Health and Care Experience Survey 2023/24 Technical Report

Annex B: Weighting Methodology

When conducting a survey, it is important to have a representative sample of the population you are interested in to reduce bias in the estimates produced. Applying weighting methods reduces the potential bias by making the results more representative of the population but also increases the variation in the estimates.

Survey weights are numbers associated with the responses that specify the influence the various observations should have in the analysis. The final survey weight associated with a particular response can be thought of as a measure of the number of population units represented by that response.

The weight for each section is calculated in two steps:

1. Weight for unequal probability of selection by weighting respondents up to the total population in each General Practice.

2. Weight for non-response by weighting up to the entire population using the service based on age and sex.

a. The number of people in the population using their General Practice will be estimated using the number of people consulting their GP or practice nurse from the 2012/13 Practice Team Information (PTI) Statistics.

b. The ‘population total’ is calculated as the whole sampling frame used for the survey.

c. The estimated population totals for all other sections will be based on the proportions of the relevant survey responses at a national level, applied at all levels of reporting.

d. Respondents will be assigned to one of 6 age groups (16-24; 25-34; 35-44; 45-54; 55-64; 65+) based on their age in the CHI database. This will be used for the weighting calculations and analysis of responses by age.

e. The sex of each respondent will be taken from the CHI database to ensure the weighting calculations are not impacted by missing data.

Survey results are published at National, Health Board, Health & Social Care Partnership, General Practice Cluster and General Practice levels. The weighting methodology will be applied to all reporting levels with the following adjustments:

• HB/HSCP Level – As described but with three age categories (rather than the six used at national level): 16-44; 45-64 and 65+.
• General Practice Cluster/Locality Level – As described but with two age categories (rather than the six used at national level): 16-54 and 55+.
• General Practice Level – As General Practice was the sample strata used, there is no need to calculate a weight for unequal probability of selection. Instead, the weights at this level should just equal the weight for non-response as described for General Practice Cluster level.

Trimming is applied such that the maximum weight is calculated as:

Mean + (x * Standard Deviation)

where x is an integer such that less than 5% of respondents have their weight trimmed for each reporting level.