The development of a financial toxicity patient‐reported outcome in cancer: The COST measure
Thoughts
Connects with: @prawitz2006 @voydanoff1990 @zafar2013
Annotations
desouza2014 - p. 3246
A literature search was followed by semistructured qualitative interviews with cancer patients that was designed to elicit themes on financial issues related to their disease and treatment
desouza2014 - p. 3246
Cancer experts (medical oncologists, oncology nurses, and mental health professionals) added items judged to be important.
desouza2014 - p. 3246
Patients were approached to evaluate the importance of the items.19 The items were presented with a 4-point Likert scale for the participants to rate importance,20 and items were ranked according to this importance score from highest to lowest.
desouza2014 - p. 3246
A quantitative item analysis was performed to evaluate the instrument’s internal consistency. Each participant completed the preliminary instrument, which was presented in a format parallel to that of the Functional Assessment of Cancer Therapy (FACT) quality-of-life instrument, with a 7-day time window and on a 5-point Likert scale from zero (“not at all”) to 5 (“a lot”). With the objective of minimizing redundancy, items were evaluated for their interitem correlation (IIC). Pairs of items with an IIC >0.7 were identified; and, within each correlated pair, the item with the highest importance score in step 2 was retained. Item-total correlations (ITCs) were calculated to ensure that the retained items impacted the instrument’s score. Items with nonsignificant ITCs (ie, not significantly related to the instrument total score; P > .05) were excluded. The Spearman rank correlation coefficient (rho) was used for all tests of association, because the item scores were ordinal rather than continuous.
desouza2014 - p. 3246
The number of final factors underlying the data was identified using the combination of a Cattell scree plot evaluation (number of factors on scree plot just before elbow), Kaiser criterion (eigenvalues >1.0), and percentage of variance criterion. Factor loadings greater than 0.5 were considered factorspecific and of statistical and practical significance
desouza2014 - p. 3249
The Kaiser-MeyerOklin value was 0.90, and the Bartlett test of sphericity reached statistical significance, supporting the factorability of the items One factor was clearly identified on inspection of the scree plot, with an eigenvalue of 5.68, explaining 89% of the variance. This factor was designated as financial toxicity. Among the 13 items, 11 had factor loadings greater than 0.5. The exceptions were item 45 (“My insurance does not provide adequate treatment coverage for my illness and care-related expenses”) and item 50 (“I rely on friends or family to help with the costs of health care”), which had factor loadings of 0.44 and 0.43, respectively.
desouza2014 - p. 3249
The Cronbach a coefficient for the 11-item COST measure was .9, indicating excellent internal consistency. The mean IIC of the 11-item COST-PROM was 0.47 (range, 0.22-0.69), and the mean ITC was 0.71 (range, 0.620.79), demonstrating nonredundancy and good construct validity