Q J Med 1999; 92: 199-206
© 1999 Association of Physicians
The winter bed crisis—quantifying seasonal effects on hospital bed usage
1 From the Belfast City Hospital, and 2 Department of Geriatric Medicine, The Queen's University of Belfast, Belfast, UK
Received 20 November 1998 and in revised form 3 February 1999
Dr K.J. Fullerton, Department of Geriatric Medicine, The Queen's University of Belfast, Whitla Medical Building, 97 Lisburn Road, Belfast BT9 7BL. e-mail: k.fullerton{at}qub.ac.uk
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionReferences |
|---|
Winter bed crises are a common feature in NHS hospitals, and have given rise to great concern. We set out to determine the relative contribution of seasonal effects and other factors to bed occupancy in a large teaching hospital over one year. There were 190 804 occupied bed-days, which we analysed by specialty groupings. There was considerable variability in bed occupancy in each specialty. A significant winter peak occurred for general medicine and orthopaedics together with a significant increase on `take-in' days. Virtually all specialties showed a significant variation in occupancy between weekdays. Geriatric Medicine had a high and fairly constant occupancy, with some seasonal effect. We conclude that seasonal trends in bed occupancy occur in `front door' specialties and are predictable. In these specialties, admission policies also make a contribution to bed usage and are amenable to modification. There is no surge in occupancy in the immediate post-Christmas period, except that attributable to the seasonal trend. In the `elective' specialties, bed occupancy fluctuates widely, with reduced occupancy at weekends and at Christmas. These differences are entirely amenable to modification. More effective bed management would make a very significant contribution to avoiding winter bed crises.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Emergency medical admissions to UK hospitals have risen substantially in recent years.1 Winter peaks in such admissions occur, probably caused by peaks in respiratory and cardiovascular disease,2 and give rise to winter bed crises in many UK hospitals.3 In 1973, an analysis of admission data for a large hospital revealed a negative relationship between admissions and temperatures.4 Winter pressure on hospital medical beds in Aberdeen between 1984 and 1988 was determined by cosinor analysis.5 No significant winter pressure was found for total admissions, but there were strong seasonal effects for cardiovascular and respiratory admissions and for patients over 75 years. Admissions for all other medical disorders were not seasonal. Recent reductions in winter mortality6 have not been paralleled by reductions in winter hospital admission. A recognition of the problem has led to increased government funding of the NHS to help deal with winter pressures.7 We decided to identify seasonal trends in hospital bed usage, and to attempt to quantify the seasonal effect, as opposed to other factors, across a range of hospital specialties.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Data were obtained from the Patient Administration System of a large teaching hospital for the period 1 April 1996 to 31 March 1997 inclusive. The data gave details of the total number of hospital beds occupied by patients from each specialty every day for the study period. These data were analysed by the technique of Cosinor Rhythmometry using the GLIM statistical package,8 and using purpose-built software, in order to determine any true seasonal variation, and to quantify its effect. This technique9 allows a test of the null hypothesis that there is no underlying seasonal sinusoidal curve (using the F(seasonality) statistic), tests the hypothesis that the curve explains all the variability (the F(fit) statistic), and gives a figure for the proportion of the variability which can be explained by the curve (the R2 statistic). The seasonal difference is given as the difference between peak and trough bed occupancy. Variations in occupancy by day of the week were explored using one-way analysis of variance. Additional analysis using logistic regression helped to identify and quantify other possible influences on bed occupancy. The fitted seasonal model was included as an independent variable, and additional factors (all categorical variables) which had a significant role were detected by a `backward conditional' iterative model, using an indicator group for each variable.10
For the purposes of this study, specialty groupings were defined as follows: General Medicine, General Internal Medicine, Respiratory Medicine, and Gastroenterology; Geriatric Medicine, Geriatric Acute, Geriatric Rehabilitation, Geriatric Continuing Care (20 beds); Other Medical Specialties, Cardiology, Oncology, Haematology, Nephrology, and Dermatology; Surgical Specialties, General Surgery, Vascular Surgery, Gynaecology, ENT, and Urology; Orthopaedics, Orthopaedic Surgery; Other Specialties, all remaining specialties including Obstetrics and Neonatology.
