Social determinants of health (SDoH) refer to the social conditions in which people are born, grow, work, live, and age that may affect their health, including the wider set of forces and systems shaping the conditions of daily life.1 A growing body of literature has indicated that SDoH are strongly associated with health care quality, utilization, and patient outcomes.2–5 For example, it is estimated that social factors account for 25%–60% of deaths in the United States.6,7 In the era of population health management and value-based payment for health care, there is a growing interest in capturing, aggregating, and analyzing SDoH data among policy makers, researchers, and providers.8–11 Current initiatives regarding SDoH include accounting for social factors in risk adjustment formulas to reimburse providers,12,13 incorporating SDoH in health care utilization and outcome predictions,14,15 and addressing social determinants to improve quality and reduce unnecessary utilization of health care.11,16
Social determinants can be broadly divided into 2 categories: individual level factors that measure the social conditions specific to a patient, such as the patient’s education, income, and housing conditions; and community-level factors that measure environmental or neighborhood characteristics, such as unemployment rate and poverty rate in the community.9,17 Patient-level social determinant information (eg, individual income and occupation) is generally not collected in a systematic manner during clinical care. Therefore, health systems and other stakeholders are increasingly leveraging community-level SDoH data to support health interventions.10,11 Incorporating community social conditions into decision-making could enable context-informed care that meaningfully accounts for neighborhood conditions that may affect patients’ health.
With the rapid expansion of pay-for-performance programs in the US health care market, providers, payers, and other stakeholders are increasingly investing in efforts to address or accommodate social needs and vulnerable social conditions to improve the value of health care. The Centers for Medicare and Medicaid Services has launched the Accountable Health Communities Model to examine whether identifying and addressing social needs will impact health care costs and reduce unnecessary health care utilization.18 Some researchers also indicated the importance of SDoH for high-cost patients, the subset of patients with high health care utilization.19 In addition, there are active debates regarding adjusting social risk factors for payment to compensate providers disproportionately serving socially vulnerable patients.20
To date, limited evidence is available regarding how SDoH influence total health care utilization or costs. Although studies have increasingly indicated that disadvantaged social conditions are associated with poor clinical outcomes and hospital readmissions,3,21 there has been little examination of other types of potentially preventable health care utilization [eg, preventable emergency department (ED) visits and preventable hospitalizations] and costs associated with preventable utilization. These utilization measures have been included in many federal and state pay-for-performance programs.22–24 Understanding the association of SDoH with total and preventable costs will be essential for health systems to target interventions for social risk factors to generate cost savings.
We examined the relationship of neighborhood social conditions with total annual and preventable health care utilization and costs among a group of Medicare patients. We used the Area Deprivation Index3 (ADI) as of a measure of community social conditions at the US census block group level. Findings of this study may inform efforts to address social determinants and improve the value of care for patients with vulnerable social conditions.
Study Population and Sample
We identified patients enrolled in Medicare fee-for-service (FFS) program or dually enrolled in Medicare and Medicaid programs during 2013 and 2014 from 6 health systems affiliated with INSIGHT. INSIGHT, previously known as the New York City Clinical Data Research Network, is one of the Patient-Centered Outcomes Research Institute-funded clinical data research networks that aggregates clinical data from health systems and makes them available to support research. INSIGHT collects data from the following health systems, the Columbia University College of Physicians and Surgeons, Mount Sinai Health System, Montefiore Medical Center, NYU Langone Medical Center, New York-Presbyterian Hospital, and Weill Cornell Medicine.25
We first identified patients who were active Medicare beneficiaries when they visited the INSIGHT health systems in 2013. We restricted our sample to patients with 9-digit residential zip codes available in New York or New Jersey (where the majority of our patients resided). We required 9-digit zip codes as these patients could be mapped to a census block group for granular community social condition information. Patients were excluded if they: (1) had any Medicare Advantage enrollment as Medicare FFS claims data do not capture all their utilization, (2) died during the study period as their limited months of enrollment may result in artificially low costs, (3) did not have continuous 12-month enrollment in Medicare Part A or Part B programs in a calendar year as their utilization and other characteristics are incomplete in the Medicare claims data, or (4) had invalid zip codes that could not be used to identify patient neighborhood. The final study sample included 93,429 patients. Sample selection process is presented in the Appendix 1 (Supplemental Digital Content 1, http://links.lww.com/MLR/C18).
