Strategies to identify medical patients suitable for management through same-day emergency care services: A systematic review

Same-day emergency care (SDEC) in unplanned and emergency care is an NHS England (NHSE) priority. Optimal use of these services requires rapid identification of suitable patients. NHSE suggests the use of one tool for this purpose. This systematic review compares studies that evaluate the performance of selection tools for SDEC pathways. Nine studies met the inclusion criteria. Three scores were evaluated: the Amb score (seven studies), Glasgow Admission Prediction Score (GAPS) (six studies) and Sydney Triage to Admission Risk Tool (START) (two studies). There was heterogeneity in the populations assessed, exclusion criteria used and definitions used for SDEC suitability, with proportions of patients deemed ‘suitable’ for SDEC ranging from 20 to 80%. Reported score sensitivity and specificity ranged between 18–99% and 10–89%. Score performance could not be compared due to heterogeneity between studies. No studies assessed clinical implementation. The current evidence to support the use of a specific tool for SDEC is limited and requires further evaluation.


Introduction
Effective delivery of acute care services requires rapid assessment and management of patients across emergency departments (ED) and acute medicine services, identifying those that can be safely discharged from hospital as well as those at risk of deterioration.
In the UK, same-day emergency care (SDEC), previously known as ambulatory emergency care (AEC), provides assessment and management of patients without overnight admission to an inpatient hospital bed. 1 Consistent provision of SDEC has repeatedly been prioritised within NHS strategy, 2 , 3 to reduce demand on inpatient services and avoid risks associated with inpatient admission. 4or unplanned medical attendances, SDEC is predominantly delivered within acute medicine services, 5 accessed via referral from primary care and emergency medicine (EM) services. 1Acute medicine services vary considerably between hospitals nationally, including in the proportion of unplanned medical attendances assessed within SDEC, and the proportion discharged without overnight admission. 6ot all patients are suitable for SDEC, for example patients requiring urgent stabilisation or treatment only available as an inpatient; additionally, SDEC services often have limited capacity due to physical or workforce limitations.For efficient service delivery that maintains patient experience, only patients suitable for assessment within SDEC should be reviewed through this pathway.Using tools to help identify suitable patients, such as the Amb score, 7 is recommended, but there is limited understanding of clinical performance of these scoring systems nationally.
This systematic review aimed to identify strategies (including scoring systems and selection criteria) currently used to identify medical admissions suitable for management within SDEC services, and to compare the populations in which they have been derived/validated and their ability to correctly identify suitable patients.

Methods
The systematic review protocol was prepared using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline 8 and registered in the PROSPERO registry of systematic reviews (registration number CRD42022351082).Table 1 shows the inclusion and exclusion criteria.There were no language or publication date restrictions.https://doi.org/10.1016/j.clinme.2024 In the UK, acute medical problems requiring secondary care assessment are managed through acute medicine services, with access via referral from other services.This is most commonly via referral from EM, following either initial triage or clinical assessment, but a proportion of patients are referred directly from primary care or community services, including general practitioners (GPs).Acute medicine services span multiple locations, with acute medical units (AMUs) providing care for patients requiring inpatient admission for at least one night, and SDEC delivering assessment, investigation and management of those requiring internal medicine services, but expected to be discharged without overnight admission. 6Studies that did not include acute presentations suitable for assessment by general or acute internal medicine clinicians were excluded.SDEC attendances are not classed as a hospital admission. 9Where studies described identification of patients for admission versus discharge in an acute medical setting, 'discharge' was assumed to include 'same day' management and these studies were included.Where studies described identification of patients for admission versus discharge in EDs, these were included if decision to discharge was equated with suitability for SDEC; studies were excluded if any patients with lengths of stay less than 12 h were classified as admitted.

