Canadian Health Measures Survey (CHMS)

Canadian Health Measures Survey (CHMS)
Variable(s)
Summary of changes
Other reference periods
Related products

Detailed information for January 2014 to December 2015 (Cycle 4)

Status:

Active

Frequency:

Every 2 years

Record number:

5071

The Canadian Health Measures Survey (CHMS) aims to collect important health information through a household interview and direct physical measures at a mobile examination center (MEC), sometimes referred to as a mobile clinic.

Data release - October 13, 2016 (First in a series of releases for this reference period); November 20, 2023 (Diabetes among Canadian Adults)

Description

The Canadian Health Measures Survey (CHMS), launched in 2007, is collecting key information relevant to the health of Canadians by means of direct physical measurements such as blood pressure, height, weight and physical fitness. In addition, the survey is collecting blood and urine samples to test for chronic and infectious diseases, nutrition and environment markers.

Through household interviews, the CHMS is gathering information related to nutrition, smoking habits, alcohol use, medical history, current health status, sexual behaviour, lifestyle and physical activity, the environment and housing characteristics, as well as demographic and socioeconomic variables.

All of this valuable information will create national baseline data on the extent of such major health concerns as obesity, hypertension, cardiovascular disease, exposure to infectious diseases, and exposure to environmental contaminants. In addition, the survey will provide clues about illness and the extent to which many diseases may be undiagnosed among Canadians. The CHMS will enable us to determine relationships between disease risk factors and health status, and to explore emerging public health issues.

CHMS data are representative of the population whether they are healthy or not and provide a better picture of the actual health of Canadians.

The following are some of the measures that the CHMS includes:

Physical measures
. Anthropometry (standing height, sitting height, weight, waist circumference, hip circumference)
. Cardiovascular health and musculoskeletal fitness (resting heart rate and blood pressure, hand grip strength)
. Physical activity (accelerometry)
. Hearing (audiometry, otoacoustic emissions, otoscopy, tympanometry)
. Lung health (spirometry)

Blood measures
. Nutritional status (e.g., folate, Vitamin D)
. Diabetes (e.g., glycated hemoglobin A1c)
. Cardiovascular health (e.g., lipid profile, red blood cell fatty acids)
. Environmental exposure (e.g., acrylamides, dioxins, furans)
. Infection markers (e.g., hepatitis)

Urine measures
. Kidney health (e.g., creatinine)
. Environmental exposure (e.g., cotinine, pesticides)
. Nutritional status (e.g., iodine, Vitamin C)

Indoor air measures (household)
. Environmental exposure (volatile organic compounds)

Tap water (household)
. Environmental exposure (fluoride, volatile organic compounds)

The CHMS stores biological samples for further analyses of measures at a later date (CHMS Biobank). The CHMS team works closely with the Health Canada and Public Health Agency of Canada Research Ethics Board and the Office of the Privacy Commissioner of Canada in order to address privacy issues and to implement proper laboratory procedures.

Reference period: Varies according to the question (for example: "over the last 12 months," "over the last 6 months," "during the last week")

Collection period: The first 2 cycles varied, but starting with cycle 3, January - December.

Subjects

  • Diseases and health conditions
  • Environmental factors
  • Health
  • Lifestyle and social conditions

Data sources and methodology

Target population

The target population for CHMS consists of persons 3 to 79 years of age living in the ten provinces.

The observed population excludes: persons living in the three territories; persons living on reserves and other Aboriginal settlements in the provinces; full-time members of the Canadian Forces; the institutionalized population and residents of certain remote regions. Altogether these exclusions represent approximately 4% of the target population.

Instrument design

Two questionnaires were used for cycle 4 of the Canadian Health Measures Survey:

1) Household questionnaire:

http://www23.statcan.gc.ca/imdb/p3Instr.pl?Function=getInstrumentList&Item_Id=152604&UL=1V
The household questionnaire content was developed with input from stakeholders (Health Canada and the Public Health Agency of Canada) and from external experts who participated as members of various advisory committees. Much of the cycle 4 household questionnaire was the same as the cycle 3 questionnaire. Prior to finalizing the questions, one-on-one qualitative test interviews were conducted to look at specific questionnaire content, particularly the content new to cycle 4. As a result of this testing, improvements were made to questionnaire wording and instructions and to the flow of questions.

2) Clinic questionnaire:

http://www23.statcan.gc.ca/imdb/p3Instr.pl?Function=getInstrumentList&Item_Id=150622&UL=1V&

Development of the clinic questionnaire proceeded in much the same way as that of the household questionnaire. Content was determined through a comprehensive consultation process and multiple iterations of the collection application were generated. Each iteration was assessed on flow within the mobile examination center (MEC) for both the respondent and staff. Quantity and quality of data collected was also assessed.

