Data and variables

We used a Generalised Linear Mixed Model (GLMM) to investigate which factors are linked to people’s likelihood of rating their care experience positively.

Factors affecting the care experience

The factors we took into account in the analyses include characteristics related to the individuals, the context and the consultation with the general practice. In particular, we analysed:

individual characteristics: how well people feel able to look after their own health, age group, gender, ethnic group, sexual orientation, impairment, disability, work status, religion, and whether day-to-day activities are limited
context: characteristics of the general practice (practice list size, whether the general practice are administered by the health board and whether they are dispensing practices) and of the place of residence (in terms of how urban/rural areas are and the level of deprivation)
characteristics of the consultations: how often people have contacted their general practice in the latest year, the kind of appointment (by phone, face to face etc.), the healthcare professional they received most of the treatment/advice from (doctor, nurse etc.) and what the treatment or advice received was for (physical problem, routine appointment, etc)

Aspects of care experience

We selected thirteen different aspects of care and ran separate analyses for each of them. The first aspect of care is overall care, and the others are related to person-centred care. The aspects of care analysed refer to people’s answers to the following survey questions:

“Overall, how would you rate the care provided by your GP practice?” (Excellent or good; fair; poor or very poor)

Thinking about that healthcare professional, how much do you agree or disagree with the following statements? (Strongly agree / agree; neither agree nor disagree; disagree / strongly disagree)

“I was given the opportunity to involve the people that matter to me”
“I was listened to”
“I was given enough time”
“I was treated with compassion and understanding”
“I knew the healthcare professional well”
“I had a chance to ask about the benefits and risks of the treatment”
“I was able to ask questions if I wanted to”
“I understood the information I was given”
“The health professional checked I understood what I had been told”
“Staff helped me to feel in control of my treatment/care”
“I was involved in decisions about my care and treatment”
“I felt able to make an informed choice about my treatment and care”

These are the answers to questions q10 and 13 on the Health and Care Experience questionnnaire.

Responses can be positive, neutral or negative. In the analysis, we explore the likelihood of positive responses.

All results are based on 2021/22 Health and Care Experience survey data.

Methods

We used a GLMM to estimate which groups of people are more or less likely to rate their experience of care positively, compared to a reference group.

GLMM is a type of statistical model that allows us to separate fixed and random effects for different groups of people.

The random effects in our model are general practices and health boards. The model takes into account the effects that particular general practices and health boards have on the likelihood of people reporting positive experiences.

We focus on the fixed effects, that estimate how each factor is associated with the likelihood of people rating their health experience positively, regardless of which general practice they are registered with or which health board they live in.

Odds ratio: The results are expressed in odds ratio. An odds ratio is a measure of how likely an outcome is, given a particular characteristic. In this context, they can be used to determine whether a characteristic (for example, age group), is a factor in the likelihood of having a positive care experience.

If a particular group has an odds ratio of more than 1, they are more likely to rate overall care positively, compared with the reference group, taking into account (controlling for) other characteristics.

If the odds ratio is less than 1, they are less likely to rate overall care positively, compared with the reference group, taking into account other characteristics.

If the odds ratio is 1, they are as likely as the reference group to rate overall care positively, taking into account other characteristics.

Confidence intervals: As the results are based on a sample of people, there is a margin of error to the results. For this reason, we report the 95 per cent confidence intervals for the odds ratio. This interval shows the range within which we would expect the true figure of the odds ratio to fall.

In this analysis, the confidence intervals tell us how certain we can be that any difference in the likelihood of rating care experience positively between the group of interest and the reference group is real, or whether the difference might be due to random variation in the data.

If an odds ratio is different from 1, but the associated confidence interval includes 1, we say that the difference between the group of interest and the reference group is not statistically significant. Hence, we cannot conclude that there is a difference in the likelihood of rating care experience positively of the group of interest and the reference group.

Example: Here we explain how to interpret odds ratio, using the results for age group as an example.

The table below shows how age group is associated with the likelihood of rating overall care positively.

Age group	Ref. group	OR	LCI	UCI	Significant
17-24	65+	0.624	0.564	0.690	Yes
25-34	65+	0.748	0.692	0.808	Yes
35-44	65+	0.855	0.795	0.919	Yes
45-54	65+	0.974	0.912	1.041	No
55-64	65+	0.902	0.855	0.952	Yes

The results in the above table show Odds Ratio (OR), Lower confidence Level (LCL), Upper Confidence Level (UCL), and whether the OR are statistically significant (Yes/No).

The reference group for age group is people who are aged 65 or older.

In this example, the odds ratio for all age groups are less than 1. This means that the model is showing people in all age groups (17-24, 25-34, 35-44, 45-54, 55-64) are less likely to rate their overall care positively, compared with the reference group.

Next, we need to consider how certain we are that the results are not due to random variation in the data. The confidence intervals for the 17-24, 25-34, 35-44, 55-64 age groups do not include one. From this, we can say that the results are statistically significant for these groups and that the difference observed is not due to random variation in the data.

For the age group 45-54, the odds ratio is less than one, but the confidence interval includes one. This means that we cannot be certain that those in the 45-54 age group would rate their care differently to those in the 65 and over age group. We would describe the results for the 45-54 age group as being “not statistically significant” and we would not conclude that there is a difference between this group and the reference group.

Throughout the report, we will only state that a group is more or less likely to rate their experience positively, compared with the reference group, if the difference is statistically significant.

For easiness of reading, when there are no statistically significant differences between the likelihood of rating care positively of different groups, we say that the groups have a similar likelihood of rating positively their care experience.

Data source

The analyses are based on the 2021/22 Health and Care Experience survey

More information is available in the Technical Report and the 2021/22 Health and Care Survey national results published on our website.

Limitations

The fieldwork for the 2021/22 Health and Care Experience survey was carried out during the COVID-19 pandemic. Therefore, there were a number of important changes to how services are provided that should be taken into account when interpreting the survey results and the results of the variation analysis in this report. More details are available on the 2021/22 Health and Care Experience survey National Results Report