 Physician leaders represent diverse career interests, clinical specialties, and leadership types. Despite the perceived value of their clinical perspectives, little is known about how physician leaders balance patient care and leadership responsibilities or how that balance varies by clinical specialty or leadership type.&ensp;We conducted what is to our knowledge the first survey evaluating physician leaders’ perspectives about these issues.

 Web-based survey administered September 2017 to February 2018 among physician leaders from the 22 health systems cited in the 2016-2018 U.S. News & World Report Best Hospitals Honor Rolls.

 Respondents were asked questions regarding their participation in patient care. We described our findings using percentages and χ

 tests to compare categorical variables across leadership type (executive, educational, clinical) and clinical specialty (procedural, nonprocedural, supportive).

 Of 447 leaders contacted, 53% responded to our survey. Of those, 84% reported some degree of participation in patient care (69%, 90%, and 85% of executive, educational, and clinical leaders, respectively; 

&thinsp;=&thinsp;.001); most (74%) spent no more than a quarter of their time on patient care, and they often achieved a balance that they deemed optimal (74%). Proceduralists were more likely to spend over a quarter of their time on patient care and less likely to achieve an optimal balance.

 Although most physician leaders engage in some degree of patient care, the balance that they achieve varies by clinical specialty and leadership type. These findings provide insight for health care organizations seeking to engage and optimize physician leadership amid ongoing health care transformation.

The American Journal of Accountable Care. 2020;8(2):26-31

American health care continues to undergo significant transformation. Changes in health policy

; consolidation among providers, payers, and other stakeholders

 have placed an emphasis on accountability in health care delivery. These changes underscore the potential value of physician leadership,

 which has been associated with improvements in operational efficiency,

 These dynamics have led to increased organizational interest in physicians with leadership skills

 and growing numbers of physicians seeking leadership training.

Importantly, physicians pursue leadership careers for different reasons

: Some seek administrative opportunities (eg, health system medical director) whereas others desire leadership in specific clinical (eg, service line chief) or academic (eg, department chair) areas. Variation also exists in each of these categories. For instance, some medical directors focus on specific projects or topic areas (eg, a certain hospital unit, an infection control program) whereas others focus on topics that cut across teams and units (eg, value-based care, population health).

Different leadership paths require different skill sets.

 For instance, medical directors of clinical units benefit from change management and process improvement skills needed to drive daily operations. Physicians with leadership in cross-cutting topics such as population health must be proficient in policy analysis, data synthesis, and change management.

Despite this diversity, a common thread is that physicians’ first-hand patient care experience lends them unique leadership perspective

 that can be highly valuable to themselves and their organizations. For example, a physician who maintains an active clinical practice may have greater credibility among clinicians whom they lead, or a physician can inform organizational contracting using their clinical perspective to balance financial and operational considerations. Conversely, the time and energy needed to maintain a clinical practice can siphon attention away from leadership work, with impacts on administrative capacity and potential leadership efficacy.

 All physician leaders face the task of balancing clinical and leadership responsibilities.

Despite these dynamics and the possibility that they vary by type of leadership role and clinical specialty, little is known about physician leaders’ perspectives on balancing clinical and leadership responsibilities. Therefore, we conducted what is to our knowledge the first survey capturing these perspectives.

We conducted a web-based survey and followed approaches used in prior work by sampling physician leaders from the 22 hospitals cited at least once in the 

Best Hospitals Honor Roll from 2016 to 2018.

 For each institution, we reviewed online content descriptions of its executive leadership teams and respective departmental websites to identify the names and contact information of the individuals who possessed a medical degree (MD or DO) and filled a position in 1 of the following categories of leadership types: executive (CEOs, vice presidents, and other health system executives), educational (medical school deans and academic department chairs), and clinical (internal medicine division or clinical service line chiefs). Division chiefs were limited to internal medicine division chiefs to promote consistency in use of title and scope of leadership responsibilities. Where online leadership information was limited (1 hospital), we followed an approach used in prior work by initiating email outreach to identify all eligible leaders.

 Individuals filling multiple leadership roles were assigned to 1 category (executive, if relevant, followed by academic and clinical, in that order). Retirees were excluded. This yielded a sample of 447 potential participants (average of 20.3 leaders per institution).

Respondents were asked what amount of time they currently spend on direct patient care, defined as serving as a physician of record or consultant on a clinical service (eAppendix A [

]). They were also asked about the optimal amount of time that they believed someone in their leadership position should spend on direct patient care to be an effective leader; the minimum amount of time someone in their clinical specialty needs to spend on direct patient care to be a safe, effective clinician; and whether they would increase or decrease their participation in patient care. Physicians participating in patient care were asked if it influences their performance in their leadership role and, if so, how. Those not currently involved in direct patient care were asked why they no longer practice. Additional questions confirmed leaders’ leadership type, clinical specialty, and years in clinical practice.

After pretesting among 32 physician leaders (which led to the addition of 1 question about number of years participating in patient care), the survey was sent to eligible respondents between September 19, 2017, and February 5, 2018. Recipients received up to 6 reminders to complete the survey. The protocol was deemed exempt by the University of Pennsylvania Institutional Review Board (protocol No. 827375). Respondents were not financially compensated for their participation.

Leaders’ clinical specialties were categorized according to the nature of their primary patient care responsibilities as procedural (obstetrics-gynecology, ophthalmology, surgery, urology), nonprocedural (emergency medicine, dermatology, family medicine, internal medicine, neurology, pediatrics, physical medicine and rehabilitation, psychiatry), or supportive (anesthesiology, pathology, radiology, radiation oncology).

We used several definitions to describe respondents’ perceptions about balance between clinical and leadership responsibilities. We defined balance as optimal (those spending an amount on patient care equivalent to the amount they deemed optimal from a leadership standpoint) or nonoptimal (all other situations). Among the latter group, we identified individuals for whom clinical burden (spending more time on patient care than was optimal from a leadership standpoint) contributed to nonoptimal balance. Additionally, we defined the feasibility of balancing leadership and clinical responsibilities as infeasible (if the optimal amount of time on patient care from a leadership perspective was less than the amount needed to maintain clinical competence) or feasible (all other situations). We used χ

 tests to compare categorical variables. All tests of significance were 2-tailed and deemed significant at α&thinsp;=&thinsp;.05. All analyses were performed using Stata version 15.1 (StataCorp).