The clinic questionnaire includes a set of self-reported health questions similar to the type of questions asked within the household questionnaire. As in cycle 3,the questions included at the MEC were related to medication use, tap water, hearing, sun exposure, indoor air and fish and shellfish consumption. In addition, the clinic questionnaire includes introductory text/instructions and screening and administrative questions related to the physical measures tests conducted at the MEC.

Sampling

This is a sample survey with a cross-sectional design.

The Canadian Health Measures Survey uses a stratified three-stage sample made up of one or two selected respondents from each dwelling selected in a sampled collection site.

SAMPLING UNIT
The sampling unit at the first stage is a collection site. A collection site is a geographical unit limited to a radius of about 50 km in urban areas and up to 75 km for rural areas. The sampling unit at the second stage is the dwelling and at the third stage, the sampling unit is the person.

STRATIFICATION METHOD
Strata are defined at every stage. At the first stage, collection sites are stratified in the 5 Canadian regions (Atlantic, Quebec, Ontario, Prairies, and British Columbia).
At the second stage, dwellings are stratified in 7 hierarchical groups defined according to the presence or not of age groups and derived using the household composition obtained from recent auxiliary information:

1) dwellings with 3 to 5 year-olds, else
2) dwellings with 6 to 11 year-olds, else
3) dwellings with 12 to 19 year-olds, else
4) dwellings with 60 to 79 year-olds, else
5) dwellings with 20 to 39 year-olds, else
6) dwellings with 40 to 59 year-olds, else
7) other dwellings without household composition or with all ages outside the ones above.

Finally, at the third stage, the persons in the household at the time of interview are stratified in two age groups prior to selection: 3 to 11 year-olds and 12 to 79 year-olds.

SAMPLING AND SUB-SAMPLING
The Canadian Health Measures Survey consists of a full sample and several subsamples.
For the full sample, at the first stage, a sample of 16 collection sites was required. The sites were allocated by region: Atlantic (2), Quebec (4), Ontario (6), Prairies (2) and British Columbia (2). Within each region, sites were sorted according to the size of their population and whether or not they belonged to a census metropolitan area. Within the Prairies and Atlantic regions, they were first sorted by province. Sites were then randomly selected using a systematic sampling method with probability proportional to the size of each site's population.

The sample size determination and allocation for the second and third stage are done together.
The target sample size for cycle 4 was 5,700 respondents for the clinic component of the survey, which worked out to approximately 356 respondents per collection site. To determine the number of dwellings to sample in each collection site to reach this target, previous response rates were used from both the CHMS and the Canadian Community Health Survey (CCHS). The CHMS and CCHS were both used to calculate:
• The expected probability that a dwelling would be eligible for the CHMS (the eligibility rate)
• The expected probability that a roster of all occupants of the household would be completed (the roster rate)
• The expected probability that a selected person would respond to the household questionnaire (the questionnaire rate)


Finally, rates from the previous CHMS sites were used to calculate the expected probability that a household questionnaire respondent would also be a respondent to the clinic (the clinic rate). Since outside CMA, inside CMA urban and inside CMA urban core (downtown) collection sites each have distinct response rates, each collection site was classified into one of these three categories and the previously mentioned response rates were calculated and applied separately for each category. The distinction between urban and urban core collection sites within the CMAs was based on the dissemination blocks from the census, which are the lowest level of geography used by the census. If a collection site within a CMA had at least 80% of its dissemination blocks designated as core, it was designated as an urban core collection site. If the rate was less than 80%, it was designated as an urban collection site.

Once all of the previous response rates were calculated, a simulation of 100 independent samples of dwellings was performed for the site being sampled. The goal of each of the 100 simulated samples was to use the expected response rates to predict if each sampled dwelling would result in 0, 1 or 2 people responding to the clinic. The final frame for the site was used for each simulation. The average expected number of clinic respondents for each age and sex group over the 100 independent samples was used to determine if the specified sample size and allocation were sufficient. The entire simulation of 100 independent samples was performed multiple times with varying sample sizes and allocations in order to settle on a final overall sample size and allocation strategy. The first iteration used approximate age/sex targets to come up with starting values for the overall sample size and the stratum allocation. Subsequent iterations required manually adjusting the overall sample size and stratum allocation, based on the previous iteration, to produce final values that would satisfy the clinic target counts.

The average sample size by site of cycle 4 was 533 dwellings, with a high of 615 and a low of 480. The sample was allocated amongst the 6 age group strata (3-5, 6-11, 12-19, 20-39, 40-59 and 60-79), with a small portion of the sample going to an "other" stratum. A maximum number of 35 dwellings per site were selected in this stratum, with fewer being selected for sites that had fewer dwellings in the stratum. This stratum size helped to prevent extreme dwelling sampling weights.

