The Impact of Medically Tailored Meals Interventions on Hospital Utilization in Maryland

HIGHLIGHTS

• •Medically tailored meals resulted in lower hospital costs, visits, and Potentially Avoidable Utilization.
• •High-cost clients experienced a decrease in costs, whereas all others had a slight increase.
• •Reductions in visits and Potentially Avoidable Utilization were larger for high-cost patients.
• •Medically tailored meals may offer unique hospital cost savings for those with chronic illnesses.

Abstract

Introduction

This study sought to understand the extent to which receiving medically tailored meal interventions affects participants’ hospitalization costs and visits, focusing on high-cost clients.

Methods

The analyses explored pre- and post-enrollment data from 2,943 medically tailored meal clients who experienced inpatient hospital care in Maryland between 2021 and 2023. Unadjusted and adjusted difference-in-differences analyses examined whether changes in hospital charges and visits were different between high-cost clients and the rest of the sample. Differences in demographic characteristics were also explored.

Results

On average, high-cost clients receiving medically tailored meal services experienced a 64% decrease in hospital charges from the pre- to post-enrollment period; the rest of the sample experienced an increase of 38%. High-cost clients saw an average reduction of 45% in hospital visits and a 47% reduction in Potentially Avoidable Utilization visits, whereas the rest of the sample saw a decrease of 11% in visits and no change in Potentially Avoidable Utilization visits.

Conclusions

This study reveals that those with high hospital costs enrolled in 1 of 2 medically tailored meal programs experienced significantly higher reductions in hospital charges, Potentially Avoidable Utilization, and visits than those with lower costs. Future research into the characteristics associated with higher levels of utilization and healthcare spending is warranted.

INTRODUCTION

Following a healthy diet can reduce the risk of chronic diseases.1, 2, 3 Medically tailored meal (MTM) interventions use food-based interventions to help prevent, manage, and treat diet-sensitive chronic illnesses. MTM clients are referred by healthcare providers, assessed through an eligibility process, and receive care from a registered dietitian nutritionist. A meal and care plan is tailored to the medical circumstances of the client, and meals are prepared and provided to the client by the organization.⁴ Some research has found that MTMs can, among other outcomes, increase patient adherence to dietary recommendations and food insecurity and improve clinical outcomes.5, 6, 7 Other studies have found both actual and estimated reduced hospital utilization.8, 9, 10, 11

Moveable Feast and Food & Friends are community-based organizations (CBOs) that provide MTMs to thousands of clients in Maryland. Both CBOs are accredited as MTM providers by the Food Is Medicine Coalition. In 2024, Moveable Feast served 2,485 clients with 424,000 meals in 15 Maryland counties and Baltimore City. The standard meals delivered included 10 frozen entrees and a bag with 10 servings of fresh produce. During 2024, Food & Friends delivered nearly 380,000 MTMs to 1,471 clients in 9 Maryland counties. The standard meal delivery included 18 meals per week and a produce bag.

This study sought to understand the extent to which receiving MTM services affects participants’ hospitalization costs and visits. The study focused on high-cost clients or those who were in the top 25% of hospital charges during the 6 months prior to MTM enrollment. This analysis provides insights into how MTM programs can be targeted to the populations most likely to benefit through reduced hospital utilization and supports the growing evidence for MTMs as a cost-saving intervention, particularly in high-cost, low-income populations.

METHODS

The analysis explored 3 years of data, from January 1, 2021 to December 31, 2023, representing 2,943 MTM clients who experienced hospital care in Maryland during the study period. Approximately 59% of the clients were served by Moveable Feast, and 41% were served by Food & Friends. Data were obtained from the Chesapeake Regional Information System for our Patients, Maryland’s health information exchange, and consist of hospitalization information 6 months prior to client MTM enrollment (prestudy period), 6 months after enrollment (poststudy period), and internal client records from the CBOs. A secondary data IRB application was submitted to Solutions IRB and verified as exempt (#0634).

