— June 2, 2026 · Medically Tailored Meals and Fewer Hospital Stays: A 3,238-Person Medicaid Evaluation

Table of Contents

1. The Study at a Glance
2. What Was Done
3. What Was Found, Number by Number
4. How the Comparison Was Built
5. How Meals Could Reduce Hospital Use
6. How It Fits With Earlier Research
7. What It Cannot Show
8. What the Researchers and Press Releases Said
9. What Remains Unknown
10. Sources and Primary Documents
11. Key Research Papers
12. Connections

The Study at a Glance

• Paper: Hager and colleagues, “Medically tailored meals receipt and healthcare utilization and costs in Massachusetts’ Medicaid demonstration,” Nature Medicine 32(6):2147–2154, published online June 2, 2026.
• Who did it: the team that carries out the independent evaluation of the Massachusetts Medicaid demonstration (a state medical school), a university food-is-medicine institute, and staff of the state Medicaid agency, the nonprofit meal provider and partner health-care organisations.
• Design: a non-randomised policy evaluation using Medicaid claims, with propensity-score overlap weighting and a difference-in-differences comparison of a six-month period before meals with the period on meals.
• Who was studied: adult MassHealth members under 65 with food insecurity and a qualifying health need, enrolled through 11 accountable care organisations, 2020 to March 2023.
• Size: 1,866 meal recipients (more than 90 days of meals) and 1,372 eligible non-recipients: 3,238 people.
• The meals: 10 meals a week (five lunches and five dinners) plus snacks, delivered once a week, usually for six months, at $125 per person per week.
• Main results: hospital admissions 31% lower, emergency visits 20% lower and total costs $3,433 lower per person in the meal group compared with the comparison group; 98% of meal costs offset; no change in primary care visits.
• Evidence type: observational (quasi-experimental). It shows an association measured with care, not proof that the meals caused the difference.

Back to Table of Contents

What Was Done

The programme

Medicaid is the joint federal and state health insurance programme that, in the paper’s words, covers “71 million Americans with low income or disability.” States can test new kinds of care inside Medicaid through what are called Section 1115 demonstrations, or waivers, which must be evaluated. In Massachusetts’ 2017–2022 demonstration, US$149 million was authorised for a new Flexible Services Program, which let newly created accountable care organisations (groups of doctors and hospitals responsible for a population’s care) pay social service organisations to meet members’ nutrition and housing needs. Medically tailored meals were the most intensive and expensive nutrition service in it.

To qualify for the nutrition services, a member had to be enrolled in one of the accountable care organisations, be under 65, be food insecure and meet at least one of five health criteria: a behavioural-health diagnosis; a complex physical diagnosis such as diabetes or cardiovascular disease; frequent emergency department use; a high-risk pregnancy; or difficulty with everyday activities.

The meals

All meals came from one established nonprofit meal provider in Boston, the largest provider in the demonstration. Each week, ten meals (five lunches and five dinners) plus snacks were delivered to the home, usually for six months, with renewal possible after a clinical reassessment. Dietitians chose one of 15 meal plans for each person, matched to their diagnoses and preferences (for example vegetarian or soft meals). Everyone had a first nutrition assessment and counselling session with a dietitian and could attend three more, but few did: 62% attended no further session, 23% attended one, 10% two and 5% three. Meals went to the enrolled member only, because the programme was not allowed to pay for meals for the rest of the household. The provider was paid US$125 per person per week, covering food, counselling, delivery and overheads.

The comparison group

The comparison group was made up of 1,372 members who had been screened as eligible for the nutrition services but received none. The state’s records give the reasons: they lost contact during the referral process (736 people, 53.6%), declined the service (447, 32.6%), left their care organisation or moved (33, 2.4%), or another reason (174, 12.7%). Each of them was given the start and end dates of a randomly chosen meal recipient, so that both groups were observed over matching stretches of time.

What was measured

Medicaid claims captured care anywhere in Massachusetts. The three main outcomes were unplanned hospital admissions (planned and elective surgery excluded), emergency department visits and total health-care costs (the amounts Medicaid paid). Primary care visits were a secondary outcome: the researchers expected meals not to reduce them, and a fall would have suggested that people were simply using less care of every kind. The study was carried out as part of the independent evaluation of the demonstration required by the federal Centers for Medicare & Medicaid Services, and two ethics boards classed it as not human-subjects research.

