Abstract
Background
Shortening time between office visits for patients with uncontrolled hypertension represents a potential strategy for improving blood pressure (BP).
Objective
We evaluated the impact of multimodal strategies on time between visits and on improvement in systolic BP (SBP) among patients with uncontrolled hypertension.
Design
We used a stepped-wedge cluster randomized controlled trial with three wedges involving 12 federally qualified health centers with three study periods: pre-intervention, intervention, and post-intervention.
Participants
Adult patients with diagnosed hypertension and two BPs ≥ 140/90 pre-randomization and at least one visit during post-randomization control period (N = 4277).
Intervention
The core intervention included three, clinician hypertension group-based trainings, monthly clinician feedback reports, and monthly meetings with practice champions to facilitate implementation.
Main Measures
The main measures were change in time between visits when BP was not controlled and change in SBP. A secondary planned outcome was changed in BP control among all hypertension patients in the practices.
Key Results
Median follow-up times were 34, 32, and 32 days and the mean SBPs were 142.0, 139.5, and 139.8 mmHg, respectively. In adjusted analyses, the intervention did not improve time to the next visit compared with control periods, HR = 1.01 (95% CI: 0.98, 1.04). SBP was reduced by 1.13 mmHg (95% CI: −2.10, −0.16), but was not maintained during follow-up. Hypertension control (< 140/90) in the practices improved by 5% during intervention (95% CI: 2.6%, 7.3%) and was sustained post-intervention 5.4% (95% CI: 2.6%, 8.2%).
Conclusions
The intervention failed to shorten follow-up time for patients with uncontrolled BP and showed very small, statistically significant improvements in SBP that were not sustained. However, the intervention showed statistically and clinically relevant improvement in hypertension control suggesting that the intervention affected clinician decision-making regarding BP control apart from visit frequency. Future practice initiatives should consider hypertension control as a primary outcome.
Clinical Trial
www.ClinicalTrials.gov Identifier: NCT02164331
Supplementary Information
The online version contains supplementary material available at 10.1007/s11606-021-07016-9.
Supplementary Information
The online version contains supplementary material available at 10.1007/s11606-021-07016-9.
Key Words: hypertension, blood pressure control, clinical practice guidelines, practice-based research networks (PBRNs), Community Health Centers (CHCs), Federally Qualified Health Centers (FQHCs), Stepped Wedge Cluster Randomized Clinical Trial (SW-CRCT), pragmatic trials, electronic health records (EHRs)
BACKGROUND
Hypertension is a common primary care medical diagnosis[1, 2] and a leading, modifiable risk factor for cardiovascular disease (CVD).[1, 3] Despite hypertension national guidelines, blood pressure (BP) control remains suboptimal, particularly in low-income and minority populations. Sub-optimally controlled BP is a major contributor to Black-White disparities in cardiovascular morbidity and mortality.[4]
The Seventh and Eighth Joint National Committees (JNC-7 and JNC-8) on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure recommend 4-week follow-up visits for patients not at BP goal.[3, 5] However, most hypertension patients with uncontrolled BP are not re-assessed within 4 weeks.[6–9] We hypothesized that shortening follow-up time when BP was uncontrolled would improve BP and hypertension control.
OBJECTIVE
We designed a multimodal intervention to promote guideline adoption and assessed follow-up time for patients with uncontrolled BP and systolic blood pressure (SBP) using a pragmatic, stepped wedge, cluster randomized controlled trial (SWCRT). Our premise was that an intervention that targeted primary care clinician (PCC) awareness, confidence, norms, perceived control, and routines relevant to this recommendation would (1) reduce visit follow-up time for hypertensive patients with uncontrolled BP and (2) reduce SBP.[10]
DESIGN
We conducted a SWCRT between 4/1/2014 and 12/13/2017 involving 12 federally qualified health centers (FQHCs). The stepped wedged design is a unidirectional crossover design on clusters from the control to intervention conditions on a staggered schedule, so each cluster serves as its own control. We used a SWCRT, given sites’ wishes to receive the intervention and the need for sequential rollout (Fig. 1). The 12 FQHCs were randomized to one of three phases of intervention with each phase involving 4 FQHCs. The intervention period lasted 1 year and the number of control periods ranged from 1 to 3 depending on which phase the FQHC was randomly assigned. The post-intervention was defined just after the intervention period ends (Fig. 1).
Figure 1.
Stepped wedge time periods. The stepped wedge cluster randomized trial involved rollout in three steps. Each wedge involved four FQHCs, with three time periods—pre-intervention (control), intervention, and post-intervention periods. The pre-randomization eligibility period is not included
Randomization and Blinding
An offsite study statistician (HH) conducted randomization using computer-generated random numbers using the FQHC study identification number. The study was unblinded given the use of a SWCRT.
PARTICIPANTS AND SETTING
The 12 FQHCs are located within urban, low-income neighborhoods in New York City/New Jersey. Practices were recruited through CDN (www.CDNetwork.org), recognized by NIH as a best-practice, practice-based research network and by the Agency for Healthcare Research and Quality as a Center of Excellence (P30). Practice, PCC. Patient eligibility are shown (Table 1). Patients were eligible for the inception cohort for primary analysis if their PCC consented, they had two consecutive visits with BP > 140/90 during pre-randomization, at least one visit and BP reading post-randomization. Patients without subsequent visits/readings were censored at the end of the period in which their last visit/BP reading was recorded.
