| | Judgment processes in older adults' drug benefit evaluations⋆Abstract BackgroundThe Medicare Prescription Drug, Improvement and Modernization Act of 2003 will provide drug coverage to older adults through a variety of mechanisms, including stand-alone prescription drug benefits. Variation in cost-sharing and utilization controls is permitted, leading potentially to a wide variety of prescription benefit plans. However, little is known regarding the manner in which beneficiaries will integrate information to form plan evaluations. ObjectivesThe objectives of this study were to assess and compare the use of compensatory and configural judgment models in older adults' evaluations of drug benefit plans. MethodsThree focus groups were conducted with a total of 19 seniors to elicit relevant drug plan attributes. A separate group of 32 seniors then judged the suitability of 48 drug benefit profiles based on these attributes. Within-subjects regression analysis was used to reveal each participant's judgment policy. ResultsFocus groups suggested that copayment, premium, deductible, formulary use, and mail-service use were relevant plan attributes. Regression analyses showed that while most participants used compensatory judgment models in evaluation formation, 12 (37.5%) subjects used conjunctive judgment models. ConclusionsConfigural judgment processes are used frequently by older adults when evaluating drug benefit plans. Future research using more fine-grained techniques (eg, process tracing) might help further elucidate judgment processes in this context. 1. Introduction  Recently, a sweeping change was brought to the federal Medicare program with the passage of the Medicare Prescription Drug, Improvement and Modernization Act (MMA) of 2003 (PL 108-173).1 This act brings many important changes to the Medicare program, including the addition of prescription drug benefits. To facilitate competition through consumer choice, MMA allows drug benefits to be provided by private insurers through 2 primary mechanisms. First, they may be provided as part of comprehensive managed health care plans under Medicare Part C (the “Medicare Advantage Program”). This program will replace the current Medicare +Choice program. Second, for those beneficiaries who wish to remain in the standard Medicare fee-for-service plan, MMA provides for the establishment of stand-alone prescription drug plans under Medicare Part D, a novel insurance product previously unavailable. The federal government also may offer a “fallback” plan of its own for rural areas with few coverage options. The MMA permits considerable variation in the design of stand-alone drug benefits created under Part D, with the stipulation that these plans may be no less generous than the standard drug coverage plan.1, 2 Variation in cost-sharing provisions is permitted, as are differences in administrative cost control mechanisms. Drug benefit plans may not have a deductible larger than the standard deductible ($250/y in 2006), but may have a lower deductible. Restrictive drug formularies are permitted as long as at least some medications within each therapeutic category are covered. Given that private pharmacy benefit management companies are expected to offer many of the stand-alone plans, it is likely that mail-order pharmacies will be used to lower dispensing fees.3 Thus, variation in these attributes has the potential to produce a large number of coverage options. Little is known, however, regarding the manner in which older adults (those aged 65 and older) will combine information on drug benefit attributes to form judgments, and ultimately, make choices among these plans. 1.1. Judgment processes Research suggests that individuals may use a variety of processes when they integrate information on any multiattribute alternative (eg, a prescription drug benefit plan) to form a comprehensive judgment regarding the alternative.4, 5, 6 One class of these judgment procedures is known as compensatory processes. The distinguishing feature of compensatory judgment processes is that a negative evaluation on one aspect of the alternative (eg, premium) can be offset by a positive evaluation of some other aspect (eg, copayment). Compensatory processes include those with optimal weighting applied to each attribute, as well as those with equal, or even random weights, applied to each attribute.5, 7 These models require significant effort from the decision maker because weights must be assigned to each relevant attribute and then used to arrive at a unique judgment for each alternative. To simplify the judgment process, individuals might use a second class of judgment procedures known as configural models.5, 6 Configural strategies allow one to simplify judgment formation by combining attribute information according to a predetermined pattern. Although a wide variety of these models have been identified, only 2 were considered in this study. The first of these is known as a conjunctive judgment model. In a conjunctive judgment model, the judge is assumed to set some minimum cutoff