assessment of preferences for self-treatment and ...assessment of preferences for self-treatment and...

Journal of Personality and Social Psychology1980, Vol. 39, No. 5, 977-990

Assessment of Preferences for Self-Treatment andInformation in Health Care

David S. Krantz, Andrew Baum, and Margaret v. WidemanUniformed Services University School of Medicine

It has been assumed that it is beneficial for patients to become active and in-formed participants in health care. Previous research, however, suggests thatindividuals differ in their receptiveness to information and self-care in treatmentsituations. This article reports the development and validation of the KrantzHealth Opinion Survey, a measure of preferences for different treatment ap-proaches. This measure yields a total score and two relatively independent sub-scales that measure, respectively, preferences for information and for behavioralinvolvement (i.e., self-care and active participation) in medical care. Three re-lated studies demonstrated the ability of the subscales or total score to predictwith some specificity (a) criterion group membership (clinic users and enrolleesin a self-care course), (b) reported use of clinic facilities, and (c) overt behavior(e.g., inquisitiveness, self-diagnosis) in a medical setting. Discriminant validityof the instrument is also established. Theoretical implications of the preferenceconstructs are described in terms of the concept of personal control, and prac-tical implications of the measure are presented.

In recent years, there have been a numberof calls for psychologists to become more in-volved in immediate problems of the practicalworld (e.g., Miller, 1969). It is particularlyappropriate that social psychology, with itsroots in action research (Lewin, 1948), applyits theories and methods to real-world prob-lems. Increasing numbers of social and per-sonality psychologists are therefore adoptinga more problem-oriented approach to theunderstanding of human behavior. An impor-tant example of this trend is the application

This research was supported in part by GrantHL 23674 from the National Institute of Health andClinical Grant C07206 from the Uniformed ServicesSchool of Medicine. We wish to thank LucilleLemanski, James Pennebaker, David Winer, Lorrainede Labry, Susan Geraci, and Betsy Kent for theirassistance with various phases of data collection. Wealso thank Sheldon Cohen, Karen Matthews, JamesPennebaker, and Judie Stein for their helpful com-ments on an earlier draft of this article.

Requests for reprints should be sent to David S.Krantz, Department of Medical Psychology, Uni-formed Services University of the Health Sciences,4301 Jones Bridge Road, Bethesda, Maryland 20014.

of social-psychological principles to a varietyof health care problems. The interest in thisarea has derived in part from a recognitionthat medical outcomes can be determined bythe nature of the interaction between doctorand patient, the patient's understanding ofillness, and his or her degree of participationin the health care process (cf. Korsch &Negrete, 1972; Krantz, 1980; Krantz &Schulz, 1980). At the same time, growingconsumerism and a movement away from thetraditional medical model have encouragedpatients to become more active and informedparticipants in the health care process. Al-though it is generally assumed that more in-formation and self-reliance are better (e.g.,Vickery & Fries, 1976), questions still remainas to how much patients should be told (e.g.,Mclntosh, 1974) and how much self-care andresponsibility are optimal (Linn & Lewis,1979).

Further complicating this situation is thelikelihood that some individuals may benefitmore than others from being highly informedor involved in their own treatment. Per-sonality-based expectancies and beliefs about

In the public domain

977

978 D. S. KRANTZ, A. BAUM, AND M. WIDEMAN

health and illness may determine the efficacyof patient-oriented approaches to health care.Although some research has addressed theseindividual differences (e.g., Wallston, Walls-ton, Kaplan, & Maides, 1976), most studieshave relied on clinical intuition or measuresof coping styles devised for other purposes.The studies reported in this article develop aspecific measure of individual differences inpreferred role in health care procedures.

Personal Control and Health Care

Several social-psychological concepts are po-tentially relevant to these problems. In stressresearch, the effects of providing subjectswith information and heightening participa-tion and choice have been examined in termsof the concept of personal control (cf. Averill,1973; Langer & Rodin, 1976). Heightenedparticipation and choice often lead to increasesin perceived control, since they may providesubjects with the belief (correct or not) that*hey can alter or affect outcomes. Since oneof the aims of self-care programs is to in-crease peoples' general feelings of self-relianceand control over their bodies, the psychologi-cal effects of self-care might be viewed interms of the concept of personal control. In-formation has also been conceptualized as aform of cognitive control, because it may in-crease the ability to prepare for aversiveevents and often also results in the interpre-tation of events so that threat is lessened(Averill, 1973; Seligman, 197S). The effectsof providing information in medical settingsmay also be viewed in this conceptual frame-workwork (cf. Johnson, 1975).

Although control over a stressor frequentlyresults in less negative effects on the organism,several studies suggest that there are notalways direct beneficial effects of providingcontrol (cf. Averill, 1973; Mills & Krantz,1979). Instead, the context and meaning of aparticular "control response" determineswhether it will be effective in reducing stress.This principle is also supported by individual-difference research (e.g., Houston, 1972) indi-cating that individuals perform better instressful situations where there is congruencebetween general expectations for control andthe locus of control in that particular situ-

ation. Because the relationship between per-sonal control and favorable outcomes is oftencomplex, research has been conducted in medi-cal settings to determine the effects of pro-viding for patients' control over some aspectsof treatment. (See Krantz, 1980, for fullerdiscussion of these issues.)

Accordingly, researchers have providedpatients with information that allows themto prepare for medical procedures (e.g.,Egbert, Battit, Welch, & Bartlett, 1964;Johnson & Leventhal, 1974) and have trainedpatients in coping strategies that facilitateprocesses involved in treatment and recovery(e.g., Kendall et al., 1979; Langer, Janis, &Wolfer, 1975). In these studies, providing in-formation has generally reduced emotionalreactivity and facilitated the recovery process,although these effects often involve complexmediating processes. Other research suggeststhat directly providing patients with actualor perceived control over outcomes is often(but not always) effective in facilitating treat-ment and recovery. Thus, there have beenstudies that made the patient a more activeparticipant in treatment (Cromwell, Butter-field, Brayfield, & Curry, 1977), that height-ened the patient's sense of choice (Mills &Krantz, 1979), or that provided for self-monitoring or self-care (e.g., Berg & LoGerfo,1979; Mahoney & Thoreson, 1974).

