Abstract
Background:
Costs incurred by health systems when caring for populations with social or behavioral complexity are poorly understood. We compared the frequency and costs of unreimbursed care among individuals with complexity factors (homelessness, a history of county jail incarceration, and/or substance use disorder or mental illness [SUD/MI]).
Methods:
We conducted a cross-sectional analysis using electronic health record data for adults aged 18 and older between 1/1/2016 and 12/31/2017 from a Midwestern safety-net health system. Zero-inflated negative binomial regression models were used to assess risk-adjusted associations between complexity factors and care coordination encounters, missed appointments, and excess inpatient days.
Results:
Our sample included 154,719 unique patients; 6.8% were identified as homeless, 7.8% had a history of county jail incarceration, and 20.6% had SUD/MI. Individuals with complexity factors were more likely to be African-American, Native American, or covered by Medicaid. In adjusted models, homelessness and SUD/MI were significantly associated with care coordination encounters (RR 1.8 [95% CI,1.7–2.0]; RR 1.9 [95% CI,1.8–2.0]), missed appointments (RR 1.5 [95% CI,1.4–1.6]; RR 1.7 [95% CI,1.7–1.8]), and excess inpatient days (RR 1.5 [95% CI,1.3–1.8]; RR 2.8 [95% CI,2.5–3.1]). County jail incarceration was associated with a significant increase in missed appointments. In 2017, SUD/MI accounted for 81.8% ($7,773,000/$9,502,000) of excess costs among those with social or behavioral complexity.
Conclusions:
Social and behavioral complexity are independently associated with high levels of unreimbursed health system resource use.
Implications:
Future payment models should account for the health system resources required to care for populations with complex social and behavioral needs.
Introduction
Improving the health of populations with complex social and behavioral needs is of critical importance to policymakers and health systems.1,2 Social and behavioral factors are key drivers of health inequities in the United States, and it is increasingly recognized that optimal health outcomes for vulnerable populations will not be achieved if these factors are not addressed.3–5 Some payers have incorporated social and behavioral factors into their payment models to provide resources to help health systems address these issues.6 For example, Minnesota’s Integrated Health Partnerships model provides a payment enhancement to health systems based on the proportion of the population that has social and behavioral risk factors.7 Massachusetts’ Medicaid payment model for managed care plans includes risk adjustment for unstable housing, mental illness, substance abuse, and a measure of neighborhood stress.6
To date, algorithms to estimate the additional resources needed to care for populations with high levels of social risk are limited because they only include resource use that has been traditionally reimbursed by payers (e.g., office visits, emergency department visits, capitated hospital stays, and quality metrics).8,9 They do not account for additional costs incurred by health systems when caring for individuals with social or behavioral complexity. For example, a hospital may keep an individual who is homeless and medically stable because there is not a safe place for the individual to transition into. The resulting excess length of stay is often a cost borne by the hospital, since reimbursement does not generally vary by length of stay under the predominant payment methodologies used by public payers.10 Current modeling strategies generally consider the hospitalization itself, not the extended length of stay required to provide safe care.6
In this analysis, we examined three social and behavioral complexity factors using data from an integrated electronic medical record within a county-affiliated safety-net health system: homelessness, a history of county jail incarceration, and substance use disorder or mental illness (SUD/MI). We quantified the relationship between these factors and three types of health system resources that are generally unreimbursed and used at greater levels by complex populations: encounters with care coordination professionals, appointments that are booked but not attended by the patient (i.e., missed appointments), and excess hospital days.11,12 Finally, as a secondary objective, we estimated excess costs associated with these types of health system resource use among patients with social and behavioral complexity. This approach can inform payment strategies that more accurately account for the costs incurred by health systems when caring for complex populations.
METHODS
Study population and data
We used data from a county-affiliated electronic health record (EHR) system that spans health care, criminal justice, and social service agencies. Our study population included individuals aged 18 and older with any clinic or hospital encounter between 1/1/2016 and 12/31/2017. We excluded individuals whose only encounter(s) in the EHR were for stand-alone services such as pharmacy or ambulance or those who were incarcerated at the end of our study period (12/31/2017) or at the time of data extraction (3/25/2020). Limits on incarceration data were due to institutional review board policies related to jail EHR data.
