Methodology
Panel recruitment and incentives
Opinion Health has built its proprietary panels of GPs and patients to represent a well-balanced cross-section of the total GP and patient population in the UK. We are currently working to further expand these panels by developing panels of secondary-care doctors, pharmacists and nurses. Our patient panels are a particular focus of development through our network of 75,000 patient groups across the world. The panels are actively managed to ensure ongoing member cooperation and quality responses.
As part of the registration process, each panellist completes an initial profile survey to collect demographic, health and behavioural information. The data collected at the registration point and via the initial profile questionnaire, is then stored as master data to use it to select representative samples of healthcare professionals or patients to provide clients with pre-segmented samples according to selection criteria. The wealth of information kept on each panellist gives us the possibility to quickly run sample feasibility. Profiling survey updates are sent out on a regular basis to make sure we have the most up-to-date information on each panellist.
In order to guarantee a diverse flow of panellists, Opinion Health recruits its panels through a wide range of marketing activity and partnerships —online and offline advertising, direct mail, public relations, search marketing, events and word of mouth. Through this large range of recruitment activity we can ensure a consistent flow of diverse and broadly based panellists to guarantee that the appropriate socio-demographic balance is kept on the panel.
Verification of Respondents
Each GP is required to enter his/her GMC number along with name, practice address and postcode. As far as patients are concerned, each panelist verifies their email address and reward cheques are sent to the address provided in their profile, thus guaranteeing the authenticity of the identity of each panellist. We also conduct spot checks on profile questions to make sure consistency of responses. Should this not be the case, we ask the respondent to provide the right information and if we think this is still not the case, we’ll proceed to remove the panelist.
Sampling
When we conduct a survey with such a large and diverse panel, we select a sub-group that is representative of the UK GP/consumer or patient population and invite this sample to complete the survey. Respondents cannot complete the survey more than once.
We also make sure that panellists are not over contacted and over surveyed. For that reason, we try, where possible, to limit the surveys they take part to one per week. We also regularly monitor the composition and response behavior to identify any persistent non-responders or those panellists who give quick and random answers.
Data Weighting
To be able to make inferences about the population from the information we have collected in the sample, the data is weighted to the profile of the population we are surveying, thus ensuring that the results properly reflect the population we are trying to measure. For example, male GPs comprise 60% of the UK GP population and females 40%. If however, a project resulted in a split of say 58 to 42% we would then weight the results to reflect the true demographic split. All other relevant socio-demographic or behavioral variables are simultaneously taken into consideration for weighting purposes.
Statistical reliability and sample size
Because results are subject to the normal error margins for sample surveys, we run our Omnibus with not less than 400 GPs, as tabulation and routing used can result in relatively small sub-samples for which the error margin is higher than for the total sample. The table below illustrates the possible variations between the sample results and the "true values" (if the entire population were surveyed). On a question where 50% of the people in a sample of 1000 give a particular answer, we can be 95 percent confident that the result would not vary by more than 3 percent (plus or minus) from surveying the entire population.
| 10% or 90% | 30% or 70% | 50% | |
|---|---|---|---|
| Sample size | ± | ± | ± |
| 100 | 6 | 9 | 10 |
| 200 | 4 | 6 | 7 |
| 500 | 3 | 4 | 4 |
| 1000 | 2 | 3 | 3 |
