Effect of Patient Care Coordination on Hospital Encounters and Related Costs

Abstract

Given high costs of hospital encounters, providers have increasingly turned to patient-centered health care programs to improve cost-effectiveness and population health for patients with high needs. Yet, evidence is mixed about program effectiveness. This pre–post comparative analysis assessed whether the number of hospital encounters and related costs decreased for patients who received care coordination services funded through Texas's 1115(a) Medicaid waiver incentive-based payment model, under which providers created new programs to improve care quality, population health, and cost-effectiveness. This study compared hospital records for patients who were frequent emergency department users at 4 urban safety net hospitals in Texas and state data for hospital encounters for the calendar years 2013 through 2015. The study included 9061 patients who frequently used emergency departments: 4117 patients who received waiver-funded care coordination services at 2 hospitals and 4944 patients who received usual care at 2 comparison hospitals. Regression models compared changes in patients' hospital use and length of stay for the 2 groups. Patients receiving waiver-funded care coordination had a 19% lower probability of hospitalization after receiving care coordination relative to patients who received usual care, for a mean savings of approximately $1500 per year per patient. Receiving care coordination was not associated with a change in length of stay. Care coordination developed by hospitals to meet the needs of their most vulnerable patients can reduce their use of hospital resources through better preventive care. These findings bolster the business case for care coordination, which may help ensure service continuation.

Introduction

When people are unable to manage their health conditions proactively or lack alternative treatment options, they often use hospitals as a major source for health care.^1

–4 This use of hospital resources is costly for patients, providers, and payers. In addition, although hospitals are often the appropriate source of care for emergent conditions, they provide inherently episodic care, with limited follow-up capacity. Better integration of care has become a major focus in efforts to improve population health and decrease costs of care.^5,6

Care coordination is one strategy among several^7,8 for integrating care for patients with chronic or complex diseases and high needs,⁹ specifically seeking to improve patient disease self-management and increase use of preventive health-related resources¹⁰ Care coordination, often referred to as care navigation, care planning, care management, or care transitions, generally entails identification by a designated staff member of factors causing a patient to visit the hospital emergency department (ED) frequently; development of a care plan tailored to that individual's needs, including both medical needs for care from multiple providers as well as needs for nonmedical services such as transportation, housing, and dietary planning; and facilitation of the primary health care and prevention services needed to reduce use of EDs and inpatient services, collectively termed “hospital encounters” in this study.^5,11

In usual care in hospitals, case managers typically focus on single-episode aspects of care such as transportation, medication, and immediate post-discharge care. By contrast, care coordinators tend to assume more holistic and potentially ongoing responsibility for supporting patients beyond a hospital encounter, including connecting patients with social services and community resources.¹² Prominent models for care coordination^13,14 highlight care coordinators' repeated interactions with patients over a period of weeks or months. Follow-up calls or visits ensure that this combination of resources and support provides continuity of care.^6,15 Collectively, this concentration of communication, coordination, and resources in response to a patient entering the health care system at a time of health crisis is intended to engage patients at the point when they are most amenable to learning from,¹⁶ and most likely to benefit from,¹⁷ care coordination.

One key indication of care coordination's success is how often people need hospital care (ED or inpatient services).^18,19 In addition, for those who are admitted to the hospital, length of hospital stay is a commonly used indicator of related resource use. Both measures relate to providing care more cost-effectively. Evidence associating care coordination services and decrease in hospital use is mixed. Some studies have found care coordination to be associated with reduced ED use and readmission rates,^18

–22 although others have not.^23

–29 Overall, the impact of care coordination services on hospital use and related costs has not been established.

Given the high costs of hospital encounters for both patients and payers, the objective of this study was to assess whether the number of hospital encounters and related costs decreased for patients who received care coordination services funded through Texas's 1115(a) Medicaid waiver incentive-based payment model. Under this program, providers could earn payments for new projects designed to address the Triple Aim to improve care quality, population health, and cost-effectiveness. The study team hypothesized that patients who received waiver-funded care navigation services would have fewer hospital encounters or shorter lengths of stay compared to similar patients who received usual care.

