Analysis of Factors Affecting Prefectural Suicide Rates by Using Support Vector Regression Method

Abstract

近年のわが国の重大社会問題の一つである自殺には地域差が存在するため，都道府県別の自殺 死亡率に有意な影響を与える要因の解明を目的とする実証研究を試みた．47都道府県の男女別年 齢調整自殺死亡率を目的変数，それとの関連が推測される健康，経済，社会，自然分野の指標54 種を説明変数としてサポートベクター回帰分析を行い，自殺死亡率に対する決定要因を探索し， その相対的影響度を推定した．その結果，男女別にそれぞれ12種の要因が得られ，男性では精神 保健福祉士数，家計収入，患者数などの要因，女性では悩み相談，出生率，残業時間などの要因 の影響が大きいことを見出した。また，これまで未検証の精神保健福祉士数や残業時間，精神状 態が有意の影響を与えるが，自殺率との関係が深いとされてきた失業率や離婚率は決定要因には ならなかった．さらに，自殺率が最も高い秋田県について決定要因の結果に基づき自殺対策の提 言を試みた． Suicide, one of the most serious social problems in modern Japan, is known to be caused by many complex factors. Regional correlation study on local suicide rates and various affecting factors attracts much attention from the viewpoint of regional countermeasure for decreasing suicide rates. In this study, a statistical approach to get information on reasons of suicide in Japan has been carried out by searching key factors affecting suicide rates. A support regression method as a nonlinear regression procedure was applied to male and female suicide rates of 47 prefectures as dependent variables and 54 kinds of various regional factors as explanatory variables. Twelve determinants for each men and women, respectively, were obtained which reproduce the suicide rates of prefectures with a statistical significance level. Among those determinants, psychiatric social worker, household income, and patient rate for men, and consultation situation of trouble and stress, birthrate, and overtime for women dominantly affect suicide rates. New