Research in Nursing ond Health, 1984, 7, 275-285

Empirical Test of the Interaction Model of Client Health Behavior

Cheryl L. Cox and Klaus J. Roghmann

The general concepts, variables, and relationships described by the Interaction Model of Client Health Behavior (IMCHB) were used to guide a secondary data analysis of 203 womens decisions to request an amniocentesis. Step-wise multiple regression explained 58% of the variance in these decisions by jointly examining the factors which define the client as unique in her responses to an at-risk pregnancy, as well as factors describing the client-provider interaction. The women could be correctly classified as users or nonusers of prenatal diagnosis with 87% accuracy in a discriminant analysis based on the variables derived from the conceptual model. Finally, structural equation modeling was used to estimate the causal paths described by the general model. While the relationships proposed in the model need further empirical evauation, this first application serves to demonstrate the testability of the IMCHB and its potential to direct nursing inquiries.

The utility of conceptual models in advancing nursing knowledge has been scrutinized in recent years. Theorists have been challenged to develop models which provide a well-described focus on the nursing process, have direct clinical application, and demonstrate the potential to predict the effect of nursing interventions on client health outcomes (Brower & Baker, 1976; Chance, 1982; Downs, 1982; Hardy, 1978; Jacox, 1974). Several critics of nursing theory (Chance, 1982; Hardy, 1978; Stevens, 1979) suggested that the major theorists, while providing frameworks which depict ways to think about the client and nurse, have not specified their conceptual models sufficiently to generate tools and hypotheses which can be empirically tested. This study is a first attempt to empirically evaluate a new conceptual model of health behavior, the Interaction Model of Client Health Behavior

(IMCHB) (Cox, 1982). The model identifies the dynamics by which the clients uniqueness on a constellation of physio-psycho-socio-environmental factors combine with the professional interactioniintervention to produce health outcomes. While the model can be applied in most any health care setting and to any professional health care provider, the concepts and relationships described in the model strongly depict the philosophy of nursing practice and the dynamics inherent in the nursing process. IMCHB incorporates many of the variables included in earlier models of health behavior (Anderson, 1968; Becker, Drachman, & Kirscht, 1972; Hochbaum, 1958; Rosenstock, 1966; Suchman, 1966). The IMCHB additionally includes broad conceptual variables which can be operationalized to address a specific health care probledsituation as well as specific constructs

Dr. Cheryl L. Cox is on assistant professor in the Department of Public Health Nursing, College of Nursing, University of Illinois at Chicago. Dr. Klaus J. Roghmann is ossociote professor in the Department of Sociology and Pediatrics, University of Rochester, NY. This research was supported in port by DDHS Grant MC-R-360440 awarded to Klaus J. Roghmann and by Division of Nursing Grant NU-05442.02 awarded to Cheryl Cox. The authors wish to recognize the support of Dr. Richard Doherty, Medical Director of the Rochester, New York Regional Prenatol Diagnostic Services Program, and Dr. Judith Sullivan, Professor, Department of Public Health Nursing, College of Nursing, University of Illinois at Chicago. This article was received November 14, 1982, was revised, and was accepted for publication on September 29, 1983. Requests for reprints may be addressed to Dr. Cheryl L. Cox, Deportment of Public Health Nurstng, University of Illinois at Chicago, College of Nursing, 845 S. Damen Avenue, Chicago, IL 60612.

1984 Wiley 0160-68911841040275-1 1 $04.00

275

ELEMENT OF CLIENT SINGULARITY

ELEMENT OF CLIENT

ELEMENT OF HEALTH

BACKGROUND VARIABLES

I
-Non-Recutsive Block-

Figure 1.

