PLEASE SCROLL DOWN FOR ARTICLE

(1)

PLEASE SCROLL DOWN FOR ARTICLE

This article was downloaded by: [Leiden University Library]

On: 26 January 2009

Access details: Access Details: [subscription number 907217933]

Publisher Taylor & Francis

Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Injury Control and Safety Promotion

Publication details, including instructions for authors and subscription information:

http://www.informaworld.com/smpp/title~content=t713734346

Non-fatal violence-related injuries in Kingston, Jamaica: a preventable drain on resources

Namvar Zohoori; Elizabeth Ward; Georgiana Gordon; Rainford Wilks; Deanna Ashley; Terrence Forrester

Online Publication Date: 01 December 2002

To cite this Article Zohoori, Namvar, Ward, Elizabeth, Gordon, Georgiana, Wilks, Rainford, Ashley, Deanna and Forrester, Terrence(2002)'Non-fatal violence-related injuries in Kingston, Jamaica: a preventable drain on resources',International Journal of Injury Control and Safety Promotion,9:4,255 — 262

To link to this Article: DOI: 10.1076/icsp.9.4.255.13675 URL: http://dx.doi.org/10.1076/icsp.9.4.255.13675

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

(2)

Abstract

Using data for a one-year period from the Kingston Public Hospital (KPH) in Jamaica, we describe patterns of non-fatal violence-related injuries, and carry out simulation analysis to estimate rates of hospital admission under various injury reduction scenarios, and the potential savings that can be realized by reducing violent crimes. In this period there were 6107 registered violence-related visits to the KPH representing 11.5% of all recorded visits. Of these 16.6% (1001) were admitted. The most common methods of inflicting injury was by stabbing (52.1%), blunt injuries (37.9%) and gunshot wounds (7.3%). Multivariate analyses indicated that gunshot injuries, stab injuries, being male between the ages of 15 and 44 years, receiving the injury in November or December, and being injured by a stranger or unknown assailant, were significant correlates of a higher probability of admission. Simulation analysis with various injury reduction scenarios indicated decreases in the probability of admission ranging from 12% to 44%, with estimated savings of up to 31% of the annual supplies budget of KPH.

Keywords: Injuries; violence; resources; simulation;

Jamaica.

Introduction

Intentional and non-intentional injuries are a major public health problem¹and cost societies a great deal in immediate monetary costs, in individual and societal opportunity costs

of lost lives and productivity, and also in the pain and suf- fering of victims and their families. An understanding of the patterns, causes and costs of violence-related injuries in any given community is important and a recognized factor for the development of programs and strategies for reducing the incidence of these injuries.^2–8

In Jamaica, violence-related injuries (VRI’s) are a major cause of morbidity and mortality.⁹ The number of violent crimes and deaths have been on the rise in recent years. According to the Jamaica Constabulary Communica- tions Network (personal communication), during 2001 there were 1139 homicides in Jamaica, of which 659 (58%) were in the capital city area of Kingston and St. Andrew. This translates into a national homicide rate of about 51 per 100,000 which is 112% higher than Jamaica’s 1991 homicide rate of 24 deaths per 100,000. By comparison, between 1991 and 1999, homicide rates declined from 10.5 to 6.2 per 100,000 in the USA, and from 71.4 to 59 per 100,000 in Columbia.¹⁰

While homicides are particularly difficult for the depen- dents and family members of the victims, non-fatal VRI’s can be even more costly for the victims, their families and the society. Emotional and physical damage can lead to significant loss of personal productivity and drain family resources.

At the societal level, the immediate costs associated with hospitalisation and medical care can be quite significant.¹¹ However, while data on mortality are more easily available, to date there has been a lack of reliable information on patterns and causes of non-fatal VRI’s in Jamaica.

Accepted 19 August 2002.