Table. 1.
Eligibility Criteria
| Inclusion criteria | Exclusion criteria |
|---|---|
| Federally Qualified Health Center (FQHC) | |
| Membership in Clinical Directors Network (CDN) Practice-Based Research Network (PBRN) | Practices planning to change electronic health record (EHR) systems or merge with a non-participating organization during the study period |
| Leadership commitment to full participation | |
| Designation of practice champion | |
| Use of the EHR for at least 6 months prior to beginning the study | |
| Ability to successfully generate and export patient-level data from the EHR | |
| PCCs | |
| Provision of primary care (both full and part-time) to adult patients within a CDN participating FQHC* | Primary Care Clinicians (PCCs) planning to leave the practice within six months of the study starting |
| Patients | |
| Age 18 years and older | |
| A hypertension diagnosis based on (ICD-9 401.x-405.x or ICD-10 I10.x) either based on billing codes or the patient problem listed in the EHR during pre-randomization. | |
| Two consecutive visits with BP > 140/90 during pre-randomization, at least one visit and BP reading during the pre-intervention, and at least one visit and BP reading with consented PCCs, post-randomization, i.e., pre-intervention, intervention, or post-intervention periods | |
| A primary care visit with the consenting PCC during post-randomization, i.e., pre-intervention, intervention, or post-intervention periods | |
*Primary care clinicians (PCCs) include family physicians, internal medicine physicians, geriatricians, nurse practitioners, and physician assistants
Ethical approval was obtained from the University of Rochester, CDN, and New York University Institutional Review Boards. We obtained waivers of informed consent for accessing de-identified data for all patient participants, with encrypted numbers for patients and PCCs.[11, 12] The study is registered at www.ClinicalTrials.gov (NCT 02164331).
INTERVENTION
The intervention targeted both the PCCs and the practices and involved three core strategies: three PCC trainings on hypertension guidelines (JNC-8) and application, monthly PCC audit and feedback reports regarding their performance, and monthly practice champion calls. Strategies targeted PCC attitudes, norms, and perceived control and included training in hypertension follow-up visits, BP targets, antihypertensive agents, and patient-centered counseling, in addition to practice-level processes, e.g., treatment algorithms and EHR templates.[13] Supplementary strategies included outreach to patients with uncontrolled BP not seen within 90 days, informational newsletters, and PCC access to a hypertension specialist (SW). We trained practice staff in BP measurement using automated office blood pressure (AOBP) devices, widely available in the practices.[14]
Usual Care (Control)
During the control (pre-intervention) period, FQHCs continued with their usual management of patients with uncontrolled hypertension.
Data Collection
The unit of randomization was the practice, and the unit of analysis was the patient, nested within the PCC and practice. Following PRECIS-2 domains for pragmatic trials,[15] we used EHR data for patient selection, assessment of intervention implementation, and outcomes. Patient and site measures were assessed through export, transfer, loading, cleaning, and standardization of data from the EHR. We obtained continuous EHR data to cover relevant time periods using unique encrypted identifiers for practices, PCCs, and patients. Using these data, we assessed patient age, sex, race/ethnicity, insurance, smoking, BMI, comorbidity conditions (based on ICD9/10 codes), and codes for hypertension (ICD9 codes 401x - 405x and ICD-10 I10.x), as well as dates of primary care visits and corresponding recorded BP readings.
MAIN MEASURES
Our two primary outcomes were (1) time (# days) to subsequent visit following a visit with uncontrolled BP defined as ≥ 140/90 mmHg and (2) systolic blood pressure at the last visit of the period. A secondary outcome was BP control (< 140/90 mmHg), based on the last recorded BP during each period. We compared the intervention and post-intervention periods to the pre-intervention period by examining the changes in the measures over time. When more than one BP reading was available for a visit, BPs were averaged. As previously reported,[13] we estimated 80% power with an alpha of 0.05 to detect a 0.8 mmHg improvement in SBP with a sample size of 200 patients per site.
Statistical Analysis
Our primary analysis used an intention-to-treat (ITT) analysis based on an inception cohort of eligible patients who had two consecutive visits with BP ≥ 140/90 during pre-randomization and ≥ one visit with consented PCCs during post-randomization control periods. For patients with loss-of-follow-up, they were censored at the end of treatment condition period in which their last visit occurred.