value for each attribute. As a result, any alternative with any attribute value failing any of these cutoffs automatically is rejected. For example, an individual might decide that to be judged “acceptable” or better, a drug benefit must have a monthly premium of $100 or less, a copayment of no more than $10, and must not use a formulary. Any drug benefit failing any one of these tests would receive a poor evaluation, regardless of the attractiveness of its other features. Disjunctive judgment models are the second type of configural judgment model considered here. When using a disjunctive process, the decision maker judges an option to be acceptable if it has an exceptional value on any one attribute, regardless of its values on other attributes. To whit, an individual might decide that any drug benefit with a very low monthly premium, a low copayment, or a low deductible would be acceptable, regardless of all other attributes. 1.2. Past studies To date, only 2 studies known to the authors have examined choices among stand-alone drug benefits.8, 9 Using a convenience sample of 130 consumers (most younger than 65 years), Holdford and Carroll8 used conjoint analysis to study the effects of varying 3 drug plan attributes (copayment amount, freedom to choose one's pharmacy, and use of a restrictive formulary) on the likelihood of prescription benefit plan choice. Their results showed that all 3 of these attributes were associated significantly with likelihood of choice, with use of a restrictive formulary being the most important factor in these decisions. Not surprisingly, lower copayment amounts were related positively to choice likelihood, as was the ability to choose one's pharmacy. Cline and Mott9 used conditional logit modeling to study hypothetical drug benefit plan choices in a mail survey of 1086 older adults in Wisconsin. Respondents were asked to choose a plan that would best meet their needs from a menu of 4 drug benefit plans. These plans varied with regard to 4 attributes (copayment amount, monthly premium, use of a restrictive formulary, and required use of a mail-service pharmacy). These authors concluded that monetary considerations, such as premium and expected out-of-pocket costs, were significant predictors of observed choices, with higher copayment and premium amounts inversely related to plan choices. These studies have provided significant insights into the factors that impact drug benefit choice. However, they suggest a number of avenues for research that have yet to be explored. For example, these investigations have implicitly assumed that individuals use a linear, compensatory judgment model in their choice processes. They did not attempt to assess participants' use of other judgment (ie, configural) strategies in their analyses. In addition, previous studies did not assess the substantive contribution (if any) that use of configural judgment rules made in individuals' evaluations of prescription drug benefits. Another path that has received relatively little attention is the analysis of variability in drug benefit decisions at the individual level (ie, idiographic analysis). Holdford and Carroll8 estimated attribute importance weights for each individual, but they presented these data in aggregate form. More detailed analysis of drug benefit judgments among a small number of individuals may serve as a complement to the aggregate (ie, nomothetic) analyses used previously. Idiographic analysis may provide insights into drug benefit decision-making in a number of ways. For example, variability in attribute weighting can be studied by examining the means and standard deviations associated with attribute weights. A small standard deviation would imply that the attribute was used consistently across participants in arriving at summary judgments, whereas a relatively large standard deviation would suggest the opposite. Nomothetic analysis yields only estimates of the average importance of a given attribute across all participants. In addition, idiographic analysis also permits variation in judgment processes to be examined, so that questions about the proportion of persons using various judgment processes and the relative importance of these processes in explaining judgments can be addressed. 1.3. Study purpose The purpose of this study is to better understand the manner in which drug plan attribute information is integrated in evaluation formation among older adults. The specific objectives of this study are to (1) describe the frequency with which compensatory and configural judgment models are used in older adults' drug benefit evaluation processes and (2) assess the substantive importance of configural model use among these individuals. In contrast to the largely nomothetic analyses used in prior studies, we approach these objectives using an idiographic analysis framework, permitting us to exploit individual variability in judgment processes.8, 9 This study was approved by the Research Subjects Protection Program at the University of Minnesota. 