This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) outbreak may motivate smokers to attempt to stop in greater numbers. However, given the temporary closure of UK stop smoking services and vape shops, smokers attempting to quit may instead seek out mobile health support, such as smartphone apps.
We examined, using an interrupted time series approach, whether the SARS-CoV-2 outbreak has been associated with a step change or increasing trend in UK downloads of an otherwise popular smoking cessation app, Smoke Free.
Data were from daily and nondaily adult smokers in the United Kingdom who had downloaded the Smoke Free app between January 1, 2020, and March 31, 2020 (primary analysis), and January 1, 2019, and March 31, 2020 (secondary analysis). The outcome variable was the number of downloads aggregated at the 12-hourly (primary analysis) or daily level (secondary analysis). The explanatory variable was the start of the SARS-CoV-2 outbreak, operationalized as March 1, 2020 (primary analysis), and January 15, 2020 (secondary analysis). Generalized additive mixed models adjusted for relevant covariates were fitted.
Data were collected on 45,105 (primary analysis) and 119,881 (secondary analysis) users. In both analyses, there was no evidence for a step change or increasing trend in downloads attributable to the start of the SARS-CoV-2 outbreak. Calculation of Bayes factors (BFs) indicated that the data for the primary analysis favored the null hypothesis compared with large associations (for level, BF=0.25; for slope, BF=0.26) but were insensitive to the detection of small associations (for level, BF=0.78; for slope, BF=1.35).
In the United Kingdom, between January 1, 2020, and March 31, 2020, and between January 1, 2019, and March 31, 2020, there was no evidence that the SARS-CoV-2 outbreak has been associated with a large step change or increasing trend in downloads of a popular smoking cessation app. Findings on the association of the SARS-CoV-2 outbreak with a small step change or increasing trend were inconclusive.
The global outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes the coronavirus disease (COVID-19) [
Definitive evidence on whether current smokers are at increased risk of disease, morbidity, and mortality from COVID-19 are not yet available. However, researchers and clinicians have emphasized biological (eg, reduced respiratory immune defense) and behavioral (eg, repetitive hand-to-mouth movements) factors that may mean that smokers are at an increased risk of contracting SARS-CoV-2 [
Timeline of key UK events in 2020 related to smoking and COVID-19, including government policies and legislation [
Date | Key event |
January 31 | First two cases reported in the United Kingdom |
February 7 | Primary care guidance updated to state that the virus is most likely to be seen in travelers returning from China, Hong Kong, Japan, Macau, Malaysia, Republic of Korea, Singapore, Taiwan, or Thailand |
February 21 | At least 17 temporary GPa closures occur |
February 24 | 4 UK passengers repatriated from a Japanese cruise ship test positive for COVID-19b, bringing the total number of UK cases to 13 |
March 2 | GP practice closures continue |
March 6 | First UK patient dies from COVID-19 |
March 8 | Prime Minister Boris Johnson unveils the Coronavirus Action Plan |
March 11 | World Health Organization declares the severe acute respiratory syndrome coronavirus 2 outbreak a pandemic |
March 12 | UK COVID-19 risk level raised from moderate to high |
March 16 | UK government advises on social distancing (eg, avoiding nonessential travel and contact with others); pregnant women, adults 70 years or older, and individuals with health conditions are urged to self-isolate |
March 18 | UK government announces that schools will close to most students, with children of key workers and vulnerable children still able to attend |
March 20 | Restaurants, pubs, clubs, and indoor sport and leisure centers ordered to close; Action on Smoking and Health launches the stop smoking campaign “Today Is The Day” and #QuitForCovid |
March 23 | UK government announces restrictions made on freedom of movement, enforceable by law |
April 3 | Public Health England releases a news story emphasizing the increased risk of severe respiratory disease due to COVID-19 in smokers |
aGP: general practice.
bCOVID-19: coronavirus disease.
