miércoles, 31 de julio de 2019

A Community-Wide Collaboration to Reduce Cardiovascular Disease Risk: The Hearts of Sonoma County Initiative

A Community-Wide Collaboration to Reduce Cardiovascular Disease Risk: The Hearts of Sonoma County Initiative



PCD logo

A Community-Wide Collaboration to Reduce Cardiovascular Disease Risk: The Hearts of Sonoma County Initiative

Allen Cheadle, PhD1; Michelle Rosaschi, MPH2; Dolores Burden, MSN, RN3; Monica Ferguson, MD, MSHP4; Bo Greaves, MD5; Lori Houston6; Jennifer McClendon, MPH7; Jerome Minkoff, MD2; Maggie Jones, MPH1; Pam Schwartz, MPH8; Jean Nudelman, MPH8; Mary Maddux-Gonzalez, MD, MPH2 (View author affiliations)

Suggested citation for this article: Cheadle A, Rosaschi M, Burden D, Ferguson M, Greaves B, Houston L, et al. A Community-Wide Collaboration to Reduce Cardiovascular Disease Risk: The Hearts of Sonoma County Initiative. Prev Chronic Dis 2019;16:180596. DOI: http://dx.doi.org/10.5888/pcd16.180596external icon.
PEER REVIEWED
Summary
What is already known on this topic?
Clinical and community collaborations are foundational to primary care transformation efforts, but it has proved challenging to build sustainable, effective collaborations.
What is added by this report?
Several lessons from the experience of the successful Hearts of Sonoma County (HSC) collaborative, including 1) start small and focused to build trust among participants and demonstrate value, 2) work within the framework of a larger effort, and 3) providing long-term, open-ended backbone support.
What are the implications for public health practice?
The HSC experience may provide a roadmap for other, similar efforts.

Abstract

Purpose and Objectives
Collaboration across multiple sectors is needed to bring about health system transformation, but creating effective and sustainable collaboratives is challenging. We describe outcomes and lessons learned from the Hearts of Sonoma County (HSC) initiative, a successful multi-sector collaborative effort to reduce cardiovascular disease (CVD) risk in Sonoma County, California.
Intervention Approach
HSC works in both clinical systems and communities to reduce CVD risk. The initiative grew out of a longer-term county-wide collaborative effort known as Health Action. The clinical component involves activating primary care providers around management of CVD risk factors; community activities include community health workers conducting blood pressure screenings and a local heart disease prevention campaign.
Evaluation Methods
The impact of the clinical improvement efforts was tracked using blood pressure data from the 4 health systems participating in HSC. Descriptive information on the community-engagement efforts was obtained from program records. Lessons learned in developing and maintaining the collaborative were gathered through document review and interviews with key informants.
Results
Favorable trends were seen in blood pressure control among patients with hypertension in the participating health systems: patients with controlled blood pressure increased from 58% in 2014 to 67% in 2016 (P < .001). Between 2017 and 2019, the community engagement effort conducted 99 outreach events, reaching 1,751 individuals, and conducted 1,729 blood pressure screenings, with 441 individuals referred to clinical providers for follow-up care. HSC scored highly on 6 essential elements of an effective coalition and achieved a degree of sustainability that has eluded many other collaboratives.
Implications for Public Health
Factors contributing to the success of HSC include 1) starting small and focused to build trust among participants and demonstrate value, 2) working within the framework of a larger effort, and 3) providing long-term, open-ended backbone support.

Introduction

Improving the health of a population requires a multi-faceted approach that includes both community and clinical strategies (1). Implementing these clinic/community strategies successfully often requires multi-sectoral collaborations that bring together a broader range of organizations and institutions than are part of typical public health coalitions (2). For example, multi-sector Accountable Communities of Health have been part of many State Innovation Model (3) health improvement projects that are attempting to bring together a range of partners to work on health system transformation (4).
Although effective collaboration is needed to bring about health system transformation, doing it well has proved challenging. In a recent study by Siegel et al (5) of 145 health system improvement collaboratives that had a reputation for being mature and effective, as few as 10 were judged to be mature enough to make true progress toward supporting a transformed health system. Some of the challenges that have limited the effectiveness of previous public health–oriented coalitions (6) are accentuated in these newer, larger collaboratives encompassing more sectors (ie, reaching agreement on goals, approaches, and steps to action among varied organizations with competing organizational objectives).
One way of overcoming these challenges is to learn from successful collaborative efforts. Substantial literature on what makes a successful coalition exists (2,7,8), but we are aware of few published examples in which multi-sector collaborative efforts have been sustained over an extended period, and long-term sustainability is critical for creating a transformed, integrated health care system.
In this article, we describe the Hearts of Sonoma County (HSC) initiative, a county-wide, multi-sector collaborative effort to reduce cardiovascular disease (CVD) risk in Sonoma County, a medium-sized county in northern California. HSC grew out of Health Action, a larger multi-sector effort that has existed for more than 10 years. HSC is being evaluated using 1) a process evaluation to capture milestones in initiative development and factors associated with success, and 2) an outcome evaluation documenting changes in CVD outcomes (eg, blood pressure control) using pooled county-level provider data. This article describes the initiative and outcomes to date and identifies lessons learned and recommendations that may be useful for other, similar initiatives.

Purpose and Objectives

Sonoma County is the northwestern-most county in the 9-county San Francisco Bay Area region, with a population of 502,000 in in 2016 (9). Its county seat and largest city is Santa Rosa. The county is near the average for California in terms of income/poverty: the median household income of $61,000 is below the $67,700 statewide median, but the federal poverty rate is lower than the state as a whole — 11.2% versus 14.3% (9). The largest racial/ethnic groups are white (66%) and Hispanic (25%) (9). From 2015 through 2017, 31% of adults in Sonoma County had ever been diagnosed with high blood pressure, and 7% had ever been diagnosed with heart disease (10). Health care providers include Kaiser Permanente, St. Joseph Health, Sutter Health, and several federally qualified community health centers.
In 2007 the Sonoma County Department of Health Services, which includes the public health department, approached the county board of supervisors with a proposal to form a collaborative to address social determinants of health and health equity. Health care was at the top of the county agenda because of a public hospital closing, and there was a growing recognition that health involves more than just health care. Therefore, the board of supervisors adopted the proposal and formed the Health Action collaborative.
Health Action brought together organizations in education, business, health care, labor, and public health to focus on social determinants of health and health equity and justice. Three focus areas were chosen: health care, education, and economic wellness, and standing committees created in each area. Education, known as Cradle To Career, focused on educational and social strategies to support children and youth reaching their fullest potential at every stage of life, such as coordinating a campaign focused on school attendance to address the effect of student absenteeism and working to develop agreed-upon local standards for college and career readiness in Sonoma County. Economic wellness focused on addressing local economic conditions and issues to support families becoming better able to make ends meet, such as affordable housing and helping low-income families take advantage of the earned income tax credit. The health care committee (the Committee for Healthcare Improvement) focused initially on primary care, addressing a shortage of primary care physicians and working to increase Patient-Centered Medical Home (PCMH) capacity. Over time, the committee recognized that a broadened focus was needed and shifted their attention first to end-of-life care and then, after a community health needs assessment, to reduce CVD risk. The initial collaborations around assessing local primary care capacity and PCMH were critical in establishing trust across health care entities in a competitive market. This trust extended to sharing workforce data.
HSC was formed by the committee in 2014 as a result of the new focus on CVD risk reduction. Drawing from the Centers for Disease Control and Prevention’s Million Hearts campaign (11) and work being done by Kaiser Permanente to implement an effective algorithm for reducing CVD risk (12), individual provider organizations participating in the committee began implementing improved practices in their clinics in 2014. The fact that the community health centers and Kaiser Permanente groups served more than half of the population and were both engaged in cardiovascular health initiatives was a major factor in deciding to focus on CVD across health systems.
In 2016 the county applied for and received a California Accountable Communities of Health Initiative (CACHI) grant to transform and operationalize the work of Health Action by piloting accountable communities of health principles to address CVD with HSC through health care system and community-based interventions. Community outreach, education, and engagement efforts coalesced in the It’s Up to Us campaign, a partnership between the United Way of the Wine Country and the Northern California Center for Well-Being (Center for Well-Being), which was launched in 2017. It’s Up to Us has 3 primary goals: 1) educate the community about CVD risk factors, 2) conduct community-based blood pressure screenings, and 3) link high-risk individuals to primary care to reduce risk of heart attacks and strokes.