An interaction model o client health behavior. f

which must be operationalized and tested as the model undergoes empirical analysis (e.g., intrinsic motivation). The elements of the IMCHB are represented in Figure 1; they include client singularity (environmental and physio-socio-demographic), client-professional interaction, and health outcome. The client singularity variableslconstructs represent the assessment factors addressed by the nursing process and those factors which can be operationalized as the clients human responses to actual or potential health problems (American Nurses Association, 1980, p. 9). The clientprofessional interaction represents the operational interaction factors important to the practice of nursing and reported within the nursing literature. The health outcomes are broad concepts and will accommodate multiple outcome measures which are both operational and sensitive to nursing interventions. The object of the IMCHB is to identify explanatory relationships between the three major elements and the characteristics which define these elements. The model is nonrecursive in the statistical sense; it demonstrates a multidirectional causal flow with feedback mechanisms to suggest the mutual influence of one set of elements upon another. The major hypothesis of the IMCHB is that health outcome is determined by the fit of the provider reactions to the clients responses to

a health concern, as well as to the clients physical and mio-envimnmental characteristics. This conceptualization gives greater weight to what occurs in the client-professionalinteraction rather than shifting the total responsibility for outcome to the client. The IMCHB was formulated at an abstract level; on that level it cannot be tested directly. Specific models on clearly defined health care issues must rm first be derived f o the general model. Because the IMCHB is designed for application to a variety of health care decisions and behaviors, the strength of the elements will vary according to the issue under consideration. For example, client attitudes and values might play a far more important role in a decision on abortion than they would in a decision about the use of emergency room services. It is for this reason that the model must be specified in accordance with a particular health care issue. Specifically, this study evaluated the IMCHB as a descriptor of at-risk (for fetal abnormalities) womens decisions to have, or not have, an amniocentesis. The IMCHB guided the selection of variables and specified the relationships for testing in a secondary data analysis. According to the conceptual model (IMCHB), there would be various interrelationships between the demographic and other background variables. These background variables would then determine the

HEALTH BEHAVIOR / COX AND ROGHMANN

277

clients attitudes? knowledge/beliefs, and motivation about amniocentesis and abortion. These later factors and aspects of the client-provider interaction would contribute directly and indirectly to the clients decisions to request or not request an amniocentesis.
METHOD

primary data collection and analysis are described in detail elsewhere (Roghmann & Doherty, 1983; Roghmann, Doherty, Newhouse, Nitzkin, & Sell, 1983).
Sample

Since 1977, survey data have been collected in a northeastern community to monitor the acceptance of prenatal screening by women at risk for fetal abnormalities and by their physicians. Most of the data were collected through mail and telephone surveys; in addition, data from birth certificates, program records, and census data from the Health System Area were used. Mail surveys of recent mothers over 34 years of age measured the awareness and knowledge of prenatal diagnosis and genetic counseling, previous history with birth defects, future reproductive plans, and attitudes toward amniocentesis and abortion. In addition, these women provided information on whether or not they were informed of the availability of prenatal diagnostic services prior to the 20th week of pregnancy, and on their decision to participate or not if so informed. Another mail survey asked physicians (general practitioners, family practitioners, internists, obstetricians, and gynecologists) the same question as were asked of the mothers. Additionally, the physkians responded to a detailed questionnaire about their practice patterns. A file of birth certificates allowed the investigators to compare statistics on women who received prenatal diagnostic services with statistics on all women delivering in the study period. Additionally, the birth certificates provided information on the clients first medical contact for prenatal care. This linkage of confidential birth certificate and medical record information was done by computer program without having to identify the subjects. The study described here is a secondary analysis of the original survey data. Secondary data analyses to test conceptual models are rarely possible because crucial variables are usually missing. This study is unique in that several surveys were linked to create a data set containing the major variables. Although complete congruency of the survey variables and the conceptual model variables was not possible, sufficient agreement existed to justify use of the surveys as the data base on which to test the model for the first time. The specific methodological approaches of the

All women over 34 years of age residing within the Health Systems Area (HSA) and delivering a liveborn infant within one year prior to the initial data collection and each year thereafter, were sent the mail survey questionnaire. Similarly, all physicians delivering obstetrical services within the HSA were mailed the Providers Survey. Through linking the multiple data sets for the 1980 data collection year, a pool of 410 respondents was created. A sample of 203 of these 410 clients who responded to the survey could be matched to (a) their specific physicians through data from the 1980 Providers Survey, (b) birth certificate information, and (c) service logs indicating whether or not the client had received genetic counseling, amniocentesis, both or none of these services. Comparative analyses showed the nonrespondents not statistically different demographically from the respondents.
Measures

