A community‐wide quality improvement initiative to improve hypertension control and reduce disparities

Abstract

Initiatives to improve hypertension control within academic medical centers and closed health systems have been extensively studied, but large community‐wide quality improvement (QI) initiatives have been both less common and less successful in the United States. The authors examined a community‐wide QI initiative across 226 843 patients from 198 practices in nine counties across upstate New York to improve hypertension control and reduce disparities. The QI initiative focused on (a) providing population and practice‐level comparative data, (b) community engagement, especially in underserved communities, and (c) practice‐level quality improvement assistance, but was not designed to examine causality of specific components. Across the nine counties, hypertension control rates improved from 61.9% in 2011 to 69.5% in 2016. Improvements were greatest among whites (73.7%‐81.5%) and more modest among black patients (58.8%‐64.7%). The authors noted a considerable improvement in BP within the group of patients with the highest risk (defined as a BP ≥ 160/100) and a decrease in disparities within this group. The quality collaborative identified five key lessons to help guide future community initiatives: (a) anticipate a plateauing of response; (b) distinguish the needs of disparate populations and create subpopulation‐specific strategies to address and reduce disparities; (c) recognize the variation across low SES practices; (d) remain open to the refinement of outcome measures; and (e) continually seek best practices and barriers to success. Overall, a large community‐wide QI initiative, involving multiple different stakeholders, was associated with improvements in BP control and modest reductions in some targeted disparities.

1. BACKGROUND

Hypertension affects approximately 30% of adults in the United States and is one of the largest contributors to morbidity and mortality.1, 2, 3, 4 Despite the availability of effective treatments, only about half of patients with hypertension are effectively managed.1 Quality improvement (QI) initiatives in closed health care systems and academic medical centers benefit from a common structure and governance guiding the intervention. However, methods to implement a broad community‐wide QI intervention across a region involving multiple health care systems are less well defined.

In 2009, a unique community‐wide collaboration was formed in Upstate New York to address the care of county residents with hypertension. The collaborative was led by the Greater Rochester Chamber of Commerce and Common Ground Health. The explicit goals of the collaborative were to (a) increase the control rate of those with high blood pressure and (b) reduce existing disparities in blood pressure control.5

While similar multifaceted approaches have been effectively implemented in closed health care systems, such as Kaiser Permanente,6, 7, 8 community‐wide projects, bringing together more diverse, complex populations, have been both less common and less successful in the United States.8 Large community‐wide initiatives present substantial logistic challenges related to organization, governance, coordination, data collection and management, reporting of actionable data, and outreach. Understanding how to evaluate and improve the implementation of large community‐wide QI initiatives is essential since most of the population is cared for in communities outside of closed health care systems.

To contribute to our understanding of complex community‐wide QI initiatives, we examined a large multifaceted initiative across 226 843 patients from 198 practices encompassing nine counties across the Finger Lakes Region of Upstate New York. The broad community‐wide QI collaborative was designed to engage as many practices as possible, but was not designed to assess the degree to which specific interventions were responsible for the results. Rather, the collaborative focused on fostering community engagement, providing population‐level comparative data, and supporting practice‐based quality improvement. The collaborative QI initiative aimed to (a) improve overall hypertension control rates across the community, (b) reduce disparities based on patient race/ethnicity and socioeconomic status (SES), and (c) share best practices and barriers identified during this project to guide future community interventions.

2. METHODS

2.1. Description of community collaborative

The initiation and early years of this project have been previously described in detail.5 To summarize, the intervention was designed and implemented using Wagner's chronic disease model,9, 10, 11 and Deci and Ryan's self‐determination theory of motivation.12 Self‐determination theory (SDT) is a general theory of human motivation that assumes people have psychological needs for autonomy, competence, and relatedness. Autonomy encourages participants to create their own interventions to the desired outcome. Competence involves supporting the development of mastery toward the desired outcome. Relatedness acknowledges the desire for interpersonal attachments as a fundamental component of human motivation. For example, we engaged the Common Ground Health African American Coalition to help design and guide interventions in the African American community that would be most successful in achieving blood pressure control. Using this theory, the community‐wide hypertension QI intervention was designed to continuously support and intrinsically motivate both practitioners and patients to improve hypertension control rates.

