Self-referral in Point-of-Service Health Plans FREE

Perceived barriers to obtaining specialty care are one of the greatest sources of consumer dissatisfaction with today's health maintenance organizations (HMOs).¹ Primary care physicians report difficulties obtaining specialty referrals in health plans with "gatekeeping" arrangements² and capitated payment.³ In response to these consumer and practitioner concerns, most HMOs now offer products, such as the point-of-service (POS) plan, with loosened restrictions on patients' access to specialty care. The triple-option POS plan is a blend of an HMO, preferred provider organization, and indemnity plan. Point-of-service members who use the principal HMO network and obtain authorization for referral services from their physician "gatekeeper" have minimal levels of cost sharing. Patient self-referral within the plan's network of practitioners is associated with moderate patient cost sharing, whereas self-referral to out-of-network practitioners has out-of-pocket payments comparable with indemnity plans.

From 1993 to 2000, US employer-based enrollment in POS health plans increased from 5% to 22% of those covered, one of the fastest growth rates among any type of managed care plan.⁴ By 2000, 44% of workers had a choice of a POS health plan,⁴ and in 1999, more than three quarters of HMOs offered a POS option.⁵ One reason for this growth is the hybrid nature of POS plans, which can facilitate an individual's transition from indemnity insurance to managed care.⁶ Patients may view POS plans favorably, because they maintain practitioner choice while offering lower premium costs than traditional insurance. Despite these perceived benefits, POS plans may have greater actuarial uncertainty due to out-of-network utilization, potentially poorer coordination of care, and higher administrative costs associated with claims processing.⁷

The limited information available suggests that POS health plan members infrequently exercise their option to self-refer. Using 1990-1991 data from 1 POS health plan, Wong and Smithen⁸ reported that 12% of claims accounting for 9% of expenditures were due to out-of-network utilization. In an analysis of 1994-1995 data from another POS health plan, Kapur et al⁹ found that 7% of expenditures were due to patient self-referral.

The growing importance of POS and other direct-access managed care models calls for more comprehensive evaluations to better understand this new type of managed health plan. Some basic questions about POS plans remain unanswered: Which patients are most likely to opt for self-referral when seeking specialty care? Why do patients bypass their primary care physicians? What is the impact of self-referral on patient outcomes? Using data from 3 commercial insurers in separate markets, this study addresses these questions and provides a detailed assessment of the self-referral process within POS health plans. Our aims were to quantify self-referred service use and charges, to identify patients most likely to obtain specialty care via self-referral, to characterize patients' reasons for self-referral, and to test the association between self-referral and satisfaction with specialty care.

Administrative databases from 3 commercial POS health plans were the primary data sources. The study period was calendar year 1996. All 3 were triple-option POS health plans, which gave members 3 alternatives at the "point of service": option 1, service use approved by the primary care physician involving the lowest level of cost sharing; option 2, self-referral within the provider network with an intermediate level of cost sharing; and option 3, self-referral to an out-of-network practitioner involving the highest level of cost sharing. For example, a member in a typical POS plan in this study had option 1 services of a $10 co-payment, no deductible, and no coinsurance; option 2 services of a $20 co-payment, a $0 to $100 deductible, and 20% coinsurance; and option 3 services of a $200 to $1000 deductible and 20% coinsurance.

Two of the POS plans were insurance products offered by not-for-profit BlueCross/BlueShield insurers in the Midwest and Northeast. Both had provider networks composed of most physicians and hospitals in the states they served. The third plan was a for-profit mid-Atlantic insurer that contracted with a more limited subset of practitioners in the state it served.

The study sample was composed of individuals aged 0 to 64 years enrolled for 6 to 12 months in 1996. We excluded mental health practitioner, substance abuse, and outpatient pharmaceutical claims because these services were separate "carved out" benefits that were administered by other organziations and not included in the administrative database. The claims systems of each plan permitted us to include all claims in the analysis, regardless of whether the member's deductible (if applicable) had been met.

Plan membership files were used to assign age and sex, calculate months of enrollment, determine if enrollees were new to the plan, and categorize members by their policy's level of office visit co-payment and deductible, if any, for option 1 (gatekeeper-approved) services. The ZIP codes of enrollees' residences were linked to the US Department of Health and Human Services Area Resource File to determine urban/rural residence.¹⁰International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis codes were matched to a set of 182 diagnosis clusters, which our research team expanded and modified from the original set of 92 by Schneeweiss et al.¹¹ A subset of clusters was categorized as chronic medical conditions using the criterion of expected continuous duration of 12 months or more.