Differences in Receptiveness to Informationand Self-Care

There appear, however, to be substantialindividual differences in reactions to thesekinds of treatment interventions. Amongpatients given preparatory information, sev-eral studies (Andrew, 1970; Shipley, Butt,Horwitz, & Farbry, 1978) have found thatindividuals who tend to deny or repressanxiety-provoking stimuli respond more poorlythan those who tend to be sensitized tostressful stimuli. Other studies using locusof control as an individual difference variablehave found that internals (those who feel theycan control reinforcement) show better ad-justment when given specific information,whereas externals fare best with general in-formation (Auerbach, Kendall, Cuttler, &Levitt, 1976). Further, Cromwell et al. (1977)

HEALTH CARE SELF-TREATMENT AND INFORMATION 979

found that heart patients given treatmentscongruent with control beliefs showed thebest outcomes on several rehabilitation out-come measures, and several studies using ahealth-specific locus of control measure(Lewis, Morisky, & Flynn, 1978; Wallstonet al., 1976) found that subjects in treatmentconditions congruent with control beliefs ex-pressed more satisfaction and reported highercompliance with a medical regimen. In sum,these data provide some evidence to supportthe conclusion that individuals fare best whengiven treatments that are congruent with theirexpectations and with their usual copingstyles. It should be noted, however, that therehave been some studies in this area (e.g.,Cromwell et al., 1977; Marston, 1970) thathave not produced clear-cut or consistentresults.

Patient Attitudes

Personality variables are only one of a com-plex of factors that may determine the opti-mum approach for particular individuals.Health beliefs together with attitudes towardphysicians and health care might also makepeople more or less suitable for specific pro-grams and treatments. Studies of illness be-havior (cf. Becker & Maiman, 1975; Rosen-stock & Kirscht, 1979) demonstrate thatbeliefs about the efficacy of medical care in-fluence use and response to health services.It has also been shown that when patients'expectations are not met, the result may belowered satisfaction and decreased compliancewith medical advice (cf. Korsch & Negrete,1972). Despite the increasing tendency topromote self-care, provide patients with in-formation, and increase their responsibility forhealth maintenance, relatively little is knownabout the relationship between patient atti-tudes and the outcomes of these types ofapproaches.

Rationale

The importance of patient expectations forhealth care outcomes suggests the need for ameasure of individual attitudes toward dif-ferent treatment approaches. Previous researchand anecdotal evidence (e.g., Mclntosh, 1974;

Springarn, 1978) suggest that some patientswould prefer more or less active participation,some want more or less information, somemore or less direction from physicians, andso on. We reasoned that a valid measure ofpatient preferences would relate to treatmentoutcomes and therefore facilitate research onmedical behavior. Such a measure of thesepreferences was recently developed (Krantz,Note 1). Since the two most common typesof psychological interventions involve pro-viding information and encouraging active in-volvement or self-care, this measure was con-structed to yield scales for each of these com-ponents. The first study reported in thisarticle provides data on the development ofthis instrument. The studies that follow pro-vide validation for the scale, which was spe-cifically designed to measure attitudes towarddifferent treatment approaches.

Scale Development

MethodPilot Work

An item pool of 40 statements about aspects ofmedical care was written to encompass the domainof preferences for an active and informed versus arelatively inactive and trusting role in the healthcare process. A number of these items were adaptedfrom a questionnaire developed to measure physicianattitudes toward self-care (Linn & Lewis, 1979). Theset of face-valid items tapped (a) beliefs in theefficacy and benefits of self-care, (b) frequency ofinformation-seeking and questioning of physiciansand nurses, (c) beliefs about the benefits or dis-advantages of making one's own medical decisions,(d) attitudes toward use of a physician versus one-self as a health-care provider, (e) frequency of self-diagnosis, and so on. A binary agree-disagree formatwas used, and items were keyed so that high scoresrepresented favorable attitudes toward self-directedcare.

The 40-item test was administered to 200 under-graduates, and a preliminary item analysis was con-ducted. Items were eleminated if they correlated lessthan .20 with the total score or showed a narrowdistribution of response alternatives. From theoriginal pool, 26 items remained. This second versionof the test was then administered to 159 under-graduates. Factor analysis (principal componentsusing varimax rotation) was used to obtain a roughidea of the components of the scale and to derivetwo meaningful subscales as planned. The first scale,called Behavioral Involvement (B), consists of nineitems concerned with attitudes toward self-treat-ment and active behavioral involvement of patients


Table 1Items and Statistics for the Krantz Health Opinion Survey

Item

Information subscale

I usually don't ask the doctor or nurse many questions about what they'redoing during a medical exam. (l)a

I'd rather have doctors and nurses make the decisions about what's best thanfor them to give me a whole lot of choices. (3)"

Instead of waiting for them to tell me, I usually ask the doctor or nurseimmediately after an exam about my health. (4)

I usually ask the doctor or nurse lots of questions about the procedures duringa medical exam. (8)

It is better to trust the doctor or nurse in charge of a medical procedure thanto question what they are doing. (10)"

I usually wait for the doctor or nurse to tell me the results of a medical examrather than asking them immediately. (15)"

I'd rather be given many choices about what's best for my health than to havethe doctor make the decisions for me. (16)

Behavioral Involvement subscale

Except for serious illness, it's generally better to take care of your own healththan to seek professional help. (2)