The collaborative included over 30 community organizations and multiple local health care systems in Upstate and Western New York.5, 13 The project's leaders employed a multifaceted intervention focusing on (a) collecting and sharing de‐identified population‐level high blood pressure data to participating systems and practices, (b) focusing community engagement most aggressively on underserved and low socioeconomic status (SES) patients, and (c) practice‐based quality improvement directed at achieving the initiative's goals of improving hypertension control while reducing socioeconomic and racial disparities.

2.2. Practice‐based quality improvement team

To assist in practice‐based quality improvement, the initiative developed and implemented a program to provide resources to participating practices. That practice‐based quality improvement team, referred to as Performance Improvement Consultants (PICs), included two internists, four family physicians, and two clinical pharmacists. The team members underwent 4 days of training that included instruction in academic detailing through a contract with the Alosa Foundation and communication skills training by a fellow of the Academy of Communication in Healthcare (HB). Training included knowledge and skills in promoting internal motivation, encouraging practice autonomy, providing techniques in successful communication and time management, and the process of academic detailing. Successful communication focused on carefully listening to practitioners and asking reflective questions that encouraged individualized practice initiatives to improve control and reduce disparities (autonomy support). PICs were assigned practices to visit and to review their practice hypertension data reports. The PICs met as a group by phone (monthly at first then bimonthly) to discuss site results, share successes, and identify best practices. Clinicians’ needs were supported by providing, organizing, and delivering data in a systematic yet nonjudgmental approach. The quality improvement component with PICs was evaluated based on thematic analysis of reports by the project's best practice committee about practice visit experiences.

2.3. Development of registries

Beginning in 2010, a high blood pressure registry was built using data reported from local health care systems and independent practices twice yearly. The registry began with an initial group of 63 practices that volunteered to submit de‐identified clinical data. As of December 31, 2017, the number of practices voluntarily submitting de‐identified data had increased to 198 practices spread over the nine county Finger Lakes region, which includes Monroe, Ontario, Chemung, Livingston, Schuyler, Seneca, Steuben, Wayne, and Yates counties in New York State. The health departments in these counties, with the collaboration of local primary care practices, also voluntarily contributed their data to the registry.

Upon receipt of the practice data by Excel file, Common Ground data analysts inspected the data and requested corrections or additional information where appropriate. Data from 13 different electronic health records (EHRs) were incorporated into the dataset. Because the data were derived from multiple EHRs, the data were standardized to ensure that the data fields aligned. Race and ethnicity were self‐identified. Socioeconomic status was assigned based on the patient's reported residence zip code. Once analysis was completed, the BP registry results were reported back to health system and/or practices biannually. Practices and systems could view detailed results for their system or practice and blinded community results based on the remaining community practices. An example of a poster provided each practice to provide peer comparison and promote quality improvement data is shown in Appendix S1.

2.4. Definitions

We examined hypertension based on a modification of the 2015 HEDIS standards to account for community attribution. The HEDIS measure defined adequate blood pressure (BP) control as a BP <140/90 in hypertensive individuals aged 18‐59 years and a BP <150/90 in individuals 60 years of age and over during the measurement period. We employed a 3‐year look back in the electronic health record problem list to define the population (denominator) with high blood pressure. The numerator included the number of eligible patients with their most recent blood pressure taken in the prior 12 months, under control. In traditional health plan settings, HEDIS reporting is based on patients being continuously enrolled in the designated health plan. Because of limitations in attribution at a community‐level, continuous enrollment in an insurance product was not viewed as an appropriate requirement for inclusion in the community‐wide registry.

Within our community initiative, we tracked patients diagnosed with hypertension during the prior 3 years but without a measured blood pressure in the preceding 12 months and categorized them as “no reads.” This group was reported separately from those with available blood pressures readings within the past year. Those de‐identified patients with missing readings were reported back to the practices but were not included in the calculation of the control rate (see Appendix S1). This is a modification from the traditional HEDIS definition to account for limitations of attribution in community interventions. Both definitions were reported to practices.