Cox proportional hazard regression models were conducted to examine the effects of the intervention on the time to the follow-up visits with the 3-level treatment condition included as the primary predictor. The relative risk for follow-up visit is assessed by the hazard ratio defined as the exponential of the regression coefficient. The hazard ratios defined as the exponential of the regression coefficient were provided to assess the relative risk of follow-up visit for comparison of groups. In the Cox regression models, the control is treated as the reference level, so the estimated hazard ratio is the relative risk of follow-up visit comparing the intervention/post-intervention period to the control period. The primary predictor is the three-level indicator of the treatment period, 0=pre-intervention (control), 1= intervention period, and 2=post-intervention periods. A hazard ratio > 1.0 means that the follow-up visit time is shorter than the control period. We included site as a covariate to account for clustering. We applied a shared frailty model to take into account the correlations for visits among the same PCC using PCC as random effect.[16] We also used empirical variance-covariance estimate to address correlations arising from repeated measures due to multiple visits for individual patients. By employing the fixed effect of the site, PCC random effect, and empirical variance-covariance estimate, the multi-level, clustering effects are well accounted for in the analysis. We controlled for patient demographics, i.e., age, race and ethnicity, smoking status, health insurance, comorbidity, and BMI in the models.
To assess the intervention effect on the change of SBP at the last visit during the intervention period and during the follow-up period SBP, we conducted linear, mixed-effects models with the same set of predictor/covariates in the aforementioned Cox model as fixed effects. To account for the multilevel clustering effects arising from the site, PCC, and repeated measures, similar to the Cox model, in addition to including the site as fixed effect, we included a random PCC intercept in the model and applied empirical variance-covariance estimate to address correlations from repeated measures and any residual effects due to clustering. We examined the heterogeneity of treatment effects (HTEs) for several pre-specified subgroups using the same methods for patients with higher SBP ≥ 160 mmHg at baseline. We also applied the JNC-8, age-based definitions of SBP control, i.e., patients age ≥ 60 years with SBP ≥ 150 or age < 60 years with SBP ≥ 140. Last, we assessed the relationship between changes in follow-up time and SBP.
In planned secondary analyses,[13] we replaced SBP with BP control. We used a generalized linear mixed-effect model (GLMM) with an identity link in the model, so the coefficient can be interpreted as a change of probability of having the BP controlled with positive (negative) coefficient indicating a higher (lower) probability of having the BP controlled. Similar strategies as those for the primary outcomes are used to address the multi-level clustering effect. Both unadjusted and adjusted analyses were conducted to examine the intervention effect with and without potential confounders controlled. The control variables are the same set of variables used in the primary analysis. We examined the potential confounding by wedge by including wedge in the model and observed the same estimated treatment effects. The results reported here do not include wedge controls. Last, we surveyed PCCs at baseline and post-intervention regarding their follow-up time for patients with uncontrolled BP and blood pressure goals.
RESULTS
We successfully extracted EHR data from seven different EHR vendor systems across all 12 practices. Pre-randomization hypertension control ranged from 47 to 69%. PCCs per practice ranged from 5 to 10. Seventy-seven PCCs consented: two-thirds were physicians and one-third were nurse practitioners or physician assistants. Two-thirds of the sample were female; two-thirds were Black or Hispanic; and the majority had worked at the FQHC for 5 or more years. PCC turnover by site ranged from none to 80%. Between 40 and 100% of consented PCCs attended training sessions. All received audit and feedback reports (Figures S1, S2). All practice champions regularly attended monthly calls and half of practices conducted patient outreach.
Patient characteristics by step and overall are shown (Table 2). In unadjusted analyses, the overall median (interquartile range) times in days to follow-up were 34 (69), 32 (59), and 32 (54) for pre-intervention, intervention, and post-intervention periods, respectively. The overall unadjusted mean (median) SBPs in mmHg at the last visit of each period were 140.96 (140), 139.69 (138), and 140.47 (138) in the pre-intervention, intervention, and post-intervention periods, respectively. Within patient SBP, visit-to-visit variation was high (standard deviation = 14.4 mmHg).
Table. 2.
Practice and Patient Factors at Baseline by Step with the SWCRT for Consented Patients
| Patient characteristics (%) | Step 1 | Step 2 | Step 3 | Overall |
|---|---|---|---|---|
| Practices (fixed effects) | 4 | 4 | 4 | 12 |
| Patient | 1597 | 1087 | 1593 | 4277 |