Back to Table of Contents

What Was Found, Number by Number

Who the participants were

Meal recipients averaged 47.8 years of age, about six in ten were women and 37.7% were non-Hispanic White. Before weighting, recipients had more cardiometabolic illness than the comparison group: 54.5% had diabetes (vs 27.1%), 56.7% high blood pressure (vs 39.9%), 54.5% obesity (vs 41.1%) and 15.3% chronic kidney disease (vs 12.6%). Rates of depression (30.3% vs 30.5%) and anxiety disorders (60.8% vs 61.0%) were similar. In the six months before meals (unweighted), recipients averaged 0.7 hospital admissions and $17,639 in health-care costs, against 0.8 admissions and $14,513 in the comparison group. After weighting, every measured characteristic was the same in both groups (diabetes, for example, 38.3% vs 38.4%).

Main results

Mean enrolment in meals was 6.7 months (range 3.0 to 33.4), and the mean meal cost per person was $3,512. The weighted rates below are per six months, before and during the meal period.

• Hospital admissions: meal group 0.7 before and 0.5 during; comparison group 0.7 before and 1.0 during. Adjusted incidence rate ratio 0.69 (95% CI 0.58 to 0.82; P < 0.001), reported as 31% fewer admissions.
• Emergency department visits: meal group 0.9 before and 0.7 during; comparison group 0.9 before and 1.0 during. Ratio 0.80 (0.72 to 0.89; P < 0.001), 20% fewer visits.
• Total health-care costs: meal group $15,211 before and $13,427 during; comparison group $15,211 before and $16,645 during. Difference in change −$3,433 (95% CI −$5,651 to −$1,215; P < 0.01).
• Primary care visits: 5.5 before and 5.4 during with meals; 5.5 before and 5.6 during without. Ratio 0.99 (0.97 to 1.01): no change.
• Per 1,000 people: the paper reports absolute reductions of 443 hospital admissions (95% CI 216 to 671) and 403 emergency visits (184 to 621) per 1,000 people receiving meals, which the paper says reflects both the size of the relative effect and the high rates of hospital use in this population.
• Net cost: the lower medical spending offset 98% of the $3,512 meal cost, which the authors express as a net cost of $14.90 per person per month during the meal period.

Results by condition

The researchers repeated the analysis within groups of people with particular conditions, re-weighting each time.

• Cardiovascular disease (457 recipients, 251 comparators): admissions ratio 0.78 (0.66 to 0.96); emergency visits 0.84 (0.72 to 0.99); costs −$10,450 (−$18,597 to −$2,303).
• Diabetes (1,028 and 372): admissions 0.71 (0.55 to 0.91); emergency visits 0.81 (0.68 to 0.97); costs −$4,123 (−$8,078 to −$169).
• Chronic kidney disease (239 and 173): admissions 0.79 (0.68 to 0.92); emergency visits 0.97 (0.78 to 1.20), not a significant difference; costs −$12,312 (−$23,526 to −$1,098).
• Depression and anxiety disorders (1,147 and 837): admissions 0.68 (0.57 to 0.81); emergency visits 0.82 (0.73 to 0.92); costs −$5,597 (−$9,152 to −$2,041).
• Overall level of illness: among the least-ill third (by a standard morbidity score), there were no significant reductions in hospital use; among the most-ill third, costs were $9,733 lower (−$15,505 to −$3,961).
• Calendar period: the admissions ratio was 0.56 (0.41 to 0.77) in 2022–2023 and 0.77 (0.63 to 0.93) in 2020–2021; cost reductions were similar in both periods, and emergency visits did not change in 2020–2021.
• Everyone who got any meals: when people with 90 days of meals or fewer were included (2,882 recipients, 1,400 comparators), admissions (0.74; 0.62 to 0.89) and emergency visits (0.82; 0.71 to 0.94) were still lower, but the cost difference (−$1,086; −$3,000 to +$826) was not statistically significant.

Heart failure was not reported as a separate group; it falls within the paper’s description of the diet-sensitive conditions that medically tailored meals are designed for.

Back to Table of Contents

How the Comparison Was Built

Because no one was assigned to meals by chance, the researchers had to build a fair comparison from records. They used three main tools, each of which the paper describes.

