One trust has reduced the median length of stay for emergency medical admissions from six to five days. Hugh Rayner shows how it was done
Reducing length of stay is a key part of acute hospitals’ strategies to cope with the increasing number of acute medical admissions. Reductions can be achieved by focusing attention on various stages in the admission and discharge process. But the end result is conventionally described as a single number, the mean length of stay. This has several drawbacks.
The use of the arithmetic mean is only statistically appropriate if data is normally distributed. Unfortunately, the distribution of lengths of acute medical admissions is far from normal (figure 1). As a result, the mean value is very heavily influenced by small numbers of patients who remain in hospital for long periods of time.
Factors influencing the discharge of patients with long lengths of stay are different from those affecting the bulk of medical admissions. At Birmingham Heartlands and Solihull trust, we have investigated the long- stay issues in detail through a collaborative appropriate occupancy study.1
This study showed that the causes of delays in discharge were complex and often not the responsibility of hospitals or social services.
The trust is equally interested in improving the efficient handling of the large numbers of patients who remain in hospital for relatively short periods of time.
To this end, we have used length of stay profiles and the median lengths of stay (ie the number of days by which 50 per cent of patients have been discharged) as performance indicators to compare the acute medical teams within the trust.
Emergency medical patients are admitted to two hospitals in the trust, Birmingham Heartlands and Solihull. The two sites admitted 7,008 and 4,785 patients, respectively, in the financial year 1996-97. One site has 21 consultants, of whom eight are involved in acute medical admissions. On the other site, there are seven consultants who are all involved in medical admissions. Patients are assessed in each hospital’s accident and emergency department and counted as an admission when they are transferred to a ward. Within each site, the medical teams take equal part in a rolling rota. This means that, over the course of a year, the case-mix of emergency medical admissions is equivalent between the teams within each hospital. The two hospitals are not comparable, however, because of differences in the arrangements for the admission of elderly patients.
Each length of stay is allocated to the patient’s admitting team, even if that patient is transferred between teams later in the admission. This has the advantage that lengths of stay refer to spells rather than finished consultant episodes, and the case-mix randomisation described above is preserved.
Handover of responsibility for patients occurs between teams, which tends to equal out the performance of the separate teams. Over recent years, the FCE to admission ratio has averaged 1.10:1 and so this blurring effect is about 10 per cent. Despite this, clear distinctions between teams have been found. There are two main ones. The mode or peak of length of stay is one day for most consultants but for some it is two days. A possible explanation for this difference is variation in discharge planning on post-take rounds between consultants.
The second variation is in the rate of decline following the peak. This is more rapid for some consultants than for others. A possible explanation for this may lie in the efficiency of the team carrying out the management plan and, in particular, not waiting for a formal consultant ward round to sanction a discharge. But these are only speculative explanations.
Examples of the length of stay profiles for different teams are shown in figures 2 and 3. These can be compared with the overall trust profile (figure 1). Interesting differences appear between teams and within teams over time.
Figure 2 shows a team whose profile in 1994-95 showed a peak of patients with a length of stay of seven days. This has progressively reduced so that in 1996-97 the profile showed a smooth decline.
Figure 3 shows a team whose profile in 1995-96 peaked at two days. The 1996-97 profile shows two differences. The total number of patients admitted has increased and, second, the peak length of stay (the mode) is one day rather than two, matching the overall trust profile.
Copies of these graphs and data on activity and mean and median length of stays for all teams are distributed unanonymised within the medical directorate and discussed at the directorate meeting. This was started in June 1996. I took great care to ensure that the data was of high quality and that the underlying assumptions were made clear. I adopted as non- threatening an approach as possible and merely presented the data for consultants to digest as they wished.
There was little adverse reaction and the vast majority of colleagues found the data of great interest as a valid description of their clinical activity. I have not attempted to investigate differences between teams in any detail as I felt this would be regarded as professionally intrusive and threatening.
At subsequent discussions, I was happy to be able to show the improvement that had been achieved by a number of consultants and use this as a means of positive feedback. This naturally improved the acceptability of the data to consultants. This improvement has also been very useful in demonstrating that the efforts of the medical staff in dealing with the increasing number of emergency medical admissions are bringing real improvements in efficiency.
Although other directorates in the trust are comparing length of stay between consultants, I am not aware of them using this graphical technique.
We have found this method of measuring performance to be acceptable to clinicians and believe that focusing attention on these indicators has contributed to the 5 per cent reduction in the mean length of stay (from 11.1 to 10.5 days) for emergency medical admissions that has occurred within the trust over the past 12 months.
The median length of stay has reduced from six to five days. This figure is inevitably a whole number so one cannot interpret it as a percentage reduction.
The mean length of stay for this trust may seem long compared with other acute trusts. This is largely due to the effect of long-staying geriatric admissions. This is another reason for the median and the profile being of particular value.
1 Roberts P, Houghton M. In Search of a block buster. Health Service J 1996; 106(5532): 28-9.
Hugh Rayner is clinical director of acute medicine, Heartlands Hospital, Birmingham Heartlands and Solihull trust.