International Journal for Quality in Health Care 2000; Volume 12, Number 2: pp.

125131
Can improving quality decrease hospital
costs?
AGNES JARLIER AND SUZANNE CHARVET-PROTAT
Department of Health Economics, ANAES, Paris, France
Abstract
Purpose. To determine whether the concept of cost of quality and the techniques used for its study in the industrial sector
are also applicable to hospitals.
Data sources. We undertook a systematic review of the literature published since 1992 (ve electronic databases and a
manual search) using keywords relating to quality of health care and costs.
Study selection. We selected all articles relating cost and quality, providing indicators for quality failure, determining the
cost of failure, and itemizing the cost of quality. Twelve articles met these criteria (USA, nine; UK, one; Australia, one;
France, one); six referred to total quality management, three to hidden costs, and three to adverse events.
Data abstraction. For each article, we recorded the test hypothesis, the focus of the study and the main results on costs
and quality.
Results of data synthesis. Preventing failure by applying total quality management to a variety of projects (managed care
project, setting-up a standard procedure . . .) led to nancial savings; quality was maintained, even enhanced. Better
communication and co-ordination reduced hidden costs and also increased quality. Adverse events prolonged hospital stays
by 1.744 days and increased costs.
Conclusion. Very few detailed articles related cost and quality and, although they all noted a positive impact of an emphasis
on quality, they nevertheless had their shortcomings. Study periods were too short, the indirect aspects of costs and savings
were not taken into account, economic reference values were omitted. We conclude that more precise and strict methods
for quantifying costs are needed.
Keywords: adverse event, cost of quality, hidden costs, total quality management.
assurance programme, these items were 15%, 25%, and 60%,
Health costs are escalating in the Western world. Between
respectively, of the COQ [5]. According to Daigh [6], the 1979 and 1992, health expenditure per capita tripled in France
COQ should not exceed 37% of the total revenue or and the percentage of the gross national product devoted to
turnover and, according to Crosby, 2.5% [7]. Several theor- health increased from 7.9% to about 10%. Of this, 42% went
etical models reveal that investing in quality beyond a certain to hospitals [1]. Paradoxically, costs might be contained by a
point is no longer nancially worthwhile [2,3,4,8]. better quality service. Although quality has a cost, the failure
Since the end of the 1980s, the concept of COQ and the to provide quality can be even more costly [2,3] because, if
techniques used in industrial operations have been transposed a product or service does not satisfy a customer, the cost of
to the health care sector. However, because service in hospitals putting the situation right after failure may be greater than
is dispensed in real-time, it is difcult to implement in- actually preventing the failure from the outset.
process controls. It is preferable to focus on continuous The concept of cost of quality (COQ) is used to quantify
quality improvement and prevention. costs. It originated in the 1950s in the USA with the im-
We have performed a review of the literature which relates plementation of the rst quality control programmes in
quality and cost in order to assess and compare the nancial industry and comprises the cost of: (i) preventing failure; (ii)
benets that have been derived, or could be derived, from performing quality controls; and (iii) rectifying internal and
external failures [4]. In the rst year (1980) of IBMs quality different kinds of quality assurance incentives.
Address reprint requests to Suzanne Charvet-Protat, ANAES, 159, rue Nationale, 75640 Paris CEDEX 13, France.
E-mail: suzanne.charvet-protat@uhp.tm.fr.
2000 International Society for Quality in Health Care and Oxford University Press 125
A. Jarlier and S. Charvet-Protat
Total quality management and cost
Methods
TQM attempts to bring about continuous improvement in
In January 1998, we searched ve databases (MEDLINE,
quality of care through a system of training, communication
HealthSTAR, EMBASE, PASCAL and FRANCIS) to identify
and organization which prevents failure. The workload is
papers (English or French) published since 1992 on the COQ
redened so that there is roomfor change and re-organization.
in hospitals. We used the following key-words for the rst
The six selected articles on TQM concerned a managed care
three databases: Quality assurance, health care OR Quality
project [9], a project to prevent pressure sores [10], the
of health care ORManagement quality circles ORTotal quality
setting-up of a standard procedure called a patient care
management together with Hospital cost(s) OR Hospital pathway for a given surgical operation [11] and a set of
charge(s) OREconomics, hospital ORFinancial management, sector-specic projects such as organizing a nursing pool
hospital OR Hospital billing OR Hospital nance OR Hos- and improving hospital admission and discharge procedures
pital running cost OR (Cost allocation OR Cost benet [1214].
