台灣性罪犯靜態再犯危險評估量表(TSOSRAS)之建立及其外在效度之研究

The Establishment of Taiwan Sex Offender Static Risk Assessment Scale (TSOSRAS-2004): A Validity Study by Various Samples

林明傑(Min-Chieh Jay Lin) ； 董子毅(Tzu-Yi Dong)

《亞洲家庭暴力與性侵害期刊》 1卷1期 (2005/12) Pp. 49-110

https://doi.org/10.29804/AJDVSO.200512.0003

摘要

本研究之目的在於以不同之樣本建立台灣性罪犯靜態再犯危險評估量表(TSOSRAS-2004)。共分兩次研究，研究一為建立之研究，研究二為外部效度之研究。
在研究一，收集於1994年至1996年從台北及高雄監獄出獄之性罪犯共423位為樣本，填入由RRASOR、Static-99、及MnSOST-R收集之危險因素且依據台灣資料現況而建立之15項因素量表初稿，追蹤至2003年2月查閱刑案資料註記以了解其有無再犯，平均追蹤期為7.6年。篩選危險因素之統計方法為卡方、Somer'sd、羅吉斯迴歸、及Cox迴歸。發現共有八個因素可以列入。此八項分別為性犯行遭起訴加上判刑確定的次數、過去被判刑確定之任何犯行次數、在保護管束中又犯下性犯行、該次性犯行中的「非性暴力行為」、該次性犯行被害者有13至15歲少女且小加害人5歲以上、該次性犯行被害者之性別、該次性犯行的被害者人數、及預估出獄時的年齡。以分別追蹤一年、三年、及七年及各所篩選顯著之因素建立量表，發現預測效度各為r=.238(ROC=.767)、r=.328(ROC=.811)、及r=.312(ROC=.752)，均為中度且滿意之效度。然以成人強暴犯、家外兒童性侵害犯、家內兒童性侵害犯三類性罪犯觀之，其預測效度各為r=.231(ROC=.736)、r=.380(ROC=.765)、及r不顯著(ROC=.590)。並將此八題之量表作為定稿。
在研究二，收集於1997年至1999年從台北、台中、及高雄監獄出獄之性罪犯共421位為樣本，填入此八題之量表，追蹤至2004年10月平均追蹤期為7.2年。以全體樣本追蹤三年有無再犯，發現r=.232(ROC=.763)。而以1997年追蹤七年共七人再犯，r=.000(ROC=.693)，其可能因為部分樣本追蹤未達七年致樣本不夠所致。另外，研究之發現及未來研究之建議亦在文中有所討論。

並列摘要

The purpose of this study was to establish sex offender static risk assessment scale in Taiwan by using various samples. In Study 1, the sex offenders in Taiwan released from 1994 through 1996 were collected as sample (total 423), and they were tracked to the year of Fubuary 2003 to check whether or not they recidivated. The risk items from RRASOR, Static-99, and MnSOST-R were screened by the data availability in Taiwan and statistic methods, including chi-square, Somer's d, logistic regression, and cox regression. It was found that 8 items could be included in the scale. They were the number of past sex offense conviction, number of any past crimes, sex offense under probation/parole, non-sexual violence during sex offense, victimizing 13-15 girls, any boy victim, number of victims, and age of release. One, three, and seven years of tracking terms were predicted by various sets of screened items. The predictive validity of 1-year scale, 3-year scale, and 7-year scale were r=.238 (ROC=767), r=.328 (ROC=.8l1), and r=.312 (ROC=.752) respectively. Based on predicting various types of sex offenders in 7-year scale, it was found that the predictive validity on adult rapist, and extrafamilial child molesters, were r=.231 (ROC=.736) and r=.380 (ROC=.765), respectively, whereas the Pearson's r did not reach significance on intrafamilial child molesters (ROC=.590).
In Study 2, the sex offenders in Taiwan released from 1997 through 1999 were collected as sample (total 421), and they were tracked to the year of 2004 to check whether or not they recidivated. The predictive validity of 3-year scale was r=.232 (ROC=763), whereas the Pearson's r of 7-year scale did not reach significance, but the ROC still showed median-low predictive validity (ROC=.693). The findings and suggestions would both discussed.