Do Specialist Self-Referral Insurance Policies Improve Access to HIV-Experienced Physicians as a Regular Source of Care?

10.1177/107755870527931 MCR&R Heslin et 62:5 al. /(October Access to2005) HIV-Experienced 1 Physicians

Do Specialist Self-Referral Insurance Policies Improve Access to HIV-Experienced Physicians as a Regular Source of Care? Kevin C. Heslin

Charles R. Drew University of Medicine and Science

Ronald M. Andersen Susan L. Ettner Gerald F. Kominski Thomas R. Belin

University of California, Los Angeles

Hal Morgenstern

University of Michigan

William E. Cunningham

University of California, Los Angeles Health insurance policies that require prior authorization for specialty care may be detrimental to persons with HIV, according to evidence that having a regular physician with HIV expertise leads to improved patient outcomes. The objective of this study is to determine whether HIV patients who can self-refer to specialists are more likely to have physicians who mainly treat HIV. The authors analyze cross-sectional survey data from the HIV Costs and Services Utilization Study. At baseline, 67 percent of patients had insurance that permitted self-referral. In multivariate analyses, being able to self-refer was associated with an 8–12 percent increased likelihood of having a physician at a regular source of care that mainly treats patients with HIV. Patients who can self-refer are more likely to have HIV-experienced physicians than are patients who need prior authorization. Insurance policies allowing self-referral to specialists may result in HIV patients seeing physicians with clinical expertise relevant to HIV care.

Medical Care Research and Review, Vol. 62 No. 5, (October 2005) 583-600 DOI: 10.1177/1077558705279311 © 2005 Sage Publications

583

584

MCR&R 62:5 (October 2005)

Keywords:

human immunodeficiency virus (HIV); specialists; insurance plans; utilization

Managed care organizations have been criticized for their administrative controls on utilization, particularly for policies that prohibit enrollees from self-referring to specialist physicians (Newcomer 2000). To obtain a specialist appointment that will be covered by their health plan, patients frequently must receive prior authorization from a primary care physician or other “gatekeeper” (Kongstvedt 1997). State Medicaid programs can impose this same restriction on beneficiaries by requiring them to enroll in managed care organizations (Westmoreland 1999). Obtaining authorization may be only an inconvenience for persons in relatively good health, but the time and effort involved could have adverse effects on patients with HIV. By increasing transaction costs, this insurance requirement may deter patients from seeing physicians with demonstrated clinical expertise in dealing with HIV. Because generalist physicians by definition see patients with a wide range of conditions, they are less likely to stay informed about new and effective HIV treatments. Patients who can self-refer to specialists may be better able to establish ongoing relationships with physicians who are more experienced with HIV care and thus attain better health outcomes. One of the difficulties in studying the impact of insurance coverage on access to HIV specialists is the lack of consensus on exactly which type of physician is best suited for treating patients with advanced disease (Hecht et al. 1999; Turner and Laine 2001; Stone et al. 2001). The U.S. Department of Health and Human Services (2001) has issued federal guidelines recommending that antiretroviral therapy be supervised by “an expert”; however, the definition of that term, as it pertains to HIV care, is still evolving. In 2003, the American Academy of HIV Medicine removed infectious diseases certification from its This article, submitted to Medical Care Research and Review on June 10, 2004, was revised and accepted for publication on October 7, 2004. The HIV Cost and Services Utilization Study was conducted under cooperative agreement U-01HS08578 between RAND and the Agency for Health Research and Quality. Additional funding was provided by the Health Resources and Services Administration, the National Institute of Mental Health, National Institute on Drug Abuse, National Institute of Dental Research, National Institute on Aging, Merck and Company, Glaxo-Wellcome, the Office of the Assistant Secretary for Planning and Evaluation, and the Robert Wood Johnson Foundation. Manuscript development was supported by the National Center for Research Resources, grant G12-RR03026-15, and the National Center on Minority Health and Health Disparities, grant 1 P20MD00148-01. Data analysis and interpretation were supported by the California Program on Access to Care (ZNN02M) and the University-wide AIDS Research Program, University of California (D01-LA-080).

