The long-term rise in emergency admissions, and seasonal ups and downs, continue to produce winter pressure crises and adversely affect elective work.1, 2 The problem of large differences in admission rates between hospitals is equally important in its implications for resource use and outcomes.3, 4 Both these problems have beset the NHS for many years and are still not understood in sufficient detail to provide policy solutions.5
An ageing population and changes in levels of disease in the community have been shown to have relatively minor effects on the long-term rise in admissions and on differences in hospital admission rates.5, 6 Factors such as relative supply of hospital resources, differences in clinicians' practice, changing medical technology and patient expectations, and differences between hospitals in their admission processes may be significantly more important.2, 3, 5
All these factors combined produce differences in admission severity thresholds between hospitals.3, 5 Thresholds are the general levels of severity of illness and of need for medical treatment in hospitals above which patients are normally admitted.
Increasing the number of beds as a response to rising admission pressures may only be a short-term panacea. As a more practical long-term solution, local demand management policies are probably needed. These policies require an adequate knowledge of the admission thresholds of local hospitals. But remarkably few studies have tried to measure thresholds.
Six hospitals in west London, Ealing, Hammersmith and Hounslow health authority and the London Health Economics Consortium collaborated in an exploratory study whose aims were to measure admission thresholds and to examine how they are affected by clinical management and admission processes.
To measure patients' admission thresholds, 1,516 adult emergency admissions from specialties other than obstetrics, gynaecology and psychiatry were surveyed in the six hospitals over three weeks from 26 January 1998. Experienced nurses abstracted medical and nursing data from patient case notes. Care was taken to ensure good data quality and comparability between the study hospitals. Extensive testing found high levels of reliability.
The intensity, severity and discharge (ISD) utilisation review instrument, currently used in the US, was used as the framework. The ISD instrument has been approved for use in Britain. Data was collected on severity of illness - specific signs, symptoms, imaging and laboratory findings; and intensity of treatment - specific medical and nursing treatments and interventions the patient receives on admission.
The ISD instrument measures patients' need for acute hospital admission against guidelines developed by doctors. An admission is deemed to meet the ISD guidelines if the patient has specified combinations of both severity of illness and intensity of treatment.
We used two indicators derived from the data collected to measure admission thresholds. First, we used the percentage of study patients in each hospital whose admission met the ISD guidelines on need for admission. A high percentage indicated a high admission threshold.
Second, we derived our own admission threshold scores from the patients' severity of illness and intensity of treatment, which were assessed and given a score by a physician from one of the study hospitals. The hospital threshold scores were calculated as the mean of the threshold scores of individual patients. A high score indicated a high threshold.
We interviewed A&E consultants, consultant physicians, clinical nurse managers, nurse practitioners, business or directorate managers and bed managers at each of the hospitals.
The impact of clinical management
We found significant differences between the hospitals on both threshold indicators. For the first indicator, the high-threshold hospitals had around 80 per cent of study patients meeting the ISD admission guidelines - nearly 20 per cent higher than for the low-threshold hospitals (see figure 1). These differences were statistically significant.
For the second indicator, based on severity of illness and intensity of treatment, the high-threshold hospitals had scores of 11 or more, more than 20 per cent higher than the low-threshold hospitals, which had scores of around nine (see figure 2). These differences were also statistically significant. The hospital rankings for the two threshold indicators were similar.
Only a small part of the threshold differences between the hospitals was due to case-mix differences. And relatively few patients not meeting ISD admission guidelines were admitted for social or nursing care only. But several clinical management factors did contribute to significantly lower thresholds. Admission for investigations which could have been undertaken in outpatients was common among patients not meeting the ISD guidelines. Had such outpatient investigations been available, admission might sometimes have been avoided. Some examples included Doppler scanning for deep-vein thrombosis, chest x-ray, abdominal ultrasound and angiograms.
Access to treatment was also important. Some patients not meeting the ISD guidelines had treatment which could have been offered without admission. These included patients with a long-standing history of anaemia who were admitted for blood transfusions, and admission of nursing home residents for intravenous fluids and/or antibiotics.
One interesting finding was that although some patients had the necessary severity of illness to meet the ISD admission guidelines, they did not receive enough specified treatment elements to meet the ISD guidelines. Sometimes this occurred when a terminally ill patient was receiving palliative care only. But sometimes this situation highlighted variations in the management of certain conditions. For example, patients suffering from asthma, with apparently similar severity of illness levels, received varying intensities of treatment.
We ranked the hospitals in terms of the above factors. We then checked whether hospitals which ranked highly on a factor also had high admission thresholds. Only the availability of alternative-to-admission services seemed to be connected with thresholds to any extent. Where these alternative services were available seven days a week - for longer hours - and were comprehensive and easily accessible, there were, in general, higher admission thresholds.
We found little link between thresholds and the other factors, such as the seniority of the medical team member making the decision to admit, and the extent of medical registrars' other commitments. We also found very little link between thresholds and the use of assessment or admission management protocols and variations in bed management systems.
There were also issues of comparability of data recording between hospitals. There were differences in the way an admission was defined. How long does a trolley wait have to be before being recorded as an admission? Some potentially useful data, such as time of admission, was frequently missing from medical records. There was also no consistency in the way the hospitals recorded bed availability. Either no computerised record was kept of the number of beds available each day, or the computerised statistics were inaccurate.
It is very important to continue to improve admission management and we are carrying out follow-up work.
The case notes of comparatively low severity admissions are being examined by some hospitals to establish whether alternative ways of management which avoid admission can be found. The case notes of patients with similar problems are being examined to see how far clinical management differs between hospitals. One hospital, which is changing the functions of its A&E department, is going to measure thresholds again when the changes have taken effect. We are working with the study hospitals to try to measure bed availability in a comparable way.
Hospitals differ markedly in their admission thresholds. Such differences have important implications for the use of resources. If all our study hospitals had had the admission thresholds of the highest (around 80 per cent of patients meeting the ISD guidelines), then about 7 per cent of admissions would not have occurred. This would have made these beds available for more severely ill patients at times of extreme pressures, representing a more cost-effective use of existing resources.
The vast majority of patients who were below the 80 per cent admission threshold required medical or nursing care in a non-inpatient setting. But if alternative facilities are not available, setting them up may not be cheaper than using existing hospital beds, though it might be more cost-effective. We suggest that routine measurement of thresholds is essential to provide an evidence base for demand-management policies and to monitor progress.
Although the comparative validity of our results is high, the threshold indicators we have used are experimental, and further development of measurement methods is needed. British guidelines on need for medical admission should be drawn up.
There appear to be few obvious links between the differences in hospital admission thresholds and differences in admission processes. But analysing these differences can be useful in making proper assessment and admission more efficient. Hospitals need more sophisticated monitoring of demand.
Further studies are necessary to explore the links between admission thresholds and bed availability, and between admission rates and community alternatives to hospital admission.