Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Estimating cost ratios and unit costs of public hospital care in South Africa revisited
1. Corresponding Author: Health Economics Unit, Department of Public Health and Family
Medicine, Health Sciences Faculty, University of Cape Town, Anzio Road, Observatory,
7925, South Africa. Tel: +27-21-4047701; Fax: +27-21-4488152; E-mail:
John.Ataguba@uct.ac.za; Qualifications: B.Sc.; MPH.
Abstract
Background
Reliable hospital unit cost estimates are limited in developing countries. Usually a simple rule
of thumb based on the assumption that the cost of an outpatient visit is equivalent to a fixed
proportion of the cost of an inpatient day is used to disaggregate unit costs. The objectives of
the paper are to obtain the ratio of cost of an outpatient visit to an inpatient day,and the
associated unit costs for different levels of public hospitals in South Africa.
Methods
Four levels of public hospitals were considered. A simplified model was used on data from
the South African District Health Information System to compute the ratio of the cost of an
outpatient visit to an inpatient day and the associated average financial costs at each hospital
level.
Results
An outpatient visit costs about 0.37 (district hospitals) to 0.64 (specialized hospitals) of an
inpatient day. Also the average financial cost of a visit (an inpatient day) ranges from R313
(R487) –district hospitals to R810 (R1441) – central or provincial tertiary hospitals.
Conclusions
The ratios of unit cost of outpatient to inpatient utilization used in computing the unit costs
vary across public hospital levels in South Africa. The need to continually update these ratios
and unit costs is noted.
Keywords: Unit costs; Public hospitals; Patient day equivalent; Outpatient visit; Inpatient
day.
1
Introduction facilities, usually district
hospitals. Kirigiaet al. [13]
Hospital unit costs are key
showed in the context of treating
ingredients in many policy
malaria in Kenya that the
decision making processes in
selective choice of hospitals
the health care sector. They may
could bias the results and
be used in assessing efficiency
conclusion about unit costs and
of units, treatments, and
cost saving. The most widely
facilities as well as for
available hospital level
budgeting and resource
expenditure data relates to
allocation [1-4]. They can also
recurrent and capital
be inputs to further analysis
expenditure breakdown across
such as benefit incidence
different expenditure categories
analysis (BIA) and economic
such as salaries and wages,
evaluation of health care
medical supplies, pharmacy,
programmes (e.g. cost-
training, purchase of equipment,
effectiveness analysis) [3, 5, 6].
etc. [4, 7, 11].
Despite the importance of such
costs, reliable estimates are
The standard methods of
rarely available in developing
allocating costs are the step-
countries [1-4, 7, 8]. This is
down method (or macro costing)
largely due to unavailability of
and the ingredients (or micro
reliable data, poor infrastructure
costing) approach. The step-
and poor record keeping culture
down method is usually
[4, 7, 8].When they are
considered as the ‘gold
available, they are often limited
standard’ [1, 2] while the
to a specific facility or facility
ingredients approach is
type [7, 9-11], or for specific
considered difficult to
diseases [12-14]or not regularly
implement because it may be
updated. Only few studies in
costly and time consuming and
developing countries [7, 11]
requires extensive data [15]. To
obtain detailed unit costs at
implement the step-down
facility level. Even at that, they
approach, several methods have
are still limited to selected
2
been proposed [2]. Conteh and country [7, 16]. It is well known
Walker [8] provide a summary that hospital costs “depend on
on how to proceed with the step- factors such as size, the level of
down approach. The most sophistication of services, the
straightforward of these is the quality of services, the types of
patient day equivalent (PDE) services offered, case-mix and
method. This is the popularly case-severity, and occupancy
used method in many African rates” [4 p.208]. Therefore such
countries [11]including South “simple rules of thumb do not
Africa. It is often routinely prove to be an accurate basis for
produced from hospital cost estimates … [as] ratio of
utilization and expenditure data. inpatient to outpatient unit costs
The key assumption upon which varies with …hospital size” [1
this is built is a simple rule of p. 1700]. As an alternative to
thumb that is seldom the rule of thumb, cross-country
empirically derived. Such a rule models may be fit to estimate
of thumb is based on the country specific costs [1].
assumption that the cost of an However, it is not able to
outpatient visit is equivalent to a account for potential differences
fixed proportion of the cost of across countries such as
an inpatient day [1, 7, 12, 16]. political, cultural, economic or
Usually, an inpatient day is social differences, and other
considered to be equivalent to factors that might have an
three or four outpatient visits [1, impact on unit costs.
