Injection drug use continues to account for a significant proportion of the HIV burden in the United States and Canada [1
]. In addition to having increased risk of HIV acquisition and transmission, injection drug users (IDUs) tend to have more limited engagement in HIV care and treatment. Data from some clinical and community-based observational cohorts have indicated that active IDUs have inferior virologic [3
] and immunologic [5
] responses to antiretroviral therapy (ART) compared with former IDUs and non-drug using patients.
Despite this, we and others have observed survival benefits of ART among IDUs with advanced HIV/AIDS that approaches that observed in other risk groups [7
]. Studies in other contexts with universal access to ART have demonstrated similar mortality and rates of antiretroviral resistance among IDUs and non-IDUs [9
], suggesting that the availability of interventions to support adherence and address co-morbid substance abuse may effectively eliminate the ART-related disparities observed in other settings [11
Mechanisms proposed to mediate the association between injecting drugs and poor HIV treatment outcomes include delayed diagnosis and treatment initiation [12
], poor retention in outpatient care [15
], and inadequate medication adherence [5
]. Additional individual-level correlates of delayed entrance or disengagement from care are older age, black race, and distrust of the medical care system [18
]. While individual-level and behavioural variables have been the focus of most previous research, some authors have proposed expanding the paradigm used for studying sub-optimal HIV care to encompass social and structural factors, such as stigmatization, drug policies and health care delivery considerations [20
]. In a nationally representative study, drug-injecting patients whose HIV providers had negative attitudes toward IDUs were significantly less likely to receive ART [22
]. IDUs who have favourable perceptions of the relationship with their HIV providers tend to have more appropriate ART utilization and better virologic response [23
]. Beyond individual providers, contextual factors that have been associated with improved treatment outcomes include aggregate HIV care experience and clinic site specialization [25
Whether underutilization of ART by IDUs is driven to a greater extent by providers' decisions to not recommend treatment or by patients' refusal is unclear, although both scenarios are known to occur [28
]. The practice of withholding ART solely on the basis of injection drug use runs counter to clinical guidelines issued by the World Health Organization, which explicitly state that drug injecting should not disqualify patients from ART eligibility, and that addiction treatment should not be required before ART initiation [29
]. We investigated this issue by evaluating North American HIV providers' willingness to initiate ART in the context of active injection drug use by their patients. Toward the goal of informing future provider-level interventions to reduce health disparities for HIV-infected drug users, we sought to identify characteristics that distinguish providers who are likely to defer for ART to medically eligible, active IDUs from those who would be likely to prescribe ART despite on-going injecting.
Between October 2009 and May 2010, we recruited health care practitioners who self-identified as regular providers of HIV care to complete an anonymous, 120-item, Internet-based survey. Because one aim of this study was to investigate barriers to optimal HIV care for a cohort of injection drug users in Baltimore, Maryland, USA, we specifically targeted HIV providers in the Baltimore metropolitan area who were identified through local provider databases and clinic staff directories. These Baltimore-based HIV providers were individually invited to participate with up to two mailed letters and three reminders via email, which contained an Internet address and password needed to access the survey.
The sampling frame was then broadened to include providers throughout North America, who were recruited using three complementary approaches: 1) a database of antiretroviral medication prescribers in the state of Maryland and the Washington D.C. metropolitan area identified through the American Medical Association Masterfile; 2) venue-based recruitment at national and international HIV conferences; and 3) an email list of subscribers to an electronic HIV clinical care resource. Providers in Maryland and the Washington D.C. area were sent mailed invitations with instructions for accessing the web-based survey as email addresses were not available.
Venue-based sampling occurred at: annual meetings of the Infectious Diseases Society of America in Philadelphia, PA; the International Antiviral Society-USA in New York, NY; and a clinical HIV care conference hosted by the Johns Hopkins HIV Service in Baltimore. An advertisement inviting eligible providers to complete the survey was circulated to subscribers of an electronic HIV clinical care guide, who could access the survey through a link contained in an email message. Surveys were created and administered using SurveyMonkey (Portland, Oregon). The Johns Hopkins Bloomberg School of Public Health Institutional Review Board reviewed the study protocol and granted it exempt status.