It is better to rely on the judgements of doctors (who are the experts) than torely on "common sense" in taking care of your own body. (5)"

Clinics and hospitals are good places to go for help since it's best for medicalexperts to take responsibility for health care. (6)"

Learning how to cure some of your own illness without contacting a physicianis a good idea. (7)

It's almost always better to seek professional help than to try to treatyourself. (9)°

Learning how to cure some of your illness without contacting a physicianmay create more harm than good. (11)'

Recovery is usually quicker under the care of a doctor or nurse than whenpatients take care of themselves. (12)"

If it costs the same, I'd rather have a doctor or nurse give me treatments thanto do the same treatments myself. (13)a

It is better to rely less on physicians and more on your own common sensewhen it comes to caring for your body. (14)

Corrected itemcorrelations

Total Assignedscore subscale

.46 .59

.36 .36

.30 .48

.47 .59

.35 .36

.35 .58

.35 .37

.16 .24

.44 .49

.36 .36

.23 .32

.42 .50

.38 .45

.42 .49

.45 .43

.39 .36

Note. Correlations are point-biserial item-remainder correlations based on 200 cases. These correlations areall positive because scoring is reversed for negatively worded items, which are denoted by superscript a's.Numbers in parentheses indicate the sequence of items on the scale. Verbatim instructions may be obtainedfrom the first author.a Negatively worded item.

in medical care. The second scale, called Information(I), is denned by seven items measuring the desireto ask questions and wanting to be informed aboutmedical decisions. Remaining items not correlatingwith either of these scales were eliminated. Thefinal version of the test, called the Krantz HealthOpinion Survey (HOS), therefore consists of 16items, each rated in a binary, agree-disagree format.The test yields scores for two a priori subscales anda total score measuring composite attitudes towardtreatment approaches. High scores represent favor-

able attitudes toward self-directed or informed treat-ment. It should be noted that these items generallyrefer to routine aspects of medical care and do notrefer to severe or traumatic illness. In addition,future refinement of the scale may require elimina-tion of some redundancy in the wording of theitems.

To determine discriminant validity, the test wasthen administered to 100 male and 100 femaleundergraduates at the University of Southern Cali-fornia together with the Crowne-Marlowe Social


Desirability Scale (Crowne & Marlowe, 1964) andthe Health Locus of Control Scale (Wallston et al.,1976). An independent sample of 38 undergraduatescompleted the HOS along with the Minnesota Multi-phasic Personality Inventory (MMPI) Hypochon-driasis scale, and a third sample (« = 87) was giventhe HOS together with the Ullman (1962) versionof the Repression-Sensitization (R-S) scale. An-other sample of 80 students were given the HOStwice over a 7-week period to determine test-retestreliability.

Results and Discussion

Table 1 presents the items grouped by sub-scales. The total HOS scale has a Kuder-Richardson 20 reliability of .77. Reliabilitiesof the Behavioral Involvement (B) and In-formation (I) subscales were .74 and .76,respectively. Kuder-Richardson 20 reliabilityof the HOS for two subsequent college sam-ples remained over .74 for subscales and thetotal scale. Test-retest reliabilities for theHOS components were .74, .71, and .59 forthe total score, Behavioral Involvement scale,and Information scales, respectively, over a7-week period. There was a slight but non-significant tendency for females to score some-what higher than males on all HOS scales.

Correlations With Other Scales

Correlations among the HOS total score, itssubscales, and four relevant individual dif-ference measures are presented in Table 2.The two subscales of the HOS correlated onlyslightly with one another, sharing less than9% of the variance. The HOS showed onlya moderate correlation (.31) with the Wall-ston Health Locus of Control (HLC) Scale,a more established questionnaire that wasdesigned to measure expectancies about theability to control one's health.1 Associationswith the HLC were even lower for individualsubscales, indicating that the HOS and HLCare probably measuring relatively independentprocesses. (Further discriminant validity willbe demonstrated in later studies in this arti-cle.) The HOS also is very modestly corre-lated with repression-sensitization. Finally,the HOS and subscales show low or near-zerocorrelations with social desirability and hypo-chondriasis.

Validation StudiesOverview

The remainder of the research presented inthis article is concerned with establishing thepredictive, construct, and discriminant validi-ties of the Krantz Health Opinion Survey. Toaccomplish this, the scale was administered tothree samples: unselected residents of a col-lege dormitory, students reporting to a collegeinfirmary for routine treatment of minor ill-nesses, and students enrolled in a medicalself-help course at the same school. For clarityof presentation, data are organized in theform of three studies. The first study reportsscores of criterion groups, the second dealswith reported use of infirmary services forminor illnesses, and the third examines overtbehavior while undergoing treatment at theclinic.

Study 1: Scores of Criterion Groups

Method

One means of establishing the validity of an in-strument is by predicting how predetermined groupsof individuals will score on the instrument. Ac-cordingly, the Health Opinion Survey was adminis-tered to criterion groups chosen to represent ex-tremes in preferences for different approaches. Thesescores were then compared to other unselected sam-ples from the same population.

The HOS was administered to 149 students atTrinity College in Hartford, Connecticut. Thissample consisted of 56 freshmen dormitory residents,12 students enrolled in a medical self-help courseoffered by the physical education department, and81 students reporting on weeknights to a collegemedical office for routine health care. The self-helpcourse covered topics such as basic first aid, cardio-pulmonary resuscitation, and so on. We reasonedthat a criterion group of students who voluntarilyenrolled for such a course would be likely to havegreater interest in self-care and in obtaining infor-mation about medical settings. It was thereforepredicted that the self-help course en-rollees wouldscore higher on the behavioral involvement, infor-mation, and total scores of the HOS.