The allocation of the dwelling sample to each of the age group strata was done to allow for the best chance of meeting the age and sex clinic respondent targets for cycle 4 without going too far over. Where possible, the sample was allocated in a way that emphasized the strata where more sample was required to meet the targets.
Once the sample of dwellings was in the field, when the household interviewer made contact with a sampled dwelling, the goal was to create a roster for the household. A roster is a list of all persons residing in the household and includes pertinent information such as age, sex and whether the individual works full-time for the Canadian Forces. With this information, the computer application randomly selected one or two persons to take part in the remaining part of the survey, including the questionnaire and the clinic visit. The number of persons selected depended on the composition of the household:
• If there was at least one child between the ages of 3 and 11, two people were selected: one child between the ages of 3 and 11 and one other person between the ages of 12 and 79.
• If there were no children between the ages of 3 and 11, only one person in the 12 to 79 age group was selected.
• If there was no one eligible for the survey, no one was selected. This included households where all in-scope persons were under the age of 3, over the age of 79 and / or were full-time members of the Canadian Forces.

When the roster was completed, the computer application assigned a sampling factor to each eligible member of the household and this information was used to determine the probability of selection. The sampling factor assigned to each individual was based on their age group and sex and the factors varied between groups in order to do a better job of reaching the clinic targets for each age group by sex. In households where two people were selected, the selection of the child (aged 3-11) was done independently of the person aged 12 to 79.

Households with a large number of members can have very large survey weights and therefore be overly influential when calculating survey estimates. To prevent this, two reset values were put in place that would reset some of the sampling factors to 1 when certain conditions were met. In both cases described below, each member of the 12 to 79 age group would have had an equal chance of being selected:

1. If a household contained six or more people in the 12 to 79 age group, all sampling factors for the 12 to 79 members were reset to 1.

2. If a household contained three or more people in the 12 to 19 age group, all sampling factors for the 12 to 79 members were reset to 1.

With the person selection method described above, the average number of persons selected by site for the full sample of cycle 4 was 513 with a high of 559 and a low of 432.

Among the full sample respondents, several subsamples were selected.

For the Fasted subsample, each sampled dwelling was randomly flagged to indicate whether a respondent should fast prior to the MEC appointment. It required that respondents fast for at least 10 hours, whereas shorter eating restrictions were imposed on those with non-fasted appointments. Pregnant women, people with diabetes, youth less than 6 years old and other special cases were not asked to fast, even if the dwelling was flagged to be fasted. This random allocation reduced the potential for bias, which could occur if respondents were given the option to fast. During collection, the sampling rates were adjusted to obtain approximately half of the sample where respondents were selected to fast and were actually fasted prior to the MEC appointment.
For the Fatty Acids subsample, Respondents aged 20 to 79 attending the MEC were randomly selected to be included or excluded from the fatty acids subsample. The targeted subsample size was 2,000 respondents; 333 respondents by sex for the following age groups: 20 to 39, 40 to 59 and 60 to 79.

For the urine fluoride subsample, respondents aged 3 to 79 attending the MEC were randomly selected to be included or excluded from the urine fluoride subsample. The targeted subsample sizes were 500 respondents for 3 to 5 year olds, 250 respondents by sex from each of the following age groups: 6 to 11 and 12 to 19 and a total of 1,000 respondents equally distributed by sex and age groups for the following age groups (20 to 39, 40 to 59 and 60 to 79). The goal was to have a total subsample of 2,500 respondents. A MEC respondent could be selected for the subsample only if tap water was collected at his dwelling to measure fluoride in tap water. All households with a member in the 3 to 5 age group participating in the survey were selected to have the level of fluoride measured in their tap water. A little over half of the households with a member in the 6 to 11 age group participating in the survey were selected to have the level of fluoride measured in their tap water. For the other CHMS respondent households having only one person selected to participate in the CHMS, the probability of selection was set in order to reach the targeted number of respondents for the urine fluoride subsample.

For the blood volatile organic compounds (VOCs) subsample, respondents aged 12 to 79 attending the MEC were randomly selected to be included or excluded from the blood VOCs subsample. The selection of each respondent was done based on age, sex and the number of persons selected from the household to participate in the CHMS. The targeted subsample sizes were 375 respondents by sex for the 12 to 19 age group and a total of 1,750 respondents by sex for the other age groups (20 to 39, 40 to 59 and 60 to 79). The goal was to have a total subsample of 2,500 respondents. A MEC respondent could be selected for the subsample only if tap water was collected at his dwelling to measure VOCs in tap water. A little over 60% of the households with two persons selected to participate in the CHMS, were selected to have the level of VOCs measured in their tap water. For the other CHMS respondent households having only one person selected for the CHMS, the probability of selection was set in order to reach the targeted number of respondents for the blood VOCs subsample.

For the blood acrylamide subsample, respondents aged 3 to 79 attending the MEC were randomly selected to be included or excluded from the acrylamide subsample. The targeted subsample sizes were 250 respondents by sex for the following age groups: 3 to 5, 6 to 11, 12 to 19 as well as 500 respondents by sex for the 20 to 79 years old, for a total of 2,500 respondents.