16]. In some instances it could
This paper therefore presents an
be considered to be even higher
empirical method to compute
[17].
the ratios or patient day
The major criticism of this rule equivalents that are not based on
of thumb principle is that it is the simple rule of thumb. To
somewhat arbitrary especially achieve this, the paper extends
when the same ratio is applied the simple relationship model
to all types of hospitals (e.g. developed in Lombard et al.
district or central) within a [16]. The model is then used to
3
calculate the cost per inpatient democratic transition there were
day and per outpatient visit at 14 separate health departments
different levels of public operational in the country and
hospitals in South Africa. the post-apartheid government
was faced with huge challenges
The rest of the paper is of redressing existing
structured as follows. The next inequalities and inequities. All
section briefly introduces the the health administrations were
South African health system. amalgamated into one national
The section after this provides and nine provincial health
the data and methodology. departments [18] with leading
Thereafter the results are importance attached to primary
presented followed by the health care. Currently South
discussion and conclusion Africa operates a three-tier
sections. hospital structure (tertiary,
regional, and district) and
Brief overview of the South
primary health care system
African health system
(comprising clinics and
The South African health system
community health centres) run
consists of both private and
mainly by nurses [18, 20]. There
public sectors. The pre-
are over 200 district hospitals
democratic (pre 1994) period
and over 50 regional hospitals.
was characterised by the
For the public sector, the
apartheid system that impacted
national department of health is
on the South Africa health care
charged with the responsibility
system. During this period the
for overall guidance and
health care system was highly
national health policy and the
fragmented such that different
provincial departments are
groups have their own health
responsible for provincial health
department [18, 19]. Health
policy in line with the broad
services for the black majority
national policy framework.
were heavily underfunded and
They are the main providers of
the rural areas were neglected
health services through hospitals
[19]. During the period of
and primary care clinics [20].
4
Private health sector comprises It is important to note here that
general practitioners, private health facilities where there is
hospitals and traditional health no mix of inpatient and
care providers [20]. Though outpatient services, computation
total health care expenditure of unit costs are relatively
accounts for over 8% of the straightforward. This is
country’s GDP, the health sector obtained simply by dividing the
continues to face several equity total expenditure of the facility
challenges. For example over by the total number of output
50% of both financial and (e.g. total number of visits or
human resources are allocated to total number of patient days)
the private health sector [18] produced within a specified
and it is estimated that over two- period. Where there is a mix
thirds of private hospitals are between inpatients and
located within three of the nine outpatients, such as the various
South African provinces [18]. hospitals considered in this
Inequity also exists within and paper, the procedure is rather
across facility levels. While not straightforward.
about 11% of total public health
(a) Data
spending is devoted to non-
hospital primary care services The South African District
that cater mainly for the poor, Health Information System
over 44% is accounted for by (DHIS) 2006/2007 database was
academic and other tertiary used to extract information on
hospitals [18]. These challenges the total number of inpatient
further require that available days and outpatient visits (this
resource is used in an efficient includes both new and review
manner and that resource visits), the total number of
allocation issues are based on nurses and doctors in each
evidence. hospital at all hospital levels,
and the average bed occupancy
ratios. Data on recurrent
Methodology
expenditure in each hospital was
also extracted from the DHIS
5
database and supplemented with hospital levels (for example,
reports produced by the national central hospitals), data
National Treasury of South were combined and four broad
Africa. While efforts were made types of facilities were
to obtain data for all the nine distinguished for which ratios
South African provinces, two and unit costs were estimated:
provinces – North West and (1) district hospitals, (2)
Eastern Cape –were omitted regional hospitals, (3) national
from the analysis due to missing central and provincial tertiary
expenditure data in the former hospitals, and (4) specialized
and missing data on the total hospitals.
number of health care workers
in the latter. Further, some
(b) The model
hospitals within the remaining
Cost of care is related to all the
provinces had missing
characteristics of care [21].
information on one/more of the
Barnett [21] expresses this as a
variables of interest. These
linear function of the
hospitals were also omitted. The
characteristics of the patient and
dataset therefore contains
the characteristics of the health
hospitals with complete
facility.
information on the variables of
interest.