The main results on costs and quality in the six selected analysis OR Cost control OR Cost of illness OR Cost
studies are given in Table 1. In theory, TQM programme saving(s) OR Cost AND Cost analysis OR Cost(s) OR Cost
costs should, in time, be offset by savings in the areas effectiveness OR Economic value of life OR Health care
considered and by an increase in quality, even for es-
cost(s) OR Health economic(s) OR Economic aspect or
tablishments subject to severe nancial constraints. In prac-
Economic(s)) AND (Hospital planning OR Hospital de-
tice, TQM did lead to both substantial reductions in costs
partments OR Hospital units OR Hospitalization OR Hos-
and increases in quality but these were not always as great
pital(s) OR Hospital Management OR Hospital department
as might have been expected [11]. For instance, when a
OR Hospital equipment OR Hospital organization). A simpler
patients hospital stay was decreased as a result of or-
set of keywords was used for the more multidisciplinary
ganizational changes, the decrease related to the last days of
databases PASCAL and FRANCIS. We sought books and
the stay which are the least costly. Although in one study
theses on the COQ and completed our search by identifying
total savings reached $17.7 million (in $ for scal year
relevant citations in the reference lists of articles retrieved.
19901991), i.e. 7.2 times programme cost [12], in others
We devised an 18-item grid (inspired by the evaluation
savings were not statistically signicant [11] or occurred only
grids used in medical evaluation) to select papers which (i)
if indirect costs and savings were taken into account [14].
related quality and cost, (ii) and explicated the criteria of
Two studies used quantitative variables (rates of unplanned
failing to provide quality, (iii) and evaluated the cost of failure
re-admissions, mortality) to assess quality [11,13]. One study
or itemized the COQ. Point (iii) was considered especially
measured patients perceptions of quality of care; these were
important. Three people were involved in the selection pro-
signicantly improved during the study [9]. The remainder
cess: a student, an economist (A.J.) and the department head
relied only on qualitative criteria such as patient satisfaction
(S.C.P.). Each time there was some uncertainty about whether
or quality of work life for employees and physicians [12,14].
a paper should be selected or not, there was a discussion
For example, head nurses were more satised with the
between at least two of these three people.
performance of pool nurses; fewer agency nurses were em-
ployed [12]. The programme led to the design and use of a
questionnaire to measure patient satisfaction with obstetric
anaesthesia and analgesia and to the use of information
Results
available fromthe database to informpractitioners of potential
activities in their practice [14].
With the keywords that were chosen, 448 articles were
Results on savings and improved quality of care need to
retrieved electronically and 106 manually, but the true number
be judged in terms of how long the programme lasted, how
of articles was fewer because of redundant information among
large it was, the type of institution where it was implemented,
databases. More than one-half of the articles retrieved were
and what resources were allocated. Two programmes lasted
not relevant to our study because they dealt with economic
for 4 or 5 years [12,14], two 2 years [10,13] and the remaining
theory or management principles only. As health economists,
two were based on beforeafter comparisons with intervals
we did not try using other keywords; clinicians or hospital
of 2 and 7 months for setting up the programmes and study
administrators might have made a different choice. In par-
periods of 1 year [9,11]. Although there was a signicant
ticular, we did not use any keywords that reected any
difference between before and after, the stability of the
preconceived notions regarding which factors (e.g. adverse
results over time is unknown. None of the long-term studies
events) might inuence the costs of quality.
examined the time-scale required to obtain positive results
Several non-selected articles proved interesting but the
for both quality and costs in any detail. They either noted
objectives and methods were not explicit enough to judge
an overall decrease in savings and improved quality over 5
the true value of the information they contained. Only 12
years [12] or a decrease as from the third year [14]. Only
articles met our strict selection criteria. Nine were from the
one study [10] calculated quarterly costs for treatment and
USA, one was from the UK, one was from Australia and prevention, namely of pressure sores. Prevention costs per
one was from France. Six related to total quality management patient varied little whereas treatment costs decreased con-
siderably but without explanation, from $14 650 to $ 6696.2, (TQM), three to adverse events and three to hidden costs.