For the methyl mercury subsample, respondents aged 20 to 79 attending the MEC were randomly selected to be included or excluded from the methyl mercury subsample. The targeted subsample size was 1,000 respondents; 500 respondents by sex for the age group 20 to 79.

For the urine environmental contaminants subsample, respondents aged 3 to 79 attending the MEC were randomly selected to be included or excluded from the urine environmental contaminants subsample. The targeted subsample sizes were 250 respondents by sex for the following age groups: 3 to 5, 6 to 11, 12 to 19 as well as 500 respondents by sex for the 20 to 79 years old, for a total of 2,500 respondents.

For the Indoor Air subsample, please refer to the selection of households for the the blood volatile organic compounds (VOCs) subsample as the same households are selected for both subsamples.

More details on the sample design can be found in 'Labrecque F. and Quigley A. (2016), Sampling documentation for cycle 4 of the Canadian Health Measures Survey.

Data sources

Data collection for this reference period: 2014-01-07 to 2015-12-16

Responding to this survey is voluntary.

Data are collected directly from survey respondents.

Collection includes a combination of a personal interview using a computer-assisted interviewing method and, for the physical measures, a visit to a mobile examination centre (MEC) specifically designed for the survey.

The CHMS will collect data in 16 sites across the country. The collection sites are located in seven provinces: Nova Scotia, New Brunswick, Quebec, Ontario, Saskatchewan, Alberta and British Columbia. Collection is scheduled so that each region is distributed within the two-year collection period, distributed between seasons and in a way that tries to minimize the movement of staff and equipment between sites. The CHMS MEC stays in each site for five to seven weeks, collecting direct measures from approximately 350 respondents per site.

First step: personal interview at the household

The first contact with respondents is a letter sent through the mail. The letter informs people living at the sampled address that an interviewer will visit their home to collect some information about the household.

At the home, the application randomly selects one or two respondents and the interviewer conducts a separate health interview with each of them. The interview takes 45 to 60 minutes per respondent. The interviewer then assists the respondent in setting an appointment for the physical measures at the CHMS MEC.

Also, for a subsample of households, interviewers take a small sample of tap water to measure the level of fluoride and/or the level of 10 different volatile organic compounds.

Second step: visit to the CHMS MEC

Statistics Canada uses MECs to conduct the physical measures portion of the survey. Similar MECs have been used successfully for years by the National Health and Nutrition Examination Survey (NHANES) in the United States.

The MEC consists of three trailers (side by side), linked by enclosed pedestrian walkways. One trailer serves as a reception and administration area, the second trailer contains a laboratory and physical measure rooms, while the third trailer contains additional physical measure rooms.

For each respondent, the complete visit to the MEC lasts about two hours. This is an approximate time, given that each respondent is assessed for their suitability for each measure and tested accordingly.

For children under 14 years of age, a parent or legal guardian has to be present with the child at the MEC and has to provide written consent for the child to participate in the tests.

At the end of their visit to the MEC, respondents are provided with a waterproof activity monitor. This small device is worn for a week at all times except when sleeping - even when swimming or bathing. It records information about normal physical activity patterns without the respondents having to do anything special.

A subsample of households is also asked to place an indoor air sampler, a small cylindrical device, in their home for the week following their visit to the MEC. The sampler measures a number of airborne substances in order to establish national baselines for indoor air concentrations of over 80 different volatile organic compounds.

View the Questionnaire(s) and reporting guide(s).

Error detection

Most editing of the data was performed at the time of the interview by the CAI application. It was not possible for interviewers/HMS to enter out-of-range values and flow errors were controlled through programmed skip patterns. For example, CAI ensured that questions that did not apply to the respondent were not asked. Edits requiring corrective action were incorporated in the CAI application to deal with inconsistent responses. In addition, warnings not requiring corrective action were also included to identify unusual (i.e., improbable rather than impossible) values as a means of catching potential errors and allowing correction at source.

At head-office, the data underwent a series of processing steps that resulted in some of the data being adjusted. As a final validation step, the CAI edits were re-applied to the processed data. As a result, the final data were complete and contained reserve codes for responses of "less than limit of detection", "valid skip", "don't know", "refusal" and "not stated".

Table 8.1 Reserve code of responses

Reserve Code label Reserve code
Less than limit of detection 9.5, 95, 99.5 etc.
Valid Skip 6, 96, 99.6 etc.
Don't Know 7, 97, 99.7 etc.
Refusal 8, 98, 99.8 etc.
Not Stated 9, 99, 99.9 etc.

Imputation

Household income data are imputed due to a high percentage of missing values. To impute the household income, as a first step, the modelled household income (an auxiliary variable), is first created using the personal income of each member of all responding CHMS households obtained from administrative files. The personal income is then summed for each household to create the modelled household income. This variable is then used to impute the household income using nearest neighbour imputation. The modeled household income defined above is used as a distance measure to determine which pair of respondent-non-respondent records is the "nearest" within imputation classes. The data from the respondent, or donor, is then copied to the non-respondent or recipient. For respondents who provided an income range, a nearest neighbour is selected within the same income range and household size. For respondents who did not provide any income range, the donor record is selected within the same collection site and household size.