Options for UK smokers interested in quitting usually include specialist stop smoking services and vape shops. However, the introduction of social distancing policies and behavioral restrictions enforceable by law mean that the vast majority of services and shops have temporarily closed. At the same time, it is plausible that the SARS-CoV-2 outbreak has acted as a “teachable moment,” motivating smokers to attempt to stop in greater numbers than would otherwise be observed for this time of the year. Given these temporary closures of stop smoking services and vape shops, smokers attempting to quit during the SARS-CoV-2 outbreak may have to seek out digital support, such as websites and smartphone apps. If so, it would be important for public health bodies to put resources toward ramping up access to high-quality, evidence-informed digital support.
One way of assessing whether the recent SARS-CoV-2 outbreak has been accompanied by a surge in smokers accessing digital support is to use an interrupted time series approach to examine whether there has been a step change or increasing trend in downloads of an otherwise popular smoking cessation app. The Smoke Free app, designed for English speakers, includes behavior change techniques that can improve the chances of quitting and shows early evidence of effectiveness [
This study used an interrupted time series approach to assess whether the SARS-CoV-2 outbreak up to the end of March 2020 (ie, a hypothesized “teachable moment”) has been accompanied by an increase in UK downloads of the free and “pro” versions of the Smoke Free app (henceforth referred to as the Smoke Free app). Two different starting points for the outbreak were used. First, we assessed whether there has been a step change or increasing trend (linear or nonlinear) in downloads during a period of government announcements of escalating social distancing policies and legislation, with March 1, 2020, as the stipulated starting point (see
Specifically, we aimed to address the following research questions:
Over a 3-month time series (ie, January 1, 2020, to March 31, 2020), has a period of government announcements of escalating social distancing policies and legislation (ie, March 1, 2020, to March 31, 2020) been accompanied by a step change or increasing trend in UK downloads of the Smoke Free app?
Over a 15-month time series (ie, January 1, 2019, to March 31, 2020), has the SARS-CoV-2 outbreak thus far (ie, January 15, 2020, to March 31, 2020) been accompanied by an increasing trend in UK downloads of the Smoke Free app?
This was a natural experiment without active recruitment. The study protocol and analysis plan were preregistered on the Open Science Framework [
To be included in the analytic sample, daily and nondaily smokers needed to be 18 years or older and have downloaded the Smoke Free app from a UK-based app store (ie, Google Play and Apple App Store) during the respective study periods.
For the primary analysis, the outcome of interest was the number of UK Smoke Free app downloads aggregated at the 12-hourly level. This temporal unit was selected to increase the number of observations in the postintervention period.
For the secondary analysis, the outcome of interest was the number of UK Smoke Free app downloads aggregated at the daily level.
For the primary analysis, the 60-day period (or 120 observations) before the government announcements of escalating social distancing policies and legislation was coded as 0, and the 31-day period (or 62 observations) containing the announcements was coded as 1. The trend during the 31-day postintervention period from March 1, 2020, to March 31, 2020, (or 62 observations) was coded from 1-62.
For the secondary analysis, the 379-day period (or 379 observations) before the start of the SARS-CoV-2 outbreak was coded as 0, and the trend over the 77-day period containing the outbreak from January 15, 2020, to March 31, 2020, (77 observations) was coded from 1-77.
For the primary analysis, the time of day was included as a categorical covariate (midnight to 11:59 AM or “morning” vs noon to 11:59 PM or “evening”), and the day of the week was included as a cyclic cubic spline to capture nonlinear, seasonal patterns in downloads.
For the secondary analysis, the month of the year and the day of the week were included as cyclic cubic splines to capture nonlinear, seasonal patterns in downloads. Two additional covariates were included. On December 19, 2019, a new expert feature, which involved the ability to communicate directly with stop smoking counsellors, was launched for “pro” users (ie, “expert feature launch”). On December 27, 2019, a promotional offer to purchase a separate version of the app called Smoke Free Plus in a range of Boots stores across the United Kingdom was announced (ie, “national advertising campaign”). The periods before and after the expert feature launch and the national advertising campaign were coded as 0 and 1, respectively.
