Automatic linkage of vital records.

The term record linkage has been used to indicate the bringing together of two or more separately re?orded pieces of information concerning a particular individual or family (I). Defined in this broad manner, it includes almost any use of a file of records to determine what has subsequently happened to people about whom one has some prior information, The various facts concerning an individual which in any modern society are recorded routinely would, if brought together, form an extensively documented history of his 1ife. In theory at least, an understanding might be derived from such collective histories concerning many of the factors which operate to influence the welfare of human populations, factors about which we are at present almost entirely in ignorance. Of course, much of the recorded information is in a relatively inaccessible form; but, even when circumstances have been most favorable, as in the registrations of births, deaths, and marriages, and in the census, there has been little recognition of the special value of the records as a source of statistics when they are brought together so as to relate the successive events in the lives of particular individuals and families. The chief reason for this lies in the high cost of searching manually for large numbers of single documents among vast accumulations of files. It is obvious that the searching could be mechanized, but as yet there has been no clear demonstration that machines can carry out the record linkages rapidly enough, cheaply enough, and with sufficient accuracy to make this practicable. The need for various follow-up studies such as might be carried out with the aid of record linkage have been discussed in detail elsewhere (I, 2), and there are numerous examples o{ important surveys which could be greatly extended in scope if existing record files were more readily linkable (3). Our special interest in the techniques of record linkage relates to their possible use (i) for keeping track of large groups of individuals who have been exposed to low levels of radiation, in order to determine the causes of their eventual deaths (see 4, chap. 8, para. 48; 5), and (ii) for assessing the relative importance of repeated natural mutations on the one hand, and of fertility differentials on the other, in maintaining the frequency of genetic defects in human populations (see 4, chap. 6, para. 36c). Our own studies (6) were started as part of a plan to look for possible differentials of family fertility in relation to the presence or absence of hereditary disease (through the use of vital records and a register of hm-sdicapped children). The first step has been the development of a method for linking birth records to marriage records automatically with a Datatron 205 computer. For this purpose use has been made of the records of births which occurred in the Canadian province of British Columbia during the year 1955 (34, 138 births) and of the marriages which took place in the same province over the 10-year period 1946-55 (114,47 1 marriages). Fortunately, these records were already in punch-card form as a part of Canada’s National Index, and from them could be extracted most of the necessary information on names and other