Dependent variables/outcome measures
We considered three types of health care system interactions (herein referred to as health care system resource use) that represent costs borne by the health system and are generally unreimbursed by payers, though are occasionally underreimbursed: care coordination encounters13 (typically performed by a registered nurse or clinical coordinator), excess inpatient days (inpatient hospital days surpassing the geometric mean length of stay for the same condition), and missed appointments (encounters with a visit status of “No Show”). We chose these particular event types because they occur disproportionately among individuals with social or behavioral complexity and result in hidden operating costs to the health system.14,15 Totals for each event type were calculated for one calendar year (2017). However, participants were included in the study if they received health care services at any point during a two-year period from 2016 to 2017. We did this to create a cohort of users who could potentially use health care services during the outcome period. Therefore, our study may capture some patients who utilized services in 2016, but not in 2017.
Independent variables
We examined three markers of social and behavioral complexity: incarceration in the Hennepin County jail, homelessness, and any behavioral health diagnosis (i.e., a mental health and/or substance use disorder diagnosis). We identified a history of incarceration from data within our integrated EHR system, which includes documentation of all admissions and discharges and some health care services at the Hennepin County jail. An individual was determined to have a history of incarceration if they were incarcerated between 1/1/2016 and 12/31/2017. Further, we used a homelessness indicator based on patients’ addresses to identify those likely to be homeless.16 The homelessness indicator was generated by running patient addresses against a database of local shelters and general delivery addresses. The indicator has a sensitivity of 76% compared to self-report. We determined a history of homelessness based on any address flagged for homelessness between 1/1/2016 and 12/31/2017. We determined history of behavioral health disorder if an individaul had an active diagnosis of substance use or mental illness between 1/1/2016 and 12/31/2017. Substance use or mental illness diagnoses were defined using the Centers for Medicare and Medicaid Services’ Hierarchical Condition Category groupers (CMS-HCC);17 details can be found in the online Appendix (eTable 1).
We controlled for individual HCC scores in adjusted models to identify the additional contribution of complexity factors to uncompensated costs beyond their standard medical risk based on age, gender, and medical diagnoses.17 We used a risk score based on the CMS-HCC model and included an expanded set of categories to account for additional conditions not present in the CMS model, such as pediatric-specific diagnoses. We did this to account for the mismatch between the CMS model and the younger population served in our health system. A list of the additional diagnoses and associated coefficients included in our risk adjustment can be found in the online Appendix (eTable 2).
We used the highest calculated HCC risk score, generated on a monthly rolling basis, for each patient in the two-year study period. The resulting HCC score is an approximation of clinical complexity and removes some of the temporal limitations of risk calculations. It is important to note that the HCC risk score accounts for SUD/MI diagnoses; therefore any additional impact of these conditions on excess costs indicates the extent to which the model, derived from claims-based data, fails to adequately account for health system costs, but does not indicate failure of the model to recognize SUD/MI itself.
Uncompensated costs
For each type of uncompensated cost, we estimated the excess cost to the health system.
Care coordination:
We estimated the average cost of a care coordination encounter to be $50.00 by dividing the health system’s care coordination budget by the number of encounters in 2017.18 Some health systems in Minnesota, including ours, receive limited reimbursement for some of these encounters through the state Health Care Home initiative.19 This reimbursement covers approximately 20% of program costs based on the average reimbursed amount in a calendar year through the state’s Health Care Home program divided by the total cost of the care coordination department, leaving 80% of the costs as unreimbursed. We therefore used an estimated excess cost per care coordination encounter of $40.
Missed appointments:
We estimated the average excess cost of a missed appointment to be $115, which is the average payment received for an outpatient visit and the value used for financial forecasting in our health system.20,21 If the appointment is missed, the system forgoes the revenue for that visit, but the majority of the costs incurred (e.g., provider time, other staff time) are not reduced.
Excess inpatient days:
We used $600 per excess inpatient day to estimate the direct cost of days beyond the mean length of stay per DRG category.22,23 Direct costs were provided by the health system finance department and represent the average daily cost of medical supplies and labor; drug costs are not included in this calculation.
Statistical analysis
We first examined the demographic characteristics of our entire study population, as well as populations with a given social or behavioral complexity factor. Next, we examined the overlap of social and behavioral complexity factors and described the rates of inpatient, emergency department, outpatient, and jail use across complexity factors to determine the reimbursable health care use patterns of each population.
