Predictors and Consequences of Negative Physician Attitudes Toward HIV-Infected Injection Drug Users

Background  We evaluated physicians’ training, experience, and practice characteristics and examined associations between their attitudes toward human immunodeficiency virus (HIV)–infected persons who are injection drug users (IDUs) and quality of care.

Methods  Cross-sectional surveys were conducted among a probability sample of noninstitutionalized HIV-infected individuals in the United States and their main HIV care physicians. Physician and practice characteristics, training, HIV knowledge, experience, attitudes toward HIV-infected IDUs, stress levels, and satisfaction with practice were assessed. The main quality-of-care measures were patient exposure to highly active antiretroviral therapy, reported problems, satisfaction with care, unmet needs, and perceived access to care.

Results  Nationally, 23.2% of HIV-infected patients had physicians with negative attitudes toward IDUs. Seeing more IDUs, having higher HIV treatment knowledge scores, and treating fewer patients per week were independently associated with more positive attitudes toward IDUs. Injection drug users who were cared for by physicians with negative attitudes had a significantly lower adjusted rate of exposure to highly active antiretroviral therapy by December 1996 (13.5%) than non-IDUs who were cared for by such physicians (36.1%) or IDUs who were cared for by physicians with positive attitudes (32.3%). Physician attitudes were not associated with other problems with care, satisfaction with care, unmet needs, or perceived access to care.

Conclusions  Negative attitudes may lead to less than optimal care for IDUs and other marginalized populations. Providing education or experience-based exercises or ensuring that clinicians have adequate time to deal with complex problems might result in better attitudes and higher quality of care.

As of December 2002, a reported 34% of all individuals with AIDS and 43% of women with AIDS were injection drug users (IDUs).¹ Medical complications and behavioral problems that accompany substance abuse can complicate the medical management of human immunodeficiency virus (HIV)–positive individuals,^2,3 and many IDUs do not receive optimal HIV therapy.^4-7 This may be in part because of concern that IDUs who are HIV infected may not adhere to treatment plans^7,8 or that medications may interact if taken concomitantly with “street drugs.”^2,9 Discomfort in or negative attitudes toward treating IDUs may also be factors.¹⁰ Researchers have studied physician attitudes toward drug users, especially those with HIV infection,^11,12 but few have explored the associations of physician characteristics and attitudes with health care quality.

Experience and expertise generally are important predictors of health care quality for HIV-infected persons^13,14 and IDUs.^4,5,10 Laine and colleagues⁴ found that receiving care from an HIV-focused provider and regular substance abuse treatment predicted adequate antiretroviral therapy intensity. Gerbert and colleagues¹⁰ found that physicians who had treated 10 or more HIV-infected patients were less negative toward stigmatized groups, such as IDUs. Strathdee and colleagues⁵ found that HIV-infected IDUs in alcohol and other drug treatment programs and those with physicians who had more experience treating HIV infection were more likely to be receiving antiretroviral drugs.

Experience treating IDUs might help physicians develop the understanding and skills that make it easier to treat such patients more effectively. Therefore, we hypothesized that such experience would be associated with positive attitudes. Because attitudes might be linked to concerns about compliance, we also hypothesized that attitudes would be associated with quality of care, especially with respect to the use of highly active antiretroviral therapy (HAART) for IDUs. To study these issues, we analyzed data from a national study of HIV-infected persons and their clinicians. These data are cross-sectional, so the analyses can show whether the data are consistent with our hypotheses but cannot confirm them.

Subjects were HIV-infected patients participating in the HIV Cost and Services Utilization Study (HCSUS) and their primary HIV care physicians. The HCSUS evaluated a nationally representative sample of 2864 noninstitutionalized adults infected with HIV in the contiguous United States who had made at least 1 visit to a nonmilitary, nonprison medical provider other than an emergency department. The HCSUS enrolled 2864 patients who completed a baseline interview and 2 subsequent interviews about their care.^15-17

In the second patient interview, conducted in 1997, subjects were asked to identify their physicians. If a patient was unable to identify a principal HIV care physician, the physician whom the patient saw most recently for HIV care was selected. If there was no such physician, a non-HIV primary care physician whom the patient identified was selected. This process was repeated using the baseline survey if no physician was identified in the follow-up survey. Physicians were identified for 2642 patients (92.2% of the sample).¹³ Study procedures were approved by institutional review boards at the primary participating institutions.

