目的：探讨不同类型离焦眼镜(HAL和DIMS)与常规单光眼镜(SVS)在儿童近视防控中的效果差异，基于随机森林模型分析其相关影响因素。

方法：回顾性分析2022年6月至2023年12月在成都市第三人民医院就诊的近视患者190例380眼，按照近视矫正方式的不同将患者分为3组：SVS组84例168眼配戴SVS； HAL组58例116眼配戴HAL； DIMS组48例96眼配戴DIMS，均选取右眼数据进行分析。随访时间为12 mo，比较三组患者一般资料和戴镜前后等效球镜度(SE)和眼轴长度(AL)指标变化，并基于随机森林模型和多Logistic回归分析筛选影响近视控制效果的因素，采用ROC曲线评估预测效能。

结果：三组患者戴镜前一般资料比较无差异(P>0.05)。HAL组和DIMS组在控制SE和AL变化方面优于SVS组，且HAL组控制效果优于DIMS组(均P<0.05)。随机森林模型显示，当影响因素为3时，平均袋外估算误差率最低，为0.1988。前三位影响因素依次为AL、戴镜年龄、裸眼视力。将其纳入多因素Logistic回归分析显示AL是离焦眼镜防控效果的保护因素，戴镜年龄、裸眼视力是离焦眼镜防控效果的危险因素，各指标预测离焦眼镜防控效果的AUC分别为0.802、0.747和0.720。

结论：HAL组近视防控效果优于DIMS组和SVS组，患者AL、戴镜年龄和裸眼视力是影响离焦眼镜防控效果的因素。

METHODS:A retrospective analysis was conducted on 190 myopic children(380 eyes)who attended the Ophthalmology Department of the Third People's Hospital of Chengdu between June 2022 and December 2023. According to the types of spectacle correction, patients were divided into three groups: SVS(84 cases, 168 eyes), HAL(58 cases, 116 eyes), and DIMS(48 cases, 96 eyes), all data from the right eye were selected for analysis, with a follow-up period of 12 mo. General information and changes in spherical equivalent(SE)and axial length(AL)were observed and compared among the three groups. A random forest model combined with Logistic regression analysis was used to identify factors influencing the effectiveness of treatment, and the predictive performance of these factors was evaluated using receiver operating characteristic(ROC)curves.

RESULTS:There was no statistical significance in the general characteristics among the three groups of patients before wearing lenses(P>0.05). Both the HAL and DIMS groups showed better control of SE progression and AL changes compared to the SVS group, and the HAL group demonstrated superior control effect compared to the DIMS group(all P<0.05). The random forest model indicated that the lowest mean out-of-bag estimation error rate, which was 0.1988, was achieved when the number of variables was 3, and the top 3 influencing factors were AL, spectacle-wearing age, and uncorrected visual acuity. Multivariate Logistic regression analysis revealed that AL was a protective factor for the effectiveness of defocus spectacles, while spectacle-wearing age and uncorrected visual acuity were risk factors. The area under the curve(AUC)values for predicting the effectiveness of defocus spectacles were 0.802, 0.747, and 0.720 for AL, spectacle-wearing age, and uncorrected visual acuity, respectively.

CONCLUSION:The HAL group demonstrated better effectiveness in myopia prevention and control compared to the DIMS and SVS groups. AL, spectacle-wearing age, and uncorrected visual acuity were identified as independent factors influencing the effectiveness of defocus spectacles.