In 2017, the City of Dallas partnered with federal and state law enforcement agencies to reduce recent increases in violent crime, especially firearm violence. The Project Safe Neighborhoods: Dallas (PSND) task force identified a target area for intervention based on the area’s high crime rate and vulnerable population.
The target area largely overlaps with the authors’ institutional service area, with the trauma center located centrally in the target area. This study examined if violent crime in the target area was reduced after PSND launched in April 2018. Additionally, we examined what effect PSND had on the rate of patients presenting to the trauma center for firearm and assaultive injuries.
Violent crime includes robbery, aggravated assault, & murder/non-negligent homicide.
Data on these index violent crimes were obtained from Dallas Police & from all other municipalities in Dallas County from January 2015 through December 2020.
Nonlinear models controlling for seasonality, temporal dependency, and geospatial dependency were fit to the data. These models were then used to derive crime rates per 100,000 population. The model predictions are plotted with 95% simultaneous confidence intervals. Notably, the first four months of data are not shown on the figures since violent crime rates from the four prior months were used as predictors in the model to account for strong temporal dependency.
Estimates of the first derivatives were used to find and determine periods of significant changes in violent crime.
|
 |
The significant difference in violent crime rates between the target and non-target regions of Dallas ended following significant increases in the non-target area between Feb. 2018 and Feb. 2019. However, violent crime never significantly changed in the target area.
Vertical black line represents the launch of Dallas’s Project Safe Neighborhoods initiative in April 2018. Shaded area represents the simultaneous 95% confidence interval (CI) of the difference in estimated violent index crime rates. The curve within the CI represents the point estimate of the difference. Emphasized regions of the CI (thick outer lines and darker shade) represent periods of time where the two regions had significantly different crime rates (January 2015–November 2016, November 2017–June 2018, and May 2020–December 2020). Estimated violent index crime rates controlled for violent index crime rates in the prior four months, spatial dependency, seasonal effects of month, and the county-wide average temporal trend.
Vertical black line represents the launch of Dallas’s Project Safe Neighborhoods initiative in April 2018. Shaded area represents the simultaneous 95% confidence interval (CI) of the derivative of the estimated violent index crime rate. The curve within the CI represents the point estimate of the derivative. The first four months of the derivative are not shown because the violent index crime rate in the prior four months was included in the model. First derivatives >0 indicate violent index crime is increasing with higher values representing faster increases; first derivatives <0 indicate violent index crime is decreasing with lower values representing faster decreases. When the 95% CI for the first derivative does not contain 0, that indicates a period of significant change; these periods are shown with darker grays and bolded lines. There were no time periods observed where the estimated violent index crime rate in target area was significantly changing. Estimated violent index crime rates controlled for violent index crime rates in the prior four months, spatial dependency, seasonal effects of month, and the county-wide average temporal trend.
Vertical black line represents the launch of Dallas’s Project Safe Neighborhoods initiative in April 2018. Shaded area represents the simultaneous 95% confidence interval (CI) of the derivative of the estimated violent index crime rate. The curve within the CI represents the point estimate of the derivative. The first four months of the derivative are not shown because the violent index crime rate in the prior four months was included in the model. First derivatives >0 indicate violent index crime is increasing with higher values representing faster increases; first derivatives <0 indicate violent index crime is decreasing with lower values representing faster decreases. When the 95% CI for the first derivative does not contain 0, that indicates a period of significant change; these periods are shown with darker grays and bolded lines. The non-target area experienced significant changes in the estimated violent index crime rate in April 2017–July 2017 (decrease), February 2018–February 2019 (increase), and September 2019–February 2020 (decrease). Estimated violent index crime rates controlled for violent index crime rates in the prior four months, spatial dependency, seasonal effects of month, and the county-wide average temporal trend.
When comparing the 12 months pre/post PSND implementation, exploratory analyses with spatial point pattern testing showed violent crime only significantly changed in seven Dallas police beats and two cities adjacent to the target areas. Only those close to the target area are shown here.
The most notable findings from this analysis are that there was a 7-fold increase in violent crime just outside the target area (Beat 651) and a nearly 50% reduction in a small portion of the target area (Beat 216).
Click here for full results in an interactive map.
To assess trauma volume, we queried the hospital’s trauma registry for monthly counts of 1) patients presenting with intentional non-self-inflicted gunshot wounds and 2) patients presenting with any intentional/assaultive injuries (e.g., aggravated assault, stab wounds, gunshot wounds).
These data were then modeled using methods analogous to those used for violent crime. These models were then used to derive injury rates per 100 trauma encounters. The model predictions are plotted with 95% simultaneous confidence intervals.
Estimates of the first derivatives were used to find and determine periods of significant changes.
|
 |
| Assaultive Injuries | Firearm Injuries |
|---|---|
| Assaultive injuries increased significantly during January 2015–May 2015 and in May 2019–February 2020. | Firearm injuries increased significantly from October 2018–December 2020. |
 |
 |
First derivatives >0 indicate the injury rate is increasing. First derivatives <0 indicate the injury rate is decreasing. When the 95% CI for the first derivatives does not contain 0, that indicates a period of significant change; these periods are shown with darker grays and bolded lines.