126
Quality and cost of failure
Table 1 Results on costs and quality in selected TQM papers
Study Hypothesis Focus of study Costs Quality
.............................................................................................................................................................................................................................
Coffey et al. [12] TQM improves both Incremental costs TQM benets Increase in:
University of quality and nancial and benets [costs in 17/19 quality of work life
Michigan Medical performance in 19 quality projects. Overall for all staff
Center improvement benets in 1990 patient satisfaction
projects (4-year /1991 7.2 times
programme) incremental costs
Kerrigan [13] Implementing TQM Reorganization of Savings: Low rates of:
Royal Victoria decreases costs workload to $351 589 in the FY unplanned hospital
Eye & Ear Hospital, (implicit hypothesis) increase quality and 1991/1992 (estimated re-admissions
Melbourne thus avoid $1.75m/5 years) postoperative
redundancies from $326 787 in the FY infections
budget cuts 1992/1993 hospital acquired
(2-year programme) (estimated $1.63m/ infections
5 years)
Eagle et al. [14] Quality assurance Measurement of direct Programme costs Substantial results in
Foothills Hospital, benets may costs and benets for decreased quality
Canada be greater than 53 projects in from $85 000 to
suggested from anaesthesia $48 000/year
cost analysis alone department over study period
(5-year programme) Benets: $14 000/year
Beck & Larrabee [10] Investing in Comparison of Costs: Hospital stay:
Regional Medical prevention decreases prevention and failure: $6696 14 days when
Center, Memphis failure costs and leads failure costs of prevention : $879 treating sores
to savings pressure sores 10 days when
(17-month preventing sores
programme)
Blegen et al. [9] Savings from Comparison of total Before: $3950 Patients perception of
University of Iowa implementing costs before and after (1541) quality of care
Hospitals & managed care implementing managed After: $3432 unchanged or enhanced
Clinics (maternity) might alter care in surgery (1737) (P<0.05) (questionnaire score:
quality of care department (18-month Overall decrease 4.26 before;
programme) =8% of total hospital 4.41 after, P<0.05).
costs Mean stay decreased
from 5.35 to 4.62 days.
Wright et al. [11] Patient care pathways Before/after 8% reduction in mean Mean stay reduced by
Massachusetts General (detailed daily goals comparison on cost per patient 3.1 days (P=0.03)
Hospital for patient and care setting up a standard ($1271) (P=0.47) 68 versus 52% of
team) could reduce patient pathway for a patients discharged
length of stay and particular operation in within 7 days of surgery
control costs a chest surgery Unchanged re-
department (1-year admission and mortality
programme) rates
FY, Financial year.
during the last two quarters of the study. However, no department or activity [11]; others included as many as 19
or even 53 projects covering topics as diverse as the in- quarterly results for quality are given nor is there any indication
of how the nal outcome regarding quality was measured troduction of new surgical procedures or services, phar-
maceutical contract purchasing by groups of hospitals, medical and whether it was signicant.
Programme size varied according to context, aims, and administration, staff training programmes, etc [12,14]. Pro-
grammes were set up in a single 44-bed unit [11] or in large available means. Some programmes focused on a single
127
A. Jarlier and S. Charvet-Protat
Table 2 Hidden costs
Hypothesis Focus of study Costs
.............................................................................................................................................................................................................................
Eisenberg et al. [15] Just paying attention to the Study of hidden costs in Savings from once-a-day
Six Pennsylvania hospitals purchase price of drugs antibiotic use antibiotic: 2.41 doses,
is not enough cost of dispensing i.e. $5.42/patient/day
savings from earlier Savings from earlier use
discharge of home parenteral
cost of nephrotoxicity antibiotics: $1042/patient
(1-year study) Cost of nephrotoxicity:
$2501/patient
Brita-Rossi et al. [16] Interdisciplinary professional Savings from reorganising Savings per patient:
Beth Israel Hospital, Boston collaboration could contain care in an inpatient $7600 for partial hip
costs and improve orthopaedic unit replacement
quality of care (Before: July 1992March $6000 for total knee
1993, replacement
After: May 1994August $7200 for hip and femur
1994) procedure with
complications/comorbidity
Thouverez [17] Socio-economic Savings from better Hidden costs:
French hospitals management could communication, Before: 88 000 FF/
decrease hidden hospital organization and employee/year
costs management in an After: 55 000 FF/
operating theatre employee/year
(18-month study)
FF, French francs [1FF=0.1801 US$ (1994)].
public, and sometimes university, hospitals (e.g. 884 beds a national average cost. No indication was given, however,
on the cost of setting up the programme. In the third study, [12]).
