Emergency department referral patterns of Australian general practitioner registrars: a cross-sectional analysis of prevalence, nature and associations

Additional keywords: emergency medicine, family practice, general practice, physician practice patterns, referral and consultation.

Up to 8% of emergency department (ED) presentations are referred from general practice.¹ Compared with general practice, ED offers expertise and equipment more closely targeted to the management of acute serious illness. However, for low-acuity morbidities not requiring this intensive level of acute care, presentation to the ED generates avoidable healthcare costs.² Furthermore, excessive ED demand can lead to compromised health care outcomes.³ Thus, the decision of general practitioners (GPs) to refer acute presentations to the ED rather than manage them themselves (in the community) demands careful consideration of patient safety and care, as well as health resource use.

Determining whether to refer a patient for specialised management involves a complex interplay of factors related to the individual GP, specialist, practice, patient and their relatives.^4,5 Particularly for acute presentations, this may occur within a time-pressured environment. Considerable inter-GP differences have been observed in referral rates to the ED,⁶ with some research exploring links between this variation and referral ‘appropriateness’,⁵ with appropriateness to be determined by the potential urgency or severity of the presentation rather than the final diagnosis within ED.⁷ In Australia, the Australian College of Emergency Medicine (ACEM) has determined that poor GP decision making leading to inappropriate referrals is not a substantive factor in ED ‘overcrowding’.⁸ Nevertheless, there is strong opinion that improved understanding of the epidemiology of ED referral patterns of Australian GPs can inform GP-relevant strategies to address ED demand: for example, training, feedback and peer support of appropriate GP ED referral decision making have been advocated.⁹

A segment of the GP population of particular interest with regard to referral behaviour is that of GP registrars (trainees). Registrars are at a formative stage of their clinical careers, establishing potentially lifelong practice patterns. Limited international studies indicate that ED referral rates of registrars¹⁰ and referral rates of other early career GPs¹¹ are significantly higher than those of longer-established GPs. However, Australian studies, and any studies establishing associations of registrars’ ED referrals, are lacking.

The aim of the present study was to address this evidence gap by establishing the prevalence, nature and associations of Australian GP registrar ED referrals.

The present study was a cross-sectional analysis of the in-practice clinical experiences of GP registrars participating in the Registrar Clinical Encounters in Training (ReCEnT) study. The ReCEnT study has been described in detail elsewhere.¹² Briefly, this ongoing multisite cohort study captures clinical consultation data of GP registrars from five of Australia’s 17 general practice regional training providers (RTPs) across five states and encompassing urban, rural, remote and very remote practices.

Registrars in general practice training posts operate within an apprenticeship-like model but with considerable autonomy, including billing, prescribing, test ordering and referral capabilities equivalent to their more experienced GP peers. They do have recourse to advice and assistance at any time from a senior GP supervisor (or another GP within the practice as their delegate).

The present analysis involves data from 12 data collection rounds between 2010 and 2015.

ReCEnT data collection is undertaken as an integral educational component of each participant’s GP training program.^13,14 Registrars may also provide optional informed consent for their data to be made available for research purposes. Registrars complete three 6-month (full-time equivalent) compulsory general practice training terms. Midway through each term, registrars record (via paper-based patient encounter forms) demographic, clinical and educational data from 60 consecutive, office-based general practice consultations. Prior to each data collection round, each registrar provides demographic data of themselves and their current practice. Some registrars at one RTP also completed data collection during a fourth optional GP term.

The outcome factor in the present study was whether, for a diagnosis or problem managed (hereafter ‘problem’), the registrar made a referral to an ED. Non-ED hospital referrals were excluded from the analysis.