More details on the method of imputation can be found in 'HSMD (2016), Household Income Imputation for Cycle 4 of the Canadian Health Measures Survey'

Estimation

In order for estimates produced from survey data to be representative of the population covered and not merely of the sample itself, users must incorporate weighting factors (survey weights) into their calculations. A survey weight is assigned to each person included on the final dataset, that is, in the sample of persons who responded to the entire survey. This weight corresponds to the number of people represented by the respondent in the population as a whole.
The survey weight is calculated as the inverse of the probability that the respondent was selected for the survey. The Canadian Health Measures Survey (CHMS) is a multi-stage sample. The probability of selection for the survey is determined by multiplying the probability of selection at each stage.

In accordance with the weighting strategy, the selection weights for collection sites are multiplied by the selection weights for dwellings (households) and adjusted for non-response. The weight of non-respondent households is redistributed to respondents within homogeneous response groups (HRGs). In order to create these HRGs, a method based on logistic regression is used: first a logistic regression model is created to estimate the response probability, and then these probabilities are used to divide the sample into groups with similar response properties. The logistic regression models are created from the limited amount of information available for all households. This includes data from the frame such as the strata, and geographic location, and paradata about the data collection such as the number of attempts to contact the household and the elapsed time between the first and last contact. An adjustment factor was then calculated within each HRG. The weight of respondent households is multiplied by this adjustment factor to produce the adjusted household weight.

Since the final sampling unit for the CHMS is the person, the adjusted household weight up to this point must be converted into a person weight. This is obtained by multiplying the adjusted household weight by the inverse of the probability of selection of the person selected in the household.

The selected person is asked to complete an interview. In some cases, interviewers do not succeed in completing it either because they cannot contact the person(s) selected, or because the person or persons selected refuse to be interviewed. Such cases are defined as non-responses at the questionnaire level, and an adjustment factor must be applied to the weights of respondent persons to compensate for this non-response. Just as for non-response at the dwelling (household) level, the adjustment is applied within classes defined by a method using response probabilities from a logistic regression model. The model is based on the characteristics available for all respondents and non-respondents, which includes all the characteristics collected when the members of the household are listed, such as the number of persons in the household, in addition to geographic information and paradata. An adjustment factor is calculated within each class. The weight of respondent persons is multiplied by this adjustment factor.

Respondents to the questionnaire are then invited to go to the CHMS Mobile Examination Centre for physical measurements. In some cases, people refuse to participate or do not keep their appointment at the MEC. Such cases are defined as non-responses at the MEC level, and to compensate for this non-response, an adjustment factor must be applied to the weights of the MEC participants. Just as for non-response at the dwelling (household) and questionnaire levels, the adjustment is applied within classes defined by their probability of attending the MEC. This probability is obtained from a logistic model using the characteristics available for respondents and non-respondents. All the characteristics collected on the questionnaire during the interview (such as income class, whether or not the respondent is employed, general health status, and frequency of smoking), in addition to geographic information and paradata, were made available to create the non-response models. An adjustment factor is calculated within each class. The weights of the persons participating at the MEC were accordingly multiplied by this adjustment factor.

The next step is calibration. This procedure is applied to ensure that the sum of the final weights corresponds to the estimates of populations defined at the scale of the five Canadian geographic regions, for each of the 12 age-sex groups of interest, the six age groups 3 to 5, 6 to 11, 12 to 19, 20 to 39, 40 to 59 and 60 to 79 for each sex. An additional criterion was used to calibrate the 20 to 39 age group to compensate for the fact that persons in this age group living with kids have a greater chance of being selected than those living without kids. In households where there was at least one person aged 3 to 11, a second person aged 12 to 79 was selected for the survey. The second person selected was usually a parent aged 20 to 39. To compensate for any potential bias caused by the selection method the 20 to 39 age group was split into those living with and without children aged 3 to 11. The population estimates are based on the most recent census counts, as well as on counts of births, deaths, immigration and emigration since then. The calibration was carried out using the mean of the monthly estimates (covering the survey period) for each cross-tabulation of standard regional boundaries and age-sex groups. The population estimate for the 20 to 39 age group in each region was split into those living with and without kids aged 3 to 11 based on the estimated ratio of 20 to 39 year olds living with and without kids from the sampling frame for cycles 1, 2 and 3.

Note that following a series of adjustments applied to the weights, it is possible that some units will have weights that stand out from the other weights to the point of being aberrant. Some respondents may actually represent an abnormally high proportion in their group and therefore strongly influence both the estimates and the variance. To avoid this situation, a respondent weight that contributes aberrantly to the age-sex group is adjusted downward using a method known as "winsorization." In this process, respondent weights that are considered to be outliers are replaced by the highest non-outlier weight for that age and sex group. All of the weights are then adjusted to redistribute the surplus weight (the part of the weight that is higher than the highest non-outlier weight). This is done by multiplying the non-outlier weights by an adjustment factor to create the winsor adjusted weights.

