Leveraging Health Information Exchange (HIE) Data to Demonstrate the Impact on Health Outcomes of Community Based Organization Interventions: Case Studies of Housing Placement and Medically Tailored Meal Program

Abstract

Objective

To evaluate the use of longitudinal Health Information Exchange data to assess changes in healthcare utilization and selected clinical outcomes associated with community-based organization interventions addressing housing instability and food insecurity.

Methods

A retrospective pre-post study design was used to analyze two distinct community-based organization cohorts within a regional Health Information Exchange. The housing cohort included 228 individuals who received housing placement services, and the nutrition cohort included 786 individuals enrolled in a medically tailored meal program. Healthcare utilization and clinical outcomes were compared during the 365 days before and after program enrollment. Outcomes included inpatient admissions, emergency department visits, outpatient visits, length of stay, hemoglobin A1C, and body mass index. Paired t-tests were used to assess differences between pre- and post-enrollment periods.

Results

Among housing program participants, emergency department visits decreased by 32% (p<0.05), while outpatient visits increased by 92% (p<0.001). Changes in inpatient admissions and length of stay were not statistically significant. Among nutrition program participants, inpatient admissions decreased by 20% (p<0.01), emergency department visits decreased by 18% (p<0.01), and length of stay decreased by 5% (p<0.01). Significant improvements were also observed in hemoglobin HbA1c (5% decrease, p<0.05) and body mass index (4% decrease, p<0.01).

Conclusions

Longitudinal Health Information Exchange data can be used to evaluate healthcare utilization and clinical outcomes associated with community-based organization interventions. Findings suggest that housing and medically tailored meal programs are associated with improvements in healthcare utilization and selected clinical measures, while demonstrating the value of Health Information Exchanges as data repositories supporting whole-person care and program evaluation.

Introduction

Health Information Exchanges (HIE) are technology infrastructure organizations that enable secure electronic sharing of patient health information across a broad array of different healthcare organizations including hospitals, physician practices, laboratories, pharmacies, and public health agencies. In the United States, the development of HIEs was jump-started by the Federal Government with the 2009 passage the Health Information Technology for Economic and Clinical Health (HITECH) Act, which incentivized electronic health record adoption and interoperability.^1,2 Studies have shown increasing hospital participation in HIEs and growing evidence of the clinical and operational benefits of electronic data sharing.^3,4

Today, roughly 70–90 regional and state HIEs operate across the United States in support of several inter-related goals: improving care coordination across providers; providing more complete clinical data at point of care; reducing duplicate testing; and facilitating population health, public health reporting, analytics and quality measurement. Yet, HIEs are fragmented nationally – 10 states have a single HIE (e.g., Delaware, Wyoming, Alaska), some HIEs operate across multi-state regions, and other states have multiple HIEs within their boundaries – Pennsylvania, for example, has 6 HIEs within its borders. Whatever the coverage area, HIEs function as valuable longitudinal repositories of patient records containing demographics, encounters, diagnostics, procedures, vital signs, laboratory results, medications, allergies, and care summaries (Continuity of Care Documents – CCDs). HIEs thus serve as a critical component of the national health information infrastructure.

One such HIE, the HealthShare Exchange (HSX) provides accurate, real-time, patient data to partners in the greater Philadelphia region. HSX connects hospitals, physician practices, nursing homes, post-acute care facilities, health plans, home care organizations, and behavioral health providers. Since 2017, HSX has accumulated clinical data on over 9 million patients in its Clinical Data Repository (CDR), a longitudinal repository of patient-level clinical data compiled from over 500 participating organizations, representing more than 30,000 providers. All these organizations need the most accurate and up to date information about their patients to provide the best and most efficient care possible. In the past 2 years, HSX has expanded the circle of healthcare support by including Community-Based Organizations (CBOs) - that are also HIPAA Covered Entities - into its network of caregivers. Food banks, housing agencies, benefits coordinators, transportation agencies, community centers, and other community service organizations are recognized as playing a vital role in the care of the patient.