Of 447 eligible respondents, 53% returned complete (n&thinsp;=&thinsp;233) or partial (n = 4) surveys. Nineteen percent of respondents were executive leaders, 58% were educational leaders, and 23% were clinical leaders. Executive leaders were more likely to come from nonprocedural clinical specialties (72% nonprocedural, 12% procedural, 16% supportive) relative to educational leaders (38% nonprocedural, 38% procedural, 24% supportive). Clinical leaders were primarily nonprocedural (90% nonprocedural vs 10% procedural and 0% supportive) (

&thinsp;<&thinsp;.001). Compared with nonrespondents, respondents did not differ with respect to gender, geographic distribution, leadership type, or clinical specialty (eAppendix B).

Overall, 84% of physicians reported participating in direct patient care. The majority (96%) reported that doing so somewhat or greatly improved their performance as leaders. Respondents noted a number of ways in which this occurred, including increased credibility among those they lead (98%) and insight into “frontline” issues (96%) (eAppendix C). Only 3% of physician leaders believed that direct patient care somewhat diminished their performance, citing reasons such as decreased time for leadership responsibilities. Of respondents not participating in direct patient care, the most commonly endorsed reason was “insufficient time or energy to be an effective clinician” (78%). None reported expectations from leadership as a reason for not participating in patient care.

Balancing Clinical and Leadership Responsibilities

Among physicians participating in direct patient care, most reported spending no more than a quarter of their time on it (74%) and that spending less than a quarter of their time was optimal from a leadership standpoint (80%). The majority (74%) reported achieving an optimal balance, with 15% citing clinical burden as contributing to nonoptimal balance. Only 4% reported that balance between clinical and leadership responsibilities was infeasible.

Perspectives varied by leadership type. Overall, executive leaders were less likely to participate in direct patient care (69%) relative to educational (90%) and clinical (85%) leaders (

). Among respondents participating in clinical care, executive leaders were also less likely to spend more than a quarter of their time on patient care or to believe that spending over a quarter of their time on patient care was optimal from a leadership standpoint (

). In contrast, rates of optimal balance or clinical burden and the reported feasibility of balancing clinical and leadership responsibilities did not vary by leadership type.

Participation in direct patient care varied by clinical specialty, ranging from 98% of leaders in procedural specialties to 82% of leaders in nonprocedural specialties and 76% of leaders in supportive specialties (

&thinsp;=&thinsp;.002) (Figure). The amount of and perceptions about time spent on patient care also varied by specialty type (Table). With regard to balance between clinical and leadership responsibilities, leaders in procedural specialties were least likely to achieve optimal balance (61.0% vs 78.4% and 82.1%; 

&thinsp;=&thinsp;.028) and most likely to report clinical burden (28.8% vs 9.9% and 7.1%; 

&thinsp;=&thinsp;.002) relative to leaders from nonprocedural and supportive specialties, respectively.

To our knowledge, this study is the first to document physician leader perspectives about the balance between their clinical and leadership responsibilities. It has 2 key takeaways.

First, our results suggest that many physicians serving a range of leadership roles find it feasible to balance clinical and leadership responsibilities. That many leaders believed patient care aided them in fulfilling their leadership responsibilities is informative to both physician leaders and organizations seeking to foster physician leadership.

 For these groups, it can be instructive to recognize that participation in patient care may be hampered more by insufficient time rather than external pressures or expectations.

Second, our findings suggest that fostering balance among physician leaders’ clinical and leadership responsibilities is not a one-size-fits-all task. In particular, it is relevant that perspectives and features of balance varied across leadership type and clinical specialty.

Poor balance—for instance, through infeasibility or the presence of clinical burden&mdash;can complicate physicians’ abilities to effectively meet either clinical or leadership responsibilities, much less both simultaneously. The fact that the extent and nature of imbalance can vary is exemplified by our finding that leaders from procedural specialties were more likely to report nonoptimal balance and clinical burden compared with leaders from other specialties.

Organizations can use insights about variation in the extent and nature of imbalance to tailor responsibilities and expectations for different types of leaders. For instance, organizations may require greater participation in patient care among clinical leaders, for whom clinical practice has more immediate relevance, than for executive leaders, whose primary responsibilities are often nonclinical (eg, emphasis on finance or operations

) and more removed from those of frontline clinicians. That perceptions about the feasibility of achieving balance did not vary by clinical specialty or leadership type is reassuring and underscores the ability of physician leaders from a range of clinical specialties and leadership types to achieve optimal balance.

First, the generalizability of our findings is limited by the 53% response rate. Second, although our study was descriptive in design and did not evaluate for associations between leadership behaviors and outcomes, the results represent useful early evidence that can be built upon in future work. Third, our sample of leaders may not be representative of the full breadth of physician leadership (eg, it may not capture mid-level leaders or those in other divisions not directly sampled in our study) or representative of physician leaders in all hospitals or health systems. However, we followed prior approaches in sampling leaders across a consistent set of leadership types among a group of institutions recognized for excellence across a set of clinical measures and outcomes.

 Fourth, we were unable to evaluate the presence and extent of institutional requirements, expectations, and minimum standards for physician leaders regarding patient care responsibilities. These factors should be evaluated in future work. Fifth, our study assessed self-reported perspectives among physicians but not other factors that should be studied in future research, including perspectives among those in nonleadership roles or objective measures of success (eg, staff retention, operational performance).

Our study provides important early evidence about physician leaders’ practices and perspectives balancing clinical and leadership responsibilities. Although most engage in some degree of direct patient care, what kind of balance they achieve varies by clinical specialty and leadership type. Collectively, our findings provide context and insight for health care organizations seeking to engage physician leaders from various backgrounds and optimize the value of their clinical perspectives amid ongoing health care transformation.

 Department of Medicine, Perelman School of Medicine, University of Pennsylvania (JWM, JAS), Philadelphia, PA; Institute for Health Policy, Management, and Evaluation, University of Toronto (ASD), Toronto, Ontario, Canada; Department of Medicine, University of Toronto (ASD), Toronto, Ontario, Canada; Department of Medicine, Sinai Health System and University Health Network (ASD), Toronto, Ontario, Canada; Department of Medicine, and Value and Systems Science Lab, University of Washington School of Medicine (JML), Seattle, WA.

