Alexander Avian1, Brigitte Messerer2, Andreas Frey3, Winfried Meissner4, Annelie Weinberg5, William Ravekes6, Andrea Berghold7. 1. Institute for Medical Informatics, Statistics and Documentation, Medical University of Graz, Auenbruggerplatz 2, 8036 Graz, Austria; Department of Anesthesiology and Intensive Care, Jena University Hospital, Jena, Lobeda, Germany; Department of Research Methods in Education, Friedrich Schiller University Jena, Am Planetarium 4, 07743 Jena, Germany. Electronic address: alexander.avian@medunigraz.at. 2. Division of Anesthesiology for Cardiovascular Surgery and Intensive Care Medicine, Medical University of Graz, Auenbruggerplatz 29, 8036 Graz, Austria. Electronic address: brigitte.messerer@medunigraz.at. 3. Department of Research Methods in Education, Friedrich Schiller University Jena, Am Planetarium 4, 07743 Jena, Germany; Centre for Educational Measurement (CEMO) at the University of Oslo, Postboks 1161 Blindern, 0318 Oslo, Norway. Electronic address: andreas.frey@uni-jena.de. 4. Department of Anesthesiology and Intensive Care, Jena University Hospital, Jena, Lobeda, Germany. Electronic address: Winfried.meissner@med.uni-jena.de. 5. Department of Orthopedics and Orthopedic Surgery, Medical University of Graz, Auenbruggerplatz 5, 8036 Graz, Austria. Electronic address: Annelie.Weinberg@t-online.de. 6. Division of Pediatric Cardiology, Johns Hopkins University School of Medicine, Baltimore, MD, USA. Electronic address: wravekes@jhmi.edu. 7. Institute for Medical Informatics, Statistics and Documentation, Medical University of Graz, Auenbruggerplatz 2, 8036 Graz, Austria. Electronic address: andrea.berghold@medunigraz.at.
Abstract
BACKGROUND: The Faces Pain Scale-revised (FPS-r) has been developed as an interval scale. For other pain measurement instruments, several studies found evidence for and against an interval level of measurement. OBJECTIVES: The primary aim of the current study was to evaluate the scale properties of the FPS-r using an item response theory approach. DESIGN: Secondary analysis of published data. SETTING: Three studies; Study 1 and study 2: One university hospital; Study 3: international pain registry. PARTICIPANTS: Study 1: n = 246, female: 41%, age: 11-18 years, 3 pain items; Study 2: n = 240, female: 43%, age: 11-18 years, 9 pain items; Study 3: n = 2266, female: 41%, age: 4-18 years, 3 pain items. METHODS: The rating scale model (interval scale), the graded response model (no interval scale, ordered response categories) and the partial credit model (no interval scale) were used to scale the data. RESULTS: In all three studies, the rating scale model was outperformed by the graded response model or the partial credit model in terms of model fit. Overlapping response categories were found in items associated with less pain. Response category widths were wider for categories associated with low pain intensity and smaller for categories associated with high pain intensities. Smallest response categories were 1%-67% smaller compared to the widest response category of the same item. CONCLUSION: According to these findings, the interval scale properties of the FPS-r may be questioned. Item response theory methods may help to solve the problem of missing linearity in pain intensity ratings using FPS-r.
BACKGROUND: The Faces Pain Scale-revised (FPS-r) has been developed as an interval scale. For other pain measurement instruments, several studies found evidence for and against an interval level of measurement. OBJECTIVES: The primary aim of the current study was to evaluate the scale properties of the FPS-r using an item response theory approach. DESIGN: Secondary analysis of published data. SETTING: Three studies; Study 1 and study 2: One university hospital; Study 3: international pain registry. PARTICIPANTS: Study 1: n = 246, female: 41%, age: 11-18 years, 3 pain items; Study 2: n = 240, female: 43%, age: 11-18 years, 9 pain items; Study 3: n = 2266, female: 41%, age: 4-18 years, 3 pain items. METHODS: The rating scale model (interval scale), the graded response model (no interval scale, ordered response categories) and the partial credit model (no interval scale) were used to scale the data. RESULTS: In all three studies, the rating scale model was outperformed by the graded response model or the partial credit model in terms of model fit. Overlapping response categories were found in items associated with less pain. Response category widths were wider for categories associated with low pain intensity and smaller for categories associated with high pain intensities. Smallest response categories were 1%-67% smaller compared to the widest response category of the same item. CONCLUSION: According to these findings, the interval scale properties of the FPS-r may be questioned. Item response theory methods may help to solve the problem of missing linearity in pain intensity ratings using FPS-r.