MEDINFO 2001 V. Patel et al. (Eds) Amsterdam: IOS Press © 2001 IMIA. All rights reserved
Method to Determine the Bed Capacity, Different Approaches used for the Establishment Planning Project in the University Hospital of Nancy Eliane Toussaint a, Gilles Herengt a b , Pierre Gillois a b, François Kohler a b a ANADIM - CHU de Nancy - Avenue de Lattre de Tassigny - 54000 Nancy - France b SPI-EAO - Faculté de médecine de Nancy - BP 184 - 54505 Vandoeuvre lès Nancy –France
reapportionment. This planning is applied to both private and public hospitals It also aims at developing collaboration among the various hospitals plus avoiding fruitless competition. With a global vision of care organization, this planning will optimise the use of care units and their equipment and help recognize the need for care accessibility for everyone. At the local level, the bed capacity of a hospital is an important part of its establishment project which is debated with the regional hospital agency.
Abstract In France hospital bed capacity is determined according to a national and regional authorization which has been established by the regional hospital agency. The bed capacity evolution in a hospital is fixed by considering the different proposals of the hospital in negotiation with the regional hospital agency. Types of beds are differentiated according to the patients' needs : medicine, surgery and obstetrics_ The first approach is taken at the national level and then at the regional level using a specific ratio of beds for 100 000 inhabitants in each category. For a given hospital, the authorized number of beds takes into consideration their occupation. Target bed occupation ratios were fixed in 1992 and are still in use. In the establishment project of the University Hospital of Nancy (developed over a five year period) four approaches have been formulated and their results have been compared.
The Lorraine region has too many beds [2]. The University Hospital of Nancy, in the heart of this region has too many beds as well. Its bed capacity in medicine is 1170 beds and 792 in surgery. There is no obstetric activity. The establishment project of the University Hospital of Nancy defines the main targets for the medical field and nursery care as well as for biomedical research, social policy, training plans and information systems. It sets up the hospitals' capabilities, plus the personnel and equipment required to realize their goals.
In this study, the two traditional methods of bed ratio per 100 000 inhabitants and target bed occupation have been updated according to the present conditions of hospitalisation ; the third method is based on the reapportionment of the present patients and the possible risk to the hospital for refusing patients. The last method consists of determining the expected pathologies five years in advance in Lorraine and the beds needed to treat them. These four methods have given consistent results under the accepted revised target occupation bed ratios in accordance with the reduction of the length of stay between 1992 and 1999.
There is a double link between the sanitary planning and the establishment project: : on the one hand the balance sheet prior to the agreement of the sanitary map and of the sanitary organizational scheme has to take into account the approved establishment projects and on the other hand the establishment project has to be in accordance with the aims of the sanitary organizational scheme [2, 3, 4]. At the end of 1999 the university hospital of Nancy issued his project in which a specific projected document gives the evolution of the university hospital installations for five years in advance. This study takes into account the number of beds in medicine and surgery as well as ambulatory medicine and surgery beds fitting the different pluridisciplinary and transdisciplinary projects of the hospital [2]. Four different methods have been developed to calculate the number of beds. Two of them are based on classical methods : the first method uses target bed occupation (TBO or CO). Once applied to the realised activity the principal draw back is to perpetuate the hospital behaviours. An alternative to this method uses national
Keywords: Sanitary planning, modelling, bed capacity, French DRG, GHM, rate of bed occupation
Introduction Sanitary planning regulates the care offer. During recent years it has been increasing with the most important role at the regional level [1]. The goal is to make the care offer meet the needs of the population with the best geographical
1404
Chapter 16: Data Systems
references instead of local means length of stays for each pathology (GHM : French DRG[1]. These references are given each year by the French Ministry of Health in the “Progamme de Médicalisation des Systèmes d’Information” (PMSI). The second method is based on the use of bed ratio for 100 000 inhabitants. The study has been completed by two other methods. The first uses distribution model of the present patients per day in the year and allows to determine the number of beds according to an accepted risk of saturation. The last method uses regional demographic projections to determine hospital stays for the five years in advance and before the application of the first two methods.
