Impact of state Medicaid expansion on cross‐sector health and social service networks: Evidence from a longitudinal cohort study

Abstract

Objective

To examine the impact of state Medicaid expansion on the delivery of population health activities in cross‐sector health and social services networks. Community networks are multisector, interorganizational networks that provide services ranging from the direct provision of individual social services to the implementation of population‐level initiatives addressing community outcomes.

Data Sources

We used data measuring the composition of cross‐sector population health networks 2006–2018 National Longitudinal Survey of Public Health Systems (NALSYS) linked with the Area Health Resource File.

Study Design

A difference‐in‐differences approach was used to examine the impact of expansion on organization engagement in population health activities and network structure.

Data Collection/Extraction Methods

Stratified random sampling of local public health jurisdictions in the United States. We restricted our data to jurisdictions serving populations of 100,000 or more and states that had NALSYS observations across all time periods, resulting in a final sample size of 667.

Principal Findings

Results from our adjusted difference‐in‐differences estimates indicated that Medicaid expansion was associated with a 2.3 percentage point increase in the density of population health networks (p < 0.10). Communities in states that expanded Medicaid experienced significant increases in the participation of local public health, local government, hospital, nonprofit, insurer, and K‐12 schools. Of the organizations with significant increases in expansion communities, nonprofits (7.7 percentage points, p < 0.01), local public health agencies (6.5 percentage points, p < 0.01), hospitals (5.8 percentage points, p < 0.01), and local government agencies (6.0 percentage points, p < 0.05) had the largest gains.

Conclusions

Our study found increases in cross‐sector participation in population health networks in states that expanded Medicaid compared with nonexpansion states, suggesting that additional coverage gains are associated with positive changes in population health network structure.

What is known on this topic

• While the impact of ACA's coverage expansion policies on access and outcomes has been studied extensively, less is known about the potential interaction between coverage expansion and population health initiatives

• Increased access to care under Medicaid expansion, coupled with heightened attention toward the social determinants of health in the Medicaid population, may have increased health and social services integration

• Medicaid enrollees are more likely to have complex medical and social needs that may be most effectively addressed using cross‐sector population health approaches that seek to improve community‐level health outcomes

What this study adds

• Population health networks in Medicaid expansion states experienced increases in the number of organizations participating in the delivery of population health activities when compared with their nonexpansion counterparts.

• A subset of organizations, including hospitals, insurers, local government agencies, local public health, and nonprofit organizations, increased their participation in population health networks as a result of Medicaid expansion.

1. INTRODUCTION

The Affordable Care Act (ACA) expanded health insurance coverage to nearly 30 million Americans by allowing states to expand eligibility for Medicaid and creating subsidies and standards to help people purchase private health insurance on state‐run and federally facilitated nongroup marketplaces. ¹ The ACA also included other policies designed to bolster the nation's public health system and improve population health. Most notably, the Prevention and Public Health Trust Fund created the federal government's first mandatory funding stream for prevention and public health programs. ² Other population health components within the ACA include enhanced regulatory and reporting requirements for nonprofit hospitals to engage in community health assessment and improvement activities, as well as incentives for health insurers and employers to invest in prevention and wellness activities. ³ While the ACA's coverage expansion policies have been studied extensively, much less attention focused on the impacts of the ACA's population health components. ⁴ , ⁵

A key policy question concerns the potential for interaction between the ACA's coverage expansion and population health components. One possibility is that as states have achieved improved coverage rates—particularly states expanding Medicaid eligibility—health care providers and public health agencies face reduced demand to provide safety net medical care for low‐income populations, thereby allowing these organizations to invest more effort and resources in the population health activities that are encouraged by other provisions of ACA. The increase in insured individuals accessing care under Medicaid expansion, coupled with growing awareness of the importance of the social determinants of health in the Medicaid population, may have acted as a mechanism for generating increased attention toward population health activities. ⁴ , ⁶ , ⁷ , ⁸ Medicaid enrollees are more likely to have both complex medical and social needs that may be most effectively addressed using population health activities that seek to improve community‐level health outcomes. ⁹ , ¹⁰

