Programs to improve health and lower cost by reducing food and housing insecurity: Do they work? What does the evidence say? Part two

Part Two: Our research


This is part two in a discussion of the impact of food and housing programs on health outcomes and healthcare cost.  We are looking at studies that quantify the effect of these programs.  See Part One for a discussion of the background and context for our research.

Our research

For the past year, I’ve been working with graduate students from Indiana University Purdue University at Indianapolis (IUPUI) under the direction of Dr. Josette Jones, to better understand the relationships between social determinants of health (SDOH) and health outcomes, health care utilization and health spending and to quantify the impact of benefit programs.

Many studies have confirmed the obvious:  There is a relationship between food insecurity or housing instability and health status and health cost.  On one level, that should be pretty obvious.  People who worry most about having enough money for food or housing tend to have poorer health than people who don’t worry as much about having enough money.  Conversely, people with poor health tend to have a harder time earning money for food or housing.

However, while the broad relationships is apparent, the underlying causes and effects that operate in these complex and multi-directional relationships is not well understood.  Similarly, there have been far fewer studies that measure or quantify the effect of interventions than of the broad relationships.  Studies are critical to estimating and evaluating the effect of future investments or cuts in programs.

Two questions:

  • Can we measure (quantify) the relationship between food insecurity and housing insecurity and health status or health cost? We found only a few studies that attempt to quantify the association between food and housing anxiety and health status/cost.  Most of these were limited to specific populations, such as the elderly, small children or people with AIDS/HIV.  Most of these did not allow for easy extrapolation to the nation as a whole.  So we conducted an analysis using a well-established national survey called the Behavioral Risk Factor Surveillance System (BRFSS).  The BRFSS n annual survey conducted in all 50 states plus territories.  It is managed by the Centers for Disease Control and administered by each state or territory.  We looked at data for 12 states (including the District of Columbia) from the 2015 survey and calculated that people who always or usually worry about having enough money for food or housing have a 48% probability of having poor or fair health (compared to good, very good or excellent health).  We also calculated that people who only sometimes worry about having enough money have a 27% probability of having poor or fair health.  You can read about our research in earlier blog posts here and here.

Of course, that only shows that there is a measurable relationship between food and housing anxiety and health.  It doesn’t measure the effect of benefit programs that affect one or more of the variables (food, housing or health) on the other variables.

  • Can we measure the effect of food or housing interventions on health status or health cost, and what do they show? This is the question we are pursuing through our present literature review.  We expect to complete and submit our paper for peer-review later in the fall.  However, given the present political debate about funding for benefit programs, particularly SNAP (See Part One), and now the $2 billion initiative by Mr. Bezos, I am presenting our early findings in this blog post.

The wrong pocket problem

Traditionally in the US, food insecurity and housing instability have been thought of as social conditions that are separate and distinct from health conditions, with related services provided by social service organizations that are not connected (or at most loosely connected) to the network of healthcare providers.  These separate silos contribute to what is called the “wrong pocket problem” where the organizations that fund or deliver social services do not themselves realize the financial benefit of improved health that results from their services.[i]  Conversely, organizations that pay for healthcare services, particularly safety net programs like Medicare and Medicaid may receive a benefit from social services but do not pay for or deliver the services themselves.  This structural disconnect makes it hard to evaluate the effectiveness of programs, or to align priorities, funding and service delivery.  The consequence is poorer outcomes by all measures.  And it makes studies that show the effect of interventions across the silos even more critical.


We summarized peer-reviewed literature that examined the impact of programs that address food insecurity or housing instability on health outcomes or health care spending.  We used PubMed, The Cochrane Library and Social Services Abstracts to execute our initial search and included relevant literature published in English between January, 2012 to August, 2018.  We ran a number of search strings comprised of a combination of social and health keywords including: “food insecurity,” “housing insecurity/instability,” health outcomes/status” and “health cost.”  We also conducted searches using specific programs as keywords such as “SNAP” and “WIC” as well as general types of interventions such as “home-delivered meals” and “permanent supportive housing.”  We specifically searched for studies of initiatives referred to in the Framework as well as for specific programs included in the Accountable Health Communities program. Because of the relatively small number of peer-reviewed articles identified in the initial search, particularly those looking at housing instability, we expanded our search to include reports and studies that have not been subject to peer-review but which, in our opinion, were of sufficient quality to warrant inclusion.  Most of these were independent evaluations of specific programs.