Correspondence: Dr. Namvar Zohoori, Epidemiology Research Unit, Tropical Medicine Research Institute, University of the West Indies, Mona, Kingston 7, Jamaica, W.I. Tel.: +1 876-977-6151, Fax: +1 876-927-2984, E-mail: NZohoori@UWIMona.edu.jm

ORIGINAL PAPER

Non-fatal violence-related injuries in Kingston, Jamaica: a preventable drain on resources

Namvar Zohoori, MBBS, MPH, PhD^1,4, Elizabeth Ward, MBBS², Georgiana Gordon², Rainford Wilks, DM¹, Deanna Ashley, DM²and Terrence Forrester, DM³

1Epidemiology Research Unit, Tropical Medicine Research Institute, University of the West Indies, Mona, Jamaica, West Indies; ²Division of Health Promotion and Protection, Ministry of Health, Jamaica, West Indies; ³Tropical Medicine Research Institute, University of the West Indies, Mona, Jamaica, West Indies; ⁴Carolina Population Center and Department of

Nutrition, Schools of Public Health and Medicine, University of North Carolina at Chapel Hill, Chapel Hill, NC, USA

2002, Vol. 9, No. 4, pp. 255–262 © Swets & Zeitlinger

Downloaded By: [Leiden University Library] At: 14:29 26 January 2009

(3)

256 N. Zohoori et al.

With this in mind, in 1998, the Division of Health Pro- motion and Protection of the Ministry of Health in Jamaica, with technical assistance from the Division of Violence Pre- vention of the National Center for Injury Prevention and Control of the U.S. Centers for Disease Control and Preven- tion, implemented a Violence-Related Injury Surveillance System (VRISS) in a number of hospitals in Jamaica. Details of the design and implementation of this system have been described elsewhere.¹²In this paper, we used data for a one- year period from Kingston Public Hospital (KPH), the largest tertiary care hospital in Jamaica, to describe the patterns of non-fatal VRI’s. We also carried out simulation analysis to estimate rates of hospital admission under various injury reduction scenarios, and the potential savings that can be realized by reducing violent crimes and the hospital care associated with them.

Data and methods

Data

The data came from the VRISS at the KPH. Data used were for the one-year period, August 1, 1998 to July 31, 1999.

During this period, there were 6107 registered violence- related visits to the Accident and Emergency (A&E) department of the KPH. Of this number, 62 had missing age information and another 12 had missing information for the method of injury. These were excluded from analysis. The final analysis file for regressions and simulations, therefore, consisted of 6033 subjects (99%) for whom there was complete information.

Data collection was done by trained registration clerks in the A&E department, and entered into the hospital’s com- puterized UNIX-based Patient Admission System (PAS). The VRISS form was based on a subset of questions from the International Classification of External Causes of Injury system.¹³ Questions included on the form consist of age, gender, circumstance of injury, method of injury, victim–perpetrator relationship, as well as admission status (whether admitted to the hospital or treated and discharged). For the purposes of this system, cases were those persons register- ing at the A&E department with a non-fatal VRI resulting from threatened or actual use of physical force with the intent to cause harm. Cases did not include persons with acciden- tal injuries (such as motor vehicle accidents) or those whose injuries were intentionally self-inflicted (such as suicide attempts). The development of the form and the appropri- ateness of the questions for the Jamaican situation were guided by focus group studies conducted with victims and their families prior to the implementation of the system.¹⁴

Analysis

All analyses were performed using STATA Software.¹⁵Simu- lation results are from predicted values of the dependent variable (admission status) following maximum likelihood

multiple regression models and the appropriate manipulation of relevant independent variables (see Results for more details).

Cost estimates

To the best of our knowledge, there are no published data on estimated costs of admissions due to VRI’s at the KPH or other hospitals in Jamaica. For the analyses in this paper, the best available estimates were unpublished data.¹⁶These estimates were based on a 1997 study of violence-related emergency admissions to the Department of General Survey. That study estimated mean costs of JA$ 7971 for admissions without use of an operating theatre, and JA$ 13,320 for those needing use of an operating theatre. These estimates included the cost of supplies and procedures, but did not include

‘salaries. . . . sundries such as bed linen, utilities including rent, equipment cost and . . . blood investigations.’ For our purposes, we used these two estimates and converted them to year 2001 dollar equivalents using national inflation factors of 7.9, 6.8, 8.2 and 8.7 respectively for the years 1998–2001.¹⁷

Results

Descriptive

During the period under study, there were 6107 registered violence-related visits to the A&E department of the KPH.