| Age mean in years (SD) | 61.09 (12.52) | 60.60 (12.43) | 58.97 (12.77) | 60.17 (12.62) |
| Sex N (%) | ||||
| Male N (%) | 483 (30.40) | 539 (49.72) | 492 (30.89) | 1514 (35.49) |
| Female N (%) | 1106 (69.60) | 545 (50.28) | 1101 (69.11) | 2752 (64.51) |
| Race N (%) | ||||
| Asian N (%) | 39 (2.45) | 3 (0.28) | 12 (0.75) | 54 (1.27) |
| Black N (%) | 1093 (68.79) | 357 (32.93) | 802 (50.35) | 2252 (52.79) |
| Other N (%) | 124 (7.80) | 66 (6.09) | 199 (12.49) | 389 (9.12) |
| White N (%) | 121 (7.61) | 313 (28.87) | 157 (9.86) | 591 (13.85) |
| Unknown race N (%) | 185 (11.64) | 345 (31.83) | 381 (23.92) | 911 (21.35) |
| Hispanic ethnicity | ||||
| Yes N (%) | 345 (21.60) | 409 (37.63) | 387 (24.29) | 1141 (26.68) |
| No N (%) | 1132 (77.83) | 675 (62.10) | 951 (59.70) | 2869 (67.08) |
| Unknown/missing* N (%) | 9 (0.56) | 3 (0.38) | 255 (16.01) | 267 (6.24) |
| Insurance | ||||
| Commercial/other N (%) | 288 (18.03) | 281 (25.85) | 392 (24.61) | 961 (22.47) |
| Medicare N (%) | 262 (16.41) | 307 (28.24) | 276 (17.33) | 845 (19.76) |
| Medicaid N (%) | 819 (51.28) | 468 (43.05) | 843 (52.92) | 2130 (59.80) |
| No insurance | 228 (14.28) | 31 (2.85) | 82 (5.15) | 341 (7.97) |
| Smoker | ||||
| Yes N (%) | 718 (44.96) | 286 (26.31) | 577 (36.22) | 1581 (36.97) |
| No N (%) | 716 (44.83) | 573 (52.71) | 613 (38.48) | 1902 (44.47) |
| Missing* N (%) | 163 (10.21) | 228 (20.98) | 403 (25.30) | 794 (18.56) |
| BMI (body mass index) mean (SD) | 31.14 (5.83) | 31.42 (6.10) | 30.92 (7.55) | 31.14 (6.51) |
| Number of comorbidities mean (SD) | 0.57 (1.33) | 0.64 (1.54) | 0.52 (1.07) | 0.57 (1.30) |
| Baseline: follow-up time to next visit (days), median (IQR†) | 37 (72) | 24 (47) | 41 (74) | 34 days (68) |
| Baseline: systolic blood pressure [SBP] (mmHg) mean (median)‡ | 144.55 (143) | 143.31 (142) | 143.86 (142) | 143.98 (142) |
*Missing data refers to data that were not present in the EHR at the time of extraction
†Visit follow-up time when the BP > 140/90 at the preceding visit for each period, i.e., baseline, intervention, post-intervention, averaged across each period; IQR is for inter-quartile range
‡Systolic blood pressures (SBP) averaged across each period, i.e., baseline, intervention, and post-intervention,N=Number, SD= Standard Deviation
The full models based on the ITT analysis for the inception cohort with each site as a fixed effect are shown in Table 3 for the time to the next visit and Table 4 for SBP measures. Compared to the control period, overall the follow-up visit time was not improved during the intervention period with hazard ratios (HR) = 1.01 (95% CI: [0.98, 1.04]), but was slightly lengthened for the post-intervention period with HR = 0.97 (95% CI: [0.93, 1.00]) (Table 3). There was variation in follow-up visit time across the study sites. Treating one of the sites (site 12), based on alphabetical order, as the reference site, HRs ranged from 1.44 (95% CI: [1.27, 1.63]) for shorter follow-up visit time to HR = 0.48 (95% CI: [0.42, 0.55]) for a longer follow-up visit time. Older patients, females, and those with more comorbidity and insurance coverage showed shorter follow-up times. Results were not significant when follow-up visits were dichotomized at 4 weeks.
Table. 3.
Intervention Effects on Time to The Next Primary Care Visit from Initial Uncontrolled BP Visit for Patients with Consented PCC Based on Proportional Hazards Model
| Parameter | Hazard ratio | 95% confidence interval | p value |
|---|---|---|---|
| Intervention (reference: pre-intervention period) | |||
| Intervention period | 1.01 | (0.98, 1.04) | 0.615 |
| Post-intervention period | 0.97 | (0.93, 1.00) | 0.058 |
| Site (reference: site 12) | |||
| Site 1 | 1.08 | (0.95, 1.24) | 0.242 |
| Site 2 | 1.17 | (1.04, 1.32) | 0.010 |
| Site 3 | 0.90 | (0.78, 1.05) | 0.185 |
| Site 4 | 0.80 | (0.70, 0.91) | <.001 |
| Site 5 | 1.35 | (1.22, 1.49) | <.001 |
| Site 6 | 0.87 | (0.77, 0.97) | 0.014 |
| Site 7 | 0.82 | (0.72, 0.94) | 0.005 |
| Site 8 | 0.48 | (0.42, 0.55) | <.001 |
| Site 9 | 0.62 | (0.53, 0.73) | <.001 |
| Site 10 | 1.02 | (0.90, 1.16) | 0.719 |
| Site 11 | 1.44 | (1.27, 1.63) | <.001 |
| SBP baseline | 1.00 | (1.00, 1.00) | <.001 |
| Smoking status (reference: non-smoker) | |||
| Smoker | 1.03 | (1.00, 1.06) | 0.043 |
| Smoker missing | 0.75 | (0.70, 0.80) | <.001 |
| Sex (reference: female) | |||
| Male | 0.97 | (0.94, 0.99) | 0.016 |
| Hispanic ethnicity (reference: non-hispanic) | |||
| Hispanic | 1.06 | (1.02, 1.11) | 0.005 |
| Hispanic missing | 1.00 | (0.92, 1.08) | 0.960 |
| Race (reference: multiple) | |||
| Black | 0.95 | (0.87, 1.04) | 0.288 |
| White | 0.95 | (0.86, 1.05) | 0.303 |
| Asian | 1.02 | (0.88, 1.18) | 0.791 |
| Unknown | 0.97 | (0.88, 1.07) | 0.514 |
| Other | 0.89 | (0.81, 0.98) | 0.019 |
| Baseline body mass index (BMI) (reference: BMI<30) | |||
| BMI>=30 | 1.03 | (1.00, 1.06) | 0.023 |
| BMI missing* | 1.09 | (1.04, 1.14) | <.001 |
| Age | 1.00 | (1.00, 1.00) | <.001 |
| Insurance (reference: none) | |||
| Commercial | 1.05 | (0.98, 1.12) | 0.157 |
| Medicaid | 1.12 | (1.05, 1.19) | <.001 |
| Medicare | 1.06 | (0.99, 1.13) | 0.094 |
| Comorbidity (reference: 0–1) | |||
| 2–3 | 1.19 | (1.15, 1.23) | <.001 |
| 3+ | 1.33 | (1.27, 1.40) | <.001 |
*BMI Missing refers to data that were not present in the EHR when data were extracted
Table. 4.
