Handover in the emergency department: Deficiencies and adverse effects
Ken Ye, BMedSci, Advanced Medical Science Student; David McD Taylor, MD, MPH, DRCOG, FACEM, Director of Emergency Medicine Research; Jonathan C Knott, PhD, GDEB, FACEM, Research Fellow; Andrew Dent, MB BS, FACEM, Director of Emergency Medicine; Catherine E MacBean, BA(hons), Research Assistant.
Abstract
Objective: To determine problems resulting from ED handover, deficiencies in current procedures and whether patient care or ED processes are adversely affected.
Methods: A prospective observational study at three large metropolitan ED comprising three components: observation of handover sessions, 2 h post-handover surveys of the receiving doctors and a general survey of ED doctors.
Results: The handovers of 914 patients were observed during 60 handover sessions in a 3-month period. Medical information, including presenting complaints, was handed over better than communication and disposition information. Seven hundred and seven (77.4%) of 914 potential post-handover interviews were undertaken. Most (88.3%) doctors thought the handover was ‘adequate/good’. However, information was perceived as lacking in 109 (15.4%) handovers, especially details of management (35, 5.0%), investigations (33, 4.7%) and disposition (33, 4.7%). There was a significant difference in the perceived quality of handovers (1–5 scale where 5 = excellent) when all required information was handed over and when it was not (median scores 4.0 vs 3.0, respectively, P < 0.001). As a result of perceived inadequate handovers, the doctor/ED and patient were affected adversely in 62 (8.8%) and 33 (4.7%) cases, respectively, for example, repetition of assessment, delays in disposition and care. Fifty doctors completed the general survey. Most believed communications made to inpatient units, inaccurate/incomplete information and disorganization were problematic.
Conclusion: Deficiencies in handover processes exist, especially in communication and disposition information. These affect doctors, the ED and patients adversely. Recommendations for improvement include guideline development to standardize handover processes, the greater use of information technology facilities, ongoing feedback to staff, and quality assurance and education activities.
Introduction
Ongoing patient management involves the transfer of care between individuals and/or departments1 and necessarily entails the ‘handing over’ of patient information and responsibility to the receiving staff.2 The goal of handover is the accurate and reliable communication of task-specific patient information across shift changes, thereby ensuring a safe and effective continuous work environment.3 Ideally, the handover process would be seamless with the receiving staff as informed as the staff handing over. However, although handover has been identified as an area of high risk for adverse events, the deficiencies associated with the process have been poorly delineated.2–7
It has been reported that the quality of handover is variable as a result of a lack of supporting framework7,8 and that a lack of formal advice and guidance consistently hampers good handover practice.9 Poor communication and omissions might lead to inaccurate handover of patient details.10 This might lead to systematic errors that result in wasted time and resources, inefficiency and frustration and has the potential for patient morbidity and mortality.1,11,12
Although nursing handover has been investigated frequently,8,10,13 there have been few investigations of medical handover, especially the handover procedures of ED doctors.2,5 Consequently, the problems associated with ED handover are not well understood. The present study aimed to examine the handover process, determine the problems that occur during handover, the deficiencies in current practice and whether patient care or ED processes are compromised as a result of ineffective handover practice.
Methods
This was a multifaceted study that comprised critical observation of patient handovers, a post-handover survey of ED doctors who received patients and a general survey of ED doctors. It was a multisite study undertaken between January and April 2006 in a tertiary referral/adult trauma ED, an urban tertiary referral ED and a metropolitan ED in Melbourne, Australia. The ED have annual patient censuses of approximately 55 000, 50 000 and 40 000 patients, respectively. The Human Research and Ethics Committees of each participating institution authorized the study.
The handover format at the study sites varied. At one site, the doctors met in a quiet room adjacent to the ED and used information technology (IT). The electronic patient log, accessed through the Emergency Department Information System (EDIS), was projected onto a screen and the patients discussed. At the second site, handover was similar but only used EDIS in the second half of the study period. At the third site, EDIS was used throughout the study period. However, in the first half of the study period, handover was undertaken in the ‘write up’ area of the ED with each doctor manning his/her computer. Subsequently, it was undertaken in the consultants' office with doctors gathering around a single computer operated by a consultant.