This represents 11.5% of all recorded visits (53,139) to the A&E department during this period. Table 1 provides descriptive statistics on the sample used in the analyses.

There were 6033 subjects in the analysis file, of which 62%

were males. About half the victims were between the ages of 25 and 44 years. Of these 6033 subjects, 16.6% (1001 subjects) were admitted to the hospital for further treatment, and the remaining 83.4% were discharged following treatment in the A&E department. The majority of the victims (57%) received their injuries from acquaintances, while relatives and strangers perpetrated 19% and 24% of the injuries, respectively. Stabbing (with a sharp instrument) was the commonest method of inflicting injury (52.1% of all injuries). Use of blunt force (such as boxing, kicking, pushing and using a blunt object) accounted for another 37.9% of the injuries, while gunshot wounds were respon- sible for 7.3% of the injuries. Other injuries (such as burns and use of chemicals) accounted for another 2.6%. Injuries were non-uniformly distributed throughout the year, with a significantly (p < 0.001) greater number of A&E department visits in the later months of the year.

Figure 1 shows the geographic distribution of reported cases of non-fatal VRI’s at the KPH during a four-month period. The KPH is the largest tertiary care hospital on the island. Located in a lower socio-economic, crime-ridden area of Kingston, the capital city of Jamaica, KPH provides services for a catchment population of 1.3 million persons living

(4)

in the Kingston and St. Andrew areas (the area shown in the map).

Multivariate regression

Using admission to the hospital as an outcome variable, and maximum-likelihood logistic regression models, we devel- oped a multivariate model for predicting the probability of admission based on age, gender, month of admission, method of injury and perpetrator–victim relationship as explanatory variables. Preliminary analysis indicated the groupings shown in Table 1 to provide the best fitting model for subse- quent multivariate analysis. These analyses also indicated that there was significant interaction between age and gender, and that another variable, ‘circumstance of injury’ (such as

robbery, sexual assault, gang violence, etc.), was not significant in the model. The latter was therefore not included in the final model.

Resulting multivariate odds ratios for hospital admission are presented in Table 2 for the whole model, and in Figure 2 for the interaction between age and gender. Compared to use of blunt force, gunshots were significantly associated with the highest risk (eight-fold) of hospital admission (p<

0.0005), and stab wounds carried an almost two-fold risk of hospital admission (p < 0.0005). After controlling for all other factors (including the method of injury: blunt, gunshot, stabbing), being injured by a spouse or relative carried a risk of hospitalisation about 28% lower (p< 0.008) than when the injury was inflicted by friends or acquaintances, while an injury from a stranger or unknown assailant carried a 32%

Table 1. Descriptive characteristics of non-fatal VRI’s presenting to the A&E department of KPH.

Number Percent of Percent

Characteristic (total = 6033) total admitted

Gender

Male 3737 61.9 21.9

Female 2296 38.1 8.0

Age (years)

<15 197 3.3 7.6

15–24 2166 35.9 16.8

25–44 3098 51.3 17.5

45+ 572 9.5 14.0

Admission status

Discharged after treatment in A&E Dept. 5032 83.4 – Admitted to hospital for further treatment 1001 16.6 – Month of injury

August 595 9.9 16.8

September 467 7.7 15.9

October 601 10.0 15.3

November 601 10.0 20.5

December 548 9.1 20.8

January 468 7.8 14.3

February 388 6.4 14.7

March 538 8.9 14.3

April 424 7.0 14.9

May 428 7.1 14.3

June 511 8.5 18.0

July 464 7.7 17.5

Relation of perpetrator

Spouse/Relative 1151 19.1 7.9

Friend/Acquaintance 3441 57.0 14.4

Stranger/Unknown 1441 23.9 28.7

Method of injury

Blunt 2289 37.9 8.2

Stab 3146 52.2 17.4

Gunshot 443 7.3 54.4

Others 155 2.6 17.4

(5)

higher risk of hospitalisation (p < 0.002). Being injured in November or December was associated with a 60% higher risk of hospitalisation compared to January (p< 0.008 and <

0.01, respectively), again controlling for all other factors in the model. None of the other months of the year were significantly different from January, or from each other, with respect to the probability of admission.

