Sac. %I. & dad.. Vol 8. PP 521)to 533 ~erpknon Press 1974 Pnnted anGreat
. Br~taln
THE
ECONOMIC
APPROACH TO FAMILY STUDIES
HEALTH
J. S. DEEBLF. Faculty of Economics and Politics. Monash University. Victoria. Australia Abstract-This paper provides a background to the use of the family as a unit in studies of health economies and to outline the statistical requirements of such studies.
I. BACKGROUND
Since both the nature of good health and the processes for producing it are unclear. health economics has been concerned mainly with the effects of its converse. illness, on: (a) productivity, through reductions in the size. qua: lity and mobility of the work force; (b) the distribution of income and its “adequacy”, through loss of earnings and/or additional outlays; (c) the allocation of resources. through their use in preventive. curative and suportive health services. Though economists tend to be most interested in market transactions. this is largely because of relative ease in data collection and need not limit the range of economic analysis. Also. concern with efficient resource allocation does not deny the importance of others factors. Studies of illness might well be justified for its influence on productivity alone, but policy issues are more likely to centre on equity in the distribution of income and well being. Since illness is an individual event, with different economic consequences for individuals. the need for standard demographic data is assumed throughout this paper. As measured by earning capacity, the economic “worth” of an individual varies with such factors as age, sex, race. education and occupation. and the same variables have been used in studies of the demand for specialized health services. But note that in the latter they serve more as proxies for characteristics less easily measured: age. sex. and perhaps race. for health standards; education and race for cultural attitudes and life styles; occupation and education for socio-economic status. 2. THE FAMIL\ Families and households are primary economic units within which incomes may be pooled expenditure patterns fixed and non-market services rendered inter-
nail>. Assessment of income losses due to illness must take account of family and household size and the degree of de’pendence amongst members. More importantly. household production may include: (a) preventive services embodied in living standards which nray influence health directly (housing and sani-
tary conditions. dietary practices. etc.); and (b) curative and supportive services which either complement or substitute for specialized external care, depending on actual and expected differences in quality, relative prices. and household purchasing power. Capacity to provide internal support may not only explain some variation in demands for organized health services. but also offer a basis for assessing the family and household as socio-economic institutions. Recognition of this role has produced at least one general model of the household as a producer of health. using as inputs “imported” services, internal services. time and other consumption items, some negatively correlated with health [l]. However, the results are not yet significant operationally. so that most data has come from more conventional studies of household expenditures and consumption. Since they reflect the socio-economic institutions of the countries of origin. not all results can be generalized; this is particularly true of the many otherwise excellent American studies. Also. the widespread manipulation of supply for social reasons means that observed consumption behaviour may embody the priorities of producers as well as consumers. Subject to these reservations, conclusions relevant to the family include: (i) household expenditures and health practices are determined mainly by the attitudes of the female head. There is some evidence that these vary with her own experience, e.g. positive attitudes tend to be associated with a history of good health. apathy with poor experience [2]; (ii) specific preventive measures rarely enter household budgets after child members pass infancy. Few include professional review when asymptomatic. However. some consumption of medicaments .is viewed as preventive (tonics. vitamins, etc.) C2-43; (iii) at least 70 per cent of all illness episodes are nlanaged without resort to external treatment or advice. These include some illnesses resulting in work loss [5.6] ; (iv) consumption of organized external services varies systematically with family size and composition. Usage per person is highest in two-person families. de-
530
J. S. DEEULE
clining until family size reaches six. after which it tends to rise. Variations are greatest in hospital use. Adjustment for age/sex composition reduces the range of variation but does not remove a pattern of economies of scale consistent with a limited capacity for internal care [4, 7-101; (v) usage. particularly hospital usage. is significantly higher amongst the widowed. the divorced. and other members of incomplete families. Family structure may also explain part of the differential use by females generally [I I. 121.; (vi) migration is associated with higher than average use. particularly wheri inter-national and inter-cultural. Psychiatric services are most affected. However. the importance of familial disruption is not clear [ 133 : (vii) family income is correlated with usage and expenditures, but the relationships are complex and have not been measured adequately. Conventional analysis of demand elasticities is complicated by the fact that: (a) ill health may be a cause of both high usage and low income. depending on its incidence within the family. the sources of family income and public or private arrangements for income support. Since the latter may be more readily available to high-income than low-income families, the major variables may not be independent. (b) pricing policies