Hospital and 11/27 admissions to Grey’s Hospital. The remaining 16 patients were identified from retrospective examination of admissions records, but no further data were available.

Patient characteristics The 41 subjects had a mean (SD) age of 27 (12) years, and 34/41 were male. The most common mechanism of injury was road traffic collision. Thirty-seven subjects sustained blunt trauma. Burns patients (n=3) suffered injuries affecting 38 - 60% body surface area, with additional inhalation injury.

Follow-up data were available for 40/41 patients to the point at which they left the Pietermaritzburg trauma system. Of these, 16/40 survived to extubation, 13/40 died in hospital in Pietermaritzburg or were transferred back to the base hospital for palliation, and 11/40 were accepted by the regional neurosurgical facility and transferred to the level 1 trauma centre.

Table 1 displays patient demographics, mechanism of injury, indi­ c ation for intubation, qualification of intubating practitioner, anticipated difficulty, medication and equipment (including checklist) used, adverse events and clinical course.

Outcomes All included patients were successfully intubated; this was achieved on first pass for 28/39 patients for whom data were available. Successful first-pass intubation was more likely in an ED than in rural hospitals (19/22 v. 2/7; p=0.003). Identified adverse events due to intubation were 1/22 (ED), 5/8 (rural hospitals) and 2/9 (roadside). Table 2 displays first-pass success and adverse event rates, Table 3 compares outcomes with and without the use of a preintubation checklist, and Table 4 details the nature of recorded adverse events.

Preintubation checklist There was wide variation in use of the preintubation checklist, as illustrated in Table 1. Overall, its use was associated with higher first-pass success rates (p=0.001) and fewer adverse events (p<0.001), as shown in Table 3. No rural hospital intubations were done using a checklist.

Pharmacological agents Medication used for intubation is displayed in Table 1. Five patients were intubated with neither sedative nor muscle-relaxing medication, and 4/5 of these procedures were deemed inappropriate owing to absence of cardiac arrest. Inappropriate drugless intubation was more prevalent in rural hospitals than EDs (p=0.019), as shown in Table 2.

Equipment and monitoring There was heterogeneity in practice between locations in the use of introducers, video laryngoscopes and end-tidal carbon dioxide monitoring, as illustrated in Table 1. In particular, none of these modalities were employed in rural hospitals.

Discussion We found that intubation practice, first-pass success rates and associated complication rates in trauma and burns patients were variable between settings. Rural hospitals recorded a particularly low first-pass success rate and a considerable adverse event rate. Use of a preintubation checklist was associated with lower intubation-related morbidity in the ED and on the roadside, but a checklist was not used in the rural hospital setting. Minimum standards for prehospital monitoring of anaesthesia[14] were typically unmet.

Preintubation checklist Recent guidelines emphasise the usefulness of a preinduction check­ list in both prehospital and ED environments.[14] Although causality cannot be established from an observational study, particularly of this small magnitude, this study found the use of a formal checklist to be associated with a higher rate of first-pass success. This is consistent with current literature internationally.[13]

Edendale Hospital’s ED emergency intubation checklist is an A4 laminated sheet attached to each airway trolley and covering

planning, medication, patient preparation and equipment required. It also depicts the Difficult Airway Society’s advanced airway algor­ ithm,[17] including rescue techniques for ‘can’t intubate, can’t ventilate’ scenarios. It is unlikely that there is sufficient ethical equipoise to justify a randomised controlled trial. However, a simplified algorithm has been proposed for the rural hospital setting.[9] This inexpensive intervention could be implemented readily and may be particularly beneficial in rural hospitals, and its value could be tested through pre- and post-intervention quality improvement data on a site- by-site basis. The data generated at each participating site could be pooled regionally to increase power, and then used to audit SA practice against international consensus guidelines with a view to standardising care. Although this would require considerable cross-

party co-operation across hospital and prehospital providers, the current study shows that it is achievable.

Adherence to best pharmacological practice in emergency intuba­ tion was variable between locations. In order to reduce intubation- related morbidity, particular attention to this issue may be war- ranted. The stimulation caused by tracheal intubation without pharma­ cological adjuncts can increase intracranial pressure, increasing morbidity in head-injured patients with a reduced level of consciousness. Stimulation of the gag reflex may also elicit vomiting and lead to aspiration in this setting. This study found that a neuromuscular blocking agent was used in only half of rural hospital intubations. In some rural facilities, suxamethonium is kept in operating theatres but not in the casualty department. Hospital policies may stipulate a doctor’s presence and signature before the drug is removed from theatres. In an emergency situation with an urgent need to secure an airway and limited medical staffing, obtaining suxamethonium in a timely manner may be impractical owing to these barriers.

