smktrans: An R Package for estimating smoking state transition probabilities from cross-sectional survey data 
This R package was developed as part of the Sheffield Tobacco and Alcohol Policy Modelling https://stapm.gitlab.io/ by the School of Health and Related Research at the University of Sheffield.
The aim of the research programme is to identify and evaluate approaches to reducing the harm from tobacco and alcohol, with the aim of improving commissioning in a public health policy context, i.e. providing knowledge to support benefits achieved by policymakers.
The two objectives of the research programme are:
- To evaluate the health and economic effects of past trends, policy changes or interventions that have affected alcohol consumption and/or tobacco smoking
- To appraise the health and economic outcomes of potential future trends, changes to alcohol and/or tobacco policy or new interventions
The STAPM modelling is not linked to the tobacco or alcohol industry and is conducted without industry funding or influence.
The code has been made open source for the following three reasons:
- Transparency. Open science, allowing review and feedback to the project team on the code and methods used.
- Methodology sharing. For people to understand the code and methods used so they might use aspects of it in their own work, e.g., because they are doing something partially related that isn’t exactly the same job and might like to ‘dip into’ elements of this code for inspiration.
- Sharing of the transition probability estimates. For people who would like to use the estimates but not run the code to produce them. There are spreadsheets containing the estimates for England and Scotland. The workflow used to produce these estimates and slides with plots of the trends in the estimates for England and Scotland are available here https://github.com/STAPM/smktrans/tree/master/transition_probability_estimates.
The starting point was to conduct a scoping review of individual based models of smoking state dynamics https://osf.io/egqvu.
The next step was to develop a workflow for estimating smoking state transition probabilities https://stapm.gitlab.io/r-packages/smktrans/articles/smoking_transition_probabilities_modelling_report.pdf
This was informed by previous work by Holford et al.
- Holford TR, Levy DT, McKay LA, Clarke L, Racine B, Meza R, Land S, Jeon J, Feuer EJ. Patterns of birth cohort–specific smoking histories, 1965–2009. American Journal of Preventive Medicine. 2014 Feb 1;46(2):e31-7. doi: https://doi.org/10.1016/j.amepre.2013.10.022.
This work fed into the development of the Sheffield Tobacco Policy Model (STPM) https://stapm.gitlab.io/projects/smoking-model-development/stpm_tech_report/stpm-technical-report.pdf
See the STPM webpage https://stapm.gitlab.io/STPM.html
The transition probability estimates have been used to inform forecasts of smoking prevalence in England:
- Gillespie D, Langley T, Agrawal S (2021) Section 2.6 - Model-based projections of alternative scenarios for the future trajectories of smoking prevalence in England and corresponding differences in quality-adjusted years of life. Tobacco Advisory Group of the Royal College of Physicians. ‘Smoking and health 2021: A coming of age for tobacco control?’ Royal College of Physicians. https://www.rcplondon.ac.uk/projects/outputs/smoking-and-health-2021-coming-age-tobacco-control
The code is actively being used in project work. It is being reviewed and developed all the time; more tests and checks are being added.
The repository is not intended to be maintained by an open source community wider than the development team.
The functions in smktrans are designed for use with:
Survey data processed by the hseclean package
- Gillespie D, Webster L, Leeming G, Morris D, Angus C, Brennan A ([YEAR]). hseclean: An R Package for Health Survey Data Wrangling. R package version [x.x.x]. University of Sheffield. https://stapm.github.io/hseclean/. doi: https://doi.org/10.17605/OSF.IO/43N7P
Mortality data processed by the mort.tools package
- Gillespie D, Webster L, Angus C, Brennan A ([YEAR]). mort.tools: An R Package for Processing Mortality Microdata. R package version [x.x.x]. University of Sheffield. https://stapm.gitlab.io/r-packages/mort.tools/. doi: https://doi.org/10.17605/OSF.IO/WN6RH
Relative risks of diseases related to smoking as encoded with the tobalcepi package
- Gillespie D, Webster L, Henney M, Brennan A, Angus C ([YEAR]). tobalcepi: An R Package for Computing the Relative Risks and Population Attributable Fractions of Diseases Related to Tobacco and Alcohol. R package version [x.x.x]. University of Sheffield. https://stapm.github.io/tobalcepi/. doi: https://doi.org/10.17605/OSF.IO/XQ8MV
Detailing all of the inputs used
Health survey data
The smktrans package is so far designed to work with cross-sectional survey data from the Health Survey for England (HSE) and the Scottish Health Survey (SHeS).
To be able to download data from these surveys from the UK Data Service, you will need to register with the UK Data Service website, which will enable you to request access to the datasets. Instructions on how to do this can be found here.
Mortality data
The smktrans package is designed to use cause-specific death rates stratified by sex and Index of Multiple Deprivation quintiles. Data for the smoking related causes detailed in this list were supplied for England by the Office for National Statistics and for Scotland by National Records Scotland.
Smoking state specific death rates are estimated by combining these death rates with smoking survey data and relative risks of diseases related to smoking.
Birth cohort survivorship data was derived from a long history of mortality rates by age and sex downloaded from the Human Mortality Database https://www.mortality.org/.