Methods a) Method using target bed occupation (TBO) : TBO is the common way of being informed about a hospital activity in order to define its bed capacity [4, 5, 6]. TBO of 85% in medicine and surgery and 95% in obstetrics have been fixed in 1992. Since then, the average length of hospital stays have been significantly decreasing. The impact of this decrease on TBO has been studied. For a given period of time, TBO is the ratio between the sum of the patient's length of stay and the potential days which are the product of the number of beds by the period length. A patient's length of stay equals the check-out date minus the check-in date. The check-out day is not included (1). This method consists in counting patients who are present at midnight.
Material and Methods
Ds = Check-out date – Check-in date Hypotheses
(1)
Potential days (2) can be split up in two parts : the days during which a patient is present (Σds) and the days without patients (Σdi)
To carry on this study, the following hypotheses have been accepted : a) Regional hospital restructuring have been taken into account with parameters set by experts. This has been done for the cessation of activity or transfers as well.
Potential Days : JP N
jp = NombreLits * DuréePériode = å (ds + di )
b) Besides the above mentioned restructuring, the impact of the university hospital in the Lorraine region is supposed not to vary for the five years ahead. The effort to suppress between 2000 and 3 000 beds in Lorraine, which has been mentioned in the sanitary map of Lorraine is expected to result in a fair sharing between all the establishments.
1
(2)
Averages : Average length of stay (3)
å DMS =
N
1
ds (3)
N
c) The incidence of hospitalisation for a given pathology is not expected to vary significantly within the next five years. This hypothesis can be globally accepted but is likely to be discussed for specific pathology, such as AIDS and three therapy [1].
Average length of stay without patient in the bed (4) N
å
=
DMI
di 1
(4)
N
Target bed occupatio : TBO or CO (5)
d) For a given pathology, the length of stay is the same or can be calculated for the next five years. Generally, a slight steady decrease can be observed.
N
CO =
e) Demographic surveys carried by the National Institute of Statistics and their related projections, have not been discussed. They are based on the 1999 population census [1].
å ds 1
jp
N
=
N
å ds 1
N
å (ds + di ) 1
=
å ds 1
N
N
1
1
å ds + å di
=
DMS DMS +DMI
(5)
This induces (6) :
Used Data
DMI
Bed ratio for 100 000 inhabitants which have been used are in accordance with those given by the regional sanitary scheme in Lorraine and the national annual statistics of health care. Target bed occupation is given by official rules [1]. Pathologies treated in the hospital are described by the uniformed medical data set used for the French DRG system (PMSI). With information recorded in a data set, each stay is first classified in a major diagnostic category to be finally affected in a specific group : GHM. For each GHM, national references : length of stay, cost,... are established each year. In this study, 1998 and 1999 version 4.5 of the classification and national references have been used [5].
=
DMS CO
− DMS
(6)
A target bed occupation in medicine of 85% and an average length of stays of 6.8 days in 1992 induces an average of 1.22 days without patients, which is a very short time. The check-out day is not included in the length of stay but is included in this 1.22 day. It seems impossible to reduce this time since bedrooms have to be prepared for other patients and the hospital must have vacancies for unexpected patients. The more the length of stay decreases, the more the target bed occupation (CO) decreases (Table 1). In 1999, for a length of stay of 4 days, the CO is 76 % whereas in 1992, it was 85 % for a length of stay of 6.9
1405
Chapter 16: Data Systems
This table shows that there was only one day without any bed occupation and it happened on a week day,. 36 beds have been occupied 30 times, which happened 8 times on a week end and 22 times on a week day. The number of occupied beds never exceeded 56. The annual number of days during which the corresponding number of beds could meet the needs can be obtained by the cumulative total. For example, 46 beds meet the needs for 269 days a year, therefore the number of days during which the needs will exceed this capacity can be calculated : 365-269 = 96 days.