Research on illness behavior (e.g., Mechanic,1968) suggests that use of medical facilities involvesthe perception and evaluation of symptoms as well

1 In a second sample of 83 subjects (noted byparentheses in Table 2), correlations between theHealth Locus of Control scale and all HOS scaleswere even lower.


Table 2Correlations Among Krantz Health OpinionSurvey (HOS) Scales and Four IndividualDifference Measures

Scale

BehavioralInvolvement (B)

Information (I)Crowne-Marlowe

Social DesirabilityWallston Health Locus

of ControlMM PI Hypo-

chondriasisRepression-Sensi-

tization (R-S)

TotalHOS

.82**

.77**

-.03.31**

(.11)

-.13

-.19

B

—.26

.08

.26**(.16)

-.17

-.05

I

——

-.14*.23**

(.00)

-.02

-.24*

Note. Correlations are based on 200 cases exceptthose involving the Hypochondriasis (n = 38) andR-S (n = 87) scales. Values in parentheses are basedon a second sample of 83 cases. Significance leveldifferences for coefficients of similar magnituderesult from differing sample sizes.*p<.QS. **p<. 01.

as the decision to act on them. Therefore, indi-viduals who seek help represent only a subset ofthose who may be ill. The behavioral involvementscale presumably measures preferences for self-care,and low self-care individuals should perhaps bemore inclined to seek professional help for rela-tively minor illness. Conversely, high self-care indi-viduals should be more likely to first treat them-selves when they perceive minor illness. The sampleof college infirmary users might therefore serve as asecond criterion group. The medical office is free tostudents and is staffed in the evenings by a nurse.Virtually all the users of this clinic report withminor illnesses such as colds, headaches, flu, and soon. It was expected that clinic users would tend toscore lower than unselected dorm residents on theBehavioral Involvement scale (but not necessarilythe other scales) of the HOS.

Results 2

One-way analyses of variance followed byDunnett's t tests were conducted to comparescores of the self-help and clinic samples withthose of the dorm residents (see Table 3). Asexpected, the self-help group scored higherthan the dorm group on preference for be-havioral involvement, t( 146) = 2.36, p <.05; total HOS score, t(146) = 2.69, p < .05;and was marginally different on information,<(146) - 1.76, p < .10. The clinic users werein turn lower than dorm residents on the B

scale, £(146) = 1.98, p < .05, but did notdiffer on either the I scale or total score. Aspredicted, the HOS successfully discriminatedbetween a criterion group of high self-caresubjects and the general student population,and low B scores were associated with use ofclinic facilities.

Study 2: Reported use of Clinic Facilitiesand Discriminant Validity

Method

Comparison of dormitory residents with clinicusers indicated that attitudes toward involvement inhealth care affect the use of health care facilities.Consistent with this finding, one might expect indi-viduals who do not prefer self-care (those with lowB scale scores) to be more likely to use clinicfacilities when they feel ill. Conversely, those withfavorable attitudes toward self-involvement shouldbe (at least initially) inclined to treat themselveswhen symptoms appear and therefore less likely touse the clinic.

To further investigate this aspect of the constructvalidity of the behavioral involvement scale, 54students from the dormitory resident sample com-pleting the HOS in Study 1 were asked to retro-spectively report the number of times they hadvisited the college medical office during that aca-demic year. To provide more evidence of discrimi-nant validity of the HOS, the Wallston HealthLocus of Control i(HLC) scale (Wallston et al.,1976) was also administered to 38 students fromthis group. The HLC scale might also be expectedto relate to use of clinic services, although compet-ing predictions can be made about the direction ofrelationship. On the one hand, HLC internals (thosewho feel they are able to control their health)might be more self-reliant and less likely to makeuse of the clinic. On the other hand, if clinic use isseen as a means of exercising control over health,HLC internals may be more likely to use theinfirmary.


The number of reported clinic visits duringthe year ranged from 0 to 20 for the 54 sub-jects. As predicted, there was a significantnegative correlation between reported visitsand Behavioral Involvement scores, r =— .35, p < .01. To determine if high B scoreswere related to increased clinic use in a linear

2 One-tailed probability levels were adopted forthis study, since directional predictions were specifi-cally made.


Table 3Normative Health Opinion Survey (HOS) Scores for Several Groups

Scale

Behavioralinvolvement Information Total HOS

Sample M SD M SD M SD

College dormitory residentsUsers of college infirmaryCollege students in medical

self-help class

5681

12

3.793.00

S.SO

2.282.25

2.58

4.054.31

5.25

2.202.13

1.82

7.847.31

10.75

3.253.45

3.79

fashion,3 the scale was broken into three ap-proximately equal groups based on the distri-bution of B scores in this sample (see Figure1). Trend analysis revealed a significantlinear association between extremity of Bscores and reported visits to the clinic, withclinic visits decreasing as B scores increased,F(l, 51) = 7.70, p< .01.

Scores on the I scale were not related tonumber of visits, r = .06, p < .66. This nullfinding indicates a degree of discriminativevalidity between the B and I scales. However,total scale scores were, like the B scale, re-lated to clinic visits, r = —.28, p < .04, andtrichotomizing the sample based on total HOSscores revealed that this was also a linearrelationship, F(l , 51) = 5.32, p < .03.

Scores on the HLC * were also negativelycorrelated with clinic visits, r = —.32, p <.05, but a trend analysis dividing the sampleinto high, medium, and low HLC groups re-vealed that this relationship was not signifi-cant, F(l, 35) = 1.81, p< .19. Examinationof Figure 1 reveals that although none of thethree HLC groups differed significantly fromone another, only the highly internal subjectsreported less use of health services than theothers.5

In sum, the B scale and total score on theHOS show a linear negative relationship toreported use of clinic facilities. This result,in conjunction with the B score differencebetween clinic and dorm samples in Study 1,supports a link between preferences for be-havioral involvement (or self-care) and useof clinic facilities. The B and I scales seemto show some discriminant validity, and theeffects of the B scale and Health Locus of

Control scale appear to be largely independentof one another.