Methods

Site selection

The current study, a subsample of a larger study, used patient data from 2 large waiver-funded care coordination projects at 2 urban hospitals, compared to data for patients served by 2 comparison hospitals selected for their similarity. All 4 hospitals were designated safety net hospitals by the Texas Health and Human Services Commission. The hospital attributes (Table 1) used to choose comparison hospitals were ED trauma level; hospital ownership and payer mix³⁰; county urbanicity (using the US Department of Agriculture's Economic Research Services Rural-Urban Continuum Codes to classify counties by population density and level of urbanization³¹); and 3 county demographic attributes believed to be potentially relevant to the nature of the need for ED care, drawn from US Census data for 2013: percentage of the population in the county living below the federal poverty level; percentage black; and percentage Hispanic.³² These hospital matches were then validated with 3 former Texas hospital chief executive officers.

Table 1.

Site Attributes

	Waiver-funded site 1	Waiver-funded site 2	Comparison site 1	Comparison site 2
Primary county served
Population^*	4,300,000	800,000	300,000	800,000
Urbanicity^†	Metropolitan >1 million population	Metropolitan 250,000 to 1 million population	Metropolitan >1 million population	Metropolitan 250,000 to 1 million population
% county < Federal Poverty Line^‡	20	25	15	25
% Black^‡	20	5	15	5
% Hispanic^‡	40	80	25	80
Facility attributes
Ownership type	Public	Public	Public	Private
ED trauma level designation by American College of Surgeons	Level I–Comprehensive ED services 24/7 all age patients	Level I–Comprehensive ED services 24/7 all age patients	Level I–Comprehensive ED services 24/7 all age patients	Level III- handles emergency resuscitation, surgery, and intensive care
Safety net designation by the state	yes	yes	yes	yes
Number of ED visits in 2013^§	175,000	56,000	40,000	110,000
% ED visits Medicare in 2013^‡	5	10	10	20
% ED visits Medicaid in 2013^‡	15	15	20	35

Rounded to nearest 100,000 to protect identity of facilities in the study.

†

Rural-Urban Continuum Codes were developed by the US Department of Agriculture to categorize counties based on population density and proximity to metropolitan areas.³¹

‡

Rounded to nearest 5% to protect identity of facilities in study.

Rounded to nearest 1000 to protect identity of facilities in study.

ED, emergency department.

Both intervention hospitals and comparison hospitals had very similar system resources devoted to primary care initiatives and integrating care for patients with chronic or complex health conditions. For example, all 4 had case managers serving patients in the ED, primary care initiatives to provide patient-centered medical homes, and management programs for patients with chronic diseases. Three sites (2 intervention and 1 comparison) had discharge care transitions projects that provided more extensive services than typical discharge planning and provided home health care directly to patients with high needs.

The intervention hospitals differed from the comparison hospitals in that, in addition to existing case management and discharge planning services, they also offered waiver-funded care coordination. To qualify for waiver funding, those care coordination programs were required to be new services that supplemented, rather than replaced, existing services. The new care coordination programs sought to reduce patient disengagement from health care services, and specifically to reduce ED use among patients who frequented the ED. Both waiver programs used algorithms, which they adapted over time, to identify patients who might benefit most from receiving care coordination, and typically initiated care coordination services in a face-to-face encounter. At one site, care coordinators received a computerized flag when a patient with a history of frequent ED use presented at the ED again; the coordinator then met with the patient in the ED or telephoned the patient at home shortly thereafter to offer care coordination services. The second waiver-funded program offered care coordination services to inpatients before discharge who were considered to be at high risk for hospital readmission or ED use. Both programs also used designated care coordinators, typically nurses or social workers, often referring patients to existing patient care services provided by other programs in the hospital, as well as to nonmedical services such as food and housing assistance.

Patient sample

All 4 hospitals provided patient identifiers, demographic data, and dates of service for patients who were frequent ED users at their facilities, defined as patients who had visited the ED 5 or more times during the previous year. In both hospitals with waiver-funded care coordination, the number of patients who fit that description exceeded the hospital's capacity to provide services, so they offered navigation services to subsets of these frequent ED users. One hospital with waiver-funded care coordination changed the eligibility criteria to 4 ED visits within the last 6 months; the other offered coordination services to hospitalized patients at high risk for subsequent ED use or hospital readmission, based on an algorithm that included social support and other nonmedical needs in addition to frequency of previous ED use. Together, the intervention sites provided data for 4117 recipients of coordination services since program inception (one in July 2013 and one in May 2014) through December 2015. To obtain a comparison population of comparable size and ED usage patterns, comparison hospitals provided rosters for the 4944 patients who had used the ED 5 or more times at their sites during calendar year 2015. Together, the intervention sites and comparison sites provided data for 9061 patients.

