2011 HSR&D National Meeting Abstract
3057 — Validating Use of a Symptom Assessment Scale in Palliative Care
Livote EE (James J Peters VA Medical Center), Verkuilen J
(City University of New York Graduate Center), Chang V
(New Jersey Health Care System), Penrod J
(James J Peters VA Medical Center), Cortez T
(New York Harbor Health Care System), Luhrs C
(New York Harbor Health Care System)
The Condensed Memorial Symptom Assessment Scale (CMSAS), developed to measure the presence and severity of 14 symptoms prevalent in patients with life-limiting diseases, is used by the VISN 3 Palliative Care program to guide care and measure program outcomes. Though considered appropriate for use in palliative care, the properties of the scale have not been examined in actual use, where incomplete and inconsistent administration could threaten the validity of the scale to measure and compare outcomes. We performed a psychometric analysis of the scale, including an analysis of missing data, based on the operational responses and investigated whether the measurement properties were equivalent to the sample in which it was developed.
The development sample included 924 veterans with cancer who had participated in symptom-related research studies. The operational sample included 886 initial assessments of VA palliative care patients between FY2006 and FY2009. Substantial missing data were found in the operational sample, especially the severity component. Therefore, only the binary (symptom presence) part of the response was used. Because the scale includes physical and psychological symptoms, we posited a multidimensional model with one primary general factor (symptom distress) and two secondary item-group factors (physical and psychological), known as the bifactor model. Exploratory and confirmatory factor analysis (CFA) was used to fit IRT models to each sample. Multiple-group CFA and MIMC (multiple-indicator, multiple-cause) modeling was used to assess measurement invariance.
The posited bifactor model provided good fit in both samples indicating good reliability and factor pattern invariance across development and operational contexts. The results of both methods indicated partial invariance after identifying a set of items exhibiting differential item functioning (DIF). Symptoms found to lack invariance included weight loss, concentration, appetite, and sleep.
The psychometric properties of the CMSAS found in development held up well in actual use in a palliative care program.
These findings provide validity evidence that the scale generalizes to palliative care but substantial missing data is a barrier to its use to measure program outcomes. Remedies to be explored are easier administration and making the scale as a whole (versus a set of stand-alone items) more useful to clinicians.