Analyse thepatient-related and non-patient-relatedcosts of intensive care using anactivity-based costing methodology.

central file server on which all data are stored. Data from the bedside

monitoring equipment and therapeutic interventions are automatically

acquired and other data - for example, drug prescriptions

- are manually entered through the bedside terminals by the

nursing and medical staff. Demographic, admission and discharge

data are also routinely collected for all patients and stored on the

PDMS.

The activity-based costings methodology has been implemented

in the ICU of the Royal Hallamshire Hospital and since

995 has been used for costing intensive care patients [2]. This system

measures the patient-related costs of delivering care, i. e. costs

attributable to individual patients. These costs include nurses, medical

ward rounds, drug treatments, disposables, equipment utilisation

and clinical support services such as physiotherapy, radiology

and laboratory services. The care delivered to a patient is partitioned

into discrete elements termed "activities of care". An activity

of care is defined as any patient-related task requiring the use of

ICU resources. There are currently over 300 activities of care configured

on the system which not only include all the drugs, treatments

and major monitoring procedures used on the ICU but

which also include the "background" nursing care and discussions

with relatives. Patient care is costed using the activities of care by

identifying the resources (and therefore costs) necessary to deliver

each activity. For example, when a bolus drug is given to a patient,

the resources used include the staff time, a syringe and needle, the

drug and any diluent used. The total cost of this activity is the sum

of these individual resource components. The activities of care

which have been delivered to an individual patient are obtained

from clinical data stored within the PDMS. The total patient-related

cost of care for an individual patient is determined from the

sum of the costs of the activities delivered to that patient.

Non-patient-related costs are also incurred in running the ICU.

Examples of patient-related and non-patient-related costs of these

are detailed in Table 1. An hourly overhead rate is calculated from

the sum of the non-patient-related costs and the total bed occupancy

for the previous financial year, which allows these costs to

be apportioned to individual patients.

For each day of stay, a "daily care record" is produced which

documents the activities of care that patient has received, together

with the costs of that care. The daily care records for each patient

are aggregated into a "summary of care" record which summarises

both the total care delivered and the cost. The daily care and summary

of care records support both audit and cost analyses which,

when combined with presenting pathology and outcome data,

make economic evaluations of intensive care feasible for the first

time.

Using data from the activity-based costings system, the relative

contribution of the different activities of care to the total patientrelated

costs of care have been analysed. The costs of patient care

are sometimes calculated "per calendar day". Using the first calendar

day to determine patient costs for the initial period of care produces

results that cannot be compared for individual patients because

the length of time they have spent on day 1 on the ICU will

vary considerably. The costing system calculates the costs both for

24-h periods and for calendar days. The difference in determining

the patient-related costs of care for calendar days and 24-h periods

for the first day of stay was analysed using aWilcoxon Signed Rank

test. Changes in the daily patient-related costs of care with duration

of stay were investigated using a Kruskal-Wallis test (a nonparametric

equivalent of the one-way analysis of variance) applied

to the costs of care measured for successive 24-h periods. Specific

differences between consecutive days were then identified by analysing

the difference in rank sums using the method of Thomas and

Nelson with the Bonferroni modification of a p ??0.05 confidence

interval for multiple comparisons [11].

In order to investigate the possibility that the costs of care during

the first 24-h for long-stay patients were different to those of

short-stay patients, the costs of care for the first 24 h for patients

646

Table 1 Examples of patient-related and non-patient costs

Major sources of

patient-related costs

Major sources of

non-patient-related costs

Drugs Energy

Fluids Heating

Consumables Building maintenance

e. g. Syringes Engineering maintenance

Needles Capital charges

Catheters Portering services

Equipment usage Cleaning and laundry

Laboratory services Rates

Medical imaging services Estates

Nursing time delivering

patient care

Nursing time not delivering

patient care

Medical time delivering

patient are

Medical time not delivering

patient care

whose duration of stay was L10 days were compared with the

costs of care for the first 24 h of care for patients whose duration

of stay was ??10 days using a Mann-Whitney U test.

Both workload and severity of illness scores have previously

been used to cost patient care on the ICU [7-9]. In order to investigate

the relation of these scores to the patient-related costs of care

determined by the activity-based costing system, a Spearman rank

correlation was used to investigate the relationship between the

patient-related costs of care and the Therapeutic Intervention

Scoring System (TISS) [12] scores and the Acute Physiology and

Chronic Health Evaluation (APACHE) II scores [13].