Hospital encounters

To assess associations between receiving waiver-funded care coordination services and related hospital encounters, the research team retrospectively compared state-maintained discharge records for patients who received waiver-funded care coordination with those for patients who frequented the ED at the 2 comparison hospitals. The research team combined the 4 lists into a single roster of patient identifiers (name; date of birth; social security number, if available; Medicaid ID number, if available; age; sex; race and ethnicity; dates of service; and insurance type). The Texas Department of State Health Services matched this patient roster to their respective state-maintained data on hospital encounters at any Texas hospital from January 2013 through December 2015, using a minimum of 3 unique identifier points for each patient to ensure correct matching. The state data on hospital encounters covered a period including a 6-month span before the first waiver-funded care coordination project began through December 2015, the last date for which state-wide discharge data were available.

The number of individuals the Texas Department of State Health Services identified in Texas Health Care Information Center records as having had a hospital encounter at any Texas Hospital at any point in the 3-year study period was 4249. Of those, 1992 were at intervention sites and 2256 were at comparison hospitals. When the state agency matched hospital patient rosters to state records of hospital encounters, they declared a match only when 3 or more unique identifiers matched exactly (name; date of birth; social security number, if available; Medicaid ID number, if available; age; sex; race and ethnicity; dates of service; and insurance type). Records with fewer than 3 matching unique identifiers between hospital rosters and state records of hospital encounters were not included among those identified as having a hospital encounter during the study period. However, as the hospital rosters contained some patient records with incomplete data, it is possible that some failed matches were because of incomplete data rather than because there was no hospital encounter for that patient. As the state could not conclusively identify which patients had no match because of incomplete data, sensitivity analyses were not possible and the current analyses were restricted to patients with at least 1 hospital encounter during the 3-year study period.

For each patient who received waiver-funded care coordination services, the first date of care coordination services (the index date) was drawn from care coordination records, producing a distribution of index dates for the study population. That distribution of index dates among patients who received the intervention was replicated among patients who received usual care at comparison hospitals, creating a hypothetical “index date” for each patient in the comparison population so that the 2 populations could be compared. The research team retained data only for patients in both groups who had a hospital encounter during the year before or after their individual index dates.

Each patient's most recent hospital encounter within the year before their index date was used to measure preintervention severity of illness, hospital charges, and length of stay. Hospital encounters post intervention were measured starting with the first hospital encounter at least 31 days after that patient's index date. The 31-day minimum duration of care coordination was used to allow waiver-funded care coordination services time to affect each patient's need for hospital resources. Texas Health Care Information Center discharge data did not reliably distinguish between outpatient and inpatient care during the study period, preventing separate measures thereof.

Illness severity and mortality

The best proxy available for patients' severity of illness and risk of mortality is the 3M™ All Patient Refined Diagnosis-Related Groups (APR-DRG).³³ The APR-DRG adjusts disease-specific diagnostic groups by patient age, 4 levels of illness severity, and 4 levels of mortality risk. The 4 Diagnosis-Related Group levels, with examples from diabetes,³³ are: Level 1–Minor (eg, uncomplicated diabetes); Level 2–Moderate (eg, diabetes with renal manifestation); Level 3–Major (eg, diabetes with ketoacidosis); Level 4–Extreme (eg, diabetes with hyperosmolar coma). 3M APR-DRGs are widely used for risk adjustment by providers, payers, and for public reporting.³⁴

Statistical analyses

A 2-stage process was used to test associations between patient receipt of waiver-funded care coordination and the probability of having a hospital encounter after the patient's individual index date; and for patients who had a hospital encounter, the length of stay.³⁵ The distribution of hospital encounters was heavily skewed, and therefore was dichotomized for regression models.