Simulation analysis

Using the estimated coefficients from the above maximum- likelihood logistic regression model, and by varying the values for selected explanatory variables in the model, we predicted the likelihood of admission to hospital for various injury reduction scenarios. Results are shown in Table 3 overall for the whole 12-month period, and in Figure 3 by month of injury. Six different scenarios were examined in the simulation analysis: (1) if all stab and gunshot injuries were blunt injuries instead (note that the model cannot test scenarios where injuries are replaced by no injury at all) – scenario A in Table 3 and Figure 3; (2) if half of stab injuries and half of gunshot injuries were blunt injuries instead – scenario B; (3) if all stab injuries were blunt injuries instead – scenario C; (4) if all gunshot injuries were blunt injuries instead – scenario D; (5) if half of stab injuries were blunt

injuries instead – scenario E; and (6) if half of gunshot injuries were blunt injuries instead – scenario F.

The overall predicted value for the likelihood of admission to hospital from the model was 16.6%, very close to the actual admission rate of 16.9%. The highest reduction in the probability of admission was, not surprisingly, achieved under scenario A (if all gunshot and all stab injuries were blunt injuries instead), which gave a probability of admission of 9.6% (a 43% reduction in the probability of hospitalisation). Short of this complete elimination of gunshot and stab injuries, the next best result was obtained equally by scenar- ios B (if half of stab injuries and half of gunshot injuries were blunt injuries instead) and C (if all stab injuries were blunt injuries instead): each of these two scenarios resulted in a 25% reduction in the probability of hospitalisation. A mod- erate reduction of about 20% in the probability of hospitali- sation was achieved by scenario D (if all gunshot injuries were blunt injuries instead). Scenarios E (if half of stab injuries were blunt injuries instead) and F (if half of gunshot injuries were blunt injuries instead) resulted in more modest reductions in the probability of hospital admission of about 15% and 12% respectively.

Table 3 also shows the estimated annual savings under the various scenarios. Estimated savings ranged from 1.34–2.24 million Jamaican dollars for scenario F, to 4.68–7.82 million Figure 1. Distribution of non-fatal violence-related injuries as recorded at the KPH, September to December 1999, Kingston and St. Andrew, Jamaica.

(6)

Jamaican dollars for scenario A. Scenarios B and C each produced estimated savings of about 2.75–4.60 million Jamaican dollars, while more modest savings were estimated for scenarios D and E.

Discussion

In this paper we have presented data for a one year period (August 1, 1998 to July 31, 1999), from a VRISS at the KPH in Kingston, Jamaica. During this period there were 6107 registered violence-related visits to the A&E department of the KPH, representing 11.5% of all recorded visits (53,139) to the A&E department during this period.

Methodological issues related to the Surveillance System itself are reported elsewhere. This paper looked at the various causes of injury and their contribution to the likelihood of admission to hospital in a multivariate model. Further, we have carried out simulation analyses of various injury reduction scenarios to examine the cost savings that would result from reductions in the probability of hospital admission under these various scenarios.

The data indicated that of the 6033 individuals presenting to the A&E Department because of non-fatal VRI’s, 1001 were admitted to the hospital for further treatment, giving an overall actual probability of admission of about 16.9%. In multivariate analyses, significant contributors to increasing Table 2. Results of multivariate logistic regression analysis for the odds ratio for

hospitalisation.

