Review of a large trauma registry in Addis Ababa, Ethiopia: insights into prehospital care and provider training for trauma quality improvement

Study setting

ALERT Hospital is a tertiary care hospital in Addis Ababa, the capital city of Ethiopia. It serves a population of approximately 3.5 million and has 20 emergency department (ED) beds, 38 inpatient trauma ward beds, and 10 critical care beds. The hospital has 10 operating rooms, 3 of which are exclusively dedicated to trauma care.

ALERT Hospital Trauma Center was inaugurated in 2015 as one of three dedicated trauma centers in the country. The hospital began a basic trauma registry in 2016 by entering all patients who presented to the trauma ED, a separate building from the medical ED, in a ledger. Data were then later transcribed into the hospital’s electronic database (online supplemental appendix 1).

Study design

This was a retrospective cohort review of a previously existing basic trauma registry at one trauma center in Addis Ababa, Ethiopia. The database was maintained by hospital personnel in Microsoft Excel and was deidentified prior to analysis for study purposes. After ethical approval was obtained by the hospital Institutional Review Board, the deidentified database was delivered to study personnel for cleaning and analysis. All entries in the database for the continuous 3 years since its inception were included in the original dataset. The registry was developed by the Ministry of Health and contains information on patient demographics, course of care (mode of arrival, referral status, triage category, time to triage, disposition after triage), diagnostic information (mechanism of injury, diagnosis), and ED outcomes (length of ED stay, disposition status from ED). No information after ED departure was included in the registry, and aside from the initial triage category, which accounts both for mechanism of injury and initial vital signs, no other physiologic or intervention data were included.

Missing data and excluded variables

The Excel format of the registry had additional duplicate or redundant variables (17 total) that were excluded and an additional 5 variables with such little data were also excluded from the analysis, leaving the database with a total of 27 variables in the final analysis (figure 1). The vast majority of patients (12 776, 99.7%) with missing information on the mechanism of injury were ultimately excluded from analysis. Although 2364 patients were recorded to have sustained a road traffic crash, only 12.1% (287) of them contained any information on the type of vehicle involved. All patient records, including those with missing data, were included in this available-case analysis, which captured missing data in respective ‘unknown’ categories. Missing data numbers are not listed in tables for brevity, but percentages are calculated from total n in each respective category, including missing data.

Request permission

Data cleaning and analysis

All patients recorded in the trauma registry between September 2016 and March 2019 were included in this study. The dataset was cleaned using Excel V.16.39. Data cleaning involved converting dates from the Ethiopian to Gregorian calendar as well as correcting errors and typos in initial data entry. In the event of an unclear entry, data clerks and hospital staff were consulted, and uninterpretable entries were discarded. Data were analyzed using Stata V.16.1.

Patients were stratified by injury severity based on their vital signs and diagnosis upon triage. Patients are first assigned a numerical score according to the Modified Early Warning Score—a scoring system intended to identify patients at increased risk of clinical deterioration and death, which has been validated in medical and surgical patients.16 17 Triage categories were organized by color, with ‘red’ as the most critical category, followed by ‘orange’, ‘yellow’, and ‘green’ in descending order of severity and urgency; ‘black’ signified patients who were dead on arrival (tables 1 and 2).

Each diagnosis entry was assigned a standardized Abbreviated Injury Score body region code. These were further sorted into injury categories within specific body regions, including head, face, thorax, abdomen, spine, upper extremity, lower extremity, and body region not otherwise specified.

Descriptive and summary statistic tables were generated using patient demographic, injury, and hospital course information. Comparative analysis included using Χ² or Kruskal-Wallis tests of significance as appropriate to analyze differences between groups. Univariate and multivariate analyses were also conducted to determine predictors of in-ED mortality.

Triage category, injuries and disposition after triage

Most patients were triaged in the ‘yellow’ triage category (n=8293, 64.7%), whereas 480 (3.7%) were ‘red’ (table 2). After initial triage, about one-third of patients were transferred to the resuscitation area (n=4291, 33.5%), and another one-third went back to the waiting area (n=4187, 32.7%). The rest of patients either went directly to the operating/procedure room (n=2188, 17.1%) or an unmonitored ED examination room (n=759, 5.9%). 18% (n=2364) of patients were involved in a road traffic crash; of these, the majority (1861, 78.7%) were pedestrians. Most (65.0%) patients were stabilized and discharged from the ED; 12.2% (1567) were admitted to the hospital, and few (57, 0.4%) patients died in the ED.

