Prehospital trauma care reduces mortality. Ten-year results from a time-cohort and trauma audit study in Iraq

Study design

The reference population consists of trauma patients in low-income countries with long pre-hospital transport times. The study was conducted with a time-period cohort design defined by a stepwise expansion of the actual trauma system: In period 1, from 1997 to 2000, the catchments area of the prehospital trauma system was the rural mine fields of Northern Iraq; in period 2, from 2001 to 2003, the trauma system was expanded to also target highway traffic accidents in the Northern sector while still being operational in the rural North; from 2004 to 2006 the trauma system developed further to include the war zones of Central Iraq, yet still in action in the previous catchments areas (Figure 1).

Intervention

Data collection and processing

All trauma patients managed by the system from January 1997 through December 2006 were consecutively included in a trauma registry. The data were collected at the first in-field encounter with a trained paramedic, and on admission at the referral hospital. The paramedics registered demographic factors, in-field response time from injury to the first encounter with the paramedic, and total prehospital transit time. They also registered physiological indicators at the first encounter in-field and again on hospital admission. All in-field data including photos were scrutinized by the main author at monthly meetings. The data for anatomical severity grading, Injury Severity Score (ISS), were collected by the trauma system supervisors at the referral surgical centers [6]. Due to the local cultural tradition, autopsies on fatal cases were not performed. The main outcome variable was trauma death defined as on-site deaths, deaths during the pre hospital phase, or trauma-related in-hospital deaths. The ISS ranges from 1, light injury, to 75, cases with ISS > 9 being defined as serious, and ISS > 15 as major trauma victims. By definition, patients with ISS = 75 have injuries incompatible with survival, and this subset (n = 238) was excluded from analysis. End-point data could not be collected in 35 patients evacuated to surgical centers outside the study area or cross-border to Iran; also these patients were excluded from the study (Figure 2).

The Physiological Severity Score (PSS) was used for estimation of physiological severity. The PSS is a simplified version of the Revised Trauma Score for triage (RTS) where the Glasgow Coma Scale element is replaced with a five-grade conscious level indicator [7]. The two other indicators, respiratory rate and systolic blood pressure, were rated according to the standard RTS guideline [8]. The PSS score ranges from 0, lifeless, to 12, normal physiological condition. The PSS on admission were compared to the PSS at the first in-field encounter; cases with negative ΔPSS were defined as prehospital treatment failures. Tests of inter-rater reliability in PSS scoring were not undertaken. Audit of patients with unexpected outcomes is an established method of trauma system quality assurance [9]. To identify and review unexpected survivors and unexpected fatalities, a model of death risk prediction was constructed based on the study data. Unexpected survivors were defined as survivors with predicted probability of trauma death (Pd) ≥ 0.5. Unexpected fatalities were defined by two criteria: Pd < 0.25, and in-field PSS ≥ 6.

Data analysis

Assumed continuously and symmetrically distributed variables are expressed by mean values with 95% confidence intervals (95% CI) constructed by the Student procedure. Due to the irregular shape of several continuous variables, comparisons were undertaken using nonparametric methods [10]. Proportions were described using the exact 95% calculated confidence interval [11]. Receiver Operating Characteristics (ROC) analysis was used to estimate the accuracy of mortality predictors. A predictor is considered accurate if the area under the ROC curve (ROC-AUC) is larger than 0.8 [12]. Most probabilistic models reported in the literature for estimation of trauma mortality risks are based on urban cohorts managed by advanced trauma systems. To develop a risk predictor with optimal fit in the actual study cohort, a logistic regression model was used to identify patients with unexpected outcome. All assumed predictors of trauma death were included using a backward selection process with inclusion at significance level of 5%. The logistic model was evaluated using the Hosmer-Lemeshow test and ROC analysis.

Ethical considerations

The Directorate of Health Suleimaniah, Ministry of Health, Kurdistan Region gave ethical approval for the study (Ref. no. 22082). There is no other authorized committee for medical research ethics in North Iraq.