First, a logit regression tested associations between patients' receipt of waiver-funded care coordination and annual probability of a hospital encounter. The logit coefficient was translated into an average marginal effect to characterize the result as a change in probability. An average marginal effect estimates the population-averaged marginal difference in the probability of an outcome associated with a 1-unit change in a predictor. The difference in probability was estimated using the margins command in Stata 14.0 (StataCorp LLC, College Station, TX).

Regression modeling for length of stay was then restricted to patients from all 4 sites who had a hospital encounter after their individual index dates. Length of stay was modeled using ordinary least squares regression, with this outcome log-transformed to reduce skew in its distribution.³⁶ All regressions used Huber/White corrections to obtain robust standard error estimates. Analyses were conducted with STATA 14.0.

The mean cost per hospitalization in the year before each patient's index date was estimated by obtaining from state data the actual charges for each patient's hospitalization. The mean cost change related to difference in frequency of hospital encounters after receiving care coordination services was estimated by multiplying the difference in frequency of hospitalizations by the mean preintervention cost per hospitalization, using an established methodology for estimating hospital prices from state administrative data that is applied to both Medicare and Medicaid patients.^37,38

The overall net revenue percentage of charges for each hospital in the study for both years was obtained from Medicare cost reports, which includes a total facility income statement for each. Each patient's recorded charges were then multiplied by the Medicare payment rates for each case by hospital by year to estimate actual payments received for each patient's hospitalization. Costs from 2013 and 2014 were adjusted to 2015 dollars using the medical inflation rate of 2.3% for 2013–2014 and 2.5% for 2014–2015.³⁹ To estimate the mean change in effective cost attributable to decreased frequency of patients' hospitalization after receiving waiver-funded care coordination services compared to patients receiving usual care at comparison hospitals, the mean payment per hospitalization was then calculated.

The study protocol, surveys, and interview scripts were approved by the UTHealth Institutional Review Board (IRB); the IRB granted a consent waiver for hospitals to provide retrospective administrative and clinical data for patients who frequently used their EDs and for the Texas Department of Health Services to provide matching records on hospital encounters. All data were de-identified before analyses.

Results

From among the 9061 patients for whom hospital encounter data were requested, the 4249 who had at least 1 hospital encounter during the study period are descriptively profiled in Table 2. At waiver-funded care coordination sites, the percentage of patients without insurance were 50% and 93% for the 2 sites, respectively; the percentage of patients without insurance at comparison sites were 18% and 16%, respectively. Consistent with previous research,⁴⁰ most hospital encounters among all patients were for minor conditions (eg, uncomplicated diabetes) to moderate severity conditions (eg, diabetes with renal manifestation).

Table 2.

Attributes of Patients Who Had a Hospital Encounter During Study Period

	Waiver-funded Site 1 (N = 773)		Waiver-funded Site 2 (N = 1220)		Comparison Site 1 (N = 464)		Comparison Site 2 (N = 1792)		Overall (N = 4249)
Characteristic	N	%	N	%	N	%	N	%	N	%
Sex
Male	466	60	570	47	203	44	576	32	1815	43
Female	307	40	650	53	261	56	1216	68	2434	57
Age, yrs mean ± SD, (range)	50 ± 13	19–95	58 ± 14	19–94	54 ± 17	18–85	49 ± 22	18–101	53 ± 18	18–101
Race
White	202	26	828	68	294	63	1182	66	2506	59
Nonwhite	571	74	392	32	170	37	610	34	1743	41
Ethnicity
Hispanic	203	26	1068	88	75	16	1431	80	2777	65
Non-Hispanic	570	74	152	12	389	84	361	20	1472	35
Insurance status
Uninsured	385	50	1137	93	84	18	279	16	1885	44
Insured
Medicaid	59	8	18	1	30	6	204	11	311	7
Medicare	96	12	36	3	183	39	619	35	934	22
Other insured	233	30	29	2	167	36	690	39	1119	26
Baseline 3M APR-DRG^* Level
Level 1 (Minor)	187	24	183	15	83	18	568	32	1021	24
Level 2 (Moderate)	363	47	559	46	219	47	706	39	1847	43
Level 3 (Major)	169	22	388	32	138	30	414	23	1109	26
Level 4 (Extreme)	54	7	90	7	24	5	104	6	272	6

3M APR-DRG is All Patients Refined Diagnosis Related Groups: levels correspond to a patient's illness severity and risk of mortality: level 1 = minor severity of illness and risk of mortality; level 2 = moderate severity of illness and risk of mortality; level 3 = major severity of illness and risk of mortality; level 4 = extreme severity of illness and risk of mortality.