• Before and after, in both groups (difference-in-differences). The study does not simply compare the two groups during the meal period. It compares how each group’s hospital use and costs changed from the six months before meals to the meal period. Differences between the groups that stay constant over time cancel out. The same method underlies the Chile food-label study reported on this site for June 11, 2026.
• Weighting (propensity-score overlap weights). For each person, a statistical model estimated the likelihood of receiving meals from demographics, social risks, 24 diagnoses, risk scores, care organisation, other programme enrolment and the previous six months of hospital use and costs. Overlap weights then give the most weight to people who were about equally likely to end up in either group, which the paper says mimics “a key attribute of clinical trials.” After weighting, all measured characteristics matched.
• A test period with no meals (negative control). The researchers ran the same analysis on two earlier six-month periods, 7 to 12 months and 1 to 6 months before the real baseline, when no one was receiving meals. If the groups had simply been drifting apart anyway, a “difference” would show up there too. None did: “no significant differences were observed between the treatment and comparison groups for any outcomes.”

The paper adds four sensitivity analyses: a crude analysis without weights; an analysis that adjusted for the same factors instead of weighting; weights that left out the baseline values of the outcomes, to test whether regression to the mean (an unusually bad period tending to be followed by a more ordinary one) could explain the result; and a second comparison group of Medicaid managed-care members aged 18 to 64 with high blood pressure, diabetes, cardiovascular disease or kidney disease who had never been screened or referred, and so could not have selected themselves out of the programme. The authors report the results were consistent across all of them. The detailed numbers sit in supplementary tables that were not part of this review.

Back to Table of Contents

How Meals Could Reduce Hospital Use

The study measured hospital use and costs, not what people ate or how their bodies responded. The authors propose several pathways and state plainly that “our study was not designed to tease out these causal pathways”:

• Better nutrition for the specific condition. The paper chose cardiovascular disease, diabetes and kidney disease for separate analysis because “these conditions have causal nutritional pathways and their clinical management is highly sensitive to diet.” Site pages describe the dietary research for each: the DASH diet for blood pressure, the kidney-friendly diet, and the type 2 diabetes page.
• Less food insecurity and fewer trade-offs. Meals may reduce what the paper calls “financial tradeoffs between food and other urgent needs such as rent or utilities.”
• More stable mental health. Depression and anxiety disorders were common in the study group.
• Greater medication adherence, the fourth pathway the authors list.

The study had no dietary data, so it cannot confirm that the meals were eaten. The authors point to an earlier randomised trial of a nearly identical meal service from the same provider, in people with diabetes and food insecurity, which found large improvements in diet quality (described in the next section). Because the comparison group received no counselling either, the authors describe the result as “a joint effect of healthy food delivery and nutrition counseling,” while noting that few people used the extra counselling sessions.

Back to Table of Contents

How It Fits With Earlier Research

Earlier observational studies pointed the same way. In a 2019 JAMA Internal Medicine study, Berkowitz and colleagues linked Massachusetts all-payer claims from 2011–2015 with records from a nonprofit meal provider, the same one as in the 2026 study, and matched 499 meal recipients to 521 non-recipients using an instrumental-variable method. Meal receipt was associated with fewer hospital admissions (incidence rate ratio 0.51; 95% CI 0.22 to 0.80), fewer skilled-nursing admissions (0.28; 0.01 to 0.60) and lower monthly costs ($3,838 vs $4,591; difference −$753, 95% CI −$1,225 to −$280). A 2018 Health Affairs study by the same lead author, in adults eligible for both Medicare and Medicaid, found that medically tailored meals were associated with fewer emergency visits, fewer admissions and lower spending, while delivered food that was not medically tailored was associated with fewer emergency visits and lower spending but not fewer admissions.

A small randomised trial showed better diets. In a 2019 randomised crossover trial in the Journal of General Internal Medicine, 44 food-insecure adults with diabetes and HbA1c above 8.0% received 10 meals a week for 12 weeks and usual care for another 12 weeks, in random order. Diet quality on the Healthy Eating Index was 71.3 with meals vs 39.9 without; food insecurity (42% vs 62%) and self-reported low blood sugar (47% vs 64%) were also lower.