Intervention Approach

This section describes the clinical care and community engagement components of HSC, as well as the structure, and operations of the HSC collaborative. Figure 1 shows the structure of HSC and its position within Health Action.
Sonoma County’s Health Action Collaborative, formed in 2007, focused on activities to improve health systems, education, and economic well-being among residents. In 2014, Hearts of Sonoma County was established as part of the health system improvement efforts building on Kaiser Permanente’s PHASE program and including a focus on clinical quality improvement for cardiovascular disease in 4 Sonoma County health systems. In 2017, the community engagement workgroup was formed to deploy community health workers to conduct blood pressure screenings in the community, improve community and clinical linkages, and build and launch a communitywide media campaign called It’s Up to Us.
Figure 1.
Sonoma County Health Action Collaborative, overall structure and health care activities, the Hearts of Sonoma County Initiative, Sonoma County, California. Abbreviations: CVD, cardiovascular disease; PHASE, Preventing Heart Attacks and Strokes Everyday. [A text description of this figure is available.]

Improving clinical care

The goal of the clinical care effort is to activate primary care providers around evidence-based interventions, including improved identification and management of hypertension and more consistent screening for other CVD risk factors, coupled with more robust smoking cessation support. With the funding from CACHI, the scope of the clinical effort was expanded to include secondary prevention modeled on the Kaiser Permanente Preventing Heart Attacks and Strokes Everyday (PHASE) initiative, which encompasses standardized, comprehensive care management and cardio-protective medications for people with CVD and those who have had a heart attack or stroke. The PHASE strategies being implemented include adoption of evidence-based clinical guidelines and standardized procedures for registered nurses; capacity building for population health management; provider/clinician/medical staff education and training; primary care workflow improvements; and extended team-based care. The population health framework introduced through PHASE is used by each entity, wherein the population at risk is identified and stratified and interventions and results are tracked on the population as a whole and by individual providers and, in some cases, by care team staff. This approach is effective in influencing clinical practice and improving outcomes.

Community engagement

The second part of the HSC strategy was to engage the community around CVD risk reduction and help link efforts in the clinical domain with interventions across the community, policy, systems, and environmental domains. Activities have included convening a new Community Engagement workgroup, training community health workers (CHWs) to conduct community-based education and blood pressure screenings, and convening a media workgroup to partner in a localized heart disease prevention media campaign. The following provide a brief summary of those activities.
Community Engagement workgroup. A Community Engagement workgroup facilitated by Center for Well-Being staff planned and implemented the campaign, including listening sessions (15 listening sessions, engaging 170 participants) to ensure the subsequent campaign spoke to populations at greatest risk for heart disease. Community was integral in shaping campaign messaging to shift their perception of risk and motivate them to take action.
Community-based education and blood pressure screenings. The Center for Well-Being developed a training module for Promotores de Salud/CHWs to be trained in blood pressure screening, identifying risk factors and warning signs, and learning what to do when they encounter residents with blood pressure outside the normal range, including how to link residents to care. Once trained, the Center for Well-Being leveraged existing partnerships to begin outreach in nontraditional settings. The Center for Well-Being developed a protocol to contact community members found to have high (140–169 mm Hg systolic or 90–99 mm Hg diastolic) or very high (≥170 mm Hg systolic or ≥100 mm Hg diastolic) blood pressure readings a few days after the screening to learn if they followed through with scheduling an appointment with their medical provider or contacting a clinic if they were out of care. The Center for Well-Being made arrangements with one community health center site in Santa Rosa, enabling CHWs to use a direct phone line to schedule medical appointments for people as soon as possible. Center for Well-Being staff links residents to additional support services, including health insurance assistance and behavioral change classes to prevent heart disease.
Localized heart disease prevention campaign. Listening session results were developed into 3 campaign concepts, further tested with residents from our target populations and revised based on their feedback. The goal of the It’s Up to Us campaign, launched in August 2017, is community empowerment, encouraging people to take ownership of their health, with a first action of checking their blood pressure. Images, taglines, and the corresponding website (CheckYourBP.org) provide a cohesive media and messaging campaign. Collateral material such as the blood pressure cards and posters were designed and distributed to health care partners.

HSC collaborative identity and functioning

The HSC collaborative has evolved over time, from starting as an initiative of the Committee for Healthcare Improvement (a committee of the larger Health Action collaborative) to piloting how Health Action will function as an Accountable Community for Health. Table 1 lists the HSC partner organizations and their role on the project, as defined by their membership in workgroups and committees. Table 2 lists these same organizations and shows which parts of the organization are represented regularly at meetings (eg, clinical representatives, organizational leadership, administration/program managers).
The Sonoma County Department of Health Services provides backbone support for HSC, and the It’s Up to Us community engagement work is backboned by the Center for Well-Being with funding from United Way. The Department of Health Services provides approximately 20 to 30 hours per week or 75% of a full-time position to coordinate HSC associated meetings. On average, the Center for Well-Being estimates 16 hours per week on work associated with HSC. However during the height of the campaign initiation (2017–2018), it was closer to 25 hours per week on campaign planning, coordination, and evaluation.
In 2018, HSC members assumed oversight of the CACHI Portfolio of Interventions, which includes management of mutually reinforcing clinical and community-based strategies that support the overall goal of improving cardiovascular health throughout Sonoma County. The clinical improvement and community engagement tracks operate independently but inform each other’s activities, with several operational connections now, including a Clinical–Community Linkages workgroup. For example, patients identified with high blood pressure at the community screenings are linked to health care providers who are represented on the Committee for Healthcare Improvement.

