APPENDIX 15
Joint memorandum by Dr Thierry Chaussalet,
Dr Elia El-Darzi and Professor Peter Millard, from The Health
and Social Care Modelling Group (DD 25)
1. MAIN RECOMMENDATION
1.1 The Health Committee should recommend
government to develop statistically valid models to test the possible
short and long-term outcome of the proposed changes in health
and social care systems.
1.2 In reviewing evidence concerning the
benefits and disadvantages of speedy hospital care the Health
Committee needs to take into account the outcome of the policy
on the individuals as well as on the total system of care.
2. THE UNIVERSITY
OF WESTMINSTER
HEALTH AND
SOCIAL CARE
MODELLING GROUP
(HSCMG)
2.1 HSCMG is an interdisciplinary group
of academic staff, research assistants and PhD students with expertise
in mathematical modelling, computer programming, operational research,
decision sciences and statistics.
2.2 Dr Thierry J Chaussalet is a principal
lecturer in Management Science at the University of Westminster.
His main research interests are in modelling biological and health
care processes, and the development of decision support tools
that can aid decision making in the care of older people.
2.3 Dr Elia El-Darzi is chair of the department
of Software and Information Systems at the University of Westminster.
His main research interest is the application of decision modelling
and IT to business problems.
2.4 Professor Peter H Millard is a visiting
professor to the University of Westminster. He held the Eleanor
Peel Chair of Geriatric Medicine at St George's Hospital Medical
School (1979-99). 1989 and 1993 doctoral thesis developed and
tested a theoretical and practical way of measuring flow through
health care systems
2.5 Recent peer reviewed HSCMG publications
concern: modelling the decisions of a local authority panel for
admission of older people to nursing and residential home (Xie
et al 2001); simulating interaction between acute, rehabilitative
and long stay care (El-Darzi et al 1999, 2000); public holiday
effect theory to explain post-Christmas bed crises (Vasilakis
& El-Darzi 2001); measuring and modelling patient flow in
hospital (El-Darzi et al 1998; Christodoulou & Millard 2000)
and surgical services (Millard et al 2000); use of flow modelling
to demonstrate how rehabilitation changed bed usage in a Spanish
hospital (Garcia-Navarro & Thompson 2001); modelling the flow
of elderly patients, clients and residents through health and
social care services in an English Health District (Millard et
al 2001); a decision tree approach to evaluating options for case
finding, diagnosing and treating older people with dementia (Millard
& Chaussalet 1998; Chaussalet et al 1998); Markov models and
data requirements for natural history modelling of Alzheimer's
disease (Chaussalet & Thompson 2001); sub-group analyses of
cost of care in Alzheimer's disease (Thompson & Chaussalet
2001); simulation study on evaluating screening policies for Alzheimer's
diseases (Chaussalet et al 2000); application of a continuous
time hidden Markov model to model geriatric inpatient behaviour
(Christodoulou & Taylor 2001).
3. SEEING THE
BIG PICTURE
3.1 There are 64 ways that a hospital system
providing acute, rehabilitative and long stay care can be changed—not
all of these changes are beneficial. The corollary of increasing
numbers of potential long-term patients in the community, single
parent families and an ageing population is that the number of
so-called "social admissions" to hospitals will increase.
For, acute hospitals are the community's ultimate safety net.
3.2 The system is overheated, because the
Operational Plan that underpinned the NHS has been forgotten.
Prolonged rehabilitation of potential long stay patients was a
key component of the original plan. (Millard 1994). In direct
contrast, the proposed solution to the bed crisis focuses attention
on blocking "social" admissions and increasing speed
of treatment.
3.3 In this submission we illustrate modern
techniques of data analysis that could be used to facilitate understanding
and planning.
4. BLACK-BOX
PLANNING
4.1 Mistakes are being made in the planning
of hospital services because "black-box models" are
used to underpin health and social care decision-making. In black-box
models (Figure 1) all measures of activity are taken without the
system understudy.