C = c(Xp,Yf)
6
obtain data on total expenditure whereCjis the unit cost of an
or total costs at the facility. inpatient day spent at facility
Facilities with a mix of inpatient levelj; Ij is the aggregate annual
admissions and outpatient number of inpatient days at
attendances, a procedure is facility level j, Oj is the
therefore needed to allocate aggregate annual number of
these expenditures to inpatient outpatient visits to facility level
and outpatient services, through j, and EXPjis the total annual
some proportioning, in order to recurrent expenditure at facility
calculate the unit costs (i.e. the levelj. Here the denominator is
average cost per inpatient day or the patient day equivalent in
outpatient visit). In order to inpatient days’ terms.
achieve this, this paper borrows
Similarly, we expressC'j– the
from an initial study in South
unit cost of an outpatient visit
Africa by Lombard et al.[16] as:
and extends the arguments
C'j= EXPj / (πj-1Ij+ Oj)
further by accounting for the
The denominator in equation
relative size of each facility in
measures the patient day
the estimation sample.
equivalent in outpatient visits’
Annual utilization (inpatient terms. We can manipulate
days and outpatient visits) and equation slightly to obtain
expenditures are considered to
EXPj= 1Ij+2Oj
avoid any fluctuation in use and
Where 1= Cj ;2= πjCj.
spending over the year. We then
If this identity holds, then we
assume that there is some ratio
can easily verify that
πjof the cost of an inpatient day
2/1 = πj
to an outpatient visit which is
specific to the facility level j This gives us back the initial
ratio πj.
(e.g. the district hospital). The
unit cost can be estimated as: A multiple regression model can
be fitted on equation such that
Cj= EXPj / (Ij + πj Oj)
the parameter estimates of 1
and 2can be recovered to verify
7
the appropriate ratioπj. However, controlling for the total number
this method does not control for of doctors and the total number
the variations in the sizes or of nurses (the sum of
scale of the facility. It assumes professional nurses, student and
that all facilities within a certain pupil nurses, and assistant
level are the same in scale and nurses) in each facility.Average
size (i.e. all district hospitals for bed occupancy ratio rather than
instance are of the same size). the number of beds was
considered as an indication of
To improve on this, and to
effective size. However, its
control for variations in scale
inclusion was not statistically
and size, we expressed as:
significant for some levels of
care. In the cases where it was
yij = 1xij1 + 2xij2 +kkzijk +
ijfor each j significant, the impacts on the
relevant ratios were not
whereyij is the total annual
significant. The regression is
recurrent expenditure in hospital
then estimated for each facility
i at level j (e.g. district
type j using theStata® 11 routine
hospitals), xij1and xij2 are the
with robust standard errors and
respective aggregate annual
the model was assessed for
number of inpatient days and
multicolinearityand
number of outpatient visits, 1
heteroscedasticity.
and 2are their respective
coefficients, zijkcomprises
After estimating equation for
variables that account for the each facility level (e.g. district
size of the facilities and k the hospital), we obtainπj and apply
associated coefficients, and k is equations and to obtain the unit
the total number of variables cost per inpatient day and
you wish to control for. These outpatient visit respectively.
zijk variables are used to account
for the relative size
heterogeneity across facilities. Results
Here, ij represents the error
In table 1 we present the results
term. We implement equation
of the regression output based
8
on equation. Though we report inpatient day (or the cost of one
robust standard errors, it is inpatient day is equivalent to 1.8
important to note that our times the cost of an outpatient
interest is not in the absolute visit). In specialized hospitals,
magnitudes of the coefficients an outpatient visit is about 0.64
but their relative magnitudes. times the cost of an inpatient
The coefficients on outpatients day (or the cost of one inpatient
and inpatients are all positive day is equivalent to 1.6 times
which implies that increasing the cost of an outpatient visit).
the number of inpatients days
will increase expenditure at each
facility level. If we apply
equation we obtain the ratios
contained in table 2.
9
Table 1: Regression output based on selected health facility variables
Dependent variable = Public hospitals categories
Total recurrent expenditure District Hospitals Specialized Central and Provincial Regional
(DH) Hospitals (SH) Tertiary Hospitals (CPH) Hospitals (RH)
208.6*** 493.1*** 495.7*** 283.5**
Outpatients
(54.81) (154.50) (167.11) (125.65)
562.3*** 766.8** 882.4** 677.9***
Inpatients
(93.66) (375.47) (365.57) (185.75)
540043.3*** 1088743.4*** 1033221.0*** 598368.0***
Number_of_doctors
(132263.91) (140931.57) (134898.98) (204372.85)
43923.6** -96614.4 -117343.5 18024.0
Number_of_nursesa
(21902.31) (105818.78) (94492.40) (39512.87)