Mailing addresses were confirmed for 551 (79.6%) of 692 identified physicians.¹⁸ A survey with a $25 check for study participation was sent to them between September 1998 and July 1999. Nonrespondents received 3 follow-up telephone calls and were offered the chance to complete the survey by telephone.¹³

Of the 551 clinicians contacted, 412 returned the survey and 411 (74.6% of the 551 contacted clinicians and 59.4% of the 692 identified clinicians) completed all the questions used in the analyses herein. Thirty-three respondents were eliminated who were nurses, nurse practitioners, or physician assistants (who cared for 76 HCSUS patients). Patients who identified 5 other physicians were deemed ineligible because they were cared for at multiple sampling sites, leaving a total of 373 responses from physicians who cared for 1820 patients.

The physician survey asked about age, sex, race, ethnicity, income, years practicing as a physician, and sexual preference. Physicians were also asked what proportion of their HIV practice comprised patients with the following risk factors for HIV transmission: male-to-male sexual contact, injection drug use, and high-risk heterosexual behavior. Three groups of physicians were defined using information about medical training and self-rated HIV expertise, including infectious disease specialists, general medicine HIV experts, and general medicine non–HIV experts.¹⁵ Fourteen physicians who identified themselves as having “other” medical backgrounds were combined with the general medicine group and classified either into the general medicine HIV experts or general medicine non-HIV experts group, depending on their self-rated HIV expertise.

Knowledge of HIV treatment and management was assessed using an 11-item scale (score range, 0-11; Cronbach α = .64). Physicians were also asked about their current HIV caseload, number of patients seen per week, hours spent on patient care per week, and number of times they attended local or national professional meetings related to HIV treatment.

Physician stress was measured by 3 questions about personal and professional stress, long work hours, and demoralization about the state of medical practice. Responses to the 3 items were summed up to create a stress score (score range, 3-15; Cronbach α = .85). Physicians were also asked to rate satisfaction with their practice as poor, fair, good, very good, or excellent.

Two questions measured attitudes toward HIV-infected IDUs. The first asked if “treating IV drug users seems futile” and the other asked whether “when given a choice, I would not treat intravenous drug users with HIV infection” (response options: strongly disagree, disagree, unsure, agree, or strongly agree). The sum of these 2 items was the negative attitude score (score range, 2-10; Cronbach α = .85). We also created a dichotomous variable indicating if physicians agreed or strongly agreed with 1 or both of these items. If they did, we refer to the physicians as having a negative attitude; if not, we say the physicians have a positive attitude.

Patient survey data were used to create 5 measures of health care quality, including the following: (1) a yes or no question about whether a patient was exposed to HAART by December 1996,¹⁸(2) the sum of responses to 10 questions about problems with care (problem score) (score range, 0-100; Cronbach α = .75),¹⁹(3) a question about patient satisfaction with medical care (score range, 0-100),¹⁹(4) the number of unmet nonmedical needs (income assistance, a place to live, home health care, mental counseling, and alcohol and other drug treatment) (Cronbach α = .42),²⁰ and (5) six questions about perceived access to care, including hospital admission or emergency department care, avoidance of care because of cost, access to medical specialists, conveniently located medical care, and ability to get medical care when needed (score range, 0-100; Cronbach α = .75).¹⁸ Three hundred eighteen patients who reported using amphetamines, cocaine, crack cocaine, or heroin during the previous 12 months were classified as IDUs. Those who reported using only sedatives, analgesics, marijuana, hashish, inhalants, lysergic acid diethylamide (LSD), or hallucinogens were not defined as IDUs for the analyses herein.

We present means and frequencies to describe physician characteristics. We also present physician characteristics weighted by the number of patients treated. Those data indicate what percentage of HIV-infected patients in the country meeting HCSUS eligibility criteria have physicians with the characteristics presented.

We assessed the relationships between physician characteristics and attitudes using Pearson product moment correlations, t tests, and F tests. To assess the independent predictors of negative attitude scores, we estimated linear regression models. We considered all variables that had a significant bivariate association with attitudes as potential independent variables and used a forward selection procedure (P<.05) to select independent predictors.

