精神分裂症病人住院期間的暴力行為軌跡類別及其與血脂值之關係

Abstract

目的：本研究之目的在於藉由精神分裂症病人於住院期間的暴力軌跡進行分類，並探索暴力行為軌跡類別與血脂值之間的關係，進一步創造能預測出高危險暴力的精神分裂症族群之模型。 方法：本追蹤型研究於2002年至2003年採用中文版的暴力行為量表(含18個項目)蒐集精神分裂症住院病人的暴力行為資料，並於2012年回顧病歷蒐集個案入院時的血脂值資料，包括總膽固醇、三酸甘油酯、低密度脂蛋白及高密度脂蛋白。首先，利用軌跡分析(trajectory analysis)中的zero-inflated Poisson (ZIP)模型依據病人的暴力行為資料將個案分類，再藉由多元羅吉斯迴歸模型探索暴力行為軌跡類別與血脂值之間的關係，最後採用逐步羅吉斯迴歸分析挑選出預測高暴力行為的最佳預測變項，並藉由ROC曲線下的面積(AUC)來評估這些預測變項的預測效度。 結果：共有四組不同暴力行為變化的組別，其中第一類(無暴力)佔有37.4%，第二類(低暴力-漸趨平緩)佔有39.2%，第三類(高暴力-下降快速)佔有10.3%，第四類 (高暴力-下降緩慢)佔有13.1%。多元羅吉斯迴歸分析指出四組暴力行為軌跡類別的血脂值並無統計上的顯著差異，但三酸甘油酯與低密度脂蛋白隨著暴力程度的增加，似乎存在邊緣性的下降趨勢。而最佳的高危險暴力精神分裂症住院病人之預測模型為-1.21 + [1.90 x女性] – [0.1 x發病年齡] + [0.20 x正性症狀分數] – [0.10 x負性症狀分數] – [0.90 x總膽固醇值] (AUC=0.85, 95% CI=0.72-0.97, p<0.0001)。 結論：精神分裂症病人於住院期間的暴力行為存在明顯不同的軌跡，但此暴力行為軌跡類別與血脂值間無統計上的顯著相關。臨床上的變項(包括女性、較早的發病年齡、較嚴重的正性症狀、較輕微的負性症狀及較低的總膽固醇值)也許能有效的預測精神分裂症住院病人的高暴力行為，以達到早期預防的效果。

Background/Aims: The present study aimed to (a) characterize trajectory patterns of violence in schizophrenia inpatients, (b) explore the relationship between violence trajectories and lipid levels, and (c) further generate a model to predict more violent behavior in schizophrenia inpatients. Method: In a prospective study of schizophrenia inpatients from 2002 to 2003 (n=107), violent behavior was assessed using an 18-item Chinese version of the Violence Scale. Lipid levels at admission, including total cholesterol (TC), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C), were retrospectively collected from chart review in 2012. In trajectory analysis, we adopted the zero-inflated Poisson (ZIP) model to classify subjects based on their violent information. After trajectory analysis was conducted, multinomial logistic regression analyses were performed to clarify the association between violence trajectories and lipid levels. Finally, we conducted stepwise logistic regression to select the best predictor variables for violence and used the area under the curve (AUC) of the receiver operative characteristics (ROC) analysis to assess the predictive validity of predictor variables. Results: We identified four distinctive trajectories of violent behavior: class 1 (no violence, 37.4%); class 2 (low-leveling off, 39.2%); class 3 (high-falling sharply, 10.3%); and class 4 (high-falling slowly, 13.1%). The result of multinomial logistic regression analysis indicated that no significant differences in lipid levels were found among the classes of violence trajectories in schizophrenia inpatients. Yet, it seemed that triglycerides showed borderline descending trend with higher level of violent behavior. The best predictive model for more violence in schizophrenia inpatients was -1.21 + [1.90 x female] – [0.10 x age at onset] + [0.20 x positive symptoms scores] – [0.10 x negative symptoms scores] – [0.90 x total cholesterol level] (AUC=0.85, 95% CI=0.72-0.97, p<0.0001). Conclusion: Distinct violence trajectories existed in schizophrenia inpatients, but no significant associations were found between violence trajectory classes and lipid levels. Some clinical variables including female, earlier age at onset, severe positive symptoms, mild negative symptoms, and lower total cholesterol level might effectively predict more violent behavior in schizophrenia patient during hospitalization to achieve the early prevention.