Heslin et al. / Access to HIV-Experienced Physicians

585

credentialing requirements for HIV specialists, citing a statement by the Infectious Diseases Society of America that board certification “does not guarantee sufficient knowledge to assure that an ID specialist will remain an expert in HIV disease over time” (see www.aahivm.org/new/index.html). High HIV patient volume, or “experience,” is now a defining component of HIV expertise (Turner et al. 1994; Markson, Cosler, and Turner 1994; Markson et al. 1998). Using various caseload thresholds for categorizing providers as “experienced” or not, a number of studies have shown that higher HIV patient volume is inversely associated with mortality risk (Kitahata et al. 1996; Laine et al. 1998). More recent work shows that HIV-experienced physicians have more accurate knowledge of standards of HIV care (Stone et al. 2001; Landon et al. 2002). In response to concerns about the potential barriers of prior authorization requirements, state legislators have authored bills that give managed care enrollees with serious chronic illnesses the right to self-refer to specialists. One such piece of legislation in California, the so-called Standing Referral Law, was specifically designed to improve access to specialists for persons with HIV in that state (Gallegos and Shelley 2000). However, there is no evidence that giving HIV patients the option to self-refer has an effect on the types of regular physicians they have. There is substantial literature on the effect of insurance coverage on access to HIV care, but most studies have used broad definitions of health insurance (public, private, none) and service utilization (medical, dental, inpatient; Mor et al. 1992; Crystal 1994; Joyce et al. 1999; Heslin et al. 2001). NEW CONTRIBUTION After a managed care backlash in the late 1990s that led to less stringent utilization controls, health plans are now reportedly resuming the use of referral authorization and other forms of utilization management to control costs (Mays, Claxton, and White 2004). From the patient’s perspective, prior authorization requirements may represent a barrier to obtaining care from physicians with HIV-related expertise (Kitahata et al. 1996; Laine et al. 1998). In this study, we used data on a nationally representative sample of HIV patients to determine whether patients who can self-refer are more likely to have HIVexperienced physicians as a regular source of care, as compared with patients who need prior authorization. Understanding the impact of specialist selfreferral policies on access to care may be helpful in designing health insurance products that are beneficial to persons living with HIV disease.

586

MCR&R 62:5 (October 2005)

METHOD DESIGN The cohort from the HIV Cost and Services Utilization Study (HCSUS) is a nationally representative probability sample of HIV-positive adults receiving care in the contiguous United States. For practical reasons, the reference population was limited to persons at least eighteen years old with a known HIV infection who made at least one visit for regular or ongoing care to a nonmilitary, nonprison medical provider other than an emergency department during a specified “population definition period.” This period was January 5 to February 29, 1996, in all but one metropolitan area, where the start was delayed until March. Full details of the design are available elsewhere (Frankel et al. 1999). The HCSUS used a multistage design in which geographic areas, medical providers, and their patients were sampled. In the first stage, investigators sampled with certainty the eight metropolitan statistical areas (MSAs) with the largest AIDS caseloads, plus an additional twenty MSAs and twenty-four clusters of rural counties. In the second stage, fifty-eight urban and twentyeight rural providers were sampled from lists of providers identified by informants as HIV providers. To ensure that all HIV providers were represented, investigators sampled another eighty-seven urban and twenty-three rural providers who had affirmed caring for HIV patients in a screening survey of approximately four thousand physicians randomly selected from the Physician Master File of the American Medical Association. In the third stage, respondents were sampled from anonymous lists of all eligible patients who visited participating providers during the population definition period. We created several weights to adjust for differential selection probabilities across subgroups of the population—one to adjust for nonresponse and one to adjust for the fact that some patients had more than one opportunity to enter the sample. Applying the weights permits inference to the reference population. DATA A prospective cohort design was used for this survey, with patients participating in three waves of structured interviews conducted by the National Opinion Research Center. After obtaining informed consent, interviews were conducted using computer-assisted personal interviewing programs (91 percent) or computer-assisted telephone interviewing programs (9 percent). Of the 4,042 eligible subjects, 2,864 (71 percent) completed baseline interviews. Each interview lasted for approximately ninety minutes and included items

Heslin et al. / Access to HIV-Experienced Physicians

587

on insurance coverage and other benefits programs, use of services, measures of health status, as well as social and demographic characteristics. The start and finish dates of these interviews were baseline, January 1996 to April 1997; and first follow-up, December 1996 to July 1997. Because patients without health insurance cannot have a self-referral policy, we excluded uninsured patients from the analysis. In addition, we used data from only the baseline and first follow-up interviews because data on the dependent variable were not collected at the second follow-up interview.