A second calibration (an exact repetition of the first calibration) is done on the winsorized weights to produce the final weight.

The CHMS uses a complex sampling design to select the sample and there are no simple formulas that can be used to calculate the variance of the survey estimates. Instead, a re-sampling approach known as the bootstrap method is used to approximate the sample variance. The bootstrap method involves creating subsamples of the full sample by randomly selecting « n-1 » collection sites with replacement among the « n » collection sites in each region. An adjusted weight is then calculated for each respondent in the selected subsample. This is repeated 500 times to create the bootstrap weights. To calculate the variance of a point estimate (such as the mean), the estimate for each of the 500 replicates is calculated using the bootstrap weight. The variability among the 500 estimates gives the variance estimate.
For the subsamples, additional weighting steps are done.

The fasted subsample was selected when the sample of dwellings were selected, and thus occurred prior to completion of the household questionnaire. To create the fasted subsample weights, the subsample flags that were assigned to the dwellings were attributed to the selected person(s). Before adjusting for non-response at the questionnaire level, the person weight of those selected for the fasted subsample was adjusted to incorporate the subsample sampling weight. An additional step was required to adjust for persons who were selected for the subsample but who did not fast or did not provide blood. Such cases were defined as non-respondents to the fasted subsample and to compensate for this non-response an adjustment factor was applied to the weights of the persons with a valid measure.

For the Fatty Acids subsample, three additional adjustments were applied to the weights from the full sample to adjust for respondents not selected for the subsample and to account for non-response to the subsample. First, the weight of the respondents not selected for the subsample was redistributed to the weights of the selected respondents using the following adjustment factor within each combination of site, age group and sex. The next two adjustments were applied on weights to account for non-response to the subsample, which occurred when a respondent did not provide blood or a valid measure could not be obtained on at least one of the laboratory tests.

To create the urine fluoride subsample weights, two adjustments to the household weights are applied to take into account the fact that a respondent could only be selected for the urine fluoride subsample if tap water was collected at the dwelling. First, the weights of the households selected for the tap water fluoride subsample were adjusted to incorporate the subsample sampling weight. Secondly, an adjustment was made to the household weights to account for tap water fluoride non-response. Then household weights were converted to person weights and the person weights were adjusted for questionnaire and MEC non-response.Weights for respondents selected for the urine fluoride subsample were adjusted to incorporate the subsample sampling weight. The weights were then adjusted for non-response to the fluoride measure in urine.

To create the blood VOCs subsample weights, two adjustments to the household weights are applied to take into account the fact that a respondent could only be selected for the blood VOCs subsample if tap water was collected at the dwelling. First, the weights of the households selected for the tap water VOCs subsample were adjusted to incorporate the subsample sampling weight. Secondly, an adjustment was made to the household weights for tap water VOCs non-response. Then household weights were converted to person weights and the person weights were adjusted for questionnaire and MEC non-response. Weights for respondents selected for the blood VOCs subsample were adjusted to incorporate the subsample sampling weight. The weights were then adjusted for non-response to the VOC measures in blood.

A separate weight was created for the activity monitor data for respondents who had at least 4 days of valid data (3 days of valid data for youths 3 to 5 years old). Respondents who did not have the required number of valid days were treated as non-respondents to the activity monitor. The weights of respondents with the required number of days were adjusted to compensate for any bias due to this non-response.

To create the blood acrylamide subsample weights, two additional adjustments were applied to the weights from the full sample to adjust for respondents not selected for the subsample and to account for non-response to the subsample. First, the weight of the respondents not selected for the subsample was redistributed to the weights of the selected respondents using the following adjustment factor within each combination of site, age group and sex. The second adjustment was applied to the weights of the respondents selected for the subsample to account for non-response, which occurred when a respondent did not provide blood or a valid measure could not be obtained for at least one of the laboratory tests.

To create the blood methyl mercury subsample weights, two additional adjustments were applied to the weights from the full sample to adjust for respondents not selected for the subsample and to account for non-response to the subsample. First, the weight of the respondents not selected for the subsample was redistributed to the weights of the selected respondents using the following adjustment factor within each combination of site, age group and sex. The second adjustment was applied to the weights of the respondents selected for the subsample to account for non-response, which occurred when a respondent did not provide blood or a valid measure could not be obtained for methyl mercury.

To create the urine environmental contaminants subsample weights, two additional adjustments were applied to the weights from the full sample to adjust for respondents not selected for the subsample and to account for non-response to the subsample. First, the weight of the respondents not selected for the subsample was redistributed to the weights of the selected respondents using the following adjustment factor within each combination of site, age group and sex. The second adjustment was applied on the weights of the respondents selected for the subsample to account for non-response, which occurred when a respondent did not provide urine or a valid measure could not be obtained on at least one of the laboratory tests.