Intervention Effects on SBP for Patients with Consented PCC by Linear Mixed-Effect Model
| Effect | Estimate | 95% confidence interval | p value |
|---|---|---|---|
| Intercept | 97.56 | (93.40, 102.72) | <.001 |
| SBP baseline | 0.29 | (0.26, 0.31) | <.001 |
| Intervention (reference: pre-intervention period) | |||
| Intervention period | −1.13 | (−2.10, −0.16) | 0.022 |
| Post-intervention period | −029 | (−1.37, 0.78) | 0.591 |
| Site (reference: site 12) | |||
| Site 1 | −2.29 | (−4.52, −0.05) | 0.045 |
| Site 2 | −2.57 | (−4.06, −1.08) | <.001 |
| Site 3 | −6.32 | (−7.66, −4.98) | <.001 |
| Site 4 | −2.40 | (−4.66, −0.14) | 0.038 |
| Site 5 | 0.80 | (−1.45, 3.06) | 0.486 |
| Site 6 | 0.00 | (−1.44, 1.45) | 0.996 |
| Site 7 | −0.54 | (−4.57, 3.50) | 0.794 |
| Site 8 | −3.35 | (−5.06, −1.64) | <.001 |
| Site 9 | 8.58 | (4.92, 12.23) | <.001 |
| Site 10 | 0.15 | (−1.82, 2.12) | 0.881 |
| Site 11 | −3.40 | (−5.22, −1.58) | <.001 |
| Smoking status (reference: non-smoker) | |||
| Smoker | −0.20 | (−1.11, 0.72) | 0.673 |
| Smoker missing* | 1.72 | (−0.13, 3.56) | 0.069 |
| Sex (reference: female) | |||
| Male | 0.32 | (−0.61, 1.25) | 0.503 |
| Hispanic ethnicity (reference: non-hispanic) | |||
| Hispanic | −0.21 | (−1.46, 1.04) | 0.740 |
| Hispanic missing* | 1.66 | (−3.05, 6.38) | 0.490 |
| Race (reference: multiple) | |||
| Black | −2.29 | (−4.33, −0.26) | 0.027 |
| White | −4.62 | (−6.98, −2.26) | <.001 |
| Asian | −4.21 | (−8.05, −0.37) | 0.032 |
| Unknown | −3.21 | (−5.48, −0.93) | 0.006 |
| Other | −3.57 | (−5.99, −1.16) | 0.004 |
| Baseline body mass index (BMI) (reference: BMI<30) | |||
| BMI>=30 | 0.42 | (−0.41, 1.25) | 0.323 |
| BMI missing* | 1.65 | (0.39, 2.91) | 0.010 |
| Age | 0.11 | (0.07, 0.15) | <.001 |
| Insurance (reference: none) | |||
| Commercial | −1.69 | (−3.47, 0.09) | 0.063 |
| Medicaid | −1.11 | (−2.79, 0.57) | 0.195 |
| Medicare | −0.71 | (−2.79, 1.38) | 0.507 |
| Comorbidity (reference: 0–1) | |||
| 2–3 | −0.41 | (−1.83, 1.01) | 0.571 |
| 3+ | −0.42 | (−2.42, 1.59) | 0.684 |
*Missing refers to data that were not present in the EHR when data were extracted
Compared to the control period, the intervention was associated with a very small but statistically significant reduction in SBP −1.13 mmHg (95% CI: [−0.16, −2.10]) (Table 4). However, this reduction in SBP was not sustained during follow-up with reduction in SBP of 0.29 mmHg (95% CI: [−1.37, 0.78]). The changes in SBP among practices ranged from a reduction of 6.32 mmHg (95% CI: [−7.66, −4.98]) to an increase of 8.58 mmHg (95% CI: [4.92, 12.23]) compared to the reference site (Table 4).
Sensitivity analyses, including use of age-based thresholds (from the panel appointed to JNC-8) of < 60 vs. ≥ 60 years, showed no significant reductions in time between visits and the reductions in SBP observed during the intervention, 0.80 mmHg (95% CI: [−1.91, 0.30]), increased to 1.52 mmHg (95% CI: [−2.75, −0.29]), and became statistically significant during the follow-up period. Sensitivity analyses based on SBP ≥ 160 mmHg at baseline showed no significant reduction in time between visits for these patients with higher baseline SBP. Among this subset of patients, SBP reductions were only 0.82 mmHg (95% CI: [−1.73, 0.10]) and increased to 2.48 mmHg (95% CI: [−4.38, −0.58]) at follow-up.