may be specifically designed to counter income difference;. If services are provided publicly. without charge and from taxation, income is significant only to the extent that such attitudinal factors as social class and education are correlated with it, or that. at the aggregate level. regional differences in income are associated with variations in supply [ 141. Insurance financing has similar effects. although the demand for insurance is itself income related [IS]. However, voluntary insurance coverage reflects demand for financial security as well as health care and is also affected by self-assessment of health status and differential pricing in favour of large families and highrisk groups [IS]. As might be expected. income appears to be most relevant when fee-for-service charges fall directly on users. but concessions for the poor may again result in interdependence [ 161. These measurement problems all arise from inability to measure degrees of ill health directly and thus to remove the main determinants of use. Self-assessment of health status has been used with some success as a predicter ofconsumption behaviour, but it has obvious defects as an objective measure. Consequently, studies to date have not greatly extended the u priori expectation that usage and expenditures are positively. but not strongly. associated with income. the association being strongest for dental care and preventive measures, less relevant to such “essentials” as hospitalization. Data on famial influences has arisen mainly as a by-product of attempts to estimate income erects. Though efforts have been made to include earnings loss as an element in the choice between time-intensive (home care) and service intensive (cxtcrnal) treatment
methods. they have been limited to employed su&rers only [17]. Little is known of the other opportunit! costs which enter farnil\ decision-making. Since the\ may include the poteniial earnings loss of members other than the sufferer. analysis b! sources of incomc for different family structures \vould be mot-c rele:\,ant to our purposes.
3.
STATISTIC.AL
REQI. IRE\IESTS
Statistical requirements will vary with the purpose of the studies envisaged. Routine collection of all potentially useful data would be unnecessary and expensive. but it would also be wasteful to duplicate. through nd hoc surveys. data alread) collected but not identified. In practice. the linking ‘of existing records and statistical sources would be of great importance. Table I (Appendix) lists some of the data headings used in or suggested by studies to date. Classification into individual, linked family. and population data implies the use of sample (rather than whole population) surveys and is related more to possible methods of collection than to the nature of the information required. The amount of “benchmark” information available for whole populations will. of course. vary between countries. Some technical difficulties are discussed briefly in Section 4 below. However. more general questions are raised by the choice of unit to be studied. Three alternatives are : (i) the household. defined as “a group of persons who eat their meals together or whose meals are prepared by a common cook from a common pot”; (ii) thrfirnily. defined as “one person or a group of people living together and related to one another by blood, marriage or adoption. When two related married couples live in a single dwelling unit, each couple and its unmarried children are treated as a seperate unit” [ IS]; or more narrowly as “a married couple. alone. or with their never married children (of any age). A family may also be a lone parent with hisiher never married child or children.. .” [ 191. (iii) tiw incow unit. defined in one study as “either: (a) a married couple with their dependent children. i.e. their children of less than fifteen years and also any children aged fifteen or over engaged in full-time secondary education: or (b) a single person aged fifteen or older. no longer engaged in full-time secondary education” [20]. As economic units. the three categories differ mainlq in the degree of income pooling and joint decisionmaking implied. The ho~srMtl is a residential unit. implying a common domestic environment but including such unrelated income-carriers as boarders whose contribution to the common budget may be limited and for whom no reciprocal obligations of care exist. The fi~rrdy may (depending on the definition used) extend beyond the nuclear stage and implies a ‘greater resource pooling and more clearly defined internal
The
economic
approach
obligations. Howcvcr. this may not be true for all nondependent children and grandparents. The incornr urrit is. by comparison. tightly defined by reference to dependence and is limited to the nuclear family as a maximum. All spending decisions are presumed to be taken jointly. No one definition is likely to meet all requirements. The income unit is the most suitable for economic studies, as representing the least equivocal definition of a basic earning, spending and decision making group. It also corresponds with the concept of family most commonly used for other purposes. e.g. for taxation concessions. family allowances. social security benefits, health insurance coverage. and so on (the age of dependence will, of course, vary between societies). However. it may be too narrow for epidemiological and sociological work. unless accompanied by information about interactions with other. related individuals and groups. Since the identification of “related units” would be equivalent to a wider definition of the family. it may in fact be preferable to retain the distinction between the two groups. In this way a family concept sufficiently wide for other purposes could be preserved without attributing to it a spurious economic unity. Given the increasing proportion of nuclear families, the need for additional information might be expected to decline.