Eligibility criteria included: 1) inclusion of an intervention addressing food insecurity or housing instability, 2) quantitative measurement of a health outcome, healthcare costs or both, 3) well-documented study design.  We included experimental studies and natural experiments or quasi-experimental studies. Cross‐sectional studies were excluded as correlations are not able to shed light on changes in outcomes. Studies reporting only socio‐economic outcomes such as employment or productivity, were also excluded.

This preliminary search yielded 94 unique articles, 53 (56.4%) addressed food insecurity and 41 (43.6%) addressed housing instability.  Articles were independently screened and critically appraised by two review authors who met frequently to review decisions and discuss disagreements, which were resolved through negotiated consensus.  16 of the 94 articles met eligibility criteria based on a review of abstracts.

Intervention and health impact data were extracted by one review author and checked by a second review author. Studies were grouped according to broad intervention categories and context before synthesis. Because the 16 articles reported a range of health outcomes and cost metrics, the data were unable to accommodate a statistical meta-analysis.


Of the 16 studies that met all of our criteria, 10 addressed food insecurity while 6 addressed housing instability.  Of the 10 studies focusing on food interventions, 9 focused on SNAP as the intervention while 1 focused on home meal delivery.  All 6 studies that addressed housing instability focused on permanent supportive housing as the intervention.  Again, we anticipate expanding the number of studies in our review, particularly as we learn of non-peer-reviewed reports that meet our criteria which will change the number and balance of these results.

Study table 1

Study table 6

Study table 3

Study table 4

Study table 5

Study table 6

Study table 2

Summary of study designs

Food insecurity interventions

SNAP is by far the biggest food assistance program in the U.S. with over 42 million recipients so it is not surprising that the largest number of food studies would focus on SNAP.  While it is a federal program that is consistent across the nation, eligibility criteria can vary by state.  Because of its close association with Medicaid, data is available for various natural experiments that compare the receipt of SNAP food benefits with health information in Medicaid systems.

A natural experiment, also called a “quasi-experiment” is an empirical interventional study used to estimate the causal impact of an intervention on target population without random assignment. In other words, it’s a study where there is an intervention but where the control group, those who did not participate in the intervention, was not determined by random selection.  Natural experiments are subject to concerns regarding internal validity, because the treatment and control groups may not be comparable at baseline.  Nonetheless, natural experiments are a common and accepted approach when it is not possible to follow the practices associated with a true experimental study, as is the case with SNAP.

There were several variations of natural experiment study designs for SNAP.  In some studies, people who received SNAP benefits were compared to people who were eligible to receive SNAP benefits but who did not participate in the SNAP program.  In other studies, pre- and post-comparisons were made using the same people before becoming SNAP beneficiaries and while they were SNAP beneficiaries.  And in form of pre- post-comparison, people’s experience while on SNAP was compared to their experience after leaving the SNAP program.  Because SNAP eligibility is so closely aligned with Medicaid, some of these studies looked at not only the short term effect but also the mid- and long-term effect of participation in the SNAP program as expressed through Medicaid experience.

Some studies looked at health outcomes while others looked at health service utilization or health spending.  Some looked at aggregate data across the nation, while other studies looked at specific sub-populations, such as children, the elderly, people in a specific state, or specific health conditions such as HIV/AIDS.  (I have not included those limited to a specific health condition in this post.)

We also found one natural experiment that looked at the effect of home delivered meals, comparing eligible participating people with eligible non-participating people.

Housing insecurity interventions

Housing supports include permanent supportive housing, housing vouchers, low-income housing, heating credits and programs to improve or upgrade the quality of housing.  Broadly, programs can be distinguished between those designed to address chronic homelessness including PSH programs and those that help people with less severe forms of housing insecurity, offering heating credits or other improvements to property.

At this stage in our research, all of the studies we have included evaluate the effect of PSH programs that are designed to assist people with chronic homelessness and the complications that are associated with homelessness, including mental health conditions, substance use disorder and chronic physical health conditions.


