D. Feldman-Stewart, Radiation Oncology Research Unit, Apps Level 4, Kingston General Hospital, Kingston, Ontario, K7L 2V7, Canada.
Objective To determine the information that men diagnosed with early-stage prostate cancer think is necessary to choose their treatment (surgery, radiotherapy or ‘watchful waiting’).
Patients and methods All men diagnosed with early-stage prostate cancer in a regional urology practice in an 18-month period were surveyed. Respondents considered the treatment decision for a specific hypothetical case and judged whether each of 59 specific information items would be ‘necessary’ to know to make the treatment decision. Respondents provided basic demographic information (age, education and marital status).
Results Of 89 patients, 71 (80%) responded; the mean (sd, range) number of items that individual respondents identified as necessary was 32 (12, 9–58). Each item was considered necessary by 54 (18, 20–93)% of respondents. Of the 59 items, only five were necessary to 80% of respondents; 23 were necessary to 67% of respondents and five were not necessary to 67% of respondents. Patients were divided about the remaining 31 items. There were no strong correlations (none accounted for > 20% of the variance) between the responses and demographic characteristics.
Conclusions There appears to be large variation among patients with early-stage prostate cancer in the number of information items deemed necessary to make a treatment decision, and little agreement on the need for most individual items. Demographic characteristics were not useful as predictors of the information needs of each patient. The process of informing patients with early-stage prostate cancer (and their decision aids) needs to be flexible and able to accommodate a wide range of patients' information needs.
Decision aids are designed to facilitate the participation of patients in decisions about their healthcare, e.g. for screening or treatment. Although aids vary in their fundamental approaches, they all provide information . We are unaware of any decision aids that provide information identified empirically. The present study was designed to determine from patients with early-stage prostate cancer what information should be incorporated into a decision aid designed to assist them with their treatment decisions.
For early-stage disease, patients usually have three treatment options, i.e. watchful waiting, surgery or radiation. Although there have been no controlled clinical trials providing unequivocal evidence, it is thought that the active treatments confer some survival benefit  and that the side-effects differ for the two interventional treatments. Thus, there is a great deal of information related to the options that might be important to the treatment decision. However, because people have a finite amount of energy for cognitive processing , processing irrelevant information will necessarily reduce the amount of energy available for decision making. Thus, we wished to limit the information provided in the aid to that which is important to the decision, and thus limit unnecessary cognitive demands.
Although there is an earlier study of the information that patients with prostate cancer need for decision-making  it was unclear how to use those results as a basis for our decision aid. That study included patients with all stages of prostate cancer, which makes the needs of the early-stage patients difficult to identify. In addition, the study identified categories of information that were important to the decisions, rather than the details that are most relevant.
The goal of the present study was to determine the extent of agreement among patients on what information is considered important to the decision. If there was ‘adequate’ agreement, we would create an objective standard of disclosure  and use that as the information provided in the decision aid. If there was no ‘adequate’ agreement, we would determine if we could predict information needs of subgroups based on demographic characteristics which would allow us to tailor the contents of the aid to specific patient profiles reliably. Finally, with no adequate agreement and with only weak predictors of patient needs, we would conclude that the only viable standard would be a subjective standard of disclosure, and would need to focus the design of the decision aid on accommodating the needs of individual patients.
Patients and methods
Participants were recruited from a university urology department practice of six urologists providing secondary and tertiary care. Patients were eligible if their biopsy was taken in Kingston during an 18-month period that ended 6 months before the survey, if they had stage T1N0 or T2N0 disease, and if their attending urologist agreed to send a letter of invitation.
The data were gathered by a survey, which included a scenario describing the case of a 65-year-old otherwise healthy man presenting with moderately well-differentiated stage T1cN0 prostate cancer, who was offered three treatment options; watchful waiting, radiation therapy or surgery . Respondents were asked to imagine that they were in the same situation as the man in the scenario. The scenario was used to reduce the potential for variation in responses arising from considering different case scenarios.
The survey presented 59 items of information which were derived from our earlier study addressing the general information needs of patients with early-stage prostate cancer . Items were selected from the earlier study if they had been identified as ‘essential’ for decision-making by least 10% of the respondents. An ‘answer’ was developed for each of the 59 questions by three radiation oncologists and a urologist who agreed, through a consensus-building process, on the final answers. Data for the answers (e.g. the probability of developing symptomatic disease) were taken from a structured literature review and decision analysis  when possible, and otherwise they were the consensus of the four physicians. To make the information as easy to understand as possible, all probabilities were presented as frequencies (out of 100) .
