As described previously, 26 regions in Colorado and 16 regions in New Mexico were randomized to standard versus intervention study arms. A total of 312 practices were initially given information about the study and invited to submit a practice enrollment application (Figure 2). A total of 239 practices completed initial enrollment for the project; 211 actually began the intervention and completed baseline surveys. Twelve practices dropped during the intervention; 199 practices completed the 9-month interventions. Characteristics of practices that dropped out were similar to those that remained (ie, not significantly different, all P > .15). Despite using methods to assure balance on population characteristics, study arms differed somewhat (Table 2), largely reflecting the higher percentage of FQHC and rural practices in the standard intervention arm. The external comparison group also differed somewhat from ENSW practices. Among external comparison practices, 68.5% were clinician owned, 25.5% hospital owned, and 6% FQHCs; 15.5% were in rural areas. On average, DARTNet practices had 2.6 full-time equivalent clinicians and were located in areas with 7.6% Hispanic and 5.5% black populations.
Table 3 shows results of the adjusted analyses of the ABCS measures over time for standard versus enhanced intervention arms, including odds ratios with 95% confidence intervals from the beta regression models. Change over time did not differ by study arm for any CQM (Table 2) (aspirin use, P = .114; blood pressure, P = .178; cholesterol, P = .078; smoking cessation, P = .596). In the absence of significant differential change, study arms were combined to assess overall improvement, which was significant for aspirin use (odds ratio [95% confidence interval]) (1.04 [1.01, 1.07], P < .001), cholesterol composite (1.05 [1.03, 1.08], P < .001), and smoking (1.03 [1.01, 1.06], P = .014), but not blood pressure control (1.01 [0.998, 1.03], P = .084).
Because practices were allowed to report cholesterol measures for subpopulations of patients instead of the composite measure, additional analyses were performed for these measures. Subpopulations included atherosclerotic CVD (ASCVD) (n = 109), LDL > 190 (n = 36), and diabetes (n = 93). Analyses of subpopulation measures showed overall improvement for ASCVD (P = .009) and LDL subpopulations (P = .004) and marginal improvement for diabetes (P = .077).
Comparison to External Comparison Group
To better understand whether improvement in ENSW practices might be due to temporal trends in cardiovascular care, ENSW practices were compared with the external comparison group on ABCS measures over time. Even after using procedures to maximize comparability between comparison practices and ENSW practices, there were differences in baseline levels of the ABCS measures, with ENSW practices having higher (much higher for aspirin use) values on all ABCS measures than the external comparison group. Baseline and 12-month follow-up estimates for ENSW and external comparison practices, back-transformed to represent actual proportions and adjusted for rural location and practice type, are shown in Table 4. Compared with external comparison practices, ENSW practices demonstrated greater improvement on all 4 ABCS measures, including blood pressure (all P < .05).
Implementation of the Building Blocks for High-performing Primary Care
The Building Blocks constituted an important intermediate outcome for both practice transformation interventions; we also hypothesized that improvement in certain key Building Blocks (especially Patient-Team Partnership) would be greater in the enhanced intervention arm. Table 5 shows results of the unadjusted and adjusted analyses of Implementation Tracker scores over time. When they were adjusted for practice type and rural location, there was a significant overall improvement in Team-Based Care, Patient Team Partnership, and Population Management (all P < .001, as indicated by improvement in the standard group in Table 5), and improvement was greater in enhanced intervention practices (all P < .05, indicated by differential improvement in the enhanced group in Table 5). There was also significant overall improvement in Leadership and Data-Driven Improvement (both P < .001), but improvement did not differ between standard and enhanced practices (both P > .3).
Based on the conceptual model and the framework for the Practice Transformation interventions, we hypothesized that more successful implementation of the Building Blocks would, in turn, result in improved delivery of cardiovascular care. To explore these relationships, we examined the association between change from baseline in the Building Blocks during the active 9-month intervention period and change from baseline in the ABCS CQMs at 9 to 15 months in a mediational analysis. Improvements in the overall Building Blocks (overall mean score), Data-Driven Improvement, Team-Based Care, and Population Management were significantly associated with improvement in blood pressure CQMs. Coefficients represent additional improvement in CQMs per 10-point increase in Building Block scores (coeff [SE]) (overall Building Blocks, 0.0128 [0.0054], P = .019; Data-Driven Improvement, 0.0010 [0.0004], P = .023; Team-Based Care; 0.0102 [0.0041], P = .013; Population Management, 0.0109 [0.0037], P = .003). Improvement in Population Management was also significantly associated with improvement in aspirin use CQMs (0.0074 [0.0036], P = .0415). Improvement in Building Blocks was not significantly associated with cholesterol or smoking CQMs (all P > .05).
J Am Board Fam Med. 2020;33(5):675-686. © 2020 American Board of Family Medicine