The cost of setting up a TQM programme was mentioned a group of consultants investigated the way an operating
theatre was run, interviewed the staff and analysed hospital in only two studies [12,14]. In one [12], only incremental
costs (the costs of quality improvement training, materials, accounts. They found hidden costs of the equivalent of
US$15 850/employee per year. The interviews pinpointed and co-ordination, and the costs associated with quality
improvement teams, including facilitators salaries) were in- aspects of the running of the theatre that could be modied
[17,18]. cluded because TQM was seen as a way of life in which it
is impractical, if not impossible, to delineate how much
time each person spends on TQM. In the other [14], it Adverse events and cost
corresponded to the cost of the staff involved only (quality
The three studies on adverse events meeting our selection
assurance co-ordinator, data entry clerk, programmer/ana-
criteria hypothesized that adverse events circumstances or
lyst).
events which would suggest that a lapse in the standard of
care has occurred could be an indication of quality failure
Hidden costs
and be responsible for increases in length of hospital stay,
mortality rate and/or costs. The main results of these studies When organization is poor, there are costs for rectifying
failure, or hidden costs, which often do not appear in the are summarized in Table 3. Adverse events prolonged hospital
stays by 1.744 days but the relationship between adverse account-books. Three articles, summarized in Table 2, sought
to discover how organizational changes might help to reduce events and length of stay deserves further investigation be-
cause patients who are more severely ill on admission are hidden costs. The most detailed study [15] used a owchart
to describe the steps involved in preparing and administering more likely to suffer certain adverse events and are also likely
to stay in hospital longer than other patients. In addition, an antibiotic in order to identify all real costs. It was only
necessary to bring about a few changes in antibiotic use to the longer a patient stays in hospital, the greater the op-
portunity for iatrogenic disease [19]. Only one study had obtain savings although the clinical impact of these changes
was not specied. The second study used the reduction in mortality rate as an end-point; it decreased from 3.5 to 1.05%
[20]. In terms of costs, two studies distinguished a cost hospital stay to quantify the decrease in hidden costs [16]
which were obtained by comparing departmental costs and attributable to an adverse drug event [20] or an avoidable
128
Quality and cost of failure
Table 3 Results on quality and costs in studies on adverse events
Focus of study Quality Costs
.............................................................................................................................................................................................................................
Walshe et al. [19] Impact of adverse events on Mean stay: Cost per case3.80 (at
Royal Sussex County Hospital, length of stay in without adverse events= 1992 price levels)
Brighton, UK ophthalmology department 2.74 days Savings from reduced length
(study of 1088 les over 3 overall=2.92 days of hospital stay not
years) with 1 event=3.06 estimated
with 2 events=3.16
with 3 events=6.22
(P < 0.05)
with 4 adverse events=
up to 4 days (P < 0.04)
Classen et al. [20] Comparison of two Decrease in: Mean cost of hospitalization/
LDS Hospital, Salt Lake City populations: mean stay from 7.69 to patient:
one with adverse drug 4.46 days $10 010 (case)
events mortality rate from 3.5 to $5355 (matched control)
one without (study of all 1.05% (P < 0.001)
admissions over 4 years) Cost attributable to adverse
drug event: $2013 (P < 0.001)
Bates et al. [21] Comparison of two Avoidable increase in length Avoidable cost due to
Two Massachusetts hospitals populations: of stay: 4 days (P=0.03) adverse events:
one with adverse drug $5857 (P=0.07)
events
one without
(6-month study)
cost due to adverse events obtained by comparing a popu- lengthen hospital stay, increase mortality rate and adversely
lation with adverse events with a control population [21,22]. affect costs [1921].
