa: Gundersen - Decorah—JanetRyan, MD; Miller - Gundersen—Ken Miller, MD;Kansas: Robert Moser, MD, Wendell Ellis, DO; Atchin-son Family Medicine—John R. Eplee, MD, DanielSontheimer, MD; Louisiana: Family Medicine Centerof Baton Rouge—Linda Stewart, MD; Michigan: OSFMedical Group—John Hickner, MD; Oakwood FamilyMedical Center- Westland—Linda French, MD; Min-nesota: Camden Physicians—Richard Gebhart, MD;Coon Rapids Medical Center—Ashlesha Tamboli, MD;Fairview Eagan Clinic—Timothy Komoto, MD; Fair-

view Uptown Clinic—Dave Bucher, MD, Family Med-icine of Winona—William Davis, MD, Thomas Retz-inger, MD; Lake Crystal Clinic—Anthony Jaspers, MD;Stillwater Medical Group—Paul Spilseth, MD;Tweeten Lutheran—Glenn McCarty, DO; United Fam-ily Health Care—Katie Guthrie, MD, Ravi Balasubra-man, MD, Stephen Mitrione, MD; Montana: KurtzClinic PC—Curt Kurtz, MD; Nevada: Coleen Lyons,MD; New Hampshire: Hillsboro Medical Services—Richard Douglass, MD; Lamprey Health Care—PaulFriedrichs, MD; Red Hill Health Center—Peter Hope,MD; Tilton Family Health Care—Jonathan Mishcon,MD; New Jersey: John Orzano, MD; Somerset FamilyPractice Association—Winifred Waldron, MD; NewYork: John Glennon, MD; Family Medicine - MedicalService Group—R. Eugene Bailey, MD, Lorne Becker,MD, John DeSimone, MD, James Greenwald, MD,Glenn Griffin, MD, L. Thomas Wolff, MD; LathamMedical Group—Miguel Diaz, MD, Rebecca Elliott,MD; Women’s Health - Mt. Sinai—Carmella Abraham,MD, Eileen Hoffman, MD; North Carolina: Dave Rog-ers, MD; Chatham Primary Care—Phil Sherrod, MD;Family Medical Center—Ed Bujold, MD; Family Prac-tice Associates of Roanoke - Amaranth—Atsushi Mat-subara, MD; North Dakota: Family Practice Center-Minot—Steve Mattson, MD; Ohio: First MedicalAssociates—Thomas Detesco, MD; Oklahoma: EnidFamily Medicine Clinic—Mike Pontious, MD, LauraMiller, DO; Oregon: Dept. of Family Medicine FP—L.J.Fagnan, MD; Flaming Medical Center—Jerry Flaming,DO, Tom Flaming, DO, Jeffrey Humphrey, DO; SalemFamily Physicians—Michael Kelber, MD, John Satten-spiel, MD, Douglas Eliason, DO; Pennsylvania: JohnFarmer, DO; GS Oak Street Family Practice—PenithaWilliams, MD; South Dakota: Sioux Falls Family Prac-tice Residency/Ctr. for Family Medicine—Fred Thanel,MD; Tennessee: Michael Hartsell, MD; HumboldtFamily Care Center—R. Louis Murphy, MD; UniversityFamily Health Care—Dan Brewer, MD, John Parham,MD; Texas: Scott & White Clinic—Larry G. Padget,MD, Luis Moreno, MD, Kim Patrick Bolton, MD,Robert Henry, DO, Sharon Barber, MD, Robert Cortes,MD, John Manning, MD, Michael Averitt, DO, PaulGerdes, MD, Michael Kirkpatrick, MD, Shane Max-well, MD, Peter Sullivan, MD; Utah: Canyon ViewMedical Group—Eric Hogenson, MD; Utah ValleyFamily Practice Residency Program—Scott Endsley,MD, Jim Giovino, MD, David Flinders, MD, DwayneRoberts, MD; Virginia: Duane Lawrence, MD; JuneTunstall, MD; Big Island Family Medical Center—George Wortley, MD; Tappahannack Family Practice—James Ledwith, MD; Washington: Cle Elum FamilyMedicine Center—John Anderson, MD, ElizabethWise, MD; West Virginia: New River Family HealthCenter—Dan Doyle, MD, J. Michael Herr, DO; Wis-consin: Chain O’ Lakes Family Practice—Terry Han-key, MD; Franciscan Skemp Health care-Lake TomahClinic—Rod Erickson, MD, Gundersen - Farrell—

Vol. 41, No. 2 MANAGED CARE AND SPECIALTY REFERRALS

251

Walter Boisvert, MD; Gundersen - Tomah—KevinJessen, MD; Gundersen - Wonewoc—Lea Cornell,MD; Gundersen Clinic - Onalaska—Dan Landdeck,MD; Gundersen Lutheran—Tom Frisby, MD; LakeTomah Clinic—Michael Saunders, MD; Mt. HorebClinic—Anne Eglash, MD, Vince Winklerprins, MD;Sparta Gundersen Clinic—Michael Pace, MD; Univer-sity of Wisconsin—Jon Temte, MD, Richard Anstett,MD, PhD, Dave Lonsdorf, MD, Catherine Soderqueist,MD; Vig - Gundersen—Brian Woody, MD.

References1. Grembowski DE, Diehr P, Novak LC, et al.

Measuring the “managedness” and covered benefits ofhealth plans. HSR 2000;35:707–734.

