Economic Evaluation andHealth Care - BMJ · Valueis intheeyeofthebeholder development of other techniques, in which people are asked to state their preferences among specified choices

Economic Evaluation and Health Care

Cost-benefit analysis

Ray Robinson

Cost-benefit analysis is probably the most compre-hensive method of economic evaluation availableand it can be applied in two ways. The human capitalapproach means that the value of people's contri-butions is linked to what they are paid. The approachbased on individuals' observed or stated preferencemeans that their personal valuations are placed on anactivity by assessing how much money they areprepared to accept for an increased risk or to pay fora particular service. Each method has its disadvan-tages and the most successful that has been devisedso far is the "willingness to pay" (stated preference)approach, though the response to this is to a largeextent dependent on the income of the person beingquestioned. There are still problems with its applica-tion, however, so its usefilness is limited.

Cost-benefit analysis is the most comprehensive andtheoretically sound form of economic evaluation and ithas been used as an aid to decision making in manydifferent areas of economic and social policy in thepublic sector during the last 50 years. There arenumerous textbooks dealing with the theory andpractice of the general approach.'-3The main difference between cost-benefit analysis

and other methods of economic evaluation that werediscussed earlier in this series is that it seeks to placemonetary values on both the inputs (costs) and out-comes (benefits) of health care. Among other thingsthis enables the (monetary) returns on investments inhealth to be compared with the returns obtainable frominvestments in other areas of the economy. Within thehealth sector itself, the attachment of monetary valuesto outcomes makes it possible to say whether aparticular procedure or programme offers an overallnet gain to society in the sense that its total benefitsexceed its total costs. Cost-effectiveness and cost-utility analysis do not do this because they measurecosts and benefits in different units.

Devising ways in which complex outcomes of healthcare can be reduced to a single monetary measure is noteasy and is the main reason why cost-effectiveness andcost-utility analysis have been relied on more often inthe health care sector. None the less there have been anumber of approaches which over the years havesought to place monetary values on the benefitsthat arise from health care programmes. These maybe divided into two main categories: the humancapital approach and approaches based on individuals'observed or stated preferences.

Human capital approachEarly approaches to benefit valuation were com-

monly based on the concept of "human capital." Thisconcept is designed to convey the fact that humanbeings are similar to capital equipment (at least as far astheir working lives are concerned) in the sense thatthey can be expected to yield a flow of productiveactivity in future years. If the value of this activity inany period of time is assumed to be equal to theindividual person's rate of pay, then the benefits ofhealth care can be measured in terms of the future flowof income that would otherwise have been forgone

because of ill health. Because these calculations involveadding up a stream of income that accrues over anumber of different years, the sum in each year mustbe time-discounted to take account of the precise timeprofile of benefits (methods of time discounting anddebates surrounding this practice were discussed in thesecond and third articles in this series).The human capital approach has been applied in the

valuation of health benefits in cases of both avoidablemorbidity and mortality. There have, however, beenvarious criticisms. As one of the first attempts to placemonetary values on avoided mortality it led to mone-tary values being placed on human life and manypeople have strong ethical objections to this. Incontrast, most economists think that implicit valua-tions are placed on human lives in a whole range ofdecisions about allocation of resources in the publicsector and the cost-benefit approach is simply beingexplicit about the process.4'5

Other criticisms have centred on the use of rates ofpay as a measure of value. Economic theory suggeststhat productivity is accurately measured only by ratesofpay when certain conditions in the labour market aremet. The widespread existence of restrictive practicesand other forms of imperfections in the labour marketmeans that these conditions often do not apply inpractice. At the same time, valuing benefits in terms ofrates of pay neglects the health benefits that accrueto people who are not employed-for example,non-working wives and retired people. It also ignoresthe non-financial costs of pain, suffering, and griefwhich are often associated with illness. But, from aneconomist's perspective, the main criticism of theapproach is that it is not based on an individualperson's valuations of benefits. Indeed, a third partyview is taken about people's "worth" to society interms of their productive potential. This viewpointis inconsistent with the prevailing view among econo-mists that the individual person is the best judge of hisor her own welfare.