4. METHODOLOGY
Because few studies have centred on family decisionmaking per se. no clear methodology has emerged. Obvious problems arise in assembling all the demographic. medical. sociological and economic data required. In the long run. the solution may be in a combination of record linkage (medical and other) with regular household budget and expenditure surveys. However. formal record linkage is so little developed and budget surveys so infrequent and generalized. that special collections will be necessary for some time. Though less precise methods can yield useful hypotheses. most large-scale surveys have relied on random sampling methods. However. subsequent procedures have varied with their reliance on (a) direct collection through interview or questionnaire. (h) extraction from. or reference to. records already maintained by suppliers of services and public or private agencies for administering and financing health care. Expense. and the quality and reliability of information. are the main criteria choice. Direct collection of all relevant data through interview or questionnaire interrogation of randomly selected families has the merits of (a) unbiased selection and
to family
health studies
531
(b) comprehensivcncss. but may face problems in: (i) surllp/in~-selection and identification of families within larger units (households. dwellings, etc.) requires precise definitions and car&l sample design. Though standard techniques are available (block sampling, etc.) the method is expensive. (ii) i~~prr$~-t know1rd+~ clrrd r7o~t-1’c.spo11.s~~ adequate replies require some sophistication in health matters by interviewees, hence response is variable and may be biased. Though the female head is the most reliable source of information. she may not be aware of all family health experience. incomes and expenditures. Medical data is imprecise and may be limited to overt symptoms. Refusal to divulge income and/or expenditure data may require use of proxy variables [2J]. (iii) irnper.fic.r or .s&cri~x~ rc~ll-may atTcct all data. Problem in definition of income. given short-run variations-permanent or current income preferred? Sclective recall of illness. depending on cultural attitudes. tolerance of disability. etc. Tendency to overrate major episodes. particularly hospitalization. Unreliable recall of the use and cost of specific services beyond a short period. Measures to overcome these limitations include: (i) ‘restriction of survey period to few weeks as a maximum, implies the use of expensive continuous sampling methods [22] ; (ii) multiple interviews or questionnaires. administered concurrently to several family members; (iii) continuous recording of health experience. utilization, income and expenditures through “diaries” maintained by one or more family members. Allows longer survey period. but both response and reliability may suffer without regular supervision; (iv) validation of utilization and expenditure data by reference to suppliers and financing agencies. i.e. physicians. hospitals, pharmacies. insurance organizations, public authorities etc. Though these modifications can improve both the range and the quality of data. they are all relatively costly. ’ Alternatively, samples of individuals or families may be selected from the records of public or private suppliers, insurance authorities or payment agencies, using interviews or questionnaires to collect unrecorded information for the selected units [23]. This approach provides (a) ready selection and identification of family units. (b) accurate data for the consumption and cost of specialized services. (c) better medical information. To the extent that data for several services and/or several family members are aggregated within administrative or financing agencies. limited record linkage may already exist. However. the method may suffer from: (a) irrcorrrp/ure poprflariorr corc,r.-valid population samples are possible only, when public services or national insurance programmes require universal registration and record-keeping. Otherwise. self-selection may bias results:
