Organizing home care in the future:
using system dynamics to assess organizational changes
Hans A.Kuipers
Jac.A.M.Vennix
University of Utrecht
Department of Gamma-informatics
P.O. Box 80140
3508 TC Utrecht
The Netherlands
Sjoerd Kooiker
Nivel, Utrecht
Abstract
In the near future the organization of home care in the Netherlands will be reorganized. In order to
show some of the dynamic consequences of these changes, a preliminary model was developed. In
this paper we will discuss the use of the preliminary model to elicit the ideas from policy makers
about future changes in the organization of home care. This is done by conducting a Delphi study to
keep the time investment of the policy makers as limited as possible.
Introduction
In the Netherlands the organization of home care is one of the main policy issues in primary
health care. Due to the aging of the population the demand for health-care increases because elder
people relatively need more health-care than younger people. The ageing of the population can't be
influenced but by giving care outside hospitals one can prevent the hospitals from becoming
overcrowded. One of the most likely options in this respect is to give certain care at home instead
of giving these in a hospital. Such a shift, from secondary to primary health care, will have great
impact on the organization of home care. It is unavoidable that the organization has to change
dramatically. One of the major problems in changing such a complex organization is the large
number of unknown consequences. For instance with regard to the number of workers, needed in
this new organization and with regard to possible consequences for the quality of care. These
problems are of great concern for policy makers. In order to assess potential effects of changes in
the organization of home care policy makers invoke their mental models of the system. Mental
models, however, are often incomplete and little accurate. Furthermore mental models are not
capable in establishing the difficult dynamic consequences of large complex structures, such as
the organization of primary health care (Forrester, 1961 ; Meadows et. al., 1982). A computer seems
more suited to calculate such dynamic consequences. Building a computer model can assist in
assessing potential effects of such organizational changes. But the construction and use of a model
does not necessarily induce organizational changes. In order to accomplish this it is important to
have the ‘problem-owners'’ participate in the process of model-building (Meadows et. al., 1982;
Meadows and Robinson, 1985). Most policy makers, however, have lack of time which means that
full participation is often impossible. In this paper we will discuss a way of having people
participate in the process of modelbuilding without asking too much of their time. In order to
realise this we chose for the preliminary model approach (Vennix et. al., 1988). In this approach a
project group designs a preliminary model and next asks relevant policy makers to criticize and
possibly improve this preliminary model. To elicit policy makers’ ideas on this preliminary
System Dynamics '90
597
model we will employ the Delphi method (Linstone and Turoff, 1975). The whole process consists of
four stages: (figure 1)
figure 1 The preliminary model approach
First we will develop a preliminary model (1)
of the organization of home care. This model
will be criticized by participants in the first
Delphi cycle (2). The results of the first cycle
will be fed back in the second cycle (3) in which
we will next ask for potential organizational
Sra Dagh | changes. Finally all the participants in the
Delphi will be brought together in a workshop
(4) in which they are able to simulate some of
da ptatien the organizational changes with the aid of a
STELLA model and discuss the potential im-
pact of these changes with each other.
In this paper we will first focus on the place of
the model in a scenario study of which it is part
(section 1), next we will discuss the prelimi-
nary model (section 2) and the process of con-
dee sulting a number of experts with the aid of the
Delphi technique (section 3).
‘workshop
@
1, The scenario study
The model-building process is part of a broader scenario study (Wennink, Kooiker; 1989) which is
conducted by the Netherlands Institute for primary health care (NIVEL). This scenario is con-
ducted to investigate possible scenario's about the organizational structure of future home care. In
order to make the necessary estimations two different studies are distinguished: (1) Trend study
and (2) Care study.
In the trend study a statistical model is developed to predict the future use of primary health care
facilities in the Netherlands (Kooiker, 1989). With the model three estimations are made of pri-
mary health care use for the years 1990-2005. A demographic extrapolation was used as a bench-
mark estimate. A trend extrapolation consisted of an estimate of the size of the user groups and the
expected enlistment in each group. Finally these estimations are compared with the expected sup-
ply of primary health care over the same period to find potential discrepancies.
The basic goal of the care study is to generate possible shifts in the use of home care. These shifts
may take place within primary health care or between primary and secondary health care facili-
ties. The object of investigation will be the chronically ill because they are considered most.
important for the use of future home care.
One of the estimations of the trend study is the number of workers necessary to supply a certain
amount of care in the near future. These numbers are necessary in order to be able to develop sce-
narios about the organization of future home care. However it is likely that there are more rela-
tions than the influence of future use or the demographic development which influence the number
of workers. In the preliminary model we will show some of these relations.
