The perceived meanings of cancer pain: An instrument development

Abstract

The purpose of this study was to develop a reliable and valid instrument that measures the perceived meanings of cancer pain for Taiwanese patients. A 46-item Perceived Meanings of Cancer Pain Inventory (PMCPI) which contains six scales was constructed for field testing. Two hundred cancer patients with pain were recruited from three teaching hospitals in Taiwan. The field testing of the PMCPI included three parts. An item analysis procedure using 100 randomly selected subjects was conducted to select items for the purpose of increasing the internal consistency and decreasing the item number in each scale. The selected items were cross validated with the remaining 100 subjects. A two-facet (item and occasion) generalizability study (G study) was conducted on selected items to estimate the variances associated with each facet and person. Twenty of the 200 subjects constituted the G study sample in which patients responded to study instruments twice. A series of confirmatory factor analyses with weighted least squares (WLS) estimations were performed on selected items to evaluate the construct validity for each scale and each pair combination of the scales. A total of 27 items was selected from the item analysis procedure. Cross validation shows that the fluctuation rate of alpha ranges from -9.24% to 16.25%. The alpha coefficients (n = 200) obtained from the selected items are.763,.783,.709,.556,.620, and.750 for Loss, Threat, Challenge, Blame-Other, Blame-Self and Spiritual-Awareness, respectively. The results of the G study indicate that the variances associated with "occasion" for each scale are near zero. However, the proportions of variances associated with "person-occasion interaction" were found to be 15.78% for the Threat scales. The proportion of variances associated with "item" is small for each scale except Loss (16.02%). The proportion of variances associated with "person-item interaction" ranges from 5.23% to 33.12%. The percentage of variance associated with "individual differences" ranges from 20.81% to 37.09%. The unspecified error variance which cannot be separated from "person-item-occasion interaction" accounts for 26.23 to 65.66% of total variance. The confirmatory factor analyses show that Loss, Challenge and Spiritual-Awareness scales have good model-data fit; the Threat scale also demonstrates reasonable fit after deleting one item. The measurement models of Blame-other and Blame-Self scales do not fit the data well enough. Six two-factor measurement models were formed using the four valid scales; all six models fit the data well. The four scales were tested for parallel items. The results show that the data are consistent with the parallel-item measurement models. Potential demographic and disease/treatment factors that may influence the four scales were identified.