We estimated means for each type of health system resource use for the entire sample and by several combinations of social and behavioral complexity factors. For each complexity factor in our primary analysis, we also estimated the risk of health system resource use outcomes relative to individuals without that factor. To assess whether relative risk ratios were statistically significantly different than one, we used Stata’s test function, which employs a Wald test.
To isolate the independent association of each social and behavioral complexity factor with health system resource use, we estimated zero-inflated negative binomial regression models, adjusted for HCC scores and social and behavioral complexity factors. We compared the fit of our model to zero-inflated Poisson and two-part models using Akaike and Bayesian information criteria.24
Finally, we estimated excess costs incurred by the health system that were attributed to social and behavioral complexity factors. For each type of resource use, we multiplied the difference in the adjusted frequency of resource use between those with a given social or behavioral complexity factor to those without that factor by the number of individuals with a given social and behavioral complexity factor. We then multiplied the unit cost per event by the estimated event count per factor.
RESULTS
Sample characteristics
Our sample included 154,719 unique patients with health care service use between 1/1/2016 and 12/31/2017. Of those, 6.8% were identified as homeless, 7.8% had at least one county jail admission, and 20.6% had a SUD or MI diagnosis (Table 1). Individuals with social or behavioral complexity factors were disproportionately African-American or Native American and covered by Medicaid compared with the overall study population.
Table 1.
Characteristics of study population
| Characteristic | All | Any homelessness | Any county jail incarceration | Any SUD/MI |
|---|---|---|---|---|
| Sample size | 154,719 | 10,468 (6.8%) | 12,145 (7.8%) | 31,887 (20.6%) |
| Mean Age | 42.5 | 40.4 | 35.2 | 43.3 |
| Race/Ethnicity | ||||
| Native American | 4,248 (2.7%) | 872 (8.3%) | 1,033 (8.5%) | 1,672 (5.2%) |
| Asian | 6,394 (4.1%) | 116 (1.1%) | 119 (1.0%) | 828 (2.6%) |
| Black (African American or African), non-Hispanic | 47,271 (30.6%) | 5,318 (50.8%) | 6,472 (53.3%) | 10,883 (34.1%) |
| Hispanic (Latino) | 23,785 (15.4%) | 444 (4.2%) | 686 (5.6%) | 3,022 (9.5%) |
| White, non-Hispanic | 65,609 (42.4%) | 3,453 (33.0%) | 3,502 (28.8%) | 14,671 (46.0%) |
| Other | 2,245 (1.5%) | 84 (0.8%) | 134 (1.1%) | 374 (1.2%) |
| Unknown | 5,167 (3.3%) | 181 (1.7%) | 199 (1.6%) | 437 (1.4%) |
| Female | 78,256 (50.6%) | 3,956 (37.8%) | 3,358 (27.6%) | 15,038 (47.2%) |
| Insurance Product | ||||
| Commercial | 39,037 (25.2%) | 410 (3.9%) | 685 (5.6%) | 4,718 (14.8%) |
| Medicaid | 50,364 (32.6%) | 6,591 (63.0%) | 6,094 (50.2%) | 15,003 (47.1%) |
| Medicare | 22,321 (14.4%) | 1,214 (11.6%) | 635 (5.2%) | 6,531 (20.5%) |
| MNCare | 5,389 (3.5%) | 121 (1.2%) | 134 (1.1%) | 731 (2.3%) |
| Other | 9,076 (5.9%) | 255 (2.4%) | 349 (2.9%) | 766 (2.4%) |
| Self/Unknown | 28,532 (18.4%) | 1,877 (17.9%) | 4,248 (35.0%) | 4,138 (13.0%) |
| HCC Risk Categoriesa,b | ||||
| Highest Risk (>= 2.50) | 6,711 (4.3%) | 954 (9.1%) | 622 (5.1%) | 3,894 (12.2%) |
| High Risk (1.00 – 2.49) | 19,416 (12.5%) | 2,984 (28.5%) | 2,570 (21.2%) | 11,863 (37.2%) |
| Rising Risk (0.42 – 0.99) | 36,235 (23.4%) | 2,352 (22.5%) | 2,676 (22.0%) | 11,823 (37.1%) |
| General (0 – 0.41) | 92,357 (59.7%) | 4,178 (39.9%) | 6,277 (51.7%) | 4,307 (13.5%) |
| HCC Mean (SD)a,c | 0.67 (0.97) | 1.07 (1.27) | 0.78 (0.93) | 1.37 (1.40) |
HCC: hierarchical condition category
Hennepin Healthcare derived categories used to stratify care management workflows
SD: standard deviation
There was considerable overlap among complexity factors (Figure 1). For example, among individuals who were homeless, 54% reported SUD or MI and 30% reported at least one admission to the county jail. Similarly, those with an incarceration history reported relatively high levels of homelessness (26% vs. 7%) and SUD/MI (45% vs. 21%) compared with the overall study population.
