Authors: Edward Castillo, MPH; Alan Smith, MPH; Barbara Stepanski, MPH; Leslie Upledger Ray, MA, MPPA; Dale Cooper; Patricia Murrin, RN, MPH

 

 

Spatial Analysis of Firearm-Related Deaths in a Southern California County from 1988-1999

 

 

 

Abstract:

 

The significance of charting the geographical distribution of firearm-related deaths have been underutilized. This study examines the geographic distribution of homicides and suicides associated with firearms in San Diego, County, California. ArcView GIS and spatial analytic methods, including K-Function, Weighted K-Function, and Local Gi* Statistic, were used with county medical examiner and census data to describe the geographic patterns of firearm-related deaths. The findings of this study will be important for prevention planning efforts in addition to providing support for the need of ongoing county geographical surveillance of firearm-related deaths.

 

 

 

Introduction:

 

Firearms contribute to a large proportion of violent deaths in the United States. Although there are data that suggest that there are fundamental differences between firearm-related homicides and firearm-related suicides, geographic differences and reasons for the differences between these two modes of death have not been adequately documented in most communities. Studies that can show in detail the location and trends in the occurrences of violent deaths are needed to make clear patterns that public policy analysts can address. This study uses a geographic information system (GIS) and spatial statistics to describe the geographic characteristics and patterns of firearm-related homicides and suicides in one of the most populated counties in the southwestern United States.

 

 

 

Methodology:

 

Study Population     

From January 1, 1988 through December 31, 1999, names, addresses, and zip codes of individuals who were killed with a firearm in San Diego County were identified from county medical examiner autopsy computer records. Data on San Diego county residents who died outside of San Diego County are not available. Median income was recorded from the 1990 census. Population data were recorded from the 1990 census and estimated every year from 1991 through 1998 for every zip code in the county.

 

The place of residence was defined as the victim’s principal residence as reported by the police or coroner/medical investigator, who also determined if a death was classified as a firearm-related homicide or a firearm-related suicide. A firearm-related homicide was defined as a death that occurred due to a firearm-related that was fired by a person other than the victim. A firearm-related suicide was a death that occurred due to a firearm that was found to have been fired by the victim. For each case, the following data were abstracted as available: resident address including city and zip code, time of incident, time of death, event description, cause of death, and death classification. All autopsies were performed by the San Diego County Coroner’s Office.

 

Statistical Analysis         

Mean age, median income, and population were calculated after stratifying on death type (firearm-related homicides and firearm-related suicides) for the entire study period. A test of linear trend using Epi Info 6.0 stratified by death type was performed to determine if the proportion of deaths due to firearms has increased over the study period. 

 

Three spatial analysis techniques were performed: K-Function, Weighted K-Function, and the Local Gi* Statistic. K-Function, also called second-order analysis, is used to determine the amount of clustering in a set of data points by comparing the number of observed pairs within a specified distance to a data set placed randomly in the same study area. All combinations of pairs of point are considered in the analysis. Twelve years, 1988-1999, were separated and used in the K-Function analysis for both homicides and suicides. Another K-Function analysis was performed using all firearm-related deaths stratified by homicides and suicides. The Weighted K-Function is based on the K-Function and considers both the location and value of the points in the analysis. In this study, the Weighted K-Function analysis included all firearm-related homicides by zip code, all firearm-related suicides by zip code, and population by zip code. The Local Gi* statistic indicates the extent to which a violent death location is surrounded by a cluster of high or low values. This was used for all firearm-related homicides by zip code, all firearm-related suicides by zip code, and the population by zip code. These clusters were identified as ‘hot spots’ for the item included in the analysis. The hot spot characteristics of firearm-related suicides and homicides were further investigated by calculating the mean age, median income, and the mean estimated population for the entire study period. To control for population, another Local Gi* analysis was performed using firearm-related homicide and firearm-related suicide rates for each zip code. Maps were constructed using ArcView GIS 3.2. All graphs were made using Microsoft Excel. SPSS was used to provide crosstabs, frequencies, and t-tests.

 

 

 

Results:

 

There were 6,721 violent deaths identified from San Diego County Coroner autopsy records from January 1, 1988 through December 31, 1999, which included 2574 (38%) homicides and 4147 (62%) suicides. Of all violent deaths, 3600 (53.6%) were attributed to a firearm, which included 1599 (62%) of homicides and 2001 (48%) of suicides. There were 512 (14%) isolates could not be geo-referenced due to a missing residence address, an incorrect address, or because they were not residents of San Diego County, and thus were not included in the study. This included 343 (21%) firearm-related homicides and 169 (8%) firearm-related suicides. The final study population included 3088 firearm-related deaths, including 1256 (40.7%) firearm-related homicides and 1832 (59.3%) firearm-related suicides.

 

The mean age for firearm-related homicide victims was 30 years of age, which was much lower than the mean age for firearm-related suicide victims, 50 years of age, p = <0.001. Victims of firearm-related homicide made significantly less money in addition to living in significantly more densely populated areas than those who committed suicide (Table 1).

 

Secular Trends in the Proportion of Homicides and Suicides Using Firearms

There was no overall significant increase in the proportion of homicides using a firearm in San Diego County from 1988-1999. The proportion of homicides due to firearms ranged from 59% in 1988 to 63% in 1999, c2 = 0.007, p = 0.934.  There was a significantly decreasing linear trend for the proportion of suicides using a firearm for the same time period that ranged from 53% in 1988 to 47% in 1999, c2 = 5.461, p = 0.019 (Figure 1). 