4.2 Black-box models assume that the space
within hospitals is homogenous. Thus, removing waiting lists and
freeing acute beds is seen to be simply a matter of shortening
length of stay, stopping inappropriate (social) admissions, early
discharge, rapid transfer elsewhere and time limited (six weeks)
rehabilitation.
4.3 If the model is right, the solution
is right. What-if the model is wrong? Here we use results of recent
research to show why the current policy initiative of speedy discharge
and time-limited rehabilitation will probably fail. To understand
the unique features of the NHS bed crisis new methods of data
presentation and analysis need to be introduced.
BED OCCUPANCY
5.1 Figure 2 uses a new approach named "Strata"
to show how hospital beds are actually being used. The data concerns
two years inpatient bed usage in geriatric and general medicine
in a London teaching hospital. The plot was created using a discharge
data set: this explains the rapid decrease in bed numbers at the
end of the two-year period.
Figure 2 The stratas within occupied medical
and geriatric beds
5.2 Although the majority of inpatients
are short stay, the beds are mainly occupied by patients who stay
or more than one week. The black layer beneath the day patient
layer shows the beds used by patients who stayed for less than
seven days. The other layers represent bed usage by patients with
progressively longer lengths of stay.
5.3 The choice to do one thing is a choice
not to do another thing. Given the national shortage of qualified
therapy staff, the probable detrimental effect of a policy focus
on rapid discharge and community-based rehabilitation will be:
1. A decrease in the quality of inpatient
rehabilitative care.
2. An increase in dependency of hospitial
patients.
3. An increase in the demand for long-term
care, and
4. An increase in the number of readmissions
and social admissions.
UNDERSTANDING LENGTH
OF STAY
Length of stay is skewed
The current policy focuses attention on shortening
the average length of stay of discharged patients. Yet, the data
in table 1 shows that the average stay (the mean) is an inadequate
measure of activity. Consider general medicine: the average stay
is 11 days; the most frequent length of stay (the mode) is two
days; 50 per cent (the median) of overnight patients are discharged
in four days and the distribution is skewed—because the
range is one to 207 days.
Table 1
DESCRIPTIVE STATISTICS OF LENGTH OF STAY
IN DAYS OF INPATIENTS DISCHARGED DURING THREE MONTHS: DAY CASES
ARE NOT INCLUDED
| Statistics | All cases
| Medicine | Surgery
| Geriatrics | Obstetrics
| Paediatrics |
| Mean | 10 |
11 | 8
| 31 | 5
| 3 |
| Cases | 4,916
| 1,325 | 1,219
| 525 | 1,060
| 787 |
| Minimum | 1 |
1 | 1
| 1 | 1
| 1 |
| Maximum | 463
| 207 | 96
| 463 | 121
| 32 |
| Mode | 2 |
2 | 2
| 14 | 1
| 1 |
| Median | 4 |
7 | 4
| 17 | 2
| 2 |
| Skewness | 9
| 7 | 3
| 5 | 6
| 5 |
6.1.1 Given the evidence in table 1, numerically and
clinically, the correct way to shorten the average length of stay
in all specialities is to focus attention on the prevention and
improved inpatient management of complexity. Anything that makes
the probability increase that an admitted patient will become
unnecessarily dependent is, in the long run, detrimental to the
total well being of the person and the group.
6.2 SPEED ALONE
IS NOT
ENOUGH
6.2.1 A minimum of two numbers is needed to explain differences
in hospital length of stay. Speed alone is not enough. A car travelling
at 100 miles an hour around Parliament Square is likely to crash.
Where it crashes depends on the direction the car is being driven.
Similarly, speed of hospital discharge is an inadequate indicator
of success.
6.2.2 In addition to measuring speed of treatment Ministers
need to find ways of measuring the end result of the patients'
journey through hospital care. Analysis of the pattern of discharge
destinations of patients aged 65 and over with stroke illness
showed that the risk in 1994 of being discharged to nursing homes
varied between one in a hundred and one in ten (Millard 1998).
Such differences have enormous personal and financial consequences
that should not be ignored.