Individuals' observed or stated preferencesThe observed preferences approach involves observ-

ing individuals' behaviour and using these obser-vations as a basis for valuing benefits. One method ofdoing this is to observe their behaviour towards riskand then to estimate the personal valuations implicit inthis behaviour. Some people accept "danger money"for undertaking work with an increased risk of injuryor death-for example, deep-sea divers. Others spendmoney on cars with enhanced safety features to reducethe risk of injury or death. When the extra incomereceived or expenditure undertaken is compared withthe change in the degree of risk associated with aparticular activity, it becomes possible to establish thepersonal valuations implicit in observed behaviour.Converting valuations associated with small changes inrisk into full life valuation is not straightforward, butvarious techniques associated with "scaling risks" havebeen developed.6 On a practical level, however, one ofthe main disadvantages of this approach is the limitednumber of situations under which attitudes towardsrisk can be observed and measured. This has led to the

BMJ VOLUME 307 9 OCTOBER 1993

This is thefifth in a series ofarticles that describes the waysin which methods ofeconomicevaluation may be used toassess the economic costs andconsequences associated withdifferentforms ofhealth careintervention.

Institute for Health PolicyStudies, Faculty ofSocialSciences, The University,Southampton S09 5NHRay Robinson, professor anddirector

BMJ 1993;307:924-6

924

on 20 May 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.307.6909.924 on 9 October 1993. D

ownloaded from

Value is in the eye ofthe beholder

development of other techniques, in which peopleare asked to state their preferences among specifiedchoices in monetary terms. This is known as the"willingness to pay" approach.

Outside the health sector, this method has been usedwidely in the evaluation of environmental costs andbenefits.7 In this context, probably the most famousand controversial application ofthe approach in Britaintook place over 20 years ago when the Roskill Commis-sion on the siting of the then proposed third Londonairport used the willingness to pay approach to valuethe costs of aircraft noise. The commission's approachwas to ask people how much compensation they wouldrequire to leave them equally well off if they wereforced to move house to avoid noise costs.8

In the context of health, the willingness to pay

approach seeks to establish the value that people attachto health care outcomes by asking them how much theywould be prepared to pay to obtain the benefits or avoidthe costs of illness.

Because payment or compensation is not actuallymade, however, either interviews or postal ques-tionnaires are normally used which are based onopen ended or discrete valuation questions. Withopen ended questions a form of bidding game resem-

bling an auction is the most common approach. Anopening bid is made which the respondent eitheraccepts or rejects; thereafter, bids are either raised orlowered until the respondent's maximum willingnessto pay is reached. With discrete questions, the respon-dent is presented with a series of prices and asked tooffer a yes/no answer depending on their personalwillingness to pay. This approach is less susceptible tostarting point bias, which may cause the respondent'sanswers to be influenced by the starting bid, but thereare a number ofother forms ofstrategic and compliancebias which must be guarded against whicheverapproach is used.9

Applications of the willingness to pay approach tohealth care are still relatively rare, but some interestingstudies have been undertaken. For example, in an

early application of the approach, Acton investigatedwillingness to pay for mobile coronary care units thatwould reduce the risk of death after a heart attack.'0Among other things, this study looked at a respond-ent's personal willingness to pay for a programme,given various probabilities of heart attack and death,in comparison with their willingness to pay for a

community programme. Not surprisingly, perhaps,willingness to pay was greater for perceived personalbenefits than it was for community benefits.

More recently, Thompson used willingness to pay

methods to value the effect of a hypothetical cure on

their state of health in 247 patients with rheumatoidarthritis who were already enrolled in a randomisedcontrolled drug trial." During interviews, respondentswere asked to think about the ways in which arthritisaffected their lives and their families. They were thentold to assume that there was a complete cure for theirdisease and asked how much they would be willing topay for it. The precise question that they were askedwas: "What percent of your families' (ie household)total monthly income would you be willing to pay on aregular basis for a complete cure for arthritis?" Consis-tency of answers was checked by a follow up questionabout how the family would manage to live on theremaining amount.Using these methods, Thompson obtained a 96%

response and 84% of the responders gave answers thatmet predetermined criteria for plausibility. The resultsindicated that the average responder was willing to pay22% of their household income to secure a cure forarthritis. This proportion did not vary with income.

