Contents Previous (Part I) Next (Part III Section 1)
The organizational responsibilities of a risk factor survey can usually be divided into:
There does not have to be an absolute separation of these functions. Conceptualization and Operation are often performed by the same agency. However, it is recommended that the quality control should be carried out by an organization without vested interest in the survey.
The target population for a health risk monitoring survey can be defined in various ways. Usually the recommendation is that the target population comprises all residents of a specified age group in a geographically defined population. However, the concept of resident is sometimes ambiguous. It has to be defined if the target population includes only permanent residents or also temporary residents. In some situations the use of residents in defining target population is not relevant and other possibilities such as use of citizens could be considered. If any exclusions apply, such as leaving out institutionalized persons or prisoners from the target population, they have to be carefully documented.
The recommended core age group for cardiovascular risk factor monitoring surveys is the population between 35 to 74 years of age. In addition, age groups 25-34 is relevant and highly recommended.
It is recommended to aim at nationally representative surveys.
Sampling frame is the list of sampling units from which the sample is selected. The sampling frame should represent the target population of the survey as precisely as possible. Keeping this aim and availability in mind, the most complete register that includes the persons defined as target population of the survey can be chosen among population registers, electoral lists, census lists or public health registers.
Probability sampling must be used. Otherwise the statistical accuracy of the population estimates cannot be assessed.
It is recommended to use a single-stage sampling scheme. However, if a single sampling frame representing the target population is not available, or if it is logistically necessary to concentrate the survey examinations in selected sub-areas of the population, a multistage sampling could be used. The disadvantage of multistage sampling is that, compared with a single stage sampling, the analysis of the data is more complex, and the sample size may have to be a little larger.
If analyses for specific population subgroups or geographical areas are anticipated, stratification is recommended to guarantee sufficient representation of the subgroups in the survey. By suitable stratification one can sometimes also improve the statistical power of the sample without increasing the sample size.
It is highly recommended to use a survey statistician in planning and implementation of sampling.
The sample size should be assessed according to the smallest sub-population for which information is wanted. In order to get statistically precise estimates of cardiovascular risk factors each sub-population group should include about 200-250 subjects. For example, if 10-year age groups and sex are kept as minimum sub-groups the total sample size in age group 35-74 should be at least 1600-2000 per geographical survey area. For larger samples, the achievable accuracy of measurements becomes the limiting factor.
Another factor which influences the statistical precision is the sampling scheme. If multi-stage sampling is used, the statistical precision is smaller than using a single stage sampling. However, if the sampling is well planned, the reduction in the precision is usually small.
To achieve high data quality one of the most critical points is the response rate of the survey. It is recommended that the response rate should be at least 70 % for all population sub-groups of interest.
Measures to obtain as high response rates as possible should be carefully planned beforehand as part of the survey planning process. Several contacts to non-participants are recommended to achieve a high participation rate. Contacts could be made both by mail and telephone. Some special methods to attract participants into the survey could also be considered.
Note that increasing the sample size has no effect on non-response bias, and therefore cannot be used as a method to compensate for non-response. Instead, every reasonable effort should be made to achieve a high response rate.
For repeat surveys an interval of about 5 years is recommended. More frequent surveys should be considered only if there are specific reasons for that.
There is no general rule for the optimal duration of a survey, which may last from few months to several years. A survey lasting a full calendar year has the advantage of adjusting the results for seasonal variation. If the survey lasts more than a few months, particular attention needs to be paid to the temporal coverage of quality control, re-testing and re-training of the measurers, and to distributing the dates of examination of all population subgroups (such as defined by age, sex or region) evenly over the whole survey period.
A short survey duration usually needs a relatively large temporary staff, whereas long surveys allow a more stable employment of the core staff. If a survey lasts several years, the results cannot be fresh when they become available. Also the time between sampling and examination becomes long, and therefore the representativeness of the sample will be reduced.
Repeated surveys should always be carried out at the same time of the year to minimize the seasonal effect. Also the duration of the survey should be the same. The dates of examination of all population sub-groups of interest (such as defined by age, sex or region) should be distributed evenly over the whole survey period.