Other variables in the present analysis related to the registrar, patient, practice and consultation. Registrar factors were age, gender, training term, having worked at the practice before that training term, place of medical qualification (Australia or international) and full-time/part-time status. Patient factors were age, gender, Aboriginal and Torres Strait Islander status, non-English speaking background status, new patient to the practice and new patient to the registrar. Practice factors included RTP, rurality, socioeconomic status of the practice location, practice size (i.e. the number of GPs) and whether the practice routinely bulk-billed (i.e. the patient incurred no financial cost for the consultation). Practice postcode was used to define the Australian Standard Geographical Classification–Remoteness Area (ASGC-RA) classification¹⁵ (degree of rurality) and the Socioeconomic Index for Areas (SEIFA) Index of Disadvantage¹⁶ of the practice location. The consultation factors were consultation duration, number of problems, the problem being a chronic disease, whether the problem was new to the registrar and whether the registrar ordered pathology, imaging or follow-up. Educational consultation factors included whether the registrar sought clinical information or assistance for the problem (from their supervisor, a specialist, other health professional or from electronic or hard-copy resources) and whether the registrar generated learning goals pertaining to the problem.

Problems were coded according to the International Classification of Primary Care, 2nd edition (ICPC-2) classification system.¹⁷ Chronic diseases were coded via an existing classification system derived from the ICPC-2.¹⁸

For descriptive purposes, individual ICPC-2 codes were grouped with clinically congruent codes to create clinically meaningful problem categories. This grouping was performed collaboratively by two members of the research team (PM, NC), one a GP.

The present study was a cross-sectional analysis of data from the longitudinal ReCEnT study. The proportion of problems for which the registrar made an ED referral was calculated with 95% confidence intervals (CIs), accounting for clustering of repeated measures of registrars.

Of those problems for which the registrar sought in-consultation information or assistance, the proportion of information or assistance sought from the registrar’s supervisor was calculated with 95% CIs, also accounting for clustering.

Analyses were conducted at the level of the problem rather than consultation, because registrars make ED referrals for individual problems.

Simple and multiple logistic regressions were performed with the dependent variable ‘registrar made an ED referral’. All independent variables with P < 0.20 in univariate analysis were included in the multiple regression model. Covariates with a small effect size and no longer significant (P < 0.05) in the multivariable model were tested for removal from the model. If no substantive change was observed to the resulting model, the covariate was removed from the final model. The logistic regression was within a generalised estimating equations framework to account for the repeated measures on registrars. The significance level for all analyses was set at two-tailed P < 0.05. Analyses were performed using Stata v13.1 (StataCorp, College Station, TX, USA) and SAS v9.4 (SAS Institute, Cary, NC, USA).

The ReCEnT study has approval from the Human Research Ethics Committee, University of Newcastle (Reference H-2009-0323).

In all, 1161 registrars (response rate 95.5%) contributed data on 166 966 consultations, entailing management of 258 381 problems. The characteristics of participating registrars and their practices are given in Table 1.

Table 1.  Demographics of participating registrars and their practices
Data are presented as n (%) or as the mean ± s.d. GPs, general practitioners; SEIFA, Socioeconomic Index for Areas – Index of Disadvantage¹⁵

Of all problems, 0.5% (95% CI 0.4–0.5) were referred to ED. The most common reasons for referral were chest pain, abdominal pain and fractures. These made up nearly 25% of ED referrals. The 12 most common reasons for referral (see Table 2) comprised 44.4% of all ED referrals.

Table 2.  Problems most frequently referred to the emergency department (ED) by general practitioner registrars
LRTIs, lower respiratory tract infections; SAH, subarachnoid haemorrhage

