Linking pain items from two studies onto a common scale using item response theory.

Area of Science:

• Psychometrics
• Health Outcomes Research
• Data Science

Background:

• Patient-reported outcome surveys often face limitations in sample size and item count due to recruitment challenges and patient burden.
• These limitations can result in insufficient statistical power or a restricted scope of measurement.
• Item response theory (IRT) methodology offers a way to pool data from multiple surveys, thereby increasing statistical power and expanding measurement scope.

Purpose of the Study:

• To compare two approaches for linking items from different pain surveys into a unified item bank with a common measurement scale.
• To evaluate the stability and accuracy of item parameters generated by each linking approach.
• To provide recommendations for developing robust item banks using IRT.

Main Methods:

• Secondary analysis of data from two independent pain surveys: the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMPAACT) Survey and the Center on Outcomes, Research and Education (CORE) Survey.
• Two item linking approaches were employed: simultaneous calibration of all items to a single IRT model, and separate calibration followed by scale transformation to a common metric.
• Item response theory (IRT) models were used to calibrate and link items across datasets.

Main Results:

• Both linking approaches yielded similar results for pain interference items, attributed to a sufficient number of common items and adequate sample sizes.
• Simultaneous calibration demonstrated more stable results for pain intensity items.
• Separate calibration resulted in an unsatisfactory linking for pain intensity due to a single common item and a small sample size.

Conclusions:

• Simultaneous IRT calibration is the preferred method for developing comprehensive item banks due to its ability to produce more stable item parameters across independent samples.
• This approach enhances the pooling of data from patient-reported outcome surveys, overcoming limitations of small sample sizes and limited item scope.
• The findings support the use of IRT for creating robust and powerful measurement instruments in health research.