The key measures studied included inpatient (defined as inpatient admission, emergency department, and observation visit types) hospitalization costs, visit numbers, and Potentially Avoidable Utilization (PAU) visits.¹² Outpatient visits were excluded because they likely indicate desired utilization for managing chronic illnesses.

Chi-square tests, independent t-tests, and Mann–Whitney U tests were used to assess significant differences in the descriptive statistics between high-cost clients and the remaining sample. Unadjusted and adjusted difference-in-differences analyses were then conducted to examine whether changes in each of the 3 outcomes of interest from pre- to post-intervention periods were different between high-cost clients and the rest of the sample (Figures 1 and 2). Adjusted difference-in-differences models included gender as a covariate owing to its significance in bivariate testing; although they were also significant, chronic illness indicators were omitted from the models owing to high collinearity. Models were run in STATA 18.5.

Figure 1.

Differences in hospital charges from pre to postintervention periods (n=2,943).

Figure 2.

Differences in PAU and hospital visits from pre to postintervention periods (n=2,943).

PAU, Potentially Avoidable Utilization.

RESULTS

Of the hospital visits made by MTM clients, 66% of the visits were covered by Medicare, 23% were covered by Medicaid, 9% were covered by commercial or other insurance, and 2% were by self-pay or charity. The data originated from 57 hospitals or medical centers and from all 23 Maryland counties and the City of Baltimore.

Eighty percent of clients were categorized under 5 primary illnesses, including diabetes, cancer, cardiovascular disease, chronic kidney disease, and HIV. Cardiovascular disease and chronic kidney disease were observed significantly more frequently than expected in the high-cost client group, whereas diabetes was observed significantly less frequently (Table 1). Three common nutrition-sensitive comorbidities were studied, and the average number of comorbidities did not differ significantly between the 2 groups. Those who identified as men were observed significantly more frequently than expected in the high-cost client group. Those who identified as Black or African American made up the majority of all clients studied, followed by those identifying as White. The average age of clients was around 65 years. High-cost clients were not significantly different in the number of meals received or time spent in the program. In regard to hospital utilization, significant differences were found in hospital charges, visit numbers, and PAU visit numbers between high-cost and nonhigh-cost clients.

Table 1.

Demographic and Hospital Utilization Information for Clients

Variable	Overall	High-cost client	Non–high-cost client	p-value
Key primary illnesses	% (n=2,943)	% (n=736)	% (n=2,207)	<0.001
Diabetes	30.7	23.6	33.1	a
Cancer	15.6	17.0	15.2
Cardiovascular	14.3	17.0	13.4	a
Kidney/renal	10.7	14.7	9.4	a
HIV	9.3	7.6	9.9
Other	19.3	20.2	19.1
Number of comorbidities (average) (of 3 reported)	0.73	0.78	0.71	0.062
Gender	% (n=2,783)	% (n=691)	% (n=2,092)	0.026
Woman	60.5	56.9	61.7	a
Man	39.1	43.0	37.8	a
Gender diverse	0.4	0.1	0.5
Race	% (n=2,319)	% (n=610)	% (n=1,709)	0.481
American Indian/Alaskan Native	0.6	0.5	0.6
Asian	1.5	2.1	1.3
Black/African American	68.3	68.2	68.4
Native Hawaiian/Pacific Islander	0.1	0.0	0.1
White	26.7	25.7	27.0
Multiracial	2.8	3.4	2.6
Hispanic (ethnicity)	% (n=1,993)	% (n=736)	% (n=1,469)	0.490
	9.9	9.2	10.2
BIPOC (all races but White, non-Hispanic)	% (n=2,399)	% (n=623)	% (n=1,776)	0.502
	75.9	76.6	75.6
Age	Average (n=2,943)	Average (n=736)	Average (n=2,207)	0.613
	64.7	64.9	64.6
FPL	% (n=2,483)	% (n=646)	% (n=1,837)	0.724
≤100% FPL	66.7%	67.5%	66.4%
Between 100% and 138% FPL	12.0%	12.2%	11.9%
>138% FPL	21.4%	20.3%	21.8%
Cumulative meals	Meals (n=2,943)	Meals (n=736)	Meals (n=2,207)	0.576
	285.09	279.13	287.08
Time on Service	Days (n=2,943)	Days (n=736)	Days (n=2,207)	0.539
	183.03	179.13	184.33
Hospital utilization	Average (n=2,943)	Average (n=736)	Average (n=2,207)
Prehospital charges	$25,965.54	$77,893.34	$8,648.44	<0.001
Posthospital charges	$15,754.96	$27,498.51	$11,838.67	<0.001
Previsit number	2.10	4.18	1.40	<0.001
Postvisit number	1.51	2.32	1.24	<0.001
Pre-PAU visit number	0.42	1.11	0.19	<0.001
Post-PAU visit number	0.28	0.58	0.18	<0.001