In another study, Johannesson, Jonsson, and Borg-quist 11th Nordic Health Economists' Study GroupMeeting, Stockholm, 27-28 August 1990) examinedwillingness to pay for antihypertensive therapy. Theyachieved a response rate of 67% from an elderlypopulation (mean age 64-4 years) comprising 481patients at a primary health care centre. A particularfocus of this study was an investigation of the relativemerits of an open ended questionnaire compared withone using discrete yes/no questions. The authorsconcluded that open ended questions do not workwell in willingness to pay postal surveys. Discretequestions, on the other hand, led to a lower non-

response rate and provided respondents with an easiervaluation task. This approach indicated that peoplewould be willing to pay between SKr 2500 andSKr 5000/year (about £225-1450) for antihypertensivetherapy. The authors also concluded that although theresults should be interpreted with caution, they indi-cated a large potential application for willingness to paymethods in the field ofpreventive health care.Two recent studies carried out in the United King-

dom were by Rushby (JA Rushby. Paper presented tothe Health Economists' Study Group, July 1991) andDonaldson.'2 Rushby investigated willingness to payfor improvements in the quality of life offered by heartpacemakers. Respondents were asked to state theirmaximum willingness to pay to be free of specifiedsymptoms associated with heart disease (shortnessof breath, chest pain, dizziness, fainting, and palpi-tations). Groups of respondents included those withpersonal experience of the symptoms, those withexperience of caring for patients, and those with no

more experience of the symptoms than the generalpopulation. All groups were, however, asked to basetheir answers on hypothetical cases rather than currentexperience. The answers indicated that the mean

willingness to pay varied from £45.50/month for heartpalpitations to £70.67/month for chest pain.

BMJ VOLUME 307 9 OCTOBER1993

Willingness to pay methods have beenused to value benefits arising from:* Ultrasound scans in pregnancy* Mobile coronary care units* Drug treatments for hypertension* Anaesthetics following hip surgery* Heart pacemakers* Cures for rheumatoid arthritis* Continuing care for elderly people

925

on 20 May 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.307.6909.924 on 9 October 1993. D

ownloaded from

Donaldson applied willingness to pay in an attemptto value the benefits of continuing care accommodationfor elderly people.'2 The aim was to compare NHSnursing homes with NHS hospital accommodation. Adistinctive feature of the study was that relatives ofrespondents not the residents themselves were askedhow much they thought the government should bewilling to pay for the respective accommodation; 71%of respondents provided valuations that could be usedin the analysis and their answers indicated a clearpreference for nursing home accommodation overhospital accommodation when the total willingness topay of the two options was compared with theirrespective costs.

In a recent review of the application of willingness topay studies, Morrison and Gyldmark argued that threecriteria must be met if use of the approach is to bevalid.'3 Firstly, given the uncertainty surroundingindividual subjects' needs for health care, willingnessto pay questions should ask how much a person iswilling to pay as an insurance premium so that a givenservice would be available if it is needed. Secondly,expectations should be expressed in terms of prob-abilities; that is, what is the probability of needingtreatment and of its success? Thirdly, representativesamples of the population are necessary to establish thetotal willingness to pay of the relevant population.They concluded that few of the studies carried out todate meet these criteria and that there are othermethodological issues that require attention, amongwhich is the link between willingness to pay andincome.

Income and monetary valuation ofbenefitsValuing the benefits of health care interventions in

terms of willingness to pay raises the problem that theamount people are willing to pay is often positivelyrelated to their level of income. The fact that a richperson is willing to pay £40/week for a drug whereas apoor person is willing to pay only £1 O/week is likely toreflect the value each ofthem attaches to money itself aswell as the benefits of the drug to health.