2. Methods 2.1. Theoretical framework The present study was framed within social judgment theory, a descriptive metatheory of the manner in which individuals form judgments.10, 11 Social judgment theory suggests that an individual's judgments regarding an object, service, person, etc are a function of its salient attributes or cues. Social judgment theory focuses not on discovering general rules governing judgment formation, but on the way individuals combine cues to arrive at an overall evaluation. Social judgment theory posits that in order for inferences about the manner in which attributes impact judgment formation to be valid, the task used to generate the judgments must be representative, or mirror closely, the environment, or ecology, of the judgment task as it would occur in reality.12, 13 Thus, it is important for researchers to understand not only all salient cues that an individual might consider when forming an evaluation, but also the values that those cues are likely to take.11 The judgment process itself may take multiple forms in social judgment theory, including the compensatory, conjunctive, and disjunctive. Finally, because social judgment theory is an idiographic framework, statistical analyses of judgment formation are conducted at the individual, and not group, level before any aggregation is attempted.10 An example may help to clarify the use of social judgment theory in the study of judgment processes. Consider the hypothetical example of families in need of new vehicles. A researcher might first conduct a literature review on the topic to see if any other studies (laboratory or naturalistic) had been conducted. These might yield useful data on the cues typical families might consider in this purchase decision (eg, price, mileage). However, social judgment theory suggests that a more important source of cues would be those persons actually involved in the decision. Thus, interviews with several families and perhaps salespeople would aid in understanding all relevant vehicle attributes. The researcher also might observe first-hand several families examining and purchasing new vehicles. The focus on representative design requires that the likely cue values also be elicited. Thus, only prices from $22,000 to $32,000 might be within the reach of the families under consideration, and the families might only be considering vehicles getting at least 20 miles per gallon. It would also be important to understand the relevant portion of family (ie, parents only, parents and children) likely to be involved in the purchasing decision. When all relevant cues, values, and decision makers have been elicited, the researcher could reasonably be satisfied that the judgment ecology was understood, allowing the design of a choice experiment yielding valid results. 2.2. Drug benefit attribute identification To identify the prescription benefit attributes important to seniors, a series of 3 focus groups was conducted with a total of 19 volunteers aged 65 and older from the Minnesota Senior's Federation, an advocacy group for older adults, in June 2003. Each group was composed of 6 (2 groups) or 7 (1 group) persons. The sole inclusion criterion was that the individual be aged 65 or older. Focus groups were chosen over other qualitative methodologies because they are useful for providing background information and understanding impressions about various products and services from a number of individuals relatively quickly.14 Following Krueger and Casey,15 3 groups, each with a minimum of 6 participants, were used to achieve “saturation,” or the point at which no new information was being obtained. Each group lasted approximately 2 hours. Most of the participants in these groups were female (63%) and slightly more than half (52.6%) were 75 years or older. Nearly two-thirds (63.2%) had a 4-year college degree or more. The focus groups were moderated by the researchers using a standard questioning route that centered on the (1) attributes of drug insurance most salient to the participants, (2) participants' understanding of these attributes and their functions, and (3) language used by the group members when talking about drug plan features. Each participant received a $20 honorarium. Data were collected during these focus groups using audio recordings and field notes. The audiotapes were transcribed verbatim immediately afterward to promote reliability. The authors then reviewed the transcripts and field notes closely to identify and classify mentions of drug plan attributes. The results of this analysis showed that group members mentioned cost-sharing provisions such as copayments, deductibles, and premiums frequently. Further study of the transcripts revealed that administrative utilization control mechanisms such as restrictive formularies and required use of mail-order pharmacies were salient for some group members, but were not well understood among participants who had not encountered these restrictions. Overall, the attributes identified were similar to those previously studied as determinants of drug benefit choice.8, 9 2.3. Drug benefit