Indeed, research suggests that a substantial share of health outcomes are driven by factors beyond traditional medical care. ⁵ Referred to Health Related Social Needs (HRSN) or Social Determinants of Health (SDOH), issues such as housing instability, food insecurity, transportation barriers, inadequate access to healthcare, and other unmet social needs are being addressed by CBOs. The result is shifting healthcare from a reactive, disease-centered model to a more proactive, whole-person care approach – where medical care combined with outreach programs address these social needs and lead to more cost effective, heathier outcomes. In this regard, HSX has partnered through a state sponsored program initiated in 2024, with other HIEs and FindHelp ⁶ – a web-based social services search service - to connect patients based on their self-identified Health Related Social Needs to an applicable healthcare organization or CBO. Patients use the FindHelp platform to request assistance and these referrals flow both back to HSX and to the CBO healthcare teams. Once the patient is enrolled at the CBO to receive services, HSX sends real-time notifications to the care coordination teams at the CBO for inpatient encounters at both admission and discharge. HSX also provides rich clinical data on chronic conditions, medications, laboratory results, vital signs, physician notes, and other clinical details that enable CBOs to provide informed support to the cohorts they serve.

While CBOs play a major role in delivering social programs, they face multiple challenges – not only from a changing policy and funding landscape, but also from limited resources to measure the impact of their efforts. As a way to help CBOs demonstrate their impact on health outcomes that may lead to more partnerships with healthcare providers and payers, HSX has completed impact analyses for two metropolitan-area social service providers: 1) Pathways to Housing a CBO which provides stable housing arrangement for the homeless (hereafter ‘PTH’) and 2) Metropolitan Area Neighborhood Nutrition Alliance a CBO which provides medically tailored nutritional meals to those with significant chronic conditions, (hereafter ‘MANNA’). The purpose of this analysis is to show how longitudinal HIE data can be used to meaningfully evaluate changes in healthcare utilization and clinical outcomes associated with participation in these programs.

PTH provides housing to hundreds of people who were previously chronically homeless. HSX provides real-time alerts when residents are hospitalized or used the Emergency Department (ED), allowing PTH staff to provide on-going housing support, particularly when behavioral health or substance use disorder issues are present. ⁷ In contrast, MANNA provides medically tailored meals to individuals with serious conditions such as cancer, diabetes, renal disease, and HIV/AIDS. Their 'Food is Medicine' program emphasizes nutrition as a component of recovery, disease management, and overall wellness. HSX provides real-time alerts when clients are hospitalized, which helps MANNA monitor client status and reduce food waste.

Prior studies have demonstrated associations between housing support programs and reductions in healthcare utilization, including lower emergency department use and hospitalization rates among chronically unhoused populations.^8-11 Similarly, medically tailored meal programs have been associated with lower inpatient utilization and healthcare costs.^12,13 Most prior analyses have relied primarily on claims data or focused on individual health systems. Less is known about how longitudinal regional HIE data can be used to evaluate CBO interventions across multiple healthcare organizations. This study examines how HIE-derived clinical and utilization data can be used to evaluate outcomes associated with housing placement and medically tailored meal programs.

Materials and Methods – Regulatory Determination

The analysis qualifies as exempt research under the federal human subjects regulations because it involved only secondary use of retrospective clinical data. All data were analyzed in aggregate, there was no contact with individual patients, and no attempts were made to re-identify the information. Accordingly, the activity meets the criteria for exemption at 45 C.F.R. § 46.104(d)(4). HSX received the clinical data pursuant to executed HIPAA Business Associate Agreements with each participating data-sharing organization, including PTH and MANNA (each, an HSX member). These agreements, together with the governing Population Health Use Case, expressly authorize HSX’s use of patient data for retrospective population health analyses of this type and provide strong safeguards for health information privacy and security. All data included in the analysis were used in accordance with those permissions. No data were incorporated from HSX members who opted out of participation or from patients who declined to permit their health care provider to share data with HSX.

Methods

This study employs the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational research. A pre-post analytic design is used to evaluate changes in healthcare utilization and selected clinical outcomes associated with participation in these two separate community-based organization programs.

Study Cohort Composition

The PTH and MANNA patient extracts were analyzed as distinct cohorts because they involved different populations, enrollment periods, and outcome measures. Patients were included in the analysis based on CBO-provided enrollment dates and the availability of sufficient HSX data to support complete one-year pre-enrollment and one-year post-enrollment observation windows. PTH initially provided a roster of 496 patients with housing placement dates ranging from January 2018 through July 2023. MANNA initially provided 972 patients with enrollment dates from January 2022 through December 2022. These rosters were matched to the HSX Clinical Data Repository. Patients were excluded if no relevant clinical data were present in the CDR, if available dates did not allow for a full pre/post comparison period, or patients were physically located outside the HSX coverage area. After applying these criteria, the final analytic cohorts consisted of 228 PTH patients and 786 MANNA patients.

Gender and age distributions for these two panels are summarized in Table 1 and are reflective of the populations these two programs target. Men are more likely to be experiencing homelessness and account for 63% of the Housing panel – in line with national homelessness estimates. ¹⁴ Homeless people are concentrated in the 35-64 age cohorts – 85% of the total panel.

