Format

Send to

Choose Destination
BMC Med Inform Decis Mak. 2016 Sep 13;16:118. doi: 10.1186/s12911-016-0361-9.

Predicting Japanese Kampo formulas by analyzing database of medical records: a preliminary observational study.

Author information

1
Center for Kampo Medicine, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo, 160-8582, Japan.
2
Human Genome Center, The Institute of Medical Science, The University of Tokyo, 4-6-1 Shirokanedai, Minato-ku, Tokyo, 108-8639, Japan.
3
SFC Laboratory, Keio University, 5322 Endo, Fujisawa, Kanagawa, 252-0882, Japan.
4
Division of Health Medical Data Science, Health Intelligence Center, The Institute of Medical Science, The University of Tokyo, 4-6-1 Shirokanedai, Minato-ku, Tokyo, 108-8639, Japan.
5
School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo, 113-8656, Japan.
6
Center for Kampo Medicine, Keio University School of Medicine, 35 Shinanomachi, Shinjuku-ku, Tokyo, 160-8582, Japan. watanabekenji@keio.jp.
7
Faculty of Environment and Information Studies, Keio University, 5322 Endo, Fujisawa, Kanagawa, 252-0882, Japan. watanabekenji@keio.jp.

Abstract

BACKGROUND:

Approximately 90 % of physicians in Japan use Kampo medicine in daily practice. However, it is a challenge for physicians who do not specialize in Kampo medicine to select a proper Kampo formula out of the 148 officially approved formulas, as the decision relies on traditional measurements and traditional medicine pattern diagnoses. The present study tries to evaluate the feasibility of a decision support system for frequently used Kampo formulas.

METHODS:

Our study included 393 patients who visited the Kampo Clinic at Keio University Hospital for the first time between May 2008 and March 2013. We collected medical records through a browser-based questionnaire system and applied random forests to predict commonly prescribed Kampo formulas.

RESULTS:

The discriminant rate was the highest (87.0 %) when we tried to predict a Kampo formula from two candidates using age, sex, body mass index, subjective symptoms, and the two essential and predictable traditional medicine pattern diagnoses (excess-deficiency and heat-cold) as predictor variables. The discriminant rate decreased as the candidate Kampo formulas increased, with the greatest drop occurring between three (76.7 %) and four (47.5 %) candidates. Age, body mass index, and traditional medicine pattern diagnoses had higher importance according to the characteristics of each Kampo formula when we utilized the prediction model, which predicted a Kampo formula from among three candidates.

CONCLUSIONS:

These results suggest that our decision support system for non-specialist physicians works well in selecting appropriate Kampo formulas from among two or three candidates. Additional studies are required to integrate the present statistical analysis in clinical practice.

KEYWORDS:

Decision support system; Japanese Kampo medicine; Random forests; Traditional medicine pattern diagnosis

PMID:
27619018
PMCID:
PMC5020542
DOI:
10.1186/s12911-016-0361-9
[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for BioMed Central Icon for PubMed Central
Loading ...
Support Center