Patient comorbidity was assessed using the Johns Hopkins Adjusted Clinical Group (ACG) Case-Mix Assessment System.¹² The first step in the ACG assignment process is to link ICD-9-CM diagnosis codes to 1 of 32 aggregated diagnostic groups (ADGs). Each ADG is a morbidity grouping with clinically homogeneous diagnosis codes that have similar expected need for health care resources. We used ADGs in this study for case-mix assessment. Diagnosis codes from health care claims in all care settings—except laboratory and radiology facilities, which submit a high proportion of claims for rule-out diagnoses—were used to assign ADGs.

The majority of physicians were assigned to a specialty group according to the self-reported specialties recorded in the American Medical Association Masterfile (82%-84% of all physicians across the 3 plans). The Masterfile was also used to determine years in practice. Specialty information for physicians whom we could not match to an entry in the Masterfile and nonphysician clinicians was obtained from the plans' files. Across the 3 plans, specialty information was missing for 0% to 1.3% of physicians.

Generalists included physicians whose only specialty designation was family medicine, general practice, pediatrics, internal medicine, or obstetrics/gynecology. We included gynecologists in the generalist category because health plans allowed women to select both a family physician or internist and an obstetrician/gynecologist as their primary care physicians. In the mid-Atlantic and northeastern plans, obstetricians/gynecologists had "gatekeeping" authority; in the midwestern plan, they did not. Individuals with both generalist and specialist designations were classified as specialists.

We assigned patients to a single "regular" physician for analyses that examined patient and regular physician characteristics (eg, specialty of the regular physician, years in practice) associated with self-referral to specialists. The method was based on patients' actual patterns of service use and selected the primary care physician seen most frequently. As an alternative to this utilization approach, we could have used health plan records to identify a patient's regular physician; however, 1 of the plans required patients to select a primary care group rather than a specific clinician. Thus, a utilization-based approach provided a common method for selecting a regular physician across the 3 plans.

Specifically, the method assigned a patient to the generalist physician with whom the patient had the largest number of visits. Self-referred visits were excluded from these assessments. Patients who made an equal number of visits to 2 or more generalists were assigned the 1 who provided the most resource-intensive services as measured by charges. Assignment to a specialist was permitted for individuals with no visits to a generalist. Using this method, we assigned between 90% and 93% of patients with some ambulatory care use at each POS plan to a regular physician. The 7% to 10% of unassigned patients either had no physician visits or the method was unable to resolve ties.

To assess the strength of the physician-patient relationship, continuity of care was calculated as the percentage of physician visits made to the regular physician.^13- 14

To remove the differences in fee schedules within and between plans, we standardized charges to a common set of values. Current Procedural Terminology codes were priced using the 1997 Medicare resource-based relative value scale. Relative values were converted to dollars using a conversion factor of $35 per unit. For other procedure codes, mean charges from the largest health plan were used as the standard rate. Using the principal diagnosis, inpatient stays were assigned to a Major Diagnostic Category (MDC) of the DRG system.¹⁵ Mean per diem rates for each MDC were calculated for children and adults separately in 1 health plan. To obtain standardized inpatient expenditures for a hospitalization, these MDC-specific per diem rates were multiplied by the length of stay in days.

For each patient, ambulatory charges were obtained by summing the standardized charges for each service occurring in an ambulatory setting. Inpatient charges were the sum of the standardized charges for all hospitalizations, including both institutional and physician charges. Total charges were the sum of ambulatory and inpatient charges.

Ambulatory charges were disaggregated into 8 categories: generalist visits, specialist visits, nonphysician clinician services, surgery, diagnostic/therapeutic procedures, laboratory, radiology, and other. These groupings were based on the method developed by Berenson and Holohan¹⁶ and assigned Current Procedural Terminology codes,¹⁷ level II codes in the Common Procedure Coding System,¹⁸ and revenue center codes to 1 of the 8 clinical service categories. Outpatient physician visits were divided into generalist and specialist categories according to the specialty of the billing physician.

Claims for physician, institutional, and ancillary services in each data set were designated as regular physician–approved (option 1) or self-referred within or outside of the provider network (options 2 and 3, respectively). The mid-Atlantic and northeastern plans had additional designations that allowed separation of self-referral within and outside of the provider network.

The survey protocol was approved by the Johns Hopkins School of Public Health Committee on Human Research. We conducted a telephone survey of patients in the midwestern POS health plan who had recently seen a specialist. The sample was selected using claims data and included members aged 0 to 64 years (parents responded for 0- to 17-year-old patients) who had a visit to a medical or surgical specialist 2 to 6 months before survey administration. Members with insurance coverage by more than 1 plan were excluded. The response rate was 65.4%.