Study 3: Treatment Preferences andBehavior in a Clinic

Preferences for information or self-treat-ment in medical settings should reflect them-selves in overt behavior exhibited while indi-viduals are undergoing treatment. Patientswho indicate higher preferences for informa-tion should, in fact, ask more questions insuch a setting or seek out information whengiven an opportunity to do so. Similarly, indi-viduals who prefer behavioral involvement(as opposed to passivity) when undergoingtreatment should play a more active role inthe treatment process. This may be reflected

3 Since the scale will probably be used as a tech-nique for classifying subjects based on extremity ofscores, we present these analyses based on discretegroups even though the scale is a continuousmeasure.

4 In this article, the HLC was scored in the in-ternal direction. High scores therefore indicate per-ceived ability to control one's health.

5 Even though the HLC and HOS share less than10% of variance, the possibility arises that the Bscale is related to reports of clinic visits by virtueof its association with HLC. Therefore, partial cor-relation coefficients were calculated between therespective scales and number of visits to the in-firmary, partialing out the effect of the other scale.These analyses, based on 38 available cases, re-vealed that both behavioral involvement (r = —.2l)and HLC (r = —.30) were correlated with visits in'a relatively independent manner. Although the cor-relations were not reduced substantially in magni-tude, they did not achieve reliable levels of signifi-cance due to smaller sample size.


| | Low Scores

[^j Medium Scores

• High Scores

Scale Behavioral Involvement InformationScores <2 3.4 5-9 <2 3-5 6-7

N 16 18 20 15 20 19

Total HOS

<6 7-9 10-16

17 20 17

Health Locus of Control

< 39 40-47 48-66

12 14 12

Figure 1. Relationship of reported clinic visits to high, medium, and low scores on Krantz HealthOpinion Survey component and total scales and Health Locus of Control scale.

in attempts to self-diagnose or become in-volved in treatment when given the oppor-tunity to do so.

These aspects of the construct validity ofthe Health Opinion Survey and subscales wereinvestigated in a study of users of studentmedical facilities. A clinic nurse rated avariety of patient behaviors exhibited duringa medical exam. These behaviors related toinformation seeking (e.g., asking questions)and active involvement in treatment (e.g.,self-diagnosis). Latency between appearanceof symptoms and seeking medical help wasalso recorded, and some patients who neededmedications were offered the option of choos-ing these medications if they desired. It waspredicted that the B and I scale scores wouldshow specificity in relating to the clinic mea-sures. That is, I scores should relate only toinformation-seeking behaviors and B scoresto measures of active involvement. Total HOSscores should be predictive of behaviors thatencompass a combination of both dimensions,such as asking for specific medications.

MethodSubjects

Subjects were the 81 undergraduates (45 males,36 females) at Trinity College who visited the col-

lege medical office with minor complaints such asheadache, colds, flu, and so on. Because of pro-cedural problems and the ongoing nature of thestudy, all subjects did not receive all dependentmeasures.

Procedure

As patients entered the waiting room of themedical office, they were approached by a femaleexperimenter and asked to participate in a study of"attitudes toward health care." Virtually all pa-tients agreed to participate. After informed consentwas obtained and anonymity assured, subjects wereescorted into the examination room, where a nurse(who was blind to the purpose of the experimentand subjects' HOS scores) conducted a medicalexamination or gave treatment as required. Duringthe course of the exam, the nurse closely observedand questioned the patients' behavior so that a post-exam questionnaire could be completed. The nursedetermined how long the patient had waited be-tween the appearance of the symptoms and report-ing to the clinic, whether the patient had made anattempt to self-diagnose, and the number of ques-tions asked by the patient. The nurse conducted theexam in a way that would not inhibit any act ofinquisitiveness or behavioral involvement of patients.For a subset of the sample (» = 29), if medicationwas required the nurse explicitly offered the patienta choice: He or she could either choose for him orherself from among several available medications orhave the nurse choose one.

When the exam was concluded, the patient wasescorted into the waiting room where the Health


Opinion Survey and a postexamination questionnairewere completed. A subset of patients also completedthe Health Locus of Control scale. During thistime the nurse completed the questionnaire assessingher perceptions of the patient's behavior.

Dependent Measures

Nurse ratings. For 62 of the subjects, the nurserecorded the number of questions asked by patientsduring the exam, whether patients had attempted toself-diagnose, and if the patients had requested aspecific medication. Fo.r the remainder of the sub-jects, these measures were not recorded. Latency(in days) to report to clinic after appearance ofsymptoms was coded on a 9-point scale, where0 = same day and 8 = greater than 1 week. Forthose patients given the option to choose medica-tions, the subject's decision (nurse or self) wasrecorded. In addition, the nurse was asked to evalu-ate the patient on a number of dimensions (discom-

fort during the exam, receptiveness to treatment, eyecontact, degree of detail in describing symptoms).

Patient questionnaire. Following the exam, pa-tients were asked questions regarding such things astheir satisfaction with treatment and quality ofcare, degree of discomfort, and familiarity with theinfirm a;ry.


