|Home | About | Journals | Submit | Contact Us | Français|
Substance use is a prevalent issue in primary care with wide-reaching implications, particularly for the care of HIV-infected patients. This analysis identified patient and provider characteristics associated with high comfort discussing substance use in HIV primary care clinics using multivariable logistic regression.
A total of 413 patients and 44 providers completed surveys on their comfort discussing substance use. Additional dependent variables from surveys included demographics, drug and alcohol use, self-efficacy, and activation for patients. Provider-level data included demographics, training, practice descriptors, and stress levels.
The majority of patients (76%) and providers (73%) reported high comfort. In multivariable analysis, patients with current problematic alcohol use or current drug use were half as likely to report high comfort compared to their non-substance-using peers. Higher patient self-efficacy and high levels of patient activation were independently associated with increased odds of higher patient comfort. While provider-level characteristics were not associated with provider comfort, the types of patients a provider saw were. Namely, the proportion of patients on antiretroviral therapy was inversely associated with the odds of high provider comfort, whereas the proportion of patients with high patient activation was positively associated.
Patients likely to benefit from a discussion of substance use, those with current use, are the least likely to report comfort discussing that use. Interventions that increase patient activation or self-efficacy may also increase their comfort. This research guides future interventions to increase the prevalence of discussions on substance use.
Previous work has shown 50%–75% of HIV-infected individuals used illicit drugs or had heavy alcohol use in the prior year.1,2 Illicit drug or alcohol use decreases antiretroviral (ARV) adherence, increases HIV disease progression, and increases HIV transmission behaviors.3–6 HIV-infected patients with current drug use experience decreased health-related quality of life.2
Despite its prevalence and impact, fewer than half of HIV-infected patients report discussing substance use with their provider.2,7 Provider reluctance to screen for substance use is unfortunate as patient-provider discussions are associated with decreased use and increased receipt of treatment for substance use disorders.2,8–12 Treatment improves the health of HIV-infected patients through improved ARV adherence, regular primary care visits, decreased hospital utilization, and improvement of common medical illnesses impacted by substance use (eg, liver disease).13–15
In previous studies, providers reported hesitation over correctly diagnosing substance abuse or providing resources to substance users.16,17 Providers also fear alienating their patients despite research to the contrary. Miller and colleagues found that more than 90% of patients support their provider screening for alcohol use.18 Recent studies identified providers with positive perceptions of their abilities as more likely to screen their patients with alcohol disorders.19 The quality of patient-provider communication is different between HIV-infected patients with problematic alcohol use than those with illicit drug use.19 Thus, provider perception of their own comfort and ability to discuss alcohol or drug abuse with their patients may be an important predictor of patients getting the treatment they need.
In ambulatory care settings, discussions of substance use are important yet often missed opportunities to improve the quality of health care. There is a paucity of literature on comfort when discussing substance use from the perspective of either the patient or provider. Intuitively, high comfort discussing substance use would be associated with conversations regarding use, yet there is no previous research to support this. The purpose of this study was to assess patient and provider comfort discussing substance use in a primary care setting. We chose to conduct the study in HIV primary care practices where the consequences of substance abuse and the potential impact of patient-provider substance use discussions are great.
The Enhancing Communication and HIV Outcomes (ECHO) study was a cross-sectional study that investigated the impact of patient-provider communication on HIV-related clinical outcomes. Study methods have been published previously.20 Briefly, patients and providers were recruited from four HIV primary care clinics in Baltimore, Detroit, New York City, and Portland, OR, from 2007–2008. Eligible physician, nurse practitioner, or physician assistant providers delivered primary care to HIV-infected patients. Providers completed baseline assessments and a brief after-visit questionnaire following each patient encounter. Participating providers completed informed consent and were provided with $250 at the end of the study.
Patients were eligible if they were HIV-infected, 19 years of age or older, had at least one primary care visit in the last year, and spoke English. Trained study personnel enrolled patients from the waiting room of participating clinics and conducted in-person interviews following the office visit. Participating patients completed informed consent and were provided with $50 at the end of the study. Research assistants abstracted ARVs, CD4 count, and HIV-1 RNA viral load from patient medical records. Institutional Review Boards at each facility approved the study.
Participants responded to one (patients) or two (provider) questions assessing their comfort discussing substance use. Patients responded to “How comfortable do you feel telling your HIV provider about using drugs or alcohol?” with responses ranging from not at all comfortable to very comfortable (scale 1–4). Providers responded to two similarly worded questions: “Using the scale below, indicate your level of comfort discussing illicit drug use (alcohol use for second question) with patients” with responses again ranging from not at all comfortable to very comfortable (scale 1–5). Patient and provider comfort variables were dichotomized into a “high comfort” group consisting of individuals reporting the highest level of comfort for both alcohol and drugs and a “less than high comfort” group consisting of all other responses based on the distribution of responses.
Demographic characteristics included self-reported patient gender, race/ethnicity (white, Black, Hispanic, Other), age (years), employment status (yes/no), and educational attainment (≥ high school degree/GED). Other patient-related covariates included depression score on the 10-item Center for Epidemiologic Studies scale (CESD) (tertiles),21 patient activation (tertiles),22, and communication self-efficacy from the four-item communication subscale of the HIV Self-Efficacy Questionnaire.23 High communication self-efficacy was defined as a mean score of 10 versus <10, based on score distribution.
Current problematic alcohol use was defined as at least one episode of drinking to intoxication in the past 30 days using items from the Addiction Severity Index-lite (ASI).24,25 Current drug use was defined as any use of heroin, cocaine, amphetamines, marijuana, or methadone without a prescription in the past 30 days using the ASI-lite.
Patient-reported duration of the patient-provider relationship was categorized as less than or greater than 5 years in duration. The length of the patient-provider encounter in minutes was obtained from audio-recordings. Each provider responded to the question “How would you describe your schedule today?” with responses ranging from “much busier than usual” to “slow” after each patient visit. The final variable was dichotomized into very busy versus less than very busy.
Provider characteristics included self-reported gender, age, race/ethnicity (white/non-white), type of training (physician/midlevel provider), time spent caring for HIV-infected patients (greater or less than 20 hours per week), and any history of communication training (yes/no). Provider self-reported stress was assessed over the past month using a four-item instrument, dichotomized as highest stress tertile versus less than highest stress tertiles.26 Physician busyness was assessed using the question “How would you describe your schedule today?” with responses ranging from “much busier than usual” to “slow” after each patient visit. The final variable was dichotomized into very busy versus less than very busy. The role of provider confidence on the impact of discussions on substance was assessed by asking, “How confident are you that discussing substance use/abuse with your patients will make a difference in their future behavior?” with responses dichotomized as very confident versus less than very confident.
For each provider we estimated the characteristics of that practice by calculating the proportion of the provider’s patients in the study for each characteristic (eg, the proportion of each provider’s patients that were female, non-white, etc.) We also calculated the proportion of office visits greater than or equal to the median length for each provider. These variables were scaled for ease of interpretation in final models to reflect the odds for a 10% change in proportion.
We used descriptive statistics to report patient and provider characteristics by comfort level (high/less than high). Bivariate associations between independent variables and comfort were estimated using χ2 analysis for categorical variables and two-sided t tests for continuous. The provider bivariate analysis utilized Fisher’s exact test for cases where the expected cell count was less than five.
Multivariable model development started with a base model investigating the relationship of each independent variable to patient comfort while controlling for site and provider clustering alone (base model, column 1, Table 2). Then variables identified from bivariate analysis with a statistical significance of P≤.20 and those identified based on a priori hypotheses were added. This created a full model from which likelihood ratio and Hosmer-Lemeshow goodness of fit tests were utilized to assess the relative importance of each variable in the model to create a final model.27 Conceptually important variables were reintroduced into the final model regardless of likelihood ratio testing. In all multivariable models, site was included as a fixed effect variable to adjust for potential geographic clinical differences. Patient comfort models were adjusted for clustering by provider using generalized estimation equations. All analyses were conducted on STATA (version 11.0, October 2009, StataCorp, College Station, TX).
From 2006–2007, study personnel contacted 47 of 55 eligible providers with 45 agreeing to participate (meeting recruitment goals) and two refusing. Study personnel identified 617 eligible patients during clinic visits occurring. Provider discretion excluded 18 possible participants due to feeling too rushed (n=12), the patient was too sick (n=5), or the visit was only for return of lab values (n=1). Of approached subjects (599), 434 consented to participate and completed the study (72%). The most common reason for refusal to participate was lack of time to complete the interview (106 individuals).
For the current analysis, 21 additional patients were excluded: 17 for missing or incomplete audio-recordings, three for missing all substance use data, and one for missing demographic data, leaving a final analytic sample of 413 encounters. Additionally, one provider was excluded from the provider analysis as they only had a single patient participate in the study (final provider n=44).
The patient study sample was mostly male (65%), African American (59%), with a mean age of 45 years (standard deviation [SD]= 9.3, range 20–77 years). (Table 1). The majority were on antiretroviral therapy (77%) with their last CD4 count > 200 (80%). Current illicit drug use was reported by 27% of patients (predominantly marijuana use), with 9.5% reporting current problematic alcohol use. One third did not complete high school or obtain an equivalent degree.
Providers were evenly split by gender (57% female) with a majority of white race/ethnicity (68%) and a mean age of 44 years (SD=8.6) (Table 3). Three quarters of the providers were physicians.
The majority of patients (76%) and providers (73%) reported high comfort discussing substance use.
In unadjusted analysis (Table 1), only 56% of current problematic drinkers reported high comfort compared to 78% of non-problematic or non-drinkers (X2, P=.003). Two thirds of patients with current drug use reported high comfort compared to 79% of patients without current drug use (X2, P=.005). Among patients reporting less than high comfort, 38% had used drugs in the past month, and 17% had consumed alcohol to intoxication in the past month.
Patients with high communication self-efficacy or high patient activation were more likely to report high comfort. Being on ARVs was strongly associated with high comfort, with 80% reporting high comfort compared to 65% of patients not on ARVs (X2, P=.003). Patients with greater depression symptoms were less likely to report high comfort (X2, P=.010). No provider characteristics were associated with high patient comfort discussing substance use.
In multivariable analysis (Table 2), patients reporting current drug or problematic alcohol use had lower odds of endorsing high comfort (aOR=0.46, 95% CI=0.28–0.77 for drug use, aOR=0.32, CI=0.15–0.66, for alcohol use) after adjusting for demographics, education, self-efficacy, patient activation, depression, ARV status, site, and clustering by provider. A patient on ARVs had twice the odds of reporting high comfort compared to a patient not on ARVs (aOR=2.15, 95% CI=1.18–3.40). Patients who did not graduate high school were half as likely to report high comfort compared to graduates (aOR=0.45, 95% CI=0.22–0.93). Female patients were twice as likely to report high comfort (aOR=2.05, 95% CI=1.13–3.72).
High patient activation was associated with increased odds of reporting high comfort (aOR=2.66, 95% CI=1.24–5.69). Also, each 10-minute increase in the length of the patient’s office visit was associated with nearly a quarter decrease in the odds of high comfort (aOR=0.78, 95% CI=0.64–0.99).
A sensitivity analysis limiting the multivariable analysis to only those with current drug or alcohol use yielded similar results, though there was insufficient power to detect a difference in self-efficacy or patient activation due to the smaller sample size.
In unadjusted analysis, providers with high comfort were younger (40 versus 46 years, 2-sided t test, P=.025). All midlevel providers, compared to 65% of physicians, reported high comfort (Fischer’s exact test P=.041). Providers with a greater proportion of patients with a high school degree (75% versus 62%, two-sided t test P=.015) and those with fewer patients on ARV (74% versus 85%, tow-sided t test, P=.059) more frequently reported high comfort discussing patient substance use. the odds of a provider reporting
In multivariable analyses (Table 4), no provider characteristics were associated with provider comfort. However, as a provider’s proportion of patients on ARV increased, the odds of a provider reporting high comfort discussing substance use decreased (aOR: 0.38, 95% CI: 0.15–0.97). As the proportion of patients in the highest patient activation tertile increased, so did the odds of a provider reporting high comfort discussing substance use (aOR 2.87, 95% CI: 1.01–8.14).
Patients and their providers must have the freedom to openly discuss substance use issues in primary care. There is scant research on factors influencing patients’ decisions to discuss drug or alcohol use with their provider and little on providers’ comfort with substance use discussions. The current study demonstrates that, though self-reported comfort is high for both patients and providers, those most likely to benefit are least likely to be comfortable discussing substance use.
Patients reporting higher levels of communication self-efficacy, antiretroviral use, and patient activation reported greater comfort, suggesting that interventions to improve communication and self-management skills might increase patient comfort discussing substance use in HIV primary care settings. Patients with current illicit drug or problematic alcohol use were independently less likely to report high comfort discussing with their provider.
Patients most likely to benefit from a discussion of their substance use were the least likely to report high comfort doing so. This is particularly important because patients who discuss substance use issues with their providers are more likely to receive treatment, yet this occurs in only a minority of primary care visits.2,7
Several potentially modifiable patient characteristics were independently associated with high patient comfort, suggesting potential targets for improving comfort. Patients with high communication self-efficacy, those receiving ARV, and those with high patient activation more frequently reported high comfort. Communication self-efficacy has been previously associated with high education level and younger age, yet interventions have shown improved self-efficacy across diverse groups primarily in participation in medical care.28,29 Patient-directed interventions to improve patient communication self-efficacy may facilitate comfort in communicating about substance use with their providers.