SD, standard deviation.

Descriptive statistics, shown in Table 3, indicate that on average, the frequency of hospital encounters per patient who received waiver-funded care coordination services decreased by 0.85 encounters during the year after patients' index dates, compared to patients at comparison sites who received usual care (2-sample t test P value ≤.001).

Table 3.

Hospital Encounters and Length of Stay Descriptive Statistics

	Waiver-funded care coordination sites (n = 1993)			Comparison sites (n = 2256)			P value for change pre- to post-index date
Characteristic	N	Mean ± SD	Range	N	Mean ± SD	Range	Intervention vs Comparison sites
# Hospital encounters per year							^*
During year before index date	1920	2.16 ± 2.38	(0–25)	1536	1.71 ± 2.24	(0–24)
During 1st year after index date	725	0.85 ± 1.89	(0–27)	1543	1.78 ± 2.25	(0–20)
Length of stay per hospital encounter
During year before index date	1920	5.73 ± 5.90	(1–102)	1536	5.90 ± 10.01	(1–286)
During 1st year after index date	725	5.72 ± 7.28	(1–105)	1543	4.89 ± 5.90	(1–91)

SD, standard deviation.

P ≤ .001.

The regression model shown in Table 4 indicates a −0.94 coefficient (P < .001) for the association between patients' receipt of waiver-funded care navigation services and having any hospital encounter subsequently. Average marginal effects translate this coefficient into a 19 percentage-point decrease in the probability of a hospital encounter in the first year after receiving waiver-funded care coordination services (P < .001), holding all other factors equal. Because the initial probability of a hospital encounter (expressed as a percentage) is compared to a second probability (also expressed as a percentage), that decrease in probability of a hospital encounter is reported as a decrease by the number of percentage points.

Table 4.

Logit Regression Predicting Hospital Encounters in 1st Year After Index Date (n = 4247)

	Hospital encounters in 1^st year after index date
	Coefficient	P value
Key independent variable
Receipt of waiver-funded care coordination	−0.94	***
Covariates
3M APR-DRG^* Level (severity of illness and risk of mortality) (reference: Level 2 Moderate)
Level 1 (Minor)	−0.27	**
Level 3 (Major)	0.14
Level 4 (Extreme)	−0.06
Male (reference: female)	0.18	*
Age	0.00	*
Nonwhite (reference: White)	0.22	**
Hispanic (reference: non-Hispanic)	−0.17	*
Insurance (reference: uninsured)
Medicaid	0.84	***
Medicare	1.10	***
Other insurance	0.73	***
Year/Quarter
2014/Q1	−0.23
2014/Q2	−0.28
2014/Q3	−0.28
2014/Q4	−0.42
2015/Q1	−0.82	**
2015/Q2	−0.90	***
2015/Q3	−1.49	***
2015/Q4	−3.02	***

^* P ≤ .05, ^** P ≤ .01, ^*** P ≤ .001.

The mean cost per hospital encounter for all patients before implementation of the care coordination programs at 2 sites was $8157, treated here as 100% of the mean cost of a baseline hospital encounter. Therefore, the cost savings associated with reduced probability of a hospital encounter for participants in this study after receiving waiver-funded care coordination was estimated as $8157 × 19% = $1550 per patient per year.

Results from the multiple regression model shown in Table 5 indicate that waiver-funded care coordination was not associated with any significant change in patients' length of stay.

Table 5.