Population health activities include the monitoring of community health status, investigation of health problems and hazards in the community, educating and informing individuals about prevention, developing policies and laws to protect and promote individual and community health, and linking people to needed health and social services. ¹⁰ , ¹¹ , ¹² Key in the implementation of population health activities is the development of robust networks of collaborative relationships among public, private, and nonprofit organizations that work together and coordinate their activities to address health and social needs in the community. ¹³ , ¹⁴ Previous research has categorized the strength of these population health networks along the two dimensions of extensive and intensive margins. ¹⁵ , ¹⁶ A network can increase its extensive margin by creating collaborative relationships with larger numbers and types of organizations in the community, while intensive margin increases by implementing a broader array of population health activities through these collaborative relationships.

Our study examines changes in the implementation of population health activities, and changes in the strength of population health networks, following the implementation of the ACA. Of particular interest, we examine whether states that expanded Medicaid eligibility experienced larger gains in cross‐sector participation in population health networks compared with nonexpansion states.

2. STUDY DATA AND METHODS

2.1. Data source and sample

Data measuring the composition of cross‐sector population health networks delivering health and social services were pulled from the National Longitudinal Survey of Public Health Systems (NALSYS). NALSYS is a longitudinal cohort study that follows 360 local public health jurisdictions in the United States. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Communities were surveyed in 1998, 2006, 2012, 2014, 2016, and 2018 to measure the availability of 20 nationally recommended population health activities within their jurisdictions and the range of organizations that deliver each activity. The sample includes jurisdictions that serve a population of 100,000 or more residents. While the sample used in this analysis does not capture small jurisdictions, it does capture about 70% of the US population. NALSYS data from 2006 to 2018 were used in this study to examine two time periods before and three time periods after Medicaid expansion. Average response rate ranged from 68% to 73% and no evidence of systematic bias in response was identified. This study was determined exempt by the University of Kentucky Institutional Review Board.

2.2. Measures

The NALSYS questionnaire is a validated instrument used to collect information on the cross‐sector implementation of population health activities. Activities included in NALSYS are derived from national recommendations based on the core functions of public health and are considered key in protecting community health. ²¹ , ²² , ²³ , ²⁴ , ²⁵ Activity examples include the analysis of data on community health status and determinants, linking individuals to needed health and social services, engaging and communicating with other community stakeholders, and the development of policies to address health needs. Local public health officials are asked to report on the availability, perceived effectiveness, and network of organizations in each community that participate in each activity. We focused this analysis on changes in cross‐sector engagement in population health networks in Medicaid expansion states compared with nonexpansion. We used NALSYS measures examining organization contributions to core population health activities using measures of networks' intensive and extensive margins as our outcomes of interest.

To measure the extensive margin of population health networks, we used a measure of whole network density derived from NALSYS. Density is a common measure in social network analysis and is an indicator of overall network connectivity. ²⁶ , ²⁷ We measured density by taking the proportion of the number of cross‐sector relationships that exist in the network to the total possible relationships that can exist between organizations. A network density score of 100% would indicate that all organizations included in NALSYS are participating in the delivery of population health activities.

We measured individual sector intensive margin using data from the NALSYS questions that ask about organization participation in the delivery of population health activities. Total participation for each sector was generated by summing up the number of activities each organization was reported as participating in and dividing it by the total number of activities in NALSYS. Higher total participation indicates increased intensive margin on the part of each sector. Examining intensive margin allows us to better understand how Medicaid expansion impacted individual sector participation in population health networks.

2.3. Analytical methods

We used a difference‐in‐differences approach to compare changes over time in organization engagement in population health activities and network structure among (a) health jurisdictions in states that expanded Medicaid against (b) jurisdictions in states that did not expand Medicaid. ²⁸ Using this method allowed us to examine differences in cross‐sector participation in population health networks pre and post Medicaid expansion in treatment and control observations.