6.2.3 A study, undertaken in 14 departments of geriatric
medicine in southern England, showed that three factors explained
observed differences in speed of treatment:
1. The percentage of inpatients coded for discharge to
nursing homes.
2. The consultants' opinion concerning ease of discharge
of potential long stay patients; and
3. The nurses' opinion concerning the quality of inpatient
rehabilitative care.
4. Departments with a slower turnover of short stay patients
referred fewer patients to long stay care services. Conversely,
the quicker the turnover of short stay patients, the greater the
likelihood that they would be discharged to nursing homes.
6.3 In reviewing evidence concerning the benefits and
disadvantages of speedy hospital care the Committee needs to take
into account the outcome of the policy on the total system of
care.
UNDERSTANDING FLOW
7.1 Words such as acute, rehabilitation and long stay
reflect dimensions of time, as well as dimensions of resource
use. The key to improved planning depends on understanding how
resources in the hospital and in the community are interacting
to influence staff behaviour.
7.2 Success in the running of a hospital depends on all
presenting patients having optimal treatment. Poor performance
in any part of the system is detrimental to the individual sick
person as well as to the group. Policies associated with the blocking
of certain types of patients, eg social problems, will change
the case mix and change the flow.
7.3 Flow models differ from case mix models because they
measure the time being taken by patients to flow through the system
of care.
7.4 When considering the wisdom of "blocking"
social admissions the committee should bear in mind that the perceived
wisdom of the generation of geriatricians who established the
acute aspects of the specialty was that "delay in admission
blocks therapeutic effectiveness" (Hodkinson & Jeffreys
1972).
MEASURING FLOW
8.1 The flow of patients through health and social care
systems can easily be measured using computer assisted data analysis.
Bed occupancy was considered by Yates (1982) to be a
8.2 Since 1991, when the original paper that described
the method of measuring flow was published, the new method of
measuring hospital inpatient activity has been validated and extended
by academic researches. The results of data analysis from a wide
range of hospital services in different countries have confirmed
the universal applicability of the method of data analysis.
8.3 Doctoral students working with Prof McClean at the
University of Ulster have both extended the method of analysis
to include community compartments and developed ways of predicting
the probability that older patients referred to hospital will
occupy beds. Doctoral students working with us are currently developing
methods we have illustrated here to analyse large databases and
to forecast the outcome of client referral to local authority
panels.
Table 2
MODELLED RESOURCE UTILISATION FOR THREE SPECIALITIES AND
THE OVERALL HOSPITAL, WITHOUT DAY CASES
| | General Surgery
| General Medicine |
Geriatrics | Overall
|
| Actual Beds | 97.0
| 132.0 | 140.0
| 386.0 |
| Modelled Beds | 96.8
| 132.6 | 136.7
| 380.9 |
| Admissions per day | 13.0
| 13.8 | 5.0
| 27.5 |
| Overall average length of stay | 7.4
| 9.6 | 27.3
| 13.9 |
| Fast stream |
| Average Stay (days) | 3.7
| 5.6 | 11.7
| 9.0 |
| 50% leave (½ life) | 2.2
| 3.6 | 7.8
| 5.9 |
| Percentage Discharged | 77.4
| 76.0 | 72.6
| 88.0 |
| Modelled Beds Used (%) | 48(49.9)
| 78(58.5) | 59(42.9)
| 247(64.8) |
| Slow stream |
| Average Stay (days) | 16.5
| 14.5 | 49.1
| 36.8 |
| 50% leave (½ life) | 11.1
| 9.7 | 33.7
| 25.2 |
| Percentage Discharged | 100.0
| 97.1 | 99.7
| 99.8 |
| Modelled Beds Used (%) | 48(50.1)
| 48(36.1) | 67(49.3)
| 121(31.8) |
| Long stay stream |
| Average Stay (days) |
| 76.1 | 2,543.9
| 1,685.9 |
| 50% leave (½ life) |
| 52.4 | 1,762.9
| 1,168.3 |
| Percentage Discharged |
| 100.0 | 100.0
| 100.0 |
| Modelled Beds Used (%) |
| 7(5.5) | 11(7.8)
| 13.0(3.5) |
8.4 Table 2 contains the results of a bed occupancy flow
rate analysis, which shows how inpatients were flowing through
the 386 occupied beds as well as in the general surgery, general
medicine and geriatrics beds. Bear in mind that these results
were obtained simply by analysing bed census data. Notice that
the specialities of general and geriatric medicine contained three
streams of flow, and general surgery contained two streams.