days. According to these results it appears that it is not wise presently to keep the 1992 target. Therefore, it would be advisable to diminish the target bed occupation according to the new length of stays observed in 1999. Table 1 : Evolution of CO with length of stay (DMS) DMS
CO
DMS
6,9
85,00%
5,8 82,64% 4,7
79,42%
6,8
84,81%
5,7 82,39% 4,6
79,06%
6,7
84,62%
5,6 82,14% 4,5
78,70%
6,6
84,42%
5,5 81,87% 4,4
78,32%
6,5
84,22%
5,4 81,60% 4,3
77,93%
6,4
84,01%
5,3 81,31% 4,2
77,52%
6,3
83,80%
5,2 81,02% 4,1
77,10%
6,2
83,58%
5,1 80,72%
76,66%
6,1
83,36%
6 5,9
5
CO
DMS
4
CO
80,41% 3,9
76,20%
83,13%
4,9 80,09% 3,8
75,73%
82,89%
4,8 79,76% 3,7
75,23%
If the initial number of beds in the hospital is not sufficient, the distribution may be censored on the right. It has not been the case in our study. To estimate the number of beds needed, a risk of saturation must be accepted. This risk can be determined by experts' experience or by modelling the distribution according to a classical probability distribution. Log normal distribution has been applied with a 2.5 % unilateral risk. c) Bed ratio per 100 000 inhabitants method : national target bed ratio is given for medicine, surgery and obstetrics but not for precise disciplines such as cardiology, orthopaedics… To obtain those ratios we have used the national hospital statistics (SAE) [10]. Those ratios have been applied to Lorraine region. The estimation of the number of beds in the hospital is obtained by applying the percentage of the hospital equipment to the Lorraine results. Table 3 gives the regional ratio for 1000 inhabitants
Notwithstanding these considerations, target ratios can be calculated by using, either the effective number of stays with patients or the theoretical number of days obtained by using the national length of stays applied to the case mix of the hospital. The number of beds is therefore very easy to calculate.
Table 3 : Bed ratio in France
Medicine Surgery
Number of beds = Number of days / (period length * Target bed occupation) b) Method using the distribution of present patients : first, the distribution of present patients is calculated : the result equals the number of beds with patients for a given period in the hospital. In table 2, the first column indicates the number of beds with patients. The second one indicates the number of days in which the number of patients given in the first one has been observed. This number can be split in week ends or in week days. Table 2 gives an annual example.
Week end
0
0
1
1
1
1
2
2
4
5
…
…
…
…
46
8
22
30
269
…
…
…
…
…
56
0
1
1
365
Week Total
National extrems 1,65 to 2,35 1,44 to 2,04
d) Method using demographic projections and the evolution of expected pathologies : using the regional Lorraine data base, the frequency of hospitalisation for each GHM (French DRG) can be obtained by sex and age bracket. Five year steps are used for ages. The National Institute of Statistics gives demographic projections for Lorraine population in 2002 by age bracket and gender. The frequency of hospitalisation previsions is obtained by applying Lorraine pathology frequency to the expected population. Then the national length of stays by GHM make it possible to calculate the number of hospitalisation days to which bed target occupation ratio is applied in order to determine the expected number of beds.
