Implications for policy and practice
The continuous reduction of length of stay is all the more remarkable considering two main developments with an increasing effect on the average clinical length of stay:
Since the eighties of the last century many hospitals have introduced day-care and have increasingly substituted (short-term) clinical admissions for day-care [9, 10].
Another development which had an increasing effect on the average length of stay is the ageing of the patient population. In 1978, 19% of the admissions were 65 years or older. In 2006, this increased to 48%. On average, elderly people stay longer in hospitals than younger ones; in 2006 the 0–64-year-old patient stayed an average 5.2 days in hospital and the patients aged more than 64 years stayed an average of 9.1 days.
In spite of these two developments the average length of stay decreased from year to year. We expect this to continue because in the coming years, the financing system in Dutch hospitals will more and more be based on market forces and the reimbursement through payments per diem will be abolished (as in the United States more than two decades ago ). The increased competition among hospitals will increase interest in length of stay reduction in order to increase capacity for additional admissions and improve financial performance.
Limitations of the study
Chance of underestimation
The potential reduction in length of stay may in fact be higher because of two methodological choices.
First, we have chosen to use a 1997 list of treatments that could have been performed in day care. This list could have been longer if we had used more recent data as a reference. Currently, we are planning to update the list. Probably a new list will show more possibilities to substitute inpatient care into day-care. Until now, the health care system in the Netherlands gave only few incentives to treat patients in day-care. Updating the list at this moment will also give an underestimation of the possibilities for day-care. We think that, when the changes in the financing system have been carried out entirely, an update will clearly show more possibilities for day-care.
Second, in our standardisation for patient mix, the expected length of stay was not used for patients with a length of stay of 100 hospital days and longer and for patients who died in hospital. For these two groups the realised length of stay was used instead of the expected length of stay. This means that the results are without the potential gain in efficiency for these two groups. However, it concerns a small number of patients. Only 0.1% of all patients had a length of stay of 100 hospital days and longer and 2.4% of all patients died in hospital.
Specialty as a variable for length of stay
The variation in the quotients of actual length of stay and expected length of stay shows that for several specialties the mean score is not 1. This is the case especially for cardiothoracic surgery and for 'other specialties'. For these two specialties it is 'normal' that the quotient of actual and expected length of stay is higher than 1.0. For 'other specialties' it is known that many hospitals created a special ward for patients that could not be discharged in time to next care facilities like nursing homes. The length of stay of these patients was longer because of these waiting days and the hospitals booked for these patients an administrative transfer to 'other specialties'. The code 'other specialties' is also used for geriatrics. This specialty treats patients that may have the same age group, diagnosis- and procedure group as patients treated by other specialty, but often the patients treated by geriatrics have a more complex syndrome and stay longer in hospital because of their frailty. The variables for standardisation (age group, diagnosis- and procedure group) do not seem to be sufficient for patients that are discharged by these two specialties. The variable 'specialty' should also been taken into account. Because we did our analysis for each separate specialty this was no problem for this study, but if length of stay is benchmarked on the level of hospitals, 'specialty' is a variable that should be taken into account.
Lack of data based on severity of illness
For a large part of the data, adjustment for age, primary diagnosis and procedure amounts to an adjustment for severity of illness. However, we realise that there may still be residual case-mix related variation that is not adjusted for. We did not adjust for variations in comorbidities Neither did we account for variations between elective versus emergency cases. Both parameters were recorded in the LMR, but the completeness of the registration of these items varies between hospitals. We realise that the presence or absence of a large number of comorbidities and/or emergency cases at hospital level will affect overall length of stay of a particular hospital. However, this potential residual variation that is not adjusted for is one of the reasons why we used the 15th percentile as benchmark and not the minimum. If a more sophisticated comparison data based on severity of illness were available, it would be possible to identify which subpopulations (younger, older, diagnosis, procedure, long stay, short stay) were generating the largest numbers of excess days. This could be possible in the future because the Dutch hospital information system will be upgraded in 2010.
Perspectives for future research
Length of stay is often used as an indicator of efficiency [6, 11–13]. Efficiency can be described as the relationship between input and output. From a hospital perspective a length of stay reduction may increase efficiency by increasing the output (number of patients) or decreasing the inputs (e.g. available beds for inpatient care). Both may be realised by reducing 'waiting'-days during a hospital stay or by minimising time between examinations, consultations and procedures. However, if the reduction in lengths of stay results in increased intensity of care (and consequently cost) the efficiency improvement may be smaller. In addition, the reduction of hospital days will mainly be a reduction of 'low care' days. The more intensive and expensive patients remain in the hospital.
From a health system perspective, efficiency also depends on the efficiency of other sectors and on health outcomes . When length of stay reduction is realised by a quicker transfer to follow-up care, the costs of care may be passed. Quicker discharge may increase the pressure on other health care sectors (and their cost) and as a result, the efficiency of the health care system may not improve. Therefore, more insight into the relationship between length of stay and quality of care in the hospital is needed [15–17]. Shorter lengths of stay may also lead to a better quality of care, and, conversely, a better quality of care can lead to a shorter length of stay. For example fewer hospitals days will reduce the chance for complications such as infections and fewer complications will lead to shorter lengths of stay.
On the contrary, we did not find research that showed that shorter lengths of stay in hospitals is related to adverse quality [15, 18, 1, 5]. Only for some specific procedures or diagnoses there is information concerning the limits of hospital stay reduction .
Brownell stated that 'reassuringly, shorter stays have not been found to be related to adverse patient outcomes. In fact, a study of almost 4000 US hospitals showed that hospitals that discharged patients more efficiently had lower post discharge death rates' . Finally, Harrison observed: 'Improving hospital efficiency by shortening length of stay does not appear to result in increased rates of readmission or numbers of physician visits within 30 days after discharge from hospital. Research is needed to identify optimal lengths of stay and expected readmission rates' .
If quality improvement leads to shorter lengths of stay and shorter lengths of stay can lead to a better quality of care, we are curious if hospitals with shorter length of stay have better outcomes than hospitals with a longer length of stay. In future work we will investigate the connection between length of stay and quality of care.