The survey questionnaire was long and to help respondents with this the items were organized into conceptual categories which were identified at the beginning of the survey . The categories were then presented in a logical order. Recognizing that using a single order would create an order effect that would in turn limit the interpretability of the results , the items were organized within each category into subgroups and the order of items manipulated within a subgroup independently for each survey.
Respondents were instructed to indicate, for each item, which of three categories was most appropriate: ‘necessary for the decision’, ‘not necessary for the decision’, or ‘not sure’. To help the respondent think about the hypothetical case, the survey was initiated by presenting examples of people buying cars. Three buyers were presented, each with a unique set of values, to show that values important to one buyer were not necessarily important to another. The examples were also designed to clarify the structure of questions in the survey.
After patients completed the survey, they provided information about their age, highest level of formal education, marital status, the treatments they were offered at the time of their diagnosis and the treatment they received. Responses were returned by mail. The project was approved by the Queen's University Health Sciences & Affiliated Teaching Hospitals' Research Ethics Review Board.
For each item, three percentages were calculated; the percentage of respondents who judged the item as necessary, as unnecessary and who were unsure. Agreement was assessed in two ways. First, the overall agreement was evaluated using the kappa statistic . Second, agreement on individual items was assessed by calculating an index of agreement  (the proportion of the group who agree with the majority view about whether the item is necessary or not). The index ranges from 100 when there is full agreement (either that the item is necessary or not) to 50 when there is maximum disagreement.
To determine associations between demographic characteristics and responses, two sets of analyses were conducted, one aimed at predicting the number of necessary items for particular respondents and one aimed at predicting responses to each individual item. The Pearson correlation coefficient was used to assess the association between the number of necessary items and age, and the correlation coefficient ε to evaluate its association with education and with marital status. Logistic regression was used to assess associations between demographic characteristics and responses to individual items. The amount of variance accounted for by the factors was also calculated (R) .
Of the 89 men eligible for the study, 71 (80%) responded; their mean (sd) age was 65.7 (6.6) years and 59 (88%) were married/living with a partner. Of the 66 who reported their education, 12 (18%) had not completed high school, 30 (45%) had completed high school and 24 (36%) had some tertiary education. Sixty-seven respondents reported the treatments that they were offered; 17 (25%) reported being offered only a single treatment option, 26 (39%) two options and 24 (36%) all three options. Of the offers, 54 (81%) reported being offered surgery, 52 (78%) radiotherapy and 31 (46%) watchful waiting. Sixty-seven also reported which treatment they had received; 35 (52%) were treated with surgery, 21 (31%) with radiotherapy and 10 (15%) with watchful waiting.
Figure 1 shows, for each of the 59 items, the percentages of respondents who judged it as ‘necessary’, ‘not necessary’, and ‘not sure’. Table 1 provides a brief description of each of the items in the order of Fig. 1; the items are ordered according to the percentage of ‘necessary’ responses. Table 2 presents the top three items as they were actually presented. Two respondents made it clear that they did not want to participate in the decision, and thus did not classify any of the items. Each item was deemed ‘necessary’ by a mean (sd) of 56 (18.3)%, ‘not necessary’ by 40 (18.2)%, with 4 (2.4)% of respondents ‘unsure’. Figure 1 shows that for most items > 90% of respondents had an opinion about whether the item was necessary or not.
Table 1. Information items necessary for the decision (* items showing ‘agreement’) †
Abbreviated labels are used to identify the items and they are ordered (descending) by the percentage of respondents who thought the item was necessary.
Table 2. Actual information items provided for the top three items
Treatment options if initial treatment unsuccessful
If surgery does not remove all of the cancer, some patients can then have radiation. There are more side-effects from radiation when it is given after an operation, and your doctor would discuss these with you. Hormone treatment can also be used if the surgery does not remove all of the cancer. If the cancer does not disappear after radiation, some patients can then have surgery. There are more side-effects from surgery when it is used after radiation, and your doctor would discuss these with you. Hormone treatment can also be used if the cancer does not disappear after radiation.
Treatment options if cancer progresses
If you choose no treatment for now, you might still be able to have either surgery or radiation later, as long as the cancer has not spread beyond the prostate. If the cancer does spread beyond the prostate, there is still very effective hormone treatment that can keep it under control for many months or years.
Chances of cancer disappearing, at least initially
The cancer does not disappear on its own in patients who choose no treatment for now. Surgery removes all detectable cancer cells in about 80 of 100 patients. In some patients, the cancer comes back even if it seems to be removed completely with surgery. Radiation makes the cancer disappear in about 80 patients of 100. In some patients, the cancer comes back even it seems to disappear completely with radiation.