We also tracked the percent of patients with the most recent reported blood pressure ≥160/100. This measure developed from consensus among all stakeholders and grew out of concern for those patients most at risk of morbidity and mortality from uncontrolled hypertension.

2.5. Analyses

Our results are primarily descriptive analyses of the submitted hypertension registry data. Participating practices with more than 100 patients diagnosed with hypertension in the past 3 years were included in the report. Reported blood pressure data were stratified and examined by county of residence, gender, self‐identified race/ethnicity, and SES. Co‐morbid conditions were reported based on the patient's problem list in their electronic medical record. Results for each of the variables were trended yearly over the duration of the project. This project was approved by the University of Rochester IRB.

3. RESULTS

Between 2010 and 2017, the hypertension registry increased from 59 400 patients to 226 843 patients (Table 1). Based on an expected rate of hypertension of 35% for the population 18 years of age or over, a majority (64.5%) of the estimated 351 472 hypertensive patients in the nine‐county Upstate New York region were included in the registry. The registry was composed of 52% women. As of June 2017, the racial/ethnic composition of the registry included 78% white, 14% black, 5% Hispanic, and 3% other races.

Table 1.

Hypertension registry

Registry date	Registry details			Control rates (%)
	Expected number of hypertensive patientsa	Hypertensive patients in registry	Percent in registry (%)	HEDISb	Patients with BP in last yearc
Dec 2010	202 370	59 400	29.4	61.9	68.2
Dec 2011	203 700	88 900	43.6	61.4	69.2
Dec 2012	205 065	104 323	50.9	65.4	71.9
June 2013	205 447	103 201	50.2	65.8	75.1
Dec 2013	205 954	114 198	55.5	71.8	79.9
June 2014	205 989	114 568	55.6	69.6	75.2
Dec 2014d	351 374	155 800	44.3	69.8	76.8
June 2015	351 686	170 242	48.4	69.3	77.0
Dec 2015	351 909	199 072	56.6	59.4	64.7
June 2016	352 392	207 381	58.9	68.3	77.1
Dec 2016	352 542	219 581	62.3	68.2	76.8
June 2017	351 472	226 843	64.5	69.5	78.2

^aThe expected number of hypertensive patients was calculated using 35% of the total population 18+ year of age residing within the 9‐County Finger Lakes region.

^bThe 2015 HEDIS measure defined control as a BP <140/90 in hypertensive individuals aged 18‐59 y and a BP <150/90 in individual 60 y of age. The absence of an available blood pressure measure within the preceding 12 mo was counted as an uncontrolled value.

^cControl rate based on patients with a BP reading within the preceding 12 mo.

^dAdditional counties added.

Table 1 depicts the change in hypertension control rate between 2010 and June 2017. Overall, hypertension control rates improved from 61.9% in 2011 to 69.5% in 2016. Overall, the majority of the improvement in blood pressure control was observed within the first few years, with improvements in blood pressure control rates beginning to plateau after several years.

Figure 1 depicts changes in hypertension control based on race, ethnicity, and socioeconomic status. Improvements in control were observed across all years in white patients (73.7%‐81.5%). Improvements were more modest in the black population (58.8%‐64.7%). Overall, absolute differences in hypertension control increased slightly between white and black patients (14.9%‐16.8%) and between white non‐Hispanic and Hispanic patients (11.0%‐11.3%). Improvements were seen across all socioeconomic groups, but the largest improvements were observed in the highest socioeconomic group. Disparities between the lowest and highest socioeconomic groups increased from an absolute 7.3% difference in BP control in 2010 to a 9.3% difference in 2017.

Figure 1.

