XVI.Add.II.3.2. Survey methods

A survey may be conducted to evaluate dissemination of RMM tools, risk knowledge and behavioural outcomes, provided adequate survey methodology is applied.

Sampling and recruitment of survey participants should ensure that the study population is similar to and hence representative of the target population and avoid selection bias due to dissimilarity in one or several relevant aspects. For example, where marketing authorisation holders rely on prescribers to recruit patients, efforts should be made to mitigate the potential for selection bias introduced by e.g. another source for recruiting patients. Selection bias may also occur if webbased survey technology that excludes participants less familiar with internet technology is used.

Bias may be minimised by selecting the optimal sampling frame, accounting for the expected response rate, age, sex, geographical distribution, and additional characteristics of the study population, and by achieving similar response rates across diverse participants to minimise nonresponse bias. As response rates in health surveys are generally low, continuous sampling may be necessary until the pre-defined sample size has been met, and additional measures that improve response rates (31) may be considered.

Bias may also be minimised by assuring that the sample contains appropriate diversity to allow stratification of results by key population characteristics (e.g. by oversampling a small but important subgroup). For example, in a physician survey, the sampling strategy should consider whether a general random sample would be sufficient, or if the sampling frame should be stratified by key characteristics such as specialty, type of practice (e.g. general practitioner, specialist or hospital care). In a patient survey, characteristics such as socio-economic status and education, medical condition(s), and chronic versus acute use of medicines should be considered for optimising the sampling frame.

The recruitment strategy should also consider that accurate and complete data collection is achieved. Efforts should be made to document the proportion of non-responders and their characteristics to evaluate potential effects on the representativeness of the sample.

Surveys often collect and analyse self-reported data, thus introducing misclassification of exposure or recall bias when participants do not remember previous events or experiences accurately or omit details. Respondents may also improve or modify an aspect of their reported behaviour in response to their awareness of being surveyed.

The data collection instrument should be designed to avoid desired-response-bias (e.g. multiplechoice response options with obvious desired response), to cover all relevant aspects of the RMM and to be able to identify different levels of risk knowledge and attitude. For data collection instruments to be considered reliable the following principles should be adhered:

• Pre-testing and validation: Testing the draft instrument in samples of participants that should be similar to the study population identifies questions that are poorly understood, ambiguous, or produce invalid responses. Pre-tests should be carried out using the same procedures that will be used when applying the data collection instrument to the study population.
• Content validity: Items or variables included in the data collection instrument should capture all aspects related to end-users’ risk knowledge and attitudes relevant to the RMM. It is also important that the items or variables are clear and unambiguous and that questions pertaining directly to the implemented regulatory action are avoided (e.g. “do you know that product X is contraindicated for disease Y?”) and non-leading questions are used.
• Construct validity: Items or variables in the data collection instrument should be developed in a way that they are likely to accurately measure (at different degrees) end-users’ risk knowledge and attitudes relevant to the RMM.

Surveys may be analysed quantitatively including:

• Descriptive statistics, such as:
  • Response rate (i.e. proportion of participants who responded of the total number of invited participants);
  • Rate of incomplete responses among responding participants;
  • Pooled proportion of participants responding correctly to the questions;
  • Stratification by selected characteristics such as RMM target population (e.g. healthcare professional or specialist, patient, carer), geographic region, receipt, and type of RMM;
• Comparison of responder and non-responder characteristics (if data is available);
• Comparison of responders and overall RMM target population characteristics;
• Comparison of characteristics of responders with correct and incorrect answers.

Information collected as free text may also be analysed qualitatively, e.g. using thematic content analysis techniques by identifying common recurrent themes or topics.

To obtain valid survey results, a weight may have to be attached to each respondent considering the following:

• Differences in selection, e.g. if certain subgroups were over-sampled;
• Differences in response rates between sub-groups;
• Differences of responders compared to target population (e.g. healthcare speciality, volume of prescribing);
• Clustering

Variations among healthcare settings in Member States may pose challenges to implementing survey studies in several Member States due to time constrains for determining and complying with national ethical and data protection requirements. Therefore, early feasibility assessment is paramount in the successful implementation of a survey. National (or regional) requirements for providing incentives to survey participants also need to be accounted for.

There may be also data protection requirements when healthcare professionals are contacted based on a prescriber list of a marketing authorisation holder.

Although survey studies aimed at evaluating risk knowledge and attitudes do not attempt to collect patient health-related information, patients who complete the survey are likely to have received the medicinal product revealing their condition/disease. Therefore, unless the patient response is completely anonymous, data protection regulation applies, and informed consent must be provided.

Survey studies must follow the provisions of the legislation on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, as laid down in Regulation (EU) 2016/679 (General Data Protection Regulation) and Regulation (EU) 2018/1725 of the European Parliament and of the Council and require approval(s) by the relevant body(ies) in Member States.