Demographics, injuries and outcomes by triage category

Of the most severely injured ‘red’ patients (480), 31.0% (149) presented by ambulance, compared with 4.2% (87) of ‘green’ patients (table 3). 19% of red patients (90) and 45.1% (946) of green patients presented by themselves without vehicular assistance. 32% of red patients (155) were sent to the operating/procedure room after triage, whereas 24.9% (119) returned to the waiting area and 19.8% (95) were sent to the resuscitation area. These proportions are reflected similarly in the breakdown of orange patients’ disposition after triage, with 34.8% (515) going to the operating/procedure room, 25.1% (371) to the waiting area, and 21.3% (316) to the resuscitation area.

Soft tissue injuries (5584, 43.6%) and fractures/dislocations (2489, 19.4%) were the most common injuries (table 4). ‘Red’ patients were most likely to present with head injuries (253, 52.7%) such as traumatic brain injuries and intracranial bleeds, whereas green patients were most likely to present with soft tissue injuries (1310, 62.4%) (table 3). The vast majority of upper (82.8%) and lower extremity injuries (88.3%) consisted of fractures/dislocations, most of which were assigned a yellow triage category. Although many of the differences between variables by triage category reached statistical significance, it was beyond the scope of the article to explore detailed statistical differences between individual triage category groups.

Most green patients (1665, 79.3%) were stabilized and discharged from the ED, 4.2% (89) were admitted, and 0.6% (12) died in the ED (table 3). Meanwhile, 29.2% of red patients (140) were stabilized and discharged, 31.5% (151) were admitted, and 4.6% (22) died in the ED. The ED length of stay was not reliably recorded in the registry and therefore not included in this analysis.

Predictors of ED mortality

A multivariate logistic regression analysis was performed to examine predictors of mortality in the ED. When controlling for age, gender, and triage severity category, residence in Addis Ababa (OR=0.52, p=0.049) and interfacility transfer (OR=0.17, p=0.001) predicted lower likelihood of in-ED mortality (table 5). Patients who arrived by ambulance (OR=2.20, p=0.017) and had sustained head injury (OR=3.11, p<0.001) had increased odds of death in the ED.

Prehospital care and initial patient triage

The small proportion of critically injured patients arriving by ambulance highlights gaps in the prehospital care system and elucidates opportunities for more robust ambulance services and communication networks to expedite patient transport. A recent large-scale study from Malawi noted similar problems in long transport times for patients with serious injuries and highlighted the importance of improved emergency medical services.18 Similarly in Nigeria, a high proportion of patients were transported to trauma centers by laypersons with long transport times.19 The pattern of a high proportion of road traffic accidents and intracranial injuries among those seriously injured was also seen in a recent study in Tanzania using data from the WHO model trauma registry.20 In cases of seriously injured patients with head injury, rapid transport to a trauma center with highest level of care provides the best opportunity for survival. In this setting, a potential opportunity for improvement could be greater access to ambulance services for critically injured patients to receive immediate care during transport and reach the trauma center as rapidly as possible.

Although critically injured patients should be prioritized to receive ambulance transport to trauma centers, conversely, low-acuity patients can safely be treated at lower levels of care. The high number of soft tissue injuries and low-acuity patients treated at ALERT creates difficulties in patient flow, rapid triage, and treatment at one of the few designated trauma centers in Ethiopia. Many of these minor injuries could likely be treated at lower levels of care, freeing space at the trauma center to manage more complex injuries but need to be managed through an improved prehospital triage process to ensure the correct mix of patients is optimized at each level of care. Currently, the Federal Ministry of Health is developing a centralized ambulance dispatch call center for Addis Ababa and constructing ambulance ‘hubs’ distributed throughout the city for more coordinated and timely prehospital response. The Ministry of Transport has also supported several training events for drivers to receive basic first responder training for road traffic accidents. The results from this study provide convincing support for this service expansion.

ED triage and resuscitation

Overall, in the ED, mortality was low at 0.5%. However, 12 (0.6%) of patients triaged as ‘green’ died in the ED. This would suggest that either a serious injury was missed, or these patients may have been triaged inappropriately. In such circumstances, a death audit could be performed to identify the exact details related to these patient presentations and a preventable death review by the trauma quality improvement committee. Furthermore, the disposition after triage also points to potential inefficiencies in workflow, with nearly half of ‘green’ patients being sent to the resuscitation area, which is typically intended for more critically injured patients. Conversely, about one-third of critically ill patients were sent to the waiting room after triage, emphasizing the need for staff trauma training and restructuring of triage systems so that patients are immediately allocated to the appropriate level of care on arrival. These findings could serve as a basis for incentivizing the establishment of trauma quality improvement programs to conduct routine preventable death reviews and identify opportunities for improvement, which do not currently exist in any robust form in Ethiopia.