Evaluation Methods

The evaluation design was largely retrospective and descriptive, documenting the development of the HSC initiative and its impact to the extent possible, given that this was not designed as a prospective research/evaluation study. The evaluation of HSC includes 1) documenting clinical care improvement efforts around CVD and the impact those changes have had on CVD outcomes; 2) capturing diverse community engagement efforts and their impact; and 3) working to understand the factors associated with the success of the collaborative, including challenges and lessons learned. The following is a brief description of the methods used in each of these three areas.
The long-term evaluation of the HSC clinical work is focused on tracking county-level CVD outcomes. HSC representatives recognized early the importance of sharing data, both for the continuous improvement and to document county-level outcomes. HSC clinical partner organizations signed a multi-party data sharing/nondisclosure agreement that enables them to report and aggregate data related to CVD risk factor interventions. To date, reporting partners have shared their Healthcare Effectiveness Data and Information Set (HEDIS) blood pressure control data annually to create a countywide report card that benchmarks and tracks how the local health system is doing overall with screening, diagnosing, and managing hypertension, and to track collective improvement. As of September 2017, 4 major primary care provider organizations in Sonoma County contributed 2014, 2015, and 2016 numerator and denominator totals for the 3 age groups and populations defined by the 2015 HEDIS Controlling Blood Pressure Technical Specification (control defined as blood pressure <140/90 mm Hg.). These organizations also reported their total number of adult patients for each of these years, which collectively represent about 57% of Sonoma County’s overall adult population.
The evaluation of the community engagement efforts — blood pressure screening, CHW outreach, It’s Up to Us media campaign — is a descriptive, process evaluation. Information gathered, both in real time through progress reporting and retrospectively, includes the number of screening events held, number of people screened, number of people with high or very high blood pressure, and number of individuals connected with primary care. There are several community engagement outcomes in which measures are being developed (eg, increased public awareness of CVD risk factors and community resources to help address them) (Figure 2).
Cardiovascular disease portfolio of interventions logic model, the Hearts of Sonoma County Initiative, Sonoma County, California. Abbreviations: CHW, community health worker; CVD, cardiovascular disease; ED, emergency department; PHASE, Preventing Heart Attacks and Strokes Everyday.
Figure 2.
Cardiovascular disease portfolio of interventions logic model, the Hearts of Sonoma County Initiative, Sonoma County, California. Abbreviations: CHW, community health worker; CVD, cardiovascular disease; ED, emergency department; PHASE, Preventing Heart Attacks and Strokes Everyday. [A text description of this figure is available.]
Details about the evolution of the collaborative structure and process, as well as successes, challenges, and lessons learned, were gathered through document review and interviews with 8 key participants. The data gathering was organized using a framework developed by the Center for Community Health and Evaluation (CCHE) to track key elements in coalition development (Figure 3).
Essential elements needed for effective collaboration, the Hearts of Sonoma County Initiative, Sonoma County, California. Abbreviation: QI, quality improvement.
Figure 3.
Essential elements needed for effective collaboration, the Hearts of Sonoma County Initiative, Sonoma County, California. Abbreviation: QI, quality improvement. [A text description of this figureis available.]
 

Results

Improving clinical care

Four major health systems have participated in the HSC work around implementing the PHASE protocol and other clinic-level interventions. Kaiser Permanente developed the protocol and has implemented it successfully in their 4 Sonoma County clinics. Other health systems have focused initially on pilot implementation in selected clinics or pods within clinics (eg, a large St. Joseph Health Medical Group practice in Santa Rosa). The community health centers began implementing PHASE in 2011 through a Kaiser Permanente Northern California Community Benefits program grant to the Redwood Community Health Coalition. Progress to date has included identification of nearly 25,000 patients with a diagnosis of hypertension, diabetes and/or atherosclerotic CVD across 22 clinic sites. Since baseline of March 2017, community health centers have demonstrated aggregate performance improvements in lifestyle measures including body mass index, tobacco, and depression screenings with documented follow-up plans as well as on prescription measures, including angiotensin converting enzyme/angiotensin receptor blocker and statin prescription rates, among patients aged 55 through 75 with diabetes.
County-level trends in CVD outcomes assessed by using the shared data from the 4 participating health systems have been encouraging. Figure 4 shows trends in blood pressure control for ages 18–59 years; results were similar in other age groupings. All of the year-to-year changes were significant (P < .001), increasing from 58% of participants who had their blood pressure controlled in 2014 to 67% in 2016. HEDIS benchmark trends for a comparable measure essentially did not change during that same period.
Percentage of hypertension patients aged 18 to 59 years with controlled blood pressure, the Hearts of Sonoma County Initiative, Sonoma County, California.
Figure 4.
Percentage of hypertension patients aged 18 to 59 years with controlled blood pressure, the Hearts of Sonoma County Initiative, Sonoma County, California. [A text description of this figure is available.]

Community engagement

Between 2017 and 2019, the community engagement effort has conducted 99 outreach events, reaching 1,751 individuals, and conducted 1,729 blood pressure screenings. A total of 441 of the people screened were found to have high or very high blood pressure readings and were contacted for follow-up by bilingual Center for Well-Being staff to evaluate the effectiveness of the screening and to motivate them to connect with their doctor or a referred provider. Partners such as St. Joseph Health have integrated It’s Up to Us into their community-based screenings, with staff adapting the campaign to meet the needs of the populations they serve. Future outreach opportunities being explored include senior centers, school parent groups, and grocery stores located in low-income neighborhoods. Work is underway to launch a blood pressure clinic with Santa Rosa Community Health’s Fiesta site, piloting a faster point of entry to care for residents out of care found to have high or very high blood pressure readings. The Center for Well-Being and Santa Rosa Community Health are looking to expand the pilot to other clinic sites in the future.

Collaborative impact, sustainability, and success factors

The HSC collaborative provides the overall structure and support for the clinical and community activities, including forming relationships for interventions linking clinics and communities. The collaborative has achieved sustainability, a result that eludes many collaboratives. Understanding why it has been successful may provide lessons for other similar collaboratives.
Figure 3 summarizes the 6 elements in the coalition model that were used to enumerate and understand the success of the collaborative. Quotes were drawn from interviews with key HSC stakeholders, which included someone from each of the major participating organizations. HSC successfully fulfilled all 6 of the essential elements in the model. The shared purpose of reducing CVD risk through clinical and community approaches used was agreed to by all, and the language was revisited and updated as the initiative continued. A component of success was a strategic approach in aligning the existing goals, interests, and requirements of individual primary care organizations with the shared communitywide goal of improving CVD health. For example, all of the primary care organizations are evaluated on HEDIS or HEDIS-like measurements. Measurements were developed that would most closely match the specifications of required performance metrics to take advantage of data the organizations were already collecting. This approach and alignment meant that improvements resulting from the collaborative work of the HSC initiative translated to improved outcomes on performance measures that are important to the individual entities participating in the initiative, which in turn supported ongoing investment in the process.
The essential people and organizations were generally present within HSC although several informants noted the absence of community residents to provide a consumer perspective: “We’re struggling with having resident involvement . . . [for example] neighborhood organizations.” Community is at the center of the coalition model to emphasize that the efforts are ultimately designed to improve the health of community residents, who should therefore be engaged to define what matters to them in the way of health and how their health can be improved. The It’s Up to Us campaign is working to increase the level of community engagement.
Effective leadership of HSC has generally been present in the form of a rotating group of clinical leaders from the different health systems. Several informants noted that leadership has come from many of the participating organizations: “Yes, we have strong leaders from all sectors — public health, clinical providers, and community-based organizations.”
Informants were unanimous in praising the staff person from the Department of Health Services for providing more than adequate staffing and support, leveraging the small amount of county funding and the CACHI grant to support the growing number of HSC activities: “Our health department [Department of Health Services] has provided consistent critical support. They have competing priorities but have always been engaged in this effort.”
Active collaboration is a critical but hard-to-define property of effective collaboratives: people and organizations set aside their more narrow organizational interests in support of the whole group. All of the informants understood the concept and agreed that over time people had seen the value of collaboration. As one informant said, “The tone of the meetings is sharing what works so that others can benefit from it. How can we improve care for all, recognizing differences and helping each other. It’s in the nature of how the organization came about. We’re charged with improving quality of life in a number of domains.”
Finally, taking action occurred initially in the form of health care organizations bringing back what they learned from the collaborative to their own organizations to be implemented: “Moving quickly to action to demonstrate value has been important; for example, one of the medical groups implemented a team-based care pilot in her own practice based on their HSC experience. People take things back.” The value of those early learnings helped catalyze other activities, including CACHI and community engagement.
In addition to validating the elements of the CCHE model, we asked respondents in a more open-ended way about the key accomplishments of HSC and why they thought the effort had lasted. They did not often mention specific activities or clinical improvements, but rather that they appreciated the collaboration itself and being able to step outside the competitive realm of their different health systems to focus on what could be done to improve patient and community health. The dialog and shared learnings at the monthly meetings built trust and promoted the active collaboration. Respondents attributed the sustainability of the HSC effort to the building of that level of trust.