Short trials after a hospital stay were mixed. The 2026 paper notes that “healthcare utilization has not been reduced in short-term trials testing brief (2−10-week) MTM programs after hospital discharge.” In a 2022 pragmatic trial in Medical Care, 1,977 adults leaving hospital with heart failure, diabetes or chronic kidney disease in an integrated health system received 10 weeks of meals or usual care; all-cause hospitalisation at 90 days did not differ (adjusted hazard ratio 1.02; 0.86 to 1.21), though in exploratory analyses deaths (0.65; 0.43 to 0.98) and heart failure admissions (0.53; 0.33 to 0.88) were lower. In the 2018 GOURMET-HF trial, 66 patients discharged after heart failure were randomised to four weeks of sodium-restricted DASH meals (1,500 mg sodium a day) or usual care; the main quality-of-life score improved similarly in both groups (P = 0.38), and 30-day heart failure readmissions (11% vs 27%; P = 0.06) did not reach statistical significance. From this and its own shorter-enrolment analysis, the 2026 paper suggests that longer enrolment may matter.

Wider Medicaid nutrition programmes. A 2025 Health Affairs evaluation by the same Massachusetts team looked at all of the state’s nutrition services together, of which medically tailored meals were the most intensive: among 20,403 participants in 17 care organisations compared with 2,108 eligible non-participants, participation was associated with 23% fewer hospitalisations and 13% fewer emergency visits, with cost reductions that were not statistically significant overall. In North Carolina’s Healthy Opportunities Pilots, a Medicaid 1115 programme in three regions of the state, 85% of services were food services such as healthy food boxes. A 2025 JAMA study compared 13,227 enrollees with 73,469 members who reported a social need but lived in counties without the programme: spending rose by $687 at enrolment, then trended $85 per person per month lower, matching the estimated spending without the programme by month 8 and falling below it afterwards. The emergency-visit trend was lower; the hospitalisation trend did not differ significantly.

National modelling. A 2022 JAMA Network Open simulation by Hager and colleagues estimated that if all of an estimated 6,309,998 insured US adults with a diet-sensitive condition and a limitation in daily activities received medically tailored meals, 1,594,000 hospitalisations (95% uncertainty interval 1,297,000 to 1,912,000) could be averted in a year, with estimated net savings of $13.6 billion (95% UI $0.2 billion to $28.5 billion). These are model estimates, not observed results.

Policy context. The 2026 paper reports that 13 US states had approved Medicaid Section 1115 demonstrations that include food-is-medicine programmes such as medically tailored meals, and three more had waivers awaiting federal approval; among the 16 states with nutrition waivers, Massachusetts was the first with large-scale meal programmes. A different food-is-medicine approach, weekly grocery deliveries to Medicaid patients with type 2 diabetes, was tested in a randomised trial reported on this site for July 21, 2026.

Back to Table of Contents

What It Cannot Show

• It is not randomised. The authors’ first limitation: “because receipt of MTMs was not randomly allocated, selection bias is possible.” If people who signed up for meals were more motivated or more engaged in their care in ways the records did not capture, the comparison “may overstate effects”; other unmeasured differences “could bias findings in either direction.”
• Who ended up in the comparison group. More than half of the comparison group had lost contact during referral and a third had declined. People who cannot be reached or who turn down a service may differ from those who accept it in ways that claims data do not record. The second comparison group, people never referred at all, was included to address this kind of selection.
• Weighting balances only what was measured. After weighting, every recorded characteristic matched. Characteristics that were never recorded cannot be balanced by weighting; randomisation is the design that balances them, on average.
• Part of the gap is a rise in the comparison group. The method compares changes, so a rise in hospital use among comparators (weighted admissions went from 0.7 to 1.0 per six months) counts toward the difference just as a fall among recipients (0.7 to 0.5) does. The negative-control and regression-to-the-mean checks were designed to test whether such patterns arise without meals; the paper reports that they did not.
• One provider, one state. All meals came from one established nonprofit with high nutrition standards. The authors write that “results may be less favorable with a lower-quality program or among healthier patients with less social and economic adversity.”
• Only while meals were arriving. Claims after the programme ended were not yet available, so the study cannot say whether lower hospital use lasted.
• No dietary or clinical measures. There were no data on what people ate, their blood sugar, blood pressure or kidney function; only health-care use and costs.
• A limited population. Participants were Medicaid members under 65, enrolled in accountable care organisations, all food insecure. Results for older adults on Medicare or for people with private insurance were not studied.
• Unusual years. The programme began in January 2020, and results differed by calendar period (for example, emergency visits did not change in 2020–2021).
• Short enrolment did not show cost savings. When everyone who received any meals was included, the cost difference was not statistically significant. A small share of start dates (under 5%) were estimated from enrolment quarters rather than exact dates.