We used Medicare FFS claims data for the period of 2013 and 2014. Specifically, we used the following files: Carrier, Outpatient, MedPAR for inpatient care, Skilled Nursing Facility, Home Health Agency, Hospice, Durable Medical Equipment (DME), Part D Drug Event, and Master Beneficiary Summary. Community social conditions measured by the 2013 ADI were provided from the University of Wisconsin-Madison’s Neighborhood Atlas.26 Appendix 2 (Supplemental Digital Content 1, http://links.lww.com/MLR/C18) provided details of the data merge.
Measuring Disadvantaged Neighborhood Social Conditions
We used the ADI of 2013 to measure the characteristics of patient residential neighborhood at the level of the US census block group, a granular geographic unit typically containing between 600 and 3000 residents.3,4 The ADI is a multidimensional composite index measuring the neighborhood disadvantages that have been shown to influence health outcomes and utilization, such as hospital readmissions.3,4,21 It synthesizes 17 variables from American Community Survey that reflect important demographic and socioeconomic aspects of a neighborhood, including income, education, employment, poverty, and housing conditions. We rescaled the ADI state decile ranks for census block groups to quintiles. Communities in higher quintiles have more disadvantaged social conditions. We mapped each patient into a census block group based on the 9-digit residential zip code from INSIGHT using a commercial crosswalk. We applied the 2013 ADI to the 2013 and 2014 utilization data under the assumption that there is minimal change of neighborhood social conditions across 2 years.
Calculating Total and Potentially Preventable Utilization and Costs
We first calculated the standardized total annual Medicare costs across all service categories for 2013 and 2014, including physician services, outpatient services, inpatient services, durable medical equipment, Part D prescription, and postacute care (eg, skilled nursing facility and home health) and hospice services. We standardized the allowed payment amount at the claim level to account for differences in costs across patients due to price variations.
We focused on 3 types of utilization that were considered potentially preventable in the previous literature: preventable ED visits, preventable hospitalizations, and unplanned 30-day hospital readmissions. The utilization of these services could be prevented through improved care coordination, increased access to primary care, and high-quality outpatient care.23,27,28 In addition, health systems are targeting these services to meet the quality and cost benchmarks required by alternative payment and delivery models, such as accountable care organizations.
To identify potentially preventable ED visits, we applied an algorithm created by Billings and colleagues to ED visits that did not result in hospitalization.29 This algorithm classified each ED visit into 1 of 4 categories based on the discharge diagnosis code: nonemergent; emergent but primary care treatable; emergent, ED care needed but preventable; emergent, ED care needed, and not preventable. Given the potential uncertainty of the diagnosis and the variation across patients, an ED visit is assigned 4 probabilities, the likelihoods that the visit was in 1 of the 4 emergency status categories. We defined an ED visit as preventable if the combined probabilities of “non-emergent,” “emergent but primary care treatable,” and “emergent, ED care needed but preventable” was 75% or higher.30
To identify preventable hospitalizations, we used an algorithm from the Agency for Healthcare Research and Quality’s Prevention Quality Indicators.31 The Prevention Quality Indicators include a set of measures to identify hospitalizations for ambulatory care-sensitive conditions (ACSCs). Hospitalizations for these conditions, such as hypertension and uncontrolled diabetes could potentially be prevented with appropriate ambulatory care or other interventions. Previous studies found that hospitalizations for ACSCs are largely influenced by factors outside the hospital setting.32,33 For example, patients with appendicitis may develop perforated appendix and require hospitalization if they do not have access to surgical evaluation and experience delays in receiving needed care.33
To identify unplanned 30-day readmissions, we used the CMS algorithm for 30-day all-cause unplanned readmissions,34 which identifies readmissions that are unplanned and therefore could be considered as potentially preventable utilization.
