Starr T B, Levine R J
Am J Epidemiol. 1983 Dec;118(6):897-904. doi: 10.1093/oxfordjournals.aje.a113707.
The use of indirect standardization in the assessment of the fertility of occupationally exposed workers is briefly reviewed and critiqued. The calculation of expected births in the method of Levine et al. (J Occup Med 1980;22:781-91) is modified to eliminate negative bias. An example is given using data from a 1977 survey of 60 male workers at a chemical manufacturing plant in Denver, Colorado, who were exposed to dibromochloropropane. The example illustrates how in-plant nonexposed reproductive experience provides a valuable supplement to US fertility tables which are specific only to race, birth cohort, age, and parity. It is also shown, however, that explicit control for potential confounding factors not included in the tables, such as marital status and surgical sterilization, can actually create rather than alleviate confounding error. This occurs when the additional factors co-vary in the reference population with the factors already included in the tables. For martial status, the control-induced error was readily minimized by restricting analysis to married experience at parity one or greater. For surgical sterilization, the corresponding error could not be reduced without severely compromising sample size, and hence control of this potential confounder in similar circumstances is not recommended.
简要回顾并批判了在评估职业暴露工人的生育力时使用间接标准化的情况。对莱文等人（《职业医学杂志》1980年；22:781 - 91）方法中预期出生数的计算进行了修正，以消除负偏差。给出了一个例子，使用的数据来自1977年对科罗拉多州丹佛市一家化学制造厂60名接触二溴氯丙烷的男性工人的调查。该例子说明了工厂内未暴露者的生殖经验如何为仅针对种族、出生队列、年龄和产次的美国生育表提供有价值的补充。然而，研究还表明，对表中未包括的潜在混杂因素（如婚姻状况和手术绝育）进行明确控制，实际上可能会产生而不是减轻混杂误差。当这些额外因素在参考人群中与表中已包括的因素共同变化时，就会出现这种情况。对于婚姻状况，通过将分析限制在产次为一或更高的已婚经验上，可轻易将控制引起的误差降至最低。对于手术绝育，若不严重影响样本量，相应误差就无法降低，因此不建议在类似情况下控制这种潜在的混杂因素。