Observation of handovers
A single investigator (KY) observed 10 morning and 10 afternoon handover sessions, on 10 randomly selected weekdays, at each site (total of 60 sessions). For each patient, a preformatted Handover Observation Checklist was used to record whether 22 items of information that might have been relevant to the patient's handover were handed over. These items included demographics, presenting complaints, details of past medical history, examination and investigation findings, communications made, and management plans. The number of items handed over therefore ranged from 0 to 22. The checklist was compiled using information in the medical literature and revised in a focus group of emergency physicians. The penultimate list was trialled extensively in handover sessions to ensure that it captured all items of information that were handed over. The checklist was not a ‘gold standard’ of what should be handed over but rather a list of items potentially handed over.
This process determined the nature and amount of the information handed over. All doctors were aware that their handovers were being observed. No formal sample size calculation was undertaken. However, for this descriptive observational activity, a sample of at least 500 patient handovers was expected to provide a rich source of data.
Post-handover survey
Structured interviews were undertaken, with doctors who had received patients, 2 h after each handover session. This allowed the doctors to become familiar with their patients. The interviews comprised a number of researcher-administered questions each evaluated for content and face-validity, trialled and revised before use. Additional comments on the handover were also permitted.
The interviews documented the seniority of the doctor handing over and perceived ‘quality of handover’ scores (5-point ordinal scale: 1 = very poor, 5 = excellent). They also determined whether all information required for a seamless handover was handed over (yes/no), what required information was not handed over and whether or not (and how) the receiving doctor, the ED or the patient had been affected adversely by a lack of information. The interviews were ‘matched’ to the patients whose handovers had been observed initially. The nature of the information handed over was examined against the quality of handover score. A sample size of at least 168 interviews was required (84 where all required information was handed over and 84 where it was not) in order to demonstrate a difference in mean quality of handover score of 0.5 (SD 1.0, power 0.9, level of significance 0.05).
General handover survey
Registrars and consultants were asked to complete a purpose-designed self-administered questionnaire. This survey explored perceived problems regarding ED handover processes, information poorly handed over, adverse effects associated with poor handovers, the minimum information thought necessary and issues related to the improvement of the handover process. For each question, respondents were required to tick one or more of a number of possible responses, with an option of recording additional responses. Suggestions were also requested, in free text format, as to how handover could be improved. Each question was evaluated for content and face-validity, trialled and revised before use. No formal sample size calculation was done as attempts were made to survey all registrars and consultants working in the participating ED.
Most data have been analysed descriptively. The Mann–Whitney U-test was used to compare ordinal data, Spearman's Correlation test for data not normally distributed and Fisher's Exact test for comparison of proportions. The level of significance was 0.05. spss statistical software14 was used for all analyses.
Results
Observation of handovers
A total of 914 patient handovers were observed (mean 15.2 ± 5.8 patients/session). The sessions had a mean duration of 23.2 ± 9.2 min (range 6–58 min) and a mean number of doctors in attendance of 11.5 ± 3.5. Overall, each handover consumed a mean of 4.6 ± 2.7 doctor-hours.
The median number of items of information handed over for each patient was six (range 1–11) (Table 1). The patient's name was given in two-thirds of cases. In most cases, at least one item of patient identification (name, sex, cubicle number) was used. The presenting complaint was the item of medical information most commonly handed over. Information regarding patient disposition and communication were handed over less often.
| No. patients (%) | (95% confidence interval) | |
|---|---|---|
| Items of information handed over (914 patient handovers observed) | ||
| Sex | 872 (95.4) | (93.8, 96.6) |
| Presenting complaints | 863 (94.4) | (92.7, 95.8) |
| Patient name | 618 (67.6) | (64.5, 70.6) |
| Cubicle number | 514 (56.2) | (53.0, 59.5) |
| Usual medication/treatment | 356 (39.0) | (35.8, 42.2) |
| Investigations results | 350 (38.3) | (35.1, 41.5) |
| Patient age | 340 (37.2) | (34.1, 40.4) |
| Past medical history | 332 (36.3) | (33.2, 39.6) |
| Examination findings | 295 (32.3) | (29.3, 35.4) |
| Inpatient unit contacted | 208 (22.8) | (20.1, 25.6) |
| Investigations ordered | 197 (21.6) | (19.0, 24.4) |
| Inpatient registrar contacted | 120 (13.1) | (11.0, 15.5) |
| Bed arranged | 48 (5.3) | (3.9, 7.0) |
| Time of ED presentation | 21 (2.3) | (1.5, 3.6) |
| Inpatient specialist contacted | 15 (1.6) | (1.0, 2.8) |
| Follow up arranged | 15 (1.6) | (1.0, 2.8) |
| Other† | 29 (3.2) | (2.2, 4.6) |
| Required information not handed over (707 post-handover surveys) | ||
| Management progress and plans | 35 (5.0) | (3.5, 6.9) |
| Investigations ordered and/or results | 33 (4.7) | (3.3, 6.6) |
| Disposition plans | 33 (4.7) | (3.3, 6.6) |
| Important past history | 27 (3.8) | (2.6, 5.6) |
| Follow-up plans | 22 (3.1) | (2.0, 4.8) |
| Demographics | 12 (1.7) | (0.9, 3.0) |
| Other‡ | 24 (3.4) | (2.2, 5.1) |
- † Planned time of departure, general practitioner contacted, medications prescribed, allergies, family contacted, investigations pending.
