Research on how individuals formulate overall opinions on health protection measures indicates that the various components of the HBM differ in their degree of influence on final attitudes. Thus, the four component scores must be weighted according to their expected leverage on provider or lay behavior. The weights used in this analysis were as follows: risk (1); severity (2); benefits (2); barriers (−3). The negative sign for the barriers weight reflects the fact that a high score is less conducive to acceptance (Becker, 1974; Becker et al., 1977a,b,c).
Given these weights, both weighted additive and weighted multiplicative methods of combining scores are plausible. A weighted additive method would treat each factor as independent; a weighted multiplicative method might allow for the fact that two or more factors have an interactive effect on the overall likelihood of acceptance. The two methods considered for combining categories within the lay or provider domain were:
Weighted Additive Method
Weighted Multiplicative Method
As shown in Tables 6.2 and 6.3, the two methods give similar results, at least in terms of the ranking of the vaccines with respect to utilization. However, two theoretical considerations favor the multiplicative combination. First, the various HBM dimensions are actually subjective estimates or probabilities of some occurrence or outcome (e.g., the perceived likelihood of contracting a condition), and therefore the overall HBM estimate is appropriately the product of the individual-component probabilities. Second, it would not make conceptual sense to construct an HBM formula that would yield some predictive probability for a situation in which the estimate for any given component was zero (e.g., in the case of an individual who felt there was no possibility at all of contracting the condition). An additive model would still yield an overall HBM estimate by summing the values for the remaining model components; however, a multiplicative model would yield an overall estimate of zero. (The multiplicative approach is illustrated in Haefner and Kirscht, 1970.) Both methods have been carried forward to illustrate that they ultimately yield similar results.