The data were stored and processed according to ethical permission from the Norwegian Social Science Data Service (ref. no. 2006/13702).

Trauma system outcomes by time cohorts

The epidemiology of trauma shifted during the study period with a massive increase in the numbers of road traffic casualties in period 3 (table 4). There was a reduction in overall mortality from 17% in period 1 (95% CI: 15 - 19) to 4% in period 2 and 3 (95% CI: 3.5 - 5). Prehospital mortality rates were reduced from 16% in period 1 to 1.7% and 1.3% in period 2 and 3 (95% CI diff: 11 - 18). The main contributions to improved survival were reduced mortality in critical area and multiple major injured. In burn patients the mortality increased from period 2 to period 3 (95% CI diff:11.6% - 26.8%)(Figure 3). Due to reduction of mean injury severity from time cohort 1 onwards (table 5), regression analysis was used to adjust for severity variations. A model combining ISS, PSS and the time cohorts explained 70% of the variations in trauma mortality, ISS being the heaviest predictor with ROC-AUC value > 95%, but also time cohorts contributing significantly. The in-field response times were reduced from 1.6 hours in period 1 to 0.7 hours in period 3 (95% CI diff: 0.7- 1.1), and there was a reduction in total out-of-hospital time from 4.4 hours to 2.3 hours (95% CI diff: 1.8 - 2.4).

	Time period 1	Time period 2	Time period 3
Casualties from mines and	268	209	329
actions of war	33%	26%	41%
Road traffic accident	27	115	1153
casualties	3%	8%	89%
Burn casualties	26	83	164
	10%	30%	60%
Other trauma casualties	84	64	266
	20%	16%	64%
Total	405	471	1912

Life-support interventions enhancing survival

Diagnosis, category (blunt/penetrating), ISS, and PSS explained 77% of the variation in trauma mortality and gave a good fit with a ROC-AUC value of .99; ISS was the dominant predictor, alone yielding a ROC-AUC value of .98. Twenty seriously injured patients with ISS from 9 to 30 were identified as unexpected survivors, and there were 44 unexpected fatalities, all of them major trauma victims with ISS > 15 (Figure 4). In the group of unexpected survivors, all patients were in poor physiological condition at the first in-field encounter with a PSS ≤ 6 but had improving physiological indicators during the prehospital phase. Twelve patients with traumatic brain injury were among the unexpected fatalities with critical area injuries, all twelve dying within 48 hours after injury. These deaths occurred before neurosurgical service was established at the referral hospitals in 2006. Also in the group of unexpected fatalities were three cases with abdominal hemorrhage dying immediately on admission after two-hours' prehospital transit time. Six patients diagnosed as "extremity injury" suffered unexpected deaths due to associated head injuries. Among the 13 unexpected deaths with multiple major injuries, seven patients were admitted with close to normal physiological scores but died from internal hemorrhage in hospital hours after admission, one of them a patient with traumatic brain injury who did not undergo neurosurgery; four of the seven patients were injured by fragmentation mines. Ten burn fatalities with probability of death > 0.25 had PSS > 10 on admission but died within one week after the injury from infectious complications and/or organ failure.

There were 36 "prehospital treatment failures" defined as seriously injured on-site survivors with deteriorating out-of-hospital physiological severity scores despite care being provided. In eight cases diagnosed in the field as "extremity injury", limb bleeds were efficiently controlled but still the level of consciousness deteriorated during the prehospital phase due to undiagnosed brain injuries. In the other cases in the treatment-failure group, the main reason for deteriorating PSS values was worsening respiratory rate scores.

