PropertyValue
is nif:broaderContext of
nif:broaderContext
is schema:hasPart of
schema:isPartOf
nif:isString
  • Participants in this study comprised 20 patients, 10 physicians, and 10 non-medical professionals. The patient group consisted of 20 men, mean age 67 years (SD = 7), newly diagnosed (i.e., within the preceding 4 weeks) with clinically localized PCa, staging T1-2N0M0. Patients had not yet started any treatment, had no significant comorbidities, and had a life expectancy greater than 10 years. Physicians were 10 urologists (2 female and 8 male), mean age 50 years (SD = 9), working at the urology units of two main clinical facilities in northeast Italy. Finally, the non-medical professional group consisted of 10 architects (2 female and 8 male), mean age 45 years (SD = 8). The study was approved by the research ethics committee of the APSS (the regional healthcare provider within the Provincia Autonoma di Trento, Italy). Written informed consent was obtained from all individual participants included in the study. All procedures were in accordance with the Helsinki declaration. In accordance with the guidelines of Multiple Criteria Decision Analysis for Health Care [46], we generated our list of attributes so as to satisfy the following desiderata: to capture the complexity of the available clinically localized PCa treatments (completeness), to include only relevant and no double-counted attributes for the purposes of comparing existing PCa treatments (non-redundancy and non-overlapping), and to ensure that the values of one attribute would not affect, nor would they depend on, the values of other attributes (mutual preferential independence, see also [47,48]). This last is a fundamental condition of multi-attribute utility theory that prevents the interaction between attributes and allows the additive decomposition of the utility function. Its fulfillment was made easier by our focus on treatments rather than health states, since the values of treatments (e.g., their expected effectiveness or side effects) are typically fixed and independent from each other, while the same does not hold for most attributes expressing how the patient feels (e.g., mood or pain, as used in [44,45]). Overall, our analysis generated eight attributes pertaining to four macro-dimensions, as described in Table 1. Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0200780.t001 Attributes of the treatments. Each attribute was accompanied by a short description that included a specification of its worst and best levels (see S1 Appendix). This was motivated by the consideration that importance weights depend on the ranges of attributes [49]. Such clarification also helped us make the attributes more understandable to patients and architects, who might otherwise be unfamiliar with this decision problem. It also indirectly confirmed that our list of attributes made sense to physicians, given that none of them raised any concerns. Attributes were presented to participants in two different orders (a randomly chosen one and its reverse). The attributes were presented to each patient-urologist/patient-architect pair in the same order. Patients were assigned to urologists by the hospital administration based on the timing of the outcome of the biopsy and on the urologists’ shift schedule. Data were collected only after at least one meeting between urologist and patient, during which the urologist had informed the patient about the positive biopsy result, discussed with him the treatment options, and provided him with his/her own treatment advice. Since physicians were interviewed only after they had already advised a patient with a specific treatment, it seems reasonable to assume that they should have been aware of the patient’s importance weights. To minimize possible memory interference, each half of the urologist-patient pair was interviewed within three days of the other, without any other encounter taking place between the two in the meantime. Even within such a small time window, whenever a urologist raised a concern about her/his memory of a specific patient whose importance weights s/he had been asked to estimate, we did not force her/him and simply did not include that pair in our data collection. This reduced the number of participants, but it also helped us to avoid interference with normal clinical activity. In particular, it attenuated the risk that urologists would alter their patterns of practice with patients due to their anticipation of our questionnaire. Each architect was matched for age and gender to a unique physician and was paired with the same patient(s). Architects never met the patients; they were told only a patient’s age and that he had received a recent diagnosis of clinically localized PCa. After being introduced to the eight attributes, participants were asked to provide their judgments, that is, importance weights for patients and estimates of patients’ importance weights for urologists and architects. More specifically, participants in all three groups (patients, urologists, and architects) were asked to rate each attribute independently on a scale ranging from 0 (= “not important at all”) to 100 (= “extremely important”). This elicitation method is known as Direct Rating (hereinafter DR, [50]) and is a simple technique with relatively low demands on time. For patients only, we also employed a version of the Hierarchical Point Allocation technique known as Value Hierarchy (hereinafter VH, [51]), in which the attributes are grouped into fewer macro-dimensions (in our case, the eight attributes were arranged in pairs into four macro-dimensions, see Table 1); participants have to distribute 100 importance points among the macro-dimensions (in our case, four, as in Table 1) and then, they have to further distribute 100 importance points between the attributes within each macro-dimension (two in our case). Importance weights are calculated by aggregating the two judgments and rescaling the result (i.e., by multiplying the within- and between-dimension point scores obtained for each attribute, and dividing the result by 100). Given that hierarchical techniques involve a number of pairwise comparisons, they are supposed to reflect people’s propensity for relative judgments, as well as to help them organize complex goal structures into hierarchical clusters [52]. Patients’ weights were elicited first with DR and then with VH. We decided to use only DR with urologists because they were already familiar with the structure of the attributes under consideration and, furthermore, because it was not realistic for them to complete a lengthy interview during their working hours. On the other hand, the use of two different techniques with patients allowed us to assess the consistency of results. Indeed, although both methods yield a cardinal scale of importance, DR reportedly yields a lower spread of weights than VH [49]. We performed descriptive statistics first, in order to summarize patients’ importance weights for each attribute. The general agreement between patients’ importance weights and their estimates as provided by urologists and architects was assessed by computing Kendall rank-order correlations. To determine whether these correlations exceeded the critical value of .8, one-sample t-tests were performed. Paired t-tests were used to determine if the agreement with patients’ importance weights differed significantly between urologists and architects. To quantify the evidence in support of the null hypothesis in these two comparisons, we computed the corresponding Bayes factors (using JASP 0.7.5.6; www.jasp-stats.org). We also converted participants’ judgements into ranks and, for each attribute, determined by means of a binomial test whether urologists’ and architects’ ranks in agreement with patients’ ranks significantly exceeded those in disagreement. Finally, we assessed by Kendall rank-order correlations the within-subject agreement in patients’ weights elicited using the two DR and VH methods.
rdf:type