Consistent with the previous literature, we calculated the standardized episodic costs of all potentially preventable utilization and all services delivered within 30 days after a preventable ED visit or hospitalization for each patient.35
We first compared differences in demographics, comorbidities, and other patient characteristics across patients in different ADI quintiles. Demographic characteristics included age, sex, and race. Comorbidities included the number of chronic conditions, frailty, serious mental illness (eg, depression), serious physical illness (eg, end-stage cancer), and end-stage renal disease.36,37 We used the Chronic Condition Warehouse-defined conditions to count the number of the chronic conditions and categorized the count into 5 categories (0, 1–2, 3–5, 6–8, and ≥9). We calculated the CMS Hierarchical Condition Categories score for each patient to measure the overall health status. Finally, we also examined dual-eligible status and Medicare Part D coverage across different ADI quintiles.
We used χ2 tests and t tests to test for differences in patient characteristics across ADI quintiles. We then used generalized linear models to examine the association of ADI quintiles with (1) total annual Medicare costs; (2) Medicare costs for by service categories, including physician services, outpatient services, inpatient services, durable medical equipment, Part D prescription, and postacute care (eg, skilled nursing facility and home health) and hospice services; (3) cost associated with preventable utilization. We also tested the association between ADI quintiles and the probability of having the 3 types of the preventable utilization by using logistic regressions. We focused on the neighborhoods with the least disadvantaged conditions (quintile 1) and the most disadvantaged conditions (quintile 5), as compared with the neighborhoods with the intermediate level conditions (quintile 3). All regressions were adjusted for all demographic, comorbidities, and other patient characteristics of each year. We additionally added year-fixed effects to control for secular changes in outcomes and indicators for hospital referral regions to account for supply side variations across regions (eg, physician practice pattern and availability of health resources). Additional details of model specification are available in Appendix 3 (Supplemental Digital Content 1, http://links.lww.com/MLR/C18). All health cost numbers were inflation-adjusted using the medical care Consumer Price Index from the Bureau of Labor Statistics. All numbers are in 2014 dollars.
We conducted several sensitivity analyses to examine the robustness of our results and to examine the potential pathway between SDoH and health care costs and utilization. We first use national ADI ranks rather than state ranks for analysis. We then excluded dual-eligible patients and patients under 65 as these patients have different characteristics and utilization patterns compared with Medicare patients 65 or older without Medicaid enrollment. Finally, we additionally controlled for resources and utilization pattern of primary care in the analysis to examine if the availability of primary care could explain the observed associated between SDoH and health care utilization.
Our study included 93,429 patients from New York and New Jersey (Fig. 1) who were continuously enrolled in Medicare Parts A and B for 12 months in both 2013 and 2014. Of these patients, 52.4% (N=48,974) resided in census block groups at the lowest quintile (the least disadvantaged) of the ADI score as these areas have a larger population size, 19.1% (N=17,838) to quintile 2, 14.4% (N=13,480) to quintile 3, 5.4% (N=5010) to quintile 4, and 8.7% (N=8127) to the top quintile (the most disadvantaged) (Appendices 4 and 5, Supplemental Digital Content 1, http://links.lww.com/MLR/C18). Compared with patients in other quintiles, those in the most disadvantaged were younger, more likely to be female, from racial/ethnicity minority groups, and dually eligible for Medicaid. Patients in the most disadvantaged quintile also were more likely to have end-stage renal disease, frailty, serious mental illness, and be seriously ill, and more chronic conditions. Finally, patients in more disadvantaged quintiles tend to have slightly higher Hierarchical Condition Categories scores, indicating their poorer health status relative to patients in other quintiles (Table 1).
Unadjusted Association of ADI Quintiles With Total and Potentially Preventable Utilization and Costs
Before adjusting for demographics and comorbidities, the average total and preventable costs were higher among patients with most disadvantaged social conditions, as compared with other patients (Appendix 6, Supplemental Digital Content 1, http://links.lww.com/MLR/C18). For example, among patients in the highest ADI quintile, average total annual Medicare costs were more than 50% higher than those in the lowest ADI quintile. Similarly, the average potentially preventable Medicare costs among patients in the highest ADI quintile were nearly 2.5 times higher than those in the lowest quintile (Appendix 7, Supplemental Digital Content 1, http://links.lww.com/MLR/C18).
Adjusted Association of ADI With Total and Potentially Preventable Utilization and Costs
Compared with the neighborhoods with intermediate level social conditions (quintile 3), the least disadvantaged neighborhoods (quintile 1) was associated with higher total Medicare costs ($427 or 3.2% higher, P<0.001) whereas the most disadvantaged neighborhoods (quintile 5) had similar total Medicare costs ($22 or 0.2% lower, P=0.89) after adjusting for patient demographics, comorbidities, and other characteristics (Fig. 2).