Of 368 providers practicing in Baltimore who were individually invited to participate, 157 (43%) completed surveys. Basic demographic data were available for 168 (80%) of the 211 Baltimore-based non-responders. Compared with participating providers, non-responders were more likely to be male (66% vs. 44%, p < 0.001) and have an MD or DO degree (92% vs. 78%, p = 0.002) and were less likely to be trained in infectious diseases (8% vs. 54%, p < 0.001). Non-Baltimore providers contacted by mail had a response rate of 8.5%. As convenience sampling was utilized to recruit survey participants at professional conferences and via the online list-serve, we could not ascertain response rates relevant to the broader recruitment strategies.
Compared with Baltimore-based providers, those recruited via conferences and the HIV list-serve were more likely to be male (57% vs. 40%, p < 0.001), to have specialty training in infectious diseases (76% vs. 56%, p < 0.001), to care for 50 or more HIV-infected participants (58% vs. 44%, p < 0.001), and were less likely to work at a clinic serving a patient population with a high lifetime prevalence (> 25%) of injection drug use (6% vs. 26%, p < 0.001)
This cross-sectional survey assessed demographics and professional characteristics, such as degree type, specialization, years of clinical experience and volume of patient care activities. A series of questions addressed characteristics of the provider's primary practice site, including its geographic location (urban vs. suburban vs. rural), whether it provides mostly primary care or specialty care, and whether it is affiliated with an academic institution. Providers were asked to estimate the proportion of patients seen in their practices who were HIV-infected, had used injection drugs, were underrepresented minorities, and lacked health insurance. Other questions addressed the availability of certain services at the primary practice site, including HIV testing, social work services, substance abuse treatment and transportation assistance.
Provider knowledge, attitudes and beliefs related to injection drug use were assessed based on the degree to which the respondent agreed or disagreed (using a 5-point Likert scale) with a series of statements designed to assess negative or prejudicial attitudes toward IDUs. The statements were drafted by a committee of HIV clinicians and researchers and pilot tested on 15 providers who treat patients in an urban, university-affiliated HIV clinic. Revision and refinement resulted in nine items incorporated into the final survey. Examples include: "I feel uncomfortable talking to my patients about their injection drug use practices"; and "HIV-infected IDUs have themselves to blame for their illness". The full text of all questionnaire items is available in the Additional file 1
Deferral of ART was assessed by asking respondents to self-rate the likelihood of prescribing ART to a series of hypothetical patients. The range of patient scenarios included a patient with no drug use history, a former IDU who had been abstinent for three months, a current IDU who injects several times per month, and a daily injector. Within each of these four categories of drug use, respondents were further asked to rate the likelihood of deferring ART if the patient in question had a CD4+ cell count of 200 cells/mm3, 350 cells/mm3 or 500 cells/mm3.
Descriptive statistics related to demographic, provider and clinic characteristics were analyzed using Chi-squared and Fisher's exact tests for categorical data. After sensitivity analyses of potential categorizations, Likert scale variables related to provider attitudes and beliefs were categorized into "agree" (strongly agree and agree) and "do not agree" (neutral, disagree or strongly disagree). Attitude and belief items that showed significant bivariate association with deferral of ART (p value for Chi-squared test less than 0.10) were included in subsequent analyses.
The main, binary outcome of interest in this study was a provider report that one would probably defer ART for patient with a CD4+ cell count equal to 200 cell/mm3
if he or she reported any active (i.e., daily or occasional) injection drug use. This approach was chosen because there is near-universal consensus that ART initiation is indicated for patients with this degree of immunosuppression, regardless of symptoms, in both resource-rich and resource-limited settings [30
]. A decision to defer ART in this context, therefore, likely indicates that a provider believes active injecting precludes effective ART. Logistic regression models were constructed using forward and backward stepwise approaches to identify variables that were independently associated with deferring ART in the setting of any active IDU. Because revised guidelines regarding when to start antiretroviral therapy were published by the US Department of Health and Human Services (DHHS) during the study period [32
], we included a covariate indicating whether respondents completed the survey prior to or after 1 December 2009, when the guidelines were released.
In secondary analyses, we repeated these model-building steps for the separate outcomes of ART deferral for drug injecting patients with CD4+ cell counts of 350 cells/mm3 and 500 cells/mm3. We next compared the proportion of providers who would defer ART within each of the three strata of CD4+ cell count and the four strata of drug-injecting status, allowing examination of prescribing behaviour across 12 scenarios characterized by varying drug use and disease severity. Finally, we used these data to characterize providers whose decision to defer differed depending on intensity of injecting (i.e., daily vs. occasionally). Statistical analyses were performed using STATA Version 11 (StataCorp, College Station, TX).