Theories and measurement in disability · Theories andmeasurementin disability activities of the...

Epidemiology and Community Health, 1979, 33, 32-47

Theories and measurement in disabilityR. G. A. WILLIAMSFrom the Medical Research Council Medical Socology Unit, University ofAberdeen

In an earlier paper (Williams et al., 1976) a modelof disability was described which gave rise to anappropriate measurement technique. The modelwas presented as 'an intuitive construct or conceptof disability', and attention was concentrated onvalidating the measurement technique in terms of it.The present paper, by contrast, attempts to drawout the theoretical assumptions of that intuitiveconstruct, and to compare them with two othertheoretical positions. The aim is, firstly, to developand demonstrate the explanatory power of themodel, and, secondly, to derive from its explanatoryprinciples further implications for the measurementof disability. It will be evident that the originalconception and the first results of the scalingtechnique already published have had a reciprocalinfluence on one another, and the theoreticaldevelopment is not easily characterised as eitherdeductive or inductive, but it is presented in aquasi-deductive form in order to test it againstfresh data and analysis.The scaling technique described in- the earlier

paper had both strengths and weaknesses. Itsstrengths lay in its objectivity, and in the veryminimal assumptions it required. It concerned theway in which disabled people demonstrablybehaved, yet in most cases their patterns ofbehaviour turned out to be such that uncontentiousjudgements of disadvantage could be made.Moreover, the scale had the unusual virtue ofdetecting those individuals to whom its assumptionsdid not apply.The weaknesses were in part those of the modest

level of rank-order measurement achieved. Certainquestions still had to be left to arbitrary judgement,including the relative ranking of differences betweenranks in the scale. An interval scale would probablybe needed for this task. There remained also thequestion of how to assign cases to whom the scaleassumption did not apply. Finally, it was not clearhow the theoretical model would determine therelevant universe of disability items. This paperattempts to make progress in these areas.

It is usual to distinguish between impairment,disability, and handicap, but there is an intermediatearea which requires the distinction of a fourth

category. Briefly, a condition may be described asbelonging to

1. A class of diagnoses ('impairment').2. A class of restricted activities defined in

relation to a limb or organ which performsthem, such as the movement of joints, or theflow of air from the lungs (sometimes'impairment' or 'impaired function' but alsosometimes 'disability').

3. A class of restricted activities defined inrelation to the purposes of ordinary dailylife, such as dressing or cooking (sometimes'disability' and sometimes 'handicap').

4. A rank of disadvantage assigned to a personon the basis of his performance on any of thefirst three descriptions, usually the third('handicap').

This paper is concerned with disability in thesense of the third description, as a limitation onactivities of ordinary daily life. It seeks to explainthe patterns of disabilities usually experienced. Itfirst considers them as products of impairment, inthe sense of either of the first two descriptions above.Then, after an intervening section, it considersthem as products of the way in which people rankdisadvantages when confronted with impairment.Old terms are used in order to avoid coiningneologisms, but the reader is asked throughoutto bear in mind these special definitions.The next section focuses on the apparently

commonsense position that the pattern of disabilityis determined in part by the impairments suffered.

THE MECHANISTIC THEORYThere is a view which assumes that as disabilityis caused by impairment, so the pattern of disabilityis caused by the pattern of impairment. If this canbe described as crude mechanism, the theory whichI outline here is refined mechanism. Refinedmechanism points out that the same impairmentcan cause disability in one case and not in another.A common example is the finger injury which ruinsthe career of a pianist or violinist but does not disablethe ordinary person. To explain the pattern ofdisability suffered, one must look not only at thetype of impairment but also at the customary

32

Protected by copyright.

on August 29, 2021 by guest.

http://jech.bmj.com

/E

pidemiol C

omm

unity Health: first published as 10.1136/jech.33.1.32 on 1 M

arch 1979. Dow

nloaded from

http://jech.bmj.com/

Theories and measurement in disability

activities of the person impaired. But since theimpairment remains basic, its presence in a limbor body system creates disability in the person whouses that limb or body system for an importantactivity. Consequently, if a disability appears atall, it is referable to the locus of impairment.

There is a range of examples which lendcredibility to this theory. Characteristically, anamputation or an injury is cited, but several otherconditions are so obviously local as to make goodexamples. With many impairments, however, it ismore difficult to define a specific 'locus' in the samesense. What is the site of a stroke, or of a circulatoryproblem, in this context? There are at least two waysof answering such a question. One is to concentrateon the site of symptoms, the other is to pursue theorigin of the impairment in diagnostic terms, and toconcentrate on the site of the disease. Neitherapproach will, however, define uniquely the locusof many impairments. Symptoms are often widelydistributed through the body, and the numberof possible disease sites is so large that furthergrouping of them is necessary. A mechanistic theoryof disability patterns would have to fix on crucialsymptom sites: to rank symptom sites in order ofimportance for ordinary activities, so that patternsof symptoms which included several sites wereclassified under the most important. Alternatively,pursuing the diagnostic approach, the theory wouldhave to fix on crucial disease sites: to specify thebody systems on which different activities primarilydepend, so that disease sites would be classifiedunder the body system to which they mainly con-tributed.

Supposing that all this can be done withreasonable agreement, the mechanistic theory isnevertheless not nearly as obvious as it appearsto be in its characteristic examples. Moreover,some impairments which have an unambiguous sitein one sense, such as a damaged lung or heart,are not easily seen as impinging on a specific groupof activities. These impairments obviously putan end to athletics, but for the most part theysimply require that activities should be carried outat slower speeds according to the time and energyavailable.

Finally, the very obviousness of the link betweenimpairment and disability may be suspicious, for thevalidity of a commonsense axiom depends on thepurpose of its use. The usual context in which thelink between impairment and disability is made iswhen someone is offering an 'account' of hisfailure to meet his normal duties (Scott and Lyman,1968). He justifies a specific disability on the groundsof his impairment; and in some circumstances,particularly when he has a highly visible and locally

specific impairment, his claim may be more easilyadjudicated than when a general loss of energy isinvolved. Consequently, the locus of impairmentbecomes a well-known resource for adjudicatingthis kind of claim. However, when the purposeat hand is explanation, not justification, and whenthe whole pattern of disability is involved and notone salient activity which happens to be relevantin the context, the link between impairment anddisability is axiomatic only in the very broad sensethat impairment sets some form of limit to activityin general. It may not be surprising, then, that arecent survey of research on rehabilitation remarksthat 'We are thus left with the uncomfortableconclusion that traditional clinical measures ofoutcome are unrelated to functional* capacity'(Nichols, 1975). The next section explores analternative way of conceiving the relationship ofimpairment to disability which may account forthis.

RATIONAL CHOICE THEORYIncreasing breathlessness, because of malfunction ofthe circulation or the lungs, is a good example of animpairment which acts mainly to slow down thespeed at which the sufferer can do things, to forcehim to divide into several stages an activity whichhe would otherwise do in one, and hence to causehim to cut down the daily outgoings from hisbudget of time and energy. The analogy with amoney budget is obvious. There is a fixed 'income'of time to be spent, perhaps also (although thisneeds more qualification) a fixed 'income' ofenergy. It is also plausible that there is a normal'budget' of expenditure on different items ofactivity, which is reached by a complicated trade-off of the varying 'utilities' of these items ofactivity in relation to the time and energy eachdemands. In this rendering of reality, impairment islike an increase in income tax, to which theconsumer must respond by cutting down onaspects of his activities that he can best afford tolose.More precisely, a disabled person who tries to

match the cost of his effort to the importanceof his different activities is very like the managerof the simple economy with which most economicstexts begin. First he produces his own activities invarying combinations. At any point he is able tosay how many units of activity A he would add tohis day by giving up a unit of activity B. He isable to state the ratio at which he wouldsubstitute A for B, whether (1) in terms of what theycost him in time and effort, or (2) in terms of whatincrement of A would just compensate him inFunction here means behaviour.

33



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


R. G. A. Williams

utility for a loss of B, so as to leave him indifferentbetween the two combinations. He can state (1)a 'cost ratio' and (2) an 'indifference ratio'.