 The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

 Concept and design (JWM, ASD, JAS, JML); acquisition of data (JWM, ASD); analysis and interpretation of data (JWM, ASD, JAS, JML); drafting of the manuscript (JWM, JML); critical revision of the manuscript for important intellectual content (JWM, ASD, JAS, JML); statistical analysis (JWM); provision of study materials or patients (JWM); administrative, technical, or logistic support (JWM); and supervision (JML).

 John W. Morgan, MD, Department of Medicine, Perelman School of Medicine, University of Pennsylvania, Blockley Hall, 13th Floor, 423 Guardian Dr, Philadelphia, PA 19104. Email: john.morgan@pennmedicine.upenn.edu.

1. Obama B. United States health care reform: progress to date and next steps. 

. 2016;316(5):525-532. doi:10.1001/jama.2016.9797

2. Dafny LS. Does CVS—Aetna spell the end of business as usual? 

. 2018;378(7):593-595. doi:10.1056/NEJMp1717137

3. Innovative mergers will disrupt health care. 

 Catalyst. Published February 16, 2018. Accessed May 28, 2018. https://web.archive.org/web/20181030200056/https:/catalyst.nejm.org/innovative-mergers-will-disrupt-health-care/

4. Liao JM, Emanuel EJ, Navathe AS. Six health care trends that will reshape the patient-provider dynamic. 

. 2016;4(3):148-150. doi:10.1016/j.hjdsi.2016.06.006

5. Angood P, Birk S. The value of physician leadership. 

6. Sarto F, Veronesi G. Clinical leadership and hospital performance: assessing the evidence base. 

. 2016;16(suppl 2):169. doi:10.1186/s12913-016-1395-5

7. Goodall AH. Physician-leaders and hospital performance: is there an association? 

. 2011;73(4):535-539. doi:10.1016/j.socscimed.2011.06.025

Articles

Although most physician leaders from 22 organizations saw patients and felt that it improved their leadership performance, their perceptions of the optimal balance of clinical and leadership responsibilities varied.

Physician Leaders’ Perspectives About Balancing Clinical and Leadership Responsibilities

When we think of medicine to treat diabetes, we typically think of pills and insulin injections to improve glycemic control. Evolving science is now pointing to a new medicine for diabetes and other chronic conditions: food. Although it is encouraging to offer alternatives to medications to individuals who face chronic conditions such as type 2 diabetes, many Americans are negatively affected by social determinants of health and do not have access to healthy foods. Existing concerns about comorbidities are heightened during the coronavirus disease 2019 pandemic, as those with diabetes who are diagnosed with the virus face particularly poor prognoses. This article addresses what employers need to know about food as medicine and offers practical steps to reduce barriers for their employee populations.

The American Journal of Accountable Care. 2020;8(2):34-37

When we think of medicine to treat diabetes, we typically think of pills and insulin injections to improve glycemic control. Evolving science is now pointing to a new medicine for diabetes and other chronic conditions: food. Paying attention to calories, fat, and carbohydrates is important to a person with diabetes, but food as medicine goes further. Affordable, nutritious food that is anti-inflammatory, is aligned with an individual’s genetic makeup and microbiome, and complements rather than conflicts with medication not only affects a person’s future health but also affects ongoing diabetes treatment and outcomes.

In helping individuals prevent and manage diabetes, we need to consider food as medicine in addition to the more standard treatment modalities. Unfortunately, 49 million people in United States lack access to affordable, nutritious food, which is a major cause for concern given the impact of food on both the prevention and management of diabetes.

 It is difficult to use food as medicine when barriers exist to nutritious food. These barriers are part of what we call the social determinants of health (SDOH), which, as defined by the World Health Organization, include the conditions in which people are born, grow, live, work, and age.

 These factors can affect our health and health care outside of the traditional treatment environment.

Food insecurity is defined as when a person cannot meet their dietary needs on a consistent basis regardless of socioeconomic status or employment status.

 This factor and other SDOH are not just an issue for those of low socioeconomic status. In fact, approximately 50% of those with food insecurity are employed. Individuals experiencing food insecurity often consume a nutrient-poor diet, which may contribute to the development of obesity, heart disease, hypertension, diabetes, and other chronic diseases.

 People who live in food-insecure households also have difficulties in managing diet-related chronic conditions.

Food insecurity can be due to a variety of factors including reliable and affordable transportation, mobility issues, inconvenient access to healthy foods, financial struggles, lack of time, and inadequate cooking expertise. These needs are fluid and may change over time. Studies about food insecurity have shown that more than 80% of those who cannot routinely meet their dietary needs will need varying levels of nutrition support over a 12-month time frame.

 People with type 2 diabetes may find themselves limited to purchasing inexpensive, high-calorie, nutritionally poor foods, including foods high in refined carbohydrates, instead of items that are more healthful, such as vegetables, lean proteins, and whole grains.

 During the coronavirus disease 2019 (COVID-19) pandemic, millions face temporary or permanent job loss and more have sought food assistance, exacerbating the documented 2-way connection between diabetes and COVID-19, which can contribute to the worsening of glucose control.

So, how can employers realistically have an impact on SDOH, including food insecurity? There are small steps employers can take, as well as many public and community resources available to positively affect the workforce.

Take for example, a midsize manufacturer located in the Midwest with a diverse, predominantly male employee population in which a large percentage ranges from having prediabetes to having uncontrolled diabetes. This employer realized that despite offering healthy food options in the onsite cafeteria, employees routinely left campus to buy food across the street at a local ethnic grocery store. Instead of fighting the competition, the employer partnered with the local grocer to offer greater accessibility to affordable nutritious foods by launching its own healthy meal preparation program.

This service provides a once-a-week healthy meal that serves a family of 4 for under $20. Employees can choose from 3 healthy meal options—they receive the meal fixings and recipes, all packaged and ready to take home at the end of the workday. The employer works with an onsite wellness provider to develop weekly menus/recipes, and the onsite cafeteria features a healthy meal each week. In this example, participation in the program grew through word of mouth, a comprehensive communications campaign, and weekly raffle prizes for participants.

