Table 1 Study descriptions

1. Wyatt [23] (1980), PNG, Case series Retrospective

121 RTI-related forensic autopsies PMGH, 1975-78.

Road user type, injury distribution, day and time, BAC.

Drivers 21%, passengers 38%, pedestrians 41%, males 85%, males 20-39 years 66%. Injuries: brain 25%, spinal cord 6%, multiple sites 54%. Driver fatality per 10,000 vehicles: cars (4.6), motorcycle (21.3). Pedestrian 20% aged <10 years, passengers 25% fell/jumped from moving vehicle. Alcohol associated with 76% of those aged >10 years. BAC > 80 mg/dl observed in 1/3 drivers and 69% male pedestrians, BAC > 120 mg/dl 62% fatalities. 46% of fatalities occurred on Friday and Saturday

Method to identify eligible cases not stated. Measurement bias reported for incomplete or delayed blood alcohol testing.

2. Sinha et al [24]. (1981), PNG, Case series Retrospective

305 trauma-related forensic autopsies, (171 RTI-related) PMGH, 1976-80.

Road user type, age, sex, injury distribution, day and time, risk factors, BAC.

Drivers 17%, passengers 46%, pedestrians 36%, males 86%, 82% < 35 years. Injuries: head 65%, chest 51%, abdomen 37%, spine 18%, fracture skull 57%, brain injury 53%, and fracture ribs 39%. Chest injuries higher in drivers (65%), head injuries higher in passengers (71%). Risk factors: driver lost control 60%, passengers fell off a truck 60%. Alcohol associated with 85% of drivers and 90% of adult pedestrians. 53% BAC >80 mg/dl. 57% occurred over weekend, 44% occurred between 6 pm and 6 am. 66% died at the crash site or soon after.

Post mortem records supplemented with hospital admission notes. Incomplete records excluded, measurement bias reported for incomplete or delayed blood alcohol testing. Comparison between RTI and non-RTI related deaths limited to those with spleen injury.

3. Palmer [25] (1982), PNG, Case series Retrospective

353 forensic autopsies (97 RTI-related) Goroka hospital, 1978-82.

Injury cause, distribution

Dead on arrival to hospital 81% (n = 79), head injuries 71% (n = 56), most deaths followed ejection or jumping from a moving vehicle. Of those who died during hospital admission, 61% had head injuries.

Method to identify eligible cases not stated

4. Gee et al [26]. (1982), PNG, Case series Retrospective

36 trauma-related spinal cord injury admissions PMGH, Lae, and Madang hospitals,1978-81.

Age, sex.

RTI-related spinal cord injuries 34% (n = 12). For all trauma related spinal cord injuries: 88% Male, mean age 26 years

Incomplete information in case records reported. Selection bias not reported.

5. Lourie et al [27]. (1983), PNG, Case series Prospective

209 RTI-related admissions A&E PMGH, 1982-83.

Road user and vehicle type, passenger seating, seatbelt use

Drivers 15%, passengers 72%, pedestrians 13%, males 82%, those aged > 16 years 87%, 3 dead on arrival, 22% admitted, 40% drivers/front seat occupants injured, 13% front-seat occupants wore seatbelt. 28% passengers on the back of utility trucks, and all injured. Vehicles: cars 33%, utility trucks 27%. 1/3 vehicles no seat belts fitted.

Presence of investigator blinding of outcome not stated. Limitations of study such as selection and recall bias, well described, Measurement bias for alcohol association reported. While risk factors for RTI identified, study design precluded quantification of risk.

6. Sinha et al [28]. (1989), PNG, Case series Retrospective

363 RTI-related forensic autopsies PMGH, 1976-85. Total trauma-related autopsies 608, 60% RTI related.

Road user type, age, sex, BAC, injury distribution, day and time, if death occurred before arrival to hospital.