All possible cost ratios between two activitiesfor a given supply of time and energy can be drawnon a graph as a continuous line YZ (Fig. 1). The

u

a

y

S

x z

activity B

Fig. 1 Rational choice model of disability.

line could be straight, if both activities were

performed equally efficiently by any combinationof time and energy; or it could be curved, if foreither activity a special combination of time andenergy was more efficient than others. The curvechosen does not matter in the present context, andthe line has been drawn with the convex curveusually found in economics.A similar line PQ can be drawn of all possible

ratios of the two activities which leave the disabledperson indifferent compared with one given combina-tion of the two activities. The line can berepeated for more and more plentiful combinationsof the two activities, and the disabled person isassumed to prefer more plentiful combinations,that is, more of both activities, to less of bothactivities. The lines RS and TU represent in-creasingly plentiful combinations. Because a mixedpattern of activities is usually preferred to onlyone activity, this line is usually drawn with aconcave curve, as in the lines PQ, RS, and TUin Fig. 1. Again, the shape of the curve does notmatter here.The point D, where the curve RS just touches YZ,

is the point where the disabled person, using all thetime and energy at his disposal, can just achieve onecombination out of the most plentiful combinationswhich he would like equally well. This point isusually unique-the other plentiful combinationswhich he would like just as much are out of reach

-but there could be several such points for all that itmatters here. The important conclusion here is thatat all other points on the cost curve the disabledperson is still fully employed but to less advantage.Conveniently for what follows, the optimum pointis characterised by a single ratio of substitutionbetween the two activities which is the same for theirrelative costs in time and energy as for their relativevalue to the disabled person. In terms of the graph,the tangent of both curves at D is the same.

If impairment now reduces the total level ofactivity of the disabled person, what happens? Interms of Fig. I the cost curve shifts inwards, sayto the dotted line WX. A new optimum point E isestablished. Both activities are lost in a certainratio. The ratio in which activities are simultaneouslyreduced need not be the same as the ratio in whichthey are substituted for each other, but they will infact be the same if the disabled person's indifferencecurves and cost curves remain parallel at thereduced level of activity. The practical meaningof this condition can be illustrated by an example.When healthy, a woman cooking eight hot meals aweek and going out about six times a week may beable to trade one hot meal for one extra trip out ofthe house. If seven hot meals and seven trips appearjust as good to her, the odds are that her optimumpoint is somewhere between these two combinations,and for the next hot meal, which she could give upfor another trip out, she would want a little morethan one extra trip to compensate her. Now impair-ment reduces her capacity both for cooking and forgoing out of the house. On the whole it would beexpected that her best combination would still lie ina roughly equal division between cooking hot mealsand going out, say five hot meals a week and four tofive trips out of the house. This expectation isstated more precisely in the assumption that thecost curves and indifference curves of the twoactivities remain parallel at the reduced level ofproduction. The woman might, of course, stick tocooking seven hot meals and reduce her trips outto two or three, but in that case something quitestartling must have happened to her originalwillingness to trade one activity for another if itsuited her. Perhaps she always cooked seven hotmeals a week on principle, and therefore would nottrade below that level in any circumstances. If itwas not a matter of principle, but still of tradingpreferences, then either the relative costs of cookingand going out change quite sharply when only alittle of them is done, or their relative value changessharply.A 'sharp' change means something precise. If a

reduction of overall activity brings some reduction,however small, in every activity, the change in

34



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



relative cost or relative value is not sharp as theword is meant here. If any activity actually increasesor remains unchanged, then the change is 'sharp'.An increase of an activity does not mean that ittakes longer-that would be an increase in its cost.It means that more of it is performed-for example,eight hot meals are cooked instead of seven.Although this is an unlikely result, it is notimpossible for relative costs or relative values tochange sharply at low levels of subsistence. Theme isa standard economic example. As a man gets poorer,he may eat more potatoes or bread. Thus theremay be activities which, like potatoes or bread,are an 'inferior good'. Watching television isprobably one of these activities.

It will be assumed, for the moment, not that thecost curves and indifference curves remain parallel,which is too strict an assumption, but that they donot change 'sharply' in the sense of revealing anyparticular activity as an inferior good. Meanwhilethere is one other situation in which a decreasein overall activity may lead to an increase in aparticular activity. This situation is one in which adisabled person is employed in some forcedcombination of activities which is a considerableway from his optimum, and then, when his overalllevel of activity is reduced, is able to choose thecombination which he himself judges best. Thischange is represented in Fig.- 1 by, for example, amove from F to E. It will also be assumed for themoment that this situation does not hold.The two assumptions thus made allow a simple

calculus to be presented of the disability patternswhich ought to be found in practice, if impairmentdoes constitute a general increase in the costsin time and energy of all activities, and if peoplerespond to this by budgeting losses in their activitiesaccording to their own preferences. The cost increasein this calculus is presented like a sequence of higherand higher 'tax' demands which must be paid withunits of activity. The 'taxes' are in increments ofone cost unit at a time, the unit being set equal to thecost of the cheapest activity unit being considered.The calculus has the following elements.

Firstly, for the sake of simplicity, each activityis reckoned to have a critical point above which aperson is non-disabled (= 0) and below which he isdisabled (= 1 or more). Thus the differentiation ofdisability patterns does not go beyond the case whereall activities are disabled in this sense. If the criticalpoint (as in many disability measurements) is thereceipt of personal help, the present rational choicemodel would tend to predict that when allactivities are disabled, the help received with all ofthem is greater the less the importance which thedisabled person attaches to performing the activity

35

himself. Broadly, the picture would be one of a littlehelp received in everything, and nothing doneentirely for the disabled person. The meaning thusgiven to disability in all activities needs to beborne in mind, because it is theoretical and may notcorrespond to the reader's experience. The focushere, though, will be on the transitional patterns ofdisability and whether these correspond to ex-perience. If not, then this picture of overalldisability will not do so either.The transitional patterns might not be particularly

interesting if they passed by rapidly. This wouldoccur if activities were lost in very small units so thatchange was virtually continuous, as it is with mosteconomic goods. But there is plenty of evidencethat the transition is slow and that disability in allactivities is extremely rare, and this could not be so ifactivities were traded in very small units. Moreover,it is well known that many activities are performedin 'packages' which are worthless if not completed.A trip to do the shopping is wasted if the shoppingcannot be done. Hence it is necessary in studying atransition to focus more closely on the stepwiseinterchange of whole units of activity than is usualin economic applications of rational choice.Consequently the second element of the choicecalculus presented requires losses of each activityto be paid in units, not fractions of a unit.

Thirdly, since a reduction in overall activity hasbeen assumed to involve a reduction, however small,in every activity, a positive number can be assignedto each activity which represents the ratio at whichthe next unit is lost compared with the next unit ofother activities. It is not necessary to know what theratio is in fact; one need only represent the rangeof what it could be. It is easiest to present this figureas if it were a price: that is to say, if four units of Aare lost for one unit of B, it is easiest to say that oneunit of B is four times more valuable or costly thanone unit of A. In the calculus B then appears witha weight of four and A of one. Taking as an examplea world with only four activities, weights can beassigned to represent varying degrees of inequalitybetween these activities. Two schemes are presentedin Table 1, one with uniform weights, the other withunequal weights in arithmetic progression.

These two schemes of weighting naturally do notexhaust the variations in weights for disabilitywhich have so far been suggested in the literature.They do represent, however, the ordinary rangeproduced by rules of thumb. One common rule ofthumb is to assign each disability a score of one andtreat the total as the disadvantage suffered. Moresensitive schemes, on the other hand, partitiondegrees within one disability, assigning themsuccessive scores of, for example, one, two, and



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


R. G. A. Williams

three; or distinguish major and minor items withscores perhaps of one for minor and two for majoritems. The general effect of such distinctions is tointroduce unequal weighting for disabilities whichwill approximate at most to the ratios of anarithmetic progression. There are also theoreticallysophisticated approaches which claim, thoughtentatively, that professional judges weight healthstatus in similar ratios. Hence the two schemesshown here represent practice in assessing disability,and offer the justification which it requires in termsof a theory of consumer preferences.