- ‡ Need for interpreter, current health status, examinations findings, documentation errors/omissions, follow up or discharge letters incomplete, unclear documentation, changes in plans.
Post-handover survey
Of the 914 handovers observed, 707 (77.4%) were followed by a post-handover survey. In the remaining cases, the receiving doctor was unavailable because of work requirements. No doctor refused interview when available. All required information was not handed over in 109 (15.4%) cases (Table 1). Management, investigation and disposition details were most commonly omitted.
The large majority of handovers was thought to be adequate or better (Table 2). Overall, the median quality of handover score was 3.0 (range 1–5). No correlation was found between the number of items handed over and the quality of handover score (r =−0.01, P = 0.86). However, there was a significant difference (P < 0.001) between the scores for handovers where all required information was handed over (median 4.0, range 2–5) and where it was not (median 3.0, range 1–4). There were no statistically significant differences in perceived handover quality according to seniority of doctors (P = 0.17). However, registrars tended to have a higher proportion of good/excellent handovers (54.5%) compared with consultants (51.0%) and intern/residents (42.8%). Approximately 5.0% of handovers were poor or very poor.
| Quality | Intern/resident | Registrar | Consultant | All | ||||
|---|---|---|---|---|---|---|---|---|
| n | (%) | n | (%) | n | (%) | n | % (95% confidence interval) | |
| Very poor | 2 | (0.8) | 2 | (0.6) | 1 | (1.0) | 5 | 0.7 (0.3,1.7) |
| Poor | 14 | (5.6) | 15 | (4.3) | 3 | (3.1) | 32 | 4.6 (3.2,6.4) |
| Adequate | 127 | (50.8) | 143 | (40.6) | 44 | (44.9) | 314 | 44.9 (41.1,48.6) |
| Good | 96 | (38.4) | 166 | (47.2) | 41 | (41.8) | 303 | 43.3 (39.6,47.1) |
| Excellent | 11 | (4.4) | 26 | (7.3) | 9 | (9.2) | 46 | 6.5 (4.9,8.7) |
| Total | 250 | (100) | 352 | (100) | 98 | (100) | 700 | 100 |
- † Seniority missing in seven cases.
The adverse effects in the 109 cases where not all required information was handed over are described in Table 3. The doctor/ED and patient were adversely affected in 62 (56.9%) and 33 (30.3%) cases, respectively. Most adverse effects related to repetition of assessment and delays in management. No patient experienced an adverse medical event as a result of insufficient handover information.
| Doctor or ED affected (n = 62) | Patient affected (n = 33) | |||
|---|---|---|---|---|
| n | % (95% CI) | n | % (95% confidence interval) | |
| Repetition of assessment | 35 | 5.0 (3.5, 6.9) | 16 | 2.3 (1.3, 3.7) |
| Time wasted | 33 | 4.7 (3.3, 6.6) | na | na |
| Delay in disposition | 22 | 3.1 (2.0, 4.8) | 16 | 2.3 (1.3, 3.7) |
| Delay in patient care | 19 | 2.7 (1.7, 4.2) | 13 | 1.8 (1.0, 3.2) |
| Confusion regarding care | 15 | 2.1 (1.2, 3.6) | na | na |
| Adverse medical event | na | na | 0 | 0.0 (0.0, 0.7) |
| Other | 6† | 0.9 (0.4, 1.9) | 3‡ | 0.4 (0.1, 1.3) |
- † Need for further management, investigations and care, delays and confusion with inpatient unit disposition, unable to suture a wound secondary to time delay.