Across care settings, patients with the least disadvantaged neighborhood conditions had higher physician ($239 or 4.8% higher, P<0.001) and inpatient costs ($130 or 9.0% higher, P=0.004) but lower postacute care costs ($57 or 14.0% lower, P=0.003), as compared with those with the middle level neighborhood conditions. The differences in outpatient ($29 or 2.0% higher, P=0.47), DME ($7 or 4.0% lower, P=0.36), and prescription ($133 or 35.0% higher, P=0.25) costs were not statistically significant.
Compared with patients in the intermediate level, patients with the most disadvantaged neighborhood conditions had lower physician costs ($188 or 4.0% less, P=0.001). The differences in outpatient ($53 or 3.0% higher, P=0.37), inpatient ($46 or 3.0% lower, P=0.46), postacute care and hospice ($7 or 2.0% higher, P=0.81), DME ($7 or 4.0% higher, P=0.50), and prescription ($42 or 11% higher, P=0.77) costs were not statistically significant.
Disadvantaged neighborhood social conditions were associated with increased probability of having potentially preventable utilization and higher preventable costs after adjusting for patient characteristics. Compared with patients in the middle quintile, patients with the least disadvantaged neighborhood conditions had similar probability of 30-day readmission but significantly lower probability of having preventable hospitalizations (0.24-percentage-point decrease, P<0.001) and preventable ED visits (0.54-percentage-point decrease, P=0.006). Patients with the most disadvantaged neighborhood conditions had significantly higher probability of preventable hospitalizations (0.24-percentage-point increase, P=0.01) and preventable ED visits (1.28-percentage-point increase, P<0.001). Patients with the most disadvantaged neighborhood conditions had an annual average $53 or 12.0% (P=0.04) more potentially preventable costs than those with the middle level (Fig. 3).
Results were consistent when we use national ADI ranks and excluding dual-eligible patients or patients under 65. Results were attenuated after adding variables measuring resources and utilization pattern of primary care. Details are available in Appendix 8 (Supplemental Digital Content 1, http://links.lww.com/MLR/C18), and full regression results are available in Appendix 9 (Supplemental Digital Content 1, http://links.lww.com/MLR/C18).
In this study, we examined the association of neighborhood social conditions with total annual and preventable Medicare costs. Compared with the intermediate level neighborhood conditions as assessed by ADI at the census block group level, the least disadvantaged neighborhood conditions were associated with higher total Medicare costs but no different preventable costs. The most disadvantaged social conditions were associated with higher preventable Medicare costs but no different total Medicare costs.
Previous studies have found that disadvantaged social conditions are associated with increased rate of readmissions or other episodic utilization.3,21 Our study made a new contribution by examining additional preventable utilization measures and estimating the costs associated with preventable utilization. In addition, we controlled for a comprehensive set of medical, disability, and other patient characteristics that may confound the association between social conditions and health care utilization and costs. Prior studies found that patients living in disadvantaged neighborhoods are less likely to have a usual source of care and to obtain recommended preventive services.38,39 Therefore, these patients may instead use more acute care for ACSCs. For example, 1 study found that diabetic Medicare patients with disadvantaged socioeconomic status have a lapse in outpatient care for chronic condition management but have more potentially preventable hospitalizations.40 Patients from disadvantaged neighborhoods may also experience poorly coordinated care, which may lead to unnecessary health care utilization. For example, patients with acute myocardial infarction from disadvantaged neighborhoods have substantial delays in time from presentation to angiography and lower rates of referral to cardiac rehabilitation at discharge.41 Furthermore, these patients remain at higher risk of adverse in-hospital outcomes after acute myocardial infarction, such as major bleeding.41 These factors may lead to preventable hospital readmissions.
