This study uses data from the Community Tracking Study (CTS) physician survey, a longitudinal data collection effort sponsored by The Robert Wood Johnson Foundation.9
The physician survey uses a clustered sample that is representative of direct patient care physicians in the continental United States, as well as in selected communities.10–12
The survey collects information on practice arrangements and ownership, sources of practice revenue and compensation, and physicians' views on their medical practice.
In the first stage of sample selection, 60 sites (defined as counties or groups of counties using conventionally accepted definitions of economic areas) were randomly selected with probability in proportion to the population. Sites were first stratified by region of the country and according to medium and large metropolitan sites (200,000 persons or more), small metropolitan sites (less than 200,000 persons), and nonmetropolitan sites to ensure representation of these areas.
The sample of physicians was drawn from the master files of the American Medical Association and the American Osteopathic Association. The full sample included active nonfederal office and hospital-based physicians in selected specialties who spent at least 20 hours per week in direct patient care, but excluded radiology, anesthesiology, pathology, and a few nonpatient care specialties, such as legal medicine. Within the CTS physician survey sampling frame, primary care physicians were intentionally oversampled. Primary care included family practice, general practice, internal medicine, general pediatrics, general internal medicine/pediatrics, adolescent medicine, and geriatric medicine. Specialists within these self-designated specialty areas who reported spending more time practicing general internal medicine than practicing their subspecialty were categorized as primary care regardless of self-designated specialty. Thus, internists who were trained in a medical subspecialty and/or who self-identified as medical subspecialists, but who reported spending most of their practice time in general medicine were counted as general internists. Using these definitions, the analyses in this study were conducted using the subsample of primary care physicians.
All estimates presented in this article were weighted to be representative of nonfederal patient care physicians in the continental United States. Weights were constructed to produce nationally representative estimates from the 60 sites and to restore proportionality to the sample arising from survey nonresponse and the clustering of the sample into 60 sites of varying size and geographic location. All estimates were produced using SUDAAN software, which adjusted the standard errors of estimates to account for the complex sample design.13
Ability to arrange mental health referrals and admissions was part of a series of questions physicians were asked regarding being able to obtain medically necessary services for their patients. For 6 types of services, physicians were asked how often they were able to obtain medically necessary services with a 6-point response scale of always, almost always, frequently, sometimes, rarely, or never. For mental health, physicians were asked separately whether they could obtain high-quality inpatient mental health care and high-quality outpatient mental health services. Physicians also were asked about referrals to specialists, ancillary services, nonemergency hospital admissions, and diagnostic imaging. The mental health services questions were asked of primary care physicians only.14
To categorize primary care physicians' reports of experiencing barriers to mental health services, the 6-point scale was dichotomized as never, rarely, or sometimes versus always, almost always, or frequently. Sensitivity analyses were conducted to confirm that the results did not change if the 6-point scale was dichotomized as never or rarely versus always, almost always, frequently or sometimes.
The CTS physician survey elicited information on practice size, type of practice setting, ownership, and other information for each physician's practice. These characteristics were used to categorize physicians' type of practice into 5 mutually-exclusive categories: solo and small-group practice of fewer than 10 physicians, private groups of 10 or more physicians, physicians in staff or group model HMOs, physicians practicing in medical school or hospital-based practices, and other arrangements, such as clinic settings. Specialty was categorized into 3 areas: 1) pediatrics; 2) family practice and general practice; and 3) internal medicine, medicine-pediatrics, geriatrics, and other, which we refer to as “internal medicine and other.”
The gatekeeper role for primary care physicians may expand the scope of care, in which they become largely responsible for managing common, uncomplicated mental and emotional problems.6
Prior studies have shown that gatekeeping mechanisms imposed in some managed care settings impede access to specialized mental health services.15
In studies of pediatric referrals, however, gatekeeping was associated with higher rates of specialty referral.16
For the CTS physician survey, gatekeeping was defined as insurance plans or medical groups requiring enrollees to obtain permission from a primary care physician before seeing a specialist. Primary care physicians were asked the percentage of their caseload for which they served in that role. Physicians reporting gatekeeping for less than 10% of the caseload were categorized separately to identify primary care physicians with little to no managed care practice. On the other end, physicians with gatekeeping responsibilities for 70% or more of the caseload were categorized separately to identify primary care physicians with all or almost all of their caseload under managed care.
Primary care physicians may lack the requisite time to optimally manage emotional and behavioral health problems.4
This limited time availability is often cited as a factor contributing to undertreatment of major depression in primary care practice.5,6,17,18
In the telephone survey, physicians were asked whether they agree strongly, agree somewhat, disagree somewhat, disagree strongly, or neither agree nor disagree with the statement “I have adequate time to spend with my patients during typical office/patient visits.” For purposes of this analysis, physicians' perceptions of the adequacy of time with their patients was dichotomized and considered as “not sufficient time” if the physician “disagreed strongly.”
The number of psychiatrists per 100,000 population in 1997 was obtained from the Area Resource File to reflect the availability of mental health resources at the market level. Unreliable and missing counts of psychiatric beds at the county level precluded its use as a measure in this model.
Other factors, such as physician caseload, physician characteristics, and market characteristics also may influence the likelihood that primary care physicians will report problems obtaining needed mental health services. In particular, patients' insurance coverage may limit their ability to obtain mental health services, especially for the uninsured, Medicaid, and Medicare patients. Psychiatrists who are in short supply may be less willing to accept Medicaid and Medicare patients. We therefore controlled for physicians who are in the top quartile of physicians in revenue for Medicaid (Medicaid revenue exceeds 25% of practice revenue) and the top quartile of physicians in revenue for Medicare (Medicare revenue exceeds 50% of practice revenue). To control for physicians who have caseloads with a higher percentage of uninsured, we identify physicians who provide more than 10 hours of charity care per month or who work in a community health center/clinic setting.
Our analysis also controlled for relevant personal and market characteristics. We controlled for gender using information obtained from the master files of the American Medical Association and the American Osteopathic Association. Controls for race reflect self-reports where a physician is classified as minority if they do not consider themselves to be white or Caucasian. Our model also controlled for how long the physician had been practicing, because physicians with fewer years of experience may have different skills and expectations about the scope of their practice. We also controlled for the market-level managed-care penetration in the each of the 60 CTS sites, using the site average for physician practices' percentage of managed care revenue.
We used logistic regressions to model the likelihood that physicians reported problems obtaining outpatient services and inpatient care. Physicians were considered to have difficulties obtaining care if they reported never, rarely, or only sometimes being able to obtain the care sought.
Measures of years in practice, charity care provision, percentage of revenue from Medicare and Medicaid, extent of gatekeeping, managed care penetration, and the number of hospital beds and psychiatrists in a county were categorized to capture nonlinearities.
We report adjusted predicted means for each attribute in our analysis as our primary results. The adjusted predicted means were calculated from the predicted means from each logistic regression assuming all physicians in the sample had the attribute of interest, while all other characteristics were held to their original value. Statistical inferences were based on the underlying logistic regression coefficients. The full logistic regression results are also presented with the odds ratios adjusted to estimate relative risk.19