Figure 1.
Overlap of homelessness, county jail involvement, and substance use disorder/mental illness in the study population, 2016–2017
Reimbursable health care use and jail use among study population
Hospitalization and emergency department use were substantially higher among individuals with any social or behavioral complexity factor (Table 2) compared with the overall study population. ED visits were approximately 2.5 to 4.1 times more common and hospital days were 1.5 to 3.0 times more prevalent among individuals with any social or behavioral complexity factor. Compared with the overall study population, those with any homelessness or any SUD/MI had 1.2 and 2.0 times more outpatient visits, respectively, whereas those with any county jail incarceration had 33% fewer outpatient visits. Jail use was substantially higher across each complexity factor, relative to the overall population.
Table 2.
Hospital, emergency department, outpatient, and jail utilization by complexity factor, 2017
| Mean events per 1000 patients (Total events) | |||||
|---|---|---|---|---|---|
| All patients N=154,719 |
No complexity factor N = 112,240 |
Any homelessness N=10,468 |
Any county jail incarceration N=12,145 |
Any SUD/MI N=31,887 |
|
| Inpatient stays | 111 (17,192) | 66 (7,402) | 266 (2,786) | 173 (2,106) | 287 (9,160) |
| Inpatient days | 642 (99,264) | 321 (36,011) | 1,739 (18,199) | 984 (11,956) | 1,897 (60,482) |
| ED visitsa | 400 (61,885) | 218 (24,444) | 1,629 (17,049) | 1,288 (15,637) | 994 (31,704) |
| Outpatient visitsb | 2,145 (331,919) | 1,668 (187,215) | 2,656 (27,806) | 1,444 (17,534) | 4,244 (135,333) |
| Jail Days | 1,603 (247,946) | 0 | 7,609 (79,648) | 20,415 (247,946) | 4,152 (132,400) |
Includes both emergency department visits and emergency department observation visits
Office visit completed
Health system resource use among individuals with social or behavioral complexity factors
Health system resource use for the three target categories was consistently higher across populations with any social or behavioral complexity factor compared to populations without a respective complexity factor (Table 3, eTable 3, and eTable 4). Each category of resource use was higher across the three complexity factors compared to indivdiauls without a corresponding complexity factor. SUD/MI was associated with the largest relative increase in resource use, driven by the low levels of health system resource use among individuals without SUD/MI.
Table 3.
