Results of an Investigation into the Integrity of the Ontario Health Survey

The purpose of this investigation is to present results for several checks on the integrity and consistency of the Ontario Health Survey data. This information can guide the interpretation of estimates and statistical inference from OHS data. Several consistency checks were devised to investigate the data. The following conclusions were reached. In general, the survey responses appear to have been coded and processed correctly. All skip patterns were observed. Coding errors were, however, found in Form 4 variable Q17F04, a binary variable equal to one if the individual was designated as a proxy respondent for the other household members. Specifically, a significant number of juveniles (under 7 years of age) were classified as being proxy respondents. Second, intra-respondent response consistency was checked and found to be satisfactory. This applies to both the responses to the interviewer-administered (Form 3 and 4 responses) and to the self-completed questions (Form 5 responses). Finally, a comparison of estimated population-level morbidity along 8 dimensions of health status between the 1991 General Social Survey and the OHS suggested that the mode of administration of the OHS might have had some impact on responses. In particular, the prevalence of morbidity among several attributes (emotion, cognition, and pain) varied considerably between the two surveys. For these three attributes, the OHS proxy- and self-reported, interviewer-administered assessments consistently under-reported the burden of morbidity relative to the GSS. This might be partly attributable to the difference in the mode of administration between the two survey. The GSS was self-reported and telephone-administered, whereas the responses from the OHS used to estimate the incidence of morbidity were proxy-reported and interviewer-administered. Both of these factors would tend to cause an under-reporting of health attributes which are not directly observable and somewhat sensitive in nature. A comparison with the self-reported, self-administered OHS assessments of emotion and pain support this hypothesis. As a result, the proxy-responses to questions pertaining to the subjects’ emotional state, severity of limitations from pain and discomfort and cognitive capacity should not be considered interchangeable with the subjects’ responses.