- ‡ Patient agitation from delays, delay in diagnosis, changes to disposition plan. na, not applicable.
Many respondents commented on the problem of ‘secondary’ or multiple handovers of patients who had been in the ED for extended periods. The quality of the secondary handover was frequently flawed when the second doctor often failed to review the patient adequately.
General handover survey
Questionnaires for the general handover survey were completed by 17 of 28 registrars and 33 of 34 consultants invited to participate (overall response rate 80.6%). Table 4 describes information reported to be handed over poorly. Communications made with the patient, relatives, medical, allied health and the GP was reported as a problem by just over half of respondents. Management, disposition, follow-up plans and examination and investigation findings were also reported commonly.
| n (%) | (95% confidence interval) | |
|---|---|---|
| Communications made | 26 (52.0) | (37.6, 66.1) |
| ED management plan | 19 (38.0) | (25.0, 52.8) |
| Disposition plan | 19 (38.0) | (25.0, 52.8) |
| Outstanding patient issues | 19 (38.0) | (25.0, 52.8) |
| Follow-up plan | 16 (32.0) | (19.9, 46.8) |
| Examination findings | 14 (28.0) | (16.7, 42.7) |
| Demographics | 11 (22.0) | (12.0, 36.3) |
| Investigations undertaken or ordered | 11 (22.0) | (12.0, 36.3) |
| Investigation results | 9 (18.0) | (9.1, 31.9) |
| Presenting problem | 9 (18.0) | (9.1, 31.9) |
| Past history | 8 (16.0) | (7.6, 29.7) |
| ED care received | 6 (12.0) | (5.0, 25.0) |
| None | 6 (12.0) | (5.0, 25.0) |
| Other† | 4 (8.0) | (2.6, 20.1) |
- † Issues with inpatient unit communication regarding name and time of person consulted, differential diagnosis and why if diagnosis has not been reached, whether or not general practitioner letter has been written, and that all of the information above may be problems at some times and to some extent are all poorly handed over.
Other perceived problems with handover were inaccurate or incomplete information (30 respondents, 60.0%), disorganization (16, 32.0%), handover too long (8, 16.0%) or too short (4, 8.0%), and confusion (3, 6.0%). The most commonly reported adverse event from a ‘poor’ handover was delay or confusion in communication with the inpatient unit (Table 5).
| n (%) | (95% confidence interval) | |
|---|---|---|
| Communication with inpatient units delayed/confused | 37 (74.0) | (59.4, 84.9) |
| Repetition of ED management | 33 (66.0) | (51.1, 78.4) |
| ED investigations delayed/missed | 30 (60.0) | (45.2, 73.3) |
| ED therapy delayed/missed | 30 (60.0) | (45.2, 73.3) |
| Disposition delayed | 28 (56.0) | (41.4, 69.7) |
| Confusion as to ED management plan | 27 (54.0) | (39.5, 67.9) |
| Communication with patient/relatives delayed/confused | 21 (42.0) | (28.5, 56.7) |
| Follow-up arrangements delayed/incorrect | 18 (36.0) | (23.3, 50.9) |
| Communication with GP/community services delayed/confused | 17 (34.0) | (21.6, 48.9) |
| Arranged beds often not arranged | 1 (2.0) | (0.1, 12.0) |
| None | 2 (4.0) | (0.7, 14.9) |
- GP, general practitioner.
The presenting problem, management, investigation and disposition plans were most commonly reported as being among the minimum information required from handover (Table 6). Ten (20.0%) doctors believed that all items should be included as minimum information.
| n (%) | (95% confidence interval) | |
|---|---|---|
| Presenting problem | 47 (94.0) | (82.5, 98.4) |
| ED management plan | 47 (94.0) | (82.5, 98.4) |
| Investigation results | 43 (86.0) | (72.6, 93.7) |
| Disposition plan | 41 (82.0) | (68.1, 91.0) |
| Investigations undertaken or ordered | 39 (78.0) | (63.7, 88.0) |
| Past history | 36 (72.0) | (57.3, 83.3) |
| Examination findings | 36 (72.0) | (57.3, 83.3) |
| ED care received | 36 (72.0) | (57.3, 83.3) |
| Communications made | 35 (70.0) | (55.2, 81.7) |
| Follow-up plan | 32 (64.0) | (49.1, 76.7) |
| Outstanding patient issues | 30 (60.0) | (45.2, 73.3) |
| Demographics | 14 (28.0) | (16.7, 42.7) |
| Other† | 5 (10.0) | (3.7, 22.6) |
- † Inpatient unit contacts including name and pager number, patient expectations, diagnosis and any problems encountered with the list of information criteria above.