Fourthly, the interest is in the next unit of eachactivity to be lost, starting at the point where allactivities are just non-disabled. This means that-,because some activities lose that unit before others,a snapshot is required of the process of transition upto the point when the next units have been lostfrom all activities. If the next unit of activity B isfour times as valuable as that in activity A, then thecalculus must normally allow payments of fourunits of A to be made but no more, at least until thenext unit has been paid from activity B. At thispoint all activities are in any case disabled. Thusthe calculus must normally show losses of activityin the stated ratio of preference. The word'normally' is used to show the essential pattern, butreality is a little more confusing, because activitiesare lost in whole units. The point at which allactivities are just non-disabled is a point in themiddle of the process of loss, where it may be, forexample, that four units of A have been lost sincethe last unit of B was lost. Even if B is four timesmore valuable than A, it is now B's turn to lose aunit. The disabled person will give up no more ofAuntil he makes another sacrifice of B. The ratioin which he loses A and B remains the same,however, as that of another disabled person whohas just sacrificed a unit of B, and who consequentlywill sacrifice no more until he has lost anotherfour units of A. The variation in disability patternsresulting from this has limits which are describedby a formula.*

It may seem arbitrary, at first sight, to say that thecalculus starts at the point where all activities arejust non-disabled. There are people who allowsome activities to drop well below a 'normal'minimum while maintaining others well above it.The minimum 'non-disabled' level could of coursebe defined as that which each disabled personwould trade all activities to maintain, so long ashe also kept above his minimum level for anyactivity. Public definitions of disability could thenbe treated as mere approximations to, or 'averages'of, such perceptions. An alternative approach is tomake use of the fact that personal help plays a large

part in defining disability. The significance of suchhelp lies in its being 'altruistic', that is, not fullyrewarded at the expense of the recipient. Itsignifies that a social minimum has been reachedby the disabled person which his relatives (or thetaxpayer) do not want to leave unremedied evenat their own cost. Because the disabled person canswitch his resources of time and energy from oneactivity to another, his passing the social minimumon one activity will not stimulate help if he couldswitch resources to that activity without threateningto pass the minimum in others. But whatever theapproach, the non-disabled level, even if it is definedseparately for each activity, tends to be a minimumlevel of production beyond which any activity mayreceive help. This is where the calculus begins.

Table 1 represents the disability patterns whichwould rationally be chosen if impairment acts so asto increase the costs of all activities and thesufferer responds by budgeting losses of activitiessolely according to his preference. The weightsrepresent the values or costs set on each activity,and the rows show the activity units lost so as tolose minimum value for each cost increase. Therules of the calculus which have been discussed aresummarised as follows:

1. Cost increases must be paid in whole units ofactivity, not fractions.

2. Multiple units of one activity must be paidaccording to their preferred ratio of loss,allowing only for uncertainty about whichactivity has most recently lost a unit (seefootnote below).

The value of an activity unit P has been defined in relation to its rateof loss. If Pxi + Pxj + . . .Pxn = K where K is the total losssustained and x, as multiplier ofP, gives its rate ofloss, then xi is defined

by the equation xi = Pi xj. However, because the xi must be whole

units, not fractions, uncertainty enters, and either of the followingequations may be equally likely:

(1) xi = [pixj

(2) Xi = (xj-

where the square brackets signify the integral part of the value. Theprobability of all points in the interval between these two values of xiappears, in the absence of any reason to deny randomness, uniform.Hence the limits of uncertainty can be stated as the following con-straint:

Xi >[i (xj-1]

and all solutions for K which conform to this requirement can betreated as equally probable.

36



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



Provisionally, it would appear from Table 1 thatthese rational choice schemes in which the onlyconstraint on choice is a sort of budgetary limitset by impairment would generate surprisinglydiverse patterns of disability. If the patterns arein fact more restricted, however, the task will be toidentify the additional sources of constraint. Themechanistic theory described earlier may be ableto show one source of constraint. The next sectionidentifies another. It may be that a rational choicetheory could apply only within the area of freedomleft by these constraints.

Table 1 Weighting disability. Two typicalfour-itemschemes: (A) uniform and (B) arithmetic progressionSCHEME AWeights= I 1 1 1

'Tax' Scaling pattern 'Tax' Scaling pattern

1 0 0 0 1 or 3 0 1 1 1 or0 0 1 0 or I 0 1 1 or0 1 0 0 or I 1 0 1 or1 0 00 1 11 0

2 0 0 1 1 or 4 Fromthis pointall0 1 0 1 or items lose one or1 0 0 1 or more units.0 1 1 0 orI 0 1 0 or

SCHEME BWeights =4 3 2 1

'Tax' Scaling pattern 'Tax' Scaling pattern

I 0 0 0 1 6 1 0 1 0 or2 0 0 1 0 or 0 1 1 1 or

0 0 0 2 1 0 0 2 or3 0 1 0 0 or 0 0 2 2

0 0 1 1 7 1 1 0 0 or4 1 0 0 0 or 1 0 1 1 or

0 1 0 1 or 0 1 1 2 or0 0 1 2 0 0 2 3

5 0 1 1 0 or 8 1 1 0 1 or1 0 0 1 or I 0 1 2 or0 1 0 2 or 0 1 1 30 0 1 3 9 1 1 1 0 or

I 1 0 2 or1 0 1 3 or0 1 2 2 or0 I 1 4

10 1 1 1 1 or1 0 2 2 or0 1 2 3

11 1 1 1 2 or3 2 0 1 or3 1 2 0 or4 2 1 0

12 1 1 1 3 or0 2 1 4

13 1 1 2 2 or1 1 1 4 or0 2 2 3

14 1 1 2 3 or0 2 2 4

15 From this point allitems lose one ormore units.

As the amputated limb is the classic case for the

mechanistic theory, and breathlessness for therational choice theory, so there is another kind ofclassic case which suggests the third theory. I have

referred to the role of impairment in giving an'account', a justification, of disability. Localimpairment offers a simple and visible resource foradjudicating such accounts. By contrast, the freeplay of rational choice may make such accountsvery difficult. Rational choice would offer such avariety of disability patterns, based on suchabstract principles, that adjudication by families,workmates, friends, nurses, and doctors would bemade very difficult. These considerations suggestthat an important constraint on the pattern ofdisability may be its conformity to convention, aconvention which is necessary for communicatingthe legitimacy and ordinariness of the disabledperson's failures to those around him. Thedevelopment of this idea will require reference tothe sociological theory of deviance.

THE DEVIANCE THEORYTwo valuable discussions of disability or handicapas forms of deviance (Freidson, 1972) or stigma(Goffman, 1963) emphasise the restrictions placed ona disabled person by the necessity of his beingclassified or 'placed' by others. In Goffman'sanalysis of stigma the 'others' are usually met inpassing, in streets, at parties, at shops, in applyingfor a job; and the categories against which they try tomatch the disabled person are correspondingly crude.It is the failure of the disabled person to match one ofthese crude categories, and the attributes felt to be'ordinary and natural' for members of it, whichconstitutes stigma. In Freidson's argument, bycontrast, the 'others' who must classify the disabledperson are primarily agencies of rehabilitation orcare. Here it is not the failure of the disabled personto match a category but his success that is significant.Success means that he represents a certain sort ofwork for the agency; he receives a definite responsefrom the professional, whereas he receives anembarrassed or equivocal one from the stranger. Inaddition to these 'others' with whom the disabledperson must communicate, yet another group isinvolved in Freidson's argument, the 'public' or the'community', who are tempted to label the clients ofagencies in their own way, organising for them asegregated deviant role. Unlike Goffman's public,this public has a definite category for the disabledperson, and definite notions of how to respond tohim; he is to act the role of the blind man, thevillage idiot, or the cripple.

Goffman's public are strangers and Freidson'sare neighbours. Between both authors, thereemerges a nascent theory about the classificationof disabled people by strangers, neighbours, andprofessionals. But among these social groups thereis a conspicuous omission: the family. The

37



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


R. G. A. Williams

understanding of his position by members of hisfamily is obviously critical for a disabled person;but in classifying, naming, and understanding hisposition they are, of all people, the least likely simplyto follow the patterns so far set out. Of course,elements ofthose patternsmay well be present; sexualstereotypes are so strong that husbands or wivesmay continue to be embarrassed by a sense of thespoiled identity of their partners; families areclosely involved in defining their members as workfor professional agencies; and families may alsoorganise for themselves butts, scapegoats, invalids,and other deviant roles. But families in which adisabled person is seen simply as a spoiled manor woman, as a butt, or as somebody else'sbusiness, reveal a surprising crudity of classifica-tion and understanding. In what ways, then, shouldthe more intimate knowledge which families haveof their members be manifested?The chief crudity of classification, which seems to

belong more appropriately to encounters in themarket place, the agency, and the street than tofamily relationships, is the tendency to dichotomise.It is only a highly specific relationship or a verysimple social order which offers two alternativelines of response, acceptance or rejection, the offerof a fixed role or the avoidance of contact. Morediffuse relationships, involving a complex ofexpectations, not all necessarily compatible witheach other, allow the persons taking part to bematched against what is expected of them in degreerather than in kind. The argument is that a notion isheld, primarily by family members, but secondarilyalso by professions, of relative deviance from usualexpectations. This notion very clearly organisesresponse to a disabled or sick person and, indoing so, directs the sequence of disabilitiesexperienced in the course of illness and impairment.For anyone recently recovered from an illness, it is