Drivers18%, passengers 45%, pedestrians 34%, males 83%, 15-44 years70% Injuries to the head 43%, chest 30%, abdomen 8%, spine and pelvis 10%, fracture skull 15%, brain injury 17%, intracranial bleed 10%, fracture ribs14%. Raised BAC (> 80 mg/dl) in 48% of drivers tested (n = 16/33), and 67% of pedestrians (n = 20/30) 54% occurred over the weekend, 40% 6 pm to 6 am 2/3 died at the crash site or soon after

Reports on five year extension to earlier study by Sinha et al [28].; Measurement bias reported for incomplete or delayed blood alcohol testing.

7. Cooke et al [29]. (1992), PNG, Case series Retrospective

573 RTI-related forensic autopsies PMGH, 1962-89 Total autopsies 1279, 45% RTI-related

Road user type, sex.

75% males (n = 432). Total males: drivers 12%, passengers 35%, pedestrians 28% Annual crude data shows increasing RTI-related fatality trends over 4 decades

Limited reporting on methodology. Bias due to incomplete datasets, identified.

8. Posanau [30] (1994), PNG, Case series Prospective

188 RTI-related admissions A&E PMGH, 1990. Total road traffic crashes 104.

Road user type, age, sex, time and day of crash, association with alcohol.

74% males (n = 136), 48% aged 18-29 years. 49% of road traffic crashes associated with alcohol. Most crashes occur in first part of day, but for alcohol related crashes, more common at night and early morning. Also alcohol related crashes more common on weekends. Road users – 25% drivers, 61% passengers. For those admitted to hospital, 43% drivers, 54% of passengers and 31% of pedestrians.

Data source; administered questionnaire, clinical assessment, and autopsy reports. Validation of definition of alcohol-related road traffic crash and assessment of alcohol intoxication not stated. Presence of investigator blinding of outcome not stated.

9. Maharaj [31] (1996), Fiji, Case series Retrospective

140 spinal cord injury admissions Fiji medical rehabilitation unit, 1985-94.

Sex, distribution of injuries.

Trauma-related spinal cord injuries 54% (n = 75), 25% RTI related (n = 19) Fijian males comprised 87% of trauma-related spinal cord injuries.

Included both trauma and non-trauma-related spinal cord injuries. Results reported as event counts and proportions.

10. Mathew et al [32]. (1996), PNG, Case series Retrospective

454 trauma-related admissions Mendi hospital, 1993.

Injury distribution, length of hospital stay.

RTI related trauma admissions 14% (n = 63) Injuries to head 37% with 2 deaths, chest 14%, abdomen 6%, fractures and dislocations 29%, Average length of stay for RTI, 9.6 days, median 5 days, range 1-51 days, same for assaults

Methods of data collection not reported. Exclusion of 15% of patient charts due to incomplete data.

11. Liko et al [33]. (1996), PNG, Case series Prospective (x2) Retrospective (x1)

274 head injury admissions PMGH and Goroka hospital, 1984-93.

Injury distribution.

RTI related head injuries 49% (n = 134). Distribution of admissions, 55% to PMGH and 45% to Goroka hospital Head injury case fatality 21%

Method to identify eligible cases not reported. Selection bias not reported.

12. Watters et al [34]. (1996), PNG, Case series (x4) Prospective (x2) Retrospective (x1) Prospective and Retrospective (x1)

667 trauma-related admissions A&E, PMGH, (35 days over 3 months); 154 trauma admissions PMGH general surgical unit (over 1 year); 88 trauma-related deaths prior to admission to PMGH(8 months); 50 deaths during admission (over 18 months) PMGH, 1992-93

Injury cause, distribution, length of hospital stay.

RTI A&E admissions 6%, hospital admissions 30% Mean stay for RTI, 26 days, twice as long compared to assault, accidents and sports Deaths prior to admission RTI 30% (n = 26), assault 45% (n = 40). Total head injuries 37% (n = 33), RTI related n = 24.