The survey instrument contained items that separated patients based on their most recent specialist visit into those who self-referred vs those referred by their primary care physician. Patients who self-referred were asked, "Why did you choose not to have your primary care doctor authorize your referral to the specialist?" Interviewers recorded verbatim the single main reason for self-referral. The telephone survey also queried both physician-referred and self-referred patients about their satisfaction with the specialists in the health plan and their satisfaction with their most recent specialist visit.

We calculated rates of specialty self-referral as the annual percentage of health plan members who exercised their option to self-refer for a specialist visit overall, for out-of-network practitioners, and by physician and nonphysician clinician specialty categories. For inpatient and ambulatory services, average annual standardized charges and the percentage self-referred were calculated. The denominator for these analyses was all enrollees in the plan. Replication of findings using each health plan's actual "allowed" charges did not substantively alter parameter estimates or conclusions. We excluded chiropractic and eye care services from analyses related to self-referral in POS health plans because each plan allowed limited self-referral to these specialties as an option 1 benefit.

Multivariable linear regression that controlled for patient characteristics, case-mix, and months of enrollment was used to obtain adjusted mean charges. Thus, charges are presented as standardized dollars, adjusted for population differences.

We used logistic regression to identify characteristics of patients and their regular physicians associated with obtaining specialty care via self-referral. The sample was restricted to patients with at least 1 medical or surgical specialist visit in 1996. A visit was defined as an encounter that involved provision of an evaluation and management service. We excluded a specialist acting as a patient's regular physician from the pool of physicians to whom the patient could possibly self-refer. Separate regression analyses were performed for each plan. The generalized estimating equation was used to account for the correlation among patients assigned to the same regular physician.¹⁹

Table 1 presents characteristics of enrollees at each of the 3 study sites. Age and duration of enrollment selection criteria excluded 17.5% to 25.4% of all members. The northeastern POS plan was growing rapidly, with 33.2% of members newly enrolling in 1996. Virtually all enrollees in the mid-Atlantic POS health plan resided in urban areas. Compared with members in the midwestern plan, those in the mid-Atlantic plan were 2.2 times more likely to have 2 or more chronic conditions.

Rates of self-referral by practitioner specialty are presented in Table 2. Overall, 8.8% of enrollees in the midwestern POS plan, 16.7% in the northeastern plan, and 17.3% in the mid-Atlantic plan self-referred for at least 1 physician or nonphysician clinician visit. The proportions of enrollees self-referring to generalists (4.7%-8.5%) were slightly higher than the proportions self-referring to specialists (3.7%-7.2%) across all 3 plans. Orthopedic surgeons and dermatologists were the 2 most common types of specialists to whom enrollees in each plan self-referred.

Data from 2 of the plans permitted an analysis of whether self-referral was within or outside of the plans' provider networks. Just 0.2% of enrollees in the northeastern POS plan and 1.8% of enrollees in the mid-Atlantic POS plan self-referred to a specialist outside of the plan's network.

Table 3 presents mean annual standardized charges per enrollee and the percentages of each service obtained via self-referral. The proportions of total charges due to self-referral were remarkably similar between the northeastern and mid-Atlantic POS plans (16.3% and 15.7%, respectively). Across the 3 plans, self-referred charges as a percentage of the service category tended to be highest for nonphysician services, specialist visits, and invasive diagnostic and therapeutic procedures (eg, endoscopic and oncologic services), and tended to be lowest for generalist visits and laboratory services.

Among patients with at least 1 visit to a specialist, 16.6% in the Midwestern, 29.8% in the northeastern, and 24.5% in the mid-Atlantic self-referred for 1 or more of their specialist visits. Table 4 presents results from logistic regression analyses that examined the effects of patient and regular physician characteristics on the odds of making a self-referred specialist visit among users of specialty care. Individuals with unstable chronic conditions, allergies, orthopedic problems, and injuries had increased chances of a self-referral compared with their counterparts. Having no or small co-payments for option 1 services (regular physician–authorized) decreased the odds of self-referral by 12% to 21% compared with persons who had higher levels of cost sharing.

Higher levels of continuity with the regular physician were associated with decreasing odds of self-referral in an exposure-response type of relationship. This relationship held true in a sensitivity analysis that selected patients with only 1 specialist visit.

In a telephone survey conducted in the midwestern plan, we asked the patients who recently self-referred why they chose to bypass their primary care physicians. Most commonly, patients (37.5%) reported that they preferred to directly access a specialist to save time or to choose their own specialist. The second most common reason (27.8%) was that patients experienced relationship problems with their primary care physicians. Such problems most commonly occurred because physicians refused to make a requested referral. Some patients (22.9%) responded that an established relationship with a particular specialist was the reason for self-referral. A small number of patients (8.3%) were confused about insurance rules and did not realize they needed physician authorization. Just 3.5% of patients self-referred because they did not have a primary care physician.