judgment task design Both focus group data and the results of past research suggested the use of 5 drug benefit attributes (copayment, annual deductible, monthly premium, use of a restrictive formulary, and required use of a mail-service pharmacy) in the design of the judgment task. Attribute values selected were based on the then current estimates of cost-sharing amounts in Medicare prescription drug benefits promulgated.16 These values included: •Copayment: $5, $10, or $15 •Annual deductible: $100 or $250 •Monthly premium: $35 or $50 •Use of a restrictive formulary: yes or no •Required use of a mail-order pharmacy: yes or no Forty-eight (3 × 2 × 2 × 2 × 2) unique hypothetical prescription insurance plans were created by varying the 5 attributes systematically in all possible combinations. The hypothetical drug plans were ordered randomly and then incorporated into a booklet. Participants evaluated all 48 drug benefit plans. A second section of this booklet contained questions concerning current prescription drug coverage, use, and demographics. 2.4. Data collection In August 2003, 32 volunteers, aged 65 and older, were recruited from the Minnesota Senior's Federation to serve as drug benefit judges. None of these individuals had served as participants in the focus groups used to elicit relevant attributes. Again, the only inclusion criterion was that the individual be aged 65 or older. Before volunteers began the judgment task the researchers explained the definition of each plan attribute and its potential impact on prescription costs or procurement. For example, participants were told that a restrictive formulary is a limited list of medications that the plan will cover and that a formulary might require the individual to switch from a medication they are currently using to a new, but similar (ie, within the same therapeutic class), prescription medication covered by the plan. Participants were also told that plans requiring the use of a mail-order pharmacy would cover medications necessary for an acute need (eg, analgesics) at any local pharmacy, but that any maintenance drug (ie, a medicine to be taken for an indefinite period of time) would have to be obtained through a mail-order pharmacy. Volunteers were asked to evaluate the suitability of each plan for their own needs by using a 5-point Likert-type scale ranging from “1”=“This plan would not suit me at all” to “5”=“This plan would suit me perfectly” (see Figure 1). It was stressed that judgments of plan suitability were to be made with respect to the individual, and not any other family member. Any questions regarding the hypothetical plans and the judgment task were answered before subjects proceeded with the task. Members of this convenience sample also received a $20 honorarium for participation. 2.5. Data analysis The manner in which each individual integrated attribute information was examined using within-subjects multiple regression analysis.11 The participant's judgment of the suitability of each drug benefit plan served as the dependent variable and the 5 drug benefit attributes served as independent variables in this procedure. The standardized beta coefficients from these analyses were used to gauge the relative importance of each plan attribute in the individual's evaluations. Compensatory judgment processes were analyzed first using the standard linear regression model: where Y=the individual's rating of the suitability of the drug plan; a=a constant; bi=the standardized regression coefficient for the ith plan attribute; Xi=the value of the plan's ith attribute. Estimation by least squares yields an equation representing an optimally weighted average of the values assigned to each of the 5 attributes comprising the hypothetical drug benefits. A variety of statistical models have been proposed for use in the study of configural information processing.17, 18, 19 The “scatter” model was used in the current study because it has the advantage of accounting for both the elevation and scatter of an alternative's profile of attributes, while other models account only for their elevation.19, 20Elevation refers to the average of attribute values for an alternative, while scatter refers to the standard deviation of attribute values within the alternative's profile. The following regression model was used to detect the use of conjunctive or disjunctive judgment processes in the evaluation of drug benefits: where bk+1=the standardized regression coefficient representing variability in the plan's attribute profile, zi=the standardized value of the ith attribute across profiles,  =the average of the standardized attribute values within the profile, and the meanings of all other symbols are the same as in Equation (1). The first summation term of the scatter model (2) is identical to that of Equation (1) and accounts for the elevation of an alternative's set of attributes. The second summation term accounts for the scatter, or variability, among the attributes. Qualitative benefit attributes, such as formulary and mail-order pharmacy use, were not included in this summation because generally it is not meaningful to speak of a mean or variance for these variables.21
