An approach to measuring the quality of breast cancer decisions

Abstract

Objective

To explore an approach to measuring the quality of decisions made in the treatment of early stage breast cancer, focusing on patients’ decision-specific knowledge and the concordance between patients’ stated preferences for treatment outcomes and treatment received.

Methods

Candidate knowledge and value items were identified after an extensive review of the published literature as well as reports on 27 focus groups and 46 individual interviews with breast cancer survivors. Items were subjected to cognitive interviews with six additional patients. A preliminary decision quality measure consisting of five knowledge items and four value items was pilot tested with 35 breast cancer survivors who also completed the control preferences scale and the decisional conflict scale (DCS).

Results

Preference for control and knowledge did not vary by treatment. The mean of the participants’ knowledge scores was 54%. There was no correlation between the knowledge scores and the informed subscale of the DCS (Pearson r = .152, n = 32, p = 0.408). Patients who preferred to keep their breast were over five times as likely to have breast-conserving surgery than those who did not (OR 5.33, 95% CI (1.2, 24.5), p = 0.06). Patients who wanted to avoid radiation were six times as likely to choose mastectomy than those who did not (OR 6.4, 95% CI (1.34, 30.61), p = 0.04).

Conclusion

Measuring decision quality by assessing patients’ decision-specific knowledge and concordance between their values and treatment received, is feasible and important. Further work is necessary to overcome the methodological challenges identified in this pilot work.

Practice implications

Guidelines for early stage breast cancer emphasize the importance of including patients’ preferences in decisions about treatment. The ability of doctors and patients to make decisions that reflect the considered preferences of well-informed patients can and should be measured.

Introduction

Perhaps nowhere in medicine are the challenges of decision making more apparent than in breast cancer. For the decisions faced by newly diagnosed patients – treatment of ductal carcinoma-in situ, surgery for early stage breast cancer, breast reconstruction and adjuvant therapy – often no one choice clearly dominates. For early stage invasive breast cancer, the equivalence of mastectomy and lumpectomy followed by radiation for overall survival and distant disease free survival has been demonstrated in large randomized controlled trials with long follow-up [1], [2]. Although tumor size, location and other factors may limit the use of breast conserving surgery, studies show that the majority of women are eligible for breast conserving surgery [3]. However, clinical studies also found an increased risk of ipsilateral breast recurrence after lumpectomy and lumpectomy with radiation—so keeping the breast may make it more likely that these women will face another decision about treatment of breast cancer [4]. In addition, keeping the breast also requires more treatment, as breast conserving therapy may require multiple excisions to get clear margins and is generally followed by up to 7 weeks of radiation treatments.

Simple guidelines or other heuristics employed to direct clinical decisions often make assumptions about the uniformity of patients’ values for these and other outcomes of treatment. For example, the NIH Consensus conference of 1990 concluded that, “breast conservation treatment… is preferable because it provides survival equivalent to total mastectomy… while preserving the breast” [5]. However, this recommendation ignored the diversity of patients’ preferences for keeping their breast. A recent study found that increased participation by patients was associated with increased use of mastectomy, suggesting that the assumption that breast conservation is preferable does not hold across all patients [6]. In more recent guidelines for local treatment of early stage breast cancer, this shortfall has been addressed with specific language highlighting the importance of including patient's preferences in the choice of surgical treatment [7].

These types of decisions have been called, “preference-sensitive decisions” to reflect the fact that the medical evidence is necessary, but not sufficient. Patient's personal preferences are also necessary to make an appropriate decision [8]. It follows that the quality of a preference-sensitive clinical decision can been defined as the extent to which the implemented decision reflects the considered preferences of a well-informed patient. This definition, with minor variations, has been advocated by several leaders in the field [9], [10], [11], [12], [13], [14], [15], [16]. The International Patient Decision Aids Standards (or IPDAS) Collaboration consensus process found very strong support for decision quality, as defined above, as a key indicator of effectiveness of decision aids [17].

Most of the research focused on measuring the quality of decisions has been associated with clinical trials of patient decision aids and other decision support tools. A recent review of measures highlights the lack of agreement in the field about how to best measure the effectiveness and the impact of decision aids on quality [12]. The lack of agreement on measures of effectiveness is not due to the wide variety of diseases covered, as it is just as prevalent within the breast cancer literature. For example, in the twelve published trials of breast cancer decision support interventions contained in the 2003 Cochrane Systematic Review, researchers used 16 different instruments to evaluate the quality of decision making and involvement in decisions, 4 different knowledge tests, 10 different satisfaction questionnaires, and 12 different measures of stress, distress or anxiety [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29].

As detailed in an article by Sepucha et al. [16], there are problems with using the currently available measures to assess decision quality. For example, satisfaction measures are often misleading, as high satisfaction scores are more likely to be the result of low expectations than a high quality decision-making process [30]. Another common outcome, patient's desire for participation, asks in various ways whether the patient wants to lead, share or defer decisions to the doctor. However, cognitive testing of this item has revealed that patients’ responses are often not consistent with their reports of the interactions. For example, patients who report that they shared decision making, when asked to describe what happened, often say that they doctor told them what to do and they said ok—not a particularly shared process [31]. Finally, the Decisional Conflict Scale developed by Annette O’Connor asks patients to report on their state with items such as “I understand the pros of each choice” and “I made an informed decision” [14]. This scale has widespread application across conditions and situations; however, the responses are limited by the patient's perspective, as people cannot reliably report on what they do not know. None of the currently available measures adequately capture the quality of a decision as defined above.

As a result, two of the authors (Drs. Sepucha and Mulley) and colleagues have proposed developing a decision quality instrument that assesses (1) decision-specific knowledge, (2) decision-specific values and (3) treatment undergone [16]. The decision-specific knowledge will assess whether or not a patient was adequately informed. The strength of association between the stated values and treatment undergone will provide a measure of value concordance, that is, the tendency for treatment to be influenced by patients’ preferences. The calculation of value concordance follows the approach used by Barry et al. [32] where the values are included in a multivariate regression model to determine the amount of variance in the decisions that can be predicted by patients’ reports of their utilities for key outcomes. Other factors that may influence treatment decisions, such as physician recommendations, medical circumstances, social context and insurance status, can be included in the regression model in order to examine the relative influence of patient preferences. This measure of value concordance enables comparisons across populations or groups of patients, but does not provide a measure the quality of an individual decision. Populations with higher quality decisions should have higher knowledge and a larger portion of the variation in the model should be explained by patient preferences compared to other factors [32].

Here, the investigators report on efforts to identify salient knowledge and value items and pilot test them to determine (1) whether patients could adequately self report decision quality as required by the framework, (2) how the decision-specific knowledge and values compared to the generic assessment in the decisional conflict scale, and (3) describe knowledge and values in a cross-sectional sample of survivors.