CONCEPTUAL FRAMEWORK The predisposing, enabling, and need-related domains of the Behavioral Model of Health Services Utilization were used to guide our analyses (Andersen and Davidson 2001). Predisposing factors refer to demographic characteristics such as gender, race/ethnicity, and education that can directly influence an individual’s use of different types of services. Enabling factors refer to social and economic resources that facilitate the use of services, such as income, health insurance, and close friends. The need-related domain of the model refers to the presence or severity of illness. The key independent variable of this study, having insurance that allows self-referral to specialists, is conceptualized as an enabling factor. A patient’s use of HIV-experienced physicians will depend in part on the relative price, which includes the cost of obtaining a referral to such physicians. An insurance policy that allows patients to self-refer will result in decreased monetary costs and opportunity costs associated with the time and effort required to obtain a referral from a gatekeeper (Phelps 2002). This decrease in the relative price increases demand for HIV-experienced providers. DEPENDENT VARIABLES Data from the baseline and follow-up interviews were used to create two dichotomous dependent variables on access to HIV-experienced physicians. Study participants were coded as having an experienced physician if they usually saw the same physician at their regular source of care and if they answered “yes” to the following question: “Does the place where you go for most of your HIV care mainly treat people with HIV disease?” To assess the validity of this measure, we examined its correlation with HIV patient volume reported by a subset of HCSUS physicians in a separate survey (Landon et al. 2002). The average HIV caseload of physicians whose patients were coded as “yes” on our dependent variable at first follow-up was twice as large as physicians whose patients were coded as “no” (i.e., 368 vs. 183 HIV patients).

588

MCR&R 62:5 (October 2005)

KEY INDEPENDENT VARIABLE The key independent variable indicates whether patients were able to selfrefer to medical specialists. Patients were asked, “Does your insurance require you to obtain authorization before seeing a specialist for medical care?” Response options were yes/no. In this study, we were interested in comparing the average probability of having an HIV-experienced physician for patients who had self-referral insurance policies with patients who did not have such coverage, adjusting for potential confounders. COVARIATES The predisposing factors of race/ethnicity, gender, age, education, and mode of HIV infection were included in the analysis. Race/ethnicity was categorized as African American, Latino, White, and “Other” (the latter category included participants who self-identified as Alaskan Native, American Indian, Asian, Pacific Islander, or of mixed racial background). The analysis also included a variable on gender. Mode of HIV exposure was categorized as heterosexual contact, male homosexual contact, injection drug use, and other exposure (i.e., hemophilia, infected blood transfusion, or undefined). Education was categorized as bachelor’s degree or higher, associate degree or some college, high school diploma or general equivalency degree, and less than high school or no degree. The analysis included variables on enabling factors that could also affect access to physician care. We included an “insurance type” variable with categories for Medicaid, Medicare, private fee-for-service organizations, and private health maintenance organizations (HMOs). Annual household income was categorized as greater than $25,000, between $10,001 and $25,000, between $5,001 and $10,000, and less than $5,000. We included a dichotomous variable indicating whether patients had case managers because such personnel are often a source of referrals to physicians. A variable indicating the number of friends with whom respondents felt they could “talk to about private matters or call on for help” was categorized as none, one or two, three or four, and five or more close friends. Because the responsibilities of child rearing could act as a barrier to accessing physicians, we also included a dichotomous variable indicating whether respondents were living in households with children. Physician supply could affect whether patients have HIV-experienced physicians as a regular source of care. Recent modifications of the behavioral model have incorporated provider supply and other community-level variables as “contextual enabling factors” (Andersen and Davidson 2001). We created the following two variables with data from the Area Resource File: the proportion

Heslin et al. / Access to HIV-Experienced Physicians

589

of all active physicians who were infectious diseases specialists and the proportion of active physicians who were internists, aggregated at the MSA level. We also included total active physician supply, land area in square miles, and total population size to normalize the effects of the physician variables. The forty patients who did not live in MSAs (1.4 percent of the baseline sample) were assigned the mean values of the supply variables for all non-MSAs in their respective geographic regions (Midwest, Northeast, South, and West). Two clinical characteristics were included to represent the need domain of the model. We included self-reported data on helper T4 lymphocyte (CD4) count, categorized as less than 50, between 50 and 199, between 200 and 499, and greater than 500. We also included a three-category variable for disease stage that indicated whether respondents were asymptomatic, symptomatic, or had an AIDS diagnosis.

ANALYSIS We calculated frequencies and weighted proportions of categorical variables, as well as weighted means and ranges of continuous variables. Multivariate logistic regression analysis was used to test the hypothesis that patients who could self-refer to specialists were more likely to have an HIV-experienced physician than were patients who needed prior authorization, controlling for other variables. Because the prevalence of the outcome was greater than 10 percent in this analysis, odds ratios would overestimate the risk ratio (RR; Zhang and Kai 1998). For this reason, we directly calculated risk ratios using predicted probabilities from the multivariate analysis. First, we set the observations equal to one (1) for each indicator variable and generated predicted probabilities of having an experienced physician, controlling for the other variables in the model. We then set the sample equal to zero (0) on the variables and again generated predicted probabilities. The probabilities from each model were averaged, and the relative risk was estimated by dividing the average probability from the first prediction by the average probability from the second prediction. For continuous independent variables, we calculated predicted probabilities of having an HIV-experienced physician after setting each observation equal to the mean and then repeated the process after setting the observations equal to a mean-plus-one-unit increase in the independent variable. The relative risk was again estimated from the averaged quotient of the predicted probabilities from the models. Cross-tabulation of variables on self-referral policies and insurance type showed that only 24 percent of HMO enrollees had coverage that allowed selfreferral to specialists, a rate that was much lower than those for persons with