In contrast the finding that patients from the most disadvantaged neighborhoods had higher preventable costs, we also found that they had lower total Medicare costs as compared with those with the least disadvantaged conditions, after adjusting for patient demographics and comorbidities. To date, limited evidence is available regarding the relationship between SDoH and total health costs. Several potential reasons may explain this finding. The lower total health costs among these patients may indicate that they have significant barriers to access health care and are more likely to have unmet needs for necessary health care.38,42 Identifying specific social factors that lead to access barriers may be necessary to address the unmet health care needs among these patients. Studies have reported that public transit and supply of health resources are associated primary care access and utilization.40,43 Neighborhoods with disadvantaged social conditions may have poor public transportation and limited availability of health resources, which may lead to barriers of health care access. This is supported by our results from the sensitivity analysis that the differences in total and preventable costs are attenuated after controlling for availability and utilization patterns of primary care in the analysis. It is also possible that the barriers of access to health care are due to perceived discrimination or stigma as patients with disadvantaged social conditions are more likely to be from racial minority groups or have other behavioral conditions (eg, substance abuse).44,45 Besides health care access, patients from disadvantaged neighborhood may have unique sources of health care (eg, safety net) which may lead to different health outcomes.46 Policies addressing these issues are needed to improve the care for patients with disadvantaged social conditions.
One potential application of our findings would be to explore whether patient residential neighborhood characteristics have a causal relationship with unnecessary health care utilization and seek opportunities to intervene.47,48 Another application would be to leverage the associations, even if they are not causal, for predictive modeling that could identify high-risk patients.49,50 In either situation, further research is needed to assess community-level social conditions, integrate the information into care delivery, and identify community factors that may be amenable to interventions.
This study has several limitations. First, health care utilization and costs are associated with many patient characteristics that are unobservable but may be correlated with disadvantaged social conditions. Therefore, results of this study do not necessarily indicate causal relationships between social conditions and health care utilization and costs. Second, community social conditions could reflect real neighborhood characteristics that influence health care utilization or could be a proxy for individual social characteristics that are associated with health care utilization, such as median income. This study was not able to incorporate individual level SDoH in the analysis and therefore cannot examine if the associations of community SDoH are still significant after controlling for individual level SDoH. Third, although the definitions of the 3 types of preventable utilization have been widely used in the literature, they may not all be truly preventable; alternatively these services may not represent the totality of preventable utilization. Fourth, we were not able to incorporate Medicaid costs for dually eligible patients. Not incorporating Medicaid spending for the dual eligible may lead to the underestimate of total and preventable utilization for these patients which may have a greater presence in higher disadvantaged neighborhoods. Fifth, excluding patients who died within the study period was necessary to avoid erroneous cost underestimates due to incomplete follow-up, but also means that this analysis does not fully account for end-of-life costs and utilization. Finally, this study was based on a convenience sample of Medicare FFS and dual-eligible patients in New York and New Jersey areas. Findings from this study may not be generalizable to other patient groups (eg, Medicare Advantage) or other geographic areas.
Our findings suggest a nonlinear relationship between neighborhood conditions and health care costs, with different associations in affluent and disadvantaged neighborhoods. Patients living in neighborhoods with disadvantaged social conditions have higher potentially preventable health care costs after controlling for demographic, medical, and other patient characteristics. Barriers to access to primary care may prompt these patients to limit their use of preventive care and increase their use of acute care for routine problems. Health systems may leverage publicly available social conditions data to target patients at higher risk of having preventable utilization and deliver effective interventions.
1. Marmot M. Closing the health gap in a generation: the work of the Commission on Social Determinants of Health
and its recommendations. Glob Health Promot. 2009;16(suppl 1):23–27.
2. Lichtenberg FR, Sun SX. The impact of Medicare
Part D on prescription drug use by the elderly. Health Aff (Millwood). 2007;26:1735–1744.
3. Kind AJ, Jencks S, Brock J, et al. Neighborhood socioeconomic disadvantage and 30-day rehospitalization: a retrospective cohort study. Ann Intern Med. 2014;161:765–774.
4. Singh GK. Area deprivation and widening inequalities in US mortality, 1969-1998. Am J Public Health. 2003;93:1137–1143.
5. Tarasuk V, Cheng J, de Oliveira C, et al. Association between household food insecurity and annual health care costs
. Can Med Assoc J. 2015;187:E429–E436.
6. Artiga S, Hinton E. Beyond health care: the role of social determinants in promoting health and health equity. 2018. Available at: https://kff.org/disparities-policy/issue-brief/beyond-health-care-the-role-of-social-determinants-in-promoting-health-and-health-equity/
. Accessed April 16, 2019.