Moving the needle on SDOH requires instituting policies and programs that have “evidence of effectiveness” to improve health factors that lead to improved outcomes. But what does “evidence of effectiveness” really mean? Employers can use County Health Rankings & Roadmaps

 as a resource to understand the challenges faced in their communities and to launch initiatives that are scientifically proven to be effective, which is important, given the limited time and resources that employers have available to devote to workplace health (

These rankings are based on a model of population health that considers many factors that, if improved, can help make communities healthier places to live, learn, work, and play. Strategies are generally assessed in terms of their effect on health factors, those things we can modify to improve length and quality of life, rather than on health directly. For example, employers may not be able to directly impact diabetes, but they can provide their employees with ample time for lunch and snack breaks and affordable healthy foods; these efforts may lead to better nutrition and an improvement in health factors that include glycated hemoglobin (A1C) values, stress, and body weight, which will ultimately help employees manage diabetes more effectively.

 strategies are assigned an evidence rating based on the quantity, quality, and findings of relevant research. The most effective strategies get a rating of “scientifically supported” and those with insufficient results receive an “evidence of ineffectiveness” rating (

). Employers who seek the most “bang for their buck” can use these ratings to prioritize workplace initiatives.

For example, a multicomponent obesity intervention led by a trained professional and competitive pricing of healthy foods in the employee cafeteria are most likely to make a difference—both are “scientifically supported.” Other worksite strategies have “some evidence” and are likely to work but need further research, including providing fruit and vegetable taste testing, displaying point-of-purchase prompts for healthy foods, including nutrition labels on cafeteria foods, offering fresh fruits and vegetables at a workplace farmers’ market, providing incentives for adopting healthy behaviors, and taxing sugar-sweetened beverages and unhealthy snacks.

Other interventions that are recommended by credible experts but have limited research documenting their impact include community weight loss challenges and offering healthy foods at catered events—these are given an “expert opinion” rating. On the far end of the spectrum, offering nutrition education and/or cooking classes through a community kitchen and providing access to an online grocery store to purchase healthy foods are given an “insufficient evidence” rating. Unless an organization has unlimited funds for nutrition interventions, programs on this end of the scale will likely not have a significant impact on health factors, including reversing the course of prediabetes and helping those with diabetes manage their condition.

Recently, new initiatives have been launched to meet people where they live and shop, such as St Joseph Hoag Health’s “Shop with Your Doc” program in California.

 This program stations doctors, nurses, and registered dietitians in local grocery stores to provide right-on-time advice and consultation on a variety of health topics, including how to read a label and what it means to have prediabetes.

We have known for a long time that food can affect blood sugar, which is why employers often provide health plan coverage for diabetes educators and/or diabetes-focused nutritionists. This modality of nutrition support has been based on population health models but is not personalized at the individual level. What is evolving is the nutritional science, which shows that food can affect high blood sugar and A1C in ways that were not known in the past.

The unique nutritional needs of a person with diabetes depend on several factors. Blood sugar control through precision-based nutrition based on an individual’s microbiome has shown great promise. Microbiomes in humans can be found in the gastrointestinal tract, skin, nose, and other areas of the body. The microbiomes in our gastrointestinal tract influence how our bodies respond to stress, how we digest food, and the health of our immune systems. Positive results have been realized by creating a nutritional plan that will create and support healthy bacteria in the gut, allowing it to break down foods in a manner that reduces and controls blood sugar spikes. This has been shown to allow people with diabetes to reduce their medication load, and it can potentially put prediabetes into remission as well.

Even hospitals that have historically relied on traditional medications to treat chronic health conditions like diabetes have gotten into a “food as medicine” frame of mind. For example, Zuckerberg San Francisco General Hospital recently launched the Therapeutic Food Pantry. Through this clinically based prescription food program, providers prescribe fresh fruits and vegetables as well as other foods aligned to an individual’s specific therapeutic and microbiome needs.

 Upon arrival at the pantry, patients receive nutrition education, food demonstrations, and referrals to other local resources, and then walk away with approximately 25 pounds of food that aligns with their disease state and health goals.

By seeking opportunities such as this, employers can establish partnerships with local health care organizations and other community agencies to make nutritious, affordable foods a convenient reality for all employees, regardless of their socioeconomic status. Many innovative employers have established such collaboratives that are a win for the employee, the employer, and the community. An example is Kraft Foods, which has partnered with the Chicago Botanic Garden to plant and cultivate vegetable and herb gardens on company grounds. The impact of this initiative is multifaceted, including teaching employees how to plant fruits and vegetables, educating employees on healthy eating, making the harvest available within the company cafeteria, and supporting local food banks with locally grown food.

At the end of the day, it is up to an individual to decide what to eat, when to eat, and how much to eat. Employers can and should support healthy nutrition both within the corporate space and within the community. Only when the full ecosystem is on board will “food as medicine” help reverse the growing trend of diabetes, which not only costs employers significantly but also diminishes quality of life for the more than 34.2 million Americans with this chronic condition.

Midwest Business Group on Health (JEB, DG, DW), Chicago, IL.

Concept and design (JEB, DG, DW);  drafting of the manuscript (JEB, DG, DW); critical revision of the manuscript for important intellectual content (JEB, DG, DW); administrative, technical, or logistic support (JEB); and supervision (DG).

Jan E. Berger, MD, MJ, Midwest Business Group on Health, 125 S Wacker Dr, Ste 1350, Chicago, IL 60606. Email: info@mbgh.org.

1. Coleman-Jensen A, Rabbitt MP, Gregory CA, Singh A. Household food security in the United States in 2018. US Department of Agriculture Economic Research Service. September 27, 2019. Accessed June 1, 2020. https://www.ers.usda.gov/publications/pub-details/?pubid=94848

2. About social determinants of health. World Health Organization. Accessed September 20, 2019. https://www.who.int/social_determinants/sdh_definition/en/

3. Understand food insecurity. Feeding America. Accessed September 25, 2019. https://hungerandhealth.feedingamerica.org/understand-food-insecurity/

4. Heerman WJ, Wallston KA, Osborn CY, et al. Food insecurity is associated with diabetes self-care behaviours and glycemic control. 

. 2016;33(6):844-850. doi:10.1111/dme.12896

5. Herman D, Afulani P, Coleman-Jensen A, Harrison GG. Food insecurity and cost-related medication underuse among nonelderly adults in a nationally representative sample. 