590

MCR&R 62:5 (October 2005)

Medicaid (78 percent), Medicare (75 percent), and private fee-for-service (70 percent) coverage. Because the extrapolation implied by the regression model relies strongly on assumptions about linearity, we also performed the analyses using propensity score methods (Rubin 1997). The objective of propensity score analysis is to replace a large set of confounders with one summary function called the propensity score, which in this study is the predicted probability of having insurance that allows self-referral. We used logistic regression to calculate the conditional probability of having such insurance, given other patient characteristics. Rubin (1997) suggests using these scores to stratify participants into five groups, based on quintiles. Unless there is selection on unobserved characteristics that are not correlated with the variables in the propensity model, it is valid to compare patients within each stratum because they have approximately equal probability of having the self-referral option. Propensity score analysis does not address biases that may result from unobservable factors correlated with both having an HIV-experienced physician and the ability to self-refer. However, it can perform better than standard multivariate regression in adjusting for observable factors that may not be “balanced out” among patients who can and cannot self-refer (e.g., by avoiding out-of-sample prediction). Thus, it is a useful methodology that can increase the robustness of the analysis. The propensity model of being able to self-refer had a receiver operating characteristic (ROC) index of .85 and contained 103 variables (including eighteen second-order terms). Previous work using propensity models have cited ROC index values greater than .80 as indicating high predictive accuracy (Rose et al. 2000). A relative risk was calculated within each quintile as an estimate of the effect of self-referral insurance policies, and an externally standardized average of the ratios (sRR) was calculated (Greenland 1982). The standard was drawn from the set of weights that HCSUS investigators created to estimate key population prevalence parameters. These weights were used for the univariate and other multivariate analyses as well. Bootstrapping was used to estimate 95 percent confidence intervals (CI; Mooney and Duval 1993). The Stata software package was used to adjust standard errors.

RESULTS DESCRIPTIVE DATA Table 1 shows the characteristics of insured HIV patients in the United States for 1996–1997. At baseline and first follow-up interviews, 70 percent and 79 percent of participants had an HIV-experienced physician, respectively. Approximately 67 percent of HIV patients had health insurance that

Heslin et al. / Access to HIV-Experienced Physicians

591

allowed self-referral to specialists at baseline. By the follow-up interview, 59 percent of patients had such coverage. At baseline, 74 percent of patients with self-referral policies had HIV-experienced physicians, compared with a rate of 66 percent among patients needing prior authorization; by first followup interview, these rates were 83 percent and 73 percent, respectively (not shown in the table). The majority of patients were men (77.4 percent). High proportions of the sample were white (49.2 percent) and exposed to HIV through sex with other men (48.6 percent). The mean age of participants was thirty-nine years old, and the maximum age was seventy-seven. Findings on other patient characteristics suggest that a large proportion of the sample was from disadvantaged backgrounds. For example, 25 percent of participants had not finished high school. Approximately 20 percent of patients had annual incomes under $5,000 (Table 1). MULTIVARIATE RESULTS Results from the multivariate analysis are shown in Table 2. Consistent with the hypothesis, patients with self-referral coverage were more likely to have an HIV-experienced physician at baseline (RR = 1.12; 95 percent CI = 1.01, 1.25) and at follow-up (RR = 1.12; 95 percent CI = 1.02, 1.26), compared with patients who needed prior authorization. The propensity score analysis generated comparable results. Being able to self-refer was associated with having an experienced physician at baseline (sRR = 1.08; 95 percent CI = 1.01, 1.15) and at follow-up (sRR = 1.09; 95 percent CI = 1.03, 1.15; not shown in the table). Because we were not able to identify whether any individuals within the HMO category were actually in preferred provider organizations, we performed a sensitivity analysis that restricted the model to the HMO group only (n = 474). The resulting point estimates suggested a positive effect of selfreferral coverage on access to physicians; however, confidence intervals included the null value for both baseline (RR = 1.24; 95 percent CI = 0.88, 2.26) and follow-up (RR = 1.40; 95 percent CI = 0.72, 4.76). Several other patient characteristics were associated with having an HIVexperienced physician in the multivariate analysis. At baseline, African Americans were less likely than were Whites to have physicians who mainly treated HIV patients (Table 2). Female gender and age were both positively associated with having an expert physician at baseline (Table 2), as was having a case manager at follow-up (Table 3). Persons with annual household incomes between $10,000 and $24,999 were less likely to have expert physicians at follow-up than were those in the highest income category (Table 3). Several contextual variables were also important at the MSA level: land area at baseline (Table 2), as well as total population at follow-up (Table 3), suggesting that

