Summarized in Figure , we identified 327,219 publications with any mention of RCT in the title or abstract from 1975 to April 2010. Of these, 1,475 (0.5%) referred to patient-centered and/or provider-patient relationship. From review of these abstracts, we identified 75 articles (5%) for full article review and identified 13 (0.9%) meeting our criteria for analysis in this review. To maximize our results, we also searched all 33 Cochrane Systematic Reviews, 25 of which did not address patient-centered material by review of their titles and abstracts; of the remaining 8 reviews, we found no articles not already identified. We did not exclude studies rated as negative.
Selection of publications for review.
The summary table (online) summarizes the key features of the study, which we now synthesize and integrate. Nearly all studies recorded some funding, and they occurred in health centers and HMOs, private settings, and university settings. Most studies had from 20–60 providers; recruitment rates were presented in the majority of studies and varied from 7% to 100%, most falling at the extremes of this range. Providers usually were physicians, mostly primary care, but with some physician assistants, nurse practitioners, and others. Subjects were characteristically general medical patients, although many had psychosocial problems; e.g., pain management, alcohol and tobacco cessation, and medically unexplained symptoms. Recruitment rates were generally >50%. All but one study included training, the majority of which took 10 or fewer hours, typically over one to two workshop sessions. Outcomes of training were measured in less than one-half of the studies, and most showed a positive impact of training. Most interventions involved only one visit and most controls were usual care.
There was a very wide range of behaviorally defined interventions, but common features were noted. Toward generally expressed goals of achieving improved communication and provider-patient relationships, trust, and positive regard, some well-defined behavioral features were using open-ended skills, eliciting and responding to emotion, expressing support and willingness to help, exploring patient understanding and explanation of their problem, asking what the patient would like to have happen, motivating and encouraging a positive approach, giving specifics of recommended behavioral change, linking treatment to the patient’s needs and level of understanding, advising but acknowledging the patient’s choice, and accepting the patient’s choice.
All studies used these individual skills and also aggregated them in a multidimensional approach, grouping related skills together for a given purpose of the intervention (e.g., tobacco cessation); some further specified sequences of skills. In turn, to better achieve the aims of the intervention, many identified multiple such skill sets, more complex interventions identifying them as steps where they then indicated how to sequence and prioritize the steps at one visit and over multiple visits. Such steps were not skills per se and, instead, identified general goals within an overall patient-centered intervention. For example, the first step might focus on an individual patient-centered goal of a multidimensional intervention (e.g., the patient’s agenda), while another step focused on another goal (e.g., the patient’s emotion), and another focused on a third goal (e.g., stopping alcohol use). The more complex interventions outlined the amount of time for each step and some identified high or low priority steps. Parts or all of this process of using a step-wise model to learn a complex skill were present in many of the studies.
In assessing the methods and quality of the studies, most involved randomization of providers/practices (rather than patients), nearly half with details of the randomization method provided, and most recorded blinding of outcome assessors; blinding of others was rarely mentioned. Unit of analysis error usually was not reported, intention to treat evaluations occurred in about half, and power calculations for patient outcomes were provided in the majority of studies.
With fewer studies than expected, evaluation of a relationship between patient-centeredness and outcomes was difficult. There were six studies with any positive patient outcome and seven were negative. In applying our post-hoc criteria, we found that only four positive studies and one negative study were sufficiently rigorous to inform a relationship of patient-centered practices to health outcomes. In the eight rejected studies, nonrepresentative study samples, stemming from low recruitment rates, were problematic in positive studies. Additionally, negative studies were plagued by low power or lack of evidence that the intervention could have been effectively deployed. In the last column of of the summary table (online), we note our summary of the post-hoc outcome evaluations. The patient outcomes studied varied considerably: pain reduction; evaluation of the provider’s patient-centered characteristics; satisfaction, adherence, and confidence in care; alcohol and cigarette reduction; antibiotic use; reduction in mental health problems.