Ordinary Least Squares Regression Predicting Length of Stay in 1st Year After Index Date (n = 2268)

	Length of stay in 1^st year after index date
	Coefficient	P value
Key independent variable
Receipt of waiver-funded care coordination	0.02
Covariates
3M APR-DRG^* Level (severity of illness and risk of mortality) (reference: Level 1 Minor)
Level 2 (Moderate)	−0.21	***
Level 3 (Major)	0.26	***
Level 4 (Extreme)	0.96	***
Male (reference: female)	0.07	*
Age	0.00	**
Nonwhite (reference: White)	−0.09	**
Hispanic (reference: non-Hispanic)	−0.15	***
Insurance (reference: uninsured)
Medicaid	0.17	**
Medicare	−0.01
Other insurance	−0.06
Year/Quarter
2014/Q1	−0.07
2014/Q2	0.09
2014/Q3	0.03
2014/Q4	0.07
2015/Q1	0.06
2015/Q2	0.04
2015/Q3	0.08
2015/Q4	−0.06

3M APR-DRG is All Patients Refined Diagnosis Related Groups: levels correspond to a patient's illness severity and risk of mortality: Level 1 = minor severity of illness and risk of mortality; Level 2 = moderate severity of illness and risk of mortality; Level 3 = major severity of illness and risk of mortality; Level 4 = extreme severity of illness and risk of mortality.

^* P ≤ .05, ^** P ≤ 01, ^*** P ≤ .001.

Discussion

Hospital-related care coordination is intended to improve patients' disease self-management and use of routine primary health care and related services, and hence reduce the need for hospital resources. One key outcome for care coordination is therefore hospital encounters and related resource use. The current study found that 2 care coordination programs funded through Texas's Medicaid 1115(a) waiver program were associated with decreased probabilities of hospital encounters, although not with any change in length of stay. The cost savings associated with the decreased probability of hospital encounters alone was substantial, at approximately $1500 per patient annually among those who had experienced any hospital encounter during the study period. The probability of hospital encounter decreased in 2015, perhaps indicating that the care coordination projects became more effective as the projects matured, staff adapted to new roles, and patients learned from receiving care coordination.

The null results for length of stay may indicate that once patients received hospital care, the needs of patients who had received waiver-funded care coordination did not differ from otherwise comparable patients who did not receive this service. This may reflect limits to the clinical impact of care coordination, relatively consistent routines for hospital care that did not differ for patients who may have actually now had better support for earlier discharge, or both.

These findings bolster the business case for providing care coordination to individuals with histories of high utilization. Prior to inception of the Medicaid waiver care coordination projects, usual practice at all 4 study hospitals had chronic care management programs for patients with chronic disease conditions, which included case management for patients with complex needs, often focusing on discharge planning; and primary care initiatives, such as patient-centered medical homes. Three of the sites had discharge care transitions projects that provided services beyond usual discharge planning. The same 3 sites also provided, rather than referred, home health care to patients with high needs. One of the sites operated a call center staffed with nurses to assist patients with deciding whether to visit an ED or to access services through some other route.

Two of the hospitals then used Medicaid waiver funds to hire care coordinators as well, who connected patients to medical homes and referred patients to the health and social services needed to reduce their need for future hospital resources, whether in the ED or as inpatients. The Medicaid waiver's incentive payment pool is a time-limited program intended to launch innovations that will become self-sufficient. Although cost-effectiveness analysis is beyond the scope of the current study, the findings suggest that savings from reduced hospital use may offset the costs of providing these additional care coordination services. Of at least equal import, these care coordination programs appear to help patients with complex needs use fewer hospital resources.

This study entailed some limitations of note. The nonexperimental design leaves open the possibility that patients at waiver-funded care coordination sites differed in unmeasured attributes from patients at the comparison sites, despite the inclusion of patient clinical data and other attributes in analyses to reduce such confounding. Reductions in hospital encounters over time could reflect regression to a long-term mean, but the comparison group design helps mitigate against this. The sample was drawn from only 4 hospitals, although the hospitals were all large facilities that serve very large patient populations and were located in distinct regions of Texas. The percentage of patients who were Hispanic is consistent with Texas demographics; although the percentage was higher than the US average, these findings are relevant to a country whose ethnic composition is becoming increasingly Hispanic.

Care coordination for people with complex needs likely will reduce inpatient care but increase preventive outpatient care. Hence, the total costs of care may rise with effective care coordination. It was not feasible to measure the total costs of care for patients in this study, who had a variety of payers, and in many instances, no payer. Future research should examine additional cost-related implications of care coordination for both providers and patients and their families. However, the current study's findings are noteworthy in their own right, both because hospital care is a major driver of health care costs in the United States, and because uncompensated care is a major financial challenge for hospitals.