In rural hospital casualty departments, 3/8 intubations involved no sedative or muscle-relaxing medication. All took place in the context of traumatic brain injury in the absence of cardiac arrest. Plausible explanations include inability to access drugs in a timely manner, or inability to administer drugs owing to lack of intravenous access prior to intubation. The introduction of a tracheal tube without the use of pharmacological adjuncts puts these patients at considerable risk of harm, both physiological and anatomical, and raises questions about the balance of risks v. benefits of advanced airway interventions in this patient group.

To reduce these risks, appropriate drugs should be readily accessible in the emergency setting prior to intubation, and the use of supraglottic devices or simple airway manoeuvres may be more appropriate in some cases.

A standardised rapid-sequence induction protocol has been demonstrated to improve outcomes elsewhere,[18] and may be use­ f ul in rural KZN. Intraosseous administration of anaesthesia is effective in rapid-sequence induction,[19] and may also warrant consideration.

Equipment and monitoring International guidelines state that difficult intubation should be anticipated in the prehospital setting, and recommend routine use of intubating bougies.[14] Bougies were rarely used in emergency intubation of trauma and burns patients in Pietermaritzburg and its drainage area. The use of a metal stylet as an adjunct was not assessed in this study, but some providers reported that they would only employ an introducer after a failed first attempt or in a difficult airway situation. International research suggests poor diagnostic accuracy in predicting difficult tracheal intubation,[20] and this also appears true in KZN: only 1/7 of rural hospital intubations were anticipated to be in difficult airways, yet the first-pass success rate was just 2/7. Encouraging the routine use of an introducer in all emergency intubations in trauma and burns patients, irrespective of anticipated difficulty, may improve the likelihood of first-pass success and reduce intubation-related complications.

End-tidal carbon dioxide monitoring was used in 18/22 ED intubations, compared with just 7/19 of prehospital cases. As such, recommended minimum standards of emergency anaesthetic monitoring[21] were frequently unmet.

Study limitations This study is limited by its small sample, the self-administered nature of case report files, potential heterogeneity of subjects and injury severity from different sites, and incomplete follow-up. Meaningful multivariate statistical analyses could not be performed owing to sample size. For logistical reasons, patients accepted by the regional neurosurgical centre could not be followed up once they left Pietermaritzburg. A high turnover of medical staff in the Grey’s Hospital casualty department, combined with the absence of comprehensive electronic patient tracking and record-keeping systems across KZN, presented challenges to data collection. The background rate of emergency airway management in trauma across KZN is not ascertained, although regional data (e.g. Pietermaritzburg) are potentially deducible from local registries such as the Pietermaritzburg Metropolitan Trauma Service database.

Despite reassurances during briefing, the commercial nature of private ambulance services and the impossibility of data verification make the current study at risk of potential over-reporting of success and under-reporting of adverse events. Nonetheless, the high recorded incidence of adverse events, even in such a small study, highlights the opportunity for improving the safety of advanced airway intervention in these patients.

Possible adverse events following attempted tracheal intubation include misplaced tube, inappropriate cuff pressure, pneumothorax, vomiting, aspiration, hypoxaemia, hypotension, bradycardia and cardiac arrest; laryngoscopy also carries the risk of dental trauma.[12,22] Intubation-related morbidity cannot be accurately quantified through a study of this nature, as it may not be possible to establish with certainty whether events recognised at the time of intubation (e.g. dental trauma) were attributable to laryngoscopy and intubation or caused as a direct result of injury mechanism, and episodes such as hypoxaemia occurring at the time cannot be verified retrospectively.

External validity The trauma system in KZN is typical of practice across SA. It is possible that data from other developing trauma systems of simi­ lar scale and logistical framework might be in keeping with our findings. Our recommendations may serve to improve patient care in comparable situations.

Conclusions Clinical heterogeneity exists during emergency intubation for trauma and burns patients in the ED, prehospital and rural settings of KZN, SA. The use of a preprocedural checklist is associated with improved outcomes and may improve quality of care. Standardised rapid-sequence induction protocols, routine use of introducers and end-tidal carbon dioxide monitoring, and increased availability of intraosseous devices may also yield improvements in patient outcomes. These recommendations are in line with recently updated international guidelines, which also suggest monitoring of key performance indicators. Further research in this setting is both feasible and warranted.

Acknowledgements. We are grateful to the Pietermaritzburg paramedic crews and medical staff at Edendale Hospital, Grey’s Hospital and peripheral hospitals for their engagement with this project. We appreciate the support of hospital management for permitting data collection on their premises, and of the Pietermaritzburg Metropolitan Trauma Service for access to their trauma database. CTL would like to thank the Sir Arthur Thomson Charitable Trust, Association of Anaesthetists of Great

Britain and Ireland, British Medical and Dental Students’ Trust and Royal College of Anaesthetists for their support of this study through their elective bursary award schemes. Author contributions. All authors contributed to the conception and design of the study. CTL, JB and ACI collected data. CTL, DNN and NC analysed the data, interpreted results and prepared the manuscript. All authors critically revised and approved the manuscript and are accountable for its integrity. Funding. None. Conflicts of interest. None.