592

MCR&R 62:5 (October 2005)

TABLE 1 Characteristics of Insured HIV Patients in the United States, 1996 to 1997 Variable Had an HIV-experienced physician at baseline Had an HIV-experienced physician at first follow-up Insurance allowed self-referral at baseline Insurance allowed self-referral at follow-up Racial/ethnic category African American Latino White Other Female gender Primary HIV exposure Heterosexual contact Injection drug use Male homosexual contact Other Education Less than high school High school Some college Bachelor’s degree Annual household income Less than $5,000 $5,000–$9,999 $10,000–$25,000 Greater than $25,000 Type of insurance Private FFS Private HMO Medicaid Medicare Had children in household Number of friends None One or two Three or four Five or more Had a case manager

n

%

SE

1,300 1,483 1,209 1,000

70.3 78.7 66.5 59.1

4.1 3.9 2.8 4.8

959 415 1,399 91 847

32.8 14.8 49.2 3.2 22.6

3.1 1.9 2.6 0.7 2.5

578 696 1,303 287

18.4 24.1 48.6 8.9

2.4 2.9 4.3 0.9

723 805 810 526

24.9 27.4 28.4 19.3

3.1 1.1 1.8 2.4

609 740 736 779

19.7 25.8 26.0 28.5

1.6 1.7 1.3 3.0

391 474 858 544 500

16.5 15.3 29.2 19.2 14.9

3.0 1.8 2.8 0.9 1.9

280 778 564 583 1,789

12.5 35.3 26.0 26.3 60.5

0.9 2.0 1.4 2.0 2.6

(continued)

Heslin et al. / Access to HIV-Experienced Physicians

593

TABLE 1 (continued) Variable CD4 count (lowest ever) Greater than 500 200–499 50–199 0–49 Disease stage Asymptomatic Symptomatic AIDS

n

%

SE

253 1,096 854 661

9.5 37.4 29.5 23.6

0.8 1.7 1.3 1.6

214 1,324 928

10.5 53.3 36.3

0.7 1.5 1.7

M

SE

HIV volume of patients’ physicians Patient age (years)

336 38.7

50 0.3

MSA-Level Variables

M

SD

Infectious diseases specialists/doctors (× 100) Internists/doctors (× 100) Total number of doctors Population, 1995 Land area in square miles, 1990

0.5% 13.0% 4,275 1,577,216 1,269

0.3% 5.0% 7,688 2,877,454 1,412

Note: Number values are unweighted; means, percentages, and standard errors are weighted. FFS = fee for service; HMO = health maintenance organization; MSA = metropolitan statistical area.

patients living in larger MSAs with higher population density were more likely to have physicians who mainly treat HIV patients.

DISCUSSION Persons with HIV disease have criticized insurers that require prior specialist authorization because data suggest that provider HIV experience is associated with patient survival (Kitahata et al. 1996; Laine et al. 1998). Compared with the privately insured, HIV-positive beneficiaries of Medicaid and other public programs use fewer ambulatory (Mor et al. 1992; Crystal 1994) and inpatient (Joyce et al. 1999) services. Similar associations of insurance with more global measures of service use, such as patient-reported unmet need for care, are also reported in the HIV literature (Heslin et al. 2001). Several studies

594

MCR&R 62:5 (October 2005)

TABLE 2 Multivariate Logistic Regression Analysis of Having a Physician who Treats Mostly HIV at Baseline HCSUS Interview (N = 2,159)a Variable (Reference Group) Insurance allowed patient self-referral Racial/ethnic category (white) African American Latino Other Age (five-year increments) Female gender (male) HIV exposure (male homosexual contact) Heterosexual contact Injection drug use Other Education (greater than BS/BA) Less than high school High school Some college Annual Household Income (greater than $25,000) $0–$4,999 $5,000–$9,999 $10,000–$24,999 Insurance type (private FFS) Medicaid Medicare Private HMO Had children at home Number of close friends (greater than 5) None 1–2 3–4 Had a case manager CD4 count (greater than 500) 200–499 50–199 0–49 Disease stage (AIDS) Asymptomatic Symptomatic

Has HIV-Experienced Physician Risk Ratio (95% CI) 1.12

(1.02, 1.25)

0.85 1.08 0.91 1.06 1.14

(0.76, 0.94) (0.95, 1.22) (0.71, 1.13) (1.01, 1.12) (1.01, 1.30)

0.97 1.01 0.96

(0.81, 1.12) (0.90, 1.13) (0.80, 1.12)