Implications for Public Health

Although the positive results — early but encouraging countywide trends in blood pressure control and significant community engagement activities with more in the works — are important, another goal of the HSC evaluation was to understand the factors behind the staying power and impact of the collaborative. We looked in particular for structural or process factors that might be generalizable to other, similar collaboratives. Three such factors that emerged were starting small and focused, while working within the framework of a larger effort, and providing backbone support that was open-ended and not limited by funding time constraints.
Start small and focused to build trust and demonstrate value. The initial seeds of the HSC initiative were the activities of the Committee for Healthcare Improvement (CHI), operating as part Health Action starting in 2007. A small number of clinical champions from the key health organizations came together to see whether sharing lessons from others could benefit their own organizations. They were able to agree on a purpose and mission and move to action fairly quickly even though resources to implement whatever changes they identified were limited and had to come from within their own organizations. These early successes helped build trust and demonstrate the value of the collaborative.
Operate within a larger structure. Although the health care work involved a small number of people with a narrow focus, it was embedded in the larger Health Action collaborative. This had 3 long-term advantages. First, leaders on the Health Action Council approved projects undertaken by CHI, including HSC, which translated into a leadership and organizational commitment to HSC. Second, connections were created with a larger group of member organizations who were potential collaborators as the work grew in scope. Third, it was easier to secure long-term backbone support from Sonoma County, because the effort had a broad focus and therefore a wider political constituency.
The lessons about starting small but operating within a larger structure suggest a path for others seeking to ultimately create a large-scale collaborative to achieve health system transformation. Create a large, ambitious collaborative structure and membership, but be willing to focus initial activities narrowly where progress can most readily be made. This requires accepting modest results in terms of health impact, which can also help build the trust required for sustainability.
Other lessons were learned through this process. Grant-funded collaboratives are often time-limited, and it can be challenging to find funding streams to sustain the effort. A key to the success of HSC was the long-term in-kind support provided by the Sonoma County Department of Health Services. This was enough to provide support to the early focused efforts of HSC. Also, administrative and especially clinical leadership in each organization is essential to teach colleagues, guide the direction of change, and encourage the use of protocols. These can all be difficult for clinicians to accept and implement, so leadership is essential. Finally it is important to have small successes and celebrate them along the way. This keeps people interested and knowing progress is being made. Having the shared purpose, however, is key. These lessons are consistent with what others have found (8) and not revolutionary, but they are often ignored in the sense of urgency created by the need to transform the health care system and the availability of large-scale, but time-limited, funding available through State Innovation Model grants (3), Medicaid DSRIP (Delivery System Reform Incentive Payment) Waivers (13), and other sources.
Some limitations should be noted. The evaluation of the community engagement activities has been a more qualitative, process evaluation; longer-term outcome measures are still being developed. The data on CVD outcomes (eg, blood pressure) are limited to the 4 participating providers, which represent just over half of the county patient population. Finally, HSC is focused on CVD only, which, although a leading cause of illness and death, is not indicative of overall health system transformation. However, many of the issues that arise in working in CVD (eg, data sharing, collaboration across systems, linking with community resources) are present in broader transformation efforts, so the HSC lessons should apply.
The HSC collaborative members continue to work together. On the horizon are the continued expansion of the community engagement work, the creation of a clinical population health improvement collaborative to broaden and standardize the clinical improvement work, and additional population-level metrics to judge the impact. The goal continues to be implementing targeted, coordinated clinical, community, and policy interventions to improve cardiovascular health, recognizing that only by sustaining efforts over the long term can sustained health improvement be achieved.

Acknowledgments

Thanks to Sonoma County Department of Health Services for conducting the analysis of the blood pressure control data. Funding for the HSC initiative was provided by St. Joseph Health Community Benefit Programs, United Way of the Wine Country, The California Endowment, Blue Shield of California Foundation, Kaiser Permanente, The California Wellness Foundation, and the Sierra Health Foundation. Funding for the initiative evaluation was provided by Kaiser Permanente. The authors have no conflicts of interest or financial disclosures.

Author Information

Corresponding Author: Allen Cheadle, PhD, Center for Community Health and Evaluation, Kaiser Permanente Washington Health Research Institute, Seattle, WA 98101. Telephone: 206-287-4391. Email: Allen.D.Cheadle@kp.org.
Author Affiliations: 1Center for Community Health and Evaluation, Kaiser Permanente Washington Health Research Institute, Seattle, Washington. 2Redwood Community Health Coalition, Petaluma, California. 3Kaiser Permanente Medical Group, Sonoma County, California. 4St. Joseph Health Medical Group, Irvine, California. 5Vista Family Health Center, Vista, California. 6Sonoma County Department of Health Services, Santa Rosa, California. 7Northern California Center for Well-Being, Santa Rosa, California. 8Kaiser Permanente, Oakland, California.