. 2015;105(10):e48-e59. doi:10.2105/AJPH.2015.302712

6. Seligman HK, Schillinger D. Hunger and socioeconomic disparities in chronic disease. 

. 2010;363(1):6-9. doi:10.1056/NEJMp1000072

7. Pal R, Bhadada SK. COVID-19 and diabetes mellitus: an unholy interaction of two pandemics. 

. 2020;14(4):513-517. doi:10.1016/j.dsx.2020.04.049

8. Strategies. County Health Rankings & Roadmaps. Accessed September 20, 2019. https://www.countyhealthrankings.org/take-action-to-improve-health/what-works-for-health/strategies

9. County health rankings model. County Health Rankings & Roadmaps. Accessed June 1, 2020. https://www.countyhealthrankings.org/resources/county-health-rankings-model

10. Our ratings: evidence rating, disparity rating. County Health Rankings & Roadmaps. Accessed June 1, 2020. https://www.countyhealthrankings.org/take-action-to-improve-health/what-works-for-health/our-ratings

11. Gorn D. Food as medicine: it’s not just a fringe idea anymore. National Public Radio. January 17, 2017. Accessed June 1, 2020. https://www.npr.org/sections/thesalt/2017/01/17/509520895/food-as-medicine-it-s-not-just-a-fringe-idea-anymore

12. SFGH Therapeutic Food Pantry. San Francisco Health Network. Accessed June 1, 2020. https://sfhealthnetwork.org/sfghfoodpantry/

13. National Diabetes Statistics Report, 2020. CDC. Accessed June 1, 2020. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf

This article addresses what employers need to know about food as medicine and offers practical steps to reduce barriers for their employee populations.

Food as Medicine: Impact on Diabetes Management

The Delivery System Reform Incentive Payment (DSRIP) program was designed to move hospitals toward more effective and accountable care for their Medicaid and charity care patients.

Assessing potential changes among DSRIP-participating hospitals in community partnerships, data analytic capabilities related to quality measurement, and engagement in the DSRIP model may provide information for future value-based payment strategies.

 Statistical analysis of linked hospital data from 2 web surveys, supplemented with findings from 2 rounds of key informant interviews.

Data from 2 New Jersey hospital web surveys in 2015 and 2018 were analyzed to assess changes in hospital activities, practices, and perceptions related to the DSRIP program. Survey measures included the number of partnerships that hospitals developed with community organizations and outpatient providers, hospitals’ ability to share patient data with outpatient partners, hospitals’ analytic capability to leverage patient data for quality assessment, and perceptions of hospital representatives of the value of their DSRIP projects. Stakeholder feedback from key informant interviews conducted in 2014 and 2017 provided context for survey findings.

Hospital partnerships with community providers did not grow significantly over the DSRIP implementation period, which may be due to constraints in program design. Hospital capacity for collection of data to fulfill DSRIP reporting requirements increased over the study period. Data exchanges with outpatient partners facilitated use of rapid-cycle evaluation tools, and the value of data sharing for improving quality of care and population health was perceived more positively over time.

The perceived benefit of DSRIP project activities overall, for patient access to care and overall health, also increased between the 2 surveys.

In New Jersey, available evidence suggests that DSRIP was successful in catalyzing many aspects of a hospital industry transformation toward data-driven population health management. These findings can inform policy makers’ decisions about how to structure future initiatives to help hospitals progress further in delivering value-based care.

The American Journal of Accountable Care. 2020;8(2):4-12

The Delivery System Reform Incentive Payment (DSRIP) program, a 2010 federal initiative negotiated between states and CMS under section 1115 waiver authority, provides a financing mechanism for providers to transform care for beneficiaries of Medicaid, the Children’s Health Insurance Program, and charity care.

 By tying funds to measurable quality outcomes, DSRIP establishes incentives for providers (mainly hospitals) to proactively manage the health of their low-income patients, making these providers regional anchors in the Triple Aim quest for better care, better health, and lower costs. DSRIP or DSRIP-like programs have been implemented in 12 states to date and vary in size, reach, provider participation, and funding pools.

 While the original program structure is evolving, the concepts underlying DSRIP are poised to be an integral part of the future Medicaid delivery system landscape as value-based payment structures grow in prominence for health care providers.

Although some early assessments of DSRIP’s impact on patient outcomes and population health indicators are emerging,

 opportunities to learn from the implementation experiences of participating providers are more quickly available. Understanding these dynamics can yield valuable information to inform framing and implementation of future policies.

 Some available evidence includes case studies that discuss DSRIP program design and lessons learned from implementation in several states.

 These studies demonstrate the positive role that DSRIP played in providing resources and harnessing organizational motivation for improving provider data capabilities and building effective community partnerships.

 Providing incentives for these changes was a core DSRIP objective because sustainable, cost-effective health improvements for low-income populations are predicated on integrated and accountable delivery systems.

 Consequently, assessing progress made by providers in building infrastructure to support redesigned processes of care delivery is an important metric for evaluating DSRIP’s impact, yet this has not been studied in a systematic way.

In addition, DSRIP promotes engaging community and social services organizations in improving population health by providing funding for services that are not traditionally reimbursed for Medicaid patients. Under DSRIP and similar projects in other states, providers have partnered with such organizations to identify and connect patients directly to community resources addressing social needs.

 The program encouraged hospitals to address patient needs as far “upstream” as possible, shifting their efforts to innovative population health missions, for example, by developing home interventions for children with asthma or arranging transportation for patients with diabetes to aerobics classes. An assessment of hospitals’ perceptions of the effectiveness of the DSRIP model is therefore important for understanding hospital engagement in this reoriented model of patient care.

This study utilizes quantitative and qualitative information about the New Jersey program to examine hospital operational changes brought about by DSRIP and perceptions of hospital leaders about the value of these changes for improving patient health. Specifically, we describe the impact of DSRIP on New Jersey hospitals in terms of (1) increasing collaboration with clinical and community partners, (2) improving data infrastructure and analytic capabilities for meeting data reporting requirements, and (3) strengthening the perceived value and effectiveness of patient care models structured for population health management. New Jersey provides a good venue for such a study because all hospitals in the state were eligible to participate in the program, not just the safety net hospitals (SNHs) with a high proportion of Medicaid and uninsured patients.

DSRIP utilized a pay-for-performance (P4P)/reporting system to encourage hospitals to adopt a population health management approach to improving care for New Jersey’s low-income population. Hospitals had to utilize data to identify a chronic condition relevant for their patient population, adopt a care management project, form partnerships with community organizations and outpatient providers to help establish a continuum of care, and specify relevant validated outcome metrics so that the impact of DSRIP activities on patient and population health could be measured.