We defined treatment observations as population health networks in states that expanded Medicaid starting in 2014 and the control group as population health networks in those states that did not. States that did not have NALSYS observations across all time periods were not included in the analysis. Our final analytical sample included 21 states that expanded Medicaid and 14 that did not (see Table S4). While a number of states expanded Medicaid effective January 1, 2014, some states choose to implement policies that expanded Medicaid eligibility early and others have delayed. To account for this, we allowed our treatment indicator to vary based on the time of expansion adoption.

Multivariate linear regression models were used for each outcome of interest. All models included state and year fixed effects to control for unobserved time‐invariant differences across states. Standard errors were also clustered at the state level. NALSYS data were linked with county‐level information from the Area Health Resource File to control for population size, racial makeup of the population, age, the portion of the population below the poverty level, income, portion of the population that is uninsured, and the availability of hospital beds, primary care physicians, and federally qualified health centers. In most cases, this was a one‐to‐one match based on the county as most public health jurisdictions serve single county communities. In some cases, jurisdictions can be multicounty. We created a weighted average based on population size for data linked in these jurisdictions. All statistical analyses were performed using Stata 16. ²⁹

A key assumption of the difference‐in‐differences model is that the two groups of states (Medicaid expansion and nonexpansion states) experienced similar trends in outcomes prior to the implementation of ACA. While we did not have enough pre‐expansion data in NALSYS to conduct an event study to examine parallel trends, we did assess the data visually by mapping our unadjusted measures of extensive and intensive margin across the sample years (Figures 1, 2 and Figure S1) and analytically by estimating adjusted year‐specific time trends for the two groups of states using the pre‐expansion (2006 and 2012) data.

FIGURE 1.

Trends in average population health network density and degree centralization by year in expansion and nonexpansion states (%), 2006–2018. Authors' analysis of social network analysis variables from the National Longitudinal Survey of Public Health Systems, 2006–2018 data.

FIGURE 2.

Trends in average cross‐sector contributions to population health activities by year in expansion and nonexpansion states (%), 2006–2018. Authors' analysis of organizational contributions to population health from the National Longitudinal Survey of Public Health Systems, 2006–2018 data. Graphs mapping trends for additional sectors can be found in the Supplement.

3. SENSITIVITY TESTS

To examine the robustness of our results, we ran a set of sensitivity tests varying our state inclusion and exclusion criteria. Allowing the timing of expansion to vary can bias our model estimates downward. ³⁰ We tested two additional models excluding states that did not expand Medicaid on January 1, 2014 (IN, LA, MI, PA) and excluding New York because of substantial early expansion. We also ran a set of models examining changes in jurisdictions that had high poverty rates and high uninsured rates prior to 2014, as we would expect the impact of Medicaid expansion may be stronger in these communities.

4. RESULTS

Network density declined by 8.6% between 2006 and 2018, suggesting that the extensive margin in population health networks decreased during our study time period. National temporal trends in the level of individual sector participation in population health networks varied by sector. Insurers, universities, and hospitals increased their participation in population health activities by the greatest amount. Local government agencies, state health agencies, physicians, and federal government agencies average participation population health activities decreased by greater than 16% from 2006 to 2018 (Table 1).

TABLE 1.

Characteristics of cross‐sector population health networks, 2006–2018.

	2006	2012	2014	2016	2018	% change, 2006–2018
Extensive margin
Network density	17.6	13.0	13.7	13.8	16.1	−8.6
Intensive margin (multisector contributions to population health (%))
Local public health agencies	64.0	59.5	64.0	65.1	65.7	2.8
State health agencies	41.7	33.2	31.3	29.8	32.1	−23.0
Other state government agencies	14.8	11.6	10.5	10.1	14.2	−4.0
Local government agencies	45.9	23.2	28.9	29.9	34.6	−24.7
Federal government	10.7	7.8	6.3	6.1	9.0	−16.2
Physician organizations	21.9	17.8	15.8	16.2	17.9	−18.3
Hospitals	36.6	34.8	40.7	41.6	42.8	16.7
Community health centers	25.8	23.9	25.0	27.4	29.6	14.7
Faith‐based organizations	16.9	13.8	15.0	14.1	17.5	3.2
Nonprofit organizations	28.3	25.5	28.1	29.9	30.2	6.7
Insurers	8.7	8.4	9.2	10.4	13.8	58.3
Employers	14.7	11.7	12.9	13.0	13.9	−5.6
K‐12 schools	24.6	22.1	22.0	21.7	24.3	−1.2
Universities	19.2	18.7	19.7	20.3	23.0	20.2
Observations (communities)	233	241	278	332	296

