Maternity Care Survey 2025: technical report
This report contains information about the methodology and background of the 2025 Maternity Care Survey. It is intended primarily for more technical users who wish to understand how the survey was designed, delivered and analysed.
Analysis and reporting
The survey data collected and coded by IQVIA was securely transferred to the Scottish Government. The survey data was analysed by analysts at the Scottish Government using the statistical software package R.
Validation of the survey responses
As some survey respondents completed the survey on paper and it is not possible to enforce routing or other survey instructions on the paper returns, validation was applied to these responses. The validation rules covered the following situations:
- If a survey respondent did not follow the routing instructions: if they answered questions in a section that they should have been routed past, these responses were changed to ‘not applicable’ (‘NA’)
- If a survey respondent provided contradictory responses within a question, the response to that question was treated as invalid and left blank
- If a survey respondent did not answer a routing question, but answered questions within the routed section: if it was not possible to unambigiously assign the respondent to one response option in the routing question, all questions in the routed section were changed to NA.
Number of responses analysed
The number of responses that have been analysed for each question is often lower than the total number of survey responses received. This is because not all of the questionnaires that were returned could be included in the calculation of results for every individual question. Responses may be excluded from the analysis because:
- The respondent skipped or refused to answer a question. People were advised that if they did not want to answer a specific question they should leave it blank
- The specific question did not apply to the respondent and so they did not answer it. For example if their baby did not receive neonatal care and therefore they did not answer questions about their experience of it
- The respondent did not know or could not remember the answer to a particular question
- Responses may be removed following validation checks, for example if a respondent selected an invalid combination of responses. Validation rules are automatically applied at the point of data collection through the online questionnaire (for example through enforcing “tick one box only” instructions) and validation checks are applied to ensure consistency between online and paper responses, as detailed above.
Weighting
When conducting a survey, it is important to have a representative sample of the population one is interested in. Applying weighting methods reduces potential bias by making the results more representative of the target population.
Survey weights are numbers associated with the responses that specify the influence the various observations should have in the analysis. The survey weight associated with a particular response can be thought of as a measure of the number of population units represented by that response.
In this survey we apply non-response weights. Non-response weights are statistical adjustments applied to survey data to account for differences in response rates among different groups of respondents. This method aims to ensure that the results accurately reflect the population they represent, particularly when non-response is not random. In particular, we applied the following methodology:
Non-response weights were calculated dividing all women in a certain group by the number of respondents in a group. All women in that group would be the true population of women that we want to represent. The data used to estimate the true population of women the survey represents was sourced from SLiPBD.
- at national level, we weighted respondents by level of deprivation (SIMD)
- we used the national proportion of women who gave birth in hospital or a midwife-led unit to estimate the true population of women who gave birth in hospital or midwife-led unit by level of deprivation to create weights for the section on care in hospital or a midwife-led unit
- we followed the same procedure to create weights for the section on care in a neonatal unit
The weighting methodology in 2025 has changed from 2018. In 2025, we only applied weights to address non-response bias. In 2018, we applied weights for non-response bias and weights to address probability of selection. The change is a consequence of a change in the sample design. In 2025, the initial sample of women was a full census, where the probability of selection of each woman is virtually 1 and therefore weights to address probability of selection were not necessary. In 2018, 7,782 women were included in the sampling frame and 5,165 women (i.e., 66%) were sampled to take part in the survey.
The 2018 survey used a sampling approach that selected different numbers of women from each large hospital. Hospitals with lower response rates in 2015 were sampled more heavily so that results from each hospital would be accurate to within around seven percentage points at the 95% confidence level. The 2025 survey used a census approach, inviting everyone in the target population to take part, so this type of oversampling was not possible.
More information about the weighting methodology used in the previous survey is available in the Technical report of the Maternity Care Survey 2018.
In summary, to maximise comparability between surveys we addressed non-response bias by weighting results by SIMD. However, differences in survey design mean that comparisons between 2018 and 2025 should be interpreted with some caution.
Percentage positive and negative
‘Percent positive’ or ‘percentage positive’ is frequently used in reporting results from this survey. This means the percentage of people who answered that they had an experience that can be considered to be positive. For example, when people were asked to rate the care they received, if they answered “Excellent” or “Good”, their answers were counted as positive answers. Similarly, if people answered “Poor” or “Very poor”, their answers were counted as negative answers, and if they answered “Fair”, their experiences were counted as neutral answers. Annex B shows how response options have been classified into positive, neutral or negative for each of this type of question.
We report results in terms of percent or percentage positive because they are easier to interpret, compared with reporting results on the full five point scale that people used to answer the questions. Some evidence also suggests that respondents do not necessarily interpret adjacent categories on a five-point Likert scale in the same way. For example, the distinction between "agree" and "strongly agree" may vary between individuals, and some respondents consistently avoid extreme response categories regardless of their underlying attitude. As a result, differences between neighbouring response options may not represent equal differences in the underlying construct being measured (Pokropek, Khorramdel & von Davier, 2023; Ulitzsch, Henninger, Meiser, 2023)
Coverage error and sampling error
The eligible population for the survey was smaller than the target population because some women could not be contacted due to invalid address information or other issues (see the response rate section of this report for more information). This may introduce coverage error if the characteristics or experiences of women who could not be contacted differ systematically from those who were invited to participate. Some coverage error is addressed by weighting, but some might still exists and it should be taken into account when interpreting the data, although it is not clear how coverage error may have affected the results of the survey.
Survey estimates are subject to sampling error because they are based on a sample rather than the entire target population, therefore the results are affected by sampling error. Confidence intervals use the standard error to derive a range in which we think the true value is likely to lie. A 95% confidence level is used in the survey outputs.
Where comparisons have been made in the analysis of the survey results, for example between different groups or different survey years, the effects of sampling error have been taken into account by the use of confidence intervals and tests for statistical significance. The effect of this sampling error is relatively small for the national estimates, but is larger when looking at subgroups with a smaller sample size. Only differences that are statistically significant are reported as differences within the reported analysis and all significance testing is carried out at the 5% level.
More information on confidence intervals, significance testing and how they are calculated can be found in Annex C.
Quality assurance of the National Report
A small group of Scottish Government analysts were sent a draft version of the national report for quality assurance. Feedback included suggestions on ways in which to report data as well as comments about the context for the survey. These were taken into account in finalising the National Report. In addition, staff at IQVIA carried out quality checks of figures used in the report.
Timeliness of reporting
The planned date of publication was announced on the Scottish Government website in August 2026. If there had been any changes to the pre-announced date, this would have been announced alongside the reason.
The publication of results took place around eight months after the fieldwork period ended, due to prioritisation of resource available in the care experience survey team and the time required for data entry, quality assurance of the data and code and production and quality assurance of the outputs.
Contact
patientexperience@gov.scot