0.87 0.96 0.93

(0.73, 1.02) (0.83, 1.09) (0.81, 1.06)

1.09 1.00 1.02

(0.94, 1.27) (0.86, 1.16) (0.91, 1.16)

0.88 0.92 0.87 1.01

(0.74, 1.03) (0.77, 1.09) (0.74, 1.03) (0.89, 1.14)

0.88 0.93 0.94 1.00

(0.75, 1.01) (0.83, 1.03) (0.84, 1.05) (0.91, 1.09)

0.98 0.96 0.98

(0.83, 1.14) (0.81, 1.12) (0.81, 1.15)

1.06 0.99

(0.96, 1.17) (0.82, 1.18) (continued)

Heslin et al. / Access to HIV-Experienced Physicians

595

TABLE 2 (continued) Variable (Reference Group) Infectious diseases specialists/all doctors Internists/all doctors Total number of doctors Land area (square miles) Total population

Has HIV-Experienced Physician Risk Ratio (95% CI) 0.99 1.01 1.00 0.83 1.04

(0.96, 1.01) (0.99, 1.03) (0.96, 1.01) (0.74, 0.89) (1.01, 1.07)

Note: CI = confidence interval; HCSUS = HIV Cost and Services Utilization Study; FFS = fee for service; HMO = health maintenance organization. a. Reference category is “nonexperienced doctor.”

TABLE 3 Multivariate Logistic Regression Analysis of Having a Physician who Treats Mostly HIV at First Follow-Up HCSUS Interview (N = 1,791)a Variable (Reference Group) Insurance allowed patient self-referral Racial/ethnic category (white) African American Latino Other Age (five-year increments) Female gender (male) HIV exposure (male homosexual contact) Heterosexual contact Injection drug use Other Education (greater than BS/BA) Less than high school High school Some college Annual household income (greater than $25,000) $0–$4,999 $5,000–$9,999 $10,000–$24,999 Insurance type (private FFS) Medicaid Medicare Private HMO

Has HIV-Experienced Physician Risk Ratio (95% CI) 1.12

(1.03, 1.28)

1.07 1.09 0.92 1.06 1.02

(0.96, 1.23) (0.95, 1.26) (0.73, 1.14) (1.00, 1.12) (0.90, 1.17)

1.02 1.04 0.86

(0.89, 1.19) (0.92, 1.20) (0.72, 1.00)

1.06 1.04 1.10

(0.91, 1.25) (0.94, 1.21) (0.98, 1.26)

1.02 1.01 0.89

(0.88, 1.20) (0.87, 1.17) (0.80, 0.98)

1.05 1.04 0.99

(0.89, 1.27) (0.91, 1.24) (0.87, 1.13) (continued)

596

MCR&R 62:5 (October 2005)

TABLE 3 (continued) Variable (Reference Group) Had children at home Had HIV-experienced doctor at baseline Number of close friends (greater than 5) None 1–2 3–4 Had a case manager CD4 count (greater than 500) 200–499 50–199 0–49 Disease stage (AIDS) Asymptomatic Symptomatic Infectious diseases specialists/all doctors Internists/all doctors Total number of doctors Land area (square miles) Total population

Has HIV-Experienced Physician Risk Ratio (95% CI) 1.09 2.40

(0.94, 1.31) (2.21, 3.22)

0.95 1.01 1.03 1.13

(0.82, 1.11) (0.91, 1.11) (0.92, 1.16) (1.01, 1.29)

0.97 1.01 1.00

(0.83, 1.12) (0.86, 1.20) (0.84, 1.18)

0.94 0.96 1.01 0.98 0.99 0.91 1.06

(0.80, 1.13) (0.87, 1.07) (0.99, 1.03) (0.96, 0.99) (0.95, 1.00) (0.84, 0.95) (1.03, 1.10)

Note: CI = confidence interval; HCSUS = HIV Cost and Services Utilization Study; FFS = fee for service; HMO = health maintenance organization. a. Reference category is “nonexperienced doctor.”

of the general patient population have examined the use of specialists by persons in gatekeeper and point-of-service plans (Forrest et al. 2001); however, previous work has not examined whether HIV patients who can self-refer are more likely to have HIV-experienced physicians as a regular source of care. The present study suggests that HIV patients with direct access to specialists are more likely to have a physician at a regular source of care that mainly treats HIV disease. This finding may be useful to consumers and other decision makers in the HIV services system. State policy makers have targeted insurance plans that restrict the use of specialists for reform in recent years. In 1997, a law was passed in California that ensures self-referrals for any health plan enrollee with “a condition or disease that requires specialized care over a prolonged period of time and is lifethreatening, degenerative, or disabling” (Escutia 1997). Because the American Medical Association does have an HIV specialist designation, some health plans did not recognize HIV as a specialty area under this law. Legislators