Note: Authors' analysis of the National Longitudinal Survey of Public Health Systems, 2006–2018.

Stratifying communities by state Medicaid expansion status indicated variation in both extensive and intensive population health network characteristics. In the post‐period (2014 data forward), networks in expansion states experienced an increase in extensive margin, while extensive margin continued to decline in nonexpansion networks. Nonexpansion networks extensive margin increased in 2016 and average extensive margin in expansion and nonexpansion networks leveled out (Figure 1 and Table 2).

TABLE 2.

Difference‐in‐differences analysis of population health network characteristics and cross‐sector contributions to population health in Medicaid expansion and Nonexpansion states, 2006–2018.

	Expansion states		Nonexpansion states		Unadjusted difference‐in‐differences	Adjusted difference‐in differences (PP)
	Pre	Post	Pre	Post
Extensive margin
Network density	14.5	14.8	16.4	14.6	1.9	2.3*
Intensive margin (average sector contribution to population health (%))
Local public health agencies	60.5	66.7	64.2	63.3	6.0***	6.5***
State health agencies	34.5	29.2	40.4	33.0	2.4	2.3
Other state government agencies	11.8	11.0	15.0	12.0	2.5	2.7
Local government agencies	33.2	32.9	36.5	30.0	5.6*	6.0**
Federal government	8.7	6.7	10.2	7.6	1.0	1.1
Physician organizations	18.7	16.7	21.1	17.0	2.0	2.4
Hospitals	34.9	43.7	37.2	40.0	5.4***	5.8***
Community health centers	24.6	28.3	24.5	26.3	1.2	1.5
Faith‐based organizations	14.0	15.1	17.3	16.6	1.6	2.1
Nonprofit organizations	26.6	32.4	28.2	26.2	7.3***	7.7***
Insurers	9.1	13.5	8.1	8.1	4.3**	4.6**
Employers	11.9	13.0	15.0	14.2	1.7	2.0
K‐12 schools	21.7	23.0	26.2	23.1	3.9**	4.4**
Universities	18.4	21.1	20.2	21.3	1.9	2.2
Observations (communities)	157	230	115	165
Observations (community years)	241	504	188	353

Note: Pre includes two time periods (2006 and 2012), post includes three (2014, 2016, and 2018). All models include state and year fixed effects and control for community socioeconomic, demographic, and health care supply characteristics. Standard errors are clustered at the state level. PP stands for percentage point.

^*** p < 0.01;

^** p < 0.05;

^* p < 0.10.

Source: Authors' analysis of the National Longitudinal Survey of Public Health Systems 2006–2018, linked with Area Health Resource File.

Intensive margin increased in expansion states across most sectors, with the exception of state health agencies, other state government agencies, local government agencies, federal government, and physician organizations. In contrast, nonexpansion states saw increases in the intensive margin of only hospital, community health center, and university participation in population health activities between the pre and post time periods. The remaining sector intensive margins either decreased or remained stable in nonexpansion communities (Figure 2, Table 2, and Figure S1).

