Fig 1.
Causal graph illustrating the assumed causal relations between variables when considering religious/spiritual beliefs and behaviours (RSBB) 5 years post-delivery as the exposure and mental health (MH) 6 years post-delivery as the outcome.
Note the bidirectional arrows between all baseline covariates in pregnancy; as these were all measured at approximately the same time the causal relations are not known with certainty.
Fig 2.
Causal graph illustrating the assumed causal relations between variables when considering mental health (MH) 2 years post-delivery as the exposure and religious/spiritual beliefs and behaviours (RSBB) 5 years post-delivery as the outcome.
Note the bidirectional arrows between all baseline covariates in pregnancy; as these were all measured at approximately the same time the causal relations are not known with certainty.
Fig 3.
Results of the mothers analyses with categorical religious/spiritual belief and behaviour (RSBB) exposures and standardised depression and anxiety scores as outcomes (n
= 3,856). Results in black are from unadjusted analyses, and those in red from adjusted analyses (adjusting for baseline confounders, RSBB and mental health). The dashed vertical line at ‘0’ indicates a null association. Error bars denote 95% confidence intervals. Taking religious attendance, for instance, relative to the baseline category of ‘not at all’ attend, in adjusted analyses individuals who attended ‘a minimum of once a week’ were associated with a 0.09 unit (95% confidence interval between -0.22 and 0.04) decrease in standardised depression scores. Full results are in Table S8 of the Supporting Information.
Fig 4.
Results of the partners analyses with categorical religious/spiritual belief and behaviour (RSBB) exposures and standardised depression and anxiety scores as outcomes (n
= 1,940). Results in black are from unadjusted analyses, and those in red from adjusted analyses (adjusting for baseline confounders, RSBB and mental health). The dashed vertical line at ‘0’ indicates a null association. Error bars denote 95% confidence intervals. Full results are in Table S11 of the Supporting Information.
Fig 5.
Results of the interaction analyses assessing whether the adjusted mother and partner results differ, with categorical religious/spiritual belief and behaviour (RSBB) exposures and standardised depression and anxiety scores as outcomes.
Results in black are for the depression outcome, and those in red for anxiety. The dashed vertical line at ‘0’ indicates no difference between mothers and partners, with results below 0 meaning that the estimate was lower in mothers, compared to partners. For instance, for ‘Yes’ to religious belief and anxiety (relative to answering ‘No’), the mean difference in mothers was -0.05, while in partners it was 0.15, giving a standardised mean difference here of approximately -0.20. Error bars denote 95% confidence intervals. Full results are in Table S13 of the Supporting Information.
Fig 6.
Results of the mothers analyses with standardised depression and anxiety scores as exposures and categorical religious/spiritual belief and behaviour (RSBB) as outcomes (n
= 4,025). Results in black are from unadjusted analyses, and those in red from adjusted analyses (adjusting for baseline confounders, RSBB and mental health). This plot displays the predicted change in the probability of the RSBB outcome for a one-standardised-unit increase in the mental health exposure, based on the associated multinomial regression model. For instance, in adjusted analyses a one-standardised-unit increase in depression scores predicts a 1.90%-point increase in the probability of answering ‘Never’ for religious attendance (95% confidence interval = 0.42 to 3.38). The dashed vertical line at ‘0’ indicates a null association. Error bars denote 95% confidence intervals. Full results of the multinomial models are in Table S16, with predicted probabilities in Table S17, of the Supporting Information.
Fig 7.
Results of the partners analyses with standardised depression and anxiety scores as exposures and categorical religious/spiritual belief and behaviour (RSBB) as outcomes (n
= 2,120). Results in black are from unadjusted analyses, and those in red from adjusted analyses (adjusting for baseline confounders, RSBB and mental health). This plot displays the predicted change in the probability of the RSBB outcome for a one-standardised-unit increase in the mental health exposure, based on the associated multinomial regression model. For instance, in adjusted analyses a one-standardised-unit increase in depression scores predicts a 1.66%-point increase in the probability of answering ‘Never’ for religious attendance (95% confidence interval = -0.49 to 3.81). The dashed vertical line at ‘0’ indicates a null association. Error bars denote 95% confidence intervals. Full results of the multinomial models are in Table S19, with predicted probabilities in Table S20, of the Supporting Information.
Fig 8.
Results of the interaction analyses assessing whether the adjusted mother and partner results differ, with standardised continuous mental health, or binary probable depression and anxiety diagnoses, as exposures and categorical religious/spiritual beliefs and behaviours (RSBB) as outcomes.
Results in black are for the depression outcome, and those in red for anxiety. The dashed vertical line at ‘1’ indicates no difference between mothers and partners, with results below 1 meaning that the relative risk ratio (RRR) estimate was lower in mothers, compared to partners. For instance, for ‘Christian’ religious affiliation (relative to ‘None’) and continuous depression scores, a one-standardised-unit increase in depression was associated with a RRR of 0.92 in mothers, while in partners it was 1.14, giving a ratio of these RRRs here of approximately 0.81. Error bars denote 95% confidence intervals. Full results are in Table S21 of the Supporting Information.