Table 1 identifies the variables of the Prenatal Diagnosis Surveys, the interpretation, and the correspondence with the elements and factors which comprise the IMCHB. All measures were either interval level initially or were subsequently computed to create interval level indices. Exceptions included insurance coverage, religious affiliation, race, marital status, availability of physician counsel on amniocentesis, physician support/ nonsupport of prenatal diagnosis, and the clients final decision about amniocentesis. The creation of indices allowed the investigators to substantially reduce the raw data while simultaneously permitting the use of the more powerful statistical manipulations. For example, the index DISCUSS was created by adding together the number of people with whom the client had discussed amniocentesis. The independent variables are divided into three levels. Level I includes those variables which are largely fixed and not manipulated by the provider. Level I1 includes those variables representing the clients experience with birth defects and children, as well as environmental sources of information. Level 111 includes the process variables representing the clients subjective re-

278

RESEARCH IN NURSING AND HEALTH

Table 1. Correspondence of the Elements and Variables of the Interaeion Model of Client Health Behavior With the Study Variables of the Prenatal Diagnosis Surveys IMCHB Elements and Variables Background Variables Age Race Education Socio-economic Status Religion Environmental Resources Survey Variable

Interpretation

Age Black Education Sea Catholic Insurance

Expressed as an absolute age Client i s or is not Black Expressed as an absolute number of years
1 = highest

5 = lowest
Client i s or i s not Catholic Client has or does not have insurance coverage for amniocentesis Number of sources of information on amniocentesis available to the client Number of people available with whom to discuss amniocentesis ( b qides physician) Client did or did not discuss test with physicia Absolute number of living children which client had at time of interview Score representing clients total number o experiences with birth f defects herself, her own family or that of her spouse Absolute number representing total number of people supporting the clients having amniocentesis Absol Ute number representing total number of people, not supporting the clients having amniocentesis .%ore representing amount of decisional power desired by client in deciding on abortion ( 1 = none; 5 = 100%) Score representing amount of decisional power desired by client in deciding on amniocentesis (1 = none; 5 = 100%) Score representing clients total correct knowledge about amniocentesis Score representing clients knowledge about what amniocentesis can detect in the fetus
Score representing how liberal the client toward having an abortion herself

Sources Discuss

Docdis Previous Experiences Children

Expind

SociaI Influence

For

Against

Intrinsic Motivation

Amompow

Tmompow

Cognitive Appraisal

Knowind

Detect

Attitude

LegaI

HEALTH BEHAVIOR / COX AND ROGHMANN

279

Table 1. (continued) Correspondence of the Elements and Variables of the Interaction Model of Client Health Behavior With the Study Variables of the Prenatal Diagnosis Surveys
~

IMCHB Elements and Variables Affective Respose Fear

Survey Vario ble

Interpretation

Affect

Score representing clients fear for self or fetus as a result of undergoing amniocentesis Score representing clients fear of pain second0 ry to amniocentesis

Pain

CI ient-ProfessianaI Interaction
Physician Support Docpro

C Iients perceived physicians expression of support for amniocentesis

Clients perceived physicians expression of non-support for amniocentesis The extent of agreement between client and physician on how much say the client should have i n deciding on amniocentesis (0 = perfect agreement; 4 = total disagreement) The extent of agreement between client and physician on how much say the client should have in deciding on abortion (0 = perfect agreement; 4 = total disagreement) Client had neither genetic counseli ng nor amniocentesis 2 = Client had counseling only 3 = Client had amniocentesis
7

Doccon

Decisional Control

Amomtest

Amomab

Health Outcome Uti Iizati on

Log (Decision)

sponses to amniocentesislabortion and aspects of the client-provider interaction. The dependent measure is an ordinal level variable. Clients were viewed as being on a decision continuum ranging from the most conservative decision (participating in neither genetic counseling nor amniocentesis) to the most liberal decision (electing an amniocentesis). Because of the exploratory nature of the study, the investigators decided that the absence of an interval level dependent measure should not prevent proceeding with model testing and parameter estimation. This is currently a much debated issue in causal modeling; however, ample Monte Carlo studies indicated that in many cases, violations of the interval assumption is not of great consequence (Asher, 1976). While the investigators were aware of the distribution assumptions required of the variables examined, and felt comfortable with their

approach, results, and interpretation, the readers response will vary in accordance with their stand on this controversial subject.