7. McGovern L. The relative contribution of multiple determinants to health. 2014. Available at: https://healthaffairs.org/do/10.1377/hpb20140821.404487/full/
. Accessed April 16, 2019.
8. Cantor MN, Chandras R, Pulgarin C. FACETS: using open data to measure community social determinants of health
. J Am Med Inform Assoc. 2018;25:419–422.
9. Cantor MN, Thorpe L. Integrating data on social determinants of health
into electronic health records. Health Aff (Millwood). 2018;37:585–590.
10. Phillips RL, Liaw W, Crampton P, et al. How other countries use deprivation indices-and why the United States desperately needs one. Health Aff (Millwood). 2016;35:1991–1998.
11. Knighton AJ, Savitz L, Belnap T, et al. Introduction of an area deprivation index measuring patient socioeconomic status in an integrated health system: implications for population health. EGEMS (Wash DC). 2016;4:1238.
12. Roberts ET, Zaslavsky AM, Barnett ML, et al. Assessment of the effect of adjustment for patient characteristics on hospital readmission rates: implications for pay for performance. JAMA Intern Med. 2018;178:1498–1507.
13. Joynt Maddox KE, Reidhead M, Hu J, et al. Adjusting for social risk factors impacts performance and penalties in the hospital readmissions reduction program. Health Serv Res. 2019;54:327–336.
14. Morgan DJ, Bame B, Zimand P, et al. Assessment of machine learning vs standard prediction rules for predicting hospital readmissions. JAMA Netw Open. 2019;2:e190348.
15. Low LL, Liu N, Wang S, et al. Predicting frequent hospital admission risk in Singapore: a retrospective cohort study to investigate the impact of comorbidities, acute illness burden and social determinants of health
. BMJ Open. 2016;6:e012705.
16. Daniel H, Bornstein SS, Kane GC. Health, Public Policy Committee of the American College of Physicians. Addressing social determinants to improve patient care and promote health equity: an American College of Physicians Position Paper. Ann Intern Med. 2018;168:577–578.
17. National Academies of Sciences Engineering Medicine. Accounting for Social Risk Factors in Medicare
Payment. Washington, DC: National Academies Press; 2017.
18. Centers for Medicare
and Medicaid Services. Accountable Health Communities Model. 2019. Available at: https://innovation.cms.gov/initiatives/ahcm
. Accessed November 23, 2019.
19. Khullar D, Kaushal R. “Precision health” for high-need, high-cost patients. Am J Manag Care. 2018;24:396–398.
20. Krumholz HM, Bernheim SM. Considering the role of socioeconomic status in hospital outcomes measures. Ann Intern Med. 2014;161:833–834.
21. Hu J, Kind AJH, Nerenz D. Area deprivation index predicts readmission risk at an urban teaching hospital. Am J Med Qual. 2018;33:493–501.
22. Millwee B, Goldfield N, Turnipseed J. Achieving improved outcomes through value-based purchasing in one state. Am J Med Qual. 2018;33:162–171.
23. Maeng DD, Hao J, Bulger JB. Patterns of multiple emergency department visits: do primary care physicians matter? Perm J. 2017;21:16-063.
24. McWilliams JM, Hatfield LA, Chernew ME, et al. Early performance of accountable care organizations in Medicare
. N Engl J Med. 2016;374:2357–2366.
25. Kaushal R, Hripcsak G, Ascheim DD, et al. Changing the research landscape: the New York City Clinical Data Research Network. J Am Med Inform Assoc. 2014;21:587–590.
26. Neighborhood Atlas. Area Deprivation Index. 2018. Available at: https://www.neighborhoodatlas.medicine.wisc.edu/
. Accessed December 1, 2018.
27. Joynt KE, Gawande AA, Orav EJ, et al. Contribution of preventable acute care spending to total spending for high-cost Medicare
patients. JAMA. 2013;309:2572–2578.
28. Zuckerman RB, Sheingold SH, Epstein AM. The hospital readmissions reduction program. N Engl J Med. 2016;375:494.
29. New York University. Faculty & Research. 2017. Available at: https://wagner.nyu.edu/faculty/billings/nyued-background
. Accessed January 31, 2018.