Table 2 : example of distribution of present patients Present patients
Bed ratio / 1000 inhabitants 2,35 2,00
Cumulative Total
Results a) Method using target bed occupation (TBO) : three target bed ratios have been used. The statutory ones of 1992 : a 85 % bed ratio, a target bed ratio of 80 % that takes into account the decrease of average length of stays observed in 1998 and finally a 75 % bed ratio corresponding to the expected length of stays in a near future. These ratios have
1406
Chapter 16: Data Systems
By taking into account the risk thresholds of 20 and 60 days, 1 864 beds for 2 days and 1 787 beds for 60 days have been obtained The distribution can be modelled by a log normal distribution (X’= log (length of stay)). Table 6 shows the average, maximum and minimum and standard deviations. A 2.5% threshold of unilateral risk gives a number of 1 584 beds.
been applied, on the one hand to the effective occupation days and on the other hand to the theoretical days which would have been obtained by applying the national reference length of stays to the case mix of the hospital. Table 4 shows the results of this calculation. Table 4 : Expected beds calculated using the effective length of stays
Table 6 : Characteristic of present patients distribution TBO
Number Average Maxi Mini mum mum
of Beds CO 75%
1 784
CO 80%
1 672
CO 85%
1 574
Performance = Theoretical days / Effective days
1,03
Expected beds calculated with the theoretical length of stays
Beds
CO 75%
1 834
CO 80%
1 720
CO 85%
1 618
1 560
1 867 483
240
Tuesday
1 660
1 965 404
261
Wednesday
1 674
1 996 236
304
Thursday
1 657
1 976 449
245
Friday
1 536
1 834 502
219
Saturday
1 294
1 541 438
177
Monday
1 258
1 507 435
173
Week
1 618
1 996 236
1 584
Week end
1 276
1 541 435
1 259
c) Bed ratio per 100 000 inhabitants method : table 7 shows the results of the regional ratio applied to Lorraine and the expected number of beds for the hospital.. These global results can be more detailed by applying this method to the different medical activities described in the annual hospital statistics (SAE) [11].
Table 5 : Distribution of present patients
< 500 [500-1000[ [1000-1500[ [1500-1600[ [1600-1650[ [1650-1700[ [1700-1750[ [1750-1800[ [1800-1850[ [1850-1900[ [1900-1950[ [1950-2000[
Monday
Threshold log normal Maxi Mini Average mum mum distribution Risk 2,5%
b) Method using the distribution of present patients : Table 5 shows the distribution of the number of present patients. The minimum brackets of classes are included. The maximum brackets are excluded. The last three columns show the number of days during which beds will be sought for during week ends, weeks and both. In this situation, beds can be found by checking out patients earlier or by putting them in other hospitals and by registering the new ones on waiting lists.
Present Patient
Standard deviation
W Wee Bot Perc Cumu Number of days in ee k h enta lative which beds are k ge Perce needed en ntage d Week Week Bot end h 2 4 6 1,64 1,64 102 257 359 2 3 5 1,37 3,01 100 254 354 94 52 146 40,0 43,01 6 202 208 6 28 34 9,32 52,33 0 174 174 0 28 28 7,67 60,00 0 146 146 0 28 28 7,67 67,67 0 118 118 0 32 32 8,77 76,44 0 86 86 0 35 35 9,59 86,03 0 51 51 0 28 28 7,67 93,70 0 23 23 0 9 9 2,47 96,16 0 14 14 0 9 9 2,47 98,63 0 5 5 0 5 5 1,37 100,00 0 0 0
1407
Table 7 : Expected beds using bed ratio per inhabitants SAE Code
Name
Hospital Bed / bed 100 000 inhabitants percentage
Expected beds
21100
Medicine
209.41
20.82%
1 003
21200
Surgery
187.40
16.47%
710
28.28
0.00%
0
213AA Obstetrics Total
1 713
c) Method using demographic projections and the evolution of expected pathologies : in 1999, each hospital stay resulted in a standardized summary which allowed to classified it in a GHM (French DRG). From effective days
Chapter 16: Data Systems
show an increase in comparison with 1997, which modulates the diminution of beds. These global approaches smooth the results for annual activity. To cope with seasonal variations of activity linked, either to pathologies (bronchitis…) or to population movements (holidays), bed number estimation must be completed with a study of the ups and downs in order to improve programmed stays and with the personnel’s vacation in order to meet the needs.
in each category and target bed occupation of 80 %, it has been possible to determine the theoretical number of beds in each major category subdivided in medicine and surgery. The regional demographic projections for 2002, given by the National Institute of Statistics, make it possible to calculate hospital stays for these projections (table 8). Since there is no obstetrics in the university hospital of Nancy and taking into account the growing number of elderly people, the result shows an increase in the number of beds needed with an expected number of 1 650.