J. S. DEEBLE
532
(b) co~1/7lt~.sir_rin Sara@ rlesiyrl-avoidance of bias due to the form and arrangement of records may require complex sampling methods peculiar to each area and service. whereas standardized procedures can be used in household surveys: (c) rrco~&g rj.ror.s-accuracy in record-keeping may depend on its administrative significance. i.e. for payment or reimbursement purposes. insurance rating. etc. Obsolete or inaccurate identification data for individuals may prejudice sample design, also offset economies in the selection of family units: (d) irtcorup/cve or imrfjTcierdy detailed rrcords-recorded details may depend on administrative and financial arrangements. i e. fee-for-service payment will usually generate more d&a than contract or capitation methods. Aggregation within sources may be limited to the range of particular programmes. i.e. medical and hospital care may be better documented than dental and optical services, etc. Reference to multiple sources increases collection costs and also raises the probability of error. Cost may be the main factor influencing the choice of method. Household surveys can provide more comprehensive data. but the validation of information subject to imperfect recall may prove expensive. Sampling and interview procedures may also be costly, unless combined with. or based on, other social surveys. Reference to secondary sources is less expensive and avoids some validation problems. but may be too narrowly based at present. However. some of its defects can be remedied. In many countries, developments in the administration and financing of health services can be expected to broaden the base for routine statistical collections. in terms of both population coverage and the range of activities included. Though studies of family health will require more than these administrative statistics. clear specifications of need would make a very useful contribution at this stage.
5.
SUMMARY
Though family structure may bear on several aspects of health economics. this working paper has concentrated on factors influencing family decision-making in the consumption of specialized. external services.. Previous studies indicate that family size. composition and income all affect utilization. but investigation of more specific economic factors requires more data on costs (including opportunity costs) than is generally available. Since the concept of family useful for cpidemiological and sociological work may be too broad for cconomic stud&. it may be preferable to dcfne it in terms of constituent “income units” more directly related to decision-making. Though some special statistical collections stem unavoidable. the cost of interview surveys may bc prohibitivc. Attention should thcrcforc be concentrated on the dosign of routine collections. particularly in coun-
tries where prospcctl\e changes in the dcli\.cr! health care offer unique opportunities.
01
REFERESCES
I Grossman Empirictrl
M. T/w D~~trrd /or H~/t/r : -I T/w~w,~~I~‘11 Illl,rstiMtrtiorl. unpublished Ph. D disscrtatton.
Columbia Univrrsit! 1970: T/I~ RIJ/L,ot EffiL,rl~,lc,\. fl,rl~~~~fer.sit1 Comsm~ Dr~ntrd for Heir& .wII,(’ t w)’ pi’limim7ry results. Uational Bureau of EconomicResearch Inc.. New.York (mimeographed). 2. Weeks A.. Fa~lilx Slwndi,q Ptrtt~rm uttd Hc~~l/til CL~IX,. Harvard. Health Information Foundation. 1961. 3 Litman T. J.. Health cart and the fatnil! : a threc-gcncrational analysis. .\frtl. Cow IX. 6i. 1971.