instructive to review the steps that marked hisrecovery in the talk which passed between himand those looking after him. The steps are veryfrequently the product of a social activity whichmay be called 'constructing progress'. In thisactivity, a curriculum is set up which can bear, notonly to the patient but more importantly to hisnurses, the meaning of uniform advance. Thelogic by which this is achieved is that of cumulationon a single dimension. Just as children aresocialised by a long curriculum in which thecapacities gained earliest are expected to beretained while new capacities are added, so the sickor disabled deviant is socialised or rehabilitated by acurriculum in which activities are gained andretained in due order along a single dimension.Furthermore, in both cases, since the curriculum is a

negotiated order, idiosyncratic pleasures of theunsocialised person will take second place toactivities typical of his conventional position. Thecumulative and conventional characteristics of thisprocess naturally limit severely the variety ofdisability patterns available on a system ofrational choice.The benefits of the process lie in justifying and

making intelligible the pattem of the disabledperson's failures, and in restricting the immediatedemands upon him to those he can properlymanage. Just as a normal child is measured by hisprogress along the curriculum, so the deviance ofillness is made normal with reference to a similarconstruction of progress. The adult patient andthe child are both offered a succession of temporaryidentities with their own ordinary and naturalattributes which are recognised by the members oftheir social group. At the same time, there is afurther use of the curricular logic, and that is tocreate a measurement of distance between temporaryand ultimate identity as a full adult member ofsociety. (Of course childhood illnesses have theirown curricula aimed at restoring the child to theactivities appropriate to his age). The principaldifference between the way distance is createdfor the child and for the adult patient lies in thedirection of movement through the curriculum.The flow along the educational ladder is one-way,with a continuous succession of new cohorts at thebottom; the flow on the rehabilitative ladder forms agreat variety of two-way movements, with successionof new cohorts from the top. Hence, distance on theeducational ladder can be measured by average'years' of attainment, and a child is normal for asix-year-old or is abnormal in having a mentalage of four; but distance on the rehabilitativeladder is measured rather by dispersion below thenorm represented by the healthy. Where ordinaryexperience is the norm, the distance of a deviation isthe comparative rarity of its experience. Thisproposition does not make all rare experiencesdeviations; on the contrary, only those conditionsof a person which are marked by others asundesirable, and which are ordered into a re-habilitative curriculum, can be said to be deviations(in their eyes) having a property of distance fromthe norm; and it is the distance of this kind ofdeviation only which it may be appropriate toconstruct on the basis of the relative rarity of itsexperience. That sick or disabled people makesome computation of the frequency of theirexperience amongst others, and assess their good orill fortune on that basis, could be illustrated in anumber of examples. Age group enters into thecomputation, more specifically in later years, but

38



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



since older people tend to attribute a componentof their distance from the norm to age, and acomponent to illness, the essence of the computationmay be fairly uniform among adults, relating towhat they know of the incidence of incapacity in allage groups.The particular consequence for disability patterns

of the present discussion is the hypothesis thatfailures in activity which are associated with one'sidentity as a normal person in the eyes of others willbe found to cumulate along a single dimension whichalso corresponds to the frequency of occurrence ofthe component disabilities. Diagrammatically, atypical set of disability patterns and prevalencesaccording to this theory is shown in Table 2.As before, an activity performed is 0 and anactivity failed is 1.The pattern in Table 2 is important because it

corresponds to the shape taken by a Guttmanscale. One of the ways of computing a Guttman

Table 2 Deviance theory: typical disability patternsDisabilitypattern Activity items

A B C D1 0 0 0 12 0 0 1 13 0 1 1 14 1 1 1 1No. failing each item 20 40 60 80

scale is to order the items initially according to theirprevalence, and this, too, corresponds to therequirements of the hypothesis. Similarly, for a

Guttman scale to exist in a given body of data,items must be failed or passed in a cumulativeorder, and this order must be uni-dimensional forthe whole group of people who are assessed.Consequently the first test of the deviance theoryof disability is whether, when disability items areordered by their prevalence, the patterns ofindividuals fall into a Guttman scale.However it is not necessary for all individuals to

follow the same scale, but for all who can be shownto share a common role within a common socialgroup or culture to follow the same scale. No greatprecision can be shown about this. It has alreadybeen suggested that it is not merely families whoare involved in constructing the position andprogress of disabled individuals; families andprofessionals exchange cues on the subject, andfriends and neighbours are additional witnesses,commentators, and agents of conformity. But if therelevant social group is much bigger than theindividual family, it is by no means predictable how

big it would be, nor where the typical construction ofprogress made by one group might differ from thatmade by another. All that can be said is thatdifferences may be looked for, and if they arealigned with different roles or with different socialor cultural groupings, that would be additionalevidence for the deviance theory against its presentrivals.

Again, it is not necessary that any items ofactivity which could count as disabled should forma cumulative scale. On the contrary, only thoseitems should scale which threaten the conventionalidentity of an ordinary person. This definition, too,is a cultural one, although one would guess thatactivities associated with adult identity would varyless in their nature than in their relative priority.

Finally, the Guttman scale is by itself a staticconcept; but the argument has proceeded from anessentially dynamic notion of constructing progress.Change in disability patterns thercfore offers afurther test of the thesis. Should all change occuralong the scale pattern? The argument would ratherbe that all socially controlled change should occuralong the scale pattern. Rehabilitation is certainlycontrolled but deterioration may not be, althoughone would expect the organising forces of conven-tion to bring a deteriorated pattern into order aftera certain period of time. If this is so, cases whichremain static after that period should also beorganised into scale patterns.Three theoretical approaches to explaining the

pattern of disability have now been presented. Thenext step is to compare them against the resultsof old and new data.

DATATwo sets of data have helped in evaluating the theoreticalpossibilities which have been outlined. The first set,from which some relevant results have already beenpublished in the earlier paper, consists ofhome interviewswith 157 women and 88 men aged 35 to 74 living innorth Lambeth, London, in 1967. All were randomlysampled from a census of the area and identified as beingdisabled in one or more of four areas of activity: self-care, domestic duties, mobility, and occupation. Generalaspects of the survey have been reported elsewhere(Bennett et al., 1970; Garrad and Bennett, 1971).The second set of data consists of two extensive pilot

studies carried out by the Institute of Social andEconomic Research at York on disabled people selectedfor the purpose by a number of medical and socialagencies in England and Scotland in 1973-74. Bothstudies include disabled people at home, in hospital,and in local authority accommodation. The first studyinvolved two interviews separated by three months, thusintroducing evidence on disability changes, and initialdata was obtained from 154 people living at home and111 in institutions. The second study, a single interview,

39



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


R. G. A. Williams

obtained data from 206 people living at home and 133 ininstitutions. Again, a report on the survey procedure hasalready been published (Wright, 1974).

DEVIANCE OR RATIONAL CHOICE

The data can first help to illuminate the varying inferencesderived from the rational choice schemes and the deviancetheory about the prevalence of varying patterns ofdisability. If rational choices about activities to be lostare made within a 'budget' on the two schemes so farconstructed, the Guttman scale hypothesis should berejected. If activities are lost in accordance with asymbolic construction of the impaired person's positionand progress, the Guttman scale hypothesis should beconfirmed.The earlier report already referred to showed that

Guttman scales obtained among men and women inLambeth. However, it was noted that, although highlevels of confirmation were achieved, the procedure couldconceivably capitalise on chance alignments of the dataand should ideally be replicated. Here, then, thereplications performed with the two sets of York dataare reported.The choice and definition of items presented many

difficulties. The Lambeth questionnaire had not beenformulated with deviance theory in mind, and selectionbetween the many coding distinctions it offered was madeearly in the development of the ideas presented here.But in retrospect the Lambeth items seem to have beenselected consistently with the theory, although they wereunlikely to have covered all the ground, and themobility items in particular stood as proxies for ranges ofactivities which should ideally be specified in terms ofordinary purposes (Williams et al., 1976). The codingdistinctions available in the York data presented two newquestions:

1. Which definitions fitted the Lambeth items mostclosely?

2. Which additional items were relevant?The decisions on these two questions were not easy. On

two items, it was necessary to take into account thedistribution of answers in interpreting the fit of the firstYork study. There is a risk in such reasoning, because thedeviance hypothesis partly concerns the relative preva-lence of failures on each disability item. However, theanalysis of the second York study could proceedautomatically from decisions made on the first, so the riskwas not serious.The second problem presented by the York data was

to decide which additional items should be included fromthose available. On the criterion sketched earlier, thatitems included should represent activities belonging toconventional adult identity, the following decisions weremade:Included-washing without help (question different

from Lambeth).Excluded-incontinence, cutting toe-nails without

help, pursuit of personal interests.The exclusion of incontinence, an item obviously bound

up with identity, was made on the grounds of the partiallyautonomous nature of bowel and urine retention,

disqualifying it as an activity in the sense of the otheritems.When these decisions had been made, replication

involved three things: firstly, the scaling properties of theYork data with the same order imposed on the same itemsas in Lambeth; secondly, the scaling properties andcomparative sequence of the same York items whenordered, as the Lambeth items originally were, accordingto the frequency of disabled responses; and, thirdly, thescaling properties of all the relevant York items, old andnew, ordered according to the frequency of disabledresponses.The conventional criteria of a Guttman scale are the

coefficients of reproducibility (values above 0 90) and ofscalability (values above 0^60: the SPSS computation,used here, is conservative, underestimating the co-efficient; see McConaghy (1975).The coefficient of reproducibility averages item repro-

ducibilities which are also quoted here, and whichshould not fall below 0 85. The relevance ofmany of theserules in a series of actual replications, rather than inestimating the replicability of a single instance, isdubious. In clinical and administrative dealings withdisability it is more useful to study the case, the aggregatebehaviour of groups of cases, and prediction about thefuture of the case, than the item response or the scale asa whole, and it is appropriate to extend the reporting ofGuttman scale properties to meet these uses.

Initially, therefore, reliance is placed on the repro-ducibility coefficients and the coefficient of scalability assummary indicators of the extent of scaling; but afterthat it is much more useful to study the proportion ofnonscale cases occurring, the proportion in which onlyone 'error' is observed, and whether there is any sign ofrepeated nonscale patterns.The first test presented reckons the reproducibility and

scalability Qf the Lambeth items when imposed in theLambeth order on the York data (Tables 3 and 4).Seven to eight items were replicable for women, four tosix for men, since not all the questions asked in the firstYork study were asked in the second. The three samplesare referred to in the Tables as L (Lambeth), N1 (thefirst national study done from York), and N2 (thesecond York study).

Table 3 Women: disability in Lambeth and nationalsamples

Item reproducibilitiesOrder as in Lambeth L Ni N2

1. Eat without help 1-00 0-97 -

2. Get out of bed, with help 1 00 0.99 0-993. Sit and stand without help 1-00 0-92 0-974. Use WC (L: or commode)

without help 1-00 0-97 0 975. Dress without help 0-94 0.91 0-966. Cook all meals (NI and 2:

usually) 0 90 0-92 0-937. Walk out of doors unaccompanied

(NI and 2: without help) 0-90 0-93 0-938. Do all own shopping (L: and

cleaning, laundry) 0-96 0-98 0.99

Overall reproducibility 0.95 0.95 0-96Coefficient of scalability 0*69 0*69 0*78Nos. in each sample 157 119 161

40



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



Table 4 Men: disability in Lambeth and nationalsamples

Item reproducibilitiesOrder as in Lambeth L NI N2

1. Eat without help 0.99 0-972. Get out of bed, with help 0.99 0-97 -3. Use WC (L: or commode)

without help 0.99 0 94 0 934. Sit and stand without help 0-98 0-91 0-985. Dress without help 0-88 0 97 0-986. Walk out of doors unaccompanied

(NI and 2: without help) 0-93 0 94 0-93

Overall reproducibility 0 95 0 95 0-94Coefficient of scalability 0-71 0-63 0 45

Nos. in each sample 88 35 45

In terms of reproducibility and scalability, tworeplications well above the conventional levels for singleinstances were obtained for women with samples of veryadequate size. For men sample size was problematic;also, a very high proportion in both Ni and N2 were notdisabled on any item: in fact only 14 cases in Ni and thesame number in N2 were disabled on any item. Resultsfor men should therefore be regarded merely as strawsin the wind, not as replications in themselves. Ni providesa fair omen; N2 has acceptable reproducibility but poorscalability. In the remaining tests scales for men will beignored.The second test presented orders the Lambeth items

in each sample by the frequency of disabled responseswhich the item generates (Table 5). Reproducibility and

Table 5 Women: disability in Lambeth and nationalsamplesOrder by item frequencies Frequencies ofdisabled responsesL Ni N2 L Ni N2

1 2 - I% 3% -2 1 2 1% 2% 1°/3 4 3 7% 12% 4%4 3 4 8% 6% 6%5 5 5 17% 14% 12%6 6 6 32% 28% 27%7 7 7 33% 54% 45%8 8 8 78% 82% 77%Overall reproducibility 0 95 0 95 0-96Coefficient of scalability 0-69 0 70 0-78

scalability would be expected to be high, since they werehigh for women on the first test. Here the interest liesin considering the stability of item order in replications.Item frequencies are also shown, and it will be noticedthat in studies of disability several items are very rare.Because of this, the tail of the disability scale may beliable to instability until enough numbers and enoughreplications of fairly numerous samples can establish thepattern. In the present case N2 presents the same order asLambeth, while Ni switches two pairs of adjacent items.Thus, although there is some instability in the tail of thescale, it is quite circumscribed, and suggests that largersamples might eliminate it.The third test proposed adds to NI and N2 the

washing item which was considered different from theLambeth item but relevant to the deviance theory. The

York question ran 'Do you wash yourself without helpfrom anyone'? The Lambeth question was very similarbut it was followed up by questions on help with shaving,combing hair, etc., for which a separate code wassupplied and included as a disabled response. When theYork item was added to the scales for Ni and N2,ordered by item frequency, the coefficients obtained wereas follows:

Ni N2Reproducibility 0 95 096Scalability 0 62 074

In terms of the order shown in Table 5, however, the itemwas interpolated differently in Ni and N2, betweenitems 4 and 5 in N2 and between items 4 and 3 in Nl.Although in Ni the item remains between items 4 and 5as in N2, the switch of items 4 and 3 which has alreadybeen described involves the present item in the anomaly.The patterns of disability derived from deviance

theory and from the two rational choice schemes cannow be compared. The upshot is so far in favour of thedeviance theory. In Table 1, the uniform weightingscheme would produce only 21 % of its transitionalpatterns in scale form. The weighting scheme whichfollows an arithmetic progression would produce agreater proportion of scaling cases (57%), but thatproportion is still a long way below the actualproportion observed if the first four Lambeth items aretreated as a separate scale (75 %). It is clear, therefore,that if disabled women budget losses of their activitiesaccording to their preferences, they do not behave as if thecritical unit of the activities which they lose had auniform value for them. Moreover, if these women, indeciding their preferences, give unequal weights to thecritical unit of an activity the loss of which makes themdisabled, the inequality is not as gentle as the ratios of anarithmetic progression. It looks at the moment as if asteeper progression may produce a result consistent withreality, and in this case it may be necessary to integrate(where that is possible), or else adjudicate, the claims ofdeviance theory and of rational choice. An alternative isthat, at successive stages in recovery or deterioration, eachof the activities considered in the data becomes an'inferior good' in the sense defined earlier, that is, theactivity is maintained or increased in substitution forother activities (whether mentioned in the data or not)which have been drastically curtailed. This is an importantand interesting possibility in explaining many activitiescharacteristic of the disabled, such as watching televisionor doing occupational therapy. But special evidencewould be needed to allow its relevance for the activitiesin these data, which are usually presumed to be per-formed to a higher standard by healthy people. For themoment, then, deviance theory offers the most im-mediately employable explanation of disability patterns,although the rational choice theory has potential whichhas not been fully explored.These findings mean that, if the mechanistic explana-

tion of disability now proves to be false, and thebehaviour of the disabled is explained by their values,then apparently most ordinary practice in assessingdisability (which tends to imply something betweenuniform and arithmetically progressing numbers for the

41



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


R. G. A. Williams

items) does not have the justification presumed of it, ofreflecting 'consumer' values. For if the consumer choicemodel of disabled behaviour is valid, it is valid only forsome steeper progression yet to be tested.

