Collins S E, Haining R P, Bowns I R, Crofts D J, Williams T S, Rigby A S, Hall D M
University of Sheffield.
J Public Health Med. 1998 Sep;20(3):325-30. doi: 10.1093/oxfordjournals.pubmed.a024776.
Health research often seeks to associate individuals to their socio-economic circumstances by linking an individual's postcode to their Census enumeration district (ED). As part of a study into health visitor resource allocation the objective here is to quantify the errors that arise in attaching ED level deprivation scores to records and counts of records by ED when records are matched to EDs via their postcodes rather than their exact address.
The result of routine matching of postcodes to EDs was compared with the more accurate method of matching addresses to EDs. Townsend scores were then attributed to records according to the two different methods and the results compared. A sample of 4013 births registered in Sheffield in 1996 was used.
The comparative work showed that the mismatching of individual addresses arising from matching postcodes to EDs was 16.4 per cent. (The 95 per cent confidence interval is 15.1-17.7 per cent.) Over one-third of mismatched records (about 6 per cent of the total records) were found to have Townsend scores greater than +/- 2 compared with the score obtained through the more accurate process of address matching.
The evidence of the study is that it is important to recognize there are errors inherent in matching individual addresses to EDs via the address postcode. For problems involving resource allocation and for research into relationships between health outcomes or service uptake and deprivation it may be necessary to seek to quantify the level of error introduced through using postcode to ED matching.
健康研究常常试图通过将个人邮政编码与人口普查枚举区（ED）相联系，来将个人与他们的社会经济状况关联起来。作为一项关于健康访视员资源分配研究的一部分，这里的目标是量化当记录通过邮政编码而非确切地址与枚举区匹配时，在将枚举区层面的贫困得分附加到记录以及按枚举区统计记录数量时所产生的误差。
将邮政编码与枚举区的常规匹配结果与更精确的地址与枚举区匹配方法进行比较。然后根据两种不同方法为记录赋予汤森得分，并比较结果。使用了1996年在谢菲尔德登记的4013例出生样本。
比较工作表明，通过邮政编码与枚举区匹配导致的个人地址不匹配率为16.4%。（95%置信区间为15.1 - 17.7%。）与通过更精确的地址匹配过程获得的得分相比，发现超过三分之一的不匹配记录（约占总记录的6%）的汤森得分大于±2。
该研究的证据表明，认识到通过地址邮政编码将个人地址与枚举区匹配存在固有误差很重要。对于涉及资源分配的问题以及健康结果或服务利用与贫困之间关系的研究，可能有必要设法量化通过使用邮政编码与枚举区匹配引入的误差水平。
