Calibration of the two prediction models.

N = 144,018 analysis cohort with complete data on all variables included in any model. The cohort is split into 10ths of the distribution of the linear predictor for each of the two prediction models. For example, for the Templeton prediction model the observed and predicted are compared by 10th of the Templeton linear predictor.aTempleton: As in reference [7], to date the only externally validated prediction model (Text S1).bNovel model: Using the same variables as Templeton but allowing them to have different multivariable coefficients to those originally derived by Templeton and including terms for all causes of infertility (rather than just tubal versus other as in the original Templeton) and four additional predictors: type of hormonal preparation, whether egg came from patient or donor, number of treatment cycles, and whether ICSI was used with the IVF (Text S2).本分析队列共纳入144018例对象，所有纳入任一模型的变量均具备完整观测数据。针对两个预测模型，将队列按照其线性预测器（linear predictor）的分布划分为10个十分位组。例如，针对Templeton预测模型，可通过其线性预测器的十分位分组对观测值与预测值进行对比。 a Templeton预测模型：如参考文献[7]所述，是目前唯一经过外部验证的预测模型（补充文本S1）。 b 新型预测模型：采用与Templeton模型一致的变量集，但允许各变量的多变量系数与Templeton原始推导的系数存在差异；同时纳入所有不孕病因的相关项（而非原始Templeton模型仅区分输卵管性与非输卵管性不孕），并新增4项预测指标：激素制剂类型、卵子供体属性（患者自身或供体）、治疗周期数，以及体外受精（In Vitro Fertilization, IVF）是否联合卵胞浆内单精子注射（Intracytoplasmic Sperm Injection, ICSI）（补充文本S2）。