Number of Questions Asked

Nurse ratings of the number of questionsasked by the patient during the exam revealedthat higher I scores were associated withgreater inquisitiveness, n = 62, r = .28, p <.03. The B scale was not reliably related toquestions asked, r - .16, p < .22, but totalscores on the HOS were related to inquisitive-

LOW scoresMedium ScoresHigh Scores

ScaleN 30 19 12 23 18 20

ScaleN

B311516

I123020

Total HOS26 20 16

HLC7 19 5

* SO

§ 40

ScaleN

B I42 24 14 29 24 27

Tola! HOS44 16 20

HLC12228

Figure 2. Relationship of clinic variables from Study 3 to high, medium, and low scores on KrantzHealth Opinion Survey (HOS) Behavioral Involvement (B), Information (I), and total scales,and Health Locus of Control (HLC) scale.


ness, r = .30, p < .02. To further examinethe nature of these relationships, scores oneach scale (I, B, and total) were divided intothirds based on the population distribution ofthe unselected dormitory sample reported inStudies 1 and 2.e Trend analyses revealed apositive linear relationship between numberof questions and I scores, F(l, 59) = 9.56,p < .01 (see Figure 2). The high, medium,and low information subjects asked an aver-age of 5.3, 3.4, and 1.3 questions, respectively.A similar linear relationship was obtained fortotal scores, F(l, 59) = 8.41, p< .01, butthere was no reliable relationship between ex-tremity of B scores and questions asked, F =2.02, p < .15.

Self-Diagnosis

It was expected that the Behavioral In-volvement, but not the Information, subscalewould be related to whether the patient hadattempted to self-diagnose. Chi-square analy-ses were used to determine the relationshipbetween self-diagnosis (coded yes/no) andextremity of scores on the respective scales.Figure 2 presents these data. Once again, eachscale was trichotomized based on the dormi-tory sample norms. A 3 X 2 chi-square analy-sis revealed a significant association betweenB scores and self-diagnosis, x 2 ( 2 ) = 6.91,p < .04. It can be seen that the higher thereported preferences for behavioral involve-ment, the greater the likelihood of subjectsspontaneously offering diagnoses. Similar chi-square analyses revealed that total scoreswere also positively related to self-diagnosis,p < .01. Unexpectedly, the information scalewas related to self-diagnosis, p < .06; how-ever, this relationship was not a positive linearone (see Figure 2).

Requesting Specific Medications

The nurse recorded whether patients re-quested specific medications during the exam.Three by two chi-square tests revealed thatrequests were reliably related only to totalscores, *2(2) = 6.62, p < .04 (see Figure 2).Neither behavioral involvement nor informa-tion scores related to medication requests inthis analysis. A case could be made that ask-

ing for medications should relate to eitherinformation or behavioral involvement. Inretrospect, however, requesting medicationsinvolves elements of both inquisitiveness andbehavioral involvement. Therefore it is notsurprising that it is most strongly related tothe total score, the aggregate of these ele-ments.

Electing to Choose Medications

Twenty-nine patients were explicitly giventhe opportunity to either select their ownmedication or have the nurse choose for them.The smaller number of subjects with avail-able data necessitated that HOS scores bedichotomized (based on the dorm median)rather than broken into thirds. Chi-squareanalysis (Nurse/Self-Choice X High/LowHOS Scores) revealed that only 1% of sub-jects with lower B scores (0, 1, and 2) chosetheir own medications, whereas 50% of highBehavioral Involvement subjects (scores >3) elected to choose, X

2 ( l ) = 4.81, p < .03.Total scores, p < .03, but not informationscores, p < .40, were also related to choice.

Reported Latency to Come to Clinic

Neither B scores nor total scores were sig-nificantly correlated with latency to report toclinic, n — 80, rs = .10 and —.10, respec-tively. Unexpectedly, I scores were negativelycorrelated with latency, r = —.28, p < .02.Closer examination of the data revealed thatthere was a curvilinear relationship betweenextremity of Behavioral Involvement scoresand latency to report to the clinic (quadratictrend, p < .06). Subjects in the lowest thirdof the distribution reported to the clinic mostquickly after symptoms appeared (M = 1.7days), those with moderate B scores waitedthe longest (M = 3.3 days), and those withhigh B scores fell between those extremes(M = 2.3 days). These data seem to suggest

0 The dormitory student sample was used us thepopulation to determine criteria for low, medium,and high scores, since there was some evidence(.reported earlier) that clinic users may be a selectedgroup. Figure 1 presents these cutpoint scores. Useof clinic group to determine normative cutpointswould not substantially change these results.


that very strong preferences for behavioralinvolvement do not reflect themselves in theamount of time taken to seek medical help.However, in view of the relationships betweenB scores and clinic use obtained in the pre-vious studies, it seems plausible to consideranother explanation for this effect. After re-porting to a medical setting, high self-careindividuals may have distorted reports ofdelay out of reluctance to admit that they hadwaited so long. Neither I nor total scoresshowed a linear or curvilinear relationship tolatency.

Wallston HLC Scale

The HLC scale correlated -.28 with num-ber of questions asked, n = 31, p < .13, in-dicating a trend for health internals to askfewer questions than health externals. Thisrelationship is opposite to that obtained forthe HOS Information subscale. The negativecorrelation between HLC and inquisitivenessis not consistent with one previous finding(Wallston et al., 1976). However, examina-tion of Figure 2 reveals that the fewest ques-tions were asked by individuals with moderateHLC scores.

Chi-square tests indicated that HealthLocus of Control scores were not reliablyrelated to self-diagnosis (p < .55) or to re-questing medications (p<AT). Thus, thereis evidence of substantial discriminant validitybetween HLC and HOS. However, HLCshowed a curvilinear relationship to latencyto report to the clinic (quadratic trend, p <.06), although the correlation between HLCand latency was not significant, n - 42, r =.15, p < .34. This finding resembles resultsobtained for the B scale.

Patient Sell-Ratings

Following the visit with the nurse, patientswere asked to rate their degree of satisfactionwith treatment, their discomfort during theexam, and their degree of familiarity with theinfirmary. The only reliable relationship wasa positive correlation between rated satisfac-tion and behavioral involvement, »= 62,r — .27, p < .04. This correlation was notsurprising, since the nurse's behavior during

the exam was designed to facilitate the be-havioral involvement of patients.