For the Indoor Air subsample (IAS), first, the weights of the households selected for the IAS subsample were adjusted to incorporate the subsample sampling weight. A second adjustment was made to the household weights for the IAS non-response to obtain the final household level weights. To obtain the IAS person-level weights, the final household weights were multiplied by the persons' selection weight and adjusted for questionnaire and MEC non-response.

Quality evaluation

One of the unique features of the Canadian Health Measures Survey (CHMS) is that three different sets of data are collected for the same respondent: household interview data, physical measures data, and laboratory results data. Each set of data has to be processed on its own, yet they cannot be completely separated from each other because at various points during processing the three sets of data have to be used together.

The processing of the household interview data was performed in a manner similar to that of other health surveys at Statistics Canada. The data are validated first at the record level and then at the individual variable level, followed by detailed top-down editing. During data collection, processing takes place on a daily basis. The household interview responses have to be processed quickly in order for the data to be available at the mobile examination centre (MEC) in time for the respondent's visit to the MEC.

Similarly, the processing of the physical measures data begins with the data being validated first at the record level and then at the individual variable level, followed by detailed top-down editing. Also, because the laboratory tests are determined based on responses received at the MEC, the MEC data are used to generate a file containing a list of the tests for which laboratory results are expected to be received. This laboratory "control" file is used in processing the laboratory results data.

The processing of the laboratory data involves significant file manipulation due to the fact that several different file types are received from the MEC and the various reference laboratories. As with the household and the physical measures data, the laboratory data are validated at the record level and at the individual variable level, and several new variables are then derived. The laboratory data are processed as quickly as possible so that any results that have been identified as outside of a normal range at the reference laboratories and the MEC are available in a timely fashion for reporting to respondents.

Disclosure control

Statistics Canada is prohibited by law from releasing any information it collects which could identify any person, business, or organization, unless consent has been given by the respondent or as permitted by the Statistics Act. Various confidentiality rules are applied to all data that are released or published to prevent the publication or disclosure of any information deemed confidential. If necessary, data are suppressed to prevent direct or residual disclosure of identifiable data.

It is Statistics Canada practice to remove personal identifiers from statistical master files when they are no longer required for data processing or other authorized purposes.

Residual suppression is used to protect the confidentiality of the respondent when results are calculated with less than 10 respondents in that category.

Revisions and seasonal adjustment

This methodology does not apply to this survey program.

Data accuracy

In terms of data accuracy, for the full sample, the survey aims at producing unbiased national estimates with a coefficient of variation (c.v.) of 16.5% or less for each of the 5 age groups (6-11, 12-19, 20-39, 40-59, and 60-79) by sex and for 3-5 year olds of both sexes combined. Examples of estimations and accuracy measures (c.v.) based on the full sample are provided for physical measures done at the mobile examination center (body mass index, ratio of forced expiratory volume), for a selected non-environmental measure on blood (High-density lipoprotein cholesterol) and for a selected environmental contaminant measure on blood (lead). As well estimations and accuracy measures are provided for the various subsamples: the fasted subsample (glucose), the fatty acids subsample (omega 3 index), the urine fluoride subsample (fluoride adjusted for creatinine), the blood volatile organic compounds subsample (total xylenes), the activity monitor subsample (average time spent sedentary), the acrylamide subsample (acrylamide hemoglobin adduct), the methyl mercury subsample (methyl mercury), the urine environmental contaminants subsample (dimethylarsinic acid adjusted for urine creatinine) and the indoor air subsample (benzene concentration). They can be seen at the link below - " Additional documentation-Data accuracy".

Response rates
The response rate for the full sample of the Canadian Health Measures Survey (CHMS) is derived in the following manner. Of the 7,792 in-scope dwellings, 5,956 agreed to provide information on the composition of the household, for a household response rate of 76.4 %. From these respondent households, 8,200 persons were selected (one or two persons per household) to participate in the survey, of whom 7,503 responded to the questionnaire, for a response rate of 91.5%. Of these persons, 5,794 then reported to the CHMS mobile examination center (MEC) for physical measurements, for a response rate of 77.2 %. At the Canadian level, a combined response rate of 53.7 % was observed for cycle 4 of the CHMS. It is important to note that the combined response rate is not obtained by multiplying the response rates at the person and household levels (or questionnaire level and the MEC level), since two persons were selected in some households.

Not all the participants who goes to the MEC are able or willing to participate in the blood and urine components. Therefore a response rate is derived for each of these components that were supposed to be done on the full sample respondents. The response rates for these measures use the full sample response rates up to the MEC and derive the rest in the following manner. Of the 5,794 participants who reported to the CHMS MEC for physical measurements, 5,551 participants provided blood and 5,680 provided urine. The combined response rate for blood draw was 51.8% whereas the combined response rate for urine was 52.8%.

