Program Goal 1: Reduce the number of retaliatory shootings involving high-risk males aged 16-30 in Cure Violence NYC intervention precincts by interrupting the system that perpetuates violence
Research Question1: Does participation in Cure Violence NYC reduce the rate of reported retaliatory shootings per 100 high-risk individuals within 12 months of 2024, compared to a matched non-program precinct with a similar history of retaliatory gun violence?
Measure:
The primary outcome will be the rate of reported retaliatory shootings per 100 high-risk individuals per year, a variable measured at the ratio level. A retaliatory shooting is operationally defined as any reported shooting that occurs within 72 hours of a prior violent incident involving the same individual or known associates and is independently flagged as retaliatory by both NYPD CompStat data and Cure Violence conflict mediation logs. Utilizing two independent sources enhances construct validity and is consistent with prior public health evaluations of violence prevention interventions.
Data Collection and Analysis:
High-risk individuals are defined as males aged 16–30 who have a documented history of involvement in violent incidents or prior arrests related to violent offenses as indicated by police records and program participant logs. In the intervention precinct, program participants will be identified through Cure Violence case management files and verified through NYPD data. In the matched comparison precinct, high-risk individuals will be identified using the same operational criteria but solely based on publicly available CompStat and arrest records.
Data will be collected from multiple sources to ensure a robust assessment. First, NYPD CompStat incident-level reports will be used to identify the location, time, victim and suspect details, and gang-related context for each shooting. Second, Cure Violence staff conflict logs will provide documentation of interpersonal disputes, potential retaliatory motives, and broader community-level dynamics. For the comparison precinct, only CompStat data will be used, as there are no equivalent conflict logs. Third, case management notes will be reviewed to track participant involvement in or proximity to violent events, providing critical context for classification decisions. Any reported shooting that does not have a victim, or where the victim is not connected in any way to the perpetrator, will be classified as “non-retaliatory,” even if the perpetrator falls within the Cure Violence target demographic. Since this study relies on secondary data sources, time and resource efficiencies are gained. However, it should be noted that verifying the full accuracy of CompStat data poses challenges, as few other agencies collect such detailed gun violence information independently.
Ethical safeguards will be implemented throughout the study. The analysis will rely exclusively on de-identified secondary data whenever possible. All identifiable information will be removed prior to analysis, and findings will be reported in aggregate to prevent the re-identification of individuals. Institutional Review Board review and approval will be sought prior to data collection. No direct contact with individuals from the comparison group will occur unless separate informed consent procedures are undertaken.
A quasi-experimental interrupted time series design will be used to compare the two precincts. The rate of retaliatory shootings per 100 high-risk individuals will be calculated for each precinct over the study period. Multiple regression analysis will be conducted to assess the effect of Cure Violence participation, adjusting for baseline violence rates and socio-economic indicators. Since both precincts are located within New York City, external factors such as legal frameworks, weather patterns, and policing structures are assumed to remain constant, minimizing confounding influences.
Nonetheless, the analysis remains subject to several limitations. The findings will be limited to reported shootings, recognizing that underreporting of gun violence is a persistent issue. In addition, while CompStat data is widely used and considered reliable within policing and public health research, discrepancies in reporting practices across precincts may introduce variability that cannot be fully controlled.
Research Question 2: Does participation in Cure Violence NYC increase the likelihood of high-risk individuals accessing mental health services by the end of 2024 compared to similar individuals in non-program neighborhoods as self-reported by participants?
The outcome variable for Question 2 is the number of males in the precinct between the age of 16 and 30 enrolled in mental health supportive services. It is an interval ratio variable. A participant is considered to have accessed mental health services if they have successfully attended at least one session with a licensed therapist.
Measure: We will gather a list of all the participants of the target demographic who were referred to a mental health service by Cure Violence and invite them for a survey. Based on the number of responses, we will post a call on community bulletins to identify similar respondents in the non-program neighborhood. The control group will be investigated in the same way, but we will not mention the "cure violence" program to reduce potential biases. All voluntary respondents will be compensated for their time with a voucher.
The survey will contain multiple choice items that ask about the respondents’ demographic data but not their name, perceived need for mental health services, whether or not they had seen a licensed therapist in the past year, and a rating scale of how easy or difficult it was to access. An open ended question will ask how many sessions each participant attended during the year.
The survey will be sent to voluntary participants through a link to their phones or email as this age group is assumed to be relatively tech savvy. Scheduled reminders will be sent to them over a period of 6 days. We hope that administering this short survey through a link increases participation as it does not require commuting. Once the survey results are in, we will normalize the data and then run an analysis to see if the differences in access to mental health services based on our operational definition is statistically significant.
Validity and Reliability(edited)
We have taken a number of steps to ensure the authenticity of our findings. First of all, before the questionnaire is officially distributed, community volunteers will be invited to fill in and give feedback on whether the questions are easy to understand, such as changing the professional term "psychological service utilization" to the colloquial "have you seen a psychologist", so as to avoid respondents answering randomly because they do not understand. All questionnaires are anonymous throughout the process, do not record private information such as name and address, only collect answers through encrypted links, and automatically issue a $10 e-coupon as a thank you after answering, which can not only reduce the risk of lying, but also increase the enthusiasm of participation. For possible contradictory answers (for example, someone ticks "I haven't seen a psychiatrist" but writes "went 3 times" in the open question), the system will automatically mark this kind of data, and analyze the reasons separately in the later stage - if it is an input error, you can contact me to correct it, and if it cannot be verified, the outliers will be eliminated to ensure the accuracy of the final analysis. In addition, miners need additional protections to participate: respondents aged 16-17 must provide an electronic receipt of parental signature, and a link to the questionnaire will be sent to the guardian's email address to ensure parents' right to know.
In practice, it is planned to send a link to the questionnaire to both groups of people in January 2024 via SMS and email, with SMS (since the target group is mostly young men and mobile phone usage is high), and email is used as a backup channel. Reminders are automatically pushed twice on the 2nd and 5th days after sending, so as to avoid frequent interruptions (more than two times may cause annoyance). Once the data collection is complete, the first step is to "collate the data". Use some statistical tools to check the basic information of the two groups, such as age, where they live, and whether they have a criminal record, to see if they are consistent. If it is found that there are many more people over the age of 25 in the comparison group than in the project group, some samples will be randomly excluded and adjusted proportionally to ensure that the two groups have similar numbers and age groups (e.g., there are 250 people left in each group at the end, and the age distribution is similar). Finally, it is necessary to compare the proportion of people in the two groups who "have been to a psychiatrist at least once". If the program group is 30% (say, 75 people/250 people) and the control group is only 15% (about 38 people/250 people), and the statistical test shows that the difference is not accidental (the p-value is small or significant), then the difference is statistically significant.
