Tobacco trends: how changes to surveys during the pandemic affected smoking estimates
We can be confident that the trend in smoking is continuing to decline, but the coronavirus (COVID-19) pandemic affected how data about smoking were collected. This blog looks at the available sources of data about smoking in more detail. It also looks at what the data tell us about smoking trends.
The Annual Population Survey (APS)
The APS is an official measure of how common smoking is in England. The APS defines current cigarette smokers as people who said they smoke cigarettes nowadays. We give data about current cigarette smokers as a proportion of the population. We call this measure of the proportion of smokers ‘smoking prevalence’.
The APS is the most comprehensive estimate we have of smoking prevalence. It is a designated National Statistic.
The APS is used to track progress against the Tobacco Control Plan. It provides important indicators of the number of adults aged 18 years and over who smoke in the UK. Based on the APS, the smoking prevalence figure for 2020 was a lot lower than in 2019. The size of this reduction was affected by the pandemic because it changed the way we could collect data in the APS.
Before the pandemic the APS data was collected using a mixed mode approach, meaning that interviews were done face-to-face or on the phone. During the pandemic, interviews were only done on the phone. This change affected cigarette smoking estimates.
In the APS, telephone respondents have historically been less likely to report smoking compared to face-to-face respondents. When interviews became telephone only, we saw an unrealistic decrease in the number of adults who smoked cigarettes. The bias has been reduced by using statistical weighting procedures, but the smoking prevalence figure for England for Quarter 2 to Quarter 4 in 2020 was still 12.1%. This is very low compared to previous years and is unlikely to be a true indication of the real number. By contrast, the figure for Quarter 1 in 2020 was 13.5%. This is closer to previous estimates and the historic trend.
Although the APS 2020 cannot be compared with previous years it does allow us to see differences across the country and between demographic groups. Other sources show that we can be confident that the trend in smoking is continuing to decline. The data used in the Local Tobacco Control Profiles is based on Quarter 2 to Quarter 4 of 2020, which means that confidence intervals are a little wider. This means we can be less sure that the estimates are accurate, but the APS is still the best measure available for sub-national smoking prevalence.
From the APS 2020, the prevalence of current smokers was highest in the North East (13.6%) and the North West (13.4%). It was lowest in the South West (10.9%). We can also see that the prevalence of current smokers was highest in the most deprived decile (14.8%). It was lowest in the least deprived decile (9.0%). The prevalence of current smokers aged 18 to 64 in the routine and manual occupation group was 21.4%. This compares to 12.1% across the general population, and gives an odds ratio of 2:1 compared to people in other occupation groups.
The Opinions and Lifestyle Survey (OPN)
After the APS, the OPN is the next best source for estimating smoking prevalence during the pandemic. With the OPN, current cigarette smokers are people who said they smoked cigarettes, even if they only smoke occasionally.
Data from the OPN was not affected by a change in survey mode because interviews were already done online or by phone. This means that estimates for 2020 are consistent with previous years. Based on the OPN, 14.4% of people aged 16 years and over in England said they currently smoked in 2020. This was a decrease from 15.5% in 2019, but this decrease was not statistically significant.
Other sources such as the GP Patient Survey (GPPS) and the Quality Outcomes Framework (QOF) are available on the Local Tobacco Control Profiles. They also show a decline over time.
There have been interruptions before in surveys, but the pandemic was a new and unforeseen interruption to normal survey activities. It made face-to-face data collection for the APS impossible. And we were unable to calculate the likely effects of the pandemic on our data, or make a plan to ease the effects in advance. Because of this, it has not been possible to address potential biases before the data was published.
The very low prevalence figure tells us that the data is unlikely to be accurate. So we must warn users that the estimate for Quarter 2 to Quarter 4 of 2020 is not comparable to previous years. But work has been commissioned to review the 2020 methodology to produce a statistical ‘fix’ to lessen the effect of the change in interview mode on the data. It is important to note that it may not be possible to find a fix, but practical approaches will be investigated.
The future of the APS
No decision has been made yet about returning to mixed modes for the APS. The 2021 collection was done by phone because the pandemic was still ongoing. This means that 2021 data will be comparable to the data from Quarter 2 to Quarter 4 of 2020 as published on the Office for Health Improvement and Disparities’ (OHID) Local Tobacco Control Profiles. APS data from 2020 and 2021 will not be comparable with data from 2019 or earlier years.
We are aware that the Tobacco Control Plan for England includes a target of 12% of current smoking prevalence for 2022. It may not be possible to correct the post-pandemic APS data to work out whether the target has been reached. But when the 2021 APS data is released, we will we will investigate whether the data is comparable, and whether it can be used to track progress against these targets. If this is not possible, we will use other sources to work out whether the 12% target has been met.
Mark Cook and James Tucker
After completing his PhD in quantitative genetics, James Tucker has worked as an analyst in a range of government departments. He joined the Office for National Statistics (ONS) in 2013. He founded the Government Data Quality Hub in 2020, and established it as a service for anyone working with data in the public sector. The hub provides strategic direction and capability building initiatives. James is currently head of health and life events analysis at the ONS. He continues to have a strong interest in data quality and management.
James has previously served on the Board of Trustees for the Royal Statistical Society (RSS). He is also on the editorial board for the RSS’s ‘Significance’ magazine, and has chaired the United Nations Quality Expert Group.