This was a group-randomized trial launched in the fall of 2007 in the South Texas region of the United States. The study design of this trial and details about the intervention have been previously reported [9]. Briefly, the study was conducted in small, autonomous primary care clinics or ‘practices’ in South Texas. These urban, suburban, and rural practices, each with one to three clinicians, serve a population of primary care patients diverse in demographic characteristics, insurance coverage, and healthcare needs. Subjects for this study were the clinicians and staff at participating practices.
Recruitment
Due to travel distance and costs, recruitment was limited to the San Antonio Metropolitan Statistical Area and surrounding counties within a one-hour drive of the medical center. Early recruits included 10 active members of a primary care Practice-Based Research Network (PBRN), all of whom agreed to participate. PBRN enrollees were asked to recommend colleagues whom they thought might be interested in study participation who also referred colleagues (n = 25). These physicians were contacted directly by phone and in-person recruiting visits were scheduled at their offices resulting in 22 participants. In addition, 145 recruitment letters were sent to primary care physicians within the region identified from professional society membership guides. From these letters, 15 practices responded; of those, eight agreed to participate in the study, resulting in a total of 40 practices. The sample size of 40 practices was determined from measurement of the primary outcome, the CCM score, across 20 clinics resulting in a power of 0.94 to detect a change in the CCM score of at least 1.5 in response to the intervention as described in the original published protocal [9].
Randomization
A stepped-wedge study design was used with block randomization of practices in groups of 10 to either an ‘initial intervention’ or ‘delayed intervention’ arm of the study [10]. The random allocation sequence was generated by a member of the study team (RP) after each group of 10 clinics was recruited. A stepped-wedge design is a type of crossover design in which different clusters cross over (switch treatments) at different time points. For this study, 20 practices, the ‘initial’ intervention practices, were randomized to receive the PF intervention for one year,’ while the remaining 20 practices served as controls. Following the completion of the one-year intervention and withdrawal of facilitators from the initial intervention practices, the ‘control’ practices were crossed-over to receive a one-year ‘delayed intervention’ of PF. Randomization was done before the orientation visit where consent for study participation occurred because status of randomization was needed to inform length of and type of data collected during the initial visit. For example, interviews with clinicians and staff were conducted during and after the team meeting introducing the study in practices randomized to the initial intervention arm but not the delayed intervention.
Intervention
Practice Facilitators held a minimum of six one-hour team meetings within each practice over a 12-month period of time. As is common with many previous PF efforts, baseline chart audit and feedback as well as interactive consensus building and goal setting were incorporated into the intervention [1]. For the audit and feedback component, glycosylated hemoglobin (HbA1c), blood pressures, and lipid levels were audited from the medical records of a random sample of 30 patients seen in the prior 12 months with a diagnosis of type 2 diabetes. Additionally, 60 consecutive adults presenting for care in each practice completed a satisfaction survey. The PF intervention in each practice began with a review of the chart audit results, as well as results from the patient survey.
Each Practice Facilitator was trained in the use of multiple ‘tools’ within a ‘toolbox’ to assist practices with improving their chronic illness care. These tools included: group/shared medical appointments; a diabetes registry; point-of-care HbA1c testing; resources/approaches to patient education/activation; and planned diabetes visits with clinical reminders and decision support for providers and staff. After presenting results from the chart audit and patient survey, along with the five strategies in the toolkit, the facilitators worked with clinicians and staff to identify alternative approaches or to adapt strategies from the toolbox to improve the delivery of diabetes care through a process of interactive consensus building and goal setting in each practice.
Data collection and measurements
Data were collected at three points in time, approximately one year apart in each practice. Wave one or ‘baseline’ data were collected when facilitators made an initial visit to each participating initial intervention and delayed intervention practice to explain the study, obtain consent from participants, and administer a baseline survey to clinicians and staff. Following Wave one data, Wave two data was collected one year after the initiation of the PF intervention in the practices randomized to the initial intervention, and at the start of the PF intervention for those randomized to the delayed intervention (control) practices. Wave three surveys were collected one year after the completion of the intervention in the initial intervention practices, and one year after the beginning of the PF intervention in the delayed intervention (control) practices.
The clinician/staff survey included the Assessment of Chronic Illness Care (ACIC) survey to measure the extent to which the care delivered in each practice was consistent with the elements of the CCM, the ‘CCM score’ [11]. Each item is scored on a 0 to 11 scale and provides sub-scale scores for each of the six CCM components as well as a total score. Scores from 0 to 2 represent ‘limited support,’ 3 to 5 represent ‘basic support,’ 6 to 8 is ‘good support,’ and 9 to 11 represent ‘fully developed support’. Version 3.5 of the ACIC was used in this study; in addition to the six subscales, it also includes items that address how well a practice integrates the CCM elements.
Analysis
After surveys were scanned into a database, the data were examined for outliers and missing values. Missing data in the surveys were addressed using multiple imputation following the recommended procedures [12, 13]. The CCM scores from the ACIC survey were calculated as recommended by the originators of the ACIC instrument [11]. Prior work suggests that survey measures of chronic illness care through surveys of practice members may vary by role within the practice and are more valid if measured by clinicians and those involved in direct patient care compared to other office staff [14, 15]. Cognizant of these findings, we limited our analysis of the ACIC surveys to those returned by those in direct patient care (physicians, advanced nurse practitioners and physician assistants, RNs, and LVN or LPNs) in each practice. Delivery of care consistent with the CCM is a practice-level construct, therefore a mean ACIC score for each practice was calculated for each of the three waves of data collection. We used a full factorial repeated-measures analysis of variance (ANOVA) to assess the main effects (mean differences between groups) and the within-group change over time. This approach allowed us to determine whether one group changed more rapidly over time (group-by-time interaction).