The response rate for the fasted subsample uses the full sample response rates up to the responding household and the rest is derived in the following same manner. From the 5,956 respondent households, 3,453 were selected for the fasted subsample. In these households, 4,275 persons 6 to 79 years of age were selected (one or two persons per household) to participate in the fasted subsample, of whom 3,899 responded to the questionnaire, for a response rate of 91.2%. Of these persons, 2,944 then reported to the CHMS MEC for physical measurements, for a response rate of 75.5%. Of these persons, 2,426 were actually fasted and had a valid measure on at least one of the laboratory tests. At the Canadian scale, a combined response rate of 44.2% was observed.

The response rate for the fatty acids subsample uses the full sample response rates up to the MEC and derive the rest in the following manner. Of the 5,794 participants who reported to the CHMS mobile examination centre (MEC) for physical measurements, 2,122 people 20 to 79 years of age were selected for the red blood cell fatty acids subsample. Of these persons 2,041 had valid laboratory results. The combined response rate for the red blood cell fatty acids subsample was 51.4%.

The response rates for the fluoride subsamples (tap water and urine) were derived in the following manner: of the 5,956 respondent households, 3,222 were selected for the tap water fluoride subsample. Of these, 2,755 provided tap water and 2,212 had valid tap water fluoride laboratory results for a household response rate of 52.5%. From the 2,755 respondent households that provided water for the tap water fluoride subsample, 2,673 respondents were selected for the urine fluoride subsample and 2,574 of them provided urine and had a valid fluoride measure. At the Canadian scale, a combined response rate of 49.6% was observed.

The response rates for the volatile organic compounds subsamples (tap water and blood) were derived in the following manner: of the 5,956 CHMS respondent households, 4,149 were selected for the tap water VOCs subsample. Of these, 3,477 provided tap water, and among those 2,601 households had valid tap water VOCs laboratory results for a household response rate of 47.9%. From the 3,477 respondent households who provided water for the tap water VOCs subsample, 2,643 respondents were selected for the blood VOCs subsample and 2,527 of these provided blood and had valid VOCs measures. At the Canadian scale, a combined response rate of 48.5% was observed.
The response rate for the activity monitor was derived in the following manner: of the 5,794 participants who reported to the CHMS mobile examination centre for physical measurements, 5,752 participants were offered an activity monitor. Of these persons, 4,118 returned the activity monitor with at least four days of valid entries for 6-79 year olds or at least 3 days of valid entries for 3-5 year olds. The combined response rate for the activity monitor was 37.7%.

The response rate for the acrylamide subsample was derived in the following manner: of the 5,794 participants who reported to the CHMS mobile examination centre (MEC) for physical measurements, 2,704 people 3 to 79 years of age were selected for the acrylamide subsample. Of these persons 2,529 had valid laboratory results. At the Canadian scale, a combined response rate of 51.2% was observed.

The response rate for the methyl mercury subsample was derived in the following manner: of the 5,794 participants who reported to the CHMS mobile examination centre (MEC) for physical measurements, 1,074 people 20 to 79 year old were selected for the methyl mercury subsample. Of these persons 1,043 had valid laboratory results. At the Canadian scale, a combined response rate of 51.9% was observed.

The response rate for the urine environmental contaminants subsample was derived in the following manner: of the 5,794 participants who reported to the CHMS mobile examination centre (MEC) for physical measurements, 2,658 people 3 to 79 year old were selected for the urine environmental contaminants subsample. Of these persons 2,570 had valid laboratory results. At the Canadian scale, a combined response rate of 52.0% was observed.

The response rate for the indoor air subsample (IAS) was derived in the following manner: of the 5,956 responding households, 4,149 were selected for the IAS subsample. Of these households, 3,831 had a least one person complete the household questionnaire. Of those households, 2,977 households had at least one person who reported to the CHMS mobile examination centre for physical measurements and 2,951 of these persons were offered an IAS to place in their home. Of those households, 2,427 returned the IAS and had a valid exposure time between 4 and 10 days. At the Canadian scale, a combined response rate of 45.1% was observed.

Non-response bias
In addition to increasing variance (accuracy), non-response can result in biased estimates if non-respondents have different characteristics from respondents. To reduce the number of non-response cases and to ensure that procedures are followed consistently, the interviewers are all extensively trained by Statistics Canada, provided with detailed interviewer manuals, and are subject to regular observations. Refusals are followed up by an interviewer manager to encourage respondents to participate in the survey. Total non-response is handled by adjusting the weight of households that responded to the survey to compensate for those who do not respond.
Response rates for selected variables can be seen at the link below - " Additional documentation-Response rates".

Coverage error
The CHMS covers the population 3 to 79 years of age living in the ten provinces. Excluded from the survey's coverage are: persons living in the three territories; persons living on reserves and other Aboriginal settlements in the provinces; full-time members of the Canadian Forces; the institutionalized population and residents of certain remote regions. Altogether these exclusions represent approximately 4% of the target population.

Documentation

Date modified: