Ontario Health Survey 1990 - Data Sources

Data Source

The Ontario Health Survey (OHS) was carried out across the province in 1990. The survey was in 2 parts. The first was a 22-page interview conducted with 1 member of a randomly selected household, knowledgeable enough to answer questions about everyone in the household. These questions focused on recent or current health problems, use of health services and demographic information such as income and education.

Part 2 was a 26-page self-completed questionnaire filled out by each member of the household ages 12 and older. This questionnaire covered self-rated health, use of medicines and drugs, smoking, alcohol use, family relationships, social support, psychological and emotional well-being, suicide, dental health, driving and road safety, women's reproductive health, sexual health, occupational health, physical activity, and nutrition. Some of these questions were asked only of people 16 and older (i.e., question on reproductive and sexual health, disease detection and driving and safety).

The overall response rates to the OHS for Leeds, Grenville and Lanark was 85.8% to the interviewer, completed portion and 75.9% to the self-completed questionnaire.

Data Limitations

Data from the Ontario Health Survey are subject to certain limitations that can be categorized as sampling and non-sampling error.

Sampling error
When a survey is done, a sample of the population is asked questions. The answers that this sample of people give are assumed to be the same as if the entire population had been asked the same questions. But the survey result is really only an estimate that represents the population. Sampling error is the difference between the estimate derived from the sample and the result that would be obtained if the entire population were canvassed. The sampling error can not be measured directly so the coefficient of variation (cv) is used instead. The size of the cv represents the size of the sampling error and the reliability of the survey estimate. Estimates with small cv's (under 16.6%) are said to be reliable and represent the population value. Estimates with larger cv's (16.6%-25.0%) can be used but, because of their higher sampling variability, are shown with a "use with caution" warning. Estimates with large cv's (over 25.0%) are suppressed because the estimate is not reliable and may not represent the population.

Non-sampling error
Non-sampling errors include response and non-response errors.

Response errors occur when respondents answer a question inaccurately, either by accident or on purpose. Questions that ask people to recall events over a long time period are more likely to result in response errors; for example how many times they visited a health professional during the last 12 months or when they had their last PAP test. Tendencies to over or under-report should balance each other out, resulting in little or no bias in the results. However, studies have shown that some things perceived as good behaviours, such as seat-belt use, tend to be consistently over-reported while others like alcohol use tend to be under-reported. Questions answered on behalf of household members by a proxy respondent are more likely to be inaccurate.

Non-response errors occur when a selected respondent can not be interviewed (total non-response) or when part of the questionnaire is not completed (item non-response). If the characteristics of respondents (age, sex, education level, health status) differ from those of non-respondents there will be biases in the survey results. The impact of non-response errors depends both on the level of non-response and the extent of difference between the characteristics of respondents and non-respondents.

Total response rates for the OHS were very high and the results were weighted to help minimize the impact of non-response. Part 1 of the survey had an 87% response rate providing information on 61,239 Ontario residents. The response rate for the self-completed questionnaire was 77% in Ontario resulting in data for 49,164 people over age 12. Response rates this high mean that the information has a high degree of accuracy; the information reflects the health and habits of the total population not just those in the sample.

In general the item non-response was low. However, a few questions in the self-completed questionnaire had higher then average non-response, probably due to the page layout, problems understanding skip patterns or sensitivity of the issue.

Overall the data quality from the OHS is excellent. The sample design minimized sampling error and high response rates reduced non-sampling errors.