Univariate associations of ED referrals are presented in Table 3 and the results of univariate and multivariable logistic regression analyses are presented in Table 4. Significant patient factor multivariable associations with making an ED referral were: patient age 0–14 years (odds ratio (OR) 1.48, 95% CI 1.18–1.86), 35–64 years (OR 1.28; 95% CI 1.06–1.56) and ≥65 years (OR 2.21; 95% CI 1.73–2.82), compared with reference age 15–34 years, and the patient being new to the registrar (OR 1.34; 95%CI 1.13–1.59). Registrars practising in outer regional, remote or very remote areas were significantly less likely to make an ED referral compared with those within major cities (OR 0.64; 95% CI 0.44–0.94). A registrar being enrolled with one particular RTP was significantly associated with making an ED-referral (OR 1.78 (95% CI 1.39–2.28) for RTP 4 compared with RTP 1). Key consultation factors significantly associated with making an ED referral included the problem being new (OR 3.14; 95% CI 2.59–3.79), the registrar seeking information or assistance from another health professional or an electronic or hard copy resource (OR 2.77; 95% CI 2.27–3.38) and the registrar generating learning goals relating to the ED-referred problem (OR 1.35; 95% CI 1.09–1.66).

Table 3.  Characteristics associated with emergency department (ED) referrals
Unless indicated otherwise, data are presented as n (%). NESB, non-English-speaking background; RTP, regional training provider; SIEFA, Socioeconomic Index for Areas – Index of Disadvantage;¹⁵ FTE, full-time equivalent

Table 4.  Simple and multiple logistic regression analyses of associations with emergency department (ED) referrals
OR, odds ratio; CI, confidence interval; NESB, non-English-speaking background; RTP, regional training provider; FTE, full-time equivalent

In-consultation information and/or assistance was sought by registrars for the diagnosis and/or management of 45.5% (95% CI 42.7–48.4%) of those problems subsequently referred to the ED. The registrar’s supervisor (or another GP in the practice) was the source of information and/or assistance for 65.8% (95% CI 61.9–69.5%) of all sources of information or assistance sought for all ED-referred problems. Table 5 presents sources used by registrars.

Table 5.  Sources of information, assistance or advice sought by registrars for emergency department (ED)-referred problems

GP registrars referred their patients to ED at a frequency of 5 per 1000 problems seen, almost twice the rate of that of established GPs in an earlier Australian study,¹⁹ and comparable to the findings of a previous, albeit limited and non-contemporary, study of UK registrars.¹⁰ A likely contributing factor to the relatively high proportion of patients referred in the present study is the common delegation within Australian general practice of acute or unscheduled presentations to a practice’s registrars. This is reflected in the significant association of ED referrals with patients being ‘new to the registrar’. This finding is also consistent with limited patient continuity of care found in Australian GP registrars’ practice.²⁰ Lack of familiarity with individual patient contexts of these new presentations may lead to lower thresholds for referral.^10,21–23

Clinical experience has been found to build greater certainty in a clinician’s decision making.²³ Conversely, uncertainty is a known contributor to GP referral rates,^4,5,24 including those to the ED.^9,25 Higher levels of clinical uncertainty and lower tolerance of uncertainty in these early career GPs may be expected to further contribute to lower thresholds for referral than more experienced GPs. However, registrar seniority (training term) was not associated with making fewer ED referrals in the present analyses. Temporal inferences are limited from the present cross-sectional study, and other unmeasured confounders (such as allocation of more serious presentations to more senior registrars) may explain our observations. It is also plausible that registrars’ uncertainty in emergency presentations may not attenuate appreciably during the relatively short course of Australian GP vocational training (minimum 18 months within clinical general practice). This may be so because, compared with many other aspects of community-based practice, registrars are likely to have had ample exposure to ED patients and acute serious illness during prevocational training ED experience (having completed at least one ED term is a requirement for undertaking a general practice term). Levels of uncertainty regarding acute serious presentations may be relatively modest compared with less serious, but less familiar, conditions, even for Term 1 registrars.

The significant association of patients under 15 years of age (compared with those aged 15–34 years) being referred to the ED by GP registrars mirrors referral patterns of established GPs (albeit at higher rates).²⁶ This may reflect lower registrar confidence in managing paediatric patients, a consequence of reduced exposure to paediatric consultations during vocational training²⁷ coupled with the pressure of concerned parents’ expectations.^28,29 This finding may also reflect a higher proportion of paediatric presentations being of high acuity compared with those of 15- to 34-year-old patients.³⁰

The association of older age with ED referrals is consistent with the lower physiological reserve and greater health care complexity of this age group compared with the reference age group.²¹ The documented trend of younger adults using the ED as a general practice substitute³¹ and ‘bypassing’ general practice by ED self-referral could also be contributing to these results.