Results from our adjusted difference‐in‐differences estimates indicated that Medicaid expansion was associated with a 2.3 percentage point increase in the extensive margin of population health networks (p < 0.10). Communities in states that expanded Medicaid experienced significant increases in the intensive margin of local public health, other state government, local government, hospital, nonprofit, insurer, and K‐12 schools. Of those with significant increases in expansion communities, nonprofits (7.7 percentage points, p < 0.01), local public health agencies (6.5 percentage points, p < 0.01), hospitals (5.8 percentage points, p < 0.01), and local government agencies (6.0 percentage points, p < 0.05) had the largest gains in intensive margin (Table 2).

4.1. Parallel trends and sensitivity tests

Visually, we see that density diverged slightly in expansion and nonexpansion states prior to 2014, changing from 18% to 14% in expansion and 17% to 12% in nonexpansion. Other state government agencies, physicians, federal government, faith‐based organizations, and state health agencies also diverged in the pre‐expansion time period. The unadjusted line graphs also suggest that local public health and state health contributions converged prior to expansion. Nonprofits, hospitals, employers, and insurers had identical slopes prior to expansion while local government agencies, universities, community health centers, and K‐12 schools experienced slight differences (Figures 1, 2 and S1). Results from our analytical parallel trends test, however, suggest that all measures meet the parallel trends assumption when also adjusting for community characteristics (Table S2).

Following the example of previous analyses examining the impact of Medicaid expansion, we tested multiple model specifications with varying inclusion and exclusion criteria based on states that expanded late or early (Table S3). We found our results to be robust across all models. Minor changes in significance and size of the coefficients did occur as we allowed treatment to vary for late expansion states and tested stricter exclusion criteria. The most substantial variation in our estimates occurred in Model 3, where we excluded all states that expanded after January 1, 2014.

Limiting our sample to those jurisdictions serving high poverty communities produced stronger results across many of the sectors. We found that high poverty communities in states that expanded Medicaid experienced additional significant increases in the intensive margin of other state government agencies, physicians, and community health centers (Table S5). Interestingly, limiting our sample to those communities with high uninsured rates prior to 2014 did alter some of the results. Local government agencies and hospitals did not have significant increases in expansion states and our results were less precise overall. These changes may have been attributable to a substantial loss of sample size, as this model only included 626 observations.

5. DISCUSSION

Findings from our study suggest that population health networks in Medicaid expansion states experienced increases in the number of organizations participating in the delivery of population health activities when compared with their nonexpansion counterparts. Additionally, our results suggest that the gains in the population health network extensive margin may have been driven be increased participation on the part of select sectors in Medicaid expansion states. Taken together, our results suggest that Medicaid expansion is associated with greater cross‐sector engagement in population health networks. While more information on the mechanisms of change need to be explored, the additional reductions of uninsured individuals in expansion states is likely an important catalyst for change.

We found that a subset of organizations increased their participation in population health networks as a result of Medicaid expansion. Increased intensive margin on the part of hospital, insurer, local government agencies, local public health, and nonprofit organizations in expansion states is particularly salient given national efforts that emphasize the importance of integrating health and social services to address health outcomes and inequities. ³¹ , ³² Providing care to a greater number of individuals who may have complex social and medical needs can encourage engagement in population health activities with community partners, like nonprofits and local government agencies, as a way to reduce utilization and cost. As insurers and hospitals take on a larger population of Medicaid enrollees—many likely to have long‐running unmet medical and social needs—they may anticipate enrollees requiring more care and more costly procedures. To offset the financial risk to the insurer, they can adopt population health‐oriented approaches to address social factors in addition to health needs. Similarly, prior research found that decreases in the amount of uncompensated care being provided by hospitals after the ACA were greater in expansion states, but questioned how hospitals might reallocate resources. ³³ Our results indicate that hospitals in expansions states may offset reductions in uncompensated care by shifting resources toward community‐level population health initiatives.