A range of methods was used to remember patient information handed over. Thirty-two (64.0%) respondents used memory, 26 (52.0%) used written notes and 15 (30.0%) used IT (ED computers). No one used a portable device such as a palm pilot. Some doctors used more than one method.
Forty-three (86.0%) respondents believed that handover could be improved, five (10.0%) thought it could not be and two (4.0%) were unsure. Twenty-five (50.0%) suggested that guidelines would make handovers more effective and 21 (42.0%) believed that pro formas to collect minimum patient data would also be effective. The preferred handover methods were a formal, centralized process (28 respondents, 56.0%), one-to-one verbal handover (13, 26.0%), ED bedside round (7, 14.0%), and one-to-one handover using cards (4, 8.0%). Some respondents chose more than one method. Perceptions regarding the use of IT varied. It was thought to be important/very important by 32 (64.0%) respondents andunimportant/very unimportant by 11 (22.0%). Seven (14%) respondents were equivocal.
Discussion
ED handover takes time during which staff are diverted from direct patient contact. Sessions can last almost 1 h, consume a considerable amount of medical resource and have implications for ED efficiency. The challenge lies in the effective transfer of all required patient information in the most time-efficient manner. The present study demonstrates clearly that handover practices at the study sites could be improved. Although data were not specifically collected on the disposition status of patients, it is likely that access block contributed to theresource use required at handover, particularly during the morning handover.
It has been reported that gaps created in personal continuity are compensated by departmental continuity. This is affected via sharing of information and responsibility and the implementation of protocols that support a system of effective communication.1,15 However, the present study shows that deficiencies in personal continuity, through inadequate handovers, do result in adverse effects. Therefore, departmental continuity is unlikely to be a definitive safety net.
The items of information handed over are of interest. Although some items of identifying information were supplied in most cases, it is disconcerting that the patient's name was often not used. Although the presenting complaint was provided in almost all cases, examination and investigation findings and items relating to communication were handed over less frequently. These findings need to be interpreted with care. In many cases, it is likely that items of information were not relevant and therefore not handed over, for example, allergy history. Furthermore, some information might not have been available, for example, investigation results. Therefore, the absolute number of items of information might not be indicative of the quality of a particular handover. This is supported by the lack of association between number of items and the perceived quality of the handovers. Indeed, the handover of a small number of highly relevant items may be more effective than the handover of a larger number of less relevant items. It is possible that some items were not handed over because they were accessible elsewhere, for example, investigations. Whether such items should/should not be handed over is debatable and relates to issues such as the time taken to access these items through various means, whether the items would indeed be accessed if not handed over, and the time efficiency of handover sessions.
The data relating to the perceived minimum information thought necessary for a handover might be useful to inform the development of a handover ‘gold standard’. However, the data obtained in the present study are a list of items only and the proportion of respondents that identified each item as necessary information might not relate to the importance of the item. In the development of a handover ‘gold standard’, consideration should be given to the suitable ‘weighting’ of items, especially if such a standard is to be used to score the quality of a handover.
It could be expected that handover quality would increase with doctor seniority. The reasons for the lack of association with seniority cannot be determined from the study. A number of confounding factors are likely to be exist including the differing types of patients handed over by the groups (e.g. consultants might see more complicated patients) and the time of the day (e.g. consultants rarely handover after night shift).
Although only a small proportion of handovers were considered less than adequate, a greater proportion (15.4%) was thought to lack all the required information. These findings appear inconsistent because a handover lacking in required information could be expected to be less than adequate. It appears that the lack of some information might not have been of such importance as to affect significantly the overall quality of the handover.
