OpenSAFELY: COVID-19 vaccine coverage and effectiveness in kidney disease patients

This is the code and configuration for OpenSAFELY.

• A preprint of the coverage analyses is available here
• The protocol draft is here
• If you are interested in how we defined our variables, take a look at the study definition; this is written in python, but non-programmers should be able to get a relatively good idea of what is going on here
• If you are interested in how we defined our code lists, look in the codelists folder.
• Developers and epidemiologists interested in the framework should review the OpenSAFELY documentation

Summary of analysis code for factors associated with vaccine uptake

(1) analysis/data_process.R

• Reads in inputdata.csv (derived from analysis/study_definition.py)
• Sets variable classes
• Calculates eGFR from most recent creatinine reading
• Defines kidney disease subgroups based on eGFR and UK Renal Registry status (with the latter used for dialysis and kidney transplant recipients)
• Calculates sequential vaccination dates and assigns vaccine type to each date
• Outputs data_processed.rds and data_processed.csv

(2) analysis/coverage/data_selection.R

• Reads in data_processed.rds
• Applies following selection criteria to define final study population for coverage analyses: (i) Aged 16+ with eGFR<60 in 2 years before 01 Dec 2020 or in UK Renal Registry on 31 Dec 2020; (ii) No ambiguities in kidney disease status (primary care code indicating dialysis or kidney transplant but not in UK Renal Registry) (iii) No missing demographic information (sex, IMD, ethnicity, region); (iv) Maximum of 5 doses recorded and no vaccines administered at an interval of <14 days between doses 1 and 4.
• Also selects nested cohorts for logistic regression sensitivity analysis (uncensored throughout follow-up) and dose 4 analysis (at least one indicator for a third primary or second booster dose)
• Outputs data_cohort.rds, data_cohort.csv, data_cohort_coverage_logistic.rds, data_cohort_coverage_dose4.rds, and flowchart.csv

(3) analysis/coverage/table_1_coverage.R

• Generates table 1 of demographic/clinical covariates in study population (full and stratified by kidney disease subgroup)
• Outputs table1_coverage_redacted_by_CKD.html and table1_coverage_redacted_by_CKD.csv (full study population)
• Outputs table1_coverage_redacted_by_CKD_dose4.html and table1_coverage_redacted_by_CKD_dose4.csv (dose 4 subset)

(4) analysis/coverage/cox_model.R

• Reads in data_cohort.rds
• Calculates Cox proportional hazard models for time to completion of 3- or 4-dose primary vaccine series, censoring at death, de-registration, or analysis cut-off
• 25 variables of interest are defined in the vector var_list
• Minimally, partially, and fully adjusted models are calculated for each covariate, stratified by region
• Outputs data_cox_coverage_{dose}_{subset}.rds, mod_strat_coxph_redacted_{dose}_{subset}.rds, and mod_strat_coxph_redacted_{dose}_{subset}.csv

(5) analysis/coverage/logistic_model.R (replicates cox_model.R but with logistic regression)

• Reads in data_cohort_coverage_logistic.rds
• Calculates logistic regression models exploring factors associated with odds of receiving 3-dose primary series by analysis cut-off
• Outputs data_lr.rds , mod_strat_logistic_redacted.rds, and mod_strat_logistic_redacted.csv

(6) analysis/coverage/vaccine_coverage.Rmd

• Visualises the outputs of table_1_coverage.R
• Calculates and plots dose 1, 2, 3, 4, and 5 cumulative coverage over time (Kaplan-Meier estimates with step counts delayed until accrual of 10 events)
• Tabulates frequency tables for products, product combinations, and dosing intervals at doses 1–4 (rounding to nearest 5 with small number redaction)
• Tabulates and plots the outputs of cox_model.R (with rounding to nearest 5 and small number redaction)
• Tabulates and plots the outputs of logistic_model.R (with rounding to nearest 5 and small number redaction)

(6) analysis/coverage/cox_model_check_coverage.Rmd

• Run separately for each outcome (dose 3/dose 4) and each analysis subset
• Visualises survival curves and Schoenfeld residual plots for outputs of cox_model.R, enabling proportional hazards assumption to be checked

Summary of analysis code for estimating comparative effectiveness of different vaccine regimens

Under development

About the OpenSAFELY framework

The OpenSAFELY framework is a Trusted Research Environment (TRE) for electronic health records research in the NHS, with a focus on public accountability and research quality.

Read more at OpenSAFELY.org.

Licences

As standard, research projects have a MIT license.