We also examined the association between the follow-up time with uncontrolled SBP and changes in SBP. The unadjusted association between visit follow-up and SBP was 0.01 mmHg/day (95% CI: [0.006, 0.01], p < 0.001), i.e., every 1-day delay in coming back to the office for a follow-up visit was associated with a negligible 0.01 mmHg increase in SBP.
Unadjusted hypertension control improved from 45 to 50.7% during intervention and 51.5% post intervention. Adjusted analyses (Table S1) confirmed a 5% statistically significant improvement during intervention (95% [CI: 2.6%, 7.3%]) which was sustained post-intervention 5.4% (95% [CI: 2.6%, 8.2%]).
Last, we examined survey responses (96% response rate) from consented PCCs (Table S2). At baseline, 94% of PCCs reported scheduling patients with uncontrolled BP within a month. This increased to 98% at follow-up suggesting potential ceiling effects on PCC follow-up plans. At baseline and follow-up, most PCCs reported using the age-based, SBP thresholds for initiating treatment.
DISCUSSION
We designed and implemented a multimodal approach to promote monthly office visits among FQHC patients with uncontrolled hypertension and to optimize their BP management. The intervention failed to achieve a statistically significant reduction in time between visits when BP was not controlled. The intervention achieved a statistically significant but clinically insignificant improvement in SBP, which was not maintained. However, with regard to a secondary clinical outcome, BP control, we observed statistically and clinically significant improvements in BP control during the intervention that were sustained post-intervention.
Our results for our primary outcomes are consistent with a number of prior studies that have shown either small or null effects of randomized trials to improve BP within FQHCs and safety net practices in the USA.[17–19] Minimal findings for our primary outcomes likely reflect higher than anticipated commitment by PCCs to scheduling patients with uncontrolled hypertension to return in 4 weeks. More than 90% of PCCs reported that they already do so and the median follow-up time was 34 days, creating a ceiling effect and thus reducing the opportunity for further improvement. Minimal improvements in SBP seen within the inception cohort likely represent a combination of no change in visit follow-up, pre-intervention SBPs that were already close to 140 mmHg, and high patient visit-to-visit SBP variability. Potentially, these limitations resulted in patients often having SBP controlled at one visit and not at the next, potentially contributing to PCC uncertainty regarding the need for shorter follow-up intervals and pharmacological intensification. Notably, prior studies show that high visit-to-visit BP variability partly reflects suboptimal medication adherence and is an independent cardiovascular risk factor.[20, 21]
The minimal effects when SBP was used as the primary outcome as compared to statistically and clinically significant effects when BP control was used as the outcome seem paradoxical. It likely represents most patients being very close to a SBP of 140 and PCC’s addressing reduction in SBP and DBP. We chose SBP as our primary outcome because it is a continuous measure and as is strongly associated with CVD outcomes. In retrospect, BP control is arguably the more pragmatic outcome because PCCs make clinical decisions based on it and because control rate is the measure used by practices to monitor and report their hypertension control quality performance.
Notably, the 5% improvement in hypertension control occurred despite challenges faced by FQHCs. PCC turnover varied widely, with one site experiencing 80% turnover. PCC turnover reduces exposure to the intervention and impacts patients by interrupting continuity of care and disrupting patient-PCC relationships, potentially undermining patient adherence and continuity with the practice.[22–25]
Patient no-show rates within practices ranged from 30 to 50%, which likely blunted effects on actual visit follow-up time and SBP reduction. A systematic review of no-show appointments that included studies conducted internationally showed cross-sectional associations between uncontrolled BP and missed appointments.[26] Rowan et al. reported that patients with stage 2 hypertension who had no baseline office visits were twice as likely not to achieve BP control compared with patients who had visits every other month.[27]
JNC-8 guidelines released just before the study began may have blunted the effects. Most PCCs reported following these age-based treatment guidelines. Sensitivity analyses showed larger effects in the post-intervention period when age-based thresholds were used, likely reflecting PCCs’ practice patterns. Moreover, effects were larger in the post-intervention period among patients with higher SBP, suggesting possible lagged effects. Last, our findings suggest that practices may respond differently to the same intervention and raise questions regarding how to better adapt the intervention to the specific needs of each practice, although we could not discern a relationship between the practice-level factors and improved SBP.
Strengths and Limitations of the Study
Strengths include successful design, implementation, and completion of a SWCRT among FQHCs with primary outcomes and covariates from extracted EHR data. These strengths improve generalizability to other FQHCs and safety-net primary care practices. Limitations include floor effects on improvement in visit follow-up and challenges of conducting pragmatic studies. We did not use standardized research-grade procedures for measuring BP, but rather, we trained staff in appropriate BP measurement using their existing automated devices, which generally align with the gold standard of ambulatory readings.[14] We lacked reliable data on hypertensive pharmacological treatment inertia, thus precluding evaluation of this potential mechanism.
CONCLUSION
A multimodal, evidence-based intervention with 12 FQHCs did not reduce time between visits or sustain improvements in SBP in the inception cohort, potentially due to baseline ceiling effects for improvement in follow-up times and SBP and to patient visit-to-visit variability in SBP. The notable improvement in our secondary outcome of hypertension control suggests that while the intervention did not improve follow-up time or SBP levels, the multimodal intervention did improve overall hypertension control. This finding suggests that PCC training, audit and feedback reports, and practice champion meetings likely affected clinician-decision surrounding BP control within FQHC practices more broadly. Future practice-wide interventions might consider how to tailor the intervention to the practice and examine hypertension control rates as a primary outcome.
Supplementary Information
(PDF 46 kb)
(DOCX 26 kb)
(DOCX 397 kb)
Acknowledgments
Contributors
Study data were collected and managed using REDCap electronic data capture tools from NCATS/NIH. REDCap (Research Electronic Data Capture) is a secure, web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for importing data from external sources. Participating FQHCs: New York sites (1) Brownsville Multi-Service Family Health Center, Brooklyn; (2) Caribbean-American: NYU Lutheran Family Health Center, Brooklyn; (3) Community Healthcare Network – Lower East Side and CABS, New York City and Brooklyn; (4) Community Healthcare Network – Jamaica, Queens; (5) Joseph P. Addabbo Family Health Center, Queens; (6) Morris Heights Health Center, Bronx; (7) ODA Primary Health Care, Brooklyn; and (8) Urban Health Plan Inc., Bronx; New Jersey sites (1) CamCare Health Corporation, Camden; (2) Henry J. Austin Health Center, Trenton; (3) Metropolitan Family Health Network, Jersey City; and (4) Newark Community Health Centers, Newark.
Technical Assistance
Kathleen Silver provided vital assistance in editing/proofing, formatting, generation of figures, and submission.
Funding
This study was funded by a research dissemination and implementation grant (1R18HL117801) from the National Heart, Lung, and Blood Institute of the National Institutes of Health (NIH). Resources included data collection, which occurred at the University of Rochester with Clinical and Translational Science Institute grant support (UL1 TR002001) from the National Center for Advancing Translational Sciences (NCATS)/NIH) and the Network of Practice-based Research Networks (N2-PBRN) at Clinical Directors Network with funding from the AHRQ-Designated Center of Excellence (P30) for Practice-based Research and Learning: “N2: Building a Network of Safety Net PBRNs” (Grant No. 1 P30-HS-021667). We would like to acknowledge the guidance of our Program Officers, Paula Einhorn, MD (NHLBI R18) and Rebecca Roper, MS MPH (AHRQ P30) and assistance from Subrina Farah and the DARTNet Instutite for EHR data extraction and standardization.
Declarations
The study is registered at www.ClinicalTrials.gov (NCT 02164331). Ethical approval was obtained from the University of Rochester, CDN, and New York University Institutional Review Boards (IRBs). We obtained waivers of informed consent for accessing de-identified data for patient participants.
Conflict of Interest
The authors declare that they do not have a conflict of interest.
Footnotes
Prior Presentations: None.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.National Center for Health Statistics CDC. Health, United States 2012. 2013:505. Available at: https://www.cdc.gov/nchs/data/hus/hus12.pdf Accessed June 8, 2021. [PubMed]
- 2.Kit BK, Ogden CL, Flegal KM. Prescription medication use among normal weight, overweight, and obese adults, United States, 2005-2008. Ann Epidemiol. 2012;22(2):112–119. doi: 10.1016/j.annepidem.2011.10.010. [DOI] [PubMed] [Google Scholar]
- 3.Chobanian AV, Bakris GL, Black HR, et al. Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure. Hypertension. 2003;42(6):1206–1252. doi: 10.1161/01.HYP.0000107251.49515.c2. [DOI] [PubMed] [Google Scholar]
- 4.Wong MD, Shapiro MF, Boscardin WJ, Ettner SL. Contribution of major diseases to disparities in mortality. N Engl J Med. 2002;347(20):1585–1592. doi: 10.1056/NEJMsa012979. [DOI] [PubMed] [Google Scholar]
- 5.James PA, Oparil S, Carter BL, et al. 2014 evidence-based guideline for the management of high blood pressure in adults: report from the panel members appointed to the Eighth Joint National Committee (JNC 8) JAMA. 2014;311(5):507–520. doi: 10.1001/jama.2013.284427. [DOI] [PubMed] [Google Scholar]
- 6.Berlowitz DR, Ash AS, Hickey EC, et al. Inadequate management of blood pressure in a hypertensive population. N Engl J Med. 1998;339(27):1957–1963. doi: 10.1056/NEJM199812313392701. [DOI] [PubMed] [Google Scholar]
