Brief interventions for smoking and alcohol associated with the COVID-19 pandemic: a population survey in England … – BMC Public Health

Design Sample and recruitment

Data were drawn from the Smoking and Alcohol Toolkit Study (STS/ATS), a monthly repeated cross-sectional survey of a representative sample of adults (aged 18+) in England.The study population consisted of adults aged 18 and over living in households in England surveyed monthly between March 2014 and June 2022. All statistical analysis was restricted to people who smoked in the past year or who used alcohol at increasing and higher risk levels as indicated by scoring 38 in the Alcohol Use Disorders Identification Test (AUDIT) [23].

The STS/ATS uses a hybrid of random location and quota sampling to select a new sample of approximately 1,800 adults (aged18years) each month in England [24]. Sample weighting uses the rim (marginal) weighting technique, an iterative sequence of weighting adjustments whereby separate nationally representative targets are set, and the process repeated until all relevant variables match the English sociodemographic population profile relevant at the time each monthly survey was collected.

Respondents with characteristics that are under-represented receive a larger weight, while those who are over-represented receive a smaller weight. Data were collected monthly through face-to-face computer assisted interviews. However, due to the COVID-19 pandemic, from April 2020 data were collected via telephone only. A series of diagnostic analyses suggested it is reasonable to compare data from before and after the lockdown, despite the change in data collection method [25, 26].

The primary outcome measure was defined using responses to the following questions:

For smoking:

Has your GP spoken to you about smoking in the past year (i.e. last 12 months)?

Yes, heshe suggested that I go to a specialist stop smoking advisor or group

Yes, heshe suggested that I see a nurse in the practice

Yes, heshe offered me a prescription for Champix, Zyban, a nicotine patch, nicotine gum or another nicotine product

Yes, heshe suggested that I use an e-cigarette

Yes, heshe advised me to stop but did not offer anything

Yes, heshe asked me about my smoking but did not advise me to stop smoking

No, I have seen my GP in the last year but heshe has not spoken to me about smoking

No, I have not seen my GP in the last year

Dont know

Respondents who answered with any of responses a-e for smoking were classified as having received a BI. Responses of h were excluded under the sensitivity analyses which cover only those who have visited their GP.

For drinking:

In the last 12 months, has a doctor or other health worker within your GP surgery discussed your drinking?

No

Yes, a doctor or other health worker within my GP surgery asked about my drinking

Yes, a doctor or other health worker within my GP surgery offered advice about cutting down on my drinking

Yes, a doctor or other health worker within my GP surgery offered help or support within the surgery to help me cut down

Yes, a doctor or other health worker within my GP surgery referred me to an alcohol service or advised me to seek specialist help.

Dont know

Refused

Respondents who answered with any of c-e, were classified as having received a brief intervention from their GP for drinking.

For the analyses including only those who visited their GP, we excluded responses of a) in response to the question below:

You said a doctor or other health worker within your GP surgery has not discussed your drinking with you in the last 12months.

a) I have not seen a doctor or health worker within my GP surgery in last 12months.

b) I have seen a doctor or health worker within my GP surgery in the last 12months but did not discuss my drinking.

As a measure of socio-economic position, we used the National Readership Surveys classification of social grade based on occupation (ABC1: higher and intermediate managerial, administrative, and professional, supervisory, clerical and junior managerial, administrative and professional; C2DE: skilled manual workers, semi-skilled and unskilled manual workers and state pensioners, casual and lowest-grade workers, unemployed with state benefits.) [27].

Respondents were classified as having a history of a mental health condition if they reported being diagnosed by a doctor or health professional.

Respondents were asked:

Since the age of 16, which of the following, if any, has a doctor or health professional ever told you that you had?

Depression

Anxiety

Obsessive Compulsive disorder

Panic disorder or a phobia

Post-traumatic stress disorder (PTSD)

Psychosis or schizophrenia

Personality disorder

Attention Deficit Hyperactivity Disorder (ADHD)

An eating disorder

Alcohol misuse or dependence

Drug use or dependence

Problem gambling

Autism or Autism Spectrum Disorder

Bipolar disorder (previously known as manic depression)

None of these

Dont know

Prefer not to say

Responses excluding the final three options above were presented in a randomised order. For our analyses, individual responses of any of the above diagnoses were dummy coded into a composite measure of History of a mental health condition. Those who selected alcohol misuse or dependence were excluded from the alcohol BI analysis given that it is likely a confounder influencing the receipt of a BI for alcohol.

Age was treated as a continuous variable in models, but categorical to summarise the sample characteristics. Other sociodemographic covariates included identified sex (Women vs other (Men and In another way/refused)), the presence of children in the household (Yes vs No), and region of England (North, Midlands and South).

