Main findingsAll patients

Our study finds a substantial proportion of patients who had been referred with a prescription for admission by their PCPs. One-third of referred patients and almost two-thirds of all other patients were discharged from the ED. Only a small minority of referred patients were transported by EMS. Most referred patients arrived on their own, presenting as walk-ins, which may indicate that they were not deemed to be under immediate threat by their referring PCP. A substantial number of ED resources was used in both referred and unreferred patients, and referred patients were almost double as likely to receive an ultrasound examination in the ED. They also were older, more often presented during office hours, had more comorbidities, and were less likely to have presented more than 3 times during the previous year. The proportion of nursing home residents was lower among referred patients, and they were rarely admitted. A large proportion of nursing home residents presented by patient transport for minor interventions such as wound dressing, diagnostics after falls, or changing urinary catheters. There were no differences in triage urgency, the number of abnormal vital parameters, or pain scale among referred and unreferred patients.

Walk-ins

Among walk-ins, who were overall younger and had less comorbidity than patients presenting by EMS or PTS, referred patients were older and had more comorbidities. Referred walk-ins had less often parenteral medication in the ED but used more diagnostic resources than walk-ins who were not referred.

Also, referred walk-ins were more likely to receive an ultrasound examination and more often underwent laboratory examinations and computed tomography than self-presenting walk-ins.

Patients discharged from ED

Among all patients discharged from ED, there were no relevant differences in age, comorbidity, vital parameters, or MTS triage level between referred and unreferred patients, but there was a tendency to more ultrasound examinations and parenteral medications in referred patients.

Predictors of admission

In multivariable analysis of all patients, the strongest predictor of hospital admission was referral by a PCP. Living in a nursing home and an ultrasound examination in the ED were associated with a lower likelihood of admission.

In the group of walk-ins, referral remained the strongest predictor of admission (Table B, data supplement), and ultrasound remained negatively associated with admission.

If referral as a walk-in is proof of a low estimated acuity or severity, other unaccounted medical parameters may have been the reason to refer via the ED. Higher age and more comorbidity in referred versus self-presenting walk-ins but not in referred versus self-presenting patients discharged from ED may hint to the possibility that comorbidity and age rather than acute problems were important reasons for admission. Also, the decision to discharge patients from ED may often have been dependent on diagnostic resources of the ED as suggested by the high proportion patients discharged from ED after ECG, imaging, or laboratory investigations. Bureaucratic and other barriers to scheduled admission may also have played a role.

Use of ED resources

Interestingly, in all analyzed groups of referred patients, ultrasound examinations were more common than in patients who were not referred. Referral was statistically associated with ultrasound examinations after controlling for trauma, MTS category, abnormal vital parameters, and admission (Table C, data supplement).

Laboratory investigations were also overrepresented among referred patients. In our EDs, these are often ordered in a standardized fashion according to presenting symptoms and may not reflect individual needs precisely. In referred patients, trauma presentations were rare, and the larger proportion of patients with laboratory investigations may correspond to a larger proportion of medical conditions and more comorbidity.

Comparison with the literaturePatient itineraries and alternatives to ED visits

There are few studies describing patient itineraries from GP-type contacts to hospital EDs. Gries et al. found 7.7% among ED presentations to a German university hospital ED were referred patients, and 44.7% were self-referrals. OR for admission was 2.2 after referral by a PCP [19]. In that study, as in ours, trauma patients were rarely referred by a PCP. In a multicenter questionnaire study among walk-ins, 17% had been referred by their primary care physician and 8% by specialist referrals. We did not discriminate between these both groups of referring physicians because there were only few specialist referrals, and most of these were passing through referrals by PCP to EDs [5].

In times of widespread ED overcrowding, there is a focus on streaming patients with inappropriate presentations or PCP-substitutable conditions away from hospital EDs. Villareal et al. described a service model with telephone triage and involvement of PCP. Patients with face-to -face contact were more likely to be transferred to an ED than those with telephone contact only [13]. Recent studies from Sweden showed that more than half of patients presenting to a university level ED had been in contact with a PCP. A total of 81% stated they had been advised to visit the ED. Shorter symptom duration was a predictor of direct presentation to the ED [27]. Unfortunately, in our study, symptom duration was not documented. A nationwide study including all Swedish EDs found 13% of ED presentations being referred of which 27% were admitted [14]. The higher admission rate in our referred ED patients may be partially due to the fact that in Germany, referrals from PCP to EDs are always prescriptions for admission.

Barriers to boarding and use of ED resources

Barriers to planned boarding may have instigated PCPs to use EDs as access points to obtain hospital beds for patients who needed inhospital treatment but did not need ED interventions or acute care. The low triage level, and above all the fact, that most patients were not sent by EMS may hint in that direction. On the other hand, hospital resources were used to a large extent in these patients in the ED. A recent investigation of referrals from urgent care centers to hospital EDs in the USA found that advanced imaging studies in 40.7% (ultrasound 16%, CT 24.7%) and specialist consults were important reasons for ED referrals, whereas 55% of presentations did not receive any ED-specific treatment [28]. We believe, a likely explanation would be that in patients with acute symptoms, often the decision to admit requires extensive use of diagnostic means. This is the only study we identified that examined PCP — ED patient journeys with a focus on the use of ED-type resources. In this line, in our study, ultrasound was overrepresented in referrals and had a negative impact on the odds of admission. Computed tomography was more common in referrals than in non-referrals among walk-ins. From German insurance data, it has been hypothesized that increased availability of ultrasound, wound-care, and laboratory investigations in the ASHIP sector could improve out-patient emergency care [17] and reduce ED presentations. Our data regarding ultrasound would be consistent with this assumption.