 Elements of hospital projects such as improved case management, discharge planning, patient education, and a focus on outcome measurement reflect the population health management principles embedded in DSRIP projects.

After state and CMS approval of care management projects, New Jersey DSRIP-participating hospitals took part in 4 overlapping program stages with defined activities determining incentive payments across demonstration years (DYs): developing technology, tools, and human resources infrastructure (stage 1); piloting, testing, and implementing a chronic disease care project (stage 2); achieving project-based quality improvements and reporting quality metrics capturing the impact of DSRIP projects (stage 3); and reporting prespecified population-focused quality metrics (stage 4). An important stage 1 activity involved engaging project partners, such as schools, doctor’s offices, or federally qualified health centers, to support hospitals with their DSRIP projects. Stage 3 and 4 activities required hospitals to report on several project-specific and population health—related quality metrics for an attributed population of Medicaid and charity care patients. Whereas metrics based on administrative data were prepared by the state, those based on paper or electronic health records (EHRs) were calculated by hospitals, sometimes in coordination with outpatient partners. Collaboration for such clinical data sharing was thus strongly encouraged under DSRIP, and those project partners sharing data with the hospital were designated as data reporting partners. As the program progressed, a larger share of payments was tied to P4P, requiring improvements in project-related outcomes over baseline for hospitals’ attributed population. Hospitals could also receive universal performance pool (UPP) payments for maintaining quality of care in areas outside their project focus.

Leveraging patient data to guide care and assess outcomes was an integral component of the program’s data-driven population health orientation. Data related to patient utilization were used to assess key health challenges in an area, identify patient populations for directing interventions, and calculate metrics for determining incentive payments. Data reporting partnerships, in particular, were expected to promote care integration and shared accountability between hospitals and community health care providers. These data sharing relationships were seen as potentially increasing the effectiveness of hospitals’ rapid-cycle improvement tools by providing more information on patients’ progress toward health improvement goals.

We use data collected from 2 hospital web surveys administered as part of the New Jersey DSRIP evaluation that were fielded during April 2015 (DY3), shortly after implementation of hospitals’ DSRIP projects, and then again 3 years later, after the end of the first round of the demonstration in February 2018 (DY6). The first survey was sent to managers at all DSRIP-eligible hospitals in the state (n

64), regardless of whether they participated in the program, and the response rate was 65%. The second survey was sent only to hospitals that participated in the DSRIP program (n

To examine the effects of DSRIP on hospitals over time, our statistical analysis compares data linked over both surveys (n

22). We used repeated-measures analysis of variance to examine changes in data infrastructure capabilities, community partnerships, and perceptions of the value of the DSRIP program in improving patient health. We also present cross-sectional results from some questions asked only on the second survey. Pairwise deletion was used to deal with item nonresponse.

To determine whether hospitals included in the paired analytic sample differ from those excluded, we compare hospital safety net status and DSRIP performance data for these 2 groups of hospitals using unpaired 

 tests. We define SNH status based on receipt of Hospital Relief Subsidy Fund dollars in 2011. Performance data are from payment summaries posted to the DSRIP website detailing the number of stage 3, stage 4, and UPP quality metrics for which hospitals were awarded payment out of the total attempted.

 All analyses were done in SPSS with α set at 0.05. 

 values are reported in the results when differences were significant.

To provide context for our findings, we draw on 2 rounds of key informant interviews (KIIs) with DSRIP stakeholders; they include selected DSRIP personnel in participating hospitals, DSRIP program advisory committee members, officials from the New Jersey Department of Health, outpatient providers who had partnered with DSRIP hospitals, and hospital industry association representatives. We selected stakeholders who we believed were best equipped to answer questions on DSRIP’s impact, potential, and sustainability. The first round of interviews was conducted from October 2014 to February 2015 and consisted of 12 interviews with 13 key informants. The second round was conducted from October to December 2017 and consisted of 10 interviews with 29 key informants. Semistructured interview guides were used in both rounds. Research team members independently analyzed the interviews and came to a consensus on core themes.

This research was approved by the institutional review board of the authors’ home institution, Rutgers, The State University of New Jersey, and informed consent was obtained from all participants.

Of the 64 general acute care hospitals in New Jersey, 9 (14.1%) elected not to participate in DSRIP and 6 (9.4%) dropped out during the implementation period. None of these 15 hospitals were SNHs. Of the 49 hospitals that participated in both rounds of DSRIP, 22 (44.9%) responded to both surveys and were included in the paired sample. The remaining 27 either did not participate in 1 or both survey rounds or were unlinkable due to inadequate identifying information. SNHs were overrepresented in our paired sample compared with DSRIP-participating hospitals not in our analysis (64% vs 44% SNH, respectively). We further examined differences in hospital performance by survey participation that may reflect differences in hospital experience and engagement. We calculated the percentage of stage 3 and stage 4 metrics for which payment incentives were earned in DY5 and found no statistically significant differences; however, UPP achievement in DY5 was slightly lower in our paired sample than among excluded hospitals (62% vs 69%), although this difference fell short of statistical significance (

Relationships With Clinical and Community Partners

Hospital informants expressed some uncertainty about program requirements for clinical and community partners during the first round of interviews, possibly explaining limited hospital engagement with partners early on. The survey indicated that hospital partnerships did increase slightly over the DSRIP implementation period. On average, hospitals had 4.7 project partners for their DSRIP programs in the beginning of the implementation period, and this increased to 5.0 by DY6 (

). The strategy of recruiting physician practices as partners doubled over the study period, from 24% of hospitals to 48% of hospitals, and in DY6, physician practices were the most common partner type reported by hospitals, with an average of 1.7 practices per hospital (Table 2). An increasing proportion of survey respondents reported recruiting other (nonclinical) community organizations as partners, going from just under 40% in the first survey to about 60% in the second survey.

The average number of data reporting partners did not increase over the study period. However, we observe an increase in the average number of organizations with which hospitals would have liked to partner for data sharing but could not because of the complexity of data sharing requirements or because a potential partner was already in a data sharing relationship with another hospital (Table 2). The DSRIP program did not provide funding for the extra work required of data reporting partners, thus constraining hospitals’ ability to forge partnerships, according to the second round of KIIs. Finally, it is worth noting that a majority of hospitals surveyed (more than 75% in both rounds) were working with outpatient partners before DSRIP. Of those with existing clinical partners, approximately 70% felt that DSRIP had strengthened these relationships and none felt those relationships were weakened.

