A critical review of scoring options for clinical measurement tools

Abstract

BackgroundThe aim of this paper is twofold: (1) to describe the fundamental differences between formative and reflective measurement models, and (2) to review the options proposed in the literature to obtain overall instrument summary scores, with a particular focus on formative models.MethodsAn extensive literature search was conducted using the following databases: MEDLINE, EMBASE, PsycINFO, CINAHL and ABI/INFORM, using “formative” and “reflective” as text words; relevant articles’ reference lists were hand searched.ResultsReflective models are most frequently scored by means of simple summation, which is consistent with the theory underlying these models. However, our review suggests that formative models might be better summarized using weighted combinations of indicators, since each indicator captures unique features of the underlying construct. For this purpose, indicator weights have been obtained using choice-based, statistical, researcher-based, and combined approaches.ConclusionWhereas simple summation is a theoretically justified scoring system for reflective measurement models, formative measures likely benefit from the use of weighted scores that preserve the contribution of each of the aspects of the construct.Electronic supplementary materialThe online version of this article (doi:10.1186/s13104-015-1561-6) contains supplementary material, which is available to authorized users.

Highly Predictive Transdiagnostic Features Shared across Schizophrenia, Bipolar Disorder, and ADHD Identified Using a Machine Learning Based Approach

“…using the total score and/or the sub-scores according to pre-defined subscales of a given instrument 469 cannot identify the optimal set of transdiagnostic features. One explanation for this phenomenon is 470 that because the patients share a broad range of phenotypic features, the pre-defined subscales and 471 sum scores become insufficient in capturing the full dimensional structure since most of the 472 instruments are designed to measure a limited set of constructs targeting a specific patient population 473 (Avila et al, 2015). This further demonstrates the advantage of our importance-guided sequential 474 model selection approach in identifying potentially clinically relevant transdiagnostic features across 475 a large set of instruments.…”

Measuring shared decision making in oncology: Development and first testing of the iSHAREpatient and iSHAREphysician questionnaires

“…Clearly, a total score may sometimes be preferred because it can be a useful summary score. For the present questionnaires, we have no theoretical indication that one or more dimensions should be weighted differently from the others to calculate the total score …”

Design and Psychometric Evaluation of the Quality of Life in Patients With Anal Fistula Questionnaire

“…Participants' score on the QoLAF-Q could range from 14 (minimum score) to 70 (maximum score). Using a summation method, 32 the following five-category scoring and interpretation system was developed: 'zero impact'=14 points, 'limited impact'=15-28 points, 'moderate impact'=29-42 points, 'high impact'=43-56, and very 'high impact'=57-70 points.…”