Among respondents to the telephone survey, we compared satisfaction with specialists available in the plan and satisfaction with their last specialist visit between patients who obtained physician-approved referrals and those who self-referred. Results are shown in Table 5. Members who self-referred were more likely to be satisfied with their specialist visit and less likely to be satisfied with the health plan's specialist network.

This study provides a comprehensive assessment of the self-referral process within the POS health plan type. Our results show that in 3 geographically distinct markets, a minority of POS members exercised their option to self-refer; just 4% to 7% of members self-referred for a specialist visit. In 2000, employers and their workers paid an average of $623 per year more for a family POS health plan premium than a traditional HMO.⁴ This sum is the price consumers and benefit managers acting on their behalf appear willing to pay to maintain the possibility of direct access to specialists, even though patients infrequently use this option.

Surprisingly, the proportion of enrollees who self-referred to specialists was about the same as the proportion who self-referred to generalists. Point-of-service members may use their self-referral options to maintain or expand relationships with primary care physicians who do not participate in the plan or to develop new relationships. In a 1991 survey of individuals who recently switched into a POS plan, 36% reported retaining their primary care physician.²⁰

Although this study is the largest and most comprehensive analysis on POS plans to date, characteristics of the study populations influence the generalizability of findings. First, all individuals in this study had privately financed health care and were younger than 65 years. Although POS plans are still uncommon in government programs, our findings might not apply if Medicare or Medicaid implemented such a benefit structure more widely. Second, study sites were located in 3 health care markets with relatively high managed care penetration. A community where patients have less experience with "gatekeeping" arrangements could have higher rates of self-referral than those found in this study, possibly because a larger proportion of members would use the POS plan as if it were a standard indemnity plan. Third, 2 of the 3 insurers in our study were BlueCross/BlueShield plans, with very high provider network participation within the states they served. Point-of-service plans with less provider participation may have experiences that more closely resemble those of the mid-Atlantic plan in our study, which had higher rates of out-of-network utilization than the northeastern plan. Last, we excluded services provided by optometrists, ophthalmologists, and chiropractors from the self-referral analyses because plans allowed limited direct access at no added cost to these specialties. We also excluded services provided by mental health practitioners because, as is frequently the case among managed care health plans, these services were administered by separate organizations. In another study performed in 1 POS health plan, mental health services were the most common type of out-of-network care obtained.⁸

Another limitation to consider is the accuracy of administrative data for measuring self-referred utilization and practitioner specialty. In each of the 3 claims databases, a single variable defined the benefit level of the service, ie, option 1 (physician-approved within network), option 2 (self-referred within network), or option 3 (self-referred out of network). This variable was linked to payment, which may enhance its validity.²¹

This study relied primarily on specialty designation reported to the American Medical Association. Although self-designated specialty information may lead to some misclassification compared with board eligibility and certification, we elected to use the former, which characterizes how physicians present themselves in their communities.

Regression analyses of the patient and physician factors associated with self-referral to specialty care used a utilization-based algorithm to impute the regular physician. Northeastern and mid-Atlantic plan enrollees who had both an imputed regular physician and a plan-assigned physician gatekeeper were included in a sensitivity analysis. Findings revealed no substantive differences in the effect sizes from the logistic regression analyses nor were conclusions altered by using the imputed regular physician rather than plan-assigned gatekeeper.

The rate at which enrollees exercised their POS options for self-referral was markedly lower in the midwestern plan compared with the other 2 plans. In the former, gatekeeping authority resided with physician groups, many of which were organized as multispecialty practices. Utilization within the physician group was not associated with increased costs to the patient, ie, it was considered an option 1 service. Thus, some self-referral in the midwestern plan probably occurred within the large multispeciality group practices. Because these self-referrals were not captured in our database, we may have underestimated the rates for the midwestern plan. A second explanation for the lower self-referral rates in the midwestern plan is that it was located in a state with the highest managed care penetration among the three plans. Midwestern plan members may have been more accustomed to a managed care environment and restricted access to specialists. Indeed, annual rates of specialist use were lowest for midwestern POS plan members.