 

K-Function

The K-function describes the apparent clustering of firearm-related homicides and suicides. There was clustering in both firearm-related homicides and firearm-related suicides compared to a random distribution in San Diego County at 5000 feet with 9 iterations. When analyzed by year, there was no secular trend in clustering for either homicides or suicides from 1988 through 1999 (data not shown).  However, homicides were more clustered than suicides as illustrated in Figure 2.

 

Weighted K-Function

Firearm-related homicides, firearm-related suicides, and populations by zip code were analyzed using the Weighted K-Function at 25,000 feet with 9 iterations (Figure 3). All three factors were clustered when compared to a random pattern. However, firearm-related homicides were clustered more than the population throughout San Diego County compared to firearm-related suicides. Firearm-related suicides were clustered similarly as the population. 

 

Local Gi*

The local Gi* statistic provides information about where clustering is taking place by identifying hot spots. At 10,000 feet, five firearm-related homicide hot spot zip codes were identified. At this range, both firearm-related suicides and the population by zip code had fewer hot spots identified, 4 and 2 respectively.  At 20,000 feet, the number of zip codes identified as hot spots increased for all three items of interest. Fifteen hot spot zip codes were identified for firearm-related homicides, 9 for firearm-related suicides, and 8 for the population. It was determined that most zip code centroids were within 20,000 feet of at least one other centroid. Therefore, 20,000 feet was considered the best distance used for this analysis. The hot spot zip codes identified at 20,000 feet for firearm-related homicides were situated around the downtown area of the city of San Diego (Figure 4). Firearm-related suicide hot spot zip codes were more dispersed, ranging throughout the northern part of the San Diego Metropolitan area and through the Del Mar area (Figure 5). Population hot spot zip codes were dispersed evenly between the downtown area and the northern metropolitan area (Figure 6).

 

Age was significantly lower in both firearm-related homicide and suicide hot spots when compared to non-hot spots. Median income was significantly lower for both firearm-related homicide and firearm-related suicide hot spots compared to non-hot spots while the estimated mean population was significantly higher in hot spots when compared to non-hot spots for both types of death (Table 2). 

 

At 20,000 feet, 13 hot spot zip codes were identified using firearm-related homicide rates (Figure 7). Of these hot spots, 11 were adjacent to each other and were situated in the downtown and surrounding areas. At the same distance, two hot spot zip codes were identified using firearm-related suicide rates (map not shown).

 

 

 

Discussion:

 

Spatial data are presented in this study of firearm-related homicides and suicides in one of the most populated counties in the southwestern United States. Firearm-related homicides clustered more than firearm-related suicides and more than population by zip code. There were more areas with high concentrations of firearm-related homicide hot spots relative to firearm-related suicides and these hot spots were restricted to a smaller area of the county than firearm-related suicides. When controlling for the zip code population by using rates in the Local Gi* analysis, only two firearm-related suicide zip code hot spots were identified. These two zip codes are situated in the southeastern region of the county. These areas are sparsely populated zip codes (1,247 and 2,931 people) and are a far distance from each other, which are possible reasons for their significance in the analysis. However, 13 hot spot zip codes were identified when using firearm-related homicide rates. Eleven of these zip codes were primarily in the downtown and surrounding areas. These results suggest that firearm-related suicides are a function of the population distribution while firearm-related homicides occur in areas with lower socioeconomic status.

 

This study allowed for a different view of a common public health problem by evaluating firearm-related deaths in San Diego County using spatial analytic methods, thus contrasting standard analytic methods. Although the methods used in this study are relatively new to the field of epidemiology, they are sure to become more widely utilized in the future. These data support the need for ongoing county geographic surveillance of firearm-related homicides and suicides to further substantiate the role of studies that would be beneficial for resource allocation and for community-based intervention planning.

 

 

Edward M. Castillo, MPH

Biostatistician

County of San Diego

Health and Human Services Agency

Emergency Medical Services

(619) 285-6429

ecast2he@co.san-diego.ca.us

 

 

Table 1. Characteristics of firearm-related homicides and suicides, San Diego County, 1988 – 1999. 

 

 

 

 

Firearm-Related Homicides

(N=1256)

Firearm-Related Suicides

(N=1832)

 

 

P-Value

 

 

 

 

Age + SD

30 + 13.2

50 + 21

< 0.001

   Range

0 - 85

11 - 97

 

 

 

 

 

Median Income + SD

30,465 + 10,723

35,932 + 9,952

< 0.001

 

 

 

 

Mean Population + SD

48,153 + 13,779

42,597 + 15,135

< 0.001

Note: Mean population is the mean estimated population from 1990 through 1998.

Source: County of San Diego, Health and Human Services Agency, Division of Emergency Medical Services, San Diego County Medical Examiner's Data: 1988-1999.

 

 

 

Table 2. Characteristics of zip code hot spots and non-hot spots based on the local Gi* statistic, San Diego County, 1988-1999. 

 

 

 

 

Firearm-Related Homicides

Firearm-Related Suicides

 

Clustered (N=591)

Not Clustered (N=665)

Clustered (N=280)

Not Clustered (N=1552)

 

 

 

 

 

Age + SD

29 + 11

32 + 15***

47 + 20

50 + 21*

   Range

9 - 85

< 1 - 85

13 - 95

11 - 97

 

 

 

 

 

Median

   Income + SD

 

24,189 + 8,241

 

36,487 + 9,282***

 

34,377 + 9,801

 

36,235 + 9956*

 

 

 

 

 

Mean Population

   + SD

51,407 + 11,910

45,032 + 14,707***

43,976 + 9,380

42,328 + 16,006

*P < 0.05                **P < 0.01         ***P < 0.001

Mean population is the mean estimated population from 1990 through 1998.

Source: County of San Diego, Health and Human Services Agency, Division of Emergency Medical Services, San Diego County Medical Examiner's Data: 1988-1999.