Heslin et al. / Access to HIV-Experienced Physicians

597

subsequently passed an amendment that named HIV disease as one of the medical conditions for which HMOs are required to extend open referrals to their enrollees (Gallegos and Shelley 2000). Concurrently, the American Academy of HIV Medicine (2003), a nonprofit accrediting organization focusing on quality in HIV/AIDS care, began issuing standards to define “HIV specialist.” As formal education and certification programs are established, and as the definition of HIV specialty care continues to evolve, further research will be important in monitoring differences in the use of HIV specialists among persons with various types of public and private health insurance. Several limitations in this study should be mentioned. An ongoing challenge in survey research is the development of valid self-reported measures of insurance coverage. Analyses of a sample from the general population suggest that many consumers overestimate plan restrictions on use of specialists (Cunningham, Denk, and Sinclair 2001). However, persons with serious conditions such as HIV disease are likely to have accurate knowledge of their coverage, simply because they have more occasions to use it. Although it is possible that some of the respondents in the HMO category were enrolled in preferred provider organizations, we unfortunately did not have the data to make that distinction in our analyses. Self-selection is also an issue in this study. Patients may choose insurance plans that provide easier access to specialists because they fully intend to see experienced physicians. This unobserved motivation may determine both who can self-refer and who has an experienced physician, biasing upward the estimated association of these two variables. Even so, if the unobserved characteristics that partly determine insurance and physician selection are correlated with the other independent variables, then we have addressed this issue to some extent with the propensity score models. Because patient characteristics such as race/ethnicity, gender, and age were independently associated with the dependent variables, we need to emphasize that the self-referral option alone will not ensure access to HIV-experienced physicians. Finally, there may be limitations in the generalizability of these findings to the current patient population. Because of the large number of HIV medications now available, current patients may differ from their counterparts in the late 1990s. This study examined how insurance that allows self-referral relates to the use of HIV-experienced physicians, but this is only one of several relevant outcomes that could be associated with such coverage for a number of different patient populations. One justification for allowing self-referral has been the potential savings that specialists may bring to disease management by avoiding a more trial-and-error approach to care. As treatment options increase, future research should examine whether these cost savings and other efficiencies are achieved by specialists in HIV, cancer, and other serious chronic

598

MCR&R 62:5 (October 2005)

illnesses. Self-referral coverage is probably important in facilitating one-time consultations with specialists for patients whose regular physicians are generalists, and this is also an underexamined area. Finally, it would also be instructive to know whether managed care enrollees who can self-refer are more satisfied with their health plans than are those who need prior authorization because gatekeeping is often a source of patient dissatisfaction. Research on these issues may help guide the design of provider referral policies that will optimize the care of persons living with serious chronic illness and, by extension, improve their quality of life and long-term survival.

REFERENCES American Academy of HIV Medicine. 2003. Introduction to AAHIVM credentialing. www.aahivm.org/new/index.html (accessed September 4, 2003). Andersen, R. M., and P. Davidson. 2001. Improving access to care in America: Individual and contextual indicators. In Changing the U.S. health care system, 2nd ed., edited by. R. M. Andersen, T. H. Rice, and G. F. Kominski, 13–40. San Francisco: Jossey-Bass. Crystal, S. 1994. Sources of outpatient care and insurance status over the course of AIDS: Implications for health policy. AIDS & Public Policy Journal 9:129–37. Cunningham, P. J., C. Denk, and M. Sinclair. 2001. Do consumers know how their health plan works? Health Affairs 20:159–66. Escutia, M. 1997. Bill number AB 1181. Legislative Counsel’s Digest, State of California. February 28. www.leginfo.ca.gov/pub/bill. Forrest, C. B., J. P. Weiner, J. Fowles, C. Vogeli, K. D. Frick, K. W. Lemke, and B. Starfield. 2001. Self-referral in point-of-service plans. Journal of the American Medical Association 285:2223–31. Frankel, M. R., M. F. Shapiro, N. Duan, S. C. Morton, S. H. Berry, J. A. Brown, M. A. Burnam, S. E. Cohn, D. P. Goldman, D. F. McCaffrey, S. M. Smith, P. A. St. Clair, J. F. Tebow, and S. A. Bozzette. 1999. National probability samples in studies of lowprevalence diseases, part 2: Designing and implementing the HIV Cost and Services Utilization Study sample. Health Services Research 34:969–92. Gallegos, M., and K. Shelley. 2000. Bill number AB 2168. Legislative Counsel’s Digest, State of California. February 23. www.leginfo.ca.gov/pub/bill. Greenland, S. 1982. Interpretation and estimation of summary ratios under heterogeneity. Statistics in Medicine 1:217–27. Hecht, F. M., I. B. Wilson, A. W. Wu, R. L. Cook, and B. J. Turner. 1999. Optimizing care for persons with HIV infection. Annals of Internal Medicine 131:136–43. Heslin, K. C., W. E. Cunningham, M. Marcus, I. Coulter, J. Freed, C. Der-Martirosian, S. A. Bozzette, M. F. Shapiro, S. C. Morton, and R. M. Andersen. 2001. A comparison of unmet needs for dental and medical care among persons with HIV infection receiving care in the United States. Journal of Public Health Dentistry 61:14–21.