8.5 As well as identifying the streams the new method
estimates the likely outcome of admission. Overall 88 per cent
of admitted patients will be short stay. These patients will use
two-thirds of the beds (64.8 per cent) and have an average stay
of 9.0 days (50 per cent will be discharged in 5.9 days). The
average stay of the second stream patients will be 36.8 days (50
per cent will leave in 25.2 days). Overall, only two in every
1,000 admitted patients will enter the third stream of care. Their
average stay will be 1,685.9 days (fifty percent will leave in
1,168.3 days): they will occupy 13 beds.
8.6 The long stay patients are present in the St George's
Hospital department of geriatric medicine because the Bolingbroke
Hospital contains the world's first, single room, personalised,
demonstration long stay unit (Millard & Kist 1995).
9. WHAT-IF
ANALYSIS
9.1 We conclude our submission to the Health Committee
with the "What-if" figure (Figure 4) from the paper
that developed the new method of measuring and modelling bed occupancy
data (Harrison & Millard 1991). Using data derived from analysis
of the streams of flow in the occupied beds in one department
of geriatric medicine the figure shows the short and long stay
effect of different changes in the system of care.
9.2 The benefit of using bed occupancy time to model
flow are fourfold.
1. Current and historic resource use can be analysed.
2. The work function of occupied beds can be measured.
3. Dynamic and stochastic models can be created.
4. What-if analysis can be performed.
9.3 Given the unknown territory into which the government
is moving the rehabilitative component of care, the worst-case
scenario is that the actions being taken will make the problem
of providing care for the population worse. Research by Prof Sally
McClean at the University of Ulster has led to the development
of methods of measuring and modelling the total system of care
(Taylor, McClean and Millard 1998, 2000) and predicting the time
component of resource utilisation by dependent people (Marshall,
McClean et al 2001). Our group has developed methods of simulating
the effect of change on the total system of care. We argue that
the potential to solve the long-term health and social care needs
of an ageing population can only be achieved by using a new scientific
approach to the planning of health and social care services.
9.4 We are grateful for the opportunity presented by
this Inquiry to present aspects of this exciting new field of
research endeavour and would be pleased if the opportunity arises
to attend to give oral evidence or to welcome representatives
of the committee to demonstrate the computer assisted data analytical
tools that our group has been developing and testing.
Thursday 24 January 2002
REFERENCES
Chaussalet T.J., Millard P.H. and El-Darzi E. (1998) Evaluating
the costs of alternative options for dementia services. Health
Care Management Science 1, 125-131.
Chaussalet T.J., Xie, H. Thompson W.A. and Millard P.H. (2000)
A discrete event simulation approach for evaluating screening
policies for Alzheimer's disease. In Simulation and modelling—enablers
for a better quality of life: Proceedings of the 14th European
Simulation Multiconference, May 23-26, 2000, Ghent, Belgium, SCS
Publications, pp 624-630.
Chaussalet, T.J. and Thompson W.A. (2001) Data requirements
in a model of the natural history of Alzheimer's disease. Health
Care Management Science 4, 11-17.
Christodoulou, G. and P.H. Millard (2000) Measuring and modelling
patient flow. British Journal of Health Care Management 6, 463-468.
Christodoulou, G. and Taylor G.J. (2001) Using a continuous
time hidden Markov process, with covariates, to model bed occupancy
of people aged over 65 years. Health Care Management Science 4,
21-24.
El-Darzi E., Vasilakis C., Chaussalet T.J. and Millard P.H.