Long term relapse to smoking by former smokers
Annual cross-sectional survey data does not normally track individual life-histories enough to reliably estimate the probabilities of relapse to smoking by former smokers over many since since quitting smoking. To address this gap in information, the smktrans package is designed to use statistical estimates of the probabilities of smoking relapse from
- Hawkins, J., Hollingworth, W., & Campbell, R. (2010). Long-term smoking relapse: A study using the British Household Panel Survey. Nicotine & tobacco research, 12(12), 1228-1235. doi: https://doi.org/10.1093/ntr/ntq175
To estimate relapse probabilities, former smokers in the survey data are assigned a probability of relapse to smoking based on the match between the data fields in the survey data and the covariates considered in the statistical analysis.
Information governance
In ScHARR, all data is stored and processed according to the ScHARR Information Governance Policy. No sensitive individual-level survey or mortality data is included within this package.
Data checks are brief reports that show the estimates of the smoking state transition probabilities produced by the smktrans package.
The code on Github (https://github.com/STAPM/smktrans) is a mirror of the code in a private Gitlab repository where the actual development takes place (https://gitlab.com/stapm/r-packages/smktrans). The code in the Github repository is linked to a repository on the Open Science Framework, which provides the doi for the package citation (https://osf.io/ygxq9/). The OSF repository is also where the latest versions of the smoking state transition probability estimates produced by the applying the functions in the smktrans package are made available.
R package and workflow code:
- Gillespie D, Webster L, Leeming G, Opazo-Breton M, Brennan A ([YEAR]). smktrans: An R Package for estimating smoking state transition probabilities from cross-sectional survey data. R package version [x.x.x]. University of Sheffield. Open access code: https://github.com/STAPM/smktrans. url: https://stapm.gitlab.io/r-packages/smktrans/. doi: https://doi.org/10.17605/OSF.IO/YGXQ9
Smoking state transition probabilities for England:
- Smoking state transition probabilities for England. Available from https://osf.io/zy9q8 (Accessed [DATE]). Produced by code in: Gillespie D, Webster L, Leeming G, Opazo-Breton M, Brennan A ([YEAR]). smktrans: An R Package for estimating smoking state transition probabilities from cross-sectional survey data. R package version [x.x.x]. University of Sheffield. Open access code: https://github.com/STAPM/smktrans. url: https://stapm.gitlab.io/r-packages/smktrans/. doi: https://doi.org/10.17605/OSF.IO/YGXQ9
Smoking state transition probabilities for Scotland:
- Smoking state transition probabilities for Scotland Available from https://osf.io/wtz7b (Accessed [DATE]). Produced by code in: Gillespie D, Webster L, Leeming G, Opazo-Breton M, Brennan A ([YEAR]). smktrans: An R Package for estimating smoking state transition probabilities from cross-sectional survey data. R package version [x.x.x]. University of Sheffield. Open access code: https://github.com/STAPM/smktrans. url: https://stapm.gitlab.io/r-packages/smktrans/. doi: https://doi.org/10.17605/OSF.IO/YGXQ9
The motivation for smktrans was to develop a set of functions to
estimate the transition probabilities among current, former and never
smoking states (i.e. initiation, quitting and relapse) from repeat
cross-sectional survey data.
The code was developed originally for the Health Survey for England, and subsequently extended to work with the Scottish Health Survey (SHeS). Subsequent versions will be extended to Wales.
What the software does in general and how it relates to data is documented in the vignettes under “Technical Documentation”.
The inputs are:
- Several years of annual cross-sectional survey data on smoking.
- Estimates of long-term smoking relapse.
- Estimates of smoking related disease risk and mortality rates.
The processes applied by the functions in smktrans implement
calculations to estimate and forecast the transition probabilities to
get them into a form that can be used in a model that simulated the
population dynamics of smoking.
The output of these processes are the probabilities of smoking
initiation, quitting and relapse, stratified by age, period, sex and
quintiles of the Index of Multiple Deprivation. These data can be saved
so that you don’t need to run the estimation processes in smktrans
each time you want to run a model that uses the estimated transition
probabilities.
smktrans is publicly available via Github.
By default the user should install the latest tagged version of the package. Otherwise, if you want to reproduce project work and know the version of the package used, install that version.
If on a University of Sheffield managed computer, install the R, RStudio
and Rtools bundle from the Software Centre. Install Rtools - using the
installr
package can make this easier. Then install the latest or a specified
version of smktrans from Github with:
#install.packages("devtools")
devtools::install_git(
"https://github.com/stapm/hseclean.git",
#ref = "x.x.x",
build_vignettes = FALSE, quiet = TRUE)
devtools::install_git(
"https://github.com/stapm/tobalcepi.git",
#ref = "x.x.x",
build_vignettes = FALSE, quiet = TRUE)
devtools::install_git(
"https://github.com/stapm/mort.tools.git",
#ref = "x.x.x",
build_vignettes = FALSE, quiet = TRUE)
devtools::install_git(
"https://github.com/stapm/smktrans.git",
#ref = "x.x.x",
build_vignettes = FALSE, quiet = TRUE)
# ref = "x.x.x" is the version to install - change to the version you want e.g. "1.2.3"Or clone the package repo locally and use the ‘install and restart’ button in the Build tab of RStudio. This option is more convenient when testing development versions.
Then load the package, and some other packages that are useful. Note
that the code within smktrans uses the data.table::data.table()
syntax.
# Load the package
library(smktrans)
library(tobalcepi)
library(hseclean)
library(mort.tools)
# Other useful packages
library(data.table)
library(dplyr) # for data manipulation and summary
library(magrittr) # for pipes
library(ggplot2) # for plotting