598 System Dynamics '90
1.1 The model
As the previous text shows the scenario study is rather static, it does not account for possible
(system) dynamic consequences of a change in the organization of health care. However, if the
organization of a system is changed, unforseen effects often occur. The building of a system dy-
namics model seems an appropriate way to give policy makers more insight in present feedback
processes and their effect on a (possible) future change in the organization of health care.In the
trend study a trend extrapolation estimates the expected supply of primary health care. In the pre-
liminary model we will also give an estimation of the future supply in primary health care, i.e.
the number of workers necessary to supply a certain amount of care.
In the next sections we will explain the process of model-building and the use of the model in the
Delphi workshop. Before this we will explain the preliminary model.
2. The preliminary model
In order to be able to make a model which will provide insight in the processes which influence the
number of people working in primary health care, it is necessary to locate the factors which influ-
ence these numbers. According to a study of Bensing (Bensing et. al., 1987) there are six possible
factors: (1) the influence on supply, (2) the supply itself, (3) the influence on demand, (4) the
demand itself, (5) policy and (6) the influence on policy.
Of course it is not possible to take all these factors into account in a model. In the preliminary
model we will make a selection of the factors which have most influence on the number of work-
ers. The preliminary model consists of three sub-models, which are interrelated: (1) Demand, (2)
Supply and (3) Finance.
2.1 Demand
This part of the model consists of three age categories of chronically ill. The first group is between
20 and 54 years old, the second between 55 and 74 years old and the last consists of chronics who are
over 75.
figure 2 The patient categories
intlow flows tlow2 outtiow
The number of patients, who will flow from one level to the next is defined by the average time in
each level. For the category 20-54 this is 35 years for the category 55-74 this is 20 years and for the
category 75 plus this is 6 years. The flows ‘new 1,2,3' are the number of people who become chronic
after the age of 30 or 55 or 75. The flows ‘death 1,2' and ‘outflow’ are the patients who die before
reaching the next age category.
System Dynamics ‘90
figure 3 Patient flow and their demand for care
chronic 75plus.
intlow tlow2
outflow
contacts3
norm_cont_20_54 contacts1 norm_cont_7Splus
total_contacts
For each age category the amount of care they need is specified by the number of contacts they
monthly ask for (contacts 1,2,3). The sum of these contacts produces the total number of contacts,
which will be the input for the rest of the model. In the preliminary model the developments in de-
mand are still independent from the developments in supply.
2.2 Supply
As the goal of the scenario project shows, the number of workers necessary to supply a certain
amount of care is a central part of the project. Hence it is also the central part of our model. In this
model, however, we will concentrate on the number of community nurses (nurses). The processes
as described in the model are similar for other kinds of workers, such as home helps.
figure 4 Desired community nurses in primary health care and the number of contacts they can
realise.
600 System Dynamics ‘90
This part is the second sub-model: the number of nurses in primary health care and the number of
contacts they can supply. As one can see this sub-model contains two parts: (1) nurses and (2) con-
tacts.
They are however so closely interwoven that discussing them as different parts is almost impossi-
ble. The number of nurses is related to the number of hired nurses and the number of dismissed
nurses. The number of nurses is also influenced by the shortage of contacts. If the number of
nurses can not meet the demands of the patients, more nurses are needed. As will be explained
later the possibility of engaging new nurses depends on the money available. Another problem is
the impossibility to hire new nurses immediately: in order to do their work they will need train-
ing, which will take rather some time. If there are not enough nurses to provide the care which is
asked for, this is a problem because one can not give a patient less care. The consequence is that the
available nurses need to work harder, in other words the workload will increase. It will take time
before nurses recognise this higher work load as a structural problem. The zest for work will de-
crease especially when a higher workload keeps on for a while. Some nurses will ask for resigna-
tion, which will lead to an even higher workload for the remaining nurses. Literature shows that
at a structural higher workload about sixteen percent (yearly) of the nurses will ask for resigna-
tion (Dogger, 1988). Of course the increase in workload has an upper limit. The increase in the
number of contacts is maximally 50 % more than the contacts a worker will do under normal
circumstances. Also the number of nurses which will ask for resignation is limited, some of them
will keep on working even at a higher workload.
One of the striking things the model shows is that, after a cut of subsidies by the Government, the to-
tal average years of experience will decrease. More people are dismissed and more people will be
hired. This will have a major impact on the quality of care and in the end on the health of the pa-
tients.