General Discussion

The present studies demonstrate that atti-tudes toward treatment approaches can bemeasured reliably and that these preferencesinfluence a variety of health and illness be-haviors in a fairly straightforward way. Twoseparable components of these attitudes weredescribed: behavioral involvement, concernedwith attitudes toward self-care and an activerole in medical care, and information, con-cerned with desire to ask questions and to beinformed of and involved in medical decisions.These two components, which are relativelyindependent of one another, display a degreeof specificity in their ability to predict be-havior. In addition, the Krantz HealthOpinion Survey does not appear to overlapconsiderably with previously existing scalesand measures.

From a conceptual point of view, the Be-havioral Involvement and Information sub-scales of the HOS bear some resemblance tocategories of personal control described byAverill (1973) in a review of stress research.He distinguished three types of control:behavioral, encompassing direct action on theenvironment; cognitive, the interpretation ofevents, including information; and decisional,having a choice among alternative courses ofaction. Since the two HOS subscales appear tobe relatively independent of one another, thissuggests that at least in terms of health care,the preferred types of personal control do notcovary to a high degree. In addition, accord-ing to Averill (1973), the relationship of per-sonal control to stress is a function of themeaning of the control response to the indi-vidual; the meaning and effectiveness of aparticular intervention are often determinedby the situational context. Several studiesconducted in health care settings (e.g., Crom-well et al., 1977; Mills & Krantz, 1979) sup-port the view that the effects of interventionsinvolving information and behavioral involve-ment also depend on the way they are pre-sented and whether they enable individualsto satisfy their needs in that setting (cf.Krantz, 1980).


Preferences for or against behavioral in-volvement and information may therefore bean index of how the individual interpretsthose approaches that encourage patient in-volvement, self-care, and informed participa-tion. The existence of these individual differ-ences also suggests that "the more controlthe better" may not be an effective principlefor everyone in medical care. We are presentlyconducting research to determine if medicaloutcomes are most favorable when patientpreferences are matched to particular treat-ment approaches. In addition, Linn and Lewis(1979) have developed a measure of physi-cians' attitudes toward self-care. It may alsobe productive to match patient preferences tophysicians with corresponding attitudes.

It is worth noting that for several variables(e.g., self-diagnosis, reported clinic visits),the HLC and Behavioral Involvement scalesseem to predict about equally well for indi-viduals scoring in the upper third of the dis-tribution on each scale. However, of these twomeasures, only the Behavioral Involvementscale is able to predict criterion behaviorsacross the entire range of scores (see Figures1 and 2), and is not confined in usefulness tothe upper ranges of the distribution. More-over, the HOS also yields a second informa-tion scale that allows measurement of in-quisitiveness in a medical setting. It shouldbe noted that the HOS total score is probablyonly applicable to behaviors representing acomposite of behavioral involvement and in-formation seeking.

The constructs identified by the KrantzHealth Opinion Survey may have several ad-ditional implications for both researchers andpractitioners. At present the value of thecomponent scales has been demonstrated forpredicting behaviors relating to routine medi-cal care for relatively short-term, minor ill-ness. It still remains for future research todetermine whether the Health Opinion Surveyis applicable to instances of chronic or seriousillness. The present studies also highlight theimportance of beliefs about health care indetermining use of clinic facilities and otherhealth-related behaviors. Researchers con-cerned with illness behavior and symptom per-ception (e.g., Mechanic, 1968; Rosenstock &Kirscht, 1979) have sought to identify factors

that lead individuals to seek or avoid medicalhelp. The present studies suggest that indi-viduals who prefer an active role in healthcare may be less likely to seek out the helpof a physician for minor illness.

At this time, we cannot specify the exactantecedents of preferences for particulartreatment approaches. Most probably a com-plex of processes including cultural and demo-graphic factors, education, and past experi-ence with illness and with medical profes-sionals all exert a degree of influence inshaping the beliefs measured by the HOS.Personality-based expectancies may play arole, although data from the present studiessuggest little overlap between a variety ofrelevant personality measures and treatmentpreferences. To the extent that preferencesare malleable, it may be possible to alterpatient beliefs to foster increased self-relianceand decreased dependence on medical pro-fessionals. In addition, if there are strongassociations between demographic characteris-tics and health-care preferences, this may haveimplications for developing modes of treat-ment for particular population groups. Finally,we do not as yet know the relationships be-tween preferences and more general healthbeliefs (cf. Kirscht & Rosenstock, 1979;Rosenstock & Kirscht, 1979), which havebeen shown to influence use of health servicesand compliance with medical recommenda-tions.

As presented here, the Krantz HealthOpinion Survey is an experimental scale thatrequires further testing before its full validityand predictive potential can be assessed.Further research may suggest that a modifiedversion is a useful measure for matchingpatients with particular practitioners or treat-ments, or for determining suitable candidatesfor types of health-promotion activities.Nevertheless, many unanswered questions re-main regarding patients' receptiveness to pro-grams that promote informed participationand self-care. Hopefully, these questions, to-gether with the validated instrument pre-sented in this article, will provide a stimulusfor future research. The present studies alsosupplement a growing body of research deal-ing with the interface between social psy-chology and medicine, and demonstrate the


potential utility of a problem-oriented appli-cation of social-psychological principles.

Reference Note

1. Krantz, D. S. The Krantz Health Opinion Survey:A Scale for the prediction of health care prefer-ences. Unpublished manuscript, University ofSouthern California, 1978.

References

Andrew, J. M. Recovery from surgery, with andwithout preparatory instruction, for three copingstyles. Journal of Personality and Social Psychol-ogy, 1970, IS, 223-226.

Auerbach, S. M., Kendall, P. C., Cuttler, H. F., &Levitt, N. R. Anxiety, locus of control, type ofpreparatory information, and adjustment to dentalsurgery. Journal of Consulting and Clinical Psy-chology, 1976, 44, 809-818.

