Check your bookshelf. Peruse the measurement chapters in your research methods collection. Chances are you will find guidelines for operationalizing psychosocial and biobehavior concepts and symptom experiences central to nursing research. We use these sources reflexively to evaluate and to report reliability estimates and validity evidence associated with scores from many instruments. Trouble is, these pragmatic approaches reflect mid-20th Century theory and practice of psychometrics. Most guidelines are work-arounds to well-known limitations of classical test theory (CTT; single reliability estimates, equal standard errors of measurement, etc.). With the ascent of item response theory (IRT) as the basis for solving practical measurement problems (Borsboom, 2006), the CTT-based habits are obsolete. Clinging to traditional measurement practices is creating a quaint patchwork of results in nursing research journals and is blocking the accumulation of knowledge we need.
The Patient Reported Outcome Measurement Information System (PROMIS) "aims to revolutionize the way patient-reported outcome tools are selected and employed in clinical research and practice evaluation" (PROMIS, n.d.). The scope of the initiative is sweeping, and accomplishments over the course of the first funding cycle are breathtaking. Teams of U.S. psychometricians and health scientists joined together to create a domain framework that incorporates global health and physical, mental, and social health and quality of life. Legacy instruments and published research were scoured for domain-relevant items; the resulting items were classified, catalogued, and subjected to qualitative review using experts and respondent focus groups and cognitive interviews. The PROMIS Assessment Center is open for use, and work is continuing. As the second phase of PROMIS funding opens, over 60 papers have been published.
The promise of PROMIS derives from its roots in the strong measurement theory of IRT. Also known as latent trait theory, IRT is model based and focuses on understanding a person's response to an item. Item characteristic curves link the underlying level of the characteristic being measured with the probability of responding to a specific response option. The process uses elegant mathematical models with scientifically interpretable parameters in a statistical framework that supports design, administration, and scoring on stable metrics. In the process, knowledge about item parameters and persons' response patterns can be scrutinized to better understand scientific problems for which measurements are being obtained.
It's time to weed out the old books that have guided measurement in nursing science and set aside the practices they espouse. Learn the new vocabulary of IRT-based measurement-start with the item characteristic curve. Upgrade your math-master the ins and outs of logistic functions. Rethink the PhD curriculum in nursing-ever more rigor is required to tackle pressing scientific problems. Use clinical insight to better understand the person-item interaction that occurs at the point of measurement. Learn about IRT so that you can be an informed user of the PROMIS item bank. Let's get serious about modern measurement in nursing science.
Susan J. Henly PhD, RN
Associate Editor
Methods Director
Minnesota Center for Health Trajectory Research
[email protected]
References