Owusu-Edusei Kwame, Chang Brian A, Aslam Maria V, Johnson Ryan A, Pearson William S, Chesson Harrell W
National Center for HIV/AIDS, Viral Hepatitis, STD, and TB Prevention, Centers for Disease Control and Prevention, 1600 Clifton Road MS E-07, Atlanta, GA 30329, USA; and Corresponding author. Email:
Icahn School of Medicine at Mount Sinai, 1 Gustave L. Levy Place, New York, NY 10029, USA.
Sex Health. 2019 Apr;16(2):148-157. doi: 10.1071/SH17221.
Background Violent crime rates are often correlated with the hard-to-measure social determinants of sexually transmissible infections (STIs). In this study, we examined whether including violent crime rate as an independent variable can improve the quality of ecological regression models of STIs.
We obtained multiyear (2008-12) cross-sectional county-level data on violent crime and three STIs (chlamydia, gonorrhoea, and primary and secondary (P&S) syphilis) from counties in all the contiguous states in the US (except Illinois and Florida, due to lack of data). We used two measures of STI morbidity (one categorical and one continuous) and applied spatial regression with the spatial error model for each STI, with and without violent crime rate as an independent variable. We computed the associated Akaike's information criterion (AIC) and Bayesian information criterion (BIC) as our measure of the relative goodness of fit of the models.
Including the violent crime rate as an independent variable improved the quality of the regression models after controlling for several sociodemographic factors. We found that the lower calculated AICs and BICs indicated more favourable goodness of fit in all the models that included violent crime rates, except for the categorical P&S syphilis model, in which the violent crime variable was not statistically significant.
Because violent crime rates can account for the hard-to-measure social determinants of STIs, including violent crime rate as an independent variable can improve ecological regression models of STIs.
背景 暴力犯罪率通常与性传播感染（STIs）难以衡量的社会决定因素相关。在本研究中，我们检验了将暴力犯罪率作为自变量是否能提高性传播感染的生态回归模型的质量。
我们获取了美国所有毗邻州（伊利诺伊州和佛罗里达州因缺乏数据除外）各县多年（2008 - 12年）的县级横断面数据，内容包括暴力犯罪以及三种性传播感染（衣原体、淋病和一期及二期梅毒）。我们使用了两种性传播感染发病率的测量方法（一种为分类变量，一种为连续变量），并针对每种性传播感染应用空间误差模型进行空间回归，自变量中分别包含和不包含暴力犯罪率。我们计算了相关的赤池信息准则（AIC）和贝叶斯信息准则（BIC），以此作为模型相对拟合优度的衡量指标。
在控制了几个社会人口学因素后，将暴力犯罪率作为自变量提高了回归模型的质量。我们发现，除了一期及二期梅毒分类模型中暴力犯罪变量无统计学意义外，在所有包含暴力犯罪率的模型中，较低的计算得到的AIC和BIC表明拟合优度更佳。
由于暴力犯罪率可以解释性传播感染难以衡量的社会决定因素，将暴力犯罪率作为自变量可以改善性传播感染的生态回归模型。