It would be ideal to make measurements divided equally on every weekday, but this is usually inconvenient in practice. Commonly, measurements are conducted during the working days (Monday-Friday).
There is also variation in risk factors depending on time of the day. Therefore, the hours of examinations should be kept constant, and the examination times of all population subgroups should be distributed evenly throughout these hours. If the survey includes analyses that demand fasting blood samples, the examinations can be carried out conveniently only during morning hours. Therefore, it is often advisable to do measurements requiring fasting only on a subsample.
Because of the importance of easy recruitment of the participants, the survey examinations should take place near the subjects' place of work or residence. As a consequence, the examination centre needs to move often, or several survey teams may be needed. In national surveys, several examination teams are usually needed, and each of them will have to move from place to place. It is then advisable also to rotate the survey teams between different regions. This will minimize the effect of survey teams in the regional comparability of the surveys.
The mobility will need to be taken into account in reserving the equipment for the teams, and the logistics of the survey will need to be planned carefully, including the requirements of transportation the blood samples from the examination site to the laboratory, and the special needs of data management in order to avoid loss of data and to facilitate prompt data quality control.
Local health centres are often able to provide suitable premises, or specific mobile clinics may be equipped for the surveys.
The order of measurements is often constraint by logistical requirements, such as subject flow and examination duration. As much as possible, the order should be determined by:
At the planning stage of each survey, a manual of operations that considers all the requirement for local survey management needs to be prepared. Operational guidelines that assure standardized and stable measurements needed for comparison between countries and for trend monitoring are provided in Part III.
The Operational guidelines also give instructions for selection and training of personnel for the different measurements. Some of the personnel may be recruited and trained locally at the different examination sites. However, personnel for tasks where the standardization requires long training and continuity, such as the blood pressure measurers and the phlebotomists performing the blood collection, should move with the examination site. When the survey area covers several regions, it is advisable to rotate the blood pressure measurers between them. This will minimize the effect of measurers in the regional comparability of the surveys.
The quality control during the survey should be planned in such a way that it can detect problems in the measurements and data management soon after occurred. If possible, the problems should be detected early enough to be remedied, and anyway before erroneous results from the survey are reported. In principle, all situations where information moves from one form to another (e.g. person's weight from the numbers on the scale to the data collection form) or from one place to another (e.g. blood samples from the subject to the laboratory, or data from the examination site to the data centre) should be subjected to quality control.
The quality control can be relatively simple, provided that it is planned properly. Instructions for the quality control of the different measurements are given in the Operational guidelines in Part III.
At the end of the survey, the attained quality should be assessed and documented carefully. This will facilitate correct interpretation of the results and help in improving future surveys.
All survey protocols should comply with the principles outlined in the Helsinki Declaration (1). In addition, survey protocols may be required to obey more specific guidelines for the ethical conduct of research on human subjects that have been created by several countries. These guidelines apply to the development of informed consent, risk/benefit assessment, confidentiality of data, subject selection and also specify rules that guard against scientific misconduct.
Survey protocols need to be approved by local Ethics Review Committees
Surveys have legal obligations vs. their personnel and the survey participants. Obligations toward personnel involve harm that personnel may suffer in connection with their work (example: getting infected during venipuncture, suffering injury while lifting survey equipment). It also involves physical or emotional harm that personnel may inflict on survey participants as a result of survey procedures or negligence (examples: fainting and falling during venipuncture, unauthorized disclosure of confidential information).
Obligation toward survey participants involves provision of proper facilities (e.g. privacy in the examining area), minimizing risks of attending the survey (e.g. proper access for handicapped) and protecting subject information through proper data processing facilities.
In order to meet these obligations surveys organizers have to pay attention to selection and the training of personnel and to the formulation of consent forms. In addition, preventive measures such as vaccinations and protection through special insurance coverage may have to be considered.
It is recommended that the compliance with the ethical aspects of the survey protocol be monitored during the survey as part of the overall quality control system.
References
1. World Medical Association Declaration of Helsinki. Ethical principles for medical research involving human subjects. Bull World Health Organ 2001;79(4):373-4.
Contents Previous (Part I) Next (Part III Section 1)