Correlates of Patient-Centered Care Practices at U.S. Substance Use Disorder Clinics

Because patient-centered care is a dynamic process, no measure perfectly assesses it, particularly at the organizational level (19, 20). We employed two measures of patient centeredness. The first was a binary indicator of whether clinics regularly held treatment planning meetings and invited patients to participate in clinical decision-making processes. In particular, clinical supervisors were asked whether their clinics had a regular meeting of all treatment providers and whether clients were regularly invited to attend this planning meeting when their case was being discussed. The second was a composite variable that captured clinical supervisors’ perceptions of patient-centered care. We developed this measure by using ten original questions assessing the value and emphasis placed by clinical supervisors on patient-centered care practices (Cronbach’s α=0.79) [the ten questions are listed in an online supplement to this article]. Question frames and structures were adopted from two validated measures: the Person-Centered Care Assessment Tool and the Shared Decision Making Questionnaire (21, 22).

The two variables measured different aspects of clinics’ patient centeredness. The first measured a particular behavior—namely, patient engagement in the care decision-making process—regarding clinicians’ collaboration with patients in the formal treatment care process. We note that depending on treatment settings and circumstances, some clinics may use alternative venues and mechanisms to discuss treatment options and may invite regular patient input. The second composite variable was devised to capture clinical supervisors’ beliefs about and support for patient-centered care. We note that supervisors’ emphasis and promotion of shared decision making may not translate into clinicians’ daily practice of patient-centered care. Acknowledging the purpose and limitations of the two variables, we used both measures as dependent variables in this study.

NDATSS survey weights adjusted descriptive statistics to be nationally representative (18). Logistic regression analysis was used to analyze the association between explanatory variables and whether clinics invited patients into clinical decision-making processes, given the binary nature of the dependent variable. Linear regression was used for the second analysis of whether attributes correlated with the composite patient-centered care variable. To investigate field-level differences, we controlled for modality (using outpatient nonopioid treatment program clinics as a reference category) in both regression analyses. We also used one dependent variable to predict the other variable in regression analyses and vice versa, because organizational leaders’ support for patient-centered care could influence clinicians’ adoption of inclusive decision-making processes, and repeated exposure to and success with patient-centered procedures could drive senior managers’ stronger support (31, 32). Models with and without patient-centered care explanatory variables resulted in almost identical relationships between other explanatory variables and dependent variables. To preserve sample size and minimize estimation bias, missing values for the explanatory and control variables were imputed by using a stepwise regression multiple imputation method with IVEware (33). For ease of interpretation, several continuous variables (e.g., staff and patient compositions) were standardized for regression analyses.

Weighted descriptive characteristics of the overall sample of substance use disorder clinics in the United States (N=657) are presented in Table 1. About 23% of the clinics invited patients to participate in clinical decision-making discussions. Separate descriptive analyses showed that 23% of the outpatient opioid treatment programs, 18% of the outpatient nonopioid treatment programs, 26% of the inpatient clinics, and 39% of the residential clinics invited patients to participate in the clinical decision-making process. The patient-centered care composite variable had a mean±SD score of –0.08±0.91, which is close to its desired distribution characteristics (i.e., mean=0, SD=1) [see online supplement for mean scores and SDs for the 10 questions].

Roughly 66% of the clinics were outpatient nonopioid treatment programs, 21% were residential clinics, 8% were outpatient opioid treatment programs, and 4% were inpatient clinics. A majority (57%) were private nonprofit clinics and accredited by the Joint Commission or the Commission on Accreditation of Rehabilitation Facilities (53%). About 60% of the directors reported a high degree of regional competition. Roughly 7% of staff had medical training and 33% were in recovery from substance use disorder. About 50% of the patients had alcohol use disorder, 33% had opioid use disorder, and 28% had prescription opioid use disorder. More than 46% of patients were involuntary, receiving treatment services following a court order; and 40% were from racial-ethnic minority groups.

We found several significant associations between environmental and organizational factors and the two dependent variables (Table 2). Residential clinics were significantly more likely than were outpatient nonopioid treatment clinics to invite patients into care decision-making meetings (adjusted odds ratio [AOR]=3.42). Two variables related to the composition of patient populations were negatively associated with the odds of inviting patients to care decision meetings: the percentage of clients with alcohol use disorder (AOR=0.77) and the percentage of clients with opioid use disorder (AOR=0.74). When a clinic’s proportion of patients from racial-ethnic minority groups was increased by 1 SD (35%), the odds of inviting patients into meetings significantly increased (AOR=1.48), compared with clinics in which the proportion of patients from minority groups was average. The patient-centered care composite variable was positively associated with a clinic’s increased odds of inviting patients into formal care decision processes (AOR=1.27).

Several variables were significantly associated with the composite patient-centered care variable (Table 2). Directors’ reliance on professional information sources was positively correlated with the composite score (coefficient=0.19). High regional competition perceived by directors was negatively correlated with the composite score (coefficient=–0.18). Clinics that invited patients into formal clinical decision-making processes tended to score higher on the composite variable (coefficient=0.19).

In 2017, only 23% of the substance use disorder clinics invited patients into formal clinical decision-making processes. Regularly inviting patients into the decision-making process was significantly correlated with a clinic’s service modality, its patient composition, and clinical supervisors’ emphasis on patient-centered care. Clinics serving a larger proportion of patients with alcohol use disorder or opioid use disorder were less likely to invite patients to participate in the clinical decision-making process. Clinics may be more willing to close information and preference gaps between clinicians and patients with certain groups of patients, or the gaps may be smaller with some groups than others (e.g., there is no medication to treat stimulant use disorder and hence nothing for clinician and patient to potentially disagree about). Prescribing a standard treatment program may be the norm for the patient groups with alcohol or opioid use disorders (i.e., total abstinence or a support group for those with alcohol use disorder and medication-assisted treatment for those with opioid use disorder).

Direct engagement efforts may also be influenced by the demographic characteristics of various patient groups. For a long time, opioid use disorder has been highly stigmatized, associated with low-income individuals, persons from racial-ethnic minority groups, and young urban residents, all of whom have been historically framed as economically dependent and undeserving of treatment (34, 35). However, middle-class, white, and middle- and retirement-aged suburbanites constitute a significant proportion of the recent surge of patients with prescription opioid use disorder, and they are often characterized as more deserving or more worthy of treatment (36). We were thus surprised to find a positive association between the proportion of patients from racial-ethnic minority groups and clinics’ tendency to directly engage with patients. These patterns suggest a more complex dynamic. Further qualitative investigation is needed to capture patients’ and clinicians’ perspectives and behaviors.

The second regression analysis, using the patient-centered care composite variable, revealed a positive relationship between patient-focused care and managers’ reliance on professional information sources. As the senior staff member of a clinic, an administrative director’s engagement with and exposure to professional association members, conferences, journals, and seminars might lead to the clinic’s adoption of a normative practice mode (37). However, directors’ reliance on these professional information sources was not significantly associated with patients being regularly invited to participate in the clinical decision-making process. Thus, normative pressure may yield some symbolic results (e.g., emphasizing patient-clinician collaboration during staff training sessions) without producing substantive organizational change (e.g., reducing average caseloads for clinicians so that they can spend more time with patients or changing clinic hours to improve patients’ access to care decision-making processes).

We found that when administrative directors perceived high regional competition, clinics tended to have lower scores on the patient-centered care composite variable. Managers who sense greater competition may tend to deemphasize noneconomic activities (i.e., patient-centered shared decision-making processes) to lower operational costs and maintain a competitive edge. Although clinics serving less price-elastic and perhaps more quality-elastic patient subgroups may strive to accommodate patients’ concerns and preferences, when faced with stark competition—given that a majority of patients are either uninsured or Medicaid insured—clinics may prioritize survival over patient-centered care (23, 38).

Finally, the two dependent variables (i.e., whether clinics regularly invited patients to participate in clinical decision-making processes and the patient-centered care composite variable) were positively associated with each other after the analysis controlled for various observed environmental and organizational factors. The causality and direction of relationships could not be tested with cross-sectional data. These correlations may reflect the influence of common unobserved preferences and unit characteristics that influence both variables. Our results confirm that the variables were correlated with each other and captured different aspects of patient-centered care practice, given the variables’ relationships with different explanatory variable sets.

Our study had several limitations. First, the two dependent variables did not fully measure a treatment clinic’s patient centeredness. Second, although the validity and reliability of the nationally representative data have been demonstrated, the survey did not capture important patient-level information and variances (e.g., how often and how long individual patients engaged with clinicians) that could show how patient-centered efforts shaped individual patients’ experiences. Third, the possibility of response bias looms large with the questions related to clinics’ emphasis on and practice of patient-centered care. Despite the survey’s efforts to minimize social desirability bias, supervisors may have provided normative answers to these questions. Fourth, our study was based on organizational leaders’ reports on patient-centered care practices, not on the direct responses of front-line clinicians or patients. Understanding the perspectives and experiences of the main participants in patient-centered care is an important topic for future study. Fifth, with cross-sectional data, this study could reveal only associations, not longitudinal or causal relationships. Finally, our findings are not necessarily generalizable to clinics outside the United States, where a different set of environmental and organizational factors may influence organizational practices.