- 7.Guthmann R, Davis N, Brown M, Elizondo J. Visit frequency and hypertension. J Clin Hypertens (Greenwich). 2005;7(6):327–332. doi: 10.1111/j.1524-6175.2005.04371.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Turchin A, Goldberg SI, Shubina M, Einbinder JS, Conlin PR. Encounter frequency and blood pressure in hypertensive patients with diabetes mellitus. Hypertension. 2010;56(1):68–74. doi: 10.1161/HYPERTENSIONAHA.109.148791. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lichtenstein MJ, Sweetnam PM, Elwood PC. Visit frequency for controlled essential hypertension: general practitioners' opinions. J Fam Pract. 1986;23(4):331–336. [PubMed] [Google Scholar]
- 10.Ajzen I. Theory of planned behavior. Handb Theor Soc Psychol Vol One. 2011;1:438. [Google Scholar]
- 11.Sugarman J, Califf RM. Ethics and regulatory complexities for pragmatic clinical trials. JAMA. 2014;311(23):2381–2382. doi: 10.1001/jama.2014.4164. [DOI] [PubMed] [Google Scholar]
- 12.Pletcher MJ, Lo B, Grady D. Informed consent in randomized quality improvement trials: a critical barrier for learning health systems. JAMA Intern Med. 2014;174(5):668–670. doi: 10.1001/jamainternmed.2013.13297. [DOI] [PubMed] [Google Scholar]
- 13.Fiscella K, Ogedegbe G, He H, et al. Blood pressure visit intensification study in treatment: trial design. Am Heart J. 2015;170(6):1202–1210. doi: 10.1016/j.ahj.2015.08.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Roerecke M, Kaczorowski J, Myers MG. Comparing automated office blood pressure readings with other methods of blood pressure measurement for identifying patients with possible hypertension: a systematic review and meta-analysis. JAMA Intern Med. 2019;179(3):351–362. doi: 10.1001/jamainternmed.2018.6551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Loudon K, Treweek S, Sullivan F, Donnan P, Thorpe KE, Zwarenstein M. The PRECIS-2 tool: designing trials that are fit for purpose. Brit Med J. 2015;350:h2147. doi: 10.1136/bmj.h2147. [DOI] [PubMed] [Google Scholar]
- 16.Therneau TM, Grambsch PM. Modeling Survival Data: Extending the Cox Model. New York, NY: Springer-Verlag; 2000.
- 17.Hicks LS, Sequist TD, Ayanian JZ, et al. Impact of computerized decision support on blood pressure management and control: a randomized controlled trial. J Gen Intern Med. 2008;23(4):429–441. doi: 10.1007/s11606-007-0403-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ogedegbe G, Tobin JN, Fernandez S, et al. Counseling African Americans to control hypertension: cluster-randomized clinical trial main effects. Circulation. 2014;129(20):2044–2051. doi: 10.1161/CIRCULATIONAHA.113.006650. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lopez PM, Divney A, Goldfeld K, et al. Feasibility and outcomes of an electronic health record intervention to improve hypertension management in immigrant-serving primary care practices. Med Care. 2019;57(Suppl 6, Suppl 2):S164–S171. doi: 10.1097/MLR.0000000000000994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kronish IM, Lynch AI, Oparil S, et al. The association between antihypertensive medication nonadherence and visit-to-visit variability of blood pressure: findings from the antihypertensive and lipid-lowering treatment to prevent heart attack trial. Hypertension. 2016;68(1):39–45. doi: 10.1161/HYPERTENSIONAHA.115.06960. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Stevens SL, Wood S, Koshiaris C, et al. Blood pressure variability and cardiovascular disease: systematic review and meta-analysis. BMJ (Clinical research ed) 2016;354:i4098–i4098. doi: 10.1136/bmj.i4098. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kerse N, Buetow S, Mainous AG, 3rd, Young G, Coster G, Arroll B. Physician-patient relationship and medication compliance: a primary care investigation. Ann Fam Med. 2004;2(5):455–461. doi: 10.1370/afm.139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Safran DG, Montgomery JE, Chang H, Murphy J, Rogers WH. Switching doctors: predictors of voluntary disenrollment from a primary physician's practice. [comment] J Fam Pract. 2001;50(2):130–136. [PubMed] [Google Scholar]
- 24.Stransky ML. Unmet needs for care and medications, cost as a reason for unmet needs, and unmet needs as a big problem, due to health-care provider (dis)continuity. J Patient Exp. 2018;5(4):258–266. doi: 10.1177/2374373518755499. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Smith MA, Bartell JM. Changes in usual source of care and perceptions of health care access, quality, and use. Med Care. 2004;42(10):975–984. doi: 10.1097/00005650-200410000-00006. [DOI] [PubMed] [Google Scholar]
- 26.Lee RRS, Samsudin MI, Thirumoorthy T, Low LL, Kwan YH. Factors affecting follow-up non-attendance in patients with Type 2 diabetes mellitus and hypertension: a systematic review. Singapore Med J. 2019;60(5):216–223. doi: 10.11622/smedj.2019042. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Rowan CG, Flory J, Stempniewicz N, Cuddeback J, Brunelli SM. Stage 2 hypertension: predictors of failure to achieve blood pressure control and the impact of adding one additional antihypertensive class. Pharmacoepidemiol Drug Saf. 2015;24(11):1170-1179. [DOI] [PubMed]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
(PDF 46 kb)
(DOCX 26 kb)
(DOCX 397 kb)