Study selection
Identified articles were exported into EndNote (Clarivate Analytics) for removal of duplicates.Abstracts were screened blindly and independently by two authors (CA and HH/CH/RK) within Rayyan (Qatar Foundation, Qatar) . 10Disagreements were resolved by full-text review with a third reviewer (JB).
After initial screening, two authors (CA and JB) reviewed full articles using predefined inclusion and exclusion criteria; disagreements were resolved by discussion with a third reviewer.
Data extraction using a standardised form (Supplementary Table 1) was performed by one reviewer (CA).
Risk of bias was assessed by one reviewer (CA), using the Prediction model Risk of Bias Assessment Tool (PROBAST). 11

Outcome measures
The primary outcome was selection tool performance in identifying patients suitable for management through SDEC (eg patients not requiring admission to an inpatient acute hospital bed).This included sensitivity, specificity, positive and negative predictive values of the tool in selecting suitable patients, and accuracy as assessed by the percentage of patients correctly identified, calculated as the proportion of true positives and true negatives from the sample population.
Secondary outcomes were readmission rates (unplanned reattendance within 7 or 30 days of discharge) and mortality (including 30-day mortality).

Study selection
Initial searches identified 8,036 abstracts.After removing duplicates, 7,205 abstracts were screened, with 89 selected for full-text screening.Nine met the inclusion criteria ( Fig. 1 ).
Two studies described derivation and validation of a score, the Glasgow Admission Prediction Score (GAPS) , 12 and the Amb score (Ambs). 13Two studies assessed the Ambs alone. 14 , 15One study compared performance of the GAPS with triage nurse judgement. 16Two studies compared the Ambs and the GAPS, 17 , 18 and two compared the Ambs, GAPS and the Sydney Triage to Admission Risk Tool (START). 19 , 20These scores as described in their original derivation studies are shown in Table 2 .
Table 3 outlines the characteristics of the included studies.Seven studies were conducted within the UK, 12-14 , 16-19 one in Italy 20 and one in Malta. 15One study was available as conference abstract only. 19The number of patients included ranged from 54 to 322,846 (median 1,424 patients).
Five studies evaluated score performance in only medical patients, 13-15 , 17 including one study that enrolled only a subset of medical patients within a specialist frailty service. 19One study assessed performance in a cohort combining patients presenting directly to acute medicine services, and patients presenting to the ED. 12 Three studies evaluated patients within the ED, 16 , 18 , 20 including but not restricted to medical patients.
No identified studies assessed implementation of a scoring system into clinical practice.

Risk of bias and applicability
All included studies were assessed using the PROBAST tool ( Table 4 ). 11Six studies were at risk of bias; however, this was limited to specific concerns within a maximum of two domains. 12 , 15-18 , 20n four studies, the inclusion/exclusion criteria may have altered score performance compared to the intended clinical application, with three studies excluding patients requiring resuscitation or intensive care review, 15 , 16 , 18 and one study not enrolling patients at times of increased service pressure. 20One study was at risk of bias in evaluation of predictors as the score component of 'need for intravenous antibiotics' (included in the Ambs) was assessed by a varied process that included decisions made later than where the score would be used in clinical practice. 18In three studies, it was unclear whether there was risk of bias in how predictors were assessed; in two studies, the method to determine 'need for intravenous antibiotics' was not described. 13 , 14Two studies had risk of bias in analysis, 15 with differences in the proportion of missing data (and therefore participants included in analysis) when comparing scoring systems in one study. 17here were concerns regarding applicability in eight studies. 12 , 13 , 15-20This related to the included participants in five studies; four studies included ED attendances not restricted to medical patients, 12 , 16 , 18 , 20 and one study focused only on patients with frailty-Fig.1.The article review process.PRISMA diagram for the systematic review.related conditions. 19Six studies had applicability concerns regarding outcome; only two studies assessed the outcome of same-day discharge or discharge within 12 h compared to length of stay over 12 h, 14 , 19 the remainder compared 'admission' without specified timeframe compared to discharge from hospital, 12 , 16 , 18 , 20 length of stay < 12 h compared to 12-48 h, 17 < 12 h compared to > 48 h, 13 , 18 or < 24 h compared to > 24 h. 15

Results of individual studies
Results of the individual studies included in the review are shown in Table 5 .

Score performance in medical patients alone Score derivation
Ala et al derived and internally validated the Ambs to predict likelihood of discharge within 12 h versus admission for over 48 h, at a single centre in Wales. 13Patients admitted for 12-48 h were excluded.Initial derivation included only patients referred directly from general practice; validation also included referrals from EM. Scores ≥ 5 identified patients likely to be discharged within 12 h, with sensitivity 96% and specificity 62%; PPV and NPV were not reported, and could not be calculated from the published data.