Patient activation is defined as the knowledge, skill, and confidence an individual has in managing his or her disease.22 Patients with higher levels of activation exhibit more optimal health behaviors and experience improved outcomes for patients with chronic illness such as diabetes and heart disease.30,31 Patients reporting higher levels of patient activation in the current study had greater than a threefold increase in odds of reporting high comfort. Likewise, the only provider characteristic independently associated with provider comfort discussing substance use issues with their patients was the proportion of their patients with high patient activation, suggesting that greater activation may facilitate substance use dialogue for both patients and providers. Interventions that improve patient activation may increase patient and provider comfort discussing substance use.
Other patient characteristics independently associated with high comfort included female gender, receipt of ARV, and higher educational level. Patients without a high school diploma or GED were half as likely to report high comfort compared to someone with a high school diploma or higher, even after accounting for current substance use, depression, and demographics. The strong association between less educational attainment and decreased odds of high comfort likely captures the greater health disparities experienced by individuals of lower socioeconomic status.32
Also, this study observed an inverse relationship between visit length and patient comfort. Longer visits may be required to address the competing medical and social needs of a more complex patient (ie, worse HIV-related health, lower health literacy, or active drug use),which take priority over addressing potentially uncomfortable discussions of substance use. Yet length of patient-provider dialogue was shorter for patients with current and former unhealthy alcohol use compared with non-drinkers in one recent study and comparable for those with illicit drug use versus non-user.20 Patients with long visits may be addressing other issues associated with discomfort talking about substance use (eg, medication adherence).
This study observed an inverse association between the proportion of patients on ARV that a provider sees and the provider’s comfort discussing substance use.
Also, this study observed a positive association between the proportion of patients with the highest level of patient activation and provider comfort. Intuitively, a provider who sees a majority of patients who are highly activated or on ARV may feel more comfortable as their patients may be more engaged in their care or providers perceive their patients to have the skill set to try and make changes in their substance use.
Provider comfort discussing substance use is a novel research question without previous research on this specific topic. Related literature comes from patient-reported scoring of provider performance in the realms of general and HIV-specific communication.33–35 From Roter and Hall’s systematic review of gender and patient-centered communication, one would hypothesize that female providers would report greater comfort discussing substance use.34 This is consistent with work from the patient perspective. In Wilson and Kaplan, patient-rated communication skills were higher for female providers.35 However, the current study did not observe any independent association between provider gender and comfort.
As the largest study to date on patient or provider comfort discussing substance use, these findings inform future research by identifying patient and provider characteristics associated with comfort. These findings should be interpreted in light of several important limitations. First, the relatively small number of providers limits power to detect independent associations between provider characteristics and provider comfort as ECHO was powered to detect differences in patient-level characteristics. Using provider-level data, however, provides a unique insight that has often been overlooked in the literature. Second, the cross-sectional study design limits causal inferences between participant characteristics and their comfort. Prospective, controlled studies would help strengthen causal inference. Finally, the 28% of nonparticipants may have differed in activation or comfort from participants but overall our participation rate is in line with previous work in similar populations.36,37
Connecting patients to drug or alcohol treatment often falls in the hands of primary care providers. This research identifies important correlates of patient and provider comfort that can inform future interventions to encourage medical providers to discuss substance use with their patients. Patients with current alcohol and drug use, in particular, reported little comfort discussing their use with their providers. Efforts to improve the patient-physician dyad in terms of comfort having substance use discussions may improve referral of patients to the treatment they need.
This research was supported by a contract from the Health Resources Service Administration and the Agency for Healthcare Research and Quality (AHRQ 290-01-0012). Dr Ray’s time was supported through a T32 predoctoral fellowship from the Agency for Healthcare Research and Quality (AHRQ 1T32 HS017582). Dr Korthuis’ time was supported by the National Institutes of Health, National Institute on Drug Abuse (K23DA019809). Dr Beach was supported by the Agency for Healthcare Research and Quality (K08 HS013903-05), Dr Saha was supported by the Department of Veterans Affairs, and both Drs Beach and Saha were supported by Robert Wood Johnson Generalist Physician Faculty Scholars Awards. The contents of the publication are solely the responsibility of the authors and do not necessarily represent the views of the funding agencies or the US government.
Preliminary data on the correlates of provider comfort were presented at the 16th Annual NRSA Trainees Research Conference in Boston, June 26, 2010.