These costs of $2013 and $5857, respectively, were signicant. Several factors could bias the above results, in particular
The third study [19] gave only the costs of adverse event the lack of information on the indicators used to reect and
monitoring. In no instance were indirect costs calculated. measure quality. It thus becomes difcult, for example, to
determine whether how long a quality improvement pro-
gramme lasts inuences quality [10]. Most studies focussed
on visible, quantiable improvements and did not address or
Discussion
mention indirect costs and benets. One study [14] did stress
the importance of indirect non-nancial and nancial benets
Our literature search focused on studies that analysed re-
but did not venture to make an estimate of a value that is
lationships between quality of care and costs. It identied 12
indeed, by nature, difcult to estimate. In fact, real benets
articles dealing with either: (i) TQM; (ii) organization and
and costs could be higher than those given. Some studies
communication systems; or (iii) adverse events. All six articles
omitted crucial information such as baseline values. The
on TQM concluded in favour of its implementation in
impact of a measure depends on how high or low were initial
hospitals [914]. Thus, although initially an industrial initiative,
costs. For instance, the savings of $7000 per patient for a
TQM can also be applied in hospitals. The studies we analysed
partial hip replacement should be viewed in relation to the
suggested that TQM might reduce costs due to failure,
cost of the operation which was not specied [16]. If costs provide substantial returns on investment, generally improve
and benets are not expressed in terms of a reference value hospital nances, and have a positive impact on employment.
(e.g. cost of surgery, overall hospital budget, etc.), they should Decreased costs did not negatively impinge on quality of
be correlated directly as in the study where savings from care; on the contrary, quality of care and patient management
TQM amounted to 7.2 times investments [12]. Finally, the tended to be improved. Three articles focussed on reducing
costs in studies from Anglo-Saxon countries are taken from hidden costs [1517]. Reorganizing tasks and promoting
hospital accounts: are these real or standard costs? Without exchange of information and co-operation among staff helped
further information, it is difcult to know what could be the to prevent failure but with varying degrees of success ac-
implications of the results for another country. cording to study. Finally, three studies addressed a single
indicator of quality failure, namely, adverse events, which can Indicators chosen to assess quality failure must permit
129
A. Jarlier and S. Charvet-Protat
3. Finkler SA. Measuring the costs of quality. Hosp Cost Manag
rigorous comparisons. Adverse events of different origin are
Account 1996; 7: 16.
considered good indicators of quality failure but cannot
reect, in the absence of other indicators, quality of hospital
4. Lundvall DM, Juran JM. Quality costs. In Juran JM, ed., Quality
care [2224]. Our literature review identied three studies on
Control Handbook. New York: McGraw-Hill, 1974: pp. 122.
adverse events two of which concerned adverse drug events.
5. Harrington HJ. Analysis of direct poor-quality costs. In Poor
We considered that these studies provided a better overview
Quality Cost. New York: Marcel Dekker, 1987: Chapter 4.
of the relationship between quality of care and costs than
6. Daigh RD. Financial implications of a quality improvement the frequently used indicator of nosocomial infection which
process. Top Health Care Financ 1991; 17: 4252.
has limitations and would require a separate in-depth study.
Many factors intervene in the incidence of nosocomial in-
7. Crosby PB. Quality may not be what you think it is. In The Art
fections (patients individual risk of infection, site of infection,
of Making Quality Certain. New York: McGraw-Hill, 1979: Part
type of health establishment, organization of care) and have
I, Chapter 2.
to be taken into account in prevention plans and in an
8. Schneiderman AM. Optimum quality costs and zero defects:
analysis of their costs [2529].
are they contradictory concepts? Qual Prog 1986; Nov: 2831.
An important aspect insufciently discussed in the studies
9. Blegen MA, Reiter RC, Goode CJ, Murphy RR. Outcomes of
that met our selection criteria was whether the positive results
hospital-based managed care: a multivariate analysis of cost and
observed during the study period would persist over time.
quality. Obstet Gynecol 1995; 86: 809814.
Results were assessed at the close of the study. Because trends
are unknown, economically relevant, long-term extrapolations
10. Beck KL, Larrabee JH. A simultaneous analysis of nursing care
are not possible. Moreover, we are not always told whether quality and cost. J Nurs Care Qual 1995; 9: 6371.
a discount rate is applied to enable cost comparisons over
11. Wright CD, Wain JC, Grillo HC et al. Pulmonary lobectomy
different periods. A further shortcoming is the frequent
patient care pathway: a model to control cost and maintain
absence of statistical analysis.
quality. Ann Thorac Surg 1997; 64: 299302.