DEVIANCE OR MECHANISTIC EXPLANATION?The mechanistic theory did not entail a cumulative andunidimensional scale of disability, as it was propoundedearlier. But it need not be inconsistent with such a scale.Indeed, a competent authority might be able to draw out acumulative pattern in the impaired sites which aretypical of major diagnostic categories such as arthritis orstroke, and it is not inconceivable that the patterns wouldhave enough in common to allow a unidimensional andcumulative scale of impairment sites to be constructed.Attention is directed here, though, to testing the lesserbut necessary claim that particular impairment sitescoincide with particular levels of accumulated disability.The Lambeth data included an extensive documenta-

tion of impairment which was not available in the Yorkdata. The organising principle was the diagnostic onereferred to earlier, in which the disease site is named andrelated to a classification of body systems in the aspectswhich impinge on activity. Four groups of impairmentswere thus defined: locomotor, internal, sensory, anda residual category which was primarily mental.However, an examination of the data underlined thecomplexity of symptom patterns: 94% of Lambethwomen had symptoms which could provisionally beassigned under two or more of these headings, anddiagnostic skills were essential in arriving at the under-lying classification of impairment site. If, however, itwas argued that for the women their symptoms weremore important than the seat of the disease in de-termining what they could not do, there would be a needfor some different way of defining which symptoms hadpriority. This second classification has been attemptedon the basis already discussed, that symptoms evidentlyassignable to specific limbs form the classic case in themechanistic argument. Hence symptoms in upper or lowerlimbs, or in both, should coincide particularly clearlywith specific levels of accumulated disability, and thesesymptoms are frequent; indeed they are found in themajority of the disabled women.

In Tables 6 and 7 both classifications of impairmentsite are related to cases which scale perfectly among theLambeth women. These cases form the critical test for theability of the mechanistic theory to explain the basicstructure of disability patterns. The separate implicationsof nonscale cases are treated below. Symptom sites areunrelated to scale structure. The argument must thereforerest on the relation ofimpairment sites to accumulation ofdisability, and here the evidence of Table 6 is ambiguous.A relationship between impairment site and scale

structure certainly exists. It permits the prediction thatlocomotor impairments will more probably be associatedwith heavier disabilities, and that internal impairmentswill, on the whole, be associated only with the firstlevel of disability. But the prediction is far moreaccurate in reverse-given the severest accumulationsof disability, it is very probable that locomotorimpairment is involved. It appears that impairment site

Table 6 Scale types by impairment siteScale types: number of disabilities

Impairment site 1 2-4 5-11 N

A. Locomotor 16 10 13 39B. Sensory 4 3 0 7C. Internal 20 9 2 31D. Other, including mental 3 5 1 9

43 27 16 86

x2 (A versus B + C + D) 10-223 (2 df).P <0-01.

Table 7 Scale types by symptom siteScale types: number of disabilities

Symptom site 1 2-4 5-11 N

A. Lower limbs 16 10 10 36B. Mixed limbs 2 1 1 4C. Upper limbs 7 5 1 13D. Central 18 11 4 33

43 27 16 86

x2 (A + B versus C + D) = 3-916 (2 df).P >0 10.X2 (A + B + C versus D) = 1 *495 (2 df).P >050.

as a causal agent, and therefore as a predictor, issomewhat unspecific. This finding gives very uncertainsupport to a mechanistic explanation.The nature of the uncertainty may be clarified if the

role of impairment in the three theories is recapitulated.From the rational choice and the deviance perspectives,impairment acts only so as to impose a general 'tax' ontime and energy. From the mechanistic perspective,the site of the impairment is a specific agent creatingspecific patterns of disability. Table 6 seems to reflecta situation somewhere between these possibilities. TheTable is consistent with the 'tax' explanation if it is heldthat the highest rates of 'tax' are nearly always imposed bylocomotor impairment. This is the prediction fromdisability pattern to impairment which, as has been said,best suits the figures. The Table is, however, not easilymade consistent with any adequately specific version ofthe mechanistic explanation. It predicts from impairmentsite to disability level with relatively poor probability,and it distinguishes only shallow levels of disability.

It may be that the fault lies in the classifications ofimpairment which are so far available. The best hope ofimprovement would seem to be some more subtleclassification of locomotor impairment, but the obvioussubdivision according to the particular limbs in whichsymptoms occur does not appear to be promising.Meanwhile, the deviance theory remains a fairprovisional account of the structure of disability.But can the deviance theory be confirmed by further

evidence which shows not merely a structure of disabilityconsistent with it, but also variations in the structurewhich are attributable to group definitions of progressin rehabilitation? Two additional tests were suggestedearlier: the first, whether differences could be found in thedisability patterns associated with different roles orcultural groups; and the second, whether rehabilitated orstatic cases show a more precise scale pattern thanrecently deteriorated cases.

A2



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



That different scales should be needed for the differentsexes is an argument for the social construction ofdisability patterns which it is easy to overlook. Thedifference might, however, be merely between theactivities usually carried on; and this is no more than isalready conceded by what I called 'refined mechanism'.A difference of order in similar items between the sexeswould be more interesting. Such a difference was foundin the earlier paper. In addition, differences werereported between special subgroups of women and themajority of women. One special subgroup was unrelatedto site of impairment on either of the impairmentclassifications. It seemed rather to be related to theinsecure economic position of women in the old workingclass culture of central London. These women, for whomthe extended family is a major resource, were willing toaccept help with their cooking rather than limit theirmobility, by contrast with the priorities of most of thewomen. In this they accorded with the priorities of thoseLambeth men who kept house for themselves.A second subgroup of women preferred to accept help

with cleaning rather than, as was more usual, withshopping. The interpretation here was more ambiguous,for the women were distinguishable both by symptomand by household situation, and symptom site and house-hold situation were themselves interrelated. This examplecannot count as evidence for either the deviance theoryor the mechanistic theory, and the interpretation wouldhave to be resolved on the merits of the two theories in theother contexts. For further details, see Williams et al.,(1976).

Finally, differences in the scale organisation ofchanging or long static cases became accessible in theYork study Ni. Changes after three months wererecorded in women at home, and the results arepresented in Table 8. Although numbers are small,

Table 8 Disability changes after three monthsAt three months Nonscale cases °/. All cases

Static 1 2 5 40Improved 1 4-2 24Deteriorated 9 37*5 24

there is indeed a tendency for recently deteriorated casesto show more nonscale patterns, and the scaling ofrecently improved or static cases is remarkably precise.The tenor of these additional tests, then, is in favour

of the deviance theory.

AREAS UNEXPLAINED BY THE DEVIANCE THEORYIt remains possible that the deviance theory holds onekey to the general structure of disability, while themechanistic theory contributes towards explaining whatthe deviance theory cannot explain. The unexplainedarea is at its maximum defined by the nonscale cases.Can the evidence relate nonscale cases to the mechanisticperspectives?The two aspects of impairment are tabulated against

nonscale cases in Tables 9 and 10. Symptom sites areagain unrelated to scalability. Impairment sites againshow some relationship, and it is one which is intelligible

43

Table 9 Nonscale cases by impairment siteNonscale Scale

Impairment site cases cases All cases

A. Locomotor 28 53 81B. Sensory 2 8 10C. Internal 6 38 44D. Other, including mental 7 15 22

43 114 157

x2 (A versus B + C versus D) = 6 609 (2 dM.P <005.

Table 10 Nonscale cases by symptom siteNonscale Scale Anl cases

Symptom site cases cases

A. Lower limbs 16 45 61B. Mixed limbs 2 6 8C. Upper limbs 4 17 21D. Central 21 46 67

43 114 157

x2 (A + B + C versus D) = 0 605 (1 df).P >0 40.x2 (A + B versus C versus D) = I * 320 (2 df).P >0-50.

from the previous discussion. Locomotor problems are alittle more resistant to scaling than internal problems.There are two possible explanations for this, which arenot exclusive. Such problems may make conformity to ascale physically difficult, or they may constitute strikingreasons for justifying non-conformity. The secondpossibility needs consideration as well as the firstbecause of a further interesting and unexplained featureof the nonscale cases.There is a far wider potential range of nonscale

variation than in fact occurs. The majority of nonscalecases among women had a pattern where one unexpectedability was maintained amid the normal scale expectedfor someone with that number of items disabled.Represented graphically, an example on a five-item scalewould be 01011, or 01101 or 01110 where 00111 wasexpected when three items were disabled. Nonscale casesof this kind will be referred to as 'one-error patterns'.