The activity-based costing methodology determines the patient-

related or direct costs of care for individual patients. The total

costs of care for an individual patient are the sum of patient-related

costs of care and a proportion of the non-patient-related costs

associated with running the ICU. The identification and quantification

of the non-patient-related costs are described, together with

the determination of an overhead rate which was used to apportion

these costs to individual patients. From the total costs of care for

each patient in the study, the average daily costs of care on our

ICU have been determined.

Results

All 68 patients admitted to the ICU from 1 October to

8 December 1995 were included in the current study

(age range 16-82 years; mean ??SD 54.7 ??18.5 years).

The length of stay ranged from 0.2 to 35.7 days

(mean ??SD 5.3 ??7.6 days); 46% of the patients were

postoperative and the primary pathologies relating to

the ICU admission are presented in Table 2. Admitting

Apache II scores ranged from 5 to 41 (mean ??SD

21 ??8.6).

Individual cost components

The percentages of the individual components of the total

patient-related costs of the care delivered to all patients

are shown in Fig. 1. This illustrates that the background

nursing care and medical ward rounds, which

we term the general care, make the most important contribution

to the total patient-related costs of care

(44%). All patients receive these components of care

regardless of their specific treatment requirements.

Drug therapy and fluid electrolyte and nutrition therapy

constitute the next most significant components, contributing

23% and 19% of the patient-related costs of

care, respectively. Cardiovascular and respiratory therapy

contribute approximately 5% each to the total

costs, while administration, which includes the resources

necessary to admit and discharge patients, contributes

4%of the total patient-related costs of care.

First day of stay

The analysis of the difference between patient-related

costs of care for the first 24 h and for the first calendar

day of stay produced the following results: the mean and

standard deviation of the first 24-h costs for all patients

was £ 723.50 ??328, while for the first calendar day the

cost was £ 420.95 ??310, a difference in the mean values

of £ 303.45. A Wilcoxon Signed Rank test showed this

difference to be statistically significant (p ??0.001).

Comparison of costs and duration of stay

The distribution of the 24-h costs of care throughout duration

of stay (DoS) in the ICU has been calculated for

the first ten 24-h periods for all patients with a DoS

longer than 24 h (Fig. 2). The Kruskal-Wallis test

showed that the costs of care changed significantly with

24-h period of stay (p ??0.0001). Exploring this difference

by analysing the rank sums of the patient-related

647

Table 2 The principal pathologies of the 68 patients on admission

to the ICU using the pathology classes from the ICD9 disease classification

system

ICD 9 pathology class No of patients

Blood and blood-forming organs 3

Cardiovascular system 12

Digestive system 13

External causes of injury 3

Infectious diseases 4

Neoplasms 2

Nervous system 2

Poisoning 3

Respiratory disease 23

Trauma and accidental injury 3

Total 68

Fig. 1 The relative importance of the different components of care

costed (CVS cardiovascular)

costs of care for different 24-h periods showed that it arose

from differences in the patient-related costs of care

between the first and second and between the first and

third 24-h period of stay (p ??0.05). None of the other

comparisons reached statistical significance (p ??0.05).

A comparison of the patient-related costs of care for

the first 24-h period for those patients with a DoS

O 10 days (£ 743 ??294) with the patient-related costs

of care for the first 24-h period for those patients with a

DoSL10 days (£ 958 ??280) showed a significant difference.

When the patient-related costs of care for the first

and second 24-h periods were compared using a Mann-

Whitney U test with the long-stay patients removed,

the difference was still statistically significant

(p ??0.001).

Comparison with the TISS and APACHE II scores

A highly statistically significant correlation (p ??0.0001)

was found between the cost of care for each calendar

day, excluding the first day of stay, and the corresponding

daily TISS score for all patients (Fig. 3).

The APACHE II score was calculated for the first

24 h of the patient stay, and this was plotted against the

cost of the first 24-h period (Fig. 4). A highly statistically

significant correlation (p ??0.004) was found.

The total costs of patient-related and non-patientrelated

care in intensive care

The non-patient-related costs, such as rates, utilities and

energy, were calculated by apportioning the total hospital

bill by the percentage of floor area that the ICU occupies.

The non-patient-related nursing costs were calculated

using the annual nursing expenditure for the

ICU less the cost of patient-related nursing care. The

overhead cost per patient day was found to be £ 560.

The "average" daily cost of direct care was £ 592 per patient,

making the "average" cost per patient day £ 1152.