Four studies over a period of time. Recruitment dates for trauma-related admissions randomised. Method to identify eligible cases as well as selection bias not reported.

13. Ravia [35] (1999), Yap, Case series Retrospective

100 injury-related admissions A&E Yap State Hospital,1996-98

Road user type, association with alcohol, residence

Drivers 12, passengers 2, pedestrians 0, RTI-related injury admissions 14% (n = 14). Five times increase in RTI among drivers from n = 2 to n = 10. 88% of all injuries associated with alcohol

Methodology section limited. Biases not reported.

14. Ponifasio [36] (2001), PNG, Case series Prospective

213 abdominal trauma related admissions to PMGH (Feb 1992- Jun 1994), and 98 to Lae hospital,1996

Distribution of injuries

RTI-related non-fatal abdominal injuries 10% (n = 30), 311 admissions: 214 males aged 13-56 years, mean 22 years.

Method of data collection and biases not reported.

15. Prasad et al [37]. (1981), Fiji Case series – hospital and population-based Retrospective

2277 road traffic crashes Fiji, 1980; 872 road traffic crashes Western division province Fiji, 1980; 165 RTI-related admissions Lautoka hospital, 1980.

Road user type, age, sex, ethnic group, injury distribution, day and time, risk factors.

Road traffic crashes: Fiji 1%, WD 6%. WD 9% of road crash victims required hospital admission. Risk factors: dangerous driving, (66-70%); pedestrian fault (7.5%); alcohol (4%); speed (2.2-2.5%); mechanical defects. Most crashes occurred during the day and in dry weather conditions, 75-77% crashes occurred on tar-sealed roads. Hospital admissions: 73% male, 24% aged 20-29 years, 60% Fiji-Indian, 32% Fijian, 40-46% head and neck injuries. Crashes increased during school holidays, over weekend, afternoons and evenings. Children more likely to be injured as pedestrians. Falling off the back of utility vans common.

Limited reporting on methodology and quality of data. Risk factors indicated, but study design precluded quantification of risk. No biases reported.

16. Jayasuria [38] (1991), PNG, Ecological study Retrospective

3202 motor vehicle crash-related injury registrations (314 deaths), Royal PNG Constabulary database for traffic accidents, 1987; National statistical office data for motor vehicle crash-related injury trends prior to 1980. Motor vehicle registry, and Licensed drivers National statistical office

Motorisation, crashes, injuries/deaths by road user and vehicle type, age, time of crash

Injuries and deaths: drivers 18%, passengers 61%, pedestrians 19%, 23% 26-30 years. Deaths: Drivers 13%, passengers 51%, pedestrians 34%. Trends in injuries and deaths by road user type have not changed much since 1970. However, for pedestrians the fatality index (deaths/deaths and injuries) there was a significant increase from 11% to 18% for the period 1970 to 1987. Rate of pedestrian injury per 100000 population = 31 (21-30 years) 20 (0-10 years), and 18 (30-50 years). Total deaths and injuries by vehicle type: 44% utility - small pick-up trucks), 16% heavy goods vehicles - single unit trucks, 18% pedestrians, 10% cars. Pedestrians, 32% aged < 10 years. Crash rates by vehicle type (1982-87), declining, but buses increasing, to four times the rate of cars and 2.4 times that of utility vehicles. Severity of crash and vehicle type (1987); highest rates of injury and fatality: drivers (motorcycles), passengers (utility, heavy goods vehicles, buses). Increased crashes from 6 pm, and over weekends (Friday and Saturday)

Data collection, exclusion criteria, not reported. Results reported as trends in unadjusted event counts, proportions, and rates.

17. Hills et al [39]. (1993), PNG, Ecological study Retrospective

4485 motor vehicle crashes-related injury registrations, Royal PNG Constabulary database for traffic accidents, 1991. Other LMIC countries: Bandung n = 1059, (1990) Malaysia n = 37955, (1992) Karachi n = 1261 (1991) Colombo n = 946 (1991-92) Total crashes 45702.