Table 1.

Gender and Age Distributions for Each Panel

PTH Housing panel - gender			MANNA Nutrition panel - gender
Gender	Patient count	% of total	Gender	Patient count	% of total
Female	84	37%	Female	467	59.4%
Male	144	63%	Male	319	40.6%
Grand Total	228	100%	Grand Total	786	100.0%

PTH Housing panel - age cohorts			MANNA Nutrition panel - age cohorts
Age cohort	Patient count	% of total	Age cohort	Patient count	% of total
25-34	14	6%	18-34	22	2.8%
35-44	89	39%	35-44	42	5.3%
45-54	64	28%	45-54	60	7.6%
55-64	41	18%	55-64	162	20.6%
65 years & older	20	9%	65 years & older	500	63.6%
Grand Total	228	100%	Grand Total	786	100.0%

MANNA eligibility is based on serving those with serious chronic conditions who most could benefit from medically tailored meals. In practice, their client population skews to lower income, middle aged to older adults, and more female - reflected in this demographic profile of MANNA patients.

Data Collection

Clinical and utilization data were obtained from the HSX Clinical Data Repository, which aggregates encounter, laboratory, and biometric information from participating health systems. For each patient, data were extracted for the 365-day period before and 365-day period after the CBO enrollment date. Encounter-level data included inpatient admissions, emergency department visits, and outpatient or ambulatory encounters. Laboratory and biometric data included hemoglobin A1C (HbA1C) and body mass index (BMI) where available. All data were aligned relative to each patient’s enrollment date, which served as the index date for defining the pre- and post-intervention observation windows. Data were normalized and organized at the patient level to support longitudinal comparison across the two time periods.

Outcome Measures

Healthcare utilization outcome metrics included counts of inpatient admissions, emergency department visits, and outpatient or ambulatory encounters during the pre- and post-enrollment periods (see Table 2 for operational definitions). Inpatient encounters were further used to calculate length of stay (LOS), defined as the duration of each hospitalization.

Table 2.

Definitions of Outcome Measures

Outcome measure	Definition	Data source	Calculation method
Inpatient (IP) admissions	Number of hospital inpatient stays per patient during the observation period	HSX CDR - encounter data	Count of distinct inpatient encounters per patient within the pre-enrollment (−365 to 0 days) and post-enrollment (0 to +365 days) periods relative to the enrollment date
Emergency department (ED) visits	Number of emergency department visits per patient during the observation period	HSX CDR, encounter data	Count of ED encounters per patient within the pre- and post-enrollment periods
Outpatient (OP) visits	Number of outpatient or ambulatory visits per patient	HSX CDR, encounter data	Count of outpatient/ambulatory encounters per patient within the pre- and post-enrollment periods
Length of stay (LOS)	Total number of days spent in the hospital per inpatient admission	HSX CDR, encounter data (admission and discharge dates)	Calculated as discharge date minus admission date for each inpatient stay
HbA1C (%)	Glycated hemoglobin level used to assess glycemic control among patients with diabetes	HSX CDR, laboratory data	For each patient, the lowest recorded HbA1C value within each period was selected; pre- and post-enrollment values were compared at the patient level
Body mass index (BMI)	Body mass index (kg/m2) used to assess weight status	HSX CDR, vital signs data	For each patient, the mean BMI value within each period was calculated; pre- and post-enrollment means were compared among patients with baseline BMI ≥25

Clinical outcome metrics focused on changes in HbA1C and BMI within the MANNA cohort. HbA1C analyses were limited to patients with diabetes and at least one recorded value in both the pre- and post-enrollment periods. Among these patients, if multiple values were present within a given period, the lowest value was selected to represent that period. Patients whose HbA1C values remained below 6.5% throughout the observation window were excluded, as they have not met criteria for diabetes and have maintained adequate glycemic control.

Data Analysis

Encounter activity and clinical measures were compared between the one-year pre-enrollment (PRE) and one-year post-enrollment (POST) periods for each patient. For each outcome measure, patient-level values were calculated separately for the PRE and POST periods and then compared within individuals. Paired t-tests were used to evaluate whether observed differences in healthcare utilization and clinical outcomes between the PRE and POST periods were statistically significant. All data preparation, normalization, and statistical analyses were conducted in Microsoft Excel using the Data Analysis Toolpak, specifically the paired two-sample t-test for means.