References
Table 8 : demographic projections and expected beds
[1] Code de la santé publique article L712.
Beds in 1997
Expected Beds in 2002
[2] Schéma régional d’organisation sanitaire et sociale de Lorraine 1999-2004. A.R.H. de Lorraine. 1999 ; 1-142
Stays in medical GHM
988
1032
Stays in surgical GHM
592
618
[3] Holcomb B. Planners eye Michigan effort to trim 5,000 hospital beds. Health Care Week 1979 ;2(28) :1,15.
1 580
1 650
TOTAL
[4] Hinz CA. In Michigan unique agency aims to trim beds. Am Med News 1983 ;26(46) :1,9. [5] Clerkin D, Fos PJ, Petry FE. A decision support system for hospital bed assignment. Hosp Health Serv Adm 1995 ;40(3) :386-400.
Discussion – Conclusion First it seems difficult to apply a method to a specified establishment. However, considering the size of the tested hospital, the obtained results are worth being reckoned with since they derive from a study based on a great number of beds, stays, … therefore the statistical estimation are highly reliable. Table 9 summarizes the results of the different methods.
[6] Leu RE, Schaub T, Sommer JH, Gutzwiller F. Hospital census : a means for determination of hospital bed availability. Offentl Gesundheitswes 1984 ;46(7):315-9. [7] Ministère de l’Emploi et de la Solidarité. Manuel des groupes homogènes de malades, version 5 de la classification, BO, version 4.5 de la fonction groupage, 1998, Vol I(Edition corrigée), n°98-2 bis.
Table 9 : Results with the different methods Method using target bed occupation Number of CO =80% beds On effective days 1 672 On theoretical days 1 720 Distribution of present patients Risk of saturation 20 days 1 864 Risk of saturation 60 days 1 787 Risk of saturation 2,5% 1 584 Bed ratio per 100 000 inhabitants method 1 713 Method using demographic projections 1 650 4 methods characteristics Average 1 713 Standard deviation 92 Minimum 1 584 Maximum 1 864 Range 280 Variation coefficient 5,36% The obtained results are in accordance with the experts’ projections. If there is a range of 280, the variation coefficient is only 5.36 % with an average 1 713 of beds needed for our university hospital. These results confirm a too great number of beds in our university hospital and make it possible to estimate the reapportionment needs. However, the 2002 projections, for expected pathologies
[8] Cabié A. Prise en charge ambulatoire des patients séropositifs et sidéens. Concours médical 1997 ; 119 (8) : 516-520 [9] Mulic S, Counot S. Recensement de la population 1999 : premiers résultats. Economie Lorraine 1999 ;190. [10] République Française Bulletin Officiel N°83/8 bis [11] Ministère de l’Emploi et de la solidarité. Statistique annuelle des établissements 1996. D.R.A.S.S. de Lorraine, Service Statistique 1996. [12] Ministère du travail et des affaires sociales. Service des Statistiques, des Etudes et des Systèmes d’information. Cahier des Charges Technique Informatique. SAE 1996 [13] Ministère du travail et des affaires sociales. Service des Statistiques, des Etudes et des Systèmes d’information. Guide de l’utilisateur. SAE 1996. [14] Ministère du travail et des affaires sociales. Service des Statistiques, des Etudes et des Systèmes d’information. Aide au remplissage : Définitions Consignes. SAE 1996. Address for correspondence : François KOHLER, Laboratoire SPI-EAO, Faculté de Médecine de Nancy, BP184, 54505 Vandoeuvre cedex, France.
[email protected]
1408