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Bed occupancy
During the study period, there were a total of 190 804 occupied bed-days, representing an overall bed occupancy in the hospital of 85%. A breakdown of bed usage by specialty is given in Table 1
. The average available bed-pool for the year was as follows: General Medicine, 134 beds; Geriatric Medicine, 121 beds (including 20 continuing care beds); Other Medical, 107 beds; Surgery, 122 beds; Orthopaedic, 49 beds; Other Specialties, 82 beds.
|
There was considerable variability in bed occupancy in each specialty grouping. Peaks and troughs were seen at certain times of the year. The timing and pattern of these peaks and troughs was different in different specialty groupings. Only in General Medicine and Orthopaedics was a clear winter peak observed (Figure 1
). Some groups, such as Orthopaedics, Surgical Specialties and Other Medical Specialties showed wide fluctuations in occupancy, while others, particularly Geriatric Medicine (Figure 1), showed a more constant occupancy level.
|
Day effect
Variations by weekday are shown in Table 1
. All specialties except Geriatric Medicine, and the overall data, showed a significant variation in occupancy between weekdays. In General Medicine and Orthopaedics, the minimum daily occupancy occurred on Fridays, in Geriatric Medicine and the Other Medical Specialties the minimum was on Sundays, while in Surgical and Other Specialties the minimum occupancy was on Sundays. The overall minimum hospital occupancy was on Saturdays.
Take-in effect
The hospital surveyed is involved in a system whereby it and the other city teaching hospital are on alternate days the `hospital of first choice' for emergency admissions from the joint catchment area, and `hospital of last resort' for emergency patients from outside the catchment area for whom beds cannot be found elsewhere. The day when the hospital is in this role is designated `take-in day' in the results which follow. Table 2 gives details of the effects of being the `take-in' hospital and of Christmas week (seven days beginning 26 December) respectively. In General Medicine, Orthopaedics and the Other Specialties, occupancy was significantly higher on take-in days. This resulted in an overall increase in occupancy on those days. In General Medicine and Orthopaedics, occupancy was significantly increased at Christmas, while in Surgical and Other Specialties it was significantly decreased. The overall effect was of decreased hospital occupancy at Christmas compared with the rest of the year.
|
Season effect
The results for cosinor rhythmometry analysis are summarized in Table 2
, while Figure 2 gives representative plots of the data and the fitted curves for General Medicine and the Surgical Specialties, respectively. Only in the specialties of General Medicine, Geriatric Medicine and Orthopaedics is there a significant seasonal effect, which explains 55%, 24% and 32% of the variability in bed occupancy, respectively. This effect is to some extent buffered by Other Specialties, where there is no demonstrable seasonal effect, so that although the overall occupancy is significantly seasonal, this is only sufficient to explain 9% of the variability. The peak occupancy in Geriatric Medicine occurs approximately 3 weeks after that in General Medicine.
|
Composite influence
The interplay of factors that might have a significant effect on bed occupancy was investigated using linear logistic modelling. The results are summarized in Table 3
. The likelihood ratio refers to the likelihood in each group that more than the average number of beds would be occupied. Daily occupancy is compared with the occupancy on Sundays, when mean hospital occupancy was lowest. In the interests of brevity and clarity, only categories with a significant relationship to outcome (in terms of specialty bed occupancy higher than the mean) are included. None of the variables made a significant contribution to occupancy in the Other Medical Specialties.
|
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
There is little disagreement that hospitals come under pressure for acute beds, and that this pressure is worse in the winter months. These winter pressures have been attributed to increased incidence of respiratory and cardiovascular disease,5 but other factors have been suggested, including delays in hospital discharge in the period immediately after Christmas. It is proposed that this `Christmas effect', presumed to be caused by annual leave among supporting staff in the community, has a knock-on effect in terms of increased bed occupancy, which can persist for several months. It is further proposed that this effect is in itself sufficient to be the sole cause of the winter bed crisis.11,12 An interesting further contribution to this discussion is the observation that at times of national significance, such as at the time of Princess Diana's funeral, there is a diminution in demand for acute medical services, followed shortly by increased demand.13 A similar effect might be expected to apply in the period immediately around Christmas Day, with an upsurge in demand during the following week. This is another possible contributing factor to the `Christmas effect'.
