Title:Designing and Evaluating Health Systems Level Hypertension Control Interventions for African-Americans: Lessons from a Pooled Analysis of Three Cluster Randomized Trials
Volume: 11
Issue: 2
Author(s): Valory N. Pavlik, Wenyaw Chan, David J. Hyman, Penny Feldman, Gbenga Ogedegbe, Joseph E. Schwartz, Margaret McDonald, Paula Einhorn and Jonathan N. Tobin
Affiliation:
Keywords:
African-Americans, hypertension control, cluster randomized trial, effect size, health disparities, health systems
interventions, high blood pressure, minority health, practice-based research network (PBRNs).
Abstract: Objectives: African-Americans (AAs) have a high prevalence of hypertension and their blood pressure (BP) control
on treatment still lags behind other groups. In 2004, NHLBI funded five projects that aimed to evaluate clinically feasible
interventions to effect changes in medical care delivery leading to an increased proportion of AA patients with controlled BP.
Three of the groups performed a pooled analysis of trial results to determine: 1) the magnitude of the combined intervention
effect; and 2) how the pooled results could inform the methodology for future health-system level BP interventions.
Methods: Using a cluster randomized design, the trials enrolled AAs with uncontrolled hypertension to test interventions
targeting a combination of patient and clinician behaviors. The 12-month Systolic BP (SBP) and Diastolic BP (DBP)
effects of intervention or control cluster assignment were assessed using mixed effects longitudinal regression modeling.
Results: 2,015 patients representing 352 clusters participated across the three trials. Pooled BP slopes followed a quadratic
pattern, with an initial decline, followed by a rise toward baseline, and did not differ significantly between intervention
and control clusters: SBP linear coefficient = -2.60±0.21 mmHg per month, p<0.001; quadratic coefficient = 0.167± 0.02
mmHg/month, p<0.001; group by time interaction group by time group x linear time coefficient=0.145 ± 0.293, p=0.622;
group x quadratic time coefficient= -0.017 ± 0.026, p=0.525). Results were similar for DBP. The individual sites did not
have significant intervention effects when analyzed separately.
Conclusion: Investigators planning behavioral trials to improve BP control in health systems serving AAs should plan for small
effect sizes and employ a “run-in” period in which BP can be expected to improve in both experimental and control clusters.