Previous studies have found a number of positive health benefits associated with Medicaid expansion. ⁴ , ⁵ , ⁶ , ⁷ Researchers have primarily focused on the increase in insurance coverage as the catalyst for benefits accrued. Our results may indicate that strengthening of the population health system could also partially be responsible for positive benefits. Stronger connections between sectors may help facilitate access to care both through more streamlined referral between clinical providers and community organizations that provide social services like transportation. Indeed, stronger engagement of social services partners may also facilitate access to programs targeting unmet social needs that can influence both health and social outcomes impacting Medicaid enrollees. ⁹ Offering a robust package of population health activities through cross‐sector partnerships has also been linked to improved health outcomes. Positive health benefits in expansion states may result from the presence of a comprehensive population health network offering robust individual‐ and community‐level population health protections in addition to access to care through insurance coverage. ³⁴

Our study provides insight into the role Medicaid expansion plays in incentivizing population health strategies and cross‐sector engagement in population health networks. As conversations about the state of the ACA continue, how coverage and Medicaid expansion impact other systems of care should be a key component of discussions. Strengthening health and social services networks has the potential to increase health equity and improve outcomes by meeting both the social and medical needs of communities. ¹³ , ¹⁸ , ³⁵ Indeed, limited reductions in uncompensated care and additional financial challenges presented to hospitals in nonexpansion states might act as a deterrent to health and social service network engagement, crowding out resources that could be allocated toward population health activities. Similarly, insurers who have not experienced an uptick in enrollment in nonexpansion states may not be incentivized in moving toward population health initiatives. Recognizing that Medicaid expansion impacts both the medical and social systems individuals navigate adds additional complexity to the conversation about whether states should consider expansion options. Our results suggest that Medicaid expansion has the potential to increase not only coverage and access to care, but also the systems working to integrate health and social services.

6. LIMITATIONS

Our study has several limitations that should be taken into consideration. Data on the network are collected from the local public health official, meaning that we are generating network data based on one actor. Although this is a limitation, it is possible that our data underestimates the level of collaboration between organizations around population health activities. Our data capture three time periods after Medicaid expansion. Network building and the movement toward population health initiatives takes time, making it possible that longer term impact will occur. We were also limited to measures from 2006 and 2012 in our data prior to Medicaid expansion. Although this is a limitation, NALSYS is the only longitudinal data source tracking changes in population health systems and our results add an important contribution to the Medicaid expansion literature.

Our organizational measures look at broad sector categories and average tendencies; an examination of more detailed relationships and subgroup changes would add a level of nuance to the understanding of how Medicaid expansion is impacting multisector networks. We also focused our analyses on measures of cross‐sector organizational contributions to population health activities; future studies might look at more global measures of population health activity implementation in NALSYS to further understand the relationship with expansion. Medicaid expansion is not the only policy effect occurring, there are other initiatives encouraging health care organization engagement in population health. We attempted to control for other policy mechanisms that are national in scope by capitalizing on the additional effect of Medicaid expansion, which can be isolated. Lastly, our sample is limited to jurisdictions serving large populations and should not be generalized to rural communities.

7. CONCLUSIONS

Our study found increases in cross‐sector participation in population health networks in states that expanded Medicaid compared with nonexpansion states, suggesting that additional coverage gains are associated with positive changes in population health network structure.

FUNDING INFORMATION

This project was supported by grant number K01HS025494 from the Agency for Healthcare Research and Quality. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Agency for Healthcare Research and Quality. This project was also supported by the Robert Wood Johnson Foundation Systems for Action Research Program (#76689).

Supporting information

Figure S1. Trends in average multisector contributions to population health activities by year in expansion and nonexpansion states (%), 2006–2018.

Table S1. Model covariate descriptive statistics, expansion and nonexpansion states.

Table S2. Results from analytical test for parallel trends.

Table S3. Sensitivity testing of multiple DiD regression model specifications.

Table S4. State inclusion list.

Table S5. Sensitivity testing of multiple DiD regression model specifications, high poverty and high uninsured rate prior to 2014.

Hogg‐Graham R, Mamaril CB, Benitez JA, Gatton K, Mays GP. Impact of state Medicaid expansion on cross‐sector health and social service networks: Evidence from a longitudinal cohort study. Health Serv Res. 2023;58(3):634‐641. doi: 10.1111/1475-6773.14144