Health Outcomes

Five of the studies measured health outcomes.  In one national study comparing people who received SNAP benefits with people who were eligible but did not receive benefits, the percentage of people who self-reported their health was poor or fair was 10.5% lower than among those who were eligible but did not participate in SNAP and 14.5% more reported their health was good or excellent than among those who were eligible but who did not participate in SNAP.

SNAP improvements bar graphOne national study looked at the impact of the Food Stamp Program (FSP), the predecessor of today’s SNAP program, on pregnancy, from 1961 to 1975.  The study showed that pregnancies exposed to FSP three months prior to birth yielded deliveries with an increased average birth weight.  The researchers concluded that FSP improved birth outcomes for both white and African American mothers, with a larger impact for African America mothers.

In another study of the FSP from the same period by the same researchers, access to food stamps in childhood lead to a significant reduction in the incidence of metabolic syndrome and, for women, an increase in economic self-sufficiency.

One study looked at the effect of losing eligibility for SNAP because of changes in immigrant status on families with small children.  The study found that adding another year of parental eligibility between the time children were in utero to age 4 led to large improvements in health outcomes at ages 6 to 16, demonstrating that early-life resource shocks impact later life health as early as school age.

Another study led by the same researcher who conducted the FSP studies  looked at the relationship between the real value of SNAP benefits (the fixed federal benefit amount adjusted for local price levels) to determine if variations in real buying power of SNAP benefits had an effect on health outcomes or health service utilization among children.  In a mixed result, the study showed that children in regions where with lower effective buying power used significantly less health services but may have used higher rates of emergency department use.  Interestingly, they found no effect on reported health status.

Health service utilization and cost

Eight of the studies measured health service utilization and/or cost, including four SNAP, one meal delivery and three permanent supportive housing.

One study of elderly SNAP eligible participants and non-participants in Maryland who were discharged from a skilled nursing facility found that an additional $10 of monthly SNAP assistance was associated with lower odds of admission to a nursing home and fewer days staying among those admitted.  A separate study of Maryland elderly people found that an additional $10 of monthly SNAP assistance also associated with a decrease in the likelihood of hospitalization, but not a decrease in emergency department use.  Overall, the researchers estimated that increasing SNAP participation to cover all those who are eligible but not participating could result in $34 million in reduced Medicaid expenditures for nursing home services and $19 million for reduced Medicaid hospital expenditures.

A study of 6.657 SNAP beneficiaries from 2002 to 2012 showed that participation in SNAP decreased the likelihood of hospitalization by 45.8% relative to eligible non-participants.

In a study of 4,447 noninstitutionalized adults of all ages (average age of 40.3 years) calculated that the average estimated annual cost of health care was $1,409 lower for SNAP participants than for eligible non-participants.  (Note:  According to the USDA Food Nutrition Service, the average monthly SNAP benefit payment in 2017 was $125.83, or $1,509.96 per year, which, considering only these measures, makes SNAP a roughly break-even financial proposition when adjusting for the wrong-pocket problem.)

Another study involving some of the same researchers showed that home meal delivery of medically tailored meals resulted in fewer in-patient admissions and lower medical spending.

With respect to permanent supportive housing, a study from Hennepin County, Minnesota, comparing pre- and post- health care service utilization for 123 residents, participants were admitted to hospitals 16% less often, visited emergency departments 35% less often and visited psychiatric emergency departments 18% less often.  Significantly, the study also showed that participants received outpatient clinic visits 21% more often than prior to entering the PSH program.

A pre- post- study from Los Angeles, California, of 890 participants in the Housing for Health (HFH) PSH program showed an average of 1.64 fewer emergency department visits in the first year of participation in the PSH program.  Using a broader notion of service utilization and cost including criminal justice involvement and other social services, the researchers calculated that the average spending excluding the PSH cost decreased from $38,146 per person to $15,358.  Even after taking into account the PSH costs, the researchers observed a 20% net cost savings.