Road user type, age, vehicle type, collision characteristics, day and time of crash.

Pedestrians ranked 1st or 2nd for RTIs Peak age for RTIs: 21-30 years (PNG, Karachi, Colombo); 16-25 years (Malaysia, Bandung) Pedestrians: 6-10 years peak for PNG and Malaysia. Vehicle type: utility trucks - PNG 45%, other LMIC countries – mostly motorcycles. Vehicle collisions: PNG – rollover, Malaysia side or 90°, Bandung collisions with pedestrians. Number of vehicles in collision: PNG > 70% single vehicle. Most injuries occurred in rural areas (70%), associated with higher speed, and delayed medical aid. Common day and time for RTIs, PNG - Friday and Saturday, usually alcohol-related; between 8 am to 8 pm.

Methodology for data collection not reported. Results reported as unadjusted proportions comparing countries. Identified variation in defining injury severity.

18. Nelson et al [40]. (1991), PNG, Ecological study Retrospective

5772 motor vehicle crashes-related injury registrations (1921 deaths), Royal PNG Constabulary database for traffic accidents, 1984; Motor vehicle registry, and Licensed drivers National statistical office, 1980-84; Traffic survey 1984 on vehicle occupancy.

Road user type, Vehicle type, ownership and occupancy characteristics Crash severity and crash rates

Fatalities: drivers 17%, passengers 46%, pedestrians 36%. Utility responsible for 38% of crashes, 65% passenger fatalities. 34% crashes involve single vehicles, RR of fatality; Driver - utility 0.99, bus 0.88, car 1.57; Passenger: utility 3.1, bus 2.4, car 1.3. RR of crash involvement; utility 1.07, bus 2.14, car 0.87. Open back vehicles responsible for high number of fatalities per crash, occupants thrown out of vehicle. Potential 27% reduction in crash fatalities if utility and heavy goods vehicles restricted to designed passenger occupancy

Methodology for data collection, exclusion criteria, not reported. Results reported as trends in unadjusted event counts, proportions, and rates, for motorisation, crashes, deaths and injuries stratified by road user, vehicle type, age and time of crash. Reported limitations to notification, collection, assumptions, and modelling – eg multiple vehicle crashes reported as total injuries per event therefore unable to identify vehicle with casualties. Effect estimates identify RR for drivers and passengers based on type of vehicle.

19. Hills et al [41]. 1996, PNG, Case control

893 drivers undertaking roadside breath alcohol testing. 12 road sites, two different sites per night, 10 pm - 2 am, Thursday to Sunday, for 5 weeks (1990).

Sex, ethnic group, BAC

97% males, 98% PNG Nationals, BAC > 80 mg/dl 24% Males, 21% Females, 24% drivers. (Driver BAC > 150 mg/dl 12%, BAC > 215 mg/dl 4%). Friday night early Saturday morning 29% drivers BAC > 80 mg/dl, Thursday 27%, Saturday 25%, Sunday 8%. BAC > 80 mg/dl 17% after 22 cases excluded. BAC > 80 mg/dl PNG nationals (citizens) 22% Non-PNG nationals 33% sampling error significant (p < .05)

Non-randomised sampling by site, day, time. No response rate provided, nor vehicle type. 22 participants excluded due to test conducted < 20 minutes from the last drink. Measurement bias for BAC not reported.

37 RTI-related admissions A&E, PMGH undertaking BAC tests, 11 pm to 3 am, for 15 weekends (1990).

Road user type, BAC and breath alcohol concentration.

Drivers 27%, passengers 62%, pedestrians 11%. 2 fatalities, 2% hospitalised, 76% discharged. BAC > 80 mg/dl in 50% of A&E admissions, and 8 drivers (n = 8)

Little information on methodology and distribution of blood and breath alcohol concentration results among cases. Results reported as events with no adjustment for confounding and bias, including differences in testing methods.