1歳6か月児における齲蝕リスク因子のデータマイニング

• Published in: 日本健康医学会雑誌 Vol. 16; no. 2; pp. 22 - 32
• Main Authors: 戸田, 真司, 串田, 守, 荒川, 浩久, 平田, 幸夫, 木本, 一成
• Format: Journal Article
• Language: English
• Published: 日本健康医学会 2007; JAPAN HEALTH MEDICINE ASSOCIATION
• Subjects: 1-year and 6-month old children; 1歳6か月児; data mining; decision tree analysis; dental caries risk factors; データマイニング; 決定木分析; 齲蝕リスク因子
• ISSN: 1343-0025; 2423-9828
• DOI: 10.20685/kenkouigaku.16.2_22

To elucidate of dental caries risk factors that are intrinsic to life habits at the beginning stage of deciduous tooth caries, data mining was conducted of oral condition and life habits of infants. The subjects of this investigation were the children who participated in the 1-year and 6-month old children's dental health examination held in a certain city of Kanagawa Prefecture. The subjects were divided into three groups: children born between January 1978 and December 1979, children born in 1988, and children born in 1998. The decision tree analysis was adopted for data mining. Because the independent variables were qualitative, the most relevant condition was detected by Chi-squared Automatic Interaction Detection. The dependent variable was the presence or absence of dental caries. The independent variables were choices in the contents of dental health questionnaires, and the number of teeth present and dental plaque observed in oral examination. For the validation of the result of decision tree analysis, correlations with the onset of dental caries were analyzed in logistic regression analysis using univariate and multivariate analyses. This study revealed that the factor "dental plaque" was the highest risk factor for dental caries, while "tooth brushing after a meal" was little effect in all groups. These results of decision tree analysis did not conflict with those of logistic regression analysis, and screening was performed more easily with decision tree analysis. All analyses made it clear that plaque adhering to teeth was the highest risk factor for dental caries in all groups and that a correlation between the frequency of snacking and dental caries risk existed. The risk factors obtained by rapid analyses based on a vast number of samples in this study might cause alarm for the life style related habit of infants. Also, the higher the frequency of snacking was, the higher was the dental caries risk. Results of data mining are statistical results based on a vast number of samples. Whereas there are certain limits, risk factors obtained by rapid analyses can be a support to cause alarm. 乳歯齲蝕発生期の生活習慣に内在する齲蝕リスク因子を解明するために,幼児の生活習慣と口腔状態のデータマイニングを行った。本研究では神奈川県某市における1歳6か月児歯科健診受診幼児を対象とした。対象は昭和53-54年生まれ群,昭和63年生まれ群,平成10年生まれ群の3群に分けた。データマイニングには決定木分析を用いた。独立変数が質的変数であるため,x^2相互自動検出により関連の深い条件を検出した。従属変数は齲蝕の有無である。独立変数は,歯科保健に関するアンケート調査項目と,口腔内診査による現在歯数,歯の汚れとした。決定木分析結果を確認する目的で,単変量および多変最解析のロジスティック回帰分析により齲蝕発生との相関性を分析した。本研究の結果,すべての群において「歯の汚れ」が最も高い齲蝕リスク因子であり,「食後の歯みがき」は影響が少ないことが示された。これらの決定木分析の結果は,ロジスティック回帰分析の結果と矛盾するものはなく,ロジスティック回帰分析よりも容易にスクリーニングできた。すべての分析から判明したことは,すべての群で歯に付着した汚れが最も高いリスク因子であり,時代を超えて変化していないという事実である。また,1日あたりの間食の頻度が多いほど齲蝕リスクが高かった。データマイニングによる結果は多くのサンプルに基づく統計結果である。一定の限界はあるものの,迅速な分析で得られたリスク因子は第一警鐘をうながし得ると思われる。