Both rounds of KIIs revealed that preparing the chart-based stage 3 and stage 4 metrics was a challenge for all hospitals, particularly those lacking sophisticated EHR systems. Some stakeholders described the considerable efforts made to extract the data needed to comply with reporting requirements. These learning experiences may also explain improvements in the analytic capabilities of hospitals in our sample over the DSRIP implementation period. Hospitals reported that a higher percentage of stage 4 inpatient or emergency department chart-based metrics were obtainable from EHRs in the second survey than in the first (

). Hospitals also reported less difficulty in the second survey than in the first in collection of inpatient and outpatient metrics to fulfill reporting requirements (

Changes in the data capabilities of outpatient reporting partners were less evident, although this is based on the smaller sample of responding hospitals with a data reporting partner. The overall average number of reporting partners with an interoperable EHR did increase slightly from the first survey to the second, and the percentage of stage 4 metrics obtainable from reporting partners’ EHRs increased over this time period. Finally, most hospitals were using rapid-cycle evaluation tools (≥90%), and the percentage of hospitals agreeing that this was facilitated by real-time data exchanges with partners increased from 11% to 42% (

Survey results showed that hospital respondents’ perceptions of DSRIP’s value and effectiveness were generally positive, and some indicators improved significantly with DSRIP implementation. Hospitals’ reasons for applying for and remaining in the DSRIP program were almost all unchanged, except that DSRIP was increasingly viewed as an opportunity for more financial resources (

). In the first survey, hospital respondents felt that disease management programs under DSRIP had the greatest impact on quality of care and population health outcomes, and this further increased in the second survey. As the program progressed, hospital respondents reported a more positive assessment of the various DSRIP-prescribed activities in improving quality of care and population health outcomes (Table 4). They increasingly saw the value of sharing data with reporting partners, reporting stage 4 metrics, and using rapid-cycle assessment and improvement tools; by the second survey, every aspect was rated as having at least a moderately positive impact, except for reporting on stage 4 metrics. These findings are consistent with both KII rounds in which interviewees expressed enthusiasm for the chronic disease management aspect of DSRIP but saw little value in reporting measures beyond those related to their specific intervention.

The survey assessed whether hospitals regarded DSRIP activities as improving access to care, quality of care, and health, and we observed improvement in perceptions in some areas. In the first survey, hospital respondents perceived that their DSRIP activities improved patient access to health services and patient health only modestly, if at all, with an average rating just shy of “some improvement.” In the second survey, respondents attributed at least “some improvement” in all queried aspects of patient health and care to the beneficial effect of DSRIP activities. Among these, DSRIP’s effects on patient access and overall patient health were statistically significant (

The second survey further assessed hospital respondent impressions of DSRIP’s overall value and effectiveness in 5 areas: program incentives, payment methodologies, quality metrics, disease management, and community partnerships. The average rating of these statements indicated modest agreement. This was corroborated by findings from the second round of KIIs, in which some of these positive perceptions were also communicated by stakeholders. Hospital informants felt that their DSRIP initiatives underscored the importance of connecting with the community outside the hospital, although progress was sometimes slow because the outpatient component was not emphasized from the beginning of the program. Hospital informants mentioned the value in being able to use DSRIP funds flexibly to pay for services not traditionally reimbursed. By the second set of KIIs, informants were also expressing positive effects on health outcomes from their interventions.

New Jersey was 1 of 12 states implementing a DSRIP or DSRIP-like program between 2010 and 2018, and we have highlighted salient ways in which New Jersey hospitals responded to DSRIP over the first demonstration period. Our findings suggest small, positive improvements in partnership relationship building, data capabilities, and perceived impact on patient care and population health. We found statistically significant increases in hospitals’ capability for quality metric reporting and in belief by respondents that their DSRIP programs were improving patient access and health. Our results also point to areas where there is room for improvement in the design of incentive-based delivery system reforms.

Our surveys revealed that although hospitals had, on average, 4 to 5 project partners from the time of project initiation, there was not significant growth in such partnerships over the DSRIP implementation period. Nevertheless, the majority of hospitals agreed that DSRIP fostered community partnerships that had a positive impact on social determinants of health. As for data reporting partnerships, aspects of DSRIP’s design likely limited growth in these partnerships from reaching their full potential. Foremost among them was the lack of financial resources to compensate outpatient providers for the additional work around data collection or for the ongoing reporting that would be required of them. Future DSRIP and DSRIP-like programs should be designed with an awareness that partners may require external funding to enable their participation, a lesson also echoed in an examination of DSRIP in New York.

 Some of the administrative restrictions that limited outpatient collaboration to just 1 hospital may have also impeded new partnerships.

Our analysis revealed increased hospital ability and greater ease over time in extracting quality metric data from EHRs. Although this may reflect in part an industry-wide trend of improved EHR sophistication, it also suggests that DSRIP spurred growth in data and analytic capabilities for participating hospitals. Moreover, perception of the positive impact of data sharing on quality of care and health outcomes improved as the program progressed. It seems that as data sophistication and analytic capabilities grew, so too did appreciation of their value in enhancing care.

Hospitals’ survey responses indicated that their belief in the value and effectiveness of the DSRIP program in improving access to care and health increased over the implementation period. This is notable given that DSRIP represented a major transition point for the hospital industry to value-based payment arrangements under Medicaid. Notwithstanding this positive outlook, the survey indicated considerable room for improvement in the operationalization of P4P structure. Hospitals respondents only “somewhat agreed” that DSRIP used payment methodologies that “fairly” incentivized hospitals’ investments in chronic disease management processes. Also, the importance of DSRIP as a source of financial resources increased from the first survey to the second. This could indicate greater financial pressure among DSRIP hospitals following investment in their chronic disease management programs and the first P4P and UPP results. Although these findings may be driven in part by the overrepresentation of more resource-constrained SNHs in our sample and hospitals with slightly lower UPP payments, they are also consistent with the findings from KIIs that were based on the overall hospital industry.

