Abstract
Accurate prediction of the likelihood of same-day discharge could make it possible to direct one-third of the medical in-take to an ambulatory care unit, thereby facilitating bed management. In Phase 1 of this study, we identified seven independent factors that contribute to an ambulatory care score (Amb score) that can potentially be used as a tool to select ambulatory emergency care (AEC) patients from the medical emergency in-take. A high score was associated with discharge within 12 hours of assessment and treatment in hospital. In Phase 2, we verified and internally validated the performance of the Amb score in a different cohort of patients, finding that it functioned well in identifying early discharges (ie AEC patients), with an area under the receiver operator curve (AUROC) of 0.91 (95% CI 0.88–0.94). An Amb score of ≥5 has a sensitivity of 96% (95% CI 90–98) and a specificity of 62% (95% CI 55–68) in identifying potential AEC patients.
Background
The number of emergency hospital admissions and their costs continue to increase annually.1 It is appropriate that all emergency referrals from primary care should be assessed in hospital, but admission is not always necessary. Acute medical units (AMU) help to reduce length of stay and prevent unnecessary admissions.2 In addition, ambulatory care units contribute to admission-avoidance: many medical emergencies can be managed in the ambulatory care setting provided diagnostic services and assessment facilities are readily available.3,4
Notwithstanding the diagnosis, some conditions do not warrant in-patient management; indeed, up to a third of referrals seen in AMU are discharged on the same day.5,6 This constitutes a significant proportion of patients that could potentially be assessed within the ambulatory emergency care setting.
The RCP Acute Medicine Task Force7 defines ambulatory emergency care (AEC) as ‘the clinical care which may include diagnosis, observation, treatment and rehabilitation, not provided within the traditional hospital bed base or within the traditional outpatient services that can be provided across the primary–secondary care interface. In the context of acute medicine, it is the care of a condition that is perceived either by the patient or by the referring practitioner as urgent, and that requires prompt clinical assessment undertaken by a competent clinical decision maker.’ AEC patients tend to fall into one of four categories: a diagnostic exclusion group in which certain conditions, such as a possible deep venous thrombosis, must be ruled out; a low-risk group, such as those with community-acquired pneumonia with a low CURB65 score; a specific diagnostic group, who require a procedure such as pleural effusion; and finally, the group that requires a treatment that has traditionally been provided in an inpatient department, such as those with low-risk pulmonary embolism.
In hospitals where AEC patients are managed in a geographically separate unit from the general emergency medical in-take, it can be difficult for the GP or the primary care nurse to identify patients who should be referred directly to the AEC unit without a full clinical assessment and further investigation. Clinical scoring systems are currently used to gauge illness severity and mortality risks8–11 when assessing acutely unwell patients, but they are not used to assess the suitability of an emergency-referred patient for ambulatory care management.
A simple scoring system that can predict the likelihood of same-day discharge after assessment would be useful in selecting patients who fall into the AEC category. The referring primary care clinician, or the emergency medicine (EM) and AMU triage staff, should be able to calculate this score before a full clinical assessment and subsequent investigations. Hence, the score must be based on the patient's clinical, demographic and social parameters.
Being able to give the patient information on the likelihood that they will be admitted is of benefit, allowing them to make appropriate arrangements for their home life, for their work commitments or with care agencies. For the hospital, this information could be useful for bed-management purposes.
This study was undertaken in two phases—Phase 1 (derivation) and Phase 2 (validation). Factors that would be useful in predicting the likelihood of early discharge (within 12 hours) were identified in Phase 1; these were then used to derive a simple ambulatory care score, the Amb Score. 12 In Phase 2, which involved a different cohort of patients, this score was verified, and its ability to identify early discharges from the unselected general medical emergency in-take was internally validated.
Methods
This study was undertaken at The Royal Glamorgan Hospital, a 570-bed district general hospital serving a largely rural population in the South Wales valleys. With a 28-bed AMU, it admits between 30 and 50 patients each day, although more than a third of these are discharged within 12 hours. Written consent was obtained from all patients who took part in this study prior to data collection. The NHS Research Ethics Committee (REC) advised that this service-development project did not require ethical approval. A total of 625 patients were recruited, of whom 282 (45%) were studied in the derivation phase (Phase 1) and 343 (55%) in the validation phase (Phase 2).