The current results contribute to the literature through rigorous comparisons of 2 care coordination programs, relative to usual care for patients who had been using EDs frequently at 2 similar hospitals in Texas. Hence, the current study tips the balance of a mixed body of evidence in favor of reduced numbers of hospital encounters associated with care coordination in real-world settings. As the popularity of care coordination appears to continue unabated, more research in a range of real-world settings will contribute to an understanding of when and how care coordination can achieve cost-related outcomes, as well as improved patient care quality and health outcomes.

Footnotes

Author Disclosure Statement

The authors declare that there are no conflicts of interest. The authors received the following financial support: this study was funded by the Texas Health and Human Services Commission (HHSC) and the Center for Medicare and Medicaid Services (CMS) project #11-W-00278/6. The findings are those of the authors and do not necessarily represent the official position of HHSC or CMS.

References

Hoot

, Aronsky

. Systematic review of emergency department crowding: causes, effects, and solutions. Ann Emerg Med, 2008; 52:126–136.e121.

Johnson

, Ghildayal

, Ward

, Westgard

, Boland

, Hokanson

. Disparities in potentially avoidable emergency department (ED) care: ED visits for ambulatory care sensitive conditions. Med Care, 2012; 50:1020–1028.

Tang

, Stein

, Hsia

, Maselli

, Gonzales

. Trends and characteristics of US emergency department visits, 1997–2007. JAMA, 2010; 304:664–670.

Yoon

, Yano

, Altman

, et al. Reducing costs of acute care for ambulatory care-sensitive medical conditions: the central roles of comorbid mental illness. Med Care, 2012; 50:705–713.

Craig

, Eby

, Whittington

Care coordination model: better care at lower cost for people with multiple health and social needs. Boston, MA: Institute for Healthcare Improvement, 2011.

, Apostol

, Martinez

, Aslakson

. Continuity, coordination, and transitions of care for patients with serious and advanced illness: a systematic review of interventions. J Palliat Med, 2013; 16:436–445.

Hofmarcher

, Oxley

, Rusticelli

. Improved health system performance through better care coordination. Paris: OECD Publishing, 2007.

Singer

, Burgers

, Friedberg

, Rosenthal

, Leape

, Schneider

. Defining and measuring integrated patient care: promoting the next frontier in health care delivery. Med Care Res Rev, 2011; 68:112–127.

Nash

DB.

Care coordination: a crucial element in managing chronic conditions. Popul Health Manag, 2010; 13 suppl 1:S1.

10.

Ferrante

, Cohen

, Crosson

. Translating the patient navigator approach to meet the needs of primary care. J Am Board Fam Med, 2010; 23:736–744.

11.

McWilliams

JM.

Cost containment and the tale of care coordination. N Engl J Med, 2016; 375:2218–2220.

12.

Sherry

, Wolff

, Ballreich

, DuGoff

, Davis

, Anderson

. Bridging the silos of service delivery for high-need, high-cost individuals. Popul Health Manag, 2016; 19:421–428.

13.

Coleman

, Parry

, Chalmers

, Min

S-J

. The care transitions intervention: results of a randomized controlled trial. Arch Intern Med, 2006; 166:1822–1828.

14.

Coleman

, Austin

, Brach

, Wagner

. Evidence on the Chronic Care Model in the new millennium. Health Aff (Millwood), 2009; 28:75–85.

15.

Paskett

, Harrop

, Wells

. Patient navigation: an update on the state of the science. CA Cancer J Clin, 2011; 61:237–249.

16.

Berg

, Korn

, Thomas

, Klemka-Walden

, Bigony

, Newman

. Opening the door to coordination of care through teachable moments. Dis Manag, 2007; 10:285–292.

17.

Hawkins

, Ozminkowski

, Mujahid

, et al. Propensity to succeed: prioritizing individuals most likely to benefit from care coordination. Popul Health Manag, 2015; 18:402–411.

18.

Corbett

, Lim

, Davis

, Elkins

. Care coordination in the emergency department: improving outcomes for older patients. Aust Health Rev, 2005; 29:43–50.

19.

Shumway

, Boccellari

, O'Brien

, Okin

. Cost-effectiveness of clinical case management for ED frequent users: results of a randomized trial. Am J Emerg Med, 2008; 26:155–164.

20.

