Runyan D K, Gould C L, Trost D C, Loda F A
Am J Public Health. 1981 Jul;71(7):706-11. doi: 10.2105/ajph.71.7.706.
This study examined the records of the North Carolina Central Registry of Child Abuse and Neglect to determine which social, family, and child characteristics were most influential in the decision to place a child in foster care. These records contained all theoretically relevant factors as well as demographic data. Analysis included the computation of odds ratios for foster care for each of 250 variables. A maximum likelihood logistic regression model was constructed to obtain the independent and cumulative contribution of each factor. Some expected variables such as parental stress factors (substance abuse) and types of abuse (burns and scalds) placed a child at a significant risk for placement in foster care (p less than 0.01). However, less obvious factors such as referral source (law enforcement agencies) or geographic area also placed children at risk. Overall, the model explained little of the variance of these decisions (R2 = 0.168) and poorly predicted placement (sensitivity 66.3 per cent, specificity 74.6 per cent). Using existing data, we were unable to adequately describe the decision process in selecting foster care.
本研究查阅了北卡罗来纳州虐待与忽视儿童中央登记处的记录，以确定哪些社会、家庭和儿童特征在决定将儿童安置到寄养家庭中最具影响力。这些记录包含了所有理论上相关的因素以及人口统计数据。分析包括计算250个变量中每个变量进入寄养家庭的比值比。构建了一个最大似然逻辑回归模型，以获得每个因素的独立和累积贡献。一些预期变量，如父母压力因素（药物滥用）和虐待类型（烧伤和烫伤），使儿童面临被安置到寄养家庭的显著风险（p小于0.01）。然而，一些不太明显的因素，如转介来源（执法机构）或地理区域，也会使儿童面临风险。总体而言，该模型对这些决策的方差解释甚少（R2 = 0.168），对安置情况的预测效果不佳（敏感度为66.3%，特异度为74.6%）。利用现有数据，我们无法充分描述选择寄养家庭的决策过程。