Attention has been drawn to policies and practices within hospitals, which can contribute to inappropriately long hospital stays. Timing of ward rounds and operating lists, and weekend leave mean that there are disproportionate delays in discharge from hospitals at weekends and holidays.14 This effect might be expected to be greatest in specialties with a relatively high rate of elective admission. It is important to be able to disentangle these effects, since some of them are more amenable to correction than others.
Seasonal peaks in the incidence of and mortality from certain diseases during winter were described by Hippocrates,15 and it was known that there is a disproportionate effect on elderly people.16 Although most attention has focussed on cardiovascular and respiratory disease,5,17 a similar effect is observed in the incidence of femoral neck fractures.18 It might be expected, therefore, that seasonal bed pressures would be most pronounced in hospital specialties that deal with acute medical problems in older people and in Orthopaedic Surgery. The technique of Cosinor Rhythmometry has been used in the detection and quantification of seasonal differences in disease mortality9 and incidence,5 and similar methods have been used in exploring potential seasonal causes of this mortality.19–21
Some of the variables that can contribute to increased bed occupancy, for example a winter seasonal trend and the `Christmas effect', are clearly interrelated. Multivariate techniques, such as linear logistic modelling, are helpful in defining each of these effects independent of the others.
The presence of a winter bed crisis in the hospital investigated is indicated by peaks in occupancy rates in General Medicine and Orthopaedics well in excess of allocated bed numbers. We have demonstrated that a clear seasonal effect in the use of hospital beds exists. It is observed in Acute Medicine, Orthopaedics and Geriatric Medicine; these are the specialties that admit and treat respiratory and cardiovascular diseases and fractures in older people. It accounts for 56%, 32% and 24% of the variance in occupancy, respectively. It leads to a clear seasonal effect in the overall use of hospital beds, accounting for 9% of the variance in overall occupancy. The peaks in General Medicine and Orthopaedics occur in January, and are followed by a peak in Geriatric Medicine in February. This is likely to represent the transfer of older patients for rehabilitation. In these two `front door' specialties, Friday is the day with lowest occupancy, and there is significantly lower occupancy on this day (and also on Saturday in Orthopaedics) on multivariate analysis. This suggests an increased rate of discharge on Fridays, with a re-accumulation of patients over the weekend. These are the only specialties in which occupancy at Christmas is significantly raised (Table 4). This is not, however, an independent effect (Table 6) and is largely explained by the seasonal trend rather than by a `Christmas effect'. Being designated `take-in hospital' means that occupancy is three times as likely to be high in General Medicine than on other days, and the hospital overall is 2.3 times more likely to have increased occupancy.
None of the other specialty groups demonstrate a significant seasonal trend. They all demonstrate considerable variability in bed occupancy, and a much clearer `holiday and weekend' effect, with lower pressures at weekends, at Christmas, and at other holiday times (Figure 2). Overall, the hospital is less likely to have increased occupancy at weekends. In the Other Medical Specialties, the large fluctuations in occupancy could not be explained by any of the factors measured.
These observations ought to make a contribution to the debate on bed utilization and bed pressures. First, in the `front door' specialties of General Medicine and Orthopaedics, a major proportion of the variability in bed use is attributable to seasonal effects. Particularly in the case of General Medicine, this makes the demand for beds more predictable than is usually assumed. Even in these specialties, however, other factors such as the emergency arrangements in the city or the day of the week are also significant contributors to bed use, and perhaps more amenable to adjustment than the time of year.
The `weekend effect' is noticeable in nearly all specialty areas. In the `front door' specialties this may reflect a clearing of beds on Fridays in anticipation of demand at the weekend, while reduced activity in `elective' specialties leads to reduced bed occupancy on Saturdays and Sundays in comparison with the rest of the week. Perhaps a more extreme example of the same effect is seen at Christmas in the `elective' specialties, where the `weekend effect' stretches over a week. The lack of a demonstrable Christmas effect in the `front door' specialties does not accord with the view that better discharge facilities at this key time could nullify the winter bed pressures entirely. Finally, the overriding significance of the seasonal effect does not mean that medical beds will inevitably be over-filled in the winter months. We have shown that the seasonal effect is significant, but predictable. Better planning of the use of hospital beds in the wintertime, coupled with initiatives in primary care enabling more people to be treated at home, and in social services in avoiding unnecessary delays in hospital discharge will not avoid the seasonal increase in morbidity, but might avoid some of the scenes of chaos and crisis which seem endemic in larger hospitals at this time. The practice adopted by some hospitals of curtailing elective work during the winter peak, and making the beds available for acute medical patients would seem sensible. In the UK, the release of extra `winter pressures money' to allow the bed-pool to be enlarged during the winter also ought to alleviate the situation. The logic of releasing `waiting list initiative' money to increase elective work during the same period is less clear.
