Using prior information to individualize start item selection when assessing physical functioning with the EORTC CAT...

Area of Science:

• Health Outcomes Research
• Psychometrics
• Cancer Care Quality Measurement

Background:

• Computerized adaptive testing (CAT) offers personalized assessments by selecting items based on prior responses.
• The initial item (start item) in CAT is typically fixed due to a lack of a priori score estimates.
• The European Organisation for Research and Treatment of Cancer (EORTC) CAT Core assesses 15 quality of life domains.

Purpose of the Study:

• To evaluate the use of cross-domain information for individualizing the start item selection in the EORTC CAT Core physical functioning module.
• To assess the impact of individualized start items on measurement precision compared to fixed start items.

Main Methods:

• Linear regression models predicted physical functioning (PF) scores from 14 other EORTC CAT Core domains using data from 10,084 cancer patient assessments.
• Monte Carlo simulations of CAT were performed to compare measurement precision with individualized versus fixed start items.

Main Results:

• Cross-domain predictions showed correlations between 0.25-0.71, with 57-85% of predicted PF scores within 1 standard deviation of observed scores.
• Individually selected start items enhanced CAT measurement precision in the initial stages.
• The benefit of individualized start items diminished significantly after three or more items were administered.

Conclusions:

• Linear regression can effectively generate cross-domain predictions for CAT start item selection.
• Individualized start items can improve the measurement precision of the EORTC CAT Core, particularly for short test administrations.