Procedure To examine the relationships described by the IMCHB, the data were subjected to multiple regression techniques, discriminant analysis, and structural equation modeling. The latter used the maximum likelihood solution developed by Joreskog and Sorbom (1978). Such an approach allows for simultaneous evaluation of the theoretical structuring of a data set, inclusive of both direct and indirect effects of the independent variables on the dependent measure. Although structural equation models do not provide proof for causal relationships, they do evaluate the relative merit of the causal paths

280

RESEARCH IN NURSING AND HEALTH

proposed by the model. Thus, the IMCHB and the derived model for explaining clients decisions about amniocentesis were not specifically tested; rather, the consistency of the observed correlations with the associations described by the models was tested by applying the structural equation model to the observed data. In contrast to simpler path models of a recursive nature, the modeling applied here allowed for reciprocity in causation as the conceptual model demands.

predictors (beta weights > .05) of the clients decision about amniocentesis when all variables are examined simultaneously.
Table 2. Contributions of the Independent Variables to the Variance in 203 Clients Decisions on Amniocentesis Levels of Independent Va r iables

Multiple R

R2

I Socio-demographic

0.394 0.554 0.665 0.61 5 0.705

0.732

0.155 0.308 0.443 0.378 0.498 0.536 0.578

RESULTS

The individual and combined contributions of the three levels of the independent variables to the explained variance in clients decisions to have or not have an amniocentesis are presented in Table 2. Figure 2 represents the standardized partial coefficients within each level of the independent variables after the final regression step. These variables emerge as the strongest direct

characteristics II Experience and Resources 111 Subjective response and Interaction Process I and I1 I and Ill II and 111 I, I I , and 111

0.760

IMLI
INSURANCE (X, ) EDUCATION (X3) BLACK (X4)

\
LEVEL I1

I L I11 M

1
4

VARIANCE
IN

.23

DOCPRO

(xlo)

-.14

CHILDREN / (X7)

.ol

/
w o : :

SOURCES

-1.62

.18X

+ +

.1OX

-.07X

.09X 4
(.26)

.30X

+ -.09X6 + -.14X, +

.07X + m17Xg 8
(.05)
(.07)

.23X 10 (.13)

(.12)b +.16Xll

(.03) (.02) .22X12

(.05)
.O8Xl3

+-

(.08)

(-03) .06XL4

&(5) .8 b = Standard Error c N(203)

a

Figure 2.

Contributions of the independent variables to variance explained in deciding amniocentesis.

HEALTH BEHAVIOR / COX AND ROGHMANN

28 1

Two discriminant functions were computed with canonical correlations of .76 and .49. This suggests excellent discriminating ability. Wilk's lambda is a measure of the discriminating power of the variables used; the lambda's associated x2 (60) = 210.78, p 5 .OOO1 and x2 (29) = 50.678, p 5 .0076 denote that the amounts of discriminating information within these variables are statistically significant. In general, the group means and/or modal category on each independent variable suggests that those individuals who did not elect to have amniocentesis in contrast to those who did were (a) less often covered for the test by insurance, (b) more often Catholic, (c) had less experience with birth defects; (d) talked with fewer people about the test, (e) had more children, (f) were less accepting in their legal and personal attitudes toward abortion, (8) knew slightly less about amniocentesis; (h) had less social support for the test; (i) more often disagreed with the physician on how much decisional control the physician should have regarding amniocentesis, and (j) wanted more decisional control in deciding about abortion. Of the 117 clients who chose not to have am1

niocentesis, the discriminant analysis equations correctly identify 108 of them while incorrectly identify 9 as clients choosing amniocentesis. Of the 76 clients who choose amniocentesis, the equations correctly predict 64 of them, while incorrectly predict that 12 of them did not choose the procedure. The equations correctly predict 92.3% of the nonusers and 84.2% of the users. Overall, the equations are able to identify 87.2% of the cases correctly. No obvious trends are noted in the incorrect predictions. A number of causal models were estimated based on the general structure of the IMCHB. Those paths which did not show significant f statistics within the maximum likelihood program were eliminated, enhancing the fit of the model to the data. The structural model that best fits the data is described in Figure 3. The paths are illustrated and the standardized solutions (read as causal coefficients) are given for each path. The model is composed of two exogenous variables, insurance and previous experience with birth defects. These variables are causally independent of any other variables within the model and exert effects on other variables without being affected by them. Sixteen endogenous variables
GEPENDENT VARIABk

VARIABLES
LEML I1

LEMLI

Figure 3.

