Tuesday, January 7, 2014: 11:15 AM
Royal Pavilion Ballroom I-III (The Regent Hotel)

Marilyn Schapira, MD, MPH, University of Pennsylvania, Philadelphia, PA, Masako Okamoto, Ph.D., Obihiro University of Agriculture and Veterinary Medicine, Obihiro Hokkaido, Japan, Yasuchi Kyutoku, Ph.D., Chuo University, Tokyo, Japan, Yurie Sugimoto, Kyushu Institute of Technology, Tokyo, Japan, Lester Clowney, Jichi Medical University, -, Japan, Ippeita Dan, Ph.D., Chuo University, -, Japan, Tamara Miller, PT, MS, University of Wisconsin-Milwaukee, Milwaukee, WI and Cynthia M. Walker, Ph.D., University of Wisconsin, Milwaukee, WI

Although understanding numerical information is a crucial factor when people make medical decisions, health numeracy has been understudied in Japan due to the absence of adequate assessment tools. In the United States, one such tool is Numeracy Understanding in Medicine Instrument (NUMi; Schapira et al., 2012), which was developed based on empirically derived health numeracy framework. In the current study, we aimed to examine validity of NUMi for the Japanese public, and to compare the Japanese (NUMi-J) and original versions.


NUMi-J was prepared using a forward and backward translation, with context and wordings being modified based on feedback from experts in clinical medicine and statistics and pilot respondents (N=1054). In the main survey, responses from a quota sample (N=2000) approximating the composition of the Japanese population was collected. As in the original study, both classical test theory (CTT) and item response theory (IRT) were used to examine the psychometrical nature of NUMi-J. Validity was assessed by examining the association of NUMi-J scores with other measures of numeracy, literacy, psychographics and levels of understanding of medical information.


The performance of Japanese sample was better than in the original study (mean scores of 14.5 vs. 13.2, respectively). Consequently, the IRT parameter for difficulty was lower with the 20 items ranging respectively from -2.70 to 0.96 in the current, and –1.70 to 1.45 in the original study). Except for difficulty parameter, the results from NUMi-J were comparable with the original ones, with sufficient reliability (Cronbach α=0.83) and discriminability (range of IRT parameter 0.33 to 2.48). As expected, performance on NUMi-J was more strongly correlated with numeracy scores (11-item objective scale, r=0.52; Berlin Numeracy test, r=0.47) than literacy scores (Health knowledge test, r=0.35; Science literacy test, r=0.40), or psychographic scores (Rational-Experiential Index-ability, r=0.19; Self-efficacy, r=0.05). Finally, logistic regression analysis indicated that the predictive ability of NUMi-J on the understanding of quantitative medical information to be the best among all the other parameters (numeracy, literacy and psychographic; All the independent variables were standardized; odds ratio of NUMi-J, 2.3).


In its translated form, NUMi worked well in this culturally different population, and will be a useful tool for assessing patients’ health numeracy in Japan, especially for those with relatively lower numeracy levels.