Equation (2) is able to detect configural information processing in the following manner: consider 2 drug benefit plans (A and B) composed of only 2 equally important attributes, premium and deductible. Plan A's premium and deductible are both equal to $200. Plan B's premium is only $100, but its deductible is $300. Both plans' elevation will be equal, but Plan B will have more scatter. If an individual is using a conjunctive judgment rule stating that a plan's deductible should not exceed $250, he/she will attend closely to this attribute, and the bk+1, or scatter, term should be statistically significant and negative.20 If an individual uses a disjunctive strategy stating that plans with low premiums or low deductibles are acceptable he/she will pay close attention to Plan B's relatively low premium and the scatter term should be significant and positive. Standard significance tests of the bk+1 coefficients from each participant's judgment model were used to assess the prevalence of configural judgment model use. The significance of the change in squared multiple correlation coefficients (increment to R2) between the linear compensatory and the scatter regression models then was compared to gauge the substantive importance of configural processing in the individual's evaluations of drug benefit plans.22 All statistical tests were conducted at the α=0.05 level of significance. 3. Results Table 1 displays the characteristics of the 32 older adults who volunteered for the judgment analysis portion of the study. Most of the sample members were female and under the age of 75 (both 59.4%) with less than half (46.9%) completing college. Slightly more than half reported that their health was “fair” or “good” and that they used more than 3 prescription drugs in the past 30 days (53.1%). Nearly one-third currently used a mail-order pharmacy. | | |  | Variable | n | % | |
|---|
| Gender | | | | | Male | 13 | 40.6 | | | Female | 19 | 59.4 | | | Age | | | | | 65-74 | 19 | 59.4 | | | 75 and older | 13 | 40.6 | | | Marital status | | | | | Married | 13 | 40.6 | | | Not married | 19 | 59.4 | | | Education level | | | | | Less than college | 17 | 53.1 | | | Four year college or more | 15 | 46.9 | | | Household income | | | | | Less than $35,000/y | 12 | 41.4 | | | $35,000/y or more | 17 | 58.6 | | | Health status | | | | | Fair or good | 17 | 53.1 | | | Very good or excellent | 15 | 46.9 | | | Private insurance for Rx drugs | | | | | Yes | 15 | 46.9 | | | No | 17 | 53.1 | | | Number of different Rx drugs used the past 30 days | | | | | 3 or fewer | 15 | 46.9 | | | More than 3 | 17 | 53.1 | | | Monthly out-of-pocket payment for Rx drugs | | | | | $80 or less | 18 | 56.3 | | | More than $80 | 14 | 43.8 | | | Currently uses a mail-service pharmacy | | | | | Yes | 10 | 31.3 | | | No | 22 | 68.8 | | | | |
The results of within-subjects regression analyses using the linear compensatory model showed that, on average, a plan's deductible appeared to be most important in these evaluations (standardized regression coefficient=−0.52), while the use of a restrictive formulary was least important (standardized coefficient=−0.052) (Table 2). These analyses also demonstrate significant variability in the manner in which subjects weighted drug benefit attributes to arrive at a judgment of suitability. The standard deviations associated with the coefficient means often are larger than the averages themselves, implying significant heterogeneity in attribute preferences. In addition, the proportion of variation in suitability judgments explained by the 5 attributes ranged from a low of 0.318 to a high of 0.979, providing additional support for idiographic analysis in the study of drug benefit preferences. | | | | ID # | Premium | Co-payment | Formulary | Mail order | Deductible | R2 | |
|---|
| 1 | −0.044 | −0.054 | 0.000 | −0.022 | −0.987 | 0.979 | | | 2 | −0.498 | −0.239 | −0.108 | 0.022 | −0.454 | 0.523 | | | 3 | −0.809 | −0.222 | −0.140 | −0.028 | −0.195 | 0.763 | | | 4 | −0.209 | −0.047 | 0.019 | −0.019 | −0.819 | 0.717 | | | 5 | −0.735 | −0.300 | 0.000 | 0.082 | −0.041 | 0.639 | | | 6 | −0.353 | −0.233 | 0.244 | −0.136 | −0.353 | 0.381 | | | 7 | −0.135 | −0.099 | 0.027 | −0.514 | −0.622 | 0.680 | | | 8 | −0.469 | −0.221 | −0.036 | −0.108 | −0.614 | 0.659 | | | 9 | −0.189 | −0.232 | 0.126 | 0.032 | −0.819 | 0.778 | | | 10 | −0.295 | −0.253 | 0.177 | −0.059 | −0.531 | 0.468 | | | 11 | 0.000 | −0.061 | 0.000 | 0.149 | −0.867 | 0.777 | | | 12 | −0.941 | −0.144 | 0.071 | 0.024 | −0.071 | 0.916 | | | 13 | −0.075 | −0.101 | −0.193 | −0.056 | −0.515 | 0.318 | | | 14 | 0.068 | −0.104 | 0.034 | −0.034 | −0.782 | 0.630 | | | 15 | −0.073 | −0.467 | −0.036 | 0.109 | −0.726 | 0.763 | | | 16 | −0.113 | −0.138 | 0.525 | −0.413 | −0.188 | 0.513 | | | 17 | −0.730 | 0.033 | −0.081 | 0.081 | −0.568 | 0.870 | | | 18 | −0.053 | −0.089 | −0.348 | 0.073 | −0.536 | 0.432 | | | 19 | −0.248 | −0.532 | −0.041 | −0.083 | −0.497 | 0.600 | | | 20 | −0.428 | −0.116 | −0.016 | −0.111 | −0.713 | 0.718 | | | 21 | −0.164 | −0.029 | −0.164 | 0.164 | −0.539 | 0.372 | | | 22 | −0.285 | −0.349 | −0.372 | −0.285 | −0.547 | 0.721 | | | 23 | −0.577 | 0.000 | 0.000 | 0.000 | −0.577 | 0.667 | | | 24 | 0.000 | 0.097 | −0.422 | −0.422 | −0.317 | 0.466 | | | 25 | −0.091 | −0.280 | 0.000 | −0.183 | −0.822 | 0.796 | | | 26 | −0.088 | −0.217 | −0.088 | −0.029 | −0.678 | 0.523 | | | 27 | −0.146 | 0.020 | −0.243 | −0.631 | −0.081 | 0.486 | | | 28 | −0.067 | −0.519 | −0.022 | 0.022 | −0.424 | 0.455 | | | 29 | 0.076 | −0.647 | −0.034 | −0.095 | −0.252 | 0.496 | | | 30 | −0.032 | −0.353 | −0.287 | −0.351 | −0.192 | 0.368 | | | 31 | −0.174 | 0.096 | −0.181 | −0.638 | −0.474 | 0.689 | | | 32 | −0.017 | −0.118 | −0.079 | −0.207 | −0.849 | 0.789 | | | Ave | −0.247 | −0.185 | −0.052 | −0.115 | −0.520 | 0.624 | | | SD | 0.269 | 0.181 | 0.183 | 0.215 | 0.255 | 0.172 | | | | |
Estimation of the more complex scatter regression model provided evidence of configural processing in the evaluation of stand-alone drug benefits (Table 3). In this analysis, 12 of 32 (37.5%) scatter coefficients were statistically significant at the 0.05 level or more. All of these coefficients were negative in sign, implying that these individuals likely were using a conjunctive model in their judgment strategies. However, inclusion of the scatter coefficient improved the proportion variance explained, on average, less than 5% for the entire sample, and of the 12 models with statistically significant scatter coefficients only 3 (9.38% of the overall sample) produced statistically significant increments to R2. 4. Discussion One goal of this study was to understand the frequency with which older adults might use compensatory and configural information integration strategies in their assessments (and ultimately, choices) of stand-alone prescription drug plans. Our findings demonstrated that linear, compensatory processing of information was most common in this context, despite the fact that compensatory model use is, in general, more effort-intensive than other forms of information integration.23 This result may be plausible for a number of reasons. First, this was a contrived experiment with only 5 attributes taking a limited number of values, simplifying information processing. Second, many of the participants in this study likely were familiar with health insurance choices because older adults often are confronted with decisions regarding Medicare HMOs and Medicare supplemental (“Medigap”) policies. The results of this preliminary investigation suggest that the use of configural judgment rules is somewhat common, with slightly more than one-third (37.5%) of the participants using a conjunctive strategy to arrive at an evaluation of the suitability of a given drug benefit (Table 3). One interpretation of this finding is that many study volunteers found the task of simultaneously integrating information on 5 plan attributes difficult and searched for a way to simplify this task, an explanation that finds some support in focus groups on health insurance choice conducted with Medicare beneficiaries.24, 25 This finding may have implications for the Medicare prescription drug benefit, because this program is predicated on the assumption that Medicare beneficiaries have well-formed preferences for various drug plan attributes and are able (and willing) to make trade-offs among them. If many Medicare recipients use simplifying decision rules (eg, cutoff points) instead of making careful decisions regarding the weighting of and concessions among attributes, the quality and efficiency that markets promote may fail to materialize.26 Findings regarding the frequency of configural strategy use must be tempered by results regarding the substantive importance of increases in the proportion of judgment variance explained (Table 3). This finding may have several explanations. One is that the statistical significance of the scatter coefficients is simply artifactual, and that these individuals are using a compensatory process in their drug benefit evaluations. Another reasonable explanation is that these persons were using a combination of 2 strategies in their integration of plan information.6 For example, a person might first use a conjunctive strategy stating that any plan with a copayment above $10 and a premium above $40 automatically receives the lowest evaluation (“does not suit me at all”) and then proceed to evaluate the remaining plans more carefully using a more complex compensatory rule. Finally, it must be kept in mind that the linear portion of the scatter model may mask the use of many configural strategies, leading to a small increment in variance explained when the scatter term is added.7, 27 The finding that a significant proportion of participants in this study appeared to use a conjunctive judgment strategy when evaluating drug benefits may have other implications for insurers and pharmacy benefit managers who might consider offering a prescription drug plan under the forthcoming Medicare Part D program. If all