Hypertension control rate by (A) race/ethnicity, and (B) socioeconomic status

Between 2010 and 2017, among patients with diagnosed hypertension, those with a BP ≥160/100 decreased from 8.3% to 4.8%, a relative decrease of 42%. Those with a BP ≥160/100 decreased from 7.4% to 3.9% among white patients, and from 13.4% to 9.3% among black patients (Figure 2). Disparities between white and black patients decreased from an absolute difference of 6.0%‐5.4%. Similar differences were seen by SES. Those with a BP ≥160/100 decreased from 7.1% to 3.6% among high SES practices and from 10.9% to 6.0% among low SES practices (Figure 2), resulting in an absolute decrease in disparities from 3.8% to 2.4%.

Figure 2.

Changes in BP ≥160/100 by (A) race/ethnicity and (B) socioeconomic status

The percent of hypertensive patients controlled to goal correlated with practice socioeconomic status. However, the range of performance of practices with the highest percentage of low socioeconomic patients varied substantially (Figure 3). Several of the lowest socioeconomic practices were among the higher performing practices in the 9‐county community.

Figure 3.

Hypertension control by practices—highlighting the ten practices with the highest percentage of low SES

Within this community initiative, missing reads (carrying the diagnosis of hypertension but without a reported BP in the most recent 12 months) were tracked separately from control rates. Between 2010 and 2017, the percent of missing reads increased from 9.4% to 11.2%. Among black patients, the percent of missing reads increased from 8.1% to 16.2%, while the percent for white patients increased from 7.6% to 9.8%. Among low socioeconomic practices, missing reads increased from 8.6% to 14.4%. Among moderate socioeconomic status, missing reads increased from 9.7% to 10.3%, while they decreased in high socioeconomic practices from 9.7% to 8.6%.

4. DISCUSSION

Large community‐wide initiatives can play an important role in improving blood pressure control across many regions of the United States, especially for patients without access to quality improvement efforts through their local health delivery systems. Our nine‐county quality improvement intervention, coordinated across multiple health systems, was associated with substantial improvements in blood pressure control rates. In particular, we noted a considerable improvement in BP within the group of patients with the highest risk (defined as a BP ≥160/100) and a modest decrease in disparities within this group.

This broad quality collaborative was associated with multiple improvements, but was not designed to evaluate causality. Through the regional collaborative, we focused our analysis on the lessons learned from a multifaceted community‐based QI intervention to improve blood pressure control and reduce disparities in outcomes. During this effort, we identified five key lessons that may support future community‐wide interventions to improve blood pressure control: (a) anticipate a plateauing of response and be prepared to develop new, creative approaches as improvements plateau; (b) distinguish the needs of disparate populations and create subpopulation specific strategies to address and reduce disparities; (c) recognize the variation across low SES practices and identify best practices; (d) remain open to the refinement of outcome measures; and (e) continually seek best practices and barriers to success.

First, while we observed substantial improvements in blood pressure control, the rate of improvement decreased after the first 3 years. This plateau effect was largely driven by slowed improvement in minority populations. Plateau affects have also been observed in other large interventions performed in closed‐systems and academic programs. In 2009, the Kaiser Permanente Northern California (KPNC) hypertension program implemented a multifaceted initiative resulting in brisk improvements in blood pressure control rates over the first several years, but subsequent leveling of improvements.6 Kaiser Permanente Southern California (KPSC) saw similar improvements in hypertension control rates from 54% in 2004 to 84% in 2010, but then a relative plateau effect.7 Analogous findings were observed in a multidisciplinary strategy in an underserved urban practice.13 Given this predictable plateau effect, anticipating a leveling‐off of intervention effects in community initiatives permits the exploration of new creative approaches.

Second, although substantial improvements in overall hypertension control (<140/90) were achieved, overall disparities based on SES and race/ethnicity were not effectively reduced. While improvements were seen across all races and socioeconomic groups, the largest improvements were observed for white patients and those in the highest socioeconomic group. Overall control rates among black and Hispanic groups improved initially, but then plateaued, leading to increasing community‐wide disparities by race. However, among those targeted for being most at risk of morbidity from hypertension (BP ≥160/100), disparities were slightly reduced. This represents an important clinical improvement and highlights the need to monitor clinically applicable goals beyond nationally defined measures. This targeted reduction in hypertension disparity among high‐risk hypertensive patients is unique and warrants further investigation.