Unadjusted and adjusted health system resource use among adult patients by complexity factor, 2017
| Patient complexity factor | Care coordination encounters | Missed appointments | Excess inpatient days | ||||||
|---|---|---|---|---|---|---|---|---|---|
| Mean per 1000 patients with factor (95% CI) |
Mean per 1000 patients without factor (95% CI) |
Relative Risk (95% CI) |
Mean per 1000 patients with factor (95% CI) |
Mean per 1000 patients without factor (95% CI) |
Relative Risk (95% CI) |
Mean per 1000 patients with factor (95% CI) |
Mean per 1000 patients without factor (95% CI) |
Relative Risk (95% CI) |
|
| Unadjusted | |||||||||
| Any homelessness | 1,342 (1,233–1,450) | 316 (303–328) | 4.3(4.0–4.5)* | 2,331 (2,218–2,444) | 807 (807–822) | 2.9(2.8–3.0)* | 860 (741–979) | 234 (220–249) | 3.7(3.3–4.0)* |
| Any county jail incarceration | 698 (632–764) | 358 (345–372) | 1.9(1.8–2.1)* | 1,555 (1,471–1,639) | 855 (839–871) | 1.8(1.7–1.9)* | 454 (381–527) | 262 (246–277) | 1.7(1.5–2.0)* |
| Any diagnosis of SUD and/or MI | 1,238 (1,181–1,294) | 164 (155–172) | 7.6(6.8–8.3)* | 2,353 (2,287–2,418) | 535 (525–546) | 4.4(4.2–4.6)* | 917 (848–985) | 111 (102–119) | 8.3(6.9–9.6)* |
| Adjusted | |||||||||
| Any homelessness | 652 (602–702) | 352 (340–365) | 1.8(1.7–2.0)* | 1,480 (1,411–1,549) | 983 (955–1,011) | 1.5(1.4–1.6)* | 421 (364–479) | 273 (256–291) | 1.5(1.3–1.8)* |
| Any county jail incarceration | 383 (351–414) | 397 (383–411) | 1.0(0.9–1.0) | 1,212 (1,154–1,270) | 1,028 (999–1,057) | 1.2(1.1–1.2)* | 276 (236–316) | 300 (280–320) | 0.9(0.8–1.1) |
| Any diagnosis of SUD and/or MI | 515 (494–535) | 272 (260–284) | 1.9(1.8–2.0)* | 1,363 (1,325–1,401) | 799 (775–823) | 1.7(1.7–1.8)* | 438 (406–471) | 156 (146–167) | 2.8(2.5–3.1)* |
P<.001
All models adjusted for HCC score and social/behavioral complexity factors
In adjusted models (Table 3 and eTable 5), homelessness and SUD/MI were consistently and positively associated with health system resource use. Both complexity factors were associated with a significantly elevated adjusted relative risk, ranging from 1.5 to 2.8, for each type of health system resource use. County jail incarceration was associated with a significant increase in missed appointments, but not care coordination or excess inpatient days. SUD/MI was associated with a larger adjusted relative risk for missed appointments and excess inpatient days compared with homelessness and county jail incarceration, whereas it had a risk level similar to homelessness for care coordination encounters.
Sensitivity models
We estimated adjusted models using zero-inflated Poisson and two-part models and compared results to our primary zero-inflated negative binomial model. A likelihood-ratio test comparing our primary model to a zero-inflated Poisson model was statistically significant (P<.001), which indicated that our zero-inflated negative binomial model was preferred. The Akaike and Bayesian information criteria were substantially smaller for the zero-inflated negative binominal model, relative to the zero-inflated Poisson or two-part model, again indicating that our primary model was a better fit.
Adjsuted excess costs
In 2017, social and behavioral complexity factors were associated with $429,000 in excess care coordination costs, $2,923,000 in excess costs attributed to missed appointments, and $6,150,000 in excess costs related to excess inpatient days (Table 4). SUD/MI accounted for 81.8% of excess costs among those with social or behavioral complexity and was the largest driver, associated with $7,773,000 of the $9,502,000 total excess costs (Table 4).
Table 4.
Adjusted excess costs resulting from utilization of selected resources among Hennepin Healthcare adult patients by complexity factor, 2017
| Excess costs ($)a | ||||
|---|---|---|---|---|
| Care coordination | Missed appointments | Inpatient excess daysb | Total excess costs | |
| Any homelessness (n=10,468) | $126,000 | $598,000 | $930,000 | $1,654,000 |
| Any county jail incarceration (n=12,145) | −$7,000 | $257,000 | −$175,000 | $75,000 |
| Any SUD/MI (n=31,887) | $310,000 | $2,068,000 | $5,395,000 | $7,773,000 |
| Total excess costs | $429,000 | $2,923,000 | $6,150,000 | $9,502,000 |
All costs adjusted for HCC score and social/behavioral complexity factors
Days beyond the mean length of stay per DRG category
DISCUSSION
In this analysis of data from an urban safety-net health system, we found that health system resource use was significantly higher among populations with social or behavioral complexity compared to populations without these types of complexity factors. Differences in health system resource use between individuals with and without each complexity factor largely persisted, although were reduced, after application of a standard clinical risk adjustment tool (HCC) and adjusting for co-occurring complexity. Excess costs attributed to social and behavioral complexity, independent of HCC risk, totaled approximately $9.5 million in 2017 for the three outcomes measured in this study (approximately $61 per individual with a social or behavioral complexity factor per year). Taken together, our results quantify some of the unreimbursed costs incurred by health systems caring for complex populations. For safety-net health systems that disproportionately care for complex populations, this uncompensated resource use could have a substantial negative impact on financial viability under current payment models, as well as on the ability of safety-net systems to address social and behavioral complexity in order to help patients achieve better health.