This is a problem that bedevils all attempts to valuebenefits in monetary terms, both inside and outside thehealth sector and it occurs when payment is actuallymade as well as when it is based on willingness to paysurveys. In some cases attempts have been made toovercome the problem by "weighting" benefits accord-ing to the income group of recipients with lowerincome groups receiving larger weights, but this raisesmany practical problems in addition to the obvious oneof calculating the appropriate scale of weights. Anotherapproach is to take willingness to pay figures at theirface value if the programme under considerationrequires only small amounts of expenditure in relationto a person's total income. If this is done distortionsthat result from different valuations of money itself arelikely to be minimised.

ConclusionCost-benefit analysis is the most comprehensive

form of economic evaluation. By assigning monetaryvalues to both costs and benefits, it offers the potentialfor comparisons between a wide range of programmesboth within the health sector and between the healthand non-health sectors. As far as its application within

Summary

* Cost-benefit analysis places monetary values onboth the inputs (costs) and the outcomes (benefits) ofhealth care* Early methods of valuation of benefits were basedon the human capital approach in which benefits arevalued in terms of productivity gains. Rates of pay areused as a measure ofproductivity* Recent approaches have adopted the more theore-tically sound practice of basing valuations on peoples'observed or stated preferences* Stated preferences are usually based on willing-ness to pay; the value people attach to health careoutcomes is established by asking them how muchthey would be prepared to pay to obtain the benefits oravoid the costs of illness* The dependence of willingness to pay on a person'sincome does, however, create a difficulty becauseanswers may reflect the value people attach to moneyitself as well as their valuation of the benefits of healthcare* Despite their theoretical superiority, willingness topay studies are still relatively rare. Their application inthe immediate future is likely to be most appropriate inthe case of, for example, pharmaceutical productswhere qualitative health gains are achievable, wherepayments are small in relation to income, and wherepeople are familiar with cash transactions

the health sector is concemed, it is a more powerfultechnique than either cost-effectiveness or cost-utilityanalysis. Among other things, it enables people toexpress the benefits of health care in terms of theirvaluations of the quantity of life, its quality, and anyother dimension that they feel is important. And all ofthese diverse benefits are expressed in terms of a singlecommon unit of measurement.

Early methods of obtaining valuations based on thehuman capital approach have been largely supersededby more theoretically sound methods based on indi-viduals' stated preferences, but a number of theoreticaland practical problems remain and applications ofthe approach have been relatively uncommon. Inthe immediate future its application is likely to berestricted to those areas in which small improvementsin the state of health are attainable, in which paymentsare relatively small in relation to income, and in whichpatients have some familiarity with cash payments-for example, pharmaceutical products that offerimprovements in the quality of life.

1 Pearce DW. Cost-benefit analysis. 2nd ed. London: Macmillan, 1983.2 Mishan EJ. Cost-benefit analysis. 4th ed. London: Unwin Hyman, 1988.3 Walshe G, Daffern P. Managing cost benefit analysis. London: Macmillan,

1990.4 Jones-Lee MW. The value of life. An economic analysis. London: Martin

Robertson, 1976.5 Mooney GH. The valuation ofhuman life. London: Macmillan, 1977.6 Mooney GH. Human life and suffering. In: Pearce DW, editor. The valuation

ofsocial cost. London: George Allen & Unwin, 1978.7 Mitchell RC, Carson RT. Using surveys to value public goods. Washington, DC:

Resources for the Future, 1989.8 Commission on the third London airport. Report. London: HMSO, 1971.9 Johannesson M, Jonsson B. Economic evaluation in health care: is there a role

for cost benefit analysis? Health Policy 1991;17:1-23.10 Acton JP. Evaluating public programs to save lives: the case of heart atuacks. Santa

Monica: Rand, 1973.11 Thompson MS. Willingness to pay and accept risks to cure chronic disease.

AmJ7Public Health 1986;76:392-6.12 Donaldson D. Willingness to pay for publicly-provided goods. J7ournal of

Health Economics 1990;9:103-18.13 Morrison GC, Gyldmark M. Appraising the use of contingent valuation.

Health Economics 1992;l:233-43.

926 BMJ VOLUME 307 9 OCTOBER 1993

on 20 May 2020 by guest. P

rotected by copyright.http://w

ww

.bmj.com

/B

MJ: first published as 10.1136/bm

j.307.6909.924 on 9 October 1993. D

ownloaded from