Discussion

The increasing financial pressures on health care mean

that clinical services will have to become more cost efficient

in their use of resources, and intensive care is no

exception to this. Since the costs of care for individual

patients in intensive care are both high and very variable,

it is important that an accurate costing methodology

is developed which allows the costs incurred in delivering

care to be analysed in detail, if decisions are to

be made with regard to appropriate resource allocation.

An essential component of the methodology is that it

can be used to undertake economic evaluations of different

treatment modalities. In this paper, we have presented

the initial results for an activity based costings

methodology which measures the patient-related (direct)

costs of care within the ICU. The data presented

on the different components of the costs of care illustrate

the distribution of costs and provide information

on which to base financial management targets for the

environment.With 44% of the direct costs contained in

background care, this is clearly an area where the com-

648

Fig. 2 Median and interquartile ranges for the total patient-related

costs of care, measured over consecutive 24-h periods

Fig. 3 Daily TISS scores versus calendar day costs of care (excluding

day 1)

Fig. 4 APACHE II score versus first 24-h costs of care

ponents which make up that cost require further analysis.

Previous studies [14, 15] have examined the first-day

costs of care in intensive care. However, it is not clear

whether these costs are derived using the first "calendar"

day or are from the first 24-h period. Clearly,

whether the difference in the costs obtained is significant

is dependent on the costing methodology used.

Our results obtained from the activity-based methodology

show a statistically significant difference in the costs

obtained depending on which definition of the "first

day" is used. This result highlights the need to include

explicit definitions of time periods used if comparisons

between the results obtained from different studies are

to be meaningful.

The patient-related costs of care measured over consecutive

24-h periods were used to study the distribution

of costs throughout the duration of stay. The results

showed that the costs of care for the first 24-h period

were significantly higher than the costs of care for the

second 24-h period. Thereafter, there was no significant

difference between the costs of care for other adjacent

24-h periods. Although the patient-related cost of care

for the first 24-h period from patients with a duration

of stay longer than 10 days was significantly higher

than that from patients whose duration of stay was less

than 10 days, this difference is not sufficient to account

for the large difference found between the costs of care

for the first and second 24-h periods, since the statistical

significance remains when the long-stay patients are removed

from the comparison. It is commonly assumed

that high costs will be incurred at the beginning of a patient's

stay, which reduce with length of stay [14]. The

results from our study show the patient-related costs of

care to be high initially, and then fall rapidly to a constant

level.

At present, the TISS score is used as a tool for costing

intensive care patients in the United Kingdom [8].

While our results demonstrate a highly statistically significant

correlation between the TISS score values and

the daily costs of care, there is a high variance in the

data (Fig. 3), which makes the predictive power of the

TISS score poor in determining the costs of care for an

individual patient.

In common with many other ICUs, the TISS scoring

system used in our ICU has been modified to reflect

modern intensive care practice and thus cost comparisons

between ICUs based on TISS scores are fraught

with difficulty. The modified TISS score used in this

study differs from the version published by Cullen et al.

[12] in that the components for renal support and ventilation

have been expanded to take into account the

changes in clinical practice since it was last validated in

983 [16]. It would be possible to construct an activitybased

costings system based on the TISS score by determining

the costs required to deliver each element within

it. However, the elements are general (e. g. vasoactive

drugs) and would not support the detailed analysis of

care (for example, the cost consequences of using specific

drugs), which is possible with the system reported

here. In addition, the TISS score is determined once a

day, which would preclude a detailed analysis of the different

treatment regimens where the principal difference

lies in the duration of treatment.

Our results show a statistically significant correlation

between APACHE II scores and the patient-related

costs of care for the first 24-h period. As in the case of

the TISS score, the variance in the patient-related cost

data is very high, which means that the predictive power

of the correlation for the costs of care for an individual

patient are poor. Therefore, our results only partially

confirm those of Ridley et al. [17], who suggested that

the APACHE II severity of illness score taken on admission

can be used to predict the cost of treatment for

the first 24 h.

Previous studies where non-patient-related costs

have been reported have shown a wide variation in

these. Ridley et al. [18] reported fixed costs which were

1% of the total costs from a study carried out in the

United Kingdom, Holt et al. [19] reported that 26% of

the total costs for ventilated patients were fixed from a

study carried out in Australia and Halpern et al. [20] reported

that indirect costs were 33% of the total costs

from a study carried out in the United States. In this

study, we report non-patient-related costs of 49% of

the total costs of care. We use percentages to compare

these costs rather than their absolute value to remove

effects due to inflation and different currencies. The

wide variation in these reported costs suggest that there

is a need for a standardised approach for categorising

the different resource components which contribute to

the costs of care.

