2. Risk of bias assessments using ROBINS‐I tool.

Study	D1a	D2b	D3c	D4d	D5e	D6f	D7g	Overall
Seeland 2020	Serious Use propensity score matching but unclear what variables they are matching on; may not have accounted for all confounders as listed in our protocol	Moderate Selection based on COVID‐19 positivity in electronic health record, any factors affecting positivity are likely to be similar between groups	Moderate Used clear Rx Norm codes to define combined contraceptives, but no ascertainment of actual patient use of contraception during the timeframe of the outcome	Critical Actual use of contraception is not ascertained	No information	Low Outcome = mortality derived from electronic health record, likely to be accurately reported	Low	Serious Use propensity score matching but we are not certain what the actual variables were included in this; also not sure whether patients are using contraception during time of outcome
Mujumdar 2020	Serious Concern for possible residual confounding especially as there is no information on what was adjusted for and whether this was a valid adjustment	Moderate Selection based on COVID‐19 positivity in electronic health record of 1 tertiary care facility, any factors affecting positivity are likely to be similar between groups	Moderate Contraceptive prescription derived from medical record data, but no ascertainment of actual patient use of contraception during the timeframe of the outcome	Critical Actual use of contraception was not ascertained	Serious Outcome data missing for ~7% of the total sample and those individuals were excluded from analysis	Low Outcome = hospitalization derived from electronic health record, likely to be accurately reported	Low	Critical Unknown confounders adjusted for unknown if patients using contraception during timeframe of outcome
Costeira 2021	Serious Adjusted for age, BMI, and smoking status. No information on personal history of DVT. However, variables were self‐reported and uncertain if they were measured validly. Variables such as smoking status likely subject to social desirability bias in self‐reporting	Serious Participants self‐selected to use application; possible that people who are using contraception are more likely/interested in using app as they are more health savvy	Moderate Use and contraceptive type based on user self‐report	Moderate Reported by patients they were taking the intervention	Serious Participants with missing smoking status were excluded from analyses	Moderate Outcome = hospitalization, but self‐reported so subject to incorrect reporting (e.g. individual reports an emergency room visit as hospitalization)	Low	Serious Has better classification of intervention and likely adherence to intervention, but all self‐report data and subject to bias

^aD1: bias due to confounding.
^bD2: bias in selection of participants into the study.
^cD3: bias in classification of interventions.
^dD4: bias due to deviations from intended interventions.
^eD5: bias due to missing data.
^fD6: bias in measurement of outcomes.
^gD7: bias in selection of the reported result.
BMI: body mass index; DVT: deep vein thrombosis.