Prior studies have reported similar challenges reducing observed racial and socioeconomic disparities.14, 15, 16, 17, 18 Although multiple interventions exist to improve BP control in minority populations, there is limited evidence that these interventions are effective in reducing overall disparities.19 For instance, one group previously reported a multidisciplinary strategy that resulted in significant improvement in BP control in black patients, but widened disparities when compared to white patients.13 Interventions based on physician‐pharmacist collaborations have similarly reduced disparities by race, ethnicity, and socioeconomic status, but patients from typically disadvantaged backgrounds generally experienced less benefit in BP control than white patients.

Third, we observed substantial variation by practice across the range of SES, with many low SES practices performing remarkably well. This observation reinforced the importance of recognizing the variation that exists between practices with similar SES patient populations. While patient SES is important, it is only one of many potential barriers to achieving successful control rates. In some cases, practices caring for low SES populations outperformed practices with moderate and high SES populations. Conducting subgroup analysis and identifying the variation that did exist between similar SES practices provided an important opportunity to identify best practices and uncover effective interventions to reduce disparities. This sharing of best practices is a powerful contributor to improving outcomes in underperforming practices.

Fourth, as community QI interventions mature, project goals should drive the continuous refinement of outcome measures. Across our community project, we identified the need to track high‐risk patients. We saw substantial improvements in reducing BP for those with a BP ≥160/100, and a modest reduction in disparity by race and SES. A focus on those with the highest blood pressures offers the greatest potential to improve morbidity and mortality. Hence, tracking improvement in the proportion of those with a BP ≥160/100 became a valuable partner measure to the more standard BP control rates in our community‐wide interventions.

Finally, within any community‐wide QI project, best practices and barriers to success must be continually sought. Within our broad community QI initiative, physicians were excited to cross system boundaries to discuss improvements and best practices in care throughout the entire 9‐county population. Encouraging a community‐wide approach provided a sense of “relatedness” that propelled implementation and aligned goals among stakeholders. The impetus of having our project initiated by the business community, especially with their focus on reducing disparities, encouraged others in the community to become involved. We also found it valuable to have a regional health collaborative serve as a neutral convener and collector/disseminator of data and best practices. This structure facilitated the sharing of peer data and best practices in a nonjudgmental fashion.

This large community‐based QI project has several important limitations. First, by design, our quality improvement approach was not designed to assess causality and was not able to randomize practices or control for the external variables during the study period. Furthermore, using a multifaceted intervention, our methodology did not allow us to determine which component(s) of the intervention were most responsible for the improvements observed. Second, we found practice and system leadership had varied levels of engagement to share results with staff and practices. The effectiveness of academic detailing was variable, with system support for academic detailing considered not only valuable, but essential. Third, national guidelines for hypertension treatment goals generated significant debate over the course of this QI project. Based on consensus among stakeholders, we used reported a common metric for tracking purposes. Fourth, teaching of proper BP measurement technique may have resulted in improved BP readings simply due to increased accuracy of measurement.20 Finally, data collection was voluntary and de‐identified, preventing longitudinal tracking of patients. As a result, the data consists of a series of cross‐sectional views of community‐wide data.

5. CONCLUSIONS

We found that a large community‐wide QI initiative, involving multiple different stakeholders, was associated with improvements in BP control and a modest reduction in some disparities. Future large community QI initiates should (a) anticipate a plateauing of response; (b) distinguish the needs of disparate populations and create subpopulation‐specific strategies to address and reduce disparities; (c) recognize the variation across low SES practices and capitalize on the opportunity to identify best practices; (d) remain open to the refinement of outcome measures; and (e) continually seek best practices and barriers to success. Despite the challenges, large community‐wide QI initiatives can play an important role in improving BP control, reducing targeted health disparities, and sharing of best practices across many areas of the country.

CONFLICT OF INTEREST

The authors report no conflicts of interest.

Supporting information

 

Fortuna RJ, Rocco TA, Freeman J, et al. A community‐wide quality improvement initiative to improve hypertension control and reduce disparities. J Clin Hypertens. 2019;21:196–203. 10.1111/jch.13469