This study is the first to quantify changes in hospital practices and perceptions using repeated survey data on DSRIP implementation experiences supplemented by broad stakeholder KIIs. Although our statistical power was limited due to nonresponse and the inability to link data for some hospitals, the findings provide insights on the association between DSRIP program features and the observed changes in hospital operations around patient care and population health improvement. Without similar information for nonparticipating hospitals, we cannot draw inferences about the extent to which DSRIP may have contributed to the changes we observe, particularly in the growth of analytic capabilities, which may reflect an industry-wide trend. We conducted a sensitivity analysis to determine if there was a dose-response relationship between the DSRIP funding level of each hospital and the changes in analytic capabilities, but we did not find a clear pattern. Additionally, caution should be used in extrapolating our findings to all hospitals, as SNHs were overrepresented in our analytic sample. Still, these hospitals are an important target of the DSRIP, so their strong response rate is a strength of our analysis.

The DSRIP program was designed to orient hospitals to population health management using a data-driven approach to identifying community health needs, delivering targeted interventions using evidence-based models of care, and promoting an accountable payment structure based on hospital quality performance and reporting. In New Jersey, the program was successful in catalyzing many aspects of this industry transformation. Processes around data collection for quality metric reporting became better established. Exchanging data with outpatient partners has facilitated use of rapid-cycle evaluation tools, and hospital respondents reported that data sharing and DSRIP activities have positively affected the patients and communities that they serve. Results also point to areas where improvements could be made in the operationalization of a P4P structure.

The authors are grateful to officials at the New Jersey Department of Human Services and the New Jersey Department of Health for contextual information relating to the DSRIP program and to Bram Poquette, MLIS, for assistance in preparing this manuscript. We also appreciate the helpful comments and suggestions provided by the editor and anonymous reviewers of this manuscript.

Center for State Health Policy, Institute for Health, Health Care Policy and Aging Research (KL, SC, SB, JF, JCC), and Edward J. Bloustein School of Planning and Public Policy (JCC), Rutgers University, New Brunswick, NJ.

Funding for this work was provided by the New Jersey Department of Human Services (NJDHS) and the Robert Wood Johnson Foundation (RWJF). This work was prepared for NJDHS. Any opinions expressed in this report are those of the authors and do not necessarily represent the view of NJDHS or RWJF.

Concept and design (KL, SC); acquisition of data (KL, SC, SB, JF); analysis and interpretation of data (KL, SC, SB, JF, JCC); drafting of the manuscript (KL, SC, JF, JCC); critical revision of the manuscript for important intellectual content (KL, SC, SB, JF, JCC); statistical analysis (KL, SC, SB); administrative, technical, or logistic support (KL, JF); and supervision (SC, JCC).

Sujoy Chakravarty, PhD, Center for State Health Policy, Institute for Health, Health Care Policy and Aging Research, Rutgers University, 112 Paterson St, 5th Floor, New Brunswick, NJ 08901. Email: schakravarty@ifh.rutgers.edu.

1. Heider F, Kartika T, Rosenthal J. Exploration of the evolving federal and state promise of Delivery System Reform Incentive Payment (DSRIP) and similar programs. Medicaid and CHIP Payment and Access Commission. August 2017. Accessed January 23, 2019. https://www.macpac.gov/wp-content/uploads/2018/03/Exploration-of-the-Evolving-Promise-of-DSRIP-and-Similar-Programs.pdf

2. Delivery System Reform Incentive Payment programs. Medicaid and CHIP Payment and Access Commission. April 2020. Accessed May 14, 2020. https://www.macpac.gov/wp-content/uploads/2018/03/Delivery-System-Reform-Incentive-Payment-Programs.pdf

This study examines the New Jersey Delivery System Reform Incentive Payment (DSRIP) program using hospital web surveys and key informant interviews and finds progress toward data-driven population health management for low-income patients.

Hospital Responses to DSRIP Program Reforms in New Jersey

The American Journal of Accountable Care. 2020;8(2):32-33

At least 12 million adults and half a million children in the United States are living with a serious illness.

 A serious illness is a health condition that carries a high risk of mortality and commonly affects a patient for several years, such as metastatic cancer, heart failure, chronic obstructive pulmonary disease, and dementia.

 As the authors of “Gauging the Experiences of the Seriously Ill in California: Analyzing Serious Illness Beyond Medicare Fee-for Service” have highlighted in the current issue,

 individuals living with a serious illness are the disproportionate users of our health care system: 911 calls, emergency department visits, repeated hospitalizations, and polypharmacy occur 4 to 5 times more frequently in this population. Yet despite high utilization and spending, those with serious illnesses often receive low-value—even distressing&mdash;service from our health care system.

Symptoms are often inadequately addressed. For example, 64% of people with metastatic cancer report high levels of pain, despite being under the active care of an oncology team.

Communication and shared decision making are insufficient. For example, fewer than 50% of hospitalized patients with a serious illness were ever asked what their personal preferences would be if a critical situation should arise,

 and only 47% of discharged patients report fully understanding their discharge instructions.

The outsized responsibilities placed on family caregivers are taking an enormous toll. More than 1 in 5 cite adverse impacts on their own health,

 and formal studies conclude that caregivers face a 63% increased risk of mortality.

Fortunately, accountable health care organizations are increasingly recognizing that the care of individuals with a serious illness represents a prime opportunity for improvement. Value-driven providers, accountable care organizations, Medicare Advantage plans, and others are implementing effective strategies that improve quality of life while decreasing costs.

These strategies entail first identifying the population living with serious illnesses early enough to impact the course of their trajectory; the article in this issue provides important insight into a useful identification process. Once the population is identified, clinicians with appropriate training should work to clarify patients’ goals of care and assess both symptom and caregiver burden. Nurse care managers are often best situated to these conversations and can then work with the treating teams to modify care plans accordingly. For those with high levels of need, specialty palliative care teams should consult or comanage the care, including providing home-based care when appropriate.

Together, building early identification processes, training key clinicians in communication and symptom management skills, and ensuring access to specialty palliative care when warranted make a dramatic difference in both the quality and the cost of care for individuals living with a serious illness.

Accountable health care organizations are increasingly recognizing that the care of individuals with serious illness represents a prime opportunity for improvement.

June 2020

Physician Leaders’ Perspectives About Balancing Clinical and Leadership Responsibilities

Food as Medicine: Impact on Diabetes Management

Hospital Responses to DSRIP Program Reforms in New Jersey

Serious Illness: A High Priority for Accountable Care

Gauging the Experiences of the Seriously Ill in California: Analyzing Serious Illness Beyond Medicare Fee-for-Service

Accountable Care Organizations and Patient-Centered Medical Homes: Health Expenditures and Health Services