The analyses were conducted in R v.3.6.3 (R Foundation for Statistical Computing) using the
For the primary analysis, an interrupted time series analysis (segmented regression) was conducted using a generalized additive mixed model (GAMM). GAMMs take account of seasonality through the fitting of seasonal smoothing terms [
The same steps were repeated for the secondary analysis. The final model was selected in a similar way. As specified in the analytic plan, the model included terms for a secular trend, slope, month of the year, day of the week, the expert feature launch, and the national advertising campaign.
Prompted by inspection of the data and as a result of the review process, three unplanned sensitivity analyses (SAs) were conducted.
First, given uncertainties as to when the SARS-CoV-2 outbreak had started to affect smokers’ lives, we modelled the starting point of the outbreak as approximately 15 days before (ie, February 15, 2020) and after (ie, March 15, 2020) the original starting point (ie, March 1, 2020). These SAs were considered exploratory (as opposed to hypothesis testing).
Second, SAs were conducted to examine whether observed nonsignificant associations for the primary analysis could best be characterized as evidence of no effect or whether the data were insensitive to distinguish the null from the alternative hypothesis [
Third, with additional data becoming available, the primary and secondary analyses were topped up with data up to May 12, 2020. These SAs were considered exploratory.
Data were collected on 45,105 and 119,881 users, respectively.
The number of users downloading the Smoke Free app from January 1, 2020, to March 31, 2020. The dotted blue vertical line represents the stipulated starting point of the severe acute respiratory syndrome coronavirus 2 outbreak (ie, March 1, 2020). The solid red line represents the fitted values with a loess smoothing function applied and associated 95% CIs.
The number of users downloading the Smoke Free app from January 1, 2019, to March 31, 2020. The dotted blue vertical line represents the stipulated starting point of the severe acute respiratory syndrome coronavirus 2 outbreak (ie, January 15, 2020). The solid red line represents the fitted values with a loess smoothing function applied and associated 95% CIs. For clarity, the y-axis was capped at 1300 downloads.
Results from the best fitting model for each analysis.
Analysis | IRRa (95% CI) | SE | Bayes factor (small expected effect) | Bayes factor (large expected effect) | |||||||
|
|||||||||||
|
Trend (January 1, 2020-March 31, 2020) | 0.998 (0.994-1.002) | 0.002 | .23 | N/Ac | N/A | |||||
|
Level (March 1, 2020) | 1.142 (0.916-1.423) | 0.112 | .24 | 0.78 | 0.25 | |||||
|
Slope (March 1, 2020-March 31, 2020) | 0.999 (0.988-1.010) | 0.006 | .85 | 1.35 | 0.26 | |||||
|
Time of dayd | 1.780 (1.581-2.004) | 0.060 | <.001 | N/A | N/A | |||||
|
|||||||||||
|
Trend (January 1, 2019-March 31, 2020) | 0.996 (0.994-0.998) | 0.001 | <.001 | N/A | N/A | |||||
|
Slope (January 15, 2020-March 31, 2020) | 0.976 (0.908-1.049) | 0.037 | .51 | N/A | N/A | |||||
|
Slope^2 | 1.001 (0.998-1.003) | 0.001 | .52 | N/A | N/A | |||||
|
Slope^3 | 1.000 (1.000-1.000) | <0.001 | .54 | N/A | N/A | |||||
|
Expert feature launch (December 19, 2019)f | 1.385 (0.885-2.168) | 0.229 | .16 | N/A | N/A | |||||
|
National advertising campaign (December 27, 2019)f | 2.067 (1.387-3.079) | 0.203 | <.001 | N/A | N/A |
aIRR: incidence rate ratio.
bAdjusted for a second-order autoregressive term and day of the week.
cNot applicable.
dReferent=morning (vs evening).
eAdjusted for a first-order autoregressive term, month of the year, and day of the week.
fModeled as a step change.