Single score validation
Thompson calculated the Ambs retrospectively for 200 patients in a single acute medicine service. 14Limited information on patient demographics and selection process was reported; admissions via GP and ED were included.Diagnostic accuracy measures were reported for performance of the Ambs discriminating patients admitted for < 12 h from those admitted for > 12 h; 67.8% of patients were correctly identified.39% with an Ambs suggesting suitability for SDEC were discharged in < 12 h.Subgroup analysis restricted to those arriving between 09:00 and 17:00 demonstrated similar performance.Performance comparing ED and GP referrals was not reported.
Dimech et al calculated the Ambs for medical patients admitted to a single hospital in Malta, using information from inpatient records. 15heir primary aim was not to evaluate performance of the Ambs, but to identify the proportion of admissions potentially suitable for management through ambulatory pathways.Length of stay < 24 h was used to indicate potential suitability for ambulatory management.Twenty percent of admissions were discharged within 24 h, of which 90% had an Ambs ≥ 5.An Ambs ≥ 5 was also found in 32.5% of those admitted for > 24 h.Measures of diagnostic accuracy were not reported, but were calculated from the information provided.

Comparison between scores
Atkin et al retrospectively evaluated performance of the Ambs and GAPS to identify patients suitable for SDEC, indicated by discharge within 12 h of hospital arrival. 17Analysis was restricted to patients arriving between 08:00 and 16:59 and discharged within 48 h.This single centre study evaluated the scores as originally described, and when substituting NEWS2 for MEWS/NEWS.The Ambs identified 57% of patients correctly; 55% with an Ambs suggesting SDEC suitability were discharged within 12 h.Ambs incorporating NEWS2 had similar per-

Table 2
Scores evaluated in the included studies within this systematic review.Amb score 13 Glasgow Admission Prediction Score (GAPS) Score components as specified during original score derivation.MEWS: Modified Early Warning Score; NEWS: National Early Warning Score; IV: intravenous; GP: general practitioner; ED: Emergency Department; ENT: Ear, nose and throat.* GAPS uses Manchester triaging system, START uses Australasian Triage Scale formance (56% accuracy; 54.5% of 'suitable' patients admitted < 12 h).Using a binary cut-off, the GAPS identified 57.5% of patients correctly; 51.4% with a GAPS suggesting SDEC suitability were discharged within 12 h.Incorporating NEWS2 had similar performance (accuracy 57.5%; 50.5% of 'suitable' patients admitted for < 12 h).Although overall accuracy was similar for the Ambs and GAPS, the measures of diagnostic accuracy were significantly different ( Table 5 ).
Burgess et al considered score performance in a single-centre retrospective analysis of patients seen through an ED-based SDEC-equivalent frailty service, restricted to adults aged > 70 with 'frailty syndromes'. 19ver 75% of included patients were discharged the same day.Comparison of the Ambs, GAPS and START in this cohort of selected frail patients demonstrated poor performance of all scores, with sensitivity of 18%, 25% and 26% respectively, suggesting that the scores did not correctly identify the majority of patients discharged without inpatient admission.

Score derivation
The GAPS was derived by Cameron et al in 2015. 12Retrospective data for all attendances to six units (three EDs, two medical assessment units and one minor injuries unit) were included, with a binary outcome of admission or discharge.Length of stay was not reported.A binary cutoff of more than 15 was used to indicate likely admission, predicting outcome with 80.3% accuracy overall, but the authors recommended adjusting cut-offs to local populations.Although the cohort included admissions directly to medical assessment units, score performance in this subgroup, or in medical patients specifically, was not reported.

Single score validation
Cameron et al compared performance of the GAPS with triage nurse prediction of admission versus discharge in patients presenting to ED triage. 16Score cut-off was adjusted to maximise sensitivity and specificity in the enrolled cohort.Triage nurse clinical judgement of likelihood of admission had similar accuracy to the GAPS.