Consistent with prior literature, we found that inpatient and emergency department utilization were considerably higher among individuals with social or behavioral complexity.14,21,25–27 However, these data are among the first to report variation in utilization across a variety of social and behavioral complexity factors in a safety-net cohort. Thus, our data highlight the relative associations of each complexity factor with utilization and the substantial influence of both homelessness and SUD/MI on inpatient hospital use. Our finding that outpatient visits were lower among adults with jail incarceration is a new contribution to the literature and is consistent with work among adolescents with juvenile justice incarceration.28 These utilization differences suggest that intervention targets may differ by type of complexity factor.
The health system costs associated with social and behavioral complexity represented one-third of the health system’s negative operating margin in 2017, and in some years, that would determine the difference between a negative and positive operating margin. Of the factors we examined, excess inpatient days accounted for the majority of the unreimbursed costs ($6.15M), mostly related to SUD/MI. Excess costs related to SUD/MI are attributable to the large number of individuals with SUD/MI, as well as the higher levels of per capita resource use. This is consistent with a growing body of evidence that SUD, in particular, is associated with a longer length of stay.29,30 It is notable however that excess costs from missed appointments in the outpatient setting also have a substantial financial impact. Care coordination encounters are an intervention that may help address the needs of socially and behaviorally complex populations;31–33 the relatively small magnitude of excess costs due to care coordination suggests that the cost of providing such services is minor compared to other types of costs (i.e., inpatient days). Negative excess costs associated with county jail incarceration suggest that housing and behavioral health issues in this population are the primary drivers of cost and not incarceration itself.
This study builds on previous work on health care utilization among socially and behaviorally complex populations and costs to payers.1,6,34 The complexity factors in this study are known to influence health care utilization as measured by payers, but the additional resource use required by health systems to care for complex populations has not been well delineated. By quantifying the excess costs incurred by a safety-net health system, we provide additional insight into health system costs that have traditionally been difficult to measure. These excess costs are often hidden to payers and policymakers because they are visible only to the health system (i.e., missed appointments). Therefore, payers are unable to address them in their payment methodology. Our results indicate that payment methodologies that account for these costs could substantially improve the ability of safety-net systems, tasked with caring for socially and behaviorally complex populations, to financially sustain their work.
Our study has several limitations that should be considered when interpreting our results. First, we assessed associations between social and behavioral complexity and resource use within one urban, Midwest safety-net health system. Results may not be generalizable to other institutions, particularly those who are not part of the safety-net. Second, we measured only three types of health system resource use that are impacted by social and behavioral complexity. Therefore, we likely underestimate the excess hidden costs incurred by health systems caring for these populations. Even with these conservative estimates, however, the excess costs of the three categories of resource use we measured represent a substantial fraction of the health system’s operating margin. Third, the care coordination encounters for patients in this study represent the modest investment the health system could afford and not the investment necessary to meet the full demand for social and behavioral services. Fourth, our homeless indicator is based on patients’ self-reported registration addresses in their electronic health record and may undercount homelessness by 25% or more. However, we previously compared the indicator to self-reported data and found it to be accurate on a population level. Finally, we only captured criminal justice incarceration if individuals accessed jails within Hennepin County, but were not able to capture visits to jails in other counties or in state or federal prisons.
Conclusions
In an urban safety-net population, social and behavioral complexity were significantly associated with three key health system resource use measures: excess hospital days, missed appointments, and care coordination. The excess costs associated with this resource use are not fully accounted for in current payment models and contribute to the financial vulnerability of safety-net health systems.