Causal model of prenatal diagnosis utilization.

282

RESEARCH IN NURSING AND HEALTH

are included in the model. These variables may or may not be completely independent in terms of antecedents, but often are found to form causal chains. The dependent measure is always an endogeneous measure. Twenty-two of the 35 hypothesized causal paths are significant at p S .05; five paths are marginally significant at p s .lo.

DlSCUSSlON The results of this first empirical trial of the model should be interpreted with caution. The IMCHB is a very complex model with a large number of variables examined in this study. The use of stepwise multiple regression to test for the amount of variance explained, the use of discriminant analysis to predict the highly discrete outcome variable, and the use of structural equation estimates to check for direct and indirect causal paths and mutual causality may not be the most appropriate statistical techniques for testing the IMCHB. Not unlike the use of secondary data analysis, however, these methods are sufficient for a first trial of the model which emphasizes the models general structure. The IMCHB with its emphasis on process during the client-provider interaction would optimally employ prospective dynamic modeling techniques, emphasizing only a few variables at a time with maximal accounting for measurement error. The findings from the regression analyses, discriminant analysis, and the causal model were consistent in illustrating the direct effects of the variables on clients decisions regarding amniocentesis. Each step eliminated variables which previous steps had included, thus subjecting the variables to more rigorous tests for inclusion. The final result demonstrated that the availability of insurance and the clients history relative to birth defects were the strongest independent direct influences on decisions about amniocentesis. Similarly, provider support for and against the test, and the client attitude toward legal abortion had strong direct influences. In terms of the indirect effects of the variables on the dependent measure, again, the findings from each successive analytical approach more parsimoniously reinforced the findings which had preceded it. Level I variables were independent indicators which were causally associated with specific life experiences and resources (Level 11); similarly, these experiences and resources entered into a causal configuration with the Level 111 variables, which exerted direct effects on the clients decision about amniocentesis.

Seventeen of the independent variables demonstrated significant direct and/or indirect explanatory paths to the outcome measure of utilization. The paths are consistent with those described by the general model. Of the variance in clients decisions on amniocentesis, 58% was explained by the elements and variables of the IMCHB.Multivariate studies of service use have in the past found only minor contributions to the explanation of use of health services (Mechanic, 1979). For example, Aday and Andersen (1975) were able to explain only 16% to 25% of the variance in service use with their model; Kohn and White (1976) accounted f for between 4% and 10%o the variance in their measure, and Wolinsky (1978) explained only 9% to 12% of the variance in utilization. Mechanic (1979) speculated that such negligible impact may be a function of how the issues are conceptualized, the nature of the measures, the aggregation of data, and the analyses performed. Simpler, traditional studies are clearer in their conceptualization, and demonstrate in a descriptive way the impact of some of the same study variables used in the multivariate studies; they provide insight without being able to quantify the impact. On the other hand, large multivariate studies often are weak in their conceptualization and measurements, and use multiple variables to fish for significant findings. This study was able to overcome many of the problems of both these approaches to the study of use of health services. It had the quantitative advantages which accompany the large-scale studies. Additionally, the study offered a richer view of the conceptualizationof the process which exists between the multiple variables, and it identified the dependent measure as a single and ordinal behavioral sequence. Every major element posited by the general model was demonstrated as important in describing the use of amniocentesis. The only conceptual variable not supported was the agreement between client and physician on decisional control. The fact that only weak indirect effects of this variable appear in the findings may support the relationship proposed in the model. That is, the actual interaction between the client and provider in the specific model may have been much earlier than the clients final decision; thus this decision may have been influenced by personnel other than the provider originally consulted by the client. Interruption in the continuity of interaction between client and provider may weaken the impact of that interaction on the clients decision. In addition, we cannot claim that our in-

HEALTH BEHAVIOR / COX AND ROGHMANN

283

direct measure of client-provider agreement on decisional control really captured the full meaning of the conceptual variable. The general model emphasized that it is the fit of the providers intervention/interaction with the uniqueness of the client that determines health behavior. These two elements and their fit, artificially captured through indices, will be examined in light of the findings presented.
Client Singulurity

provider decisions and interactions-experience and the availability of financial resources indirectly influence provider decisions; and (c) provider opinion affects ones social groups responses to a health care issue, and these responses combine to produce a specific health care attitude and behavior.
Conclusions