Heslin et al. / Access to HIV-Experienced Physicians

599

Joyce, G. F., D. P. Goldman, A. Leibowitz, D. Carlisle, N. Duan, M. F. Shapiro, and S. E. Bozzette. 1999. Variation in inpatient resource use in the treatment of HIV: Do the privately insured receive more care? Medical Care 37:220–27. Kitahata, M. M., T. D. Koepsell, R. A. Deyo, C. L. Maxwell, W. T. Dodge, and E. H. Wagner. 1996. Physicians’ Experience with the Acquired Immunodeficiency Syndrome as a factor in patients’ survival. New England Journal of Medicine 334:701–06. Kongstvedt, P. R. 1997. Essentials of managed health care. New York: Aspen. Laine, C., L. E. Markson, L. J. McKee, W. W. Hauck, T. R. Fanning, and B. J. Turner. 1998. The relationship of clinic experience with advanced HIV and survival of women with AIDS. AIDS 12:417–24. Landon, B. E., I. B. Wilson, N. S. Wenger, S. E. Cohn, C. J. Fichtenbaum, S. A. Bozzette, M. F. Shapiro, and P. D. Cleary. 2002. Specialty training and specialization among HIV physicians in the United States. Journal of General Internal Medicine 17 (1): 12–22. Markson, L. E., L. E. Cosler, and B. J. Turner. 1994. Implications of generalists’ slow adoption of Zidovudine in clinical practice. Archives of Internal Medicine 154:1497–1504. Markson, L. E., R. Houchens, T. R. Fanning, and B. J. Turner. 1998. Repeated emergency department use by HIV-infected persons: Effect of clinic accessibility and expertise in HIV Care. Journal of the Acquired Immune Deficiency Syndromes 17:35–41. Mays, G. P., G. Claxton, and J. White. 2004. Managed care rebound? Recent changes in managed care plans’ cost containment strategies. Health Affairs. August 11, 2004. http://content.healthaffairs.org/cgi/content/full/hlthaff.w4.427/DCI Mooney, C. Z., and R. D. Duval. 1993. Bootstrapping: A nonparametric approach to statistical inference. Newbury Park, CA: Sage. Mor, V., J. A. Fleishman, M. Dresser, and J. Piette. 1992. Variation in health service use among HIV-infected patients. Medical Care 30 (1): 17–29. Newcomer, L. N. 2000. Is the gatekeeper just another buggy whip? American Journal of Managed Care 6 (11): 1253–54. Phelps, C. 2002. Health economics. Boston: Addison-Wesley. Rose, J. H., E. E. O’Toole, N. V. Dawson, C. Thomas, A. F. Connors, N. Wenger, R. S. Phillips, M. B. Hamel, D. T. Reding, H. J. Cohen, and J. Lynn. 2000. Generalists and oncologists show similar care practices and outcomes for hospitalized late-stage cancer patients. Medical Care 38: 1103–18. Rubin, D. B. 1997. Estimating causal effects from large data sets using propensity scores. Annals of Internal Medicine 127: 757–63. Stone, V. E., F. F. Mansourati, R. M. Poses, and K. H. Mayer. 2001. Relation of physician specialty and HIV/AIDS experience to choice of guideline-recommended antiretroviral therapy. Journal of General Internal Medicine 16: 360–68. Turner, B. J., and C. Laine. 2001. Differences between generalists and specialists: Knowledge, realism, or primum non nocere? Journal of General Internal Medicine 16:422–24. Turner, B. J., L. McKee, T. F. Fanning, and L. E. Markson. 1994. AIDS specialist versus generalist ambulatory care for advanced HIV infection and impact on hospital use. Medical Care 32:902–16.

600

MCR&R 62:5 (October 2005)

U.S. Department of Health and Human Services. 2001. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. Rockville, MD: HIV/AIDS Treatment Information Service. Westmoreland, T. 1999. Medicaid and HIV/AIDS policy: A basic primer. Washington, DC: The Henry J. Kaiser Family Foundation. Zhang, J., and K. F. Kai. 1998. What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes. Journal of the American Medical Association 280:1690–91.