beneficiaries use a compensatory judgment process, insurers can rely, for example, on a high premium value being balanced by low copayment and deductible values and design their drug benefit product accordingly. However, if beneficiaries use conjunctive processes, insurers must (1) understand what attributes are critical in the judgment process and (2) understand the maximum (or minimum) acceptable attribute values when designing insurance products for this market (eg, deductibles over $250/y may lead to rejection, regardless of a low copayment and deductible). A failure to understand this judgment process could result in very poor enrollment in an otherwise well-designed benefit plan. 4.1. Future research Although the estimation of statistical models from manipulated attributes and stated judgments provides a powerful technique for understanding information integration, this process ultimately is a “black box” phenomenon.28 Statistical models are, at best, simply models that allow progress to be made into the explanation of a phenomenon through abstract representation. An alternate research strategy that might provide further insight into the judgment process used in this context is that of process tracing. This method involves presenting subjects with the judgment task and then collecting concurrent verbal protocols from them describing the way in which judgments are being formed.29 Another research method that may prove useful is that of eye fixation analysis, in which participants' patterns of eye fixations on various attributes are used to infer judgment processes.30 These research strategies could be used in combination with linear models to gain a more in-depth understanding of the manner in which information is acquired, sorted, and integrated in the evaluation of drug benefit plans. 4.2. Limitations The results of this study should be viewed in light of several limitations. The data analyzed herein were collected among a small, self-selected group of older adults limiting generalizability. In addition, many of the participants were intimately involved in Medicare drug benefit advocacy efforts and thus, were better informed regarding drug coverage than a “typical” Medicare beneficiary might be. The consistency or reliability of individuals' judgments was not directly tested. This could have been done by estimating each judgment model on most plans and then comparing the predicted judgment with the stated judgment for the remaining plans. However, each participant's judgment reliability can be inferred from the multiple R2 values presented in Table 2, Table 3. For example, a person giving inconsistent weights (or for that matter, assigning random values to the suitability judgments) will exhibit a lower multiple R2 value than someone that is highly consistent in cue weighting.11 This research was conducted in only one state and as such it is unknown whether the results obtained might generalize to seniors living in other areas of the country. Finally, the drug benefits featured in this study did not include many attributes (such as benefit caps or “donut holes”) featured in the MMA that likely will impact drug benefit evaluations. 5. Conclusions The goal of this study was to better understand the processes by which older adults integrate attribute information to form drug benefit evaluations. Although compensatory processes appear to dominate in this context, conjunctive judgment rules were used by 12 (37.5%) participants. However, configural information integration strategies were substantively important for only a few individuals. A large minority of seniors appear to use configural judgment processes in order to simplify prescription drug benefit evaluations, although more research is needed to ascertain both the importance of these processes as well as the use of other judgment rules within the context of drug benefit choice. Acknowledgment Dr Cline gratefully acknowledges the American Foundation for Pharmaceutical Education–American Association of Colleges of Pharmacy for a New Investigator Program grant that made this research possible. References 1. 1Health Policy Alternatives. Prescription Drug Coverage for Medicare Beneficiaries: A Summary of the Medicare Prescription Drug, Improvement and Modernization Act of 2003. Menlo Park, Calif: The Henry J. Kaiser Family Foundation; 2003;. 2. 2Moon M. How Beneficiaries Fare Under the New Medicare Drug Bill. (Issue Brief #730). New York, NY: The Commonwealth Fund; 2004;. 3. 3Congressional Budget Office . Issues in Designing a Prescription Drug Benefit for Medicare. Washington, DC: Congressional Budget Office; 2002;. 4. 4Hoffman PJ. The paramorphic representation of clinical judgment. Psychol Bull. 1960;57:116–131. MEDLINE |
CrossRef
5. 5Einhorn HJ. The use of nonlinear, noncompensatory models in decision making. Psychol Bull. 1970;73:221–230. MEDLINE |
CrossRef
6. 6Peter JP, Olson JC. Consumer Behavior and Marketing Strategy. Boston, Mass: Irwin McGraw Hill; 1999;. 7. 7Dawes RM, Corrigan B. Linear models in decision making. Psychol Bull. 1974;81:95–106.