Although agreement overall was statistically significant, the Kappa statistic (0.11, P < 0.001) indicated poor agreement beyond that expected by chance alone . The index of agreement for each of the 59 items is shown in Fig. 2. The items are ordered by the percentage of ‘necessary’ responses of those who had an opinion. Using an arbitrary threshold of 67% to define ‘sufficient’ agreement, Fig. 2 shows that respondents agreed that 23 of the items were necessary for the decision and that five items were unnecessary. The items on which there was agreement are marked in Table 1 (as *); items at the beginning of the table are those that were necessary for the decision, while the items asterisked at the end of the table are those that were unnecessary. Fig. 2 also shows that there was disagreement on the remaining 31 items (over half).
Because information must be provided to individual patients, the variability in the number of items deemed necessary was also determined. Figure 3 shows the number of items deemed necessary by each respondent; while 10% felt that <16 items were necessary, another 10% felt that > 50 items were necessary. Figure 3 shows a gradually increasing function between the extremes. Except for the two men who indicated that no information was necessary, each respondent identified at least one item as necessary beyond the 23 items that the whole group agreed on as being necessary.
To determine whether clinicians could anticipate the information needs of particular patients, the associations between the responses and the respondents' demographic characteristics was calculated. Overall, there was no association between the number of items that respondents deemed necessary and their age (r= 0.009, n = 66), their education (ε = 0.128) or their partner status (ε = 0.149).
Although there were statistically significant associations between the responses and particular demographic characteristics for 19 of the 59 items in univariate analyses, only five of the items had associations that accounted for > 10% of the variance, of which none accounted for > 25% of the variance.
These results suggest that there is a wide variation both in the amount of information and in the particular details that patients with early-stage prostate cancer think is necessary to make their treatment decisions. The extent of the observed variation implies that the subjective standard is the most appropriate legal standard for judging the adequacy of information provided for decision making, and that our decision aid must be designed to accommodate that variation. That we identified a subset of 23 items that at least 67% of the respondents agreed were necessary implies that a set of items can be defined that could be regarded as ‘core’ information, i.e. information was identified that could be provided to all patients because most would want it. However, the core could not be considered an objective standard because each of the respondents also identified items beyond the core as being necessary for their decisions. The core would not cover all information that is material for any of the decision-making respondents.
An alternative to providing individually tailored information is to present all of the information that is material to all patients. However, the present results clearly indicate that such a strategy would result in burdening patients with too much information that is not material to their decisions. The burden is of concern because it reduces the amount of energy that the patient has to process the information which is material to them. A strategy of providing a core that most patients want, then supplementing it with the needs of the individual, has greater potential to be an efficient and effective strategy for ensuring that patients not only receive but also understand the information that is material to their decisions.
The present respondents considered the treatment decision for a hypothetical patient, so that they would consider the same medical situation. To the extent that the study was successful, the resulting variation in information priorities is probably caused more by differences in patients' values than differences in the medical situations. Considering the variety of medical situations included in the population of patients with early-stage prostate cancer, clinical practice would be expected to have to accommodate an even greater variation in information needed for treatment decision-making than apparent in the present study.
That there were no demographic factors that were highly correlated with responses replicates the results of our earlier study designed to identify the questions that patients want answered . Indeed, few of the correlations in either study accounted for >10% of the variance. The implication of these findings is that clinicians cannot expect to predict the information needs of any individual patient. The patient must be asked directly about the information that is material to his decision.
The present results also confirm our earlier finding that patients do not consider all information included in a typical category to be equally important to their decisions. Research designed to determine the information that patients need for decision making therefore should focus on the information details rather than on categories of information. The methodological concern extends to identifying the information that patients consider important for purposes other than decision making .
The present high response rate implies that there was a fairly good representation of the attitudes of patients with early-stage prostate cancer, and who were treated in the period and geographical area surveyed. It is unlikely that the present extent of variation in information needs would be unique to our area. Increasing the sample size would only add to the statistical power to detect even smaller associations and thus would not increase the potential to improve a clinicians' ability to predict the needs for the individual patient.
We conclude that the process of providing information to patients with early-stage prostate cancer must be flexible, both in the amount of information that is provided and in which particular items are provided. Creating a decision aid to meet such conditions is a challenge. Nevertheless, our findings suggest that a patient-guided process is necessary to meet the goals of informed decision making.
We thank Beth McConnell for her assistance in administering the survey and Karleen Schulze for her assistance in the statistical analyses. Deb Feldman-Stewart is supported by a Ministry of Health Career Scientist Award and the project was supported by National Cancer Institute of Canada and the Canadian Cancer Society. Part of the results were presented at the 19th Annual Meeting of the Society of Medical Decision Making, October 1998.