Looking at the aggregate costs for 51 seniors in a PSH program, researchers conducting a pre- post- study from San Francisco estimated that 16, 433 days of care in skilled nursing facilities were avoided by participation in the PSH program.  They estimated a reduction of $1.46 million in hospital-based care in the first year of participation, and the avoided cost of $9.4 million for skilled nursing care over seven years.


Based on these studies, both SNAP and permanent supportive housing appear to generate a positive return on investment through reductions in healthcare and social supports costs, although, depending on the specific sub-population being studied and the definition of which programs will benefit, the range varies from about break even (cost of SNAP is roughly equal to the savings in lower health cost across the full range of SNAP participants) to cost avoidance in the millions per year at the state level (when focusing on a specific population like children or the elderly or including a broader range of services that benefit).  These studies also suggest that in addressing the wrong pocket problem, policy makers should take into consideration not only health spending but a broader set of costs that are affected by reducing food insecurity and housing instability.  These studies suggest that programs that benefit from reduced food and housing anxiety include the criminal justice system, substance use disorder, education, employment and child care.  All should be evaluated and considered as a part of the equation.

Research opportunities

With respect to the proposed changes in eligibility for SNAP or PSH programs, these studies do not constitute sufficient models to estimate the effect of the proposed federal or state requirements.  However, they may inform new studies that can look at the specific groups that are likely to be excluded from these benefit programs.  In the version of the Farm Bill passed by the House in June, the two groups most likely to be affected are adults between the ages of 18 and 59 who are not disabled and are not raising a child under six years old and families with children older than five years old.  This is an area ripe for study.

Similarly, as the above studies show, a significant percentage of those who are eligible for SNAP or PSH do not actually participate in SNAP or PSH programs.  States and communities may want to evaluate options to expand participation among those who meet the eligibility requirements whatever they are.

Alternatively, assuming that the combination of more restrictive eligibility requirements and general budget reductions for such programs are passed and reduce access to SNAP or PSH, states, communities as well as Medicare and Medicaid should evaluate alternatives to SNAP and PSH.  It may be that home meal delivery, food banks, community gardens, healthy food coupon programs or other variations of housing supports may yield an even better return with lower investment.  Because most of these programs are smaller and have not received significant federal funding, they require a different approach to capturing data and study.  There is an important role for funding by private foundations as well as insurers and even healthcare providers as revenue models shift from fee-for-service to value-based models.

Finally, all but one of these studies fail to directly address the questions of racial and ethnic disparities.  Such studies could lead to more insights which, when integrated with other programs, may result in even greater returns in individual health, population health, total cost and even provider satisfaction.


These issues lead to a question of sustainability.  Programs that can demonstrate that they can yield a significant financial return on investment while improving non-financial measures such as population health status, employment and justice involvement will stand a better chance of weather the swings in the nation’s politic environment.  The key to sustainability is to break down the silos that create the wrong pocket problem so that all the costs can be compared to all the benefits.

Taking down the silos means creating a new model of healthcare and social supports delivery.  One such model getting increased attention is the “community health hub” model which brings together not just physical health, behavioral health and social supports, it also brings in retail and consumer products vendors to create an integrated care universe.[ii]

As important, new models like the community health hub require new technology solutions that truly break down barriers and promote engagement.  Healthcare and social services solutions must learn the technology lessons of commercial enterprises ranging from financial solutions like Square to language education applications like Rosetta Stone and Duo Lingo.  These new technologies will also dove-tail with other emerging health care initiatives, like remote patient monitoring and telehealth.  The opportunity is huge.

Developing such a health hub will require tremendous leadership.  The studies presented here form the foundation, but much more research needs to be done to make this the ubiquitous approach that it needs to be.  Ultimately, it requires leadership across the spectrum to effect a real revolution in healthcare and social improvement.

This post presents the preliminary results of our in-process literature review.  We will update this post once we complete our analysis.  Nonetheless, we look forward to your comments, presented in a gentle voice.  If you know of additional research, please tell us about it either by posting a comment below or by contacting me directly at


[i] What is the “wrong pocket” problem and why is it important?  The National Housing Conference,

[ii][ii] Zeigler B, Reddy S, Leath B, et al.  Pathways Communty HUB:  A model for coordination of community health care.  Population Health Management, 11/2014.

Posted in Uncategorized

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s