In the analyses of BIs for smoking, data were collected from March 2014 to June 2022. In the analyses of BIs for alcohol, data were collected from March 2014 to March 2022 because from April 2022 the brief intervention variable was collected every other month, and only questions related to AUDIT items one to three were collected (preventing the selection of individuals according to full 10-item AUDIT score).

For all primary analyses on BIs for smoking, and BIs for alcohol, the pre-pandemic period refers to the months up to and including February 2020, and the post-pandemic period from April 2020 onwards (no data were collected in March 2020 due to the pandemic). Characteristics of the sample for the pre- and post-pandemic periods are described in Table S1.

Regarding the analyses involving mental health data, the pre-pandemic period refers to the years 2016 and 2017, and the period from October 2020 onwards as the pandemic onset period, as these were the only periods where data on the included mental health measures were collected. Moreover, for 2016/2017 mental health was only assessed in past-year smokers, so this sample did not include any people who used alcohol at increasing and higher-risk levels but did not smoke.

The analyses were conducted in R version 4.2.1 [28] using the packages survey [29] and mgcv [30]. This analysis plan was pre-registered on the Open Science Framework https://doi.org/10.17605/OSF.IO/65FRC. The STROBE reporting guidelines were used in the design and reporting of this study. Respondents with missing data on any of the covariates of interest were excluded from the analyses (less than 5% of responses). Characteristics of the sample and descriptive statistics are presented using weighted descriptive statistics for the overall sample, and for the pre-pandemic and post-pandemic periods, respectively.

A segmented regression design was used to assess the effect of the COVID-19 pandemic on receipt of BIs for smoking and alcohol, respectively. Data was analysed at the individual-level with segmented regression using generalised additive models (GAM) [31, 32]. These allow the fitting of seasonal smoothing terms and thus seasonality to be considered (which are particularly relevant in the context of delivery of interventions for smoking and alcohol use [33]). A log link function was used so that relative risks can be reported.

Each GAM modelled the trend in the overall receipt of BIs (dependent variable) for smoking and alcohol, respectively in the pre-pandemic period, and any change in the trends in the post-pandemic period. Trend is a variable coded 1n (n being the total number of time-points to the end of the series) reflecting the time trend over time. The slope variable was defined as 0 before April 2020 of the pre-pandemic period and each month from April 2020, by increments of 1 up to m where m is the number of waves from April 2020.

Models were first fit assuming a linear underlying and post-implementation trend, followed by fits using non-linear trends to explore changes in the level of BI delivery and potential rebounding in the delivery of BIs over time. Specifically, the outcome of BI delivery refers to receipt of a BI in the previous 12months. It is therefore possible that an immediate step change in delivery would not be detected in April 2020 or in the months immediately afterwards but would be reflected by changes in the trend in the longer term. In addition, after an initial drop during heightened restrictions 2020 and 2021, rates of BI delivery may have rebounded with some GP delivery returning to normal practice. Therefore, we fit further GAMs with the independent variables for slope and trend wrapped in a smooth function (model fit using the restricted maximum likelihood method with nine basis functions specified for the underlying trend and change in slope). Models accounted for seasonality in the receipt of BIs by using a smoothing term with cyclic cubic regression splines (11 knots, one for each month in the year) and were adjusted for sociodemographic characteristics (age, sex, children in the household, and region).

Interactions were tested between social grade and the post-intervention change in slope, and results reported stratified by social grade to explore whether the post-intervention slope depends on social grade. The model fit of the linear and non-linear GAMs were compared using the Akaike Information Criterion (AIC; lower values indicating better model fit) and a likelihood ratio test.

BI delivery may have declined during the pandemic due to reduced GP contact overall, rather than reduced delivery rates among those who visited their GP. To understand whether BI delivery also declined among those still visiting their GP, all analyses were repeated with the sample of only those who smoked in the past-year and those who used alcohol at increasing and higher risk levels, respectively, who reported visiting their GP in the past year (Table S2).

Models were checked for full convergence, and for randomly distributed residuals using the gam.check() function in the mgcv package [30] in R.

We constructed logistic regression models to explore whether changes in the of receipt of BIs for smoking and alcohol, respectively, from December 2016 to January 2017 and October 2020 to June (or March in the case of BIs for alcohol) 2022, depended on history of a mental health condition. Associations were reported as odds ratios (ORs) with 95% confidence intervals. Models adjusted for sociodemographic characteristics. The inclusion of the time period*mental health interaction allowed us to explore potentially differential changes in receipt of interventions over between the two time periods according to whether an individual had a history of a mental health condition.

Read the original post:

Brief interventions for smoking and alcohol associated with the COVID-19 pandemic: a population survey in England ... - BMC Public Health