Back to Table of Contents

What the Researchers and Press Releases Said

Tufts University issued a press release, dated June 3, 2026 in the copy read for this report. Kurt Hager, the first author, an assistant professor of population and quantitative health sciences at UMass Chan Medical School: “These findings show that medically tailored meals can be both clinically effective and economically sustainable within Medicaid.”

Dariush Mozaffarian, the senior author, a cardiologist who directs the university’s Food is Medicine Institute: “Our results show that food really is medicine, with major clinical and policy implications for health-insurance coverage of medically tailored meals to impact diet-related diseases and healthcare costs.” In the same release: “It’s rare to find anything in medicine that both improves health and saves money.” In an interview with the trade publication Nutrition Insight (June 4, 2026), he said: “Until recently, large-scale evidence has not been available. Most studies were relatively small or short-term.” On the negative-control check, he said the same models applied to periods before anyone was enrolled found “no differences between the groups in those periods, which is exactly what you’d expect if the program, not some pre-existing difference in the groups, was driving the results.”

Where the press coverage and the paper differ.

• The release and the Nutrition Insight article said participants received “10 meals per week—a mix of breakfasts, lunches, and dinners—plus snacks.” The paper says “ten meals (five lunches and five dinners) plus snacks.”
• The release said “per-person healthcare costs declined by $3,433.” The paper’s discussion uses similar wording, but its table defines the $3,433 as the difference in the change in costs between the meal group and the comparison group, not a fall in the meal group’s own spending.
• The release said “Medicaid patients receiving meals for longer had the largest improvements in healthcare costs.” In the paper this comes from comparing the main analysis (more than 90 days of meals) with the analysis that included shorter enrolments, and the authors word it more cautiously: longer enrolment “may be more likely” to be associated with cost reductions.
• Nutrition Insight described the participants as “hospital patients”; they were Medicaid members screened for food insecurity, not people selected during a hospital stay.

The release’s account of the limitations matches the paper, and its statement that primary care visits did not fall is supported by the data.

Funding and interests. The study was funded by the US National Institutes of Health, with the Massachusetts Executive Office of Health and Human Services funding three of the evaluation-team authors; the paper states that the funders had no role in the design, analysis or decision to publish. Several co-authors work for the state Medicaid agency, the meal provider or partner health-care organisations and helped run the programme. The senior author’s disclosures list advisory, consulting and equity relationships with a number of health, nutrition and technology companies, and the first author reports consulting for a nonprofit food-is-medicine coalition; the other authors report none.

Back to Table of Contents

What Remains Unknown

• Would a randomised trial at this scale find the same? The largest randomised trial described above gave meals for only 10 weeks after discharge. No large randomised trial of six months or more of meals with hospital use and cost outcomes was part of this June 2026 news.
• Does the effect last after meals stop? The authors write that if lower costs continued after the programme, the meals “could be even more cost-effective”; the data to test this were not yet available.
• Which part matters? The food itself, its tailoring, the weekly contact, the counselling and relief from food worry were delivered together.
• What changed in people’s health? Blood sugar, blood pressure, kidney function, weight and diet were not measured.
• Would it work elsewhere? With other meal providers, in other states, for people over 65 or for people with private insurance, the results could differ, as the authors note.
• Which conditions benefit most? Heart failure was not analysed separately, and the subgroup results for kidney disease and cardiovascular disease rest on a few hundred people each.

Back to Table of Contents

Sources and Primary Documents

Every study figure above was taken from the full text of the paper (open access under a Creative Commons licence, read through Europe PMC), not from the press coverage. Its supplementary tables were not read; where results rest on them, the text says so.

1. Hager K, Alcusky M, Zhang FF, Sing G, Ash A, Mick E, et al. (2026). Medically tailored meals receipt and healthcare utilization and costs in Massachusetts’ Medicaid demonstration. Nature Medicine 32(6):2147-2154. — doi:10.1038/s41591-026-04407-5 · PubMed PMID: 42230753
2. Full text (open access): PubMed Central PMC13279258
3. Thomas LE, Li F, Pencina MJ (2020). Overlap Weighting: A Propensity Score Method That Mimics Attributes of a Randomized Clinical Trial. JAMA 323(23):2417-2418. Methods reference cited by the paper. — PubMed PMID: 32369102

Press coverage: Tufts University release, “Medically Tailored Meals Produce Better Health and Lower Costs” (reprint dated June 3, 2026, copy read for this report); Nutrition Insight, June 4, 2026 (interview article).

