Purpose: The purpose of this study was to validate decision-specific knowledge items to be included in decision quality instruments (DQIs) for three common breast cancer decisions: surgery for early stage disease (SURG), adjuvant systemic treatment (SYS) and breast reconstruction (REC).
Method: We conducted a cross sectional mailing survey of recent breast cancer survivors identified through cancer registries at four academic cancer centers. Patients completed two modules (either SURG and SYS or SURG and REC). Each module contained items covering knowledge of treatments and outcomes, and demographics. A subset of patient respondents completed the survey twice to examine test-retest reliability. Providers and healthy volunteers also completed the knowledge items. To assess clinical validity, we also asked providers to rate the extent to which each item was essential for patients to know. First, we eliminated items that did not perform well. Next, we calculated test-retest reliability and internal consistency (using Cronbach’s Alpha) for the reduced set. Then, we used ANOVA to determine whether the items discriminated among providers, patients and healthy volunteers.
Result: We received 445 SURG, 354 SYS and 91 REC eligible responses (overall response rate 59%). The sample was predominantly Caucasian (82.5%), college-educated (63.6%), with Stage 1 disease (60.7%) and a mean age of 54.9. We eliminated 14 items across the three instruments, 7 for being too easy, 2 for being too hard and 5 for other problems. A majority of providers rated the items as essential for patients to know. The final SURG, REC, and SYS instruments had test-retest reliability of 0.72, 0 .75, and 0.57, respectively; and internal consistency of 0.74, 0.66, and 0.81. The instruments discriminated between higher levels of provider knowledge and lower levels of patient knowledge (88% vs. 52% SURG, 73% vs. 40% SYS, 60% vs. 37% REC, p<0.001 for each). Patients were more informed than healthy volunteers (52% vs. 49% SURG, 40% vs. 28% SYS, 37% vs. 33% REC), however the difference was only significant for SYS p<0.001.
Conclusion: The three instruments covered essential information, were able to discriminate between well informed providers and less informed patients, and were reliable. The knowledge portion of the DQIs demonstrated both psychometric and clinical validity.
Candidate for the Lee B. Lusted Student Prize Competition