Reverse Record Check

In survey research, a reverse record check is a method that can be used to verify the accuracy of data that a respondent provides against an information source that contains the “true” answer. Use of this method is not always feasible, even with ample funding, because external sources of information against which a survey response can be validated often do not exist.

The effect of human memory on the response to the survey questions has been the concern of survey methodology for many years. A survey respondent [Page 764]has to be in the position of providing accurate answers to survey questions involving memory. Naturally, this must depend on the nature of questions, but a number of empirical studies have demonstrated that fallibility of memory can be an important source of measurement error. In addition to memory failures, there are other reasons that the answers respondents sometimes provide are inaccurate, thus contributing to measurement error.

Perhaps the most common method of assessing measurement error is the use of a record check study or validation study. Such a study generally assumes that information contained in the external records is without error, that is, that the records contain the true values on the survey variables. In fact this may not be true, as the records may themselves be inaccurate, incomplete, or both. Furthermore, the matching and finding of records for survey individuals is often troublesome and expensive. Errors may occur in matching an individual's survey answer(s) and record data, and if no attempt is made to reconcile differences, such mismatches may indicate response errors where none exists. There are three kinds of record check study designs: (1) reverse record check study, (2) forward record check study, and (3) full design record check study. The reverse record check study is generally based on a retrospective design in which the entire sample is drawn from the record file for persons with a trait under study, interviews are conducted containing questions about information also contained on the records, and a comparison of survey data with record data is performed to estimate the extent and nature of any measurement error. The term reverse record check is used because after the survey is completed, the researcher goes back to the records to check the survey responses. The information can be gathered by using so-called retrospective questions, referring to a certain period of time preceding the date of the interview. In retrospective surveys, many types of recall errors, such as omission, telescoping (forward or backward), confusion, and reporting load effects, can occur.

Earlier studies involving reference period questions showed forward bias in reporting. When memory is inexact, forward bias will arise in answering such questions. Even if an upper bound is not imposed by the investigator, subjects may impose such a bound on their own reports. If subjects bound their reports, it leads to truncation of the distribution of reports just as when bounds are imposed in the question; it also leads to forward bias in reporting. It is important to note that reverse record check studies by themselves fail to measure errors of overreporting. Instead, reverse record checks can measure what portion of the records sampled correspond to events reported in the survey.

...