Does It Work?

What the Evidence Shows — and How Confident We Can Be

The evidence base includes a randomized controlled trial, a 2025 observational study with a target trial emulation design, and program-level outcome data across multiple cities.

Three caveats apply across the entire evidence base. First, most documented outcomes come from a small number of well-resourced programs in major urban medical centers; rural, community hospital, and smaller-city results are largely undocumented. Second, program-reported data — which constitutes most of the available evidence — is not independently verified, and some programs may have enrolled a more motivated population that would have done better regardless. Third, the evidence on long-term outcomes beyond three years remains thin.

The Randomized Controlled Trial Evidence

Two randomized controlled trials constitute the strongest independent evidence for HVIPs.

Cooper, Eslinger, and Stolley (2006), published in the Journal of Trauma, found that the control group in a hospital-based violence intervention study was three times more likely to be arrested for a violent crime than participants who received the intervention. The study population was repeat victims of violence on parole or probation — a higher-risk adult population than most HVIP programs serve today.¹

Cheng et al. (2008), published in Pediatrics by researchers at Johns Hopkins University, conducted a randomized trial at two large urban hospital emergency departments, with a study population of youth aged 10-15 presenting with peer assault injury. The study found trends toward reduced misdemeanor activity, reduced aggression scores, and increased self-efficacy. The authors described results as “a trend toward significant program effects” — most findings did not reach conventional statistical significance.² The small sample size (113 with usable data) limited statistical power.

The Webster et al. (2022) review at Johns Hopkins, examining seven RCTs and six observational studies of HVIPs, concluded that “RCTs of the most robust HVIPs showed some evidence of protective effects, but overall evidence of reduced risks for violence was mixed” and that “RCTs were underpowered, and all but one were vulnerable to selection bias.”³

The Dosage Study: Boston University / Annals of Internal Medicine (2026)

A study published in the Annals of Internal Medicine in January 2026 adds important nuance to the evidence base. Researchers from Boston University School of Public Health examined Boston Medical Center’s Violence Intervention Advocacy Program, evaluating outcomes for young adults aged 16-34 who survived a shooting or stabbing. The study used a target trial emulation design with observational data, comparing outcomes between patients who received any engagement within one month of injury versus patients who received sustained engagement (initiating within one month and engaging more than four of the first eight weeks).⁴

The findings are both encouraging and clarifying:

Sustained engagement was associated with “considerably lower cumulative incidence” of the combined outcome of violent reinjury or violence perpetration at 1, 2, and 3 years, with a 55.3% risk reduction at three-year follow-up (6.4% vs. 14.3%).
Any engagement (even initial contact without sustained follow-up) produced outcomes “roughly equal between treatment and control strategies.” Initial contact alone did not produce durable results.
The researchers concluded that HVIPs can improve long-term violence outcomes, but that “these effects seem to require intensive participant engagement.”
The authors cautioned that because “there is no single, agreed-upon package of services for all HVIPs,” it is “unknown how our results may generalize to other HVIPs.”

The researchers concluded that HVIPs can improve long-term violence outcomes, but that “these effects seem to require intensive participant engagement.” They also cautioned that results may not generalize because “there is no single, agreed-upon package of services for all HVIPs.”⁵

Program-Level Reinjury Data

The following are program-reported outcomes, not independently verified, but specific enough to evaluate. The range across programs matters: these are not all equivalent results, and they come from different measurement windows, populations, and program designs.

Minneapolis Next Step: Reinjury rate reduced from a historical baseline of 41% (measured over five years) to 3% after one year of program operation. The difference in measurement windows requires caution in direct comparison, the 41% is a cumulative five-year figure and the 3% is a one-year figure. The magnitude of documented change is nonetheless dramatic and consistent with direction of effect found in independent studies.⁶

Indianapolis Prescription for Hope (Eskenazi Hospital): An initial study found that 3% of participants returned to the emergency department with a repeat violent injury within the first year, compared with 8.7% when the program was not operating — representing a two-thirds reduction in repeat injury requiring emergency care.⁷ This is among the most direct before/after comparisons in the field because it uses the same hospital’s own historical data as the control condition.