We tested for differences in means for each of the health care quality indicators between IDUs and non-IDUs and between patients under the care of physicians with negative and positive attitudes, adjusting for patient age, sex, race, income,insurance, lowest CD4 cell count, mental health score (Cronbach α = .80), and physical functioning (Cronbach α = .91), as well as physician specialty and expertise. The model predicting exposure to HAART controlled for whether the patient said that HAART was definitely or probably worth receiving. We included the number of needs a patient had when evaluating the effect of physician negative attitudes on unmet needs.

Analyses examining the relationship between physician characteristics and their attitudes toward IDUs were conducted with the physician as the unit of analysis. Analyses of quality of care were conducted at the patient level and incorporated weights to account for sampling probabilities, missing patient data, and physician nonresponse.¹⁷ Hierarchical models were estimated using the general linear model for mixture distributions to account for the clustering effects resulting from the complex sampling design and the clustering of patients by physicians. The link function for prediction of being offered HAART was a logit function. Least squares means were calculated to estimate adjusted means of other quality-of-care measures.

In this study, 17.1% of the patients were current IDUs (data not shown). The mean age of the physicians was 43.9 years; they had been in practice a mean of 16.5 years (Table 1). Most physicians were male (72.2%) and white (78.6%). When the data were weighted to estimate the percentage of patients treated by physicians with these characteristics, the physicians were slightly older, had more years of practice, were more likely to be male, and were less likely to be white, compared with the unweighted sample.

On average, physicians said that 23.9% of their patients had a history of injection drug use (Table 1). Ninety-four percent of the physicians reported caring for at least some patients with a history of injection drug use, and 71.3% said that 10% or more of their patients had a history of injection drug use (data not shown). Forty percent of the physicians were infectious disease specialists, and 69.0% cared for 50 or more HIV-infected patients (Table 1). The mean ± SD HIV treatment knowledge score was 8.2 ± 2.0. The mean ± SD stress score was 8.4 ± 2.7. The mean practice satisfaction score was 3.8 (very good), indicating that these physicians were generally satisfied with their practice situation.

The mean ± SD attitude score toward HIV-infected IDUs was 3.75 ± 1.79 (Table 1). About 14% of the physicians agreed or strongly agreed that treating HIV-infected IDUs seems futile, and 8.6% agreed or strongly agreed that, if given a choice, they would rather not treat HIV-infected IDUs. About 17% of the physicians agreed or strongly agreed with 1 or both statements. Those physicians cared for 23.2% of all HIV-infected patients nationally.

Several physician characteristics were significantly associated with negative attitudes toward HIV-infected IDUs (Table 2), including being in practice longer (r = 0.11, P = .03), being male (mean negative attitude score, 3.89 for men vs 3.40 for women; P = .01), having a lower percentage of HIV-infected patients with a history of injection drug use (r = −0.22, P<.001), and having a higher percentage of homosexual patients who are infected with HIV (r = 0.19, P<.001). Physicians who spent between 20 and 40 hours per week on patient care had the most positive attitudes (P = .01). Not significantly associated with negative attitudes were physician age (r=0.06, P=.22) and percentage of HIV-infected patients who have sexual contact with high-risk persons (r=−0.08, P=.10).

There were significant differences in attitudes among physicians with different types of expertise (P = .02) (Table 2). General medicine physicians who were non–HIV experts had the most negative attitudes (mean negative attitude score, 4.11), and infectious disease specialists had the most positive attitudes (mean positive attitude score, 3.47). Greater physician knowledge about HIV treatment was associated with less negative attitudes (r = −0.18, P<.001), as was attending HIV-related professional meetings (P = .04). Seeing more patients per week was associated with worse attitudes (mean negative attitude score, 4.21 for physicians who see ≥100 patients per week vs 3.22 for those who see <20 patients per week; P = .007). Having a higher level of stress was related to more negative attitudes (r = 0.12, P = .02), but satisfaction with practice was not (r=−0.03, P=.58).

Regression analysis indicated that physicians with fewer patients with a history of injection drug use, lower HIV treatment knowledge scores, and more patients seen per week had more negative attitudes (Table 3). Physicians who spent between 20 and 40 hours per week on patient care had the least negative attitudes.

There were no significant overall differences among all HIV-infected patients (IDUs and non-IDUs) cared for by physicians with negative vs positive attitudes toward HIV-infected IDUs on the aspects of health care quality measured (Table 4). After adjustment for patient and physician characteristics, IDUs had more unmet needs (P = .004) but better perceived access to care (P = .02) than non-IDUs.

