Table 1 Summary of recommendations or considerations from STROBE, ROBINS-I and Sterne et al. guidelines

Patient Selection

State eligibility criteria

State inclusion and exclusion criteria of study participants, including criteria concerning missing data

Report the number of individuals at each stage of the study

Give reasons for exclusion at each stage

Indicate the amount of individuals discarded due to missingness at each stage of the study

Give consideration to selection bias introduced by exclusion criteria

May use a flowchart to summarise

Modelling and Covariate Selection

Covariates

Detail whether included as continuous or categorical and, if relevant, detail how the quantitative covariate was categorised

Consider departures from linearity for continuous covariates and state which transformation, if any, was used

State analysis model

make it clear which method will be used to model the data

Covariate Selection

describe the procedure used to reach the final model

this includes, but is not restricted to, missing data imputation, transformation of covariates, interactions between covariates or inclusion of covariates for a priori reasons

Results

Provide unadjusted estimates and the final adjusted model

State the number of participants included in unadjusted and adjusted analyses

Missing Data

Report the number of participants with missing data

Report this for each covariate of interest or the number of complete data for the important covariates

Give reasons for missing values

Investigate if there are key differences between those observed and those with missing data - this may be compared across exposure/intervention groups.

Missing data methods (general)

Which method was used to handle missing data?

State clearly the method used

State any missing data assumptions that were made

Such as whether the data are MCAR, MAR or MNAR

Sensitivity analysis

Should investigate robustness of findings

Compare method with a complete-case analysis

If necessary, assess validity of methods if there are differences

Assess plausibility of missing data assumptions

Multiple Imputation

Give details of the imputation model

State the software used and key settings for imputation model

State the number of imputations used

State variables included in imputation model

State how non-normal or binary covariates were handled

Were interactions in analysis model included in imputation model?

If a large fraction of data are imputed, compare observed and imputed values

Missing data assumptions

Discuss if variables included in the imputation model make MAR assumption plausible

Sensitivity analyses

Compare MI results with CC results

Investigate departures from MAR assumption

If necessary, suggest explanations for why there are differences in results across sensitivity analyses