Johnson

, Jones

, Lulias

, Perry

. Practice innovation, health care utilization, and costs in a network of federally qualified health centers and hospitals for Medicaid enrollees. Popul Health Manag, 2018; 21:196–201.

21.

Biernacki

, Champagne

, Peng

, Maizel

, Turner

. Transformation of care: integrating the registered nurse care coordinator into the patient-centered medical home. Popul Health Manag, 2015; 18:330–336.

22.

Reistroffer

, Hearld

, Szychowski

. An examination of the relationship between care management with coaching for activation and patient outcomes. Am J Manag Care, 2017; 23:123.

23.

Horwitz

, Busch

, Balestracci

, Ellingson

, Rawlings

. Intensive intervention improves primary care follow-up for uninsured emergency department patients. Acad Emerg Med, 2005; 12:647–652.

24.

Block

, Ma

, Emerson

, Langley

, Torre Dde

, Noronha

. Does access to comprehensive outpatient care alter patterns of emergency department utilization among uninsured patients in East Baltimore?. J Prim Care Community Health, 2013; 4:143–147.

25.

Gardner

, Li

, Baier

, Butterfield

, Coleman

, Gravenstein

. Is implementation of the care transitions intervention associated with cost avoidance after hospital discharge?. J Gen Intern Med, 2014; 29:878–884.

26.

Kim

, Michalopoulos

, Kwong

, Warren

, Manno

. Telephone care management's effectiveness in coordinating care for Medicaid beneficiaries in managed care: a randomized controlled study. Health Serv Res, 2013; 48:1730–1749.

27.

Lee

K-H

, Davenport

. Can case management interventions reduce the number of emergency department visits by frequent users?. Health Care Manag (Frederick), 2006; 25:155–159.

28.

Peikes

, Chen

, Schore

, Brown

. Effects of care coordination on hospitalization, quality of care, and health care expenditures among Medicare beneficiaries: 15 randomized trials. JAMA, 2009; 301:603–618.

29.

Wong

FKY

, Chow

, Chang

, Lee

, Liu

. Effects of nurse follow-up on emergency room revisits: a randomized controlled trial. Soc Sci Med, 2004; 59:2207–2218.

30.

American Hospital Association (AHA). AHA Annual Survey Database. 2015. https://www.ahadataviewer.com/additional-data-products/AHA-Survey Accessed January 15, 2015 .

31.

US Department of Agriculture. Rural-Urban Continuum Codes. 2013. https://www.ers.usda.gov/data-products/rural-urban-continuum-codes.aspx#.UYJuVEpZRvY Accessed January 15, 2015 .

32.

U.S. Census Bureau. Quick Facts. 2013. https://www.census.gov/quickfacts/table/PST045216/00 Accessed January 15, 2015 .

33.

Lyons

. An Overview of 3M All Patient Refined Diagnostic Related Groups (3M APR DRG). 2012; Slide 12. http://www.fha.org/Health-Care-Issues/Finance-Reimbursement/Medicaid/Resources/Diagnosis-Related-Groups-DRGs.aspx Accessed August 20, 2018 .

34.

Goldfield

The evolution of diagnosis-related groups (DRGs): from its beginnings in case-mix and resource use theory, to its implementation for payment and now for its current utilization for quality within and outside the hospital. Qual Manag Healthc, 2010; 19:3–16.

35.

Wooldridge

Introductory econometrics: a modern approach. Mason, OH: Thomson Southwestern, 2003.

36.

Manning

, Mullahy

. Estimating log models: to transform or not to transform?. J Health Econ, 2001; 20:461–494.

37.

Van Horne

, Netherton

, Helton

, Fu

, Greeley

. The scope and trends of pediatric hospitalizations in Texas, 2004–2010. Hosp Pediatr, 2015; 5:390–398.

38.

Levit

, Friedman

, Wong

. Estimating inpatient hospital prices from state administrative data and hospital financial reports. Health Serv Res, 2013; 48:1779–1797.

39.

Bureau of Labor

Statistics

. Databases, Tables & Calculators by

Subject

. https://www.bls.gov/data/#prices Accessed April 19, 2017 .

40.

Shen

Applying the 3M All Patient Refined Diagnosis-Related Groups grouper to measure inpatient severity in the VA. Med Care, 2003; 41(6 suppl):II103–II110.