References

  1. National Academies of Sciences Engineering, and Medicine; Baciu A, Negussie Y, Geller A, Weinstein JN, editors. Communities in action: pathways to health equity. Washington (DC): National Academies Press; 2017.
  2. Kubisch AC, Auspos P, Brown P, Dewar T. Voices from the field III: lessons and challenges from two decades of community change efforts. Washington (DC): Aspen Institute; 2010.
  3. Centers for Medicare and Medicaid Services. State innovation models initiative; 2017. https://innovation.cms.gov/initiatives/state-innovations/. Accessed August 25, 2018.
  4. Tipirneni R, Vickery KD, Ehlinger EP. Accountable communities for health: moving from providing accountable care to creating health. Ann Fam Med 2015;13(4):367–9. CrossRefexternal icon PubMedexternal icon
  5. Siegel B, Erickson J, Milstein B, Pritchard KE. Multisector partnerships need further development to fulfill aspirations for transforming regional health and well-being. Health Aff (Millwood) 2018;37(1):30–7. CrossRefexternal icon PubMedexternal icon
  6. Merzel C, D’Afflitti J. Reconsidering community-based health promotion: promise, performance, and potential. Am J Public Health 2003;93(4):557–74. CrossRefexternal icon PubMedexternal icon
  7. Butterfoss FD, Goodman RM, Wandersman A. Community coalitions for prevention and health promotion: factors predicting satisfaction, participation, and planning. Health Educ Q 1996;23(1):65–79. CrossRefexternal icon PubMedexternal icon
  8. Roussos ST, Fawcett SB. A review of collaborative partnerships as a strategy for improving community health. Annu Rev Public Health 2000;21(1):369–402. CrossRefexternal icon PubMedexternal icon
  9. United States Census Bureau. 2011–2016 American Community Survey; 2016. http://factfinder2.census.gov. Accessed August 25, 2018.
  10. California Health Interview Survey. CHIS 2017–2018 release; 2018. http://healthpolicy.ucla.edu/chis/Pages/default.aspx.
  11. Centers for Medicare and Medicaid Services. Million Hearts: cardiovascular disease risk reduction model; 2017. https://innovation.cms.gov/initiatives/Million-Hearts-CVDRRM. Accessed August 25, 2018.
  12. Jaffe MG, Lee GA, Young JD, Sidney S, Go AS. Improved blood pressure control associated with a large-scale hypertension program. JAMA 2013;310(7):699–705. CrossRefexternal icon PubMedexternal icon
  13. Henry J. Kaiser Family Foundation. An overview of Delivery System Reform Incentive Payment (DSRIP) waivers; 2014. https://www.kff.org/medicaid/issue-brief/an-overview-of-delivery-system-reform-incentive-payment-waivers/. Accessed August 26, 2018.

State-Level and County-Level Estimates of Health Care Costs Associated with Food Insecurity

State-Level and County-Level Estimates of Health Care Costs Associated with Food Insecurity

PCD logo

State-Level and County-Level Estimates of Health Care Costs Associated with Food Insecurity

Seth A. Berkowitz, MD, MPH1,2; Sanjay Basu, MD, PhD3,4,5; Craig Gundersen, PhD6; Hilary K. Seligman, MD, MAS7,8 (View author affiliations)

Suggested citation for this article: Berkowitz SA, Basu S, Gundersen C, Seligman HK. State-Level and County-Level Estimates of Health Care Costs Associated with Food Insecurity. Prev Chronic Dis 2019;16:180549. DOI: http://dx.doi.org/10.5888/pcd16.180549external icon.
PEER REVIEWED
Summary
What is already known on this topic?
Food insecurity is associated with higher health care costs, on average.
What is added by this report?
We found substantial variation in state- and county-level health care expenditures associated with food insecurity. We also found that higher food insecurity prevalence is more strongly associated with higher spending than differences in health care prices or intensity of health care use.
What are the implications for public health practice?
A multi-level strategy that encompasses both area-level determinants of food insecurity (eg, local labor market factors and state-earned income tax credits) and hunger safety net programs may improve public health.

Abstract

Introduction
Food insecurity, or uncertain access to food because of limited financial resources, is associated with higher health care expenditures. However, both food insecurity prevalence and health care spending vary widely in the United States. To inform public policy, we estimated state-level and county-level health care expenditures associated with food insecurity.
Methods
We used linked 2011–2013 National Health Interview Survey/Medical Expenditure Panel Survey data (NHIS/MEPS) data to estimate average health care costs associated with food insecurity, Map the Meal Gap data to estimate state-level and county-level food insecurity prevalence (current though 2016), and Dartmouth Atlas of Health Care data to account for local variation in health care prices and intensity of use. We used targeted maximum likelihood estimation to estimate health care costs associated with food insecurity, separately for adults and children, adjusting for sociodemographic characteristics.
Results
Among NHIS/MEPS participants, 10,054 adults and 3,871 children met inclusion criteria. Model estimates indicated that food insecure adults had annual health care expenditures that were $1,834 (95% confidence interval [CI], $1,073–$2,595, P < .001) higher than food secure adults. For children, estimates were $80 higher, but this finding was not significant (95% CI, −$171 to $329, P = .53). The median annual health care cost associated with food insecurity was $687,041,000 (25th percentile, $239,675,000; 75th percentile, $1,140,291,000). The median annual county-level health care cost associated with food insecurity was $4,433,000 (25th percentile, $1,774,000; 75th percentile, $11,267,000). Cost variability was related primarily to food insecurity prevalence.
Conclusions
Health care expenditures associated with food insecurity vary substantially across states and counties. Food insecurity policies may be important mechanisms to contain health care expenditures.

Introduction

Food insecurity, or uncertain access to food because of limited financial resources, affected 12.9% of Americans in 2016 — more than 40 million individuals (1). Food insecurity is associated with numerous chronic health conditions, including diabetes mellitus, hypertension, coronary heart disease, chronic kidney disease, and depression (2–6). Perhaps for this reason, estimates from both the United States and Canada indicate that, on average, health care costs are substantially higher among food-insecure individuals than among food-secure individuals (7–9).
Although food insecure individuals in the United States experience higher health care costs on average, this average likely obscures substantial variation across states and counties. The Map the Meal Gap study (http://map.feedingamerica.org/) has shown that US food insecurity rates vary widely (10). Similarly, local pricing and intensity of health care use also differ in the United States, resulting in widespread variation in health care spending (11,12). Furthermore, these patterns do not necessarily match; an area with higher food insecurity may have lower health care prices, and vice versa. This means that estimating local health care costs associated with food insecurity is not straightforward.
Understanding variation in health care costs associated with food insecurity has substantial public health implications, because doing so can inform the implementation of new initiatives (eg, the Centers for Medicare & Medicaid Services’ Accountable Health Communities Model [13]) or state and local public health and nutrition programs. Such programs could focus scarce resources on areas where health care costs associated with food insecurity are high. Furthermore, local economic policy, particularly state-earned income tax credits, local wage conditions, and housing policies can influence food insecurity (14,15). Therefore, understanding variations in health care costs associated with food insecurity has implications beyond public health.
To help inform both policies and programs to address these issues, we sought to estimate county-level and state-level health care costs associated with food insecurity in the United States.

Methods

Study design and data sources

To generate local estimates of health care costs associated with food insecurity, we needed 3 key pieces of information: 1) the mean per-person dollar amount of excess health care expenditures among adults and children; 2) the number of food-insecure adults and children residing in each county and state; and 3) the variation, from the national average, in health care costs for each county and state. The rationale for this was that, because health care expenditures exhibit substantial geographic variability, a similar individual might have lower health care costs if they resided in a low-cost area (in terms of health care prices) and higher health care costs if they lived in a high-cost area, even if their health care needs were exactly the same. Because no single data source had information on all 3 of these factors, we needed to combine data from several sources to generate our estimates. The institutional review board at the University of North Carolina at Chapel Hill exempted this analysis of secondary data from human subjects review.