Phase 1: derivation
This phase identified factors that might be useful in assisting the referring doctor or the hospital triage staff in determining the likelihood that a emergency medical patient would be discharged from hospital within 12 hours of assessment.
Out of 612 medical emergencies who were referred by their GP, 282 were randomly selected to be studied retrospectively. The study sample was divided into patients who were discharged within 12 hours of hospital assessment (the ambulatory group) and those who were admitted for ≥48 hours (the admission group). A patient who was discharged within 12 hours of assessment was assumed to have a condition that could potentially be treated in an AEC unit. Social, demographical and clinical variables that could be associated with an early discharge (≤12 hours) were identified. A hospital stay of 48 hours or longer was arbitrarily counted as an admission, so follow up was for 48 hours. Those who died or who discharged themselves against medical advice, and those who stayed in hospital for 12–48 hours were excluded. In this phase, referrals from the EM department were excluded.
This phase was undertaken over a four-week period in May and June 2010. We used the earliest full set of recordings, be they from the accompanying GP letter, the ambulance notes, or the first set of observations after arrival in hospital, to calculate the Modified Early Warning Score (MEWS). Table 1 shows the MEWS scoring parameters used in following our local trust protocol. Information on other variables was extracted from the medical and nursing notes and by talking directly with the patients or relatives.
Modified early warning score (MEWS) used by Cwm Taf local health board.
Analysis was performed using Epi Info Version 3.5.1 (CDC Atlanta 2008). Continuous variables were presented as means with standard deviation (SD) and chi-square was used in 2×2 tables to compare categorical data. Odds ratios with 95% confidence interval (CI) are given. Logistics regression was used to identify independent variables, and the adjusted odds ratios are quoted. We regarded a p<0.05 as statistically significant.
The seven independent factors that were associated with an early discharge (≤12 hours) were then used to formulate an ambulatory care score, termed the Amb score; it was postulated that a high Amb score might be useful in identifying potential AEC patients from amongst the unselected general medical emergency in-take.
Phase 2: internal validation
This phase was undertaken prospectively over a three-week period in June and July 2011. The main aims were to verify factors used to calculate the Amb score that were derived in Phase 1, and to validate the score in a different cohort of patients by comparing the Amb scores for patients who were discharged within 12 hours (the ambulatory group) with those of patients who were admitted for ≥ 48 hours (the admission group). The Phase 2 study cohort included both GP and EM referrals, but the other criteria for exclusion remain the same as in Phase 1.
A sample size of 103 patients in each group was suggested (STATCALC) as adequate to identify differences between the two groups powered at 80% at a 95% level of confidence. In fact, a total of 343 patients who fulfilled the criteria were randomly selected from 1,015 emergency referrals in the three-week period. An Amb score was calculated for each patient on the basis of information obtained from the patient or their relatives, and from the clinical notes (GP letters, ambulance sheets, EM notes, and nursing and medical records).
Validation was carried out by applying the Amb score to the data from this phase. Chi-square analysis was used to verify the Amb score variables, and the area under the receiver operator curve (AUROC) was used to assess the effectiveness of the score in identifying those who were discharged within 12 hours (ie the potential AEC patients). An appropriate cut-off level for the optimal Amb score sensitivity and specificity was identified from the AUROC analysis.
Results
Phase 1: derivation
Data were collected on 282 patients who fulfilled the criteria. On average, the males (36.2%) were older (mean age 67, SD 16) than the females (mean age 63, SD 19), but not significantly so. About half of those in the study sample (50.7%) were discharged within 12 hours of assessment (the ambulatory group). Of those who remained for ≥48 hours (the admission group), 42.2% were not assigned to a specialty team, 24.3% stayed under the care of the elderly team and 11.2% went to the respiratory team.
The significant factors identified by bivariate analysis that were associated with patients who had an early discharge (ie potential AEC patients) are listed in Table 2. Access to transportation, having good family support, being female and being younger than 80 years old were significantly associated with discharge within 12 hours. A previous history of common medical problems, such as coronary artery disease or chronic lung diseases, was not shown to affect length of stay, neither was a possible diagnosis of significant bleeding or acute coronary syndrome. If intravenous treatment was anticipated, however, the patient was more likely to be admitted for ≥ 48 hours.
Variables studied in the derivation study (Phase 1).