In conclusion, our literature review has underscored how
12. Coffey RJ, Gaucher EM, Lyons P. TQM brings nancial benet
relevant and useful it may be to transpose the concept of
to the University of Michigan Medical Center. In Neuhauser
COQ from the industrial to health sector and pinpointed
D, McEachern JE, Headrick L, eds. Clinical CQI. A Book of
several methods of analysing relationships between quality
Readings. Oakbrook Terrace: JCAHO, 1995: pp. 237242.
and costs. Nevertheless, several questions remain open: which
13. Kerrigan JS. Quality pays and reduces your risk. J Qual Clin
hospital sectors would benet most from these methods?
Pract 1995; 15: 147157.
how relevant are studies based on just single indicators such
as nosocomial infections? is TQM adapted to every kind of
14. Eagle CJ, Davies JM, Pagenkopf D. The cost of an established
health establishment? Moreover, the results described in this
quality assurance programme: is it worth it? Can J Anaesth 1994;
article depend on quality indicators which are often not well
41: 813817.
detailed in the original papers. They must not be taken out
15. Eisenberg JM, Glick H, Koffer H. Assessing the hidden cost
of context, and have to be interpreted with care and supported
of antibiotic therapy for hospitalized patients. Drug Inform J
by further studies using more rigorous methods. In essence,
1988; 22: 459469.
they emphasize the need for well-designed quality protocols
16. Brita-Rossi P, Adduci D, Kaufman J et al. . Improving the process
in the future, especially if the aim is to reduce failures and
of care: the cost-quality value of interdisciplinary collaboration. J
decrease costs.
Nurs Care Qual 1996; 10: 1016.
17. Thouverez C. The contribution of socio-economic management
to a performance project in the operating theatre (in French).
Acknowledgements
Recherche Soins Inrmiers 1995; 42: 2535.
18. Savall H, Zardet V. The reduction of hidden costs: a means of The authors thank Dr. Herve Maisonneuve for his con-
nancing hospital quality improvement (in French). J Econ Me d
structive advice in designing this study and for critically
1996; 14: 285290.
reading the manuscript, Dr. Tiiu Ojasoo for valuable help in
writing the manuscript, Hele`ne Cordier for all information
19. Walshe K, Bennett J, Ingram D. Using adverse events in health-
retrieval, and Nathalie Preaubert for help in performing a
care quality improvement: results from a British acute hospital.
preliminary selection of the articles. Int J Health Care Qual Assur 1995; 8: 714.
20. Classen DC, Pestotnik SL, Evans RS et al. . Adverse drug events
in hospitalized patients. J Am Med Assoc 1997; 277: 301306.
References
21. Bates DW, Spell N, Cullen DJ et al. The costs of adverse drug
events in hospitalized patients. J Am Med Assoc 1997; 277:
1. Ville G. Challenges in the implementation of quality assurance
307311.
in health care organizations (in French). Paris: IESTO, 1995.
22. Bates DW, ONeil AC, Petersen LA et al. Evaluation of screening
criteria for adverse events in medical patients. Med Care 1995; 2. Finkler SA. Total quality management. Measuring costs of
quality. Hosp Cost Manag Account 1993; 5: 16. 33: 452462.
130
Quality and cost of failure
23. Sanazaro PJ, Mills DH. A critique of the use of generic screening 27. Haley RW, Culver DH, White JW et al. The efcacy of infection
surveillance and control programs in preventing nosocomial in quality assessment. J Am Med Assoc 1991; 265: 19771981.
infections in US hospitals. Am J Epidemiol 1985; 121: 182205.
24. Hill CA, Winfrey KL, Rudolph BA. Best hospitals: a description
28. Meynet R, Fabry J, Sepetjan M. Micro-economy of nosocomial
of the methodology for the index of hospital quality. Inquiry
infections (in French). Lyon: Fondation Marcel Merieux, 1991.
1997; 34: 8090.
29. Wakeeld DS, Pfaller MA, Hammons GT, Massanari RM. Use
25. Jarvis WR. Selected aspects of the socioeconomic impact of
of the appropriateness evaluation protocol for estimating the
nosocomial infections: Morbidity, mortality, cost, and pre-
incremental costs associated with nosocomial infections. Med
vention. Inf Contr Hosp Epidemiol 1996; 17: 552557.
Care 1987; 25: 481488.
26. Curie E, Maynard A. The economics of hospital acquired
infection. Discussion paper 56. University of York Centre of
Health Economics, 1989. Accepted for publication 19 November 1999
131