National Health Interview Survey/Medical Expenditure Panel Survey

To estimate the excess health care costs, if any, associated with food insecurity, we used linked data from the National Health Interview Survey (NHIS) (16) and the Medical Expenditure Panel Survey (MEPS) (17). NHIS is a nationally representative epidemiologic surveillance survey of the civilian noninstitutionalized US population (16). MEPS is a nationally representative cohort that collects detailed data on health care expenditures over a 2-year period and is drawn from NHIS participants (17). We used data collected from NHIS participants in 2011 who participated in MEPS during 2012–2013. We extracted information on the exposure of food security status from NHIS, which used a 10-item version of the United States Department of Agriculture food security survey module for adults with a 30-day look-back window (16). In accordance with standard scoring, raw scores of 0 to 2 were considered food secure and raw scores of 3 to 10 were considered food insecure (16). We used the MEPS total health care expenditures variable, which includes all health care costs (eg, inpatient admissions, outpatient visits, medication costs). Using NHIS food insecurity data and MEPS health care cost data ensures appropriate time ordering between the hypothesized exposure and outcome. More details on NHIS and MEPS data, as well as on the estimates and statistical methods used in this study, are provided at https://saberkowitz.web.unc.edu/supplemental-information/state-and-local-healthcare-costs/.

Map the Meal Gap

Data on the prevalence of food insecurity among adults and children at the county and state level came from Map the Meal Gap (MMG), which is based on US Census data (including the American Community Survey and Current Population Survey) and Bureau of Labor Statistics data. MMG methods have been published (10). MMG uses a 2-step process established by Feeding America to obtain estimates of food insecurity prevalence for all US counties. In the first step, the state-level determinants of food insecurity (for both children and adults) are estimated based on data from 2001 through 2016. The model components used are unemployment, poverty, median income, percentage Hispanic ethnicity, percentage African-American race, percentage living in owned housing, year fixed effects, and state fixed effects. These models are then used in the second step to produce food insecurity estimates at the county level, using county-specific variables. For our study, the county-specific variables were drawn from the 2016 American Community Survey 5-year estimates.

Dartmouth Atlas of Health Care

To estimate how a given county or state differed in health care spending (either based on prices or intensity of care) from the national average, we used data from the Dartmouth Atlas of Health Care (www.dartmouthatlas.org/), covering 2012–2013 because that was when cost data were collected, to calculate a “cost factor.” The resulting cost factor is greater than 1 for areas with higher-than-average costs and less than 1 for areas with lower-than-average costs (18).

Statistical analysis

Step 1 of our analysis was to determine national estimates of excess health care costs, if any, associated with food insecurity. To do this, we used NHIS and MEPS data. Analyses incorporated representativeness weights and survey design (clustering) information as appropriate. Because the mechanisms through which food insecurity may be associated with health care costs are likely different between adults and children, we stratified our data by age (≥18 years for adults and <18 years for children) and then made separate estimates in these groups. To generate the cost estimates, we drew on prior work examining the association between food insecurity and health care costs (8). Because health care cost data are notoriously difficult to analyze (19) and generalized linear models rely on certain assumptions that may not always be met in practice, we applied a targeted maximum likelihood estimation approach (TMLE). TMLE is a doubly robust analytic strategy that initially creates an estimate of the excess health care costs associated with food insecurity and then updates that estimate using a submodel that estimates the probability of being food insecure (20).
Using NHIS and MEPS data allowed us to estimate the mean per-person cost associated with food insecurity for adults and children, but NHIS and MEPS are not designed to estimate health care costs for every county or state. Therefore, in step 2, we multiplied our nationally representative per-person estimate of health care costs by the number of food insecure adults and children in each county and state (using data from MMG). Then, to account for county and state differences in health care spending, we multiplied by the cost factor for the locality. To bound the uncertainty in the estimates, we created a lower and upper bound by using the 95% confidence interval (CI) for the NHIS/MEPS estimate of average health care costs associated with food insecurity. Finally, we conducted correlation analyses to help understand whether local variations in health care costs associated with food insecurity are more closely related to food insecurity prevalence or local health care spending characteristics.
All dollar estimates were inflation adjusted to December 2016 dollars, following MEPS guidance (https://meps.ahrq.gov/about_meps/Price_Index.shtml). All analyses — with the exception of MMG estimates, which were derived using Stata version 14.2 (StataCorp LP) — were conducted in SAS version 9.4 (SAS Institute, Inc) and R version 3.4.2 (R Foundation).

Results

In the analyses of health care costs associated with food insecurity, 10,054 adults and 3,871 children were included. Both food-insecure adults and children were more likely than their food-secure counterparts to be racial/ethnic minorities, have lower income, and lack health insurance (Table 1).
In TMLE analyses that accounted for age, sex, race/ethnicity, income, education, health insurance, metropolitan residence, and region of residence within the country, model-based estimates showed that adults who were food insecure had annual health care expenditures that were $1,834 (95% CI, $1,073–$2,595) higher than adults who were food secure (P < .001). In children, the model-based estimate for health care costs associated with food insecurity was $80 annually, but this finding was not significant (P = 0.53, 95% CI, −$171 to $329). Among approximately 28,266,000 food-insecure adults and 12,938,000 food-insecure children in the United States in 2016, using these model-based point estimates of the excess cost associated with food insecurity translates to approximately $52.9 billion in excess health care expenditures associated with food insecurity in 2016 (95% CI, $31.8 billion to $74.3 billion). This represents 3% to 6% of the approximately $1.2 trillion in annual health care expenditures we estimate from MEPS data. Because the estimate for children was not significantly different from $0, taking only adult costs yielded a national estimate of $51.8 billion in excess health care expenditures in 2016 (95% CI, $31.7 billion to $74.2 billion).
Using the model-based estimates from our main analyses (eg, point estimate for adults of $1,834), we then calculated the costs associated with food insecurity for each state (including the District of Columbia) and county in the United States (Figures 1 and 2). Estimates by state are presented in Table 2 and estimates by county are presented in the Appendix. At the state level, adult food insecurity prevalence ranged from 6.8% (North Dakota) to 17.6% (Mississippi), and child food insecurity prevalence ranged from 10.3% (North Dakota) to 25.0% (New Mexico). At the state level, the mean annual model-based health care cost associated with food insecurity was $1,087,815,000 (standard deviation [SD], $1,407,496,000), and the median annual health care cost associated with food insecurity was $687,041,000 (25th percentile, $239,675,000; 75th percentile, $1,140,291,000). The state with the highest annual model-based health care cost associated with food insecurity was California, at $7,213,940,000, and the state with the lowest annual health care cost associated with food insecurity was North Dakota at $57,587,000. On a per capita basis, Mississippi had the highest health care cost associated with food insecurity, while North Dakota had the lowest. The 5 states with the highest per capita health care costs associated with food insecurity were Mississippi, Texas, Louisiana, Florida, and Oklahoma. The mean annual county-level health care cost associated with food insecurity was $17,905,000 (SD, $69,194,000), and the median annual county-level health care cost associated with food insecurity was $4,433,000 (25th percentile, $1,774,000; 75th percentile, $11,267,000).