(1998) A Simulation Modelling Approach to Evaluating Length of
Stay, Occupancy, Emptiness and Bed Blocking in a Hospital Geriatric
Department. Health Care Management Science 1(2), 143-149.
El-Darzi, E., C. Vasilakis, T.J. Chaussalet and P.H. Millard
(2000) A simulation model to evaluate the interaction between
the acute, rehabilitation, long-stay care and the community. In:
Zanakis, S.H., Doukidis, G. and Zopounidis, C., (eds) Recent Developments
and Applications in Decision Making, pp. 475-485. Kluwer Academic
Publishers.
Garçia-Navarro, J.A. and Thompson W.A. (2001) Analysis
of bed usage and occupancy following the introduction of geriatric
rehabilitation care in a hospital in Huesca, Spain. Health Care
Management Science 4, 63-66.
Harrison, G.W. and P.H. Millard (1991) Balancing acute and
long-term care: the mathematics of throughput in departments of
geriatric medicine. Methods of Information in Medicine 30, 221-228.
Harrison, G.W. (2001) Implications of mixed exponential occupancy
distributions and patient flow models for health care planning.
Health Care Management Science 4, 35-43.
Hodkinson, H.M. & Jeffreys, P.M. (1972). Making hospital
geriatrics work. British Medical Journal, 4, 536-539.
Marshall, A.H. McClean, S.I.. Shapcott, C.M. Hastie I.R and.
Millard P.H (2001) Developing a Bayesian belief network for the
management of geriatric hospital care Health Care Management Science,
4, 25-30.
McClean, S.I. and Millard P.H. (1993) Modelling in-patient
bed usage behaviour in a department of geriatric medicine. Methods
of Information in Medicine. 32, 79-81.
Millard, P.H. (1994) Meeting the needs of an ageing population.
Proceedings of the Royal College of Physicians of Edinburgh 24,
187-96.
Millard P.H. (1998) The anatomy, physiology and biochemistry
of health care for an ageing population. Health and Hygiene 19:
49-60.
Millard, P.H. and T.J. Chaussalet (1998) A modelling approach
to the development of health and social services for dementia
care. Archives of Gerontology and Geriatrics 6, 325-334.
Millard, P.H. and Kist, P. (1995) Care of the long stay elderly
patient: the Bolingbroke project. In: Denham, M.J., (Ed.) Care
of the long stay elderly patient, London: Chapman and Hall.
Millard P.H., Mackay M., Vasilakis C. and Christodoulou G.
(2000) Measuring and modelling surgical bed usage. Annals of the
Royal College of Surgeons England. 82, 75-82.
Millard P.H., Christodoulou G., Jagger C., Harrison G.W.
and McClean S.I (2001) Modelling hospital and social care bed
occupancy and use by elderly people in an English health district.
Health Care Management Science 4, 57-62.
Taylor G.J., McClean S.I., Millard P.H. (1998) Using a continuous-time
Markov model with Poisson arrivals to describe the movement of
geriatric patients. Applied Stochastic Models and Data Analysis
14:165-174.
Taylor G.J., McClean S.I., Millard P.H. (2000) Stochastic
models of geriatric patient occupancy behaviour. Royal Statistical
Society (Series A) 493-503.
Thompson W.A., Chaussalet T.J. (2001) Subgroup analyses of
cost of care in a Markov model of the natural history of Alzheimer's
Disease. In: J.G. Anderson and M. Katzper (eds.) The Proceedings
of The Simulation in The Health and Medical Sciences 2001, pp.
41-44. Simulation Councils, Inc.
Vasilakis C., El-Darzi E. (2001) A simulation study of the
winter bed crisis. Health Care Management Science 4, 31-36
Xie, H., Chaussalet, T.J., Thompson W.A, Millard P.H. (2002)
Modelling decisions of a multidisciplinary panel for admission
to long-term care, accepted for publication in Health Care Management
Science.
Yates, J. (1982). Hospital beds: a problem for diagnosis
and management. London: William Heinemann.
24 January 2002
| 