After adjusting for possible confounders (including rurality), we found that a significant difference in ED referral rates remained between registrars of two RTPs. Although inter-RTP differences in hospital accessibility (associated with greater ED admission rates²¹) may not be accounted for fully by ASGC-RA classification, this difference in ED referral rates may also reflect regional ‘cultural’ variability, both in general practice (including RTP-provided education) and in EDs.

The significant associations of ED-referred problems with being new problems, having fewer problems managed within the consultation, follow-up not being arranged and fewer tests ordered or medications prescribed were expected for high-acuity presentations requiring specialised ED management, with most of these observations also seen for established GPs.¹⁹

The significant association of ED referrals with the seeking of in-consultation information or assistance reflects the seriousness of these presentations and that the patient would still likely have been referred to the ED if seen by a senior GP rather than by the registrar. Assistance or information was sought primarily from a registrar’s supervisor or another doctor in the practice (65.8% of all assistance sought). This is consistent with previous evidence in the registrar population that suggests that in-consultation information seeking is preferentially from human as opposed to non-human (hard-copy or online) resources for more complex or challenging clinical scenarios.³²

The finding that registrars were more likely to generate learning goals from ED-referred problems indicates these were true learning experiences for the registrars, reflecting vocational training best-practice and self-directed learning rather than simple delegations of responsibility to ED colleagues. Furthermore, although expectedly fewer tests were ordered and medications prescribed in these acute presentations, demonstrable prehospital management by registrars was observed (pathology ordered in 12% of ED-referred cases and medications prescribed or administered in 25% of such cases). The statistically significant increase in consultation duration, despite fewer problems being managed, may be related to the seriousness of the problem being managed, but may also reflect registrars’ active management (as opposed to purely expedient triage and discharge to the ED).

We found a striking difference in ED referrals in outer regional, remote and very remote regions compared with major cities (OR 0.64; 95% CI 0.44–0.94). This result should be treated with caution because it may reflect structural factors (see below).

The strengths of the study included good generalisability to Australian GP vocational training given the involvement of five of 17 RTPs in five states across all rurality classifications from major city to very remote, and the markedly high response rate for GP studies.³³ The large number of variables measured in the ReCEnT study allowed for adjustment for multiple potential confounding covariates.

A basic limitation of the present study is its cross-sectional nature and that causality cannot be inferred in the associations we found. Other limitations included the lack of data on the ‘appropriateness’ of referrals. We also lack data on the ‘outcomes’ of the referrals, including diagnostic and therapeutic outcomes and hospital admission. In addition, a scenario not addressed in our analysis was that of rural hospitals where local GPs have admitting rights for in-patient management and can thus admit patients directly, bypassing the ED altogether. This could account for our finding of ED referrals being less prevalent in outer regional, remote and very remote regions compared with major cities. Quantification of these direct referrals was beyond the scope of the present study.

The marked variability in referral prevalence between two RTPs found in the present study is a suitable subject for further research because it will have important educational implications. Future research could also investigate the prevalence and associations of ‘inappropriate’ referrals and outcomes of referral (e.g. hospital admission, within-ED investigation and management, and patient satisfaction).

We have established ED referral patterns of GP registrars within five Australian RTPs, with referral rates nearly twice those of more experienced GPs. This difference may be due largely to GP practice structural factors of patient appointment scheduling. Our results of an association of ED referral with the patient being new to the registrar suggests that continuity of care in general practice may be of importance, even in acute presentations. Further associations established in our analysis highlight the learning opportunities provided to the registrars by these acute presentations. Observed inter-RTP variability in referral patterns warrants further research.

None declared.