A normal temperature (35–37.5˚C), a systolic blood pressure of 100–200 mmHg and a peripheral oxygen saturation of ≥93% on air (or on oxygen if the patient was on home oxygen) were significantly associated with an early discharge (≤12 hours). A MEWS score of ≥2 was significantly associated with admission for ≥48 hours. Conversely, a MEWS score of zero was significantly associated with a discharge within 12 hours.
A hospital discharge within the previous seven days had no effect on the length of stay but those who had been discharged within the previous 30 days were more likely to stay for ≥48 hours.
Multiple logistics regression analysis returned seven independent variables that determine whether a patient was discharged within 12 hours or admitted for ≥48 hours (Table 3). These seven variables were used to formulate the Amb score (Table 4). The weighting of each parameter was chosen to reflect the odds ratio (OR) of being discharged within 12 hours from the derivation data, such that those with an OR in favour of early discharge were given a higher value. The two parameters with an OR in favour of admission for ≥48 hours (male gender and age >80) were given negative values. Having access to transportation, not being acutely confused and not requiring IV treatment were associated with the greatest odds of being discharged within 12 hours, so these were given a higher weighting in the Amb score calculation. This gave a total Amb score out of 8. Patients with a high Amb score were more likely to be discharged within 12 hours following assessment, so these could be treated as potential AEC patients. When the score was applied to the derivation data, a score of 6 or above corresponded to an increased likelihood of being discharged within 12 hours (OR 0.3).
Independent variables in derivation study (Phase 1).
Calculating the Amb score.
Phase 2: internal validation
In this phase, 40% of the referrals were from GPs; sexes were equally represented equally (50.4% females). There were no significant gender differences in age (mean age 66, SD 20).
Approximately one-third (35.5%) of the total study sample in this phase was discharged within 12 hours. This group had a mean Amb score of 7.1 (SD 1.1), compared with a mean Amb score of 4.2 (SD 1.8) in the admission group (p<0.01) (Table 5).
Verification of Amb score parameters in the validation study (Phase 2).
Bivariate analysis comparing each of seven Amb score parameters in the two groups showed a high level of statistical significance, confirming the discriminatory ability of these variables (Table 5). The area under the receiver operator curve (AUROC) for the Amb score was 0.91 (95% CI, 0.88–0.94) (Fig 1). A cut-off point between an Amb score 4 and 5 was chosen as this gave the most favourable sensitivity of 96% (95% CI, 90–98) without compromising much of the 62% specificity (95% CI, 55–68). The specificity reflects the proportion of patients that the score had correctly predicted would be admitted for ≥48 hours. It is lower than the sensitivity because some of these patients were discharged within 12 hours. The score would have been less specific if the cut-off threshold were reduced and it would lose sensitivity at a higher cut-off.
Table 6 shows the top five diagnoses in each group in the Phase 2 cohort. Deep venous thrombosis (DVT) was the most frequent diagnosis in the ambulatory group, whereas exacerbation of chronic obstructive pulmonary disease (COPD) was the most frequent reason for admission for ≥48 hours.
Top five diagnoses in the validation cohert (Phase 2).
Discussion
This was a single-centred study aiming to identify factors that might be useful in predicting discharges within 12 hours of hospital assessment (ie potential AEC patients). These factors were then used to derive a simple scoring system (Amb score) that could be used to select out these potential AEC patients from the general medical emergency in-take. It is postulated that the score would be helpful to patients by informing the likelihood of same-day discharge, and for the hospital front-door assessment and bed management teams by facilitating in-take management. Patients who were discharged within 12 hours were chosen as surrogates for potential ambulatory care patients because specialised AEC units tend to stay open for 8–12 hours each day, and because most AEC patients who are incorporated into the general medical emergency in-take are discharged within 12 hours of arrival in hospital.
Being younger (<80 years) and female were factors that were associated with being discharged within 12 hours of assessment. In this semi-rural setting with a significant coal mining legacy, the women probably enjoy a better state of physical health than the men. Whether this gender difference in the length of stay is similar in other areas in the UK remains to be seen, although gender differences in healthcare-seeking behaviour have been reported previously in other countries,13,14 although which gender utilises healthcare systems more depends on where the studies were undertaken.
Public transport services in the South Wales valleys can be variable,15 so if a patient had no access to personal or public transport, they had to rely on the Welsh Ambulance Service. This might have delayed their discharge. Access to personal transport is a more crucial factor in determining early discharge than having good family support. In areas in the UK with better-resourced public transport services, transportation might not be such a significant factor in early discharges.