The nature of the information not handed over is important. Much information related to details of the patients' ED management although a considerable proportion related to discharge planning and follow up. Bomba and Prakash7 also reported that examination findings and details of referrals or consultations were infrequently handed over (24% and 15%, respectively). In our study, these omissions resulted in repetition, wasted time, delays and confusion. Such effects appear to decrease ED efficiency rather than lead to adverse events. It has been reported that discontinuity of care rarely leads to major management failures.1 We found that in approximately half of cases where necessary information was lacking no adverse effects eventuated. Furthermore, the adverse effects that did result did not cause major consequences. This supports the conclusions of Cook et al.15
Consistent with Roughton and Severs,9 the large majority of doctors thought handover could be improved. Many advocated handover guidelines and/or pro formas, and better handover leadership, education and training, especially for junior doctors. Recommendations for the standardization of handover procedures, especially brevity and relevance, are consistent with findings of others.7
Although the preferred handover format varied among the respondents, the majority preferred a formal centralized process, facilitated by IT. Similar findings have been reported by others.7,9 IT may have considerable potential to improve the quality of handover. An initiative being developed at one of the participating ED is the incorporation into EDIS of a ‘handover screen’. It is notable that information perceived by many to be among the minimum required for handover, is already available on EDIS (demographics, presenting problem, investigations, diagnosis, disposition). This information could automatically populate ‘required information fields’ on a handover screen, as it becomes available. Other fields would need to be completed by the doctor handing over (communications made, management, disposition, follow-up plans), either before or during handover, with each field acting as a prompt for required information. It would provide both documentation of information and avoid the need for memory that is used for almost two-thirds of respondents.
The present study has important limitations. Although the Hawthorne effect might have influenced handover behaviour, this effect would most likely have improved handover quality. Many of the study endpoints were subjective (e.g. quality of handover) and open to varying interpretations. Furthermore, the survey questions might also have been open to interpretation and recall bias. Handover to the night shift staff, when handover tends to be less formal, was not observed and the effect of this variation on the handover quality is unknown. It is possible that a handover was better than it was ultimately perceived to be. Indeed, the required information might have been handed over but not heeded by the receiving doctor. The handover format changed at two of the ED during the study period and this might have affected handover quality. However, the extent and direction of this effect could not be determined. Although the present study examined a large number of handovers given by many ED staff at three different ED, its external validity might be limited if handover processes vary considerably in other ED.
No single handover system is likely to be suitable for all situations.16 As ED differ in physical layout, IT facilities, staffing, patient number and acuity, it is reasonable to expect the format of their handover processes to be tailored accordingly. Nevertheless, we recommend that ED review their handover practice. The findings of the present study suggest that a formal, central process run by a senior ED staff member is generally preferred. The use of IT, handover guidelines and a handover ‘gold standard’ may help to ensure the handover of the minimum information required. Although not examined in the present study, subsequent feedback to a doctor who provides a poor quality handover, especially if the doctor/ED or patient had been inconvenienced, may assist in improving handover quality.
Handover is an area deserving further research. The development of a handover ‘gold standard’ would help to standardize information handed over and, if suitably weighted to account for the importance of items, would assist in handover research including better comparisons between doctor subgroups. Part of the validation of a ‘gold standard’ would be to rate it against perceived handover quality, the amount of required information handed over and adverse events. The adverse effects described in the present study were those perceived by the respondents. A more objective assessment of these effects should be made in order to better determine their nature and extent. The considerable resources required for handover, as measured by doctor-hours, begs the question of whether handover is time-efficient. Indeed, time spent not directly receiving the handover of a particular patient may well be wasted. The amount of time wasted should be determined and the correlation between handover quality and time expended during handover sessions investigated. It may be that a one-to-one handover model is ultimately more efficacious than the traditional centralized model.
Conclusion
A range of deficiencies in handover practice exists, especially in areas of communications, ED management and disposition plans. These deficiencies can result in communication confusion, repetition of management, delays or missed investigations and therapy and delayed disposition. The adoption of a formalized process run by a consultant and making greater use, where possible, of IT facilities may improve handover quality.
Acknowledgements
The present research project was supported by grants from Tattersals Limited and the University of Melbourne, Australia.
Author contributions
DT supervised the study overall. KY, DT, JK and CMacB developed the study protocol. All authors assisted with Ethics Committee applications. KY, DT and AD contributed to data collection. All authors contributed to data interpretation and preparation of the final manuscript. DT undertook the statistical analyses.
Competing interests
David Taylor and Jonathan Knott are Section Editors (Original Research) for Emergency Medicine Australasia.