Singer et al. [21] quoted average costs per patient in

the ICU as £ 1148 per patient day, but did not report

the variation in costs across patients. The average daily

cost of care in our study was £ 1152, with a standard deviation

of £ 243. While these figures are clearly comparable,

they must be treated with caution: firstly, the distribution

of costs is non-gaussian and therefore the

mean is an insufficient descriptor of the average costs

of care; secondly, the effects of casemix on these figures

are unknown and represent an area where further research

is required. In addition, there is a need for an approach

to express the costs of care which is both country-

independent and independent of inflation, so that

the results of studies undertaken in different countries

and at different times can be compared. Finally, our

findings support the conclusion of Gyldmark [22] that

there is a need for a standardised model for costing

ICU care.

649

650

References

. Office of Health Economics (1995)

Compendium of health statistics

(10 population, mortality and morbidity

statistics), 9th edn. London: Office of

Health Economics

2. Edbrooke DL, Wilson AJ, Gerrish SP,

Mann AJ (1995) The Sheffield costing

system for intensive care. Care Crit Ill

1: 106-110

3. Mennitz PGH, Lenz K (1995) Patient

data management systems in intensive

care - the situation in Europe. Intensive

Care Med 21: 703-715

4. Buist M (1994) Intensive care unit resource

utilisation. Anaesth Intens Care

22: 46-60

5. Bams JL, Miranda DR (1985) Outcome

and costs of intensive care. Intensive

Care Med 11: 234-241

6. Byrick PJ, Mindorff C, McKee L,

Mudge B (1980) Cost-effectiveness of

intensive care for respiratory failure patients.

Crit Care Med 8: 332-334

7. Loes O, Smith-Erichsen N, Lind B

(1987) Intensive care: cost and benefit.

Acta Anaesthesiol Scand 31 [Suppl 84]:

3-19

8. Atkinson S, Bihari D, Smithies M, Daly

K, Mason R, McColl I (1994) Identification

of futility in intensive care. Lancet

344: 1203-1206

9. Zimmerman JE, Shortell SM, Knaus

WA, Rousseau DM,Wagner DP, Gillies

RR, Draper EA, Devers K (1993) Value

and cost of teaching hospitals: a prospective,

multicenter, inception cohort

study. Crit Care Med 21: 1432-1442

0. Finkler SA (1982) The distribution between

costs and charges. Ann Intern

Med 96: 102

1. Thomas JR, Nelson JK (1990) Research

methods in physical activities.

Human Kinetics, Champagne, Illinois,

pp 190-191

2. Cullen DJ, Civetta JM, Burton A,

Briggs BA, Ferrara LC (1974) Therapeutic

intervention scoring system: a

method for quantitative comparison of

patient care. Crit Care Med 2: 57-60

3. Knaus WA, Draper EA, Wagner DP,

Zimmerman JE (1985) APACHE II: a

severity of disease classification system.

Crit Care Med 13: 818-829

4. Drummond MF (1980) Principles of

economic appraisal in health care. Oxford

University Press, Oxford

5. Ridley S, Biggam M, Stone P (1993) A

cost benefit analysis of intensive therapy.

Anaesthesia 48: 14-19

6. Keene R, Cullen DJ (1983) Therapeutic

Intervention Scoring System: Update

983. Crit Care Med 11: 1-3

7. Ridley S, Biggam M, Stone P (1994) A

cost utility analysis of intensive therapy.

Anaesthesia 49: 192-196

8. Ridley S. Biggam M, Stone P (1991)

Cost of intensive therapy. A description

of methodology and initital results. Anaesthesia

46: 523-530

9. Holt AW, Bersen AD, Fuller S, Pipers

RK, Worthley LIG, Vedig AE (1994)

Intensive care costing methodology:

cost benefit analysis of mask continuous

positive airway pressure for severe cardiogenic

pulmonary oedema. Anaesth

Intensive Care 22: 170-174

20. Halpern NA, Bettes L, Greenstein R

(1994) Federal and nationwide intensive

care units and healthcare costs:

986-1992. Crit Care Med 22: 2001-

2006

21. Singer M, Myers S, Hall G, Cohen SL,

Armstrong RF (1994) The cost of intensive

care - a comparison on one unit between

988 and 1991. Intensive Care

Med 20: 542-549

22. Gyldmark M (1995) A review of cost

studies of intensive care units: problems

with the cost concept. Crit Care Med 23:

964-972