CrossRef
8. 8Holdford D, Carroll NV. Consumer preferences for types of cost containment in prescription drug plans. J Man Care Pharm. 2002;8:192–198. 9. 9Cline RR, Mott DA. Demand for a Medicare prescription drug benefit: exploring consumer preferences under a managed competition framework. Inquiry. 2003;40:169–183. MEDLINE 10. 10Hammond KR, Stewart TR, Brehmer B, Steinmann DO. Social judgment theory. In: Arkes HR, Hammond KR editor. Judgment and Decision Making. New York, NY: Cambridge University Press; 1986;. 11. 11Cooksey RW. Judgment Analysis: Theory, Methods, and Applications. New York, NY: Academic Press; 1996;. 12. 12Brunswik E. Representative design and probabilistic theory in a functional psychology. Psychol Rev. 1955;62:193–217.
CrossRef
13. 13Engel JD, Wigton R, LaDuca A, Blacklow RS. A social judgment theory perspective on clinical problem solving. Eval Health Prof. 1990;13:63–78.
CrossRef
14. 14Churchill GA. Marketing Research: Methodological Foundations. 6th ed. Fort Worth, Tex: Harcourt Brace; 1995;. 15. 15Krueger RA, Casey MA. Focus Groups: A Practical Guide for Applied Research. 3rd ed.. Thousand Oaks, Calif: Sage Publications; 2000;. 16. 16Sroka CJ. Medicare: Beneficiary Cost-sharing Under Proposed Prescription Drug Benefits. Washington, DC: Congressional Research Service (order code RL31525); 2003;. 17. 17Einhorn HJ. Use of nonlinear, noncompensatory models as a function of task and amount of information. Org Behav Human Perform. 1971;6:1–27. 18. 18Goldberg LR. Five models of clinical judgment: an empirical comparison between linear and nonlinear representations of the human inference process. Org Behav Hum Perform. 1971;6:458–479. 19. 19Brannick MT, Brannick JP. Nonlinear and noncompensatory processes in performance evaluation. Org Behav Hum Decis Process. 1989;44:97–122. 20. 20Ganzach Y. Theory and configurality in expert and layperson judgment. J Appl Psychol. 1994;79:439–448.
CrossRef
21. 21Stevens SS. On the theory of scales of measurement. Science. 1946;103:677–680. 22. 22Cohen J, Cohen P. Applied Multiple Correlation/Regression for the Behavioral Sciences. Hillsdale, NJ: Lawrence Erlbaum; 1983;. 23. 23Payne JW, Bettman JR, Johnson EJ. The adaptive decision maker: effort and accuracy in choice. In: Hogarth RM editors. Insights in Decision Making: A Tribute to Hillel J. Einhorn. Chicago, Ill: University of Chicago Press; 1990;. 24. 24Public Opinion Strategies , Peter D. Hart Research Associates. Reactions to the New Medicare Law. Menlo Park, Calif: Henry J. Kaiser Family Foundation; 2004;. 25. 25Hibbard JH, Jewett JJ. What type of quality information do consumers want in a report card?. Med Care Res Rev. 1996;53:28–47. MEDLINE 26. 26Hibbard JH, Jewett JJ. Will quality report cards help consumers?. Health Aff. 1997;16:218–228. 27. 27Wiggins N, Hoffman PJ. Three models of clinical judgment. J Abnormal Psychol. 1968;73:70–77. 28. 28Hoffman PJ. The paramorphic representation of clinical judgment. Psychol Bull. 1960;57:116–131. MEDLINE |
CrossRef
29. 29Payne JW, Braunstein ML, Carroll JS. Exploring predecisional behavioral: an alternative approach to decision research. Org Behav Hum Decis Process. 1978;22:17–44. 30. 30Russo JE, Rosen LD. An eye fixation analysis of multialternative choice. Memory Cog. 1975;3:267–276. College of Pharmacy, University of Minnesota, Minneapolis, MN 55455, USA  Corresponding author. Tel.: +1 612 624 0124; fax: +1 612 625 9931.
PII: S1551-7411(04)00005-1 doi:10.1016/j.sapharm.2004.12.004 © 2005 Elsevier Inc. All rights reserved. | |
|
FULL TEXT00005-1/journalimage?img=PIIS1551741104000051.gr1.lrg.gif&fig=fig1&kwhquery=&issn=1551-7411&ishighres=false&allhighres=false&free=yes','journalimage');)