2.3 Finance
The financial sub-model is kept simple because an appropriate financial model was not avail-
able. In practice this will not be a major problem because just as in reality (in the Netherlands) up
to now all the money, which is asked for, will be paid by the Government. Furthermore, the build-
ing of a suitable financial model is of no use because major changes in the way of subsidizing will
take place in the near future.
figure 5 Financial model
comfnurses
costs_comnurs
offect_budget
dasired_budget
ae nee costs_comnurs
The major costs in primary health care are the wages of the nurses. The expenses in the financial
model are calculated by the number of nurses * wages per worker. The income (i.e. the subsidy) is
System Dynamics '90 601
calculated by the difference between the money available (budget) and the desired budget. If the dif-
ference is negative, subsidies are provided, if the difference is positive, subsidies are equal to zero.
2.4 The overall model
outtiow
norm_cont_20_54 contacts1 contacts3 — norm_cont_7Splus
total_contacts
perceived_t\t cont
desired_contacts
realised
costs_worker
wanted_finance
costs_comnurs
602 System Dynamics ’90
Although we have discussed the three sub-models, it is necessary to discuss the relations between
the sub-models. The most important relation which should be addressed is the relation between the
financial- and the nurses sub-model. This relation is the "effect-of-budget". When the balance of
budget is positive new nurses can be hired.Unfortunately in practice this is not very likely. The
reason for using the effect of budget in the model is its impact on resignations when the govern-
ment is going to cut subsidies. When there is a cut of the subsidies, one is not able to hire new
nurses.
Nevertheless the desire for care will grow in the near future, because the number of chronics is
growing (ageing effect) and it is likely that there will be a shift from the secondary to the primary
health care facilities. This means a growing workload for the remaining nurses, which as the
model shows, will lead to more resignations. If this situation keeps on for a wile, it could endanger
the quality of care as we have explained above.
3. Participation of relevant policy makers.
As noted before we will conduct a Delphi study in order to have experts criticize and improve the
preliminary model. In literature there is a distinction between two different types of Delphi, the
classical Delphi and the policy Delphi. In the process of model-building we will use the policy
Delphi. As we want to clarify the differences between the different groups involved, we use a policy
Delphi because , contrary to classical Delphi, in a policy Delphi consensus is not necessary. In this
process confrontation of the panel members is necessary, which is the reason for organizing a
workshop at the end of the Delphi study. In this section we will discuss each Delphi cycle and the
use of the workshop.
3.1 The first Delphi cycle
In the first cycle we want the panel-members to give there comment on the preliminary model.
Before they are able to give this comment they need to know something about system dynamics and
the technical aspects of STELLA (Richardson et. al.,1989). For this purpose we will develop a
workbook (Vennix et. al.,1988). The first part of the workbook will contain an explanation of the
principles of system dynamics and a brief explanation of the STELLA symbols. The second part of
the workbook will contain an explanation of the three different sub-models and the overall model.
Subsequently the participants will be asked to answer a number of questions. The questions asked
will be aimed at eliciting knowledge on concepts and relationships between concepts as contained
in the STELLA model. In other words participants will be asked whether they agree or disagree
with the relationships between the concepts, i.e. with the way we conceptualized the home care
system. The next question aims at finding out whether participants consider the model incomplete
on some aspects. They are encouraged to indicate whether certain important characteristics are
missing in the model. Next they are invited to actually draw in missing elements in the diagrams
presented in the workbook. This information will be used by the project group to adapt the
preliminary model.
3.2 The second Delphi cycle
In this second cycle we will according to the comments that were made in the first cycle report
about the results of the first cycle and the way we adapted the preliminary model to take their
comments into account. We again employ a workbook in which we will ask participants to
comment on the adapted model in order to arrive at a final model which can be used in the
workshop to conduct a number of policy experiments.
System Dynamics '90
In the second part of the workbook we will switch to the question of potential organizational
changes in the home care system.
Currently there are three options for such changes which are most likely. First, a merging between
the organizations of community nurses and home helps. Second, a merger between the
organization of home care and elderly homes and third the commercializing of home care. Either
of these changes has great impact on the number of workers necessary and the number of workers
available. For example when home care is partly commercialised especially the highly
experienced workers with a lot of experience will be hired by these institutes. Because a
commercial bureau is able to pay higher wages they will succeed in snatching the experienced
workers from social home care. This will have a negative influence on the quality of social home
care, Another problem which will arise with commercial home care is the changing financial
situation for social home care. At present everybody is paying a percentage of his salary for social
home care. When people with higher salaries are going to use commercial home care instead of
social home care it is likely they don't have to pay their contribution to social home care because
they don't make use of it. This, however, will endanger the financial reserves of social home care
. This and other consequences of organizational changes need to be prevented. In the workbook
these three scenarios and their potential consequences will be sketched.