Variable 95% Confidence Interval

(referent value

in parentheses) Odds Ratio Lower Limit Upper Limit p-value

Gender (Male)

Female 0.23 0.07 0.73 0.012

Age (<15)

15–24 0.48 0.23 0.99 0.048

25–44 0.59 0.29 1.21 0.152

45+ 0.46 0.18 1.20 0.114

Month of injury (January)

February 1.08 0.72 1.63 0.697

March 1.26 0.87 1.83 0.226

April 1.27 0.85 1.88 0.239

May 1.08 0.73 1.61 0.690

June 1.34 0.93 1.94 0.117

July 1.38 0.95 2.01 0.090

August 1.27 0.89 1.81 0.191

September 1.16 0.79 1.69 0.456

October 1.25 0.87 1.79 0.224

November 1.60 1.13 2.26 0.008

December 1.60 1.12 2.27 0.010

Relation of perpetrator (Friend/Acquaintance)

Spouse/Relative 0.72 0.56 0.92 0.008

Stranger/Unknown 1.32 1.11 1.57 0.002

Method of injury (Blunt)

Stab 2.00 1.66 2.39 0.000

Gunshot 8.05 6.18 10.47 0.000

Others 2.24 1.42 3.51 0.000

Gender ¥ Age Interaction*

15–24 11.33 3.50 36.67 0.000

25–44 9.28 2.90 29.70 0.000

45+ 9.07 2.37 34.72 0.001

* Chi-square statistic for the log-likelihood ratio test for the significance of the age-gender inter- action term in the model is 8.72 with 3 degrees of freedom (p< 0.05).

(7)

the probability of admission were gunshot injuries and stab injuries (compared to blunt injuries), being a male between the ages of 15 and 44 years, receiving the injury in the months of November or December, and being injured by a stranger or unknown assailant (compared to non-relative friends or acquaintances). In addition, being injured by a spouse or relative reduced the likelihood of admission, after controlling for age, gender or method of injury.

Given the multivariate nature of the above results, certain interesting questions arise regarding the likelihood of admission to hospital following a VRI. For example, while the higher probability of admission following both gunshot and stab injuries (compared to blunt injuries) is understandable, factors such as male gender, the relationship of the perpe-

trator, or the months of November and December indepen- dently altered the probability of admission after controlling for the nature of the injury. One explanation for this might lie in the severity of injury (a factor not explicitly included in the data collection system), so that gunshot or stab wounds received by females, or inflicted by spouses and relatives, may have been less severe than those received by males or inflicted by strangers, thus not requiring hospital admission.

Similarly, the higher probability of admission in the months of November or December may have been due to other similar factors not captured in these data.

Before discussing the simulation analysis results, a few points should be noted. In this data set, it was not possible to simulate a situation where a certain type or level of injury was replaced by no injury at all, since there was no uninjured control group, and the maximum-likelihood model devel- oped here used blunt injuries as the baseline or comparison group. In these analyses, therefore, simulated reductions in gunshots or stab wound injuries needed to be replaced by a similar level in blunt injuries. Further, the scenarios chosen for these simulations, though quite plausible, were arbi- trarily chosen: one could simulate an infinite number of situations based on various combinations of reductions in gunshot or stab wound injuries (or any other factor in the model). However, for manageability, we chose a few discrete combinations of factors here. These simulations may be interpreted as answering the question ‘What would be the overall probability of admission to hospital given the hypo- thetical changes in levels of stab or gunshot injuries, holding Figure 2. Odds ratios for admission, by age and gender from mul-

tivariate model.

Table 3. Overall probabilities of, and percent reductions in, hospitalisation under various simulated injury reduction scenarios, and the estimated savings associated with them.

Percent reduction in

Predicted probability of Number of

probability of hospitalisation patients that Annual savings – Annual savings – hospitalisation (compared to would not if no theatre with theatre

Simulated Scenario (percent) actual data) be admitted (millions JA$*) (millions JA$*)

A: If all stab and gunshot injuries 9.6 43.3 433 4.68 7.82

were blunt injuries instead

B: If half of stab injuries and half 12.6 25.5 255 2.75 4.60

of gunshot injuries were blunt injuries instead.