Results

Among the PTH cohort (n=228), a statistically significant reduction in ED visits was observed (32% decrease, p<.05), together with a statistically significant increase in outpatient visits (92% increase, p<.001) – see Table 3. Inpatient stays increased modestly, while average length of stay decreased, although neither change reached statistical significance. Among the MANNA cohort (n=786), statistically significant reductions were observed in inpatient stays (20% decrease, p<.01), average length of stay (5% decrease, p<.01), and ED visits (18% decrease, p<.01). HbA1C and BMI also improved significantly in the MANNA subset analyses – see Table 4.

Table 3.

PTH: Comparison of Key Metrics, 1-Year Pre vs. 1-Year Post Program Enrollment

Measure	One-year before	One-year after	Change	P
Number of Inpatient stays	197	207	5% increase	NS
Average Length of Inpatient stay (LOS)	8.0	7.4	7.5% decrease*	NS
Number of ED visits	616	419	32% decrease	<.05
Number of Outpatient visits	568	1,092	92% increase*	<.001

*Note. Excludes patients with LOS > 30 days.

**Note. 71% of the POST OP visits were from (homeless) patients with no PRE-placement OP Visits.

Table 4.

MANNA Comparison of Key Metrics, 1-Year Pre vs. 1-Year Post Program Enrollment

Measure	One-year before	One-year after	Change	P
Number of Inpatient stays	1,749	1,400	20% decrease	<.01
Average Length of Inpatient stay (LOS)	7.94	7.52	5% decrease	<.01
Number of ED visits	673	550	18% decrease	<.01
Number of Outpatient visits	13,076	12,745	3% decrease	NS
Average HbA1C	8.22	7.78	5% decrease*	<.05
Average BMI	33.6	32.2	4% decrease*	<.01

*Note. Patients with HbA1C values <6.5% for the duration of the 2-year analysis period were excluded.

**Note. Patients with starting BMI < 25 were excluded.

Discussion

Prior Housing First studies demonstrated reductions in emergency department utilization and hospitalizations among chronically unhoused populations.^8-11 The present analysis similarly found a 32% reduction in emergency department visits following PTH housing placement, from 616 in the year prior placement to 419 in the year post placement. Although inpatient admissions increased modestly and did not reach statistical significance, the reduction in average length of stay may still be operationally meaningful because stable housing and real-time HIE notifications may facilitate safer and timelier discharge planning.

Earlier medically tailored meal studies using claims data reported reductions in inpatient admissions and healthcare costs among program participants.^12,13 Similarly, the MANNA cohort demonstrated reductions in inpatient stays, emergency department visits, and length of stay following enrollment. Unlike prior claims-based analyses, however, the present study leveraged longitudinal HIE data across multiple healthcare organizations and additionally evaluated selected clinical biomarkers including HbA1C and BMI.

These findings suggest that regional HIE data repositories can serve as an important source for evaluating the impact of community-based interventions using longitudinal clinical data spanning multiple provider organizations and care settings. Unlike claims-based datasets, HIE data may provide more timely access to utilization, laboratory, biometric, and care coordination information across disparate healthcare systems. From a policy and payer perspective, these utilization changes are directionally encouraging. Based on estimated Medicaid reimbursements of $500 per ED visit and $600 for an inpatient day, the observed reductions suggest potential cost implications. ¹⁵

The ability to demonstrate program impact using empirical data within the broader context of whole patient health is important for CBOs to demonstrate value to healthcare partners. These findings build on the earlier claims-based studies by demonstrating directional improvements not only in healthcare utilization but also in selected clinical outcomes. At the same time, the observed PRE-to-POST changes should be interpreted with caution. Other factors beyond the scope of this analysis – including concurrent clinical interventions, behavioral change, regression to the mean, seasonality, and access to additional social supports – may also have contributed to the observed improvements. Future research should incorporate matched comparison groups, risk adjustment methodologies, and broader multisite analyses to better isolate the independent effect of community-based interventions on healthcare utilization and clinical outcomes.

Conclusions

This study demonstrates how longitudinal HIE data can be used to evaluate healthcare utilization and selected clinical outcomes associated with community-based interventions addressing housing instability and food insecurity. Across both programs, the findings suggest directional improvements in acute care utilization, while the MANNA cohort also demonstrated improvements in selected biomarkers. Although this study used a pre-post design without a control group and cannot establish causality, this analysis highlights the potential role of HIEs as enabling infrastructure facilitating closer collaboration between healthcare providers and community-based organizations. By providing longitudinal utilization, clinical, and real-time event data across multiple healthcare organizations, HIEs may help CBOs more rigorously evaluate program effectiveness and demonstrate value to healthcare partners, payers, and policymakers.