Health care costs associated with food insecurity (A) and per capita health care costs associated with food insecurity (B), by state, United States, 2012–2013.
Figure 1.
Health care costs associated with food insecurity (A) and per capita health care costs associated with food insecurity (B), by state, United States, 2012–2013. [A text description of this figure is available.]

Health care costs associated with food insecurity (A) and per capita health care costs associated with food insecurity (B), by county, United States, 2012–2013.
Figure 2.
Health care costs associated with food insecurity (A) and per capita health care costs associated with food insecurity (B), by county, United States, 2012–2013. [A text description of this figure is available.]
The components of our cost estimates were the number of food-insecure individuals and the cost factor that accounted for local care intensity and prices. We found that at both the county and state level, the number of individuals who were food insecure was strongly correlated with the total expenditure estimate (r2 = 0.99 for county cost and r2 = 0.99 for state cost). The total expenditure estimate was only weakly or moderately associated with the cost factor (r2 = 0.22 for county cost and r2 = 0.57 for state cost). This suggests that a high proportion of the variation in food insecurity–associated health care expenditures is attributable to the number of food-insecure individuals.

Discussion

We found that food insecurity was associated with higher health care spending in adults and that this spending varied substantially across locality. Although patterns of local health care use and price explained some of this difference, the number of food-insecure individuals, and in particular the number of food-insecure adults, accounted for the largest share of variation in associated costs.
These findings are consistent with and expand our knowledge about the relationship between food insecurity and health care costs. Studies in both Canada (9) and the United States (7,8) have found that food insecurity is associated with higher health care costs. Specifically, a study from our research group (8) using similar methods found higher health care costs associated with food insecurity during a period when food insecurity prevalence was higher. Furthermore, recent research in the United States has found that, for several common clinical conditions, food insecurity is associated with excess health care costs even when accounting for other demographic and clinical characteristics (21). This may be related to several factors (2,22), including worse dietary quality in food-insecure individuals (23); trade-offs between food and other basics, such as medications, that make chronic disease management more difficult (24); and psychological factors, including stress and depressive symptoms (6). This study adds to this literature by quantifying the wide variation in excess expenditures. Although this study cannot determine why this variation occurs, the correlation analyses suggest that variation in model-based estimates of local health care costs associated with food insecurity is closely correlated with food insecurity prevalence in the area and less closely correlated with the local cost factor.
In our analyses, the point estimate for health care costs associated with food insecurity in children was small and not significantly different than $0. Although this study cannot determine why that is the case, past work suggests that food insecurity may be most closely related to increased health care cost through increased prevalence of chronic disease and exacerbation of those chronic conditions when they occur (8,22). If this is the case, then children may not see short-term (eg, the 2-year period in the NHIS/MEPS data) increases in health care costs simply because they are at low risk of developing these conditions, regardless of food security status. This does not imply, however, that food insecurity does not have long-term effects on children’s health or even short-term effects on important aspects of life that do not generate short-tern health care costs, like school achievement.
This study has implications for public health. Literature has demonstrated that local and state economic policies and conditions can have a substantial effect on food insecurity prevalence (14,15). In particular, lower tax burden (but not overall tax burden) for low-income individuals is associated with lower food insecurity, with strong associations between lower food insecurity and higher state-earned income tax credits (14,15). Other factors associated with lower food insecurity include local labor conditions and ease of access to hunger safety-net programs (14,15). For this reason, an important direction for future research will be to evaluate whether polices that reduce area food insecurity prevalence also lead to lower health care spending. Because there is evidence that individual-level nutrition interventions, particularly the Supplemental Nutrition Assistance Program (SNAP) (25–27) and medically tailored meal delivery programs (28), may also be associated with lower health care costs, having area-level policy options could provide a multilevel framework for addressing high health care spending by supporting access to proper nutrition. Fewer than 40% of individuals with food insecurity in this study had private health insurance, meaning that public health care programs, particularly at the state level, are shouldering much of the cost associated with food insecurity. SNAP and other nutrition programs are funded at the federal level, so if states worked to maximize uptake of federal nutrition programs, they may not only lower food insecurity rates but also decrease health care expenditures.
This study has several limitations. The costs estimated are likely conservative, because there is evidence that MEPS underestimates health care expenditures (29), and we did not consider indirect costs (like lost productivity owing to illness). Also, the sample used for estimating health care expenditures included only civilian noninstitutionalized individuals, which excludes some groups. The association between food insecurity and health care costs may not be fully related to food insecurity causing higher costs. There is likely to be a bidirectional relationship whereby food insecurity may worsen health, thus increasing health care costs, and worse health (and attendant expenses) may lead to food insecurity by decreasing the ability to work and increasing household debt. When examining the relationship between food insecurity and health care costs, there is a small delay between food security assessment in NHIS and the beginning of cost data collection in MEPS. Although assessment of food insecurity before collection of cost data is necessary to preserve time-ordering and mitigate reverse causation, the delay could lead to some misclassification (if food security status changes in the interval), which would tend to bias results to the null. Also, because NHIS and MEPS are not designed to yield county-level estimates of food insecurity prevalence or health care costs, we had to combine NHIS and MEPS data with data sources that were designed to provide more granular estimates. In addition, this is an inherently ecological analysis. However, since both the exposure (food insecurity prevalence) and the outcome (health care spending in the locality) were area-level assessments, this type of analysis is not subject to concerns about ecological fallacy (30). Finally, we did not have the ability to look at the specific distribution of comorbidities within each locality. To the extent that differences in comorbidities reflect differences in effect modifiers, actual local spending will not match the estimates. This would occur both for areas where individuals are less healthy than expected (and thus incur greater costs) and areas where individuals are healthier than expected (and correspondingly have lower health care costs).
Our study also has strengths. We used a nationally representative, longitudinal data set to estimate the association between food insecurity and health care costs. Furthermore, we used robust and well-validated methods to provide local estimates of food insecurity prevalence.
Food insecurity is associated with substantial health care expenditures, but there is evidence that this varies widely across states and counties. This variation suggests that local and state policies could be important mechanisms for improving health and containing health care expenditures. As health care cost containment remains a national priority, state and local strategies to reduce food insecurity rates may be an important public health tool.

Acknowledgments

The research reported in this publication was supported by Feeding America. The funders did not have any role in the conduct of the research or the decision to submit for publication. S.A.B. and S.B. have no conflicts of interest to report. Both C.G. and H.K.S. serve on Feeding America’s Technical Advisory Group. H.K.S. is also senior medical advisor at Feeding America. S.A.B.’s role in the research reported in this publication was also supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health under award no. K23DK109200. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author Information

Corresponding Author: Seth A. Berkowitz, MD, MPH, 5034 Old Clinic Bldg, CB 7110, Chapel Hill, NC 27599. Telephone: 919-966-2276. Email: seth_berkowitz@med.unc.edu.
Author Affiliations: 1Division of General Medicine and Clinical Epidemiology, Department of Medicine, University of North Carolina at Chapel Hill School of Medicine, Chapel Hill, North Carolina. 2Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina. 3Research and Analytics, Collective Health, San Francisco, California. 4School of Public Health, Imperial College London, London, United Kingdom. 5Center for Primary Care, Harvard Medical School, Boston, Massachusetts. 6Department of Agricultural and Consumer Economics, University of Illinois, Urbana, Illinois. 7Departments of Medicine and Epidemiology and Biostatistics, University of California San Francisco, San Francisco, California. 8UCSF Center for Vulnerable Populations at Zuckerberg San Francisco General Hospital, San Francisco, California.