Supplementary Material
References
- 1.Alley DE, Asomugha CN, Conway PH, Sanghavi DM. Accountable health communities — addressing social needs through Medicare and Medicaid. New England Journal of Medicine. 2016;374(1):8–11. doi: 10.1056/NEJMp1512532 [DOI] [PubMed] [Google Scholar]
- 2.Blewett LA, Owen RA. Accountable care for the poor and underserved: Minnesota’s Hennepin Health model. Am J Public Health. 2015;105(4):622–624. doi: 10.2105/AJPH.2014.302432 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Daniel H, Bornstein SS, Kane GC, for the Health and 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(8):577. doi: 10.7326/M17-2441 [DOI] [PubMed] [Google Scholar]
- 4.McGovern L, Miller G, Hughes-Cromwick P. Health Policy Brief: The Relative Contribution of Multiple Determinants to Health Outcomes. Health Affairs; 2014. [Google Scholar]
- 5.Galea S, Nandi A, Vlahov D. The social epidemiology of substance use. Epidemiologic Reviews. 2004;26(1):36–52. doi: 10.1093/epirev/mxh007 [DOI] [PubMed] [Google Scholar]
- 6.Ash AS, Mick EO, Ellis RP, Kiefe CI, Allison JJ, Clark MA. Social determinants of health in managed care payment formulas. JAMA Internal Medicine. 2017;177(10):1424. doi: 10.1001/jamainternmed.2017.3317 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Integrated Health Partnerships (IHP). Minnesota Department of Human Services. Published June 7, 2019. Accessed June 26, 2019 https://mn.gov/dhs/partners-and-providers/news-initiatives-reports-workgroups/minnesota-health-care-programs/integrated-health-partnerships/
- 8.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(2):327–336. doi: 10.1111/1475-6773.13133 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Joynt KE, De Lew N, Sheingold SH, Conway PH, Goodrich K, Epstein AM. Should medicare value-based purchasing take social risk into account? N Engl J Med. 2017;376(6):510–513. doi: 10.1056/NEJMp1616278 [DOI] [PubMed] [Google Scholar]
- 10.Quinn K, Courts C. Sound Practices in Medicaid Payment for Hospital Care. Center for Health Care Strategies, Inc.; 2010. Accessed June 26, 2019 http://www.chcs.org/media/Sound_Medicaid_Purchasing_FINAL.pdf
- 11.Shimotsu S, Roehrl A, McCarty M, et al. Increased likelihood of missed appointments (“no shows”) for racial/ethnic minorities in a safety net health system. J Prim Care Community Health. 2016;7(1):38–40. doi: 10.1177/2150131915599980 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Salit SA, Kuhn EM, Hartz AJ, Vu JM, Mosso AL. Hospitalization costs associated with homelessness in New York City. N Engl J Med. 1998;338(24):1734–1740. doi: 10.1056/NEJM199806113382406 [DOI] [PubMed] [Google Scholar]
- 13.Closing the Quality Gap: A Critical Analysis of Quality Improvement Strategies: Volume 7 - Care Coordination. Agency for Healthcare Research and Quality; 2007. Accessed June 26, 2019 https://www.ncbi.nlm.nih.gov/books/NBK44015/pdf/Bookshelf_NBK44015.pdf [PubMed]
- 14.Winkelman TNA, Admon LK, Jennings L, Shippee ND, Richardson CR, Bart G. Evaluation of amphetamine-related hospitalizations and associated clinical outcomes and costs in the United States.: 13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hwang SW, Aubry T. Hospital costs and length of stay among homeless patients admitted to medical, surgical, and psychiatric services. Medical Care. 2011;49(4):5. [DOI] [PubMed] [Google Scholar]
- 16.Vickery KD, Shippee ND, Bodurtha P, et al. Identifying homeless Medicaid enrollees using enrollment addresses. Health Services Research. Published online July 3, 2017. doi: 10.1111/1475-6773.12738 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Risk adjustment. Centers for Medicare & Medicaid Services. Published July 31, 2018. Accessed July 1, 2019 https://www.cms.gov/Medicare/Health-Plans/MedicareAdvtgSpecRateStats/Risk-Adjustors.html [PubMed]
- 18.Ronis SD, Grossberg R, Allen R, Hertz A, Kleinman LC. Estimated nonreimbursed costs for care coordination for children with medical complexity. Pediatrics. 2019;143(1):e20173562. doi: 10.1542/peds.2017-3562 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wholey DR, Finch M, Shippee ND, et al. Valuation of the State of Minnesota’s Health Care Homes Initiative: Evaluation Report for Years 2010–2014. Minnesota Department of Health; 2015. [Google Scholar]