4 Andersen R.. and Anderson 0. W.. 4 5.
Dmrtk o/ H~irir Sucks. University of Chicago Press. Chicago. 1967 Last J. M. The Icebergcompleting the picttw m
htwt II, 2% 1963. C.. Epidemiolog! in Farnil! Practw. hfeci. J. Arrstrdin 1, 1101. 1969. 7 Andersen R.. Smedby B.. and Anderson 0. W. .\f~,~irt~tr/ Carr use it7 Swrrier~ rmtl the I’uftd Sttrtc.h Centrc of Health Administration Studies. Chicago. Research Series No. 17. 1970. 8. Andersen R., .-1 Br/woiownl .Cfo&/ of Ftw:/i~,.s C jL’ O/ Health SrrtYcus. Centre for Health .Adtnmtstrntton Studies. Chicago. Research Series No. 25. 196X. 9. Deeble J. S. and Scotton R. B. H&t/t CW L dcr LO/UW tar?:Imurmcr-Report of’rr SWI.~K Institute of .Applied Economic Research. L’nivrrsitL of Mclbournr. Melbourne. Technical Paper No. 1,.196X. I 0. Ashford J. R. and Pearson N. G.. Who uses health xrvices and why’? J. Roy. Stntisr. Sec.. .SWW,J .4. Part 3. p. 295-345. 1970. 11. Abel-Smith B. and Titmuss R. M. 7ltr Cost c!f t/x, Natiotlul Hedth Srrricr if7 Euyltrd trd ~I;I/L~F. Cambridge University Press. Cambiidge. 1956. I?. Simpson J. et d.. Custo,rt trrui Pvrrctice in .$fvdic~/ CMC. Nttffield Provincial Hospitals Trust. Oxford. 1968. 13. Krupinski J. and Staller A. Psychologtcal disorders. In 71~ Hcdth of’tr Mmopolis. (Edited by Stollrr and Krupinski), Melbourne Heinetnann. I97 I. 14. Cooper M. H. and Culver .A. J. An economic assrsstnrnt of some aspects of the organizttion of the national health service. In Hetrlth .S~rice\ Fimr~~cingq. London. British Medtcal Association. 1969. IS. Scotton R. B.. Membership of voluntar!, health Insttrante, Ecort. Rticord 45, 109. 1969. I6 Andersen R.. Anderson 0. W. and Stncdbe!, B. Pcrception of need and response to wmptoms in Sweden and the United States. ,Mrtl. C‘w, VI, IX. 1968. 17. Silver M. Variations in medical cupcns~s and work-loss rates. (Edited by Klarman H. E.). In E/q~ir,wtr/ Sro, wt/ic~r~/ wro ~~.\-pc~Ji~/irt,,c.\cl/~/ C’olwttrrj~ Hctrlth Imutu~~cc~ --‘I Fire- ~,YW Rc-.WIWI. Harvard University Press. Harvard. 1963. Appendix A. p. 129. 19. See Miller F. J. Sot,. Sci. & ,2lct/. 8, 1974. ‘0. Harper R. J. A. Survey of living conditions tn Melbourne 1966. T/w Ecom~rrrc Record 43. 371. 1967. general
practice.
6. Bridges-Webb
The economic
approach
21. Baker T. D. and M. Pcrlman Hculrh Mrr~~porwr ijr u Dcrc,/o/>i,q EUJWI~II~. .lohns Hopkins P&s. Baltimore,
1967. 2’. SW I .S Puhl~c Health %I-\icc.
Hcal~h
Intcr\,ic\<
Sur-
10 ramill
health
studies
vey. Na.ional Centre for Health Statistics. Series IO (Several issues). 23. For alternative. more limited. approaches, see Ashford and Pearson (II) and Dcehlc and Scotton (9).
APPENDIX
I.
For.
idit-itlutrlr
(I) age seh race marital status status in farnil! occupation industry (8) income- wage earnings business earnings other 19) health status (‘I assessment) I IO) episodes of illness (I I ) work loss (12) consumption of services
(2) (3) (4) (5) 16) 17)
current
? survey
or permanent?
period
( I I size and composition 121 residence and duration of residence (3) residence of related families 13) size of household (single or multi-family. etc.) (5) housing (type. area. amenities. etc.) 161 occupation and educational attainment of head of household (I social class ranking) (71 health historl of head and female head. (if different&? assessment. IX t no of income- units in farnil! 19) Income. h! source. from: wa,fe earnings busmess earnings current or permanent? other I (101expenditures from family budget for (at least): food housing ‘I survey period health services. by type education purchase of consumer durables 1 t I I1 transacuons with related families. (etc. e.g. gifts. assistance in kind or through services. etc.).
t I I Indi\idua)s (21 Households (3 I Families
533
x age. sex. race. marital status. occupation. industry x type. size. location x type. size. residence. income. education of head (? social class).