2. Forrest CB, Starfield B. Entry into primary careand continuity: the effects of access. Am J Public Health1998;88:1330–1336.

3. Forrest CB, Reid RJ. Passing the baton: HMOs’influence on referrals to specialty care. Health Aff (Mill-wood) 1997;16:157–162.

4. Franks P, Clancy CM. Referrals of adult patientsfrom primary care: demographic disparities and theirrelationship to HMO insurance. J Fam Pract 1997;45:47–53.

5. Forrest CB, Glade GB, Starfield B, et al. Gate-keeping and referral of children and adolescents tospecialty care. Pediatrics 1999;104:28–34.

6. Martin DP, Diehr P, Price KF, et al. Effect of agatekeeper plan on health services use and charges: Arandomized trial. Am J Public Health 1989;79:1628–1632.

7. Hurley RE, Freund DA, Gage BJ. Gatekeepereffects on patterns of physician use. J Fam Pract1991;32:167–174.

8. Ferris TG, Perrin JM, Manganello JA, et al.Switching to gatekeeping: changes in expenditures andutilization for children. Pediatrics 2001;108:283–290.

9. Ferris TG, Chang Y, Blumenthal D, et al.Leaving gatekeeping behind–effects of opening access tospecialists for adults in a health maintenance organiza-tion. N Engl J Med 2001;345:1312–1317.

10. Roulidis ZC, Schulman KA. Physician com-munication in managed care organizations: Opinions ofPCPs. J Fam Pract 1994;39:446–451.

11. Gosden T, Forland F, Kristiansen IS, et al.Impact of payment method on behaviour of primary carephysicians: a systematic review. J Health Serv Res Policy2001;6:44–55.

12. Krasnik A, Groenewegen PPA. Changing re-muneration systems: effects on activity in general prac-tice. BMJ 1990;300:1698–1701.

13. Iversen T, Luras H. The effect of capitation onGPs’ referral decisions. Health Econ 2000;9:199–210.

14. Robinson JC. Blended payment methods inphysician organizations under managed care. JAMA1999;282:1258–1263.

15. Forrest CB, Glade GB, Baker A, et al. Thepediatric primary-specialty care interface: how pediatri-cians refer children and adolescents to specialty care.Arch Pediatr Adolesc Med 1999;153:705–714.

16. Forrest CB, Glade GB, Baker AE, et al. Coor-dination of specialty referrals and physician satisfactionwith referral care. Arch Pediatr Adolesc Med2000;154:499–506.

17. Forrest CB, Nutting P, Starfield B, et al.Family physicians’ referral decisions: Results from theASPN Referral Study. J Fam Pract 2002;51:215–222.

18. Brach C, Sanches L, Young D, et al. Wrestlingwith typology: penetrating the “black box” of managedcare by focusing on health system characteristics.MCR&R 2000;57:93–115.

19. Forrest CB, Whelan EM. Primary care safety-netdelivery sites in the United States: A comparison of com-munity health centers, hospital outpatient departments,and physicians’ offices. JAMA 2000;284:2077–2083.

20. Forrest CB, Reid RJ. Prevalence of health prob-lems and primary care physicians’ specialty referral de-cisions. J Fam Pract 2001;50:427–432.

21. Mort EA, Edwards JN, Emmons DW, et al.Physician response to patient insurance status in ambu-latory care clinical decision-making: Implications forquality of care. Med Care 1996;34:783–797.

22. Shimmura K. Effects of different remunerationmethods on general medical practice: a comparison ofcapitation and fee-for-service payment. Int J HealthPlann Manage 1988;3:245–258.

23. Deleted in proof.

24. Littel RC, Milliken GA, Stroup WW, et al.SAS(R) System for Mixed Models. Cary, NC: SAS Insti-tute; 1996.

25. Cartland JDC, Yudkowsky BK. Barriers topediatric referral in managed care systems. Pediatrics1992;89:183–192.

26. Collins KS, Schoen C, Sandman DR. TheCommonwealth Fund Survey of physician experienceswith managed care. New York: Commonwealth Fund,1997.

27. Halm EA, Causine N, Blumenthal D. Is gate-keeping better than traditional care? JAMA1997;278:1677–1681.

28. Kerr EA, Hays RD, Mittman BS, et al. Priamrycare physicians’ satisfaction with quality of care in Cali-fornia capitated medical groups. JAMA 1997;278:308–312.

FORREST ET AL MEDICAL CARE

252

29. Rask KJ, Deaton C, Culler SD, et al. Theeffect of primary care gatekeepers on the managementof patients with chest pain. Am J Manag Care1999;5:1274 –1282.

30. Adolph VR, Falterman KW. Appendicitis inchildren in the managed care era. J Pediatr Surg1996;31:1035–1036.

31. Krasna IH. Abdominal pain and appendicitis: Isthere a difference in referrals between HMO pediatri-

cians and private pediatricians? J Pediatr Surg2000;35:1084–1086.

32. Grembowski DE, Martin D, Patrick DL, et al.Managed care, access to mental health specialists, andoutcomes among primary care patients with depressivesymptoms. J Gen Intern Med 2002;17:258–269.

33. Hellinger FJ, Wong HS. Selection bias inHMOs: A review of the evidence. Med Care Res Rev2000;57:405–439.

Vol. 41, No. 2 MANAGED CARE AND SPECIALTY REFERRALS

253