This restriction on the occurrence of nonscale patternsis not easily explicable as the result of the physicalimpossibility, through locomotor impairment, of per-forming specified activities. But if it is supposed thatthe restriction derives from a partial and as yetunsuccessful effort of those caring for the disabled personto bring his pattern of disabilities into conformitywith his place on a socially constructed scale, then theremaining deviation he shows may be explicable by theacceptance of locomotor impairment as a licence fornonconformity.At the same time, because the one-error pattern is likely

to be seen as near to conformity, it would minimisemisunderstanding and friction with those surroundingthe disabled person. In this way the mechanistic theorymay have a part to play in the context of the deviancetheory.

IMPLICATIONS FOR THE ASSESSMENT OFHANDICAPThe earlier paper published on the Lambeth datashowed that assessment of handicap was simplified by



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


44

the Guttman scale structure found in disability. Theperson with seven disabilities was always worse off thanthe person with six, because his seven disabilitiesincluded the other's six. Thus, arbitrary judgements werevirtually eliminated from the main task of assessment, butthey continued to be necessary in three areas: inassessing the representativeness of the items ofdisability chosen for study against the universe ofrelevant items; in assigning nonscaling cases; and(where this is necessary) in assessing the overall extentof change in an experimental and a control group whereindividuals shift from different initial levels of handicapto new levels. From the theoretical discussion andfurther evidence set out in this paper, it is now possibleto suggest partial criteria for the relevance of disabilityitems, for assigning nonscale cases, and for generating aratio scale of disability, on the assumption, which remainsprovisional, that the deviance theory offers the bestaccount of the structure of disability.The universe of relevant disabilities has been defined as

those which threaten the conventional identity of an adultmember of society. This identity includes a range ofactivities which it is taken for granted that an adultvoluntarily and consciously carries out, and the neglectof which creates embarrassment in others and apresumption that an account or justification is calledfor. Proper management of the body and properparticipation in a private household have been cited asobvious examples which account for all the items usedas data. It is easy, too, to exclude as irrelevant personalhobbies, or those services which are normally performedfor each person professionally. But there is a widemargin of activities about which it would be possibleto disagree, and considerable room for furtherclarification. The definition emerging from the presentargument constitutes a start.Some progress can also be made in assigning nonscale

cases. Just as the deviance theory seems to account forthe structure of scaling cases, so it seems to account inan important respect for the structure of nonscaling cases.One-error cases appear to represent attempted ap-proximations to the conventional scale. On the face of it,this does not help much, for there are logically twopoints of the scale, one above and one below, to which aone-error case might be an approximation. A casetaking the pattern 101, for example, might be anapproximation to 111 or to 011. However, according tothe argument previously set out, an error case ap-proximates to a scale position only in the sense that thedisabled person concerned is being encouraged by hisfriends and relatives to conform to a logic of recovery.It would contradict this logic to encourage him to getworse, that is, to move from 101 to 111. The conclusionmust be that a person showing a disability pattern 101was being encouraged to achieve the pattern 011, but forreasons inaccessible to us the pattern 101 was accepted asa substitute. Whatever the explanation of one-errorvariations from the scale, therefore, it seems reasonableto assess their intended value as being equal to thescale pattern which contains the same number ofdisabilities. Other types of error remain uninterpretable,and their assessment requires judgement. But these more

R. G. A. Williams

variable error cases are a very small proportion of thedisabled, usually amounting to less than 10% in all.The criterion of a ratio scale is necessarily offered

tentatively. Numerous approaches to deriving such a scalehave been attempted, mostly by forms of rating andexpressed preference. The utilitarian philosophy whichunderlies these attempts implies a choice of values, andconsequently of representative disability items, whichmay not coincide with the perspective and the itemsrelevant in an explanatory theory like the one sketchedhere. Besides, pain and the risk of death are con-siderations in a utilitarian classification of health stateswhich are not taken into account in the presentdisability scale. However, if the premises advancedearlier are true, there are interesting implications forwhat a ratio scale of disability ought to look like.Although restricted to one perspective, these cannot beoverlooked in more ambitious attempts to evaluatestates of health.The implication is that, in items which people order

cumulatively so that the severest accumulations ofdisability are also the rarest, they are assessingacceptability of the disability as a function of itsfrequency. For the disabled person and his family, thefrequency of his condition is experienced in terms of thefraction of their acquaintance who are similarly afflicted,including in the 'acquaintance' those of whom they areaware without necessarily having met them. Therewill be many local variations in this experience. But itcan be approximated on average by the fraction of thegeneral population with the condition. It is importantto note that this is true only of scalable conditions.

Strictly speaking, on this argument, the frequency of ascalable condition represents not a value but adescription: the degree of conformity of the disabledperson to the norm. Conformity as a descriptive termconsists in being like everyone else in some presupposedmilieu, non-conformity in being the odd man out. If thegeneral population is taken as an approximation to thismilieu, then the degree of conformity of a condition isexpressible as the proportion of the general populationwho experience that condition, and it varies from 0(theoretical total non-conformity) to 1 (the possession ofcharacteristics common to everyone). The deviancetheory, however, supposes in addition that conformity isbeing treated as a value in the behaviour of the disabled.The priorities attaching to certain activities are negotiatedby the disabled, their families, and neighbours andothers, in a common subculture; and the value attachingto conformity ensures that this pattern of priorities is thenadhered to. The next question is what relationship thedescriptive scale of conformity bears to the evaluativescale. In the same way one might ask, for example, whatrelationship an evaluative scale of obesity bore to the scaleof body weight in avoirdupois.There is some plausibility in supposing flat-footedly

that the relationship of acceptability to conformity islinear. A scale is produced by these means which givesweight to extreme disabilities in much the same way asjudges have been shown to do in court decisions(Rosser and Watts, 1972). For example, a scalablecondition experienced by I % of the population is half



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



as acceptable as one experienced by 2%, and the ratio ofacceptability between them is the same as that betweenscalable conditions experienced by half of the populationon the one hand, and all the population on the other.But it is also probable that there is a limit on the diminu-tion of acceptability as conformity drops towards zero,which is set by the size of the population about whoma disabled person normally gets personal information.Acceptability may reach zero, in fact, before conformity.Similarly, to be the same as most of the people one knowsmay be sufficiently 'normal', and acceptability may reachits maximum well before conformity approaches unity.These possibilities are represented by points A and B inFig. 2. However, there are also reasons for being wary ofsupposing that the line AB is straight. To be uniqueamong the people one knows may well be especiallyunacceptable compared with being one of two or three.As conformity diminishes, the diminution of acceptabilitymay well be subjectively magnified (the dotted curve AB).On the other hand, perhaps it is mainly the first steps indeviance that count, in which case the continuous curveAB would be appropriate. Both these very simple curvescould be defended.

In the light of these examples, it is safer to proposeonly that the evaluative scale of acceptability whichapparently underlies the behaviour of the disabled is afunction of the descriptive scale of conformity and hasa true zero. Studies of how the descriptive scale isevaluated will be needed. But if the argument of thispaper is valid, any proposed index of health states whichincludes valuations of disability would need to beconsidered in relation to the descriptive scale ofconformity defined here, and the function of that scalewhich it represents would need, if it were other than thesimple function described here, to be defended withadditional reasons.What emerges, therefore, is a possible criterion of

validity for ratio scales of disability. It seems from theforegoing arguments that the true ratio scale of theacceptability of scalable disabilities must be a functionof their perceived rarity. Comparison with other attemptsto elicit a ratio scale may show that, with suitable pre-cautions for the extreme values, acceptability may betreated as a linear function of conformity. But greatcaution is still necessary, as with all attempts to imposenumbers on social life. Moreover, it is perhapsappropriate to stress here the other corollary of thearguments presented, that acceptability of disabilities alsovaries significantly with the perceived social network ofthe disabled person, with the role he fills, and with thenegotiation of that role and its various priorities withthe group to which he belongs. These considerationsmake it inappropriate, for many purposes, that a singlescale should represent valuations of disability for a wholepopulation. Rather, a family of scales is appropriate.Ultimately, therefore, it may be possible to generate anumber of interrelated scales with varying degrees ofrefinement for different purposes. One scale could beproduced which was valid within known limits for thegeneral population. Those limits would be stated by theproportion of nonscale cases other than one-error cases,since the deviance hypothesis can account for one-error

45

°o X0 A

A

conformityFig. 2 Relation of acceptability to conformity.

cases, and by the groups within the population whichshow distinct scaling patterns, such as the Lambethwomen who are economically insecure. From this overallscale, it would be possible to branch out into a numberof scales valid for these distinct groups, or for othergroups where additional items were relevant such asoccupation or schooling. With additional items, or withmore precise knowledge of the size and morbidity of thereference group to which individuals conform, furtherdistinct subgroups might emerge, and eventually ahierarchy of generalised valuations of disability could beestablished. At the most particular level, finally, wouldbe the clinician attempting to predict from an individual'scurrent disability pattern the next critical disability whichthat individual is likely to lose or regain. Dependingon the generality of the statement required, then,different scales would be to the purpose, but they wouldhave a known relationship to each other.These possibilities are still some way from realisation.