C: If all stab injuries were blunt 12.6 25.5 255 2.75 4.60

injuries instead

D: If all gunshot injuries were 13.6 19.7 197 2.13 3.56

blunt injuries instead

E: If half of stab injuries were 14.4 14.8 148 1.60 2.67

blunt injuries instead

F: If half of gunshot injuries 14.8 12.4 124 1.34 2.24

were blunt injuries instead

* Year 2001 dollar equivalents.

(8)

22

20

18

16

14

12

10

8

6

Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul

G F

E D

C B

A

Predicted Probaility of Admission (%)

Figure 3. Predicted probabilities of admission under various injury reduction scenarios, by month of injury. A: If all stab and gunshot injuries were blunt injuries instead. B: If half of stab injuries and half of gunshot injuries were blunt injuries instead. C: If all stab injuries were blunt injuries instead. D: If all gunshot injuries were blunt injuries instead. E: If half of stab injuries were blunt injuries instead. F: If half of gunshot injuries were blunt injuries instead. G: Actual data (no simulation).

constant all other factors in the model?’ Also, our cost esti- mates were based on unpublished data from a study carried out in 1997 at the KPH. While we inflated these costs to 2001 dollars, and while these estimates do have the advantage of being based on data at the same hospital as our study, they are likely to produce underestimates of the actual costs, due mainly to two facts: they were based only on admissions to general surgical services, and they did not include personnel salaries and some other costs such as investigations and equipment costs.

While scenario A (replacement of all gunshot and stab injuries by blunt injuries) may be arguably regarded as being impractical (though theoretically possible), it serves as a useful indicator of the overall estimated cost of these injuries at the KPH during this one-year period.

The estimated total savings of 4.7–7.8 million Jamaican dollars (Table 3) represents about 19–31 percent of the 2001 annual medical supplies budget of KPH, which was 24.9 million Jamaican dollars (Ministry of Health, personal communication).

The next two scenarios (B and C) each provide exactly the same level of reduction in the probability of hospitalisation and, therefore, cost savings. That is, reducing stab injuries and gunshot injuries each by fifty percent (scenario B), or alternatively replacing all stab injuries only by blunt injuries (scenario C), both resulted in a 25% reduction in the probability of hospitalisation following presentation to the A&E department of the KPH. Again, while scenario C may be con- sidered impractical, scenario B should certainly be consid- ered as practical and achievable with proper educational campaigns and other policy measures. It is noteworthy that a

comparison of scenarios B and C indicates that a 50% reduction in gunshot injuries has about the same effect on the probability of admission as a 50% reduction in stab injuries (as also seen in comparing scenarios E and F), even though numerically stab injuries were some seven times more fre- quent than gunshot injuries. This is due to the more serious nature of gunshot injuries compared to stab injuries – as seen in the multivariate analyses – but the implication is that by concentrating campaigns on both of these types of injuries – gunshots, which are less prevalent but more serious, and stab injuries, which are more prevalent but less serious – one should be able to achieve impressive reductions in levels of hospitalisation. It should also be noted that these two types of injuries were prevalent in different situations: while gunshots were predominantly (over 80%) used by assailants who were strangers or unknown to the victim and in gang or mob confrontations and robberies, stab injuries were most prevalent (about 80%) among relatives and acquaintances, and in fights (data not shown). Again the implication here is for the targeting of different populations and situations for educational and policy measures.

The data used in these analyses record only information on non-fatal injuries, and therefore give an underestimation of the cost and the savings that can be realized by reducing actual prevalence of stab and gunshot injuries in the community. In fact, the relatively low number of gunshot injuries in the data is most likely due to the fact that more gunshot injuries are fatal and therefore do not present to the A&E department (2001 data from Jamaica indicate that 69% of homicides were due to gunshots as opposed to 18% due to stabbing – Jamaica Constabulary Communications Network,

(9)

personal communication) . Therefore, any educational and other policy measures that may be instituted are likely to have a greater benefit than is implied by these analyses and calculations.