Health center referrals

The large number of referrals from health centers is also related to the organization of prehospital transport and referral structures. Rather than being referred to receive clinical care, as most of the referrals were for patients with low triage acuity, anecdotally, many of these patients were referred to have medicolegal documentation completed, which can only be performed by physicians. Although data regarding reason for referral were not specifically documented in the registry, the frequent patient presentation from health centers for medicolegal documentation is well-known among Ethiopian physicians. The burden of this documentation being transferred to treating physicians at trauma centers can distract from the time dedicated to clinical service provision for more severely injured patients. Further study on the reasons for trauma center referral and medicolegal documentation process may help identify possible opportunities for improvement. Although the ambulance dispatch system may help to address some of these inappropriate referrals, education is needed at the health center level, as well as consideration of appointing more general physicians at health centers to alleviate the burden of injury documentation being performed at the trauma center. A more widely implemented trauma intake form that can serve as a medicolegal document nationwide would be another potential solution.

Expansion of trauma registry data

To fully use data on injured patients to inform improvement efforts at the facility level and beyond, an integrated and comprehensive trauma registry is necessary. However, this has been difficult to implement and maintain at the hospital level thus far. A study of injured patients at Yekatit 12 and Black Lion Hospitals in Addis Ababa demonstrated a high prevalence of trauma and poor outcomes for severely injured trauma patients and called for the implementation of trauma registries to better track patient care and outcomes in trauma,21 which was subsequently implemented with higher-quality data.22 Similarly, in Mekele, Ethiopia, a retrospective study found a high prevalence of traumatic injuries in EDs, primarily from interpersonal violence, falls, and road traffic accidents, and also called for the implementation of a trauma registry for higher-quality data.23 One study including patients from two referral hospitals in Addis Ababa found a higher than expected mortality rate by Injury Severity Score,21 highlighting the importance of not only cataloging injury data with comprehensive registries, but identifying opportunities for trauma quality and outcome improvement initiatives as well. As a next step at ALERT Trauma Center, a more detailed trauma registry with information on physiologic data and interventions will be implemented, and staff will be trained to improve data completeness. This registry, using the WHO trauma intake form and standardized WHO trauma registry, will allow for injury scoring and more details on ED and hospital interventions, as well as inpatient complications and discharge outcomes, all of which are essential for understanding current practices and identifying improvements at the hospital level. This more comprehensive registry will inform a facility-based quality improvement program as part of a national initiative to strengthen trauma care in Ethiopia. A similar process has been undertaken in Tanzania, lessons from which may pave the way for implementation in Ethiopia.24

Limitations

Data incompleteness posed a significant limitation in the evaluation of this trauma registry. The retrospective nature of this study made real-time improvement of data quality impossible, and as data clerks recorded entries by hand, this method produced a significant number of errors and missing information in the registry. Although the clerks were trained in data collection protocols, they were not able to consistently capture pertinent information. For example, date and time entry formats were not standardized, and many words were incorrectly transcribed from the ledger into the electronic database. These errors in data transcription and entry further support the notion of an electronic data capture mechanism. Also, the available-case approach to analyzing datasets with missing entries can introduce bias; however, in our analysis, the likelihood of such bias was minimal because the causes of missing data were unlikely to have been systematic or associated with outcomes.

Furthermore, physiologic data necessary to calculate Injury Severity Scores and interventions performed in the ED were not included in this registry. Additionally, in-hospital interventions and outcomes, including those routinely included in most trauma registries such as hospital length of stay, prevalence and cause of overall mortality, and intensive care unit outcomes are not available in this registry. Individual chart review was deemed to be too difficult and costly for the scope of this project as paper charts must be individually collected from the record room and many paper records are incomplete. It is difficult to target and prioritize areas for improvement without this more granular understanding of care provided at the patient level. Additionally, causes and timing of patient deaths were not available in the registry; therefore, it is difficult to infer detailed interventions that may have had the highest impact for these patients. This registry has been modified and at present, a more comprehensive trauma registry is being used at ALERT Hospital following the WHO trauma intake format.

Studies on trauma registry implementation in LMICs demonstrate an enormous range of form completion rates, ranging from 21% to 90%.25–27 One study in Ethiopia at Tikur Anbessa Hospital characterized the successes and challenges of standardized trauma registry form implementation, noting that data capture rate was low, and that lack of training and supervision was a key challenge to form completion.28 Furthermore, it is difficult to evaluate the types and quality of in-hospital services provided as the registry did not contain such information. Patient charts also commonly do not document all ED interventions or procedures. A more comprehensive registry is needed and currently beginning implementation to capture more information regarding in-hospital trauma care. Ideally, this would be integrated with the medical record to avoid duplication of data collection and overburdening the staff with clerical tasks of registry data collection.