Academic journal article Psychological Test and Assessment Modeling

Measurement Equivalence of the Patient Reported Outcomes Measurement Information System® (PROMIS®) Pain Interference Short Form Items: Application to Ethnically Diverse Cancer and Palliative Care Populations

Academic journal article Psychological Test and Assessment Modeling

Measurement Equivalence of the Patient Reported Outcomes Measurement Information System® (PROMIS®) Pain Interference Short Form Items: Application to Ethnically Diverse Cancer and Palliative Care Populations

Article excerpt

Background

Pain is estimated to affect between 2% and 55% of the population worldwide (Johannes, Le, Zhou, Johnston, & Dworkin, 2010); however, most estimates for chronic pain range from 22 % to 31 % (Breivik, Collett, Ventafridda, Cohen, & Gallacher, 2006; Johannes et al.; Moulin, Clark, Speechley, & Morley-Forster, 2002) and generally increase with age (Johannes et al.). Short standardized assessments of pain are desirable, given the burden of assessment, particularly in frail populations with chronic illness, those suffering from cancer, and those receiving palliative care. Short assessments are highly desirable for clinicians who frequently must assess for and manage multiple conditions in a given visit, particularly among older patients. The aim of the Patient Reported Outcome Measurement Information System® (PROMIS®) project was to provide computerized adaptive tests (CAT) and short form measures that can be used clinically and in research (Reeve et al., 2007). However, little information about measurement equivalence, particularly across ethnically diverse groups is available. The aim of this article is to examine the differential item functioning (DIF) and reliability of the PROMIS short forms across diverse socio-demographic groups using item response theory (IRT) methods.

There are as many as 11 response categories for some pain items, e.g., the Numeric Rating Scale (NRS; Cleeland, 1989). Question arises as to the validity and efficiency of numerous response categories, and some research using advanced psychometric analyses with IRT suggests a more parsimonious categorical representation of pain response (Chen, Revicki, Lai, Cook, & Amtmann, 2009). For example, Orlando-Edelen and Saliba (2010) showed that some of the 11 categories of the NRS items were overlapping and did not provide unique information. Other research (Waterman et al., 2010) confirmed the category overlap in measures with 11 response categories, and provided evidence that the response categories were not interval. Thus fewer categories may be more efficient. For example, Walton, Wideman, and Sullivan (2013) using a Rasch IRT approach found few instances of disordered threshold parameters in pain items with lower numbers (five) of response categories, suggesting a continuous interval level scale. The use of fewer response categories was the approach taken by PROMIS investigators who developed pain items with five response categories (Amtmann, et al., 2010; Cella et al., 2007).

Some efforts to develop and examine pain item banks have included analyses of DIF; e.g., back pain (Kopec et al., 2008) and pain interference in PROMIS (Amtmann et al., 2010). In a study by Amtmann et al. (2010), gender DIF was observed for the pain interference item, enjoyment of life. Age-related DIF was identified for eight items. Among the items examined by Amtmann and colleagues, only the item, ability to concentrate was included in the short form items examined in this paper. In the study by Kopec and colleagues (2008), DIF was observed for two items: pain expression and frustrated in an observational measure of pain suitable for patients with dementia (van Nispen tot Pannerden et al., 2009). Nurses had greater tendencies to report pain expression as present in residents with mild to moderate as contrasted with severe dementia; in contrast they were more likely to report "frustrated" among individuals with severe to very severe dementia as contrasted with milder dementia.

Methods

Sample

These data are from individuals with cancer who were selected from cancer registries in four regions of the United States. Details are provided in the overview article on the sample characteristics (Jensen et al., 2016). The overall sample sizes were 1,053 Hispanics, 917 Asians/Pacific Islanders, 1,122 Blacks and 2,278 non-Hispanic Whites. The focal group was males (n = 2,192) in the analyses of gender; there were 3,247 females. In the analyses of education, the reference group was graduate degree (n = 644). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.