PSYCHOMETRIC PROPERTIES OF A NEW SUBJECTIVE NUMERACY SCALE: CLASSICAL AND IRT ANALYSES

Purpose: Recent research has suggested that patient numeracy influences cancer screening and treatment decision-making. Objective numeracy measures have been constructed; however, poor psychometric properties and participant reluctance to complete math problems limit their utility. A Subjective Numeracy Scale (SNS; Fagerlin et al., 2007) was developed as a proxy for objective numeracy measures. However, little psychometric information is available on this new instrument. This study analyzed the SNS using classical and item response theory (IRT) psychometric analyses. Method: Ovarian screening participants (N = 306) completed instruments including the SNS and a 3-item objective numeracy measure (NUM). Results: Participants’ mean age was 59 (SD = 11), and 92% were at least high school educated. Exploratory factor analysis revealed two expected factors: perceived numerical ability (PNA) and preferences for numerical information (PREF). Internal consistency of the PNA subscale was strong (α = .91) but lower for the PREF subscale (α = .72). Correlations suggested modest positive associations between both SNS subscales and the NUM (ρ = .35 and .38 for PNA and PREF). IRT analyses of PNA items revealed adequate item discrimination parameters (2.83-5.75) and test information covering the continuum of perceived numerical ability from 2 SD below to 2 SD above the mean. However, IRT analyses of PREF subscale items showed lower item discrimination parameters (1.10-2.98); 3 items with minimal measurement information; and test information covering the continuum of preference for numerical information from 2 SD below to 1 SD above the mean.

Conclusions: While the concept of subjective numeracy is innovative, modest correlation with the NUM suggests that the SNS may not be equivalent to measures of objective numeracy. In addition, IRT analyses suggest elevated measurement error at the high and low ends of the subjective numeracy continuum, as well as weak measurement performance of several PREF items. Further development of this promising instrument may be warranted prior to its use as a proxy for objective numeracy.