The demographic variables are depicted in the IMCHB as the progenitors of certain life experiences and the more subjective client responses (affective response, cognitive appraisal). These variables are assumed to interact simultaneously and often interdependently. The findings support these general assertions in relation to a decision on amniocentesis. The background variables were causally related to one another. These same variables became antecedents for the cognitive appraisal of the health concern and the affective responses of that client to the concern. Although not yet causally confirmed, the bivariate correlations (Cox, Sullivan, & Roghmann, 1984) and the regression analyses suggest that these same background variables and the cognitive appraisal may well be antecedent to the clients level of intrinsic motivation in this health care issue.
Clienf-Provider Interaction

The general assertions of the IMCHB on the interaction between the components of client singularity and client-provider interaction have been supported, but with some limitations. These limitations were due primarily to measurement problems, i.e., potential systematic measurement error, the lack of accessibility to definitive interaction variables, and poor quantitative measures of those which were examined. In addition, the primary client-provider relationship in those clients choosing amniocentesis may have been confounded by these clients interaction with members of the prenatal diagnosis counseling team (multiple providers). Despite these limitations, the investigators suggest that: (a) provider opinion/behavior has a direct and significant impact on subsequent client behavior, indicating a misspecification of those models of health behavior limited to client characteristics: (b) client singularity has an impact on provider behavior; the clients educational attainment, environmental resources, and motivational expression significantly and causally influence

The major conclusions of this study are that, first, both the individuality of the client and the client-professional relationship/interaction are significant determinants of health decisions and subsequent health behavior. Second, not only do these elements directly influence behavior, but in addition, they appear to have reciprocity with one another; client singularity influences professional response, and professional response likewise influences client singularity. Both the client singularity variables/constructs and interaction variables point to important areas for health care intervention. On the basis of these findings and conclusions, a number of general implications relative to the model are apparent. First, a general model to explain health and illness behavior and professional intervention must not be limited to a single focus, but rather should include all variables relevant to the type of health care issue under consideration. As demonstrated in this study, a combination of psychological, sociological, experiential, interaction, and environmental explanatory variables are needed to capture an explanation of a clients decision about amniocentesis. Second, perhaps even more important than which combination of categorical factors are influential in a clients decision on amniocentesis, is the process by which these multiple factors are related and the suggested causal impact of one factor upon another. Through an understanding of where and how these dynamic relationships ultimately influence behavior, the clinician is in a far better position to develop interventions which will have a significant impact on increasing the use of these services. The IMCHB emphasizes this process, drawing from theory developed within the contributing scientific areas and nursing research on client-provider interaction. Although the methodology to capture these processes is still underdeveloped, their relevance must be recognized. Third, a model purporting to explain health and illness behavior must include both the clientprofessional interaction together with client char-

284

RESEARCH IN NURSING AND HEALTH

actenstics if a full understanding of the resulting behavior is the objective. Client behavior has an impact on that of the professional, and professional opinionhehavior ultimately will have an impact upon client decisions and health actions. Not to address this fact is to severely limit the explanatory force of the model as well as to narrow the base for the development of health care interventions. Fourth, a model of client health behavior must make explicit the factors within the client-professional relationship that are presumed to make the most difference in the health care issue being explained. How that provider adapts to the client in terms of supplying emotional support, depth and specificity of health or procedural information, and the clients requisite amount of decisional control are all factors which potentially can have an impact on subsequent client behavior. Fifth, a conceptual model of health behavior should be capable of addressing the kinds of health care issues which are becoming increasingly more prevalent. Issues which require persistent client self-care management in contrast to direct provider management are the health concerns of today and the future. Clients must be more active determiners of their health states and the behaviors which will address that state. The model presented here gives a prominent role to the clients motivation for behavior. This motivation is based on the need to be self-determining to the extent that ones singularity will allow. This self-determinism was demonstrated to have an impact on provider behavior and thereby indirectly influence subsequent client behavior. Such a conceptualization will increasingly demonstrate its usefulness as health care evolves more and more toward the clients assuming responsibility for their own health behavior. Self-determinism as the vital force behind the assumption of persistent and difficult behaviors will be increasingly supported and encouraged by health professionals. The complexity of an explanationof client health behavior and professional interactionhntervention has been demonstrated. There are multiple direct and indirect effects of variables on a single health outcome. To approach an explanation of health behavior through an examination of direct effects only limits the potential of the explanation to serve as a guide for practical intervention. Although one variable may be demonstrated to have powerful direct effects, that power may be derived from a number of other variables which have no direct relationship to the final outcome. Unless one has specifically looked for these in-