Charges due to self-referral in the northeastern and mid-Atlantic plans were remarkably similar overall and by type of service. About 16% of total charges and 20% of specialist visit charges were due to self-referred services in both plans. Furthermore, 7.2% of enrollees in both plans self-referred to specialists. These similarities are striking considering the plans' different geographic markets, tax status, type of ownership, provider network, and patient populations. On the other hand, out-of-network specialist self-referral rates were lower in the northeastern plan (2/1000 members per year) compared with the mid-Atlantic plan (18/1000 members per year). This disparity can be explained by the northeastern plan's large BlueCross BlueShield provider network in which virtually all physicians in the service area participated. Despite this higher rate of out-of-network use in the mid-Atlantic plan, its overall standardized ambulatory charges were lower than the northeastern plan, suggesting that a smaller provider network is not likely to lead to excessive costs due to self-referral. However, standardizing charges removes differences in fees set by out-of-network practitioners, who almost certainly will bill at higher rates than in-network physicians.

Compared with the plans in our study, Wong and Smithen⁸ reported a higher rate of out-of-network users (16% of members per year) in their study of a single POS plan using data from 1991. The patient sample in that study was substantially older than the sample for which we report results (mean age, 48 years vs 31 years in the databases we used), and in that study, patients older than 65 years had the highest rate of out-of-network services.

Many health plans are loosening their restrictions on patients' direct access to specialists. This study provides insight into how patients will use a self-referral option, even if it is associated with increased out-of-pocket costs. The strength of patients' relationships with their regular physicians was an important determinant of self-referral. In the claims analyses, greater continuity of care, as measured by the proportion of all visits made to the regular physician, was strongly associated with a lower probability of self-referral to specialists. In the patient survey, 28% reported that they self-referred because they disagreed with their physician about the need for specialty referral. In another survey of Israeli patients, greater dissatisfaction with their regular practitioner was associated with stronger preferences for self-referral.²² The implications of this finding are that some primary care physicians may need training in communication and negotiation skills with patients; health plans could also make it easier for patients to switch primary care physicians; and patients may need education about the roles of primary care physicians acting as gatekeepers.

About 1 in 5 survey respondents used their POS option to maintain ongoing relationships with specialists. This finding was supported by the claims analyses in which chronically ill patients had significantly increased odds of self-referral.

The size of the deductible in triple-option POS health plans progressively increases from option 1 (physician-approved services) to option 3 (self-referral out of network). In all 3 POS plans, having high out-of-pocket costs for physician-approved (option 1) services significantly increased the odds of self-referral. For such enrollees, the increased cost sharing associated with self-referral appeared to be a weak disincentive, probably because there were already some out-of-pocket expenses associated with gatekeeper-approved services.

The US health care system is seeking a new balance between the freedoms and excesses represented by traditional open-access indemnity insurance system, and the restrictions and coordination embodied in closed-network managed care systems. The POS plan was developed as a hybrid of these 2 systems, and its success depends on the degree to which the balance works.

The results of this study suggest that for the majority of POS enrollees, simply having the option to bypass their regular physician is enough. For those concerned that loosening gatekeeper restrictions will lead to huge increases in uncontrolled utilization, our findings suggest that this is unlikely to materialize.

This study suggests that one advantage of the POS hybrid is that sicker patients appear to make use of the options to bypass their regular physician at greater rates than others. For example, patients with more complex and numerous comorbidities, all else equal, were substantially more likely to seek self-referrals. Our survey results underscore one benefit of such freedom of choice: patients that avail themselves of self-referred specialist services report higher satisfaction. Safran et al²³ have found that survey-based measures of primary care performance were similar among members enrolled in POS, managed indemnity, and network-model HMO health plans, suggesting that the POS self-referral option does not compromise the quality of primary care. That said, a careful assessment of some of the effects of self-referrals on other dimensions of quality, particularly concerning coordination of care,²⁴ as well as overall care efficiency awaits future research.

Over the last decade, medical educators and workforce analysts have observed that tightly structured managed care plans like HMOs make greater use of generalists and less use of specialists than do classic fee-for-service plans.²⁵ Some have suggested that as POS and other open-access managed care plans increase in market share, so too will our requirement for specialist physicians relative to generalists.²⁶ A detailed assessment of the impact of POS plans on the future needs for generalists vs specialists was not a main goal of this analysis. However, based on our documentation of the relatively modest rates of out-of-plan and self-referred specialist use by POS plan members, this study suggests that the move toward POS plans will not dramatically alter the predictions of workforce forecasters that based their projections on closed-network HMOs.

For better or for worse, innovations in the private sector of the US health care system have profound impact on patients, physicians, and all other participants. As we redesign the structure of our health insurance plans in pursuit of a well-balanced model, we must make sure we get the balance between closed-network coordination and free-market flexibility right. The study reported here suggests that to date, this appears to be the case with the POS plan, though as this hybrid innovation grows in importance, so too must continued assessment of the effects such a model is likely to have on all aspects of US health care delivery.