Care during off-office hours and accessibility of hospital EDs

While a study from the Netherlands assumes that most patients during office hours visit their own PCP [29], we find a majority of walk-ins with low acuity visiting the ED without previous contact with their GP. Other studies have estimated that between 20 and 40% of patients in an ED were eligible for care by a GP [30]. Introduction of general practice cooperatives (GPC), where primary care cases can be seen during off-office hours, in Dutch EDs has resulted in a reduction of ED presentations by 30% and an increase of referrals by GPs or GPCs [31]. The rate of admissions from EDs also increased but was still comparatively low in relation to our figures. A study from Vienna also found a decrease of ED presentations after introduction of a primary care cooperative alongside the ED [32]. In Belgium, no change of ED case load has been found after installing a regional off-hours general practitioner cooperative [33]. Minderhout et al. surveyed patients’ motives to visit EDs instead of their GP and found that important issues were the subjective impression of severity of their condition but also the expectation that hospital infrastructure such as laboratory or radiological investigations might be needed [29]. In our cohort, the use of laboratory investigations and imaging was common even among discharged patients, suggesting that this perception of patients may be accurate. However, in referred patients, use of ultrasound and computed tomography was even higher, suggesting that this may have been the reason for referral, and that the decision of admission or discharge may have been dependent on this imaging.

Special populations

Referred patients in our cohort were older than direct ED presenters and were more co-morbid. Older patients have been found to be less likely to self-present with low-acuity conditions than younger patients [6]. This may be due to closer patient-physician relationships in older patients with increased medical needs [34, 35]. The proportion of frequent attenders in our cohort was similar to publications from the USA, Canada, and the UK [36,37,38]. Frequent users of the ED were rarely sent by their PCP, suggesting that in these cases the ED was rather used as a substitute for primary care than as an adjunct.

Nursing home residents were much less likely to present after contact with a PCP, and they were often cared for on an outpatient basis with little use of hospital resources. We suspect that some nursing home residents lack adequate access to primary care and are sent to emergency rooms often for simple medical assessment or simple measures as has been previously published [10, 39, 40]. In Germany, according to present regulations, this may be partly due to lacking reimbursement for patient transportation to ASHIP practices [41].

Wound care was underrepresented in referred patients. In Germany, PCP come from different professional backgrounds, but a substantial proportion of PCP do offer wound care and also suture lacerations. We were unable to obtain explicit data on this issue and did not identify any published evidence on this subject. It may be that most minor wound care is carried out by PCP, or that injured patients may primarily visit an ED.

Limitations

This retrospective study has several limitations; whereas the selection of our random sample of a large number of consecutive patients during 1 year will not be biased, referrals and self-presentation to the two participating hospitals may have been influenced by regional factors and the specific services provided by both hospitals. It has been shown that PCP access varies dramatically throughout Germany [42]. Also, selection between EDs or PCP, respectively, will be influenced by socioeconomic factors that we could not account for. Our EDs serve two Berlin boroughs with a metropolitan and suburban setting and about 500,000 inhabitants with a below-average socio-economic status. Generalizations should be made with caution.

Sample size was planned based on rule-of-thumb estimates to allow for up to 15 predictors of events occurring in 10% of cases [26, 43]. This may be disputed as may be our selection of parameters included in multivariable analysis that was partly based on group differences, univariable analysis, known predictors of admission, and interest. Whereas we think risk of overfitting will be small, relevant parameters may have been overlooked. Results should be interpreted as preliminary and validated in other studies. Furthermore, our group comparisons are at risk of multiple testing [44]. This should be kept in mind while interpreting the results and planning further studies on the subject.

Except for 58 patients who were not triaged and who — due to the structure of our electronic admission protocols — also missed transport data, data completeness was good with the following caveats: CCI was calculated from the information contained in patients’ documents, but not collected prospectively from patients. It is probable that relevant information was missing, and it is possible that this underreporting disproportionately affected patients with limited access to healthcare and patients discharged from ED. Vital parameters were taken on an “as-needed” basis and were omitted in many patients who were judged to be stable, for example, those with minor injuries. These patients were counted as having no abnormal vital parameters.

Apart from referral notes, we had no information on the PCPs motives for referral. Further study of the expectations and motives behind informal modes of cooperation across sectoral barriers in the German health system may be of interest and should include PCPs.

Clinical and policy implications

It appears that instead of a merely complementary operation of ASHIPs in the ambulatory sector, on the one hand, and hospital EDs in the in-patient sector, on the other hand — as codified in the strict regulatory separation between the sectors in Germany — in everyday life, PCP find more cooperative modes of care for their patients, thus overcoming sectoral borders and making use of ED infrastructure such as ultrasound or computed tomography for out-patient care. It is unclear if this is cost efficient. Better availability of ultrasound in primary care might help to reduce presentations to emergency departments. Provision of specialized nursing or home visits to nursing home residents may also reduce ED attendance.