Averill, J. R. Personal control over aversive stimuliand its relationship to stress. Psychological Bulle-tin, 1973, 80, 286-303.

Becker, M. H., & Maiman, L. A. Sociobehavioraldeterminants of compliance with health and medi-cal care recommendations. Medical Care, 1975, S,10-24.

Berg, A. O., & LoGerfo, J. P. Potential effect ofself-care algorithms on the number of physicianvisits. New England Journal of Medicine, 1979,300, 535-537.

Cromwell, R. L., Butterfield, E. C., B,rayfleld, F. M.,& Curry, J. J. Acute myocardial infarction: Re-action and recovery. St. Louis, Mo.: Mosby, 1977.

Crowne, D. P., & Marlowe, D. The approval motive.New York: Wiley, 1964.

Egbert, L. D., Battit, G. E., Welch, L. E., & Bart-lett, M. K. Reduction of post-operative pain byencouragement and instruction of patients. NewEngland Journal of Medicine, 1964, 270, 825-827.

Houston, B. K. Control over stress, locus of control,and response to stress. Journal of Personality andSocial Psychology, 1972, 21, 249-255.

Johnson, J. E. Stress reduction through sensationinformation. In I. C. Sarason & C. D. Spielberger(Eds.), Stress and anxiety (Vol. 2). New York:Wiley, 1975.

Johnson, J. E., & Leventhal, H. Effects of accurateexpectations and behavioral instructions on re-actions during a noxious medical examination.Journal of Personality and Social Psychology,1974, 29, 710-718.

Kendall, P. C., fit al. Cognitive-behavioral andpatient education interventions in cardiac catheteri-zation procedures: The Palo Alto medical psy-chology project. Journal of Consulting and ClinicalPsychology, 1979, 47, 49-58.

Kirscht, J. P., & Rosenstock, I. M. Patients' prob-lems in following recommendations of health ex-perts. In G. C. Stone, F. Cohen, & N. E. Adler

(Eds.), Health psychology: A handbook. SanFrancisco: Jossey-Bass, 1979.

Korsch, B., & Negrete, V. Doctor-patient communi-cation. Scientific American, 1972, 227(2), 66-78.

Krantz, D. S. Cognitive processes and recovery fromheart attack: A review and theoretical analysis.Journal of Human Stress, 1980, (J(3), 27-38.

Krantz, D. S., & Schulz, R. A model of life crisis,control, and health outcomes: Cardiac rehabilita-tion and relocation of the elderly. In A. Baum &J. E. Singer (Eds.), Advances in environmentalpsychology (Vol. 2), Hillsdale, N.J.: Eilbaum,1980.

Langer, E. J., Janis, I., & Wolfer, J. Reduction ofpsychological stress in surgical patients. Journal ofExperimental Social Psychology, 1975, 11, 155-165.

Langer, E., & Rodin, J. The effects of choice andenhanced personal responsibility for the aged: Afield experiment in an institutional setting. Journalof Personality and Social Psychology, 1976, 34,191-198.

Lewin, K. Resolving social conflicts. New York:Harper, 1948.

Lewis, F. M., Morisky, D. E., & Flynn, B. S. A testof the construct validity of health locus of control:Effects on self-reported compliance for hyper-tensive patients. Health Education Monographs,1978, 6, 138-148.

Linn, L. S., & Lewis, C. E. Attitudes toward self-care among practicing physicians. Medical Care,1979, 17, 183-190.

Mahoney, M. J., & Thoreson, C. E. (Eds,), Self-control: Power to the person. Monterey, Calif.:Brooks/Cole, 1974.

Marston, M. V. Compliance with medical regimens:A review of the literature. Nursing Research,1970, 19, 312-323.

Mclntosh, J. Processes of communication, informa-tion seeking and control associated with cancer:A selected review of the literature. Social Scienceand Medicine, 1974, 8, 167-187.

Mechanic, D. Medical sociology: A selective view.New York: Free Press, 1968.

Miller, G. P. Psychology as a means of promotinghuman welfare. American Psychologist, 1969, 24,1063-1075.

Mills, R. T., & Krantz, D. S. Information, choice,and reactions to stress: A field experiment in ablood bank with laboratory analogue. Journal ofPersonality and Social Psychology, 1979, 37, 608-620.

Rosenstock, I. M., & Kirscht, J. P. Why people seekhealth care. In G. L. Stone, F. Cohen, & N. E.Adler (Eds.), Health psychology—A handbook.San Francisco: Jossey-Bass, 1979.

Seligman, M. E. P. Helplessness. San Francisco:Freeman, 1975.

Shipley, R. H., Butt, J. H., Horwitz, B., & Farbry,J. E. Preparation far a stressful medical proce-dure: Effect of amount of stimulus preexposureand coping style. Journal of Consulting and Clini-cal Psychology, 1978, 46, 499-507.


Springarn, N. D. Doctors who tell all: They shouldunderstand that sometimes patients would prefernot to know. The Washington Post, July 30,1978.

Ullman, L. P. An empirically derived MMPI scalewhich measures facilitation-inhibition of threaten-ing stimuli. Journal of Clinical Psychology, 1962,7*, 127-132.

Vickery, D. M., & Fries, J. 'F. Take care of your-

self: A consumer's guide to medical care. Reading,Mass.: Addison-Wesley, 1976.

Wallston, B. S., Wallston, K. A., Kaplan, G. D., &Maides, S. A. Development and validation of theHealth Locus of Control (HLC) scale. Journal ofConsulting and Clinical Psychology, 1976, 44,580-585.

Received October 26, 1979 •

assessment of preferences for self-treatment and ...assessment of preferences for self-treatment and...

Documents