Back to Table of Contents

Key Research Papers

Background papers on medically tailored meals and Medicaid nutrition programmes, in the order discussed above. Each entry was checked against PubMed and Crossref before it was written down.

1. Berkowitz SA, Terranova J, Randall L, Cranston K, Waters DB, Hsu J (2019). Association Between Receipt of a Medically Tailored Meal Program and Health Care Use. JAMA Internal Medicine 179(6):786-793. — PubMed PMID: 31009050
2. Berkowitz SA, Terranova J, Hill C, Ajayi T, Linsky T, Tishler LW, DeWalt DA (2018). Meal Delivery Programs Reduce The Use Of Costly Health Care In Dually Eligible Medicare And Medicaid Beneficiaries. Health Affairs 37(4):535-542. — PubMed PMID: 29608345
3. Berkowitz SA, Delahanty LM, Terranova J, Steiner B, Ruazol MP, Singh R, et al. (2019). Medically Tailored Meal Delivery for Diabetes Patients with Food Insecurity: a Randomized Cross-over Trial. Journal of General Internal Medicine 34(3):396-404. — PubMed PMID: 30421335
4. Go AS, Tan TC, Horiuchi KM, Laws D, Ambrosy AP, Lee KK, et al. (2022). Effect of Medically Tailored Meals on Clinical Outcomes in Recently Hospitalized High-Risk Adults. Medical Care 60(10):750-758. — PubMed PMID: 35972131
5. Hummel SL, Karmally W, Gillespie BW, Helmke S, Teruya S, Wells J, et al. (2018). Home-Delivered Meals Postdischarge From Heart Failure Hospitalization. Circulation: Heart Failure 11(8):e004886. — PubMed PMID: 30354562
6. Hager K, Sabatino M, Williams J, Ash AS, Halasa-Rappel Y, Flahive JM, et al. (2025). Medicaid Nutrition Supports Associated With Reductions In Hospitalizations And ED Visits In Massachusetts, 2020-23. Health Affairs 44(4):413-421. — PubMed PMID: 40193848
7. Berkowitz SA, Archibald J, Yu Z, LaPoint M, Ali S, Vu MB, et al. (2025). Medicaid Spending and Health-Related Social Needs in the North Carolina Healthy Opportunities Pilots Program. JAMA 333(12):1041-1050. — PubMed PMID: 40014313
8. Hager K, Cudhea FP, Wong JB, Berkowitz SA, Downer S, Lauren BN, Mozaffarian D (2022). Association of National Expansion of Insurance Coverage of Medically Tailored Meals With Estimated Hospitalizations and Health Care Expenditures in the US. JAMA Network Open 5(10):e2236898. — PubMed PMID: 36251292

PubMed Topic Searches

1. PubMed: Medically tailored meals and health-care use
2. PubMed: Food is medicine and Medicaid
3. PubMed: Food insecurity, hospitalisation and chronic disease
4. PubMed: Home-delivered meals and heart failure

Back to Table of Contents

Connections

• News Archive
• Grocery Deliveries Lower Blood Sugar in Type 2 Diabetes — the 460-person randomised grocery-delivery trial
• Chile’s Food Warning Labels and Child Excess Weight — another nutrition policy studied with difference-in-differences
• Type 2 Diabetes — diabetes affected 54.5% of meal recipients before weighting
• Heart Failure — a diet-sensitive condition the meals are designed for
• Kidney Disease — the subgroup with the largest cost difference
• The Kidney-Friendly Diet — sodium, potassium, phosphorus and protein in kidney disease
• Cardiovascular Disease — heart and blood-vessel disease and its risk factors
• Hypertension — high blood pressure, present in more than half of recipients
• DASH Diet — the eating pattern used in the GOURMET-HF meals
• Mediterranean Diet — a whole-food eating pattern and its trial evidence
• Depression — common among participants, with similar reductions in hospital use

Back to Table of Contents