Comparison between scores
Cameron et al (2018) compared performance of the GAPS and Ambs for patients triaged within the ED in a prospective study at two urban teaching hospitals. 18The Ambs as used in this study precedes the published validation (Supplementary Table 1).Score performance was assessed in predicting length of stay > 48 h compared to discharge within 12 h, with optimum cut-offs to maximise sensitivity and specificity of   Risk of bias and applicability assessed using the PROBAST tool. 11Classified in each domain as high risk ( + ), low risk (-) or unclear risk.
GAPS < 18 and Ambs > 5. Cut-offs that maximised the proportion of correct predictions were GAPS < 20 (74.2% correct, 95%CI 71.9-76.5%)and Ambs > 5 (70.5% correct, 95%CI 68.1-72.9%).The AUROCs for score performance in predicting admission from the ED (vs discharge), and admission for > 48 h (versus < 48 h) were assessed ( Table 5 ).Score performance in medical patients specifically was not reported.Salvato et al compared performance of the Ambs, GAPS and START in identifying patients presenting to the ED that did not require admission, with score components recorded at ED triage and scores subsequently calculated by the research team. 20The triage scoring system used locally was mapped to that used in the GAPS and START.In their cohort, the Ambs performed better overall than the GAPS and START (assessed by AUROC) despite lower sensitivity.All three scores were outperformed by clinical judgement of trained triage nurses.Inclusion was not restricted to medical patients, and the proportion of patients presenting with medical problems or requiring admission to internal medicine, or score performance in these patient groups, was not provided.

Additional outcomes
Readmission rates were assessed in two studies ( Table 5 ). 12 , 17The original GAPS study showed a positive relationship between GAPS and likelihood of subsequent hospital admission in those initially discharged from the ED. 12 Atkin et al found an lower rate of readmission within 7 and 28 days in those with Ambs or GAPS suggesting suitability for SDEC. 17ortality was not reported in any of the included studies.

Meta-analysis
Meta-analysis of the included studies could not be performed due to study heterogeneity.

Discussion
This review describes the published evidence for criteria available to identify medical admissions suitable for SDEC.The included studies described and evaluated three scores: the Ambs, 13 GAPS 12 and START. 22hese scoring systems vary in their components, but have some common factors, namely that increasing age and previous hospital admission reduce likelihood of same day discharge.Comparison of performance between scoring systems, or within different populations, was limited by variation in score definition, inclusion criteria, and outcome definitions.
The populations assessed within the studies varied considerably, with only one study including only medical admissions without exclusion criteria based on length of stay. 14The remainder enrolled medical patients within a larger cohort of ED admissions, 12 , 16 , 18 , 20 medical admissions to a specialty SDEC service, 19 or medical admissions meeting specified length of stay criteria. 13 , 15 , 17This will impact score performance, as the proportion of patients requiring hospital admission is different in diverse patient cohorts.Similarly, scores derivation and later assessment took place in different types of hospital, for example the Amb score was derived in a more rural setting, which likely has different logistical considerations when assessing suitability for SDEC care.The proportion of patients suitable for SDEC management varied considerably in the cohorts within the included studies, ranging from 20% to 80%. 14 , 20 The outcome predicted by the scores in the included studies also varied.The GAPS and START aim to identify patients requiring admission compared to those that can be discharged from hospital, 12 , 22 while the Ambs was intended to identify patients with length of stay less than 12 h, making them suitable for selection to an SDEC service. 13In the UK, SDEC attendances are not classified as an inpatient admission, and as such evaluations of scores that assessed a decision to admit versus discharge were included where the authors equated this to a decision that patients were suitable for management through SDEC. 9The START score was developed and validated in Australia and included patients admitted to observation wards with length of stay of 4-24 h classed as 'admitted'. 21 , 22These studies therefore did not meet our inclusion criteria, as patients with lengths of stay suitable for SDEC were not differentiated from patients requiring admission.Similarly, Raita et al constructed machine learning models predicting the risk of hospitalisation at ED triage using retrospective data in the USA, 23 but patients with medical presentations and a length of stay < 24 h were included in the admitted patient cohort.
Comparing performance of the Ambs is also limited by changes in score definition used (Supplementary Table 2).Updated early warning scores have been substituted into the Ambs in some cases, accounting for changes in clinical practice. 24Substitution of score components may impact score performance, necessitating appropriate validation, particularly as NEWS2 incorporates acute confusion, which is itself a component of the Ambs.A preliminary version of the Ambs, used in one included study, 18 was presented as a conference abstract prior to the Ambs derivation and validation discussed here. 13This variation suggests there may be variation in the 'Ambs' used in clinical practice, which may not have been appropriately validated.
Several studies restricted analysis to those patients where it was felt that scoring systems may be most useful as an adjunct to clinical judgement, where there is greatest uncertainty regarding need for hospital admission, by excluding patients within resuscitation or minor injuries areas. 16 , 18Score performance in these studies may underestimate performance if applied across the whole patient cohort.
Where score performance was compared to clinical judgement, clinician judgement was as or more accurate. 16,20This suggests that the currently suggested scoring systems may not improve on judgement from appropriately experienced clinicians.
The search criteria for this review aimed to identify relevant studies describing a process used in the ED or acute medical setting to identify patients suitable for review within SDEC.It is possible that articles discussing ED disposition were not identified; however, no rel-