When modeling the starting point of the outbreak as February 15, 2020, and March 15, 2020, there was no evidence for a step change or increasing trend in downloads attributable to the start of the SARS-CoV-2 outbreak (see
The calculation of BFs indicated that the data for the primary analysis favored the null hypothesis compared with large associations (for level, BF=0.25; for slope, BF=0.26) but were insensitive to detection of small associations (for level, BF=0.78; for slope, BF=1.35; see
In the SAs repeating the primary and secondary analyses with additional data up to May 12, 2020, there was no evidence for a step change or increasing trend in downloads attributable to the start of the SARS-CoV-2 outbreak (see
This study used an interrupted time series approach to examine the impact of the ongoing SARS-CoV-2 outbreak on the uptake of an otherwise popular smoking cessation app in UK smokers. Across two different time periods and using two different conceptualizations of the starting point of the outbreak, we found no evidence for a step change or increasing trend in the number of downloads of the popular Smoke Free app. Calculation of BFs indicated that the data favored the null compared with large associations but were insensitive to detection of small associations.
First, the small number of observations since the start of the SARS-CoV-2 outbreak means that we likely had low statistical power to detect anything but large effects. Second, pinpointing the exact starting point of the outbreak was not straightforward, as it is unclear when UK smokers first became aware of or affected by the pandemic, hypothesized to act as triggers for additional quit attempts. This was partly mitigated by conducting two SAs, which also did not provide evidence for a step change or increasing trend in downloads. Third, given the current structure of the app’s database, we were unable to distinguish between downloads of the free and “pro” versions of the Smoke Free app in our analyses. Although a tally is kept for free and “pro” downloads combined, the tally kept for “pro” accounts has been designed to track weekly subscriptions (with each user counted several times) as opposed to one-off downloads. It is plausible that a greater number of smokers who otherwise would have tried to stop using e-cigarettes or stop smoking services would now be more inclined to purchase the “pro” version of the app. Fourth, the Smoke Free app is only one of many available online stop smoking platforms in the United Kingdom. We did not capture traffic on other available apps or websites such as the NHS Smokefree website [
The lack of an uptick in downloads of a popular smoking cessation app during the SARS-CoV-2 outbreak thus far suggests that smokers may not be turning to available, evidence-informed digital support in greater numbers. Evidence from controlled studies and population-level surveys indicate that smoking cessation attempts involving pharmacological or behavioral support (including digital interventions) are substantially more likely to be successful compared with unassisted attempts [
In the UK, between January 1, 2020, and March 31, 2020, and between January 1, 2019, and March 31, 2020, there was no evidence that the SARS-CoV-2 outbreak thus far has been associated with a large step change or increasing trend in downloads of a popular smoking cessation app. Findings on the association of the SARS-CoV-2 outbreak with a small step change or increasing trend were inconclusive.
Sensitivity analyses.
Akaike information criterion
autoregressive
Bayes factor
coronavirus disease
electronic cigarettes
generalized additive mixed model
incidence rate ratio
moving average
sensitivity analysis
severe acute respiratory syndrome coronavirus 2
OP and JB receive salary support from Cancer Research UK (C1417/A22962). JB, LS, AH, and OP are members of Shaping Public Health Policies to Reduce Inequalities and Harm, a UK Prevention Research Partnership Consortium (MR/S037519/1). UK Prevention Research Partnership is an initiative funded by the UK Research and Innovation Councils, the Department of Health and Social Care (England) and the UK devolved administrations, and leading health research charities.
OP, AH, LS, and JB designed the study. OP conducted the statistical analyses and wrote the first draft of the manuscript. All authors have contributed to the final version of the manuscript and agree with its submission to JMIR.
OP and JB are unpaid members of the scientific advisory committee for the Smoke Free app. LS has received a research grant and honoraria for a talk and travel expenses from manufacturers of smoking cessation medications (Pfizer and Johnson & Johnson). JB and AH have received unrestricted research funding from Pfizer to study smoking cessation.