年龄调整率的区间估计及其在宫颈癌患病率估计中的应用

摘要

目的 评价三种年龄调整率可信区间估计方法,探索适合江苏省宫颈癌筛查研究中年龄调整患病率可信区间估计的方法.方法 以二项分布正态近似法、Gamma分布法及"确切概率法"进行年龄调整率的区间估计,运用统计模拟考察多种率及阳性数情况下三种方法的区间覆盖率及宽度.结果 当样本量较小(阳性数较少)时,确切概率法的覆盖率离理论可信度的偏差及区间宽度均优于Gamma分布法,两者的覆盖率均明显优于正态近似法;随着阳性数增多,三法各自的覆盖率偏差及区间宽度均逐渐变小,方法间的差异亦逐渐缩小;当阳性数增至30以上时,确切概率法及正态近似法的覆盖率的偏差皆在±1%以内,此时两者的区间宽度接近;而Gamma分布法的覆盖率偏差若要达到1%以内,则要求总阳性数在100以上.无论样本构成是轻度偏离还是明显偏离总体构成,上述规律皆成立.结论 综合考虑区间覆盖率、区间宽度及计算便捷性,建议当总阳性数小于30时,采用确切概率法计算调整率的可信区间;当总阳性数大于等于30时,采用正态近似法.%Objective To evaluate three estimation methods of confidence interval for age-adjusted ptevalence,i. e. method based on normal approximation to binomial distribution,gamma distribution method and exact probability method,and propose an appreciate method in efficient interval estimation for age-adjusted cervix cancer prevalence. Methods Statistical simulation was conducted to evaluate three interval estimation methods for age-adjusted prevalence based upon large-scale screen data for cervix cancer by comparing coverage rate and average interval width. Results With small sample size,the coverage rate and interval width for exact probability method was superior to the gamma distribution method, and both these two methods wee apparently superior to the normal approximation method. When the sample size becomes large,the coverage deviations and interval widths were both shrink. With the total positive number greater than 30,the coverage deviations of the methods based on normal approximation and exact probability method was less than 1 ％, and the difference of the interval widths of these two methods is negligible. Meanwhile, the Gamma distribution method needed more positive numbers( more than 100)to make the coverage deviation less than 1％. This rule was stable regardless the degree of the difference between the proportion of random sample and population. ConcluSion According to the coverage rate, interval width and convenience of calculating, the exact probability method is recommended when the total positive number is less than 30,while the method based on normal approximation to binomial distribution is proposed when the total positive number is equal or greater than 30.