VCU Medical Center, Richmond, Virginia: Five-year recidivism rate of 3.6% for enrolled patients, compared to a national baseline of “almost up to 50%, if not a little bit higher.”⁸ Virginia’s 12 HVIPs statewide collectively report a re-injury rate of 3% for HVIP patients compared to a national average of 40%.⁹

Atlanta IVYY Project (Grady Memorial Hospital): Less than 2% of patients treated in coordination with the IVYY Project have returned with a gunshot wound, per Emory University School of Medicine announcement — “a reinjury rate far below the national standard of 30-40 percent.”¹⁰ First-year data from the program’s 2023 launch showed 1% reinjury among 450+ enrolled gunshot victims.¹¹

Cleveland University Hospitals: Reinjury rate reduced from 29% to 19% over five years with 600+ participants. This is the most modest reduction in the documented evidence base and represents a meaningful improvement even without reaching the 3% figures other programs report.¹²

Cincinnati Hope and Shield Network: In the first six months, 50 enrolled patients with zero documented repeat injury. The small sample size and brief window limit what can be concluded, but it is consistent with the direction of effect.¹³

Louisville Pivot to Peace: Of 628 patients presenting at University of Louisville Hospital with gunshot wounds or stab wounds between 2024 and 2025, 85% agreed to speak with an intervention specialist. In neighborhoods where the full Pivot to Peace model was operating, violent crime fell 62% in the first half of 2025 compared to the same period over the prior four years.¹⁴ This is the strongest recent city-level data in the documented evidence base, though the reduction reflects the full community violence intervention model, not the HVIP component in isolation.

Reinjury rates for HVIP participants across documented programs range from approximately 2% to 19%, compared to national baselines of 30-50%. The Cleveland University Hospitals figure — reinjury reduced from 29% to 19% over five years with 600+ participants — represents the longest documented program window in the evidence base, making it a useful counterpoint to programs reporting first-year data.¹⁵ Cincinnati’s Hope and Shield Network enrolled 50 patients in its first six months with zero documented repeat injuries, though the small sample and brief window limit what can be concluded.¹⁶ The VCU Medical Center in Richmond documented a five-year recidivism rate of 3.6% compared to a national baseline Dr. Aboutanos describes as “almost up to 50%, if not a little bit higher.”¹⁷

Employment and Educational Outcomes

A multi-program review of hospital-based violence intervention programs — summarized in HAVI field literature — found that 89% of program graduates either obtained employment, received a diploma, or completed general educational development (GED). The same review found that participants who gained employment were four times more likely to succeed in the intervention overall.¹⁸

Cost-Benefit

Virginia’s 12 hospital-based violence intervention programs are estimated to have generated over $82.5 million in health care costs avoided since 2019, nearly half of which represents direct savings to the state, through preventing and reducing re-injury rates.¹⁹ A separate analysis of HVIP cost-effectiveness found estimated cost savings ranging from $82,765 to $4 million per participant across a five-year model, driven by reduced injury recidivism, avoided trauma surgeries, reduced criminal justice costs, and maintained economic productivity.²⁰

The 2024 American Hospital Association report estimated that the total cost of violence to U.S. hospitals runs roughly $18 billion annually.²¹ The Health Alliance for Violence Intervention and Everytown for Gun Safety developed a cost calculator to help cities model their own program costs and returns.²²

What the Evidence Does Not Yet Establish

Long-term outcomes beyond three years. The Boston University study tracked participants for three years; most other studies measure first-year or first-six-months outcomes. Whether the reinjury reductions documented in year one persist through year three, year five, or longer is largely undocumented. The Boston study provides some evidence that sustained engagement produces durable effects, but replication across more programs and longer time horizons is needed.

Generalizability across program models. The Boston researchers noted that because there is “no single, agreed-upon package of services for all HVIPs,” it is unknown how their findings generalize to other programs. A program with 20 specialists working around the clock (UChicago) is structurally different from one with three staff and limited follow-up capacity.

Mechanism. The Boston University researchers documented that sustained engagement produces better outcomes than brief contact, but did not establish which specific components drive the effect — whether trauma counseling, employment support, the credible messenger relationship, safety planning, or some combination.²³

Rural and community hospital settings. Every program with documented outcomes operates in a major urban Level 1 Trauma Center. Whether the model produces comparable results in rural or community hospital settings — where patient volumes are lower, credible messenger talent pools are smaller, and community organization connections are thinner — is unknown.