For exposure to HAART, there was a significant interaction between having a physician with a negative attitude and being an IDU (Table 4). Injection drug users treated by physicians with negative attitudes were significantly less likely to be exposed to HAART than other patients. After adjusting for patient and physician covariates, the predicted percentages of patients exposed to HAART among those treated by physicians with negative attitudes were 13.5% for IDUs and 36.1% for non-IDUs; the predicted percentages of patients exposed to HAART among those treated by physicians with positive attitudes were 32.3% for IDUs and 34.4% for non-IDUs. Therefore, IDUs treated by physicians with positive attitudes had more than twice the likelihood of exposure to HAART as IDUs treated by physicians with negative attitudes. In contrast, the percentage of non-IDUs who received HAART was similar between those cared for by physicians with negative and positive attitudes (36.1% and 34.4%, respectively). In variance components analysis, physician characteristics accounted for less than 3.0% of the explained variance in satisfaction with care, 7.3% of the variance in perceived access to care and exposure to HAART rates, 8.2% of the variance in the problem score, and 12.4% of the variance in unmet needs score.

About 17.4% of the physicians surveyed, who treated 23.2% of HIV-infected patients nationally, had negative attitudes toward HIV-infected patients who used injection drugs. After controlling for other factors,^4,5,10 HIV experience and specialty training were not related to attitudes, but physicians with more experience treating IDUs and less time pressure had more positive attitudes.

Physician attitudes toward HIV-infected IDUs were not significantly related to the average HIV patient’s quality of care. Injection drug users reported more problems with care and more unmet needs than other patients, but these effects were comparable for physicians with and without negative attitudes. In multivariate analyses, IDUs had more unmet needs, but there were no other negative effects associated with being an IDU, and IDUs reported better access to care than non-IDUs.

After controlling for numerous other covariates, IDUs who were treated by physicians with negative attitudes were only half as likely to have been exposed to HAART by December 1996 as non-IDUs treated by physicians with positive attitudes. Physicians who have negative attitudes toward IDUs are perhaps more likely to believe that IDUs are less adherent to treatment regimens and to withhold antiretroviral treatment.⁷ However, whether drug users are less adherent is unclear^21,22; it is difficult to assess which patients are likely to be adherent^8,21,23,24; and physician judgment can be influenced by characteristics that are not related to adherence.²⁵ We have only cross-sectional data on experience and attitudes, and some of the associations observed may be due to selection effects. That is, physicians who are more comfortable providing care for IDUs might choose to do more of it and develop more knowledge and experience. Also, patients with a history of injection drug use may be more likely to select physicians with more positive attitudes. The associations observed may also be due to an unmeasured factor, such as the physicians’ being depressed or having other life stresses. This study was conducted during an early period in the HIV epidemic, and attitudes may have changed since these data were collected. We can think of no reason why the basic associations noted would change, however.

Although many physicians have negative attitudes toward HIV-infected IDUs, it is encouraging that those attitudes did not affect the care they provide to HIV patients overall and did not affect most of the care provided to IDUs. A second encouraging finding is that the predictors of attitudes toward IDUs are potentially modifiable. Experience caring for particular subsets of patients, such as IDUs, is critical for ensuring that care is of the highest possible quality. Also, patient load is associated with negative attitudes, which are in turn related to selected aspects of care for a subset of patients. Therefore, improving clinicians’ HIV treatment knowledge or allowing them enough time to deal with complex patients may help them develop more positive attitudes toward treating IDUs, which may lead to more effective care strategies for a group of patients who often have difficulty getting the best possible care. Our findings that attitudes toward particular patients are associated with quality of care and that experience, knowledge, time pressures, and stress are related to attitudes may well apply to other groups of difficult-to-treat patients with a broad range of conditions.

Correspondence: Paul D. Cleary, PhD, Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave, Boston, MA 02115 (cleary@hcp.med.harvard.edu).

Accepted for Publication: November 19, 2004.

Financial Disclosure: None.

Funding/Support: This study was supported by grants from The Robert Wood Johnson Foundation (026449), Princeton, NJ; Agency for Healthcare Research and Quality (HS 10-227), Rockville, Md; and Aetna Inc Quality Forum, Hartford, Conn.