Intravenous (IV) treatments, such as antibiotics for cellulitis, is mostly an in-patient service in this region, so the need or otherwise for IV treatment is a significant factor in determining length of hospital stay. Many other trusts have a well-established outpatient IV antibiotic service, so this factor might not be significant in other regions, although patients who require other IV fluids or emergency blood transfusion might still need to be admitted.
Often, the cause of acute onset confusion is not initially apparent, and it is good clinical practice to treat any possible precipitating factors while closely monitoring the patient. Such patients tended to be kept in for ≥48 hours.
A MEWS score of 0 can be useful in deciding that a patient should be managed in an ambulatory care setting. In keeping with previous studies,8,9 we have shown that a high MEWS score is associated with a longer hospital stay of at least 48 hours.
Patients who had been discharged within the previous 30 days tended to be readmitted for ≥48 hours, although not necessarily with the same diagnosis. Anecdotally, these readmissions were mostly due to difficulties in the social support services rather than a deterioration in the patient's clinical condition, so in areas with excellent community support services, this might not be such a significant factor.
In this study, we have identified some factors that can assist the referring doctor, those that receive the calls, and the EM and AMU triage staff in considering the suitability of ambulatory care management for some patients who would normally have been incorporated into the unselected general medical emergency in-take. The resulting Amb score is simple and can be calculated prior to a full clinical assessment or investigations (other than the measurements required calculate the MEWS score). The validation data show that the score performs well when an appropriate cut-off level, which has a high sensitivity at identifying potential AEC patients without significantly compromising its specificity, is selected.
Clinical factors often affect decisions to refer emergency cases to hospital but not all such cases require admission. Provided these patients are assessed as clinically stable after initial treatment in hospital, and provided they have appropriate social support, they can be discharged for further management within the ambulatory care setting. In the correct context, the Amb Score could be a useful guide to such patient management. It can help GPs to predict the likelihood of same-day discharge and it could help bed managers in planning the daily general medical emergency in-take and in avoiding over-crowding in the AMU or EM unit. Potential AEC patients could be re-directed to a designated ambulatory care unit if one exists or to a specialty-based rapid access out-patient clinic. The Amb Score could also be useful to identify general medical in-take patients who need early review by a senior doctor (specialty doctor or consultant), thereby preventing admission and unnecessary investigation on the AMU.
With bed shortages and increasing emergency referrals to hospital on the one hand and more accessible diagnostic facilities and specialty-based rapid access clinics16,17 on the other, there is an opportunity for the expansion of AEC services. Tools such as the Amb score could assist in these developments.
There are of course pitfalls in using such a score: for example, a patient with cardiac-risk factors, presenting with cardiac sounding chest pains and an abnormal electrocardiogram, should normally be admitted, although he might have an Amb score of 8; conversely, a 90-year-old male with a MEWS score of 1, a recent hospital discharge and no access to transportation might have an Amb score of 4 but could still be suitable for early discharge if he had a minor problem such a urinary tract infection. Ultimately, a decision that is based on a quick initial clinical assessment by a competent healthcare professional takes precedence over the score, which should only be used as a guide.
Conclusions
The Amb score is a novel and simple test that is sensitive in predicting discharge within 12 hours of hospital assessment, and thus might select out potential AEC patients from the unselected general medical emergency in-take before full assessment and further investigations are undertaken. The score was derived from a small study, from one hospital in a semi-rural setting. Further studies are required to validate the usefulness of the Amb score in other regions and in both primary and secondary care.
- © 2012 Royal College of Physicians
References
Jump to section
Related Articles
Cited By...
- Process of care and activity in a clinically inclusive ambulatory emergency care unit: progressive effect over time on clinical outcomes and acute medical admissions
- Using computer simulation to model the expansion needs of the ambulatory emergency care unit at Derriford Hospital
- Sepsis recognition tools in acute ambulatory care: associations with process of care and clinical outcomes in a service evaluation of an Emergency Multidisciplinary Unit in Oxfordshire
- Comparison of Glasgow Admission Prediction Score and Amb Score in predicting need for inpatient care
- Ambulatory emergency care - improvement by design
- Ambulatory emergency care: how should acute generalists manage risk in undifferentiated illness?
- Testing the AMB score - can it distinguish patients who are suitable for ambulatory care?
- The ambulatory care unit at Derriford Hospital