30. Hwang W, Liao K, Griffin L, et al. Do free clinics reduce unnecessary emergency department visits? The Virginian experience. J Health Care Poor Underserved. 2012;23:1189–1204.
31. Agency for Healthcare Research and Quality. Prevention quality indicators overview. 2017. Available at: https://qualityindicators.ahrq.gov/modules/pqi_resources.aspx
. Accessed January 31, 2018.
32. Freund T, Campbell SM, Geissler S, et al. Strategies for reducing potentially avoidable hospitalizations for ambulatory care-sensitive conditions. Ann Fam Med. 2013;11:363–370.
33. Agency for Healthcare Research and Quality. Guide to prevention quality indicators: hospital admission for ambulatory care sensitive conditions. 2001. Available at: https://ahrq.gov/downloads/pub/ahrqqi/pqiguide.pdf
. Accessed November 23, 2019.
34. Centers for Medicare
and Medicaid Services. 30-day all-cause hospital readmission measure. 2017. Available at: https://cms.gov/Medicare/Medicare-Fee-for-Service-Payment/PhysicianFeedbackProgram/Downloads/2015-ACR-MIF.pdf
. Accessed March 2, 2018.
35. Figueroa JF, Joynt Maddox KE, Beaulieu N, et al. Concentration of potentially preventable spending among high-cost Medicare
subpopulations: an observational study. Ann Intern Med. 2017;167:706–713.
36. Kim DH, Schneeweiss S. Measuring frailty using claims data for pharmacoepidemiologic studies of mortality in older adults: evidence and recommendations. Pharmacoepidemiol Drug Saf. 2014;23:891–901.
37. Weissman DE, Meier DE. Identifying patients in need of a palliative care assessment in the hospital setting: a consensus report from the Center to Advance Palliative Care. J Palliat Med. 2011;14:17–23.
38. Kirby JB, Kaneda T. Neighborhood socioeconomic disadvantage and access to health care. J Health Soc Behav. 2005;46:15–31.
39. Prentice JC. Neighborhood effects on primary care access in Los Angeles. Soc Sci Med. 2006;62:1291–1303.
40. Ryvicker M, Sridharan S. Neighborhood environment and disparities in health care access among urban Medicare
beneficiaries with diabetes: a retrospective cohort study. Inquiry. 2018;55:1–12.
41. Udell JA, Desai NR, Li S, et al. Neighborhood socioeconomic disadvantage and care after myocardial infarction in the National Cardiovascular Data Registry. Circ Cardiovasc Qual Outcomes. 2018;11:e004054.
42. National Quality Forum. Risk adjustment for socioeconomic status or other sociodemographic factors. National Quality Forum, Washington, DC; 2014.
43. Ryvicker M, Gallo WT, Fahs MC. Environmental factors associated with primary care access among urban older adults. Soc Sci Med. 2012;75:914–921.
44. De P, Pozen A, Budhwani H. Is perceived stigma in clinical settings associated with poor health status among New York City’s residents of color? Med Care. 2019;57:960–967.
45. Hausmann LR, Jeong K, Bost JE, et al. Perceived discrimination in health care and health status in a racially diverse sample. Med Care. 2008;46:905–914.
46. Hussein M, Diez Roux AV, Field RI. Neighborhood socioeconomic status and primary health care: usual points of access and temporal trends in a major US urban area. J Urban Health. 2016;93:1027–1045.
47. Barber RD, Coulourides Kogan A, Riffenburgh A, et al. A role for social workers in improving care setting transitions: a case study. Soc Work Health Care. 2015;54:177–192.
48. Hesselink G, Zegers M, Vernooij-Dassen M, et al. Improving patient discharge and reducing hospital readmissions by using Intervention Mapping. BMC Health Serv Res. 2014;14:389.
49. Kasthurirathne SN, Vest JR, Menachemi N, et al. Assessing the capacity of social determinants of health
data to augment predictive models identifying patients in need of wraparound social services. J Am Med Inform Assoc. 2018;25:47–53.
50. Lu MLR, Davila CD, Shah M, et al. Marital status and living condition as predictors of mortality and readmissions among African Americans with heart failure. Int J Cardiol. 2016;222:313–318.