References

  1. Coleman-Jensen A, Rabbitt MP, Gregory CA, Singh A. Household food security in the United States in 2017. https://www.ers.usda.gov/publications/pub-details/?pubid=90022. Accessed September 25, 2018.
  2. Gundersen C, Ziliak JP. Food insecurity and health outcomes. Health Aff (Millwood) 2015;34(11):1830–9. CrossRefexternal icon PubMedexternal icon
  3. Berkowitz SA, Berkowitz TSZ, Meigs JB, Wexler DJ. Trends in food insecurity for adults with cardiometabolic disease in the United States: 2005-2012. PLoS One 2017;12(6):e0179172. CrossRefexternal icon PubMedexternal icon
  4. Seligman HK, Bindman AB, Vittinghoff E, Kanaya AM, Kushel MB. Food insecurity is associated with diabetes mellitus: results from the National Health and Nutrition Examination Survey (NHANES) 1999-2002. J Gen Intern Med 2007;22(7):1018–23. CrossRefexternal icon PubMedexternal icon
  5. Crews DC, Kuczmarski MF, Grubbs V, Hedgeman E, Shahinian VB, Evans MK, et al. ; Centers for Disease Control and Prevention Chronic Kidney Disease Surveillance Team. Effect of food insecurity on chronic kidney disease in lower-income Americans. Am J Nephrol 2014;39(1):27–35. CrossRefexternal icon PubMedexternal icon
  6. Leung CW, Epel ES, Willett WC, Rimm EB, Laraia BA. Household food insecurity is positively associated with depression among low-income supplemental nutrition assistance program participants and income-eligible nonparticipants. J Nutr 2015;145(3):622–7.CrossRefexternal icon PubMedexternal icon
  7. Berkowitz SA, Seligman HK, Meigs JB, Basu S. Food insecurity, health care utilization, and high cost: a longitudinal cohort study. Am J Manag Care 2018;24(9):399–404. PubMedexternal icon
  8. Berkowitz SA, Basu S, Meigs JB, Seligman HK. Food insecurity and health care expenditures in the United States, 2011-2013. Health Serv Res 2017. PubMedexternal icon
  9. Tarasuk V, Cheng J, de Oliveira C, Dachner N, Gundersen C, Kurdyak P. Association between household food insecurity and annual health care costs. CMAJ 2015;187(14):E429–36. CrossRefexternal icon PubMedexternal icon
  10. Gundersen C, Engelhard E, Waxman E. Map the meal gap: exploring food insecurity at the local level. Appl Econ Perspect Policy 2014;36(3):373–86. CrossRefexternal icon
  11. Newhouse JP, Garber AM. Geographic variation in health care spending in the United States: insights from an Institute of Medicine report. JAMA 2013;310(12):1227–8. CrossRefexternal icon PubMedexternal icon
  12. Geographic variation in health care spending and promotion of high-value care. Institute of Medicine; 2018. http://www.nationalacademies.org/hmd/Activities/HealthServices/GeographicVariation.aspx. Accessed April 30, 2019.
  13. Alley DE, Asomugha CN, Conway PH, Sanghavi DM. Accountable health communities — addressing social needs through Medicare and Medicaid. N Engl J Med 2016;374(1):8–11. CrossRefexternal icon PubMedexternal icon
  14. Bartfeld J, Men F. Food insecurity among households with children: the role of the state economic and policy context. Soc Serv Rev 2017;91(4):691–732. CrossRefexternal icon
  15. Bartfeld J, Dunifon R, Nord M, Carlson S. What factors account for state-to-state differences in food security? https://www.ers.usda.gov/publications/pub-details/?pubid=44133. Accessed April 30, 2019.
  16. Centers for Disease Control and Prevention. National Health Interview Survey; 2018. https://www.cdc.gov/nchs/nhis/index.htm. Accessed April 30, 2019.
  17. Agency for Healthcare Research and Quality. Medical Expenditure Panel Survey; 2018. https://meps.ahrq.gov/mepsweb/. Accessed April 30, 2019.
  18. The Dartmouth Institute. Dartmouth Atlas of Health Care; 2018. http://www.dartmouthatlas.org/tools/downloads.aspx. Accessed April 30, 2019.
  19. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ 2001;20(4):461–94. CrossRefexternal iconPubMedexternal icon
  20. Schuler MS, Rose S. Targeted maximum likelihood estimation for causal inference in observational studies. Am J Epidemiol 2017;185(1):65–73. CrossRefexternal icon PubMedexternal icon
  21. Garcia SP, Haddix A, Barnett K. Incremental health care costs associated with food insecurity and chronic conditions among older adults. Prev Chronic Dis 2018;15:180058. CrossRefexternal icon PubMedexternal icon
  22. Seligman HK, Schillinger D. Hunger and socioeconomic disparities in chronic disease. N Engl J Med 2010;363(1):6–9. CrossRefexternal iconPubMedexternal icon
  23. Morales ME, Berkowitz SA. The relationship between food insecurity, dietary patterns, and obesity. Curr Nutr Rep 2016;5(1):54–60.CrossRefexternal icon PubMedexternal icon
  24. Berkowitz SA, Seligman HK, Choudhry NK. Treat or eat: food insecurity, cost-related medication underuse, and unmet needs. Am J Med 2014;127(4):303–310e3.
  25. Berkowitz SA, Seligman HK, Rigdon J, Meigs JB, Basu S. Supplemental Nutrition Assistance Program (SNAP) participation and health care expenditures among low-income adults. JAMA Intern Med 2017;177(11):1642–9. CrossRefexternal icon PubMedexternal icon
  26. Srinivasan M, Pooler JA. Cost-related medication nonadherence for older adults participating in SNAP, 2013–2015. Am J Public Health 2018;108(2):224–30. CrossRefexternal icon PubMedexternal icon
  27. Samuel LJ, Szanton SL, Cahill R, Wolff JL, Ong P, Zielinskie G, et al. Does the Supplemental Nutrition Assistance Program affect hospital utilization among older adults? The case of Maryland. Popul Health Manag 2017. PubMedexternal icon
  28. Berkowitz SA, Terranova J, Hill C, Ajayi T, Linsky T, Tishler LW, et al. Meal delivery programs reduce the use of costly health care in dually eligible Medicare and Medicaid beneficiaries. Health Aff (Millwood) 2018;37(4):535–42. CrossRefexternal icon PubMedexternal icon
  29. Validating the collection of separately billed doctor expenditures for hospital services: results from the Medicare-MEPS Validation Study; 2018. https://meps.ahrq.gov/mepsweb/data_stats/Pub_ProdResults_Details.jsp?pt=Working+Paper&opt=2&id=855. Accessed April 30, 2019.
  30. Subramanian SV, Jones K, Kaddour A, Krieger N. Revisiting Robinson: the perils of individualistic and ecologic fallacy. Int J Epidemiol 2009;38(2):342–60, author reply 370–3. CrossRefexternal icon PubMedexternal icon