To identify specific life support measures with effect on survival, patients with respiratory problems and external bleeds were scrutinized. Most patients with airway problems were managed by basic measures only; endotracheal intubation was done only in 19 patients, crico-thyrotomy in one. Forty-seven patients had severe chest injuries with ISS ≥ 9. In this group, 39 patients had less than optimal respiratory scores in-field but 30 of the 39 had normal respiratory rate at end-point. Eighty-two patients with severe limb bleeds had BP < 70 mm Hg at first in-field; 69 of them were normotensive on hospital admission. The only fatal case in this group of patients was one man found three hours post-injury with traumatic double amputation from a fragmentation mine.

Costs and effectiveness

Throughout the study period 180 paramedics were trained and joined the trauma system. By the end of 2006, 135 of them remained active. The treatment costs per patient (medical treatment, evacuation, data gathering and quality control) varied during the study period from US$ 130 to US$ 180.

Trauma audit

The high rate of unexpected deaths, 25% of all fatalities, should concern us; were these deaths avoidable? Some of the unexpected deaths from traumatic brain injuries could probably have been avoided if neurosurgical service had been in place throughout the study period. Most of the unexpected deaths in patients with abdominal bleeds might have been avoided if damage control surgery had been conducted at an early stage at a district hospital or immediately on admission at the referral hospital. The effect of the prehospital treatment was good also in burn cases; however, this did not have any significant impact on burn fatality rates, which remained high throughout the study period. Most burn fatalities, including the ten unexpected deaths observed in the study, are late deaths due to postinjury immune depression; in such cases survival depends on postinjury surgical care rather than prehospital life support. We should thus conclude that there is ample room for improvement of in-hospital trauma care in the study area.

Six patients diagnosed by the paramedic as "extremity injury" suffered unexpected deaths. In these patients the level of consciousness deteriorated during the prehospital phase despite efficient control of the external bleeding. The findings indicate that associated injuries (traumatic brain injury, internal hemorrhage) went undiagnosed by the paramedic. Especially in high-energy blast injuries (car bombs, fuel-air explosives) early clinical signs of brain injury and abdominal bleeds may be discrete and easy to miss [17]. We therefore recommend triage training especially for such mass casualties to help reduce miss-triage on-site and in the emergency room.

The prehospital treatment protocol is under debate and several studies question the usefulness of advanced measures [18]. Uncontrolled extremity bleeding is still a leading cause of avoidable battlefield deaths despite homeostatic agents are now being applied on wide scale in advanced trauma systems [19]. The actual simple treatment protocol - no tourniquet but sub-facial packing plus compression plus hypothermia prevention - proved effectual: 84% of extremity injured patients with severe in-field hemorrhage were normotensive on admission. We emphasize hypothermia prevention including warm IV fluids as part of the in-field treatment protocol for bleeds. In the actual study we did not gather data on core temperature, but previous studies conducted in the same study area document significant impact of simple preventive measures on body core temperature through protracted evacuations [20]. Airway block in unconscious patients is another common reason for avoidable trauma death. Very few study patients (< 1%) received advanced airway support in-field. Of four prehospital deaths from traumatic brain injury, one might have been prevented by in-field tracheal intubation; in the group of non-head injured unconscious patients we could not identify any preventable deaths caused by airway block. The findings indicate that basic airway measures are sufficient to control the airway in most risk cases. The treatment protocol did not included in-field chest tube drainage. There was one prehospital chest fatality, a patient with large chest wall wound. Among the other severe chest cases 75% had normal respiratory rate on hospital admission. Also for chest injured it seems that basic life support measures done early is the key to survival - IV ketamine pain relief, half-sitting position, and hypothermia prevention.

The intervention had a sustained impact on the quality of the EMS system in the study area. Despite adverse working conditions the overall retention rate of trained paramedics was high, 75%. The system performed on low costs; per-case costs of less than US$ 200 including systematic quality control should be a feasible price for most low-income communities. Also on national scale the model has had an impact; a two-tier dispatch system is now under implementation in the major cities in North Iraq, and there are requests from the Ministry of Health to implement the actual chain-of-survival model also in Central and South Iraq.