- 20.Kheirkhah P, Feng Q, Travis LM, Tavakoli-Tabasi S, Sharafkhaneh A. Prevalence, predictors and economic consequences of no-shows. BMC Health Services Research. 2015;16(1). doi: 10.1186/s12913-015-1243-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Frank JW, Linder JA, Becker WC, Fiellin DA, Wang EA. Increased hospital and emergency department utilization by individuals with recent criminal justice involvement: results of a national survey. J Gen Intern Med. 2014;29(9):1226–1233. doi: 10.1007/s11606-014-2877-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.How Allina saved $13 million by optimizing length of stay Health Catalyst. Published February 10, 2016. Accessed April 28, 2020 https://www.healthcatalyst.com/success_stories/allina-saved-13-million-optimizing-length-of-stay/ [Google Scholar]
- 23.Finkelstein A, Taubman S, Wright B, et al. The Oregon Health Insurance Experiment: evidence from the first year. NBER Working Paper 17190. Published online July 2011. https://www.nber.org/papers/w17190.pdf [Google Scholar]
- 24.Wang Y, Liu Q. Comparison of Akaike information criterion (AIC) and Bayesian information criterion (BIC) in selection of stock–recruitment relationships. Fisheries Research. 2006;77(2):220–225. doi: 10.1016/j.fishres.2005.08.011 [DOI] [Google Scholar]
- 25.Winkelman TNA, Choi H, Davis MM. The Affordable Care Act, Insurance Coverage, and Health Care Utilization of Previously Incarcerated Young Men: 2008–2015. Am J Public Health. 2017;107(5):807–811. doi: 10.2105/AJPH.2017.303703 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Vickery KD, Bodurtha P, Winkelman TNA, et al. Cross-sector service use among high health care utilizers in Minnesota after Medicaid expansion. Health Affairs. 2018;37(1):62–69. doi: 10.1377/hlthaff.2017.0991 [DOI] [PubMed] [Google Scholar]
- 27.Wadhera RK, Choi E, Shen C, Yeh RW, Joynt Maddox KE. Trends, causes, and outcomes of hospitalizations for homeless individuals: a retrospective cohort study. Medical Care. 2019;57(1):21–27. doi: 10.1097/MLR.0000000000001015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Aalsma MC, Anderson VR, Schwartz K, et al. Preventive care use among justice-involved and non–justice-involved youth. Pediatrics. Published online October 2, 2017:e20171107. doi: 10.1542/peds.2017-1107 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Winkelman TNA, Admon LK, Jennings L, Shippee ND, Richardson CR, Bart G. Evaluation of amphetamine-related hospitalizations and associated clinical outcomes and costs in the United States. JAMA Netw Open. 2018;1(6). doi: 10.1001/jamanetworkopen.2018.3758 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Miller AC, Polgreen PM. Many opportunities to record, diagnose, or treat injection drug–related infections are missed: a population-based cohort study of inpatient and emergency department settings. Clinical Infectious Diseases. 2019;68(7):1166–1175. doi: 10.1093/cid/ciy632 [DOI] [PubMed] [Google Scholar]
- 31.Aronson L, Bautista CA, Covinsky K. Medicare and care coordination: expanding the clinician’s toolbox. JAMA. 2015;313(8):797. doi: 10.1001/jama.2014.18174 [DOI] [PubMed] [Google Scholar]
- 32.Capp R, Misky GJ, Lindrooth RC, et al. Coordination program reduced acute care use and increased primary care visits among frequent emergency care users. Health Affairs. 2017;36(10):1705–1711. doi: 10.1377/hlthaff.2017.0612 [DOI] [PubMed] [Google Scholar]
- 33.Hempstead K, DeLia D, Cantor JC, Nguyen T, Brenner J. The fragmentation of hospital use among a cohort of high utilizers: Implications for emerging care coordination strategies for patients with multiple chronic conditions. Medical Care. 2014;52:S67–S74. doi: 10.1097/MLR.0000000000000049 [DOI] [PubMed] [Google Scholar]
- 34.Gottlieb L, Sandel M, Adler NE. Collecting and applying data on social determinants of health in health care settings. JAMA Internal Medicine. 2013;173(11):1017. doi: 10.1001/jamainternmed.2013.560 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.