In conclusion, we have shown that non-fatal VRI’s com- prise a significant proportion of admissions to KPH. There are predictable correlates of the likelihood of admission, and educational drives aimed at reducing specific risk factors can lead to large savings of limited resources.

Acknowledgements

The authors would like to gratefully acknowledge the assistance of Nadine Jones (Planning Institute of Jamaica), Matthew Taylor and Trevor MacCartney (Department of Surgery, Kingston Public Hospital), Jasneth Mullings (Tropical Medicine Research Institute), and personnel of the Medical Records and the Accidents and Emergency Departments (Kingston Public Hospital). Technical assistance for the VRISS was provided by the U.S. Centers for Disease Control and Prevention and the Tropical Medicine Research Institute of the University of the West Indies.

Funding was provided through the Ministry of Health, Jamaica.

References

1 Rosenberg ML, O’Carroll PW, Powell KE. Let’s be clear. Violence is a public health problem. JAMA.

1992;267(22):3071–3072.

2 Twemlow SW, Sacco FC. Peacekeeping and peacemaking:

the conceptual foundations of a plan to reduce violence and improve the quality of life in a midsized community in Jamaica. Psychiatry. 1996;59(2):156–174.

3 Fernald LC, Ani C, Gardner JM. Aggressive behaviour in children and adolescents. Part I: A review of the effects of child and family characteristics. West Indian Med J.

1997;46(4):100–103.

4 Mazza JJ, Reynolds WM. Exposure to violence in young inner-city adolescents: relationships with suicidal ideation, depression, and PTSD symptomatology. J Abnorm Child Psychol. 1999;27(3):203–213.

5 Soyibo K, Lee MG. Domestic and school violence among high school students in Jamaica. West Indian Med J.

2000;49(3):232–236.

6 Rosenthal BS. Exposure to community violence in adoles- cence: trauma symptoms. Adolescence. 2000;35(138):271–

284.

7 Rosenthal BS, Wilson WC. Relationship between exposure to community violence and psychological distress: linear or curvilinear? Psychol Rep. 2001;88(3 Pt 1):635–640.

8 Rosenthal BS, Hutton EM. Exposure to community violence and trauma symptoms in late adolescence: comparison of a college sample and a noncollege community sample.

Psychol Rep. 2001;88(2):367–374

9 McDonald A, Duncan ND, Mitchell DI, Fletcher PR. Trauma aetiology and cost in the Accident and Emergency Unit of the University Hospital of the West Indies. West Indian Med J. 1999;48(3):141–142.

10 PAHO/WHO Division of Disease Control and Prevention.

Statistics on Homicides, Suicides, Accidents, & Attitudes towards Violence. Available from: http://www.paho.org/

accessed in May 2002.

11 Mansingh A, Rampal P. The nature of interpersonal violence in Jamaica and its strain on the national health service. West Indian Med J. 1993;42:53–56.

12 Ward E, Durant T, Thompson M, Gordon G, Mitchell W, Ashley D, and the VRISS Working Group. Implementing a hospital-based violence related injury surveillance system:

A background to the Jamaica Experience. Injury Control and Safety Promotion. 2002;9(4):241–247.

13 WHO-Working Group for Injury Surveillance Methodology Development and Its Technical Group. International Classi- fication of External Causes of Injuries (ICECI). Amsterdam:

Consumer Safety Institute, June 1999.

14 Mullings J. A qualitative assessment of accidents and vio- lence in Kingston, Jamaica. Technical Report prepared for and submitted to the Ministry of Health (Jamaica), August 19, 1998.

15 StataCorp. Stata Statistical Software: Release 6.0. College Station, TX: Stata Corporation. 1999.

16 Taylor M, McCartney T. Intentional violence and the Kingston Public Hospital at what cost? Department of Surgery, KPH. Kingston, Jamaica. Unpublished manuscript.

17 Planning Institute of Jamaica. Economic and Social Survey, Jamaica. PIOJ, Kingston, Jamaica, 2001.