^aAdjusted standardized residuals for the association between primary illnesses or gender and high-cost/no high-cost categories. Significant values (absolute value >1.96) indicate cells with counts that are significantly different from what is expected under the null hypothesis of independence.

BIPOC, Black, Indigenous, and people of color; FPL, federal poverty level; PAU, Potentially Avoidable Utilization.

All unadjusted models (n=2,783 clients) indicated significant difference in differences for all 3 outcomes (p<0.001). This was also the case for models adjusted for gender. On average, high-cost clients experienced a $49,542 (64%) decrease in hospital charges from pre- to post-intervention periods; the rest of the sample experienced an average increase in charges of $3,265 (a 38% increase) (Figure 1). High-cost clients saw an average reduction of 1.90 visits (45%), whereas the rest of the sample saw a decrease of 0.17 visits (12%) (Figure 2). High-cost clients had 0.52 fewer PAU visits (47%), on average, with no change for the rest of the sample (a decrease of 0.005 PAU visits).

DISCUSSION

Organizations providing MTMs to clients have limited budgets to provide meals to clients in need. Although research indicates that MTMs can have an impact across multiple populations and outcomes, gaps in research about which clients are most likely to benefit remain. The analyses indicate that high-cost clients may benefit distinctly from MTMs; decreases in hospital costs, PAU, and visits were uniquely significant for this group. Although they were less likely to have diabetes, high-cost clients were more likely to have cardiovascular or kidney diseases, indicating that there may be a relationship between these disease states, hospital utilization, and receipt of MTMs. Future research may examine this relationship to identify specific drivers of change in utilization outcomes from pre- to post-intervention periods for individuals with chronic illnesses. In addition, there is a group of clients (approximately 56% of the nonhigh-cost group) who increased their hospital costs from pre- to post-intervention periods; further examination of this group is warranted as well.

Limitations

This was a 1-group pre–post design with no control group, which introduces the possibility of regression to the mean and potentially overestimated effects. The study team attempted to mitigate the likelihood of regression to the mean by conducting sensitivity analyses and ultimately omitting bivariate outliers before determining inclusion in the high-cost patient group. Future research examining subgroup differences in pre–post outcomes should include a control group to ensure that findings like those included in this manuscript are reproduced. Many of the clients had multiple comorbidities, including diabetes, cardiovascular disease (including hypertension), and chronic kidney disease, in addition to their primary diagnosis. The analyses did not include which disease precipitated the indicated hospitalization. PAU models have smaller sample sizes and may be underpowered to detect actual effects. Furthermore, all demographic information is optional and self-reported by clients, resulting in substantial percentages of participants with missing values for race, ethnicity, gender, and income.

CONCLUSIONS

This study of MTM client hospitalization costs, visits, and PAU over a 3-year period reveals that high-cost clients at enrollment in 2 MTM interventions experienced significantly greater reductions in hospital charges, visits, and PAU than those with lower costs. Although the study suggests that chronic disease state and gender may play a role in whether a client is considered high cost, future research into what characteristics are associated with these higher levels of utilization and healthcare spending is warranted and will help organizations that provide MTMs to make important decisions about outreach, referral sources, policy recommendations, and funding.