We will ask the participants if they agree with the consequences as we see them. They are encour-
aged to criticize and add consequences. The participants will also be asked to provide other possi-
ble organizational changes in the future and consequences of these changes. Finally the partici-
pants will be asked to indicate which of these changes they consider most likely in the future of
home care. The project group categorizes and summarizes the various proposals of the participants
and presents these in the final step of the Delphi, the workshop.
3.3 The workshop
In the workshop participants will be in a position to actually implement some organizational
changes and study its effects with the aid of the Stella model.
The workshop will start with an introduction in which the goals of the workshop will be explained.
After this introduction we will discuss the results of the second Delphi round, especially the likeli-
hood of the various organizational changes the participants indicated in the workbook. In com-
munication with the participants we will choose the most likely of these organizational changes to
conduct experiments with.
The experiments will be conducted by small groups of three participants. Each group will conduct a
different policy experiment. The groups will be asked to write down the results of the experiments.
The workshop will be wrapped up with a plenary session in which the subgroups will report about.
the results of the experiments they conducted.
Because, as we noted before, the time of the participants is limited and therefore the workshop
should not take any longer than about four hours. Thats why we will divide the workshop as
follows: one hour for the introduction and the discussion of the results of the second round and two
hours for conducting experiments and one hour for the plenary session.
Summary and discussion
In this paper we have discussed a system dynamics model to assess the effects of organizational
changes in the home care system in the Netherlands. In order to limit the time investment from
the part of the policy makers who are invited to participate in the model-building process we use the
preliminary model approach. The Delphi technique is employed to consult a number of policy
makers on this preliminary model. The purpose of this is threefold. First to elicit their criticism
with regard to this preliminary model. Second to extract their ideas of likely future changes in the
organization of home care and its potential consequences. Third in a workshop session
604 System Dynamics '90
participants will be in a positions to use the Stella model to calculate the effects of organizational
changes.
A number of policy makers has agreed to participate and currently we are producing the relevant
workbooks and questions and are designing the procedure for the workshop.
Literature
Dogger, J., Werk in de wijk, werk voor de wijk, in: T.S.G., nr.1, 1988
Forrester, J.W., Industrial Dynamics, Cambridge (Mass.), 1961.
Geurts, J.L.A., J.A.M. Vennix (red.), Verkenningen in beleidsanalyse: theorie en praktijk van
modelbouw en simulatie, Kerckebosch, Zeist, 1989
Kooiker, S.E., Methodologische overwegingen bij het gebruik van survey-gegevens in de trend-
studie, Utrecht, 1989
Linstone, H.A., Turoff, M., The Delphi Method "Techniques and Applications”, Massachusetts,
1975
Meadows, D.H., J. Richardson, G. Bruckmann, Groping in the dark: the first decade of global
modeling, John Wiley & Sons, Chichester/New York, 1982
Meadows, D.H., J.M. Robinson, The electronic oracle: computer models and social decisions,
John Wiley & Sons, Chicester/New York 1985
Richmond, B.M., P. Vescuso, S. Peterson, STELLA for business,. High Performance Systems
Inc., Highway, Lyme, NH 03768, 1987.
Vennix, J.A.M., J.L.A. Geurts, Communicating insights from complex simulation models: a
gaming approach, in: Simulation & Games, vol. 18, no. 3, sept. 1987, p. 321-343
Vennix, J.A.M., R.E.H.M. Smits, J.L.A. Geurts, De funktie van computersimulatie: inleiding in
de systeemdynamica. In: Geurts, J.L.A., J.A.M. Vennix (red.), Verkenningen in beleids-
analyse: theorie en praktijk van modelbouw en simulatie, Kerckebosch, Zeist, 1989, p. 75-92
Vennix, J.A.M., Mental models and computer models: design and evaluation of a computer-based
learning environment. Diss., K.U. Nijmegen, 1990.
Vennix, J.A.M. & W.J. Scheper, Modeling as organizational learning: an empirical perspective.
In: Proceedings of the 1990 International System Dynamics Conference, Boston (Ma), 1990.
Wennink, H.J., Kooiker, S.E., Onderzoek naar thuiszorg voor categorieén chronische patiénten,
Utrecht, 1989.
CC BY-NC-SA 4.0