In our own hospital, we have adopted a number of measures to address the issues raised by these data. We have brought together physicians, surgeons and bed-managers in an attempt to be more proactive in matching elective work and acute demand. We are working with other hospitals and our main purchaser in trying to smooth the day-to-day fluctuations by spreading the load at times of peak demand. With a collaborative approach to `winter pressures' spending we have been able to increase the overall acute medical winter bed-pool by 40–50 beds. In February 1999 the additional pressures of a `flu epidemic resulted in the cessation of all elective work in the city. On several occasions, the `take-in' hospital started the day `on take' with no available medical beds. Clearly, much still remains to be done!
|
Acknowledgments |
|---|
We acknowledge the help and support of Mr S. McKenna, Information Department, Belfast City Hospital.
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
1. Kendrick S. The pattern of increase in emergency hospital admissions in Scotland. Health Bull 1996; 54:169–83.
2. Kendrick S. Frame S, Povey C. Beds occupied by emergency patients: long term trends in patterns of short term fluctuations in Scotland. Health Bull 1997; 55:167–75.
3.
Blatchford O, Capewell S. Emergency Medical Admissions: taking stock and planning for the winter. Br Med J 1997; 315:1322–3.
4. Bull GM. Meterological correlates with myocardial and cerebral infarction and respiratory disease. Br J Prev Soc Med 1973; 27:108–13.[Web of Science][Medline]
5. Douglas AS, Allan TM, Rawles JM. Composition of seasonality of disease. Scot Med J 1991; 36:76-82.
6.
Donaldson GC, Keatinge WR. Mortality related to cold weather in elderly people in southeast England, 1979-1994. Br Med J 1997; 315:1055–6.
7.
Warden J. Government injects cash to reduce winter crisis in NHS. Br Med J 1997; 315:967–72.
8. Generalised Linear Interactive Modelling System, Version 8.0 (VMS platform).
9. Nelson W, Liang Tong Y, Lee J-K, Halberg F. Methods for cosinor-rhythmometry. Chronobiologia 1979; 6:305–32.[Web of Science][Medline]
10. Superior Performance Statistical Software. SPSS version 8.0 for Windows.
11. Millard P. Quoted in: Hospital Doctor 1997; 21 August:40–1.
12. Millard P, Kearney D, Schrooten P. Occupation des lits Compte de Noel. Intramuros 1997; 4(April):11.
13. Morgan-Jones MR, Smith K, Oakley P. The `Diana effect'. Br Med J 1998; 316:1751.
14. Audit Commission National Health Service Report No 5. Lying in Wait: the use of medical beds in acute hospitals. HMSO, 1992.
15. Hippocrates. Aphorisms 3:23.
16. Avicenna's Canon of Medicine. Thesis II:268.
17. Douglas AS, al-Sayer H, Rawles JM, Allan TM. Seasonality of disease in Kuwait. Lancet 1991; 337:1393–7.[Web of Science][Medline]
18. Douglas AS. Seasonality of hip fracture and haemorrhagic disease of the newborn. Scot Med J 1993; 38:37–40.
19. Stout RW, Crawford V. Seasonal variations in fibrinogen concentrations among elderly people. Lancet 1991; 338:9–13.[Web of Science][Medline]
20. Woodhouse PR, Khaw K-T, Plummer M, Foley A, Meade TW. Seasonal variations of plasma fibrinogen and factor VII activity in the elderly: winter infections and death from cardiovascular disease. Lancet 1994; 343:435–9.[Web of Science][Medline]
21. Van der Bom JG, de Matt MPM, Bots ML, Hofman A, Kluff C, Grobbee DE. Seasonal variation in fibrinogen in the Rotterdam study. Thromb Haemost 1997; 78:1059–62.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||