Table 5
Results of the individual studies included in the systematic review.evant articles were identified on hand-searching of included articles' references.
Selection criteria for specific conditions were excluded from this review.Some conditions commonly managed through SDEC services, such as pulmonary embolism, have established criteria for ambulatory management. 25Studies assessing only patients with a diagnosis or suspected diagnosis of specific disease were excluded, as clinical judgement would be needed to determine the specific condition.
These results suggest that there has been no robust validation of performance of scoring systems to select medical patients suitable for assessment and management within SDEC.The variation in approach seen in the definitions of suitability for SDEC/ambulatory management suggests a lack of consensus regarding which patients are suitable for management within SDEC.Further evaluation is needed to understand the settings where scoring systems would be most beneficial, and could be practically incorporated, the cohort of patients these scores should aim to identify, and the acceptable accuracy of any scoring system.Current policy relies on a limited evidence-base; generating the in-depth understanding necessary to rapidly inform national policy will require a multi-centre study of factors assessing SDEC suitability, involving hospital sites across diverse settings.Components of the scoring systems previously evaluated should be assessed, including physiological observations and demographic factors, alongside other features potentially available at initial triage that could be incorporated into simple tools, such as mobility, functional measures, or symptom categories, with performance compared across settings.Assessment of performance should consider a wider range of clinically relevant outcomes, including unplanned readmission to hospital, alongside likely impact on staffing and infrastructure.

Conclusion
Scoring systems (the Ambs, GAPS and START) have been suggested to identify patients suitable for management through SDEC services; however, evidence validating performance of these scores outside the setting in which they were derived is limited.There is considerable heterogeneity in how these scores have been evaluated, including in the definition of patients considered suitable for management through SDEC.

Table 1
.100230 Received 20 March 2024; Received in revised form 16 July 2024; Accepted 16 July 2024 1470-2118/© 2024 The Author(s).Published by Elsevier Ltd on behalf of Royal College of Physicians.This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ) Inclusion and exclusion criteria for the systematic review.Database searches were conducted in December 2023.The detailed search strategy is included in Supplementary Information.Search terms included both SDEC and AEC to reflect variation in naming conventions for SDEC-equivalent services.The search strategy was not designed to identify studies describing condition-specific ambulatory pathways, therefore these studies were excluded.Databases searched included the Cochrane Database of Systematic Reviews, MEDLINE (via Ovid), Embase (via Ovid), CINAHL Plus (via EBSCO), MEDLINE in Process (via Ovid), Cochrane Central Register of Controlled Trials (CENTRAL), PsycInfo (via Ovid), Healthcare Management Information Consortium database and Web of Science.Reference lists of included publications were also handsearched.
21Sydney Triage to Admission Risk Tool (START) score21

Table 3
Characteristics of studies included in the systematic review $ .

Table 4
Assessment of risk of bias and applicability of included studies.

Table 5 (
continued )Not reported PPV: positive predictive value; NPV: negative predictive value; AUROC: area under receiver operating characteristic curve.* denotes measures calculated from data provided in study.ˆto aid comparison, outcomes switched where reporting here to facilitate comparison of diagnostic accuracy measures.