The purpose of this paper has been to give an airing tothe theoretical ideas that indicate this line of development.The need is for better explanations of disabledbehaviour before attempting more sophisticated evalua-tion of it.I thank Professor A. E. Bennett and Mr. K. G. Wrightfor the use of the two sets of data referred to; Mr. J.Cairns, whose tabulations of the York data I have drawnon; and Mr. B. Sauv6 for advice on computing.

Comment: A. J. CULYERReader in EconomicsUniversity of York

I believe Williams is the first to attempt a generaltheory of disability explaining, among other things, why itis that physical impairments may lead to the variouschanges in social behaviour that are here termed'disability'. I propose to focus on Williams's third



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


R. G. A. Williams

sentence. I shall not refer to the details of measurement,Guttman scales, etc., but I shall attempt to show that,suitably extended, his second model (the so-calledrational choice model) can be generalised to embrace theothers as special cases and to provide a framework fordiscussing with greater clarity many of the famousquestions concerning disability, handicap, and impair-ment, the relationships between them, and theimplications, if any, for social policy and communitymedicine.

This more general model I term the 'choice' model inorder both to emphasise the discretionary nature of thelink between impairment and disability and, by dropping'rational', to show that the basic ideas in no way dependupon special theories about what rational choice actuallyis.We begin by assuming that the human activities of

the sort that figure in Guttman scales are those of ultimateinterest in disability, whether from a broad policy pointof view or at the level of clinical or social work. WithWilliams, let us continue to assume that, in principle,these activities can (a) be measured as a daily, weekly,etc., rate on a ratio scale, and (b) be associated with par-ticular intensities with which various limbs and otherphysical bodily functions are used. For example, runningis, among other things, lung and leg intensive comparedwith reading, which is eye intensive. Given an individual'sbodily functioning, the time he chooses for each activity(this may not be independent of the former), his talents,the value he places on various activities, the socialpressures he faces from relatives, professionals, etc., inany period of time the individual will actually engage ina measurable amount of each activity.

In Fig. 3 I suppose that, in view of these variousfactors, the individual could either perform Oa ofactivity A or Ob of activity B. Assuming that we aredealing with two activities only, for convenience of

C*0

> w C@Ect

C'-

Ef

%-.C

U 00 0

a

Fig. 3 Choice model of disabilitv.

exposition, then the feasible set of combinations of A andB which he might perform is determined by the area Oab.I have assumed ab to be a straight line although it mightbe convex from above-a relatively unimportant dif-ference. It is unlikely to be concave, although if it were,some of the conclusions of 'indifference curve' theorymay not follow. Let us suppose that individuals preferactivity to non-activity. In that case, an individual willbe located at some point on the boundary ab of thefeasible set: precisely where is of little interest at thislevel of generality.

Let us now introduce an impairment of physicalfunctioning, or an additional impairment to one thatalready existed when the individual was on ab. It maybe the loss of a limb, or a lung. The maximum amount ofB that can possibly be performed is now Od, and themaximum possible amount of A is Oc. The newfeasible boundary is cd upon which the individual mustnow settle. Two features of cd are important in comparingit with ab: (1) cd lies entirely within ab, hence less of atleast one activity must be chosen and (2) cd has asteeper slope than ab on any ray from the origin. Hencethere will be a tendency to substitute activities in whichthe individual is less impaired for those where he is moreimpaired.

This prediction does not depend upon rationality. Itis true that, if we assumed rationality of the sortpostulated by Williams (essentially concave indifferencecurves), and added the assumption that each activity is'normal' in the technical economic sense that neitheractivity is 'inferior', then less of activity B would beundertaken. But let us assume instead that individuals actirrationally-for example, that they act at random. Suchirrational individuals would be distributed evenly alongab. On the average, they would be located at e, halfwayalong ab. Similarly, impaired persons confronted withcd would likewise be located on the average at e', halfwayalong cd. Since e' lies to the left of e, highly irrationalpeople will behave (on the average) just the same asrational people.Another form of irrationality might be that people act

'conservatively', attempting to maintain the sameproportionate levels of different activities. This implieslocating at the intersection of the ray Oe and cd, at e'.Which of these various models best describes actual eventsis a matter of empirical test. My own hunch would bethat the rational choice hypothesis, as developed here,best applies to the young disabled, among whom onewould expect to see a high variance of disability givenbasic impairments-depending on tastes, determination,parental wealth, social expectations, etc. The 'conserva-tive irrational' model might be a useful starting point inmodelling the disability of elderly persons.

Rational 'tastes' are, however, well worth bringingin because they greatly enrich the insights of thismodel of disability. Suppose, for example, that socialpressures, or personal attitudes, shift 'tastes' systema-tically towards less, or more, of the activity that isintensive in the impaired function. The interestingpossibility arises that the individual might move in theregion cc' on the new frontier, involving more readingbut less walking than before; or, in the other case, in

P g d bActivity B (lower limb intensive activity pertimeperiod. For example.walking)

46



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from



the region d'd, involving more walking but less readingthan before. The model includes the logical possibilitythat a person may engage in more activities in which he isimpaired than he formerly did. This possibility is flatlydenied by the mechanistic theory. So much the worsefor that theory, since the world actually does containat least one professional but one-legged footballer. Thepossibility clearly depends, however, on Od > Og or Oc >Of.

This more general theory includes the mechanisticand deviance theories in the following ways. Themechanistic theory forms the basis for describing thechange in the feasible set, modified so that thepossibilities are no longer 'all or none' but 'more or less',depending on severity. The deviance theory comes in asone, but only one, of the influences which determinewhere on the feasible frontier individuals will be located.Another natural extension of this model is to use it

as the basis for framing social policies designed to helpthe disabled. The following are among the questions thatneed more thought. Is policy more properly concernedwith shifts in the feasible set or with points chosenwithin a feasible set? In layman's language, shouldself-help be penalised by less public help, or lack of'self-help be rewarded by giving public help? Is itpossible to devise relevant weights to combine levels offunctioning and levels of ability to form overall indicesof disability and impairment? The model makes it clearthat these matters are not merely technical, or forprofessional ajudication only. Such weights are in ahighly relevant sense 'political'.

Reprints from Dr. R. G. A. Williams, Institute ofMedical Sociology, Westburn Road, AberdeenAB9 2ZE.

References

Bennett, A. E., Garrad, J., and Halil, T. (1970).Chronic disease and disability in the community;a prevalence study. British Medical Journal, 3, 762-764.

Freidson, E. (1972). Disability as social deviance. InMedical Men and their Work. Edited by E. Freidsonand J. Lorber. Aldine: Chicago.

Garrad, J., and Bennett, A. E. (1971). A validatedinterview schedule for use in population surveys ofchronic disease and disability. British Journal ofPreventive and Social Medicine, 25, 97-104.

Goffman, E. (1963). Stigma: notes on the managementof spoiled identity. Prentice Hall: Englewood Cliffs.

McConaghy, M. J. (1975). Maximum possible error inGuttman scales. Public Opinion Quarterly, 39, 343-357.

Nichols, P. J. R. (1975). Some psychological aspects ofrehabilitation and their implications in research.Proceedings of the Royal Society of Medicine, 68,537.

Rosser, R. M., and Watts, V. C. (1972). The measurementofhospital output. International JournalofEpidemiology1, 361-368.

Scott, M. B., and Lyman, S. W. (1968). Accounts.American Sociological Review, 33, 46-62.

Williams, R. G. A., Johnston, M., Willis, L. A., andBennett, A. E. (1976). Disability: a model and measure-ment technique. British Journal of Preventive andSocial Medicine, 30, 71-78.

Wright, K. G. (1974). Alternative measures of the outputof social programmes: the elderly. In EconomicPolicies and Social Goals. Edited by A. J. Culyer.Martin Robertson: London.

47



http://jech.bmj.com

/E

pidemiol C

omm


arch 1979. Dow

nloaded from


Theories and measurement in disability · Theories andmeasurementin disability activities of the...

Documents

Transcript of Theories and measurement in disability · Theories andmeasurementin disability activities of the...