direct contributions, important cues for subsequent interventions may be missed. The IMCHB is a useful tool with which to approach an explanation of a specific health behavior. Extensive empirical evaluation of the model, inclusive of the descriptive application to various heaIth problems and populations is needed. Additionally cross-validation of the model on similar samples or subsets of the same sample are needed to address the predictive value of the model. By approaching, nursing research on health behavior under the directives of such a comprehensive model, investigators will be able to move beyond the search for isolated variables and models which represent a single academic discipline or focus, and into the development of strong predictive theory which will guide professional health care practice and intervention.

Aday, L., & Anderson, R. Development of indices of access to medical care. Ann Arbor: Health Administration Press, 1975. American Nurses Association. Nursing: A Social Policy Statement. Kansas City, MO, 1980. Anderson, R. A behavioral model of fomilies use of health services (Research Series No. 25). Chicago: Center for Health Administration Studies, University of Chicago, 1968. Asher, H. B. Causal modeling. Beverly Hills: Sage, 1976. Becker, M., Drachman, R., 8 Kirscht, J. Predicting mothers compliance with pediatric regimens. Journal of Pediatrics, 1 972, 8 7, 843. Brower, H., & Baker, 6. Using the adaptation model in a practitioner curriculum. Nursing Outlook, 1976, 24, 6 8 6 6 8 9 . Chance, K. Nursing models: A requisite for professional accountability.Advances in Nursing Science, 1982,4, 57-65. Cox, C. L. An interaction model of client health behavior; theoretical prescription for nursing. Advances in Nursing Science, 1982, 5, 41-56. Cox, C. L., Sullivan, J. A., & Roghmann, K. J. A conceptual explanation o risk reduction behavior f and intervention development: Use of prenatal diagnosis. Nursing Research, 1984, 33, 168-1 73. Downs, F. A theoretical question. Nursing Research, 1982, 3 I , 259. Hardy, M. Perspective on nursing theory, Advances in Nursing Science, 1978, I, 37-48. Hochbaum, G. M. Public participation in medico1 screening programs: A sociopsychdogicalstudy (Public Health Service Pub. No. 572). Washington, DC: U.S. Government Printing Office, 1958. Jacox, A. Theory construction in nursing: An overview. Nursing Research, 1974, 23, 4-1 2. JBreskog, K., & Sorbom, P. Lisrel IV: Analysis of linear

HEALTH BEHAVIOR / COX AND ROGHMANN

285

structural relationships by the method of maximum likelihood. Chicago: International Education Resources, 1978. Kohn, R., & White, K. Health care-An international study: Report of the World Health Organization/ International collaborative study of medical care utilization. London: Oxford University Press, 1976. Mechanic, D. Correlates of physician utilization: Why do maior multivariate studies of physician utilization find trivial psychosocial and organizational effects? Journal of Health and Social Behavior, 1979, 20, 387-396. Roghmann, K. J . , & Doherty, R. A. Reassurance through prenatal diagnosis and willingness to bear children after age 35. American Journal of Public Health, 1983, 73, 760-762.

Roghmann, K. J., Doherty, R. A,, Newhouse, J., Nitzkin, J . , & Sell, R. The selective use of prenatal genetic diagnosis: Experiences of a regional program in upstate New York during the 1970s. Medical Care, 1983, 21, 1 1 11-1 125. Rosenstock, I. M. Why people use health services. Milbank Memorial Quarterly, 1966, 44, (Part 2), 94-1 27. Stevens, B. Nursing theory: Analysis, application, evaluation. Boston: Little, Brown, 1979. Suchman, E. A. Health orientation and medical care. American Journal of Public Health, 1966,56, 97-1 05. Wolinsky, F. Assessing the effects of predisposing, enabling, and illness morbidity characteristics on health service utilization. Journal of Health and Social Behavior, 1978, 19, 384-396.