3.2 Kaplan Meier Analysis
Kaplan-Meier estimation of the survival function was conducted in R Studio using the R programming language. We plotted Kaplan Meier Curve using R Function “Survfit” of Survival Library. Initially, we plotted Kaplan–Meier curves for the entire 60 months (5 years). For simplicity, we have reduced the analysis time to 36 months. Table 2 presents the Kaplan–Meier table.
Table 1
Kaplan Meier Estimation of Survival Table
Time (months) | Number at risk | Number of events | Survival probability | Standard error | Lower 95% CI | Upper 95% CI |
1 | 5572 | 1784 | 0.68 | 0.00625 | 0.668 | 0.692 |
2 | 3788 | 553 | 0.581 | 0.00661 | 0.568 | 0.594 |
3 | 3159 | 292 | 0.527 | 0.0067 | 0.514 | 0.54 |
6 | 2381 | 450 | 0.442 | 0.00672 | 0.429 | 0.455 |
12 | 1545 | 546 | 0.328 | 0.00653 | 0.315 | 0.341 |
24 | 669 | 449 | 0.211 | 0.00616 | 0.2 | 0.224 |
36 | 296 | 105 | 0.171 | 0.00616 | 0.16 | 0.184 |
Table 2
Descriptive Statistics of Study Sample
Characteristic | N = 56821 |
Respondent current age | 31 (26–36) |
Total number of children born | 3.00 (2.00–4.00) |
Ideal number of children | 3.00 (3.00–4.00) |
Number of household members | 5.00 (4.00–6.00) |
Antenatal care provider | |
Medical doctor | 239 (4.2%) |
No Antenatal care | 129 (2.3%) |
Nurse and/or Midwife | 5–314 (94%) |
Last children delivered by | |
Medical Doctor | 1–296 (23%) |
Nurse and/or Midwife | 4–043 (71%) |
Others (CHW - Relatives - Herself) | 342 (6.0%) |
Place of delivery | |
Health center | 3–372 (59%) |
Home | 286 (5.0%) |
Private clinic/polyclinic | 117 (2.1%) |
Provincial/district hospital | 1–483 (26%) |
Referral hospital | 341 (6.0%) |
Highest Level of education | |
No Education | 645 (11%) |
Primary | 3–662 (64%) |
Secondary | 1–148 (20%) |
Higher | 228 (4.0%) |
Type of place of Residence | |
Rural | 4–682 (82%) |
Urban | 1 − 000 (18%) |
Last child delivered by Cesarean Section | 889 (16%) |
Have health insurance | 4–599 (81%) |
Respondent currently working | 4–311 (76%) |
Province & Region | |
Kigali | 770 (14%) |
South | 1–184 (21%) |
West | 1–255 (22%) |
North | 906 (16%) |
East | 1–567 (28%) |
Sex household head | |
Male | 4–298 (76%) |
Female | 1–384 (24%) |
1Median (IQR); n (%) |
Estimation of Contraceptive After Delivery
In the Kaplan-Meier survival analysis, we estimated the probability of not using the contraceptive method after a certain time following delivery. This is mostly due to the fact that our event of interest was the resumption of contraceptive use after delivery. In this sense, the probability of survival is the probability that a particular woman does not resume contraceptive use after a certain time. Using this probability, we calculated the complementary probability and obtained the probability of contraceptive use at a particular time.
one month after delivery, on average, 32% of women had already started a modern contraceptive method. This increased to 47% after three months and to 55% after six months. Note that these are the cumulative probabilities. One year after delivery, 67% of women started modern contraceptive methods. A total of 79% and 83% of women started modern contraceptives within two and three years after delivery, respectively. Table 3 and Fig. 2 show the probabilities.
Table 3
Probability for Contraceptive use at certain time after deliver
Time (Months) | Contraceptive use Probability |
1 | 0.3201723 |
2 | 0.4194185 |
3 | 0.4730842 |
6 | 0.5582429 |
12 | 0.6721959 |
24 | 0.7886684 |
36 | 0.8288316 |
3.3 Cox Hazard Proportional Model
The Cox proportional hazards model was implemented using the R Programming Language Package. The Cox Model was used to study the effects of demographic and socioeconomic factors on the timing of contraceptive use after childbirth. Covariates were selected based on the recent literature on the factors that influence family planning behaviors. Cox Model analysis consisted of univariate and multivariate analyses. In addition, the hazard proportionality assumption was tested using the scaled Schoenfeld Residuals.
Bivariate Cox Model Analysis
Initially, 24 variables were selected for bivariate Cox proportional hazard model analysis. The evaluation of the model results was based on hazard ratio, 95% confidence intervals, and its and P values. The significance level was set at P < 0.05. Covariates with P values below the significance level and with hazard ratios that were significantly different from one were selected for multivariate analysis. Table 4 shows the results of the bivariate analysis.
Bivariate analysis revealed that an increased number of antenatal visits, an increased number of children under five years, women who delivered by cesarean section, women with access to health insurance, and Catholic and Muslim women were associated with an increased usage of contraceptive methods after delivery. In addition, women who had any level of education were associated with a higher use of postpartum family planning compared to women with no education. Compared to women living in Kigali, those living in the North, South, and Eastern provinces had higher odds of using postpartum contraceptives.
On the other hand, an increase in women’s age, number of household members, number of children born, and ideal number of children were associated with reduced chances of using family planning after delivery. Women who delivered at home were significantly associated with a reduction in the use of postpartum family planning compared with women who delivered at health centers. Women who never lived in union, were separated, divorced, or widowed were associated with reduced use of postpartum family planning compared to women who were married and/or living with partners. Lastly, women who reported challenges in obtaining money for healthcare and those who reported difficulty with distance to health facilities were associated with a reduced chance of using postpartum family planning. Similarly, women who lived in households headed by Female were associated with a reduced use of postpartum family planning.
Multivariable Cox Model Analysis
Significant and non-negligible variables in the univariate analysis were selected for multivariate analysis. In total, 16 variables were selected for the multivariable analysis. The goodness-of-fit of the model was evaluated using the likelihood ratio test with a P value below 0.001. Individual covariates were evaluated based on their coefficients, hazard ratio, and associated P-values. Table 6 presents the results of the multivariable Cox Analysis. The test for the Cox proportional hazard assumption was performed using scaled Schoenfeld Residuals.
Table 4
Univariate Analysis with Cox Proportional Hazard Model
Characteristic | HR1 | 95% CI1 | P - value |
Respondent current age | 0.98 | 0.97–0.98 | < 0.001 |
Total children ever born | 0.94 | 0.93–0.95 | < 0.001 |
Number of antenatal visits | 1.09 | 1.06–1.13 | < 0.001 |
Ideal number children | 0.91 | 0.89–0.93 | < 0.001 |
Timing of 1st antenatal | 0.94 | 0.92–0.96 | < 0.001 |
Age in 5 - year groups | 0.89 | 0.87–0.91 | < 0.001 |
Number of household members | 0.93 | 0.92–0.95 | < 0.001 |
Number children under5 in household | 1.09 | 1.04–1.14 | < 0.001 |
Antenatal care provider | | | |
Medical doctor | — | — | |
No antenatal care | 0.69 | 0.52–0.91 | 0.008 |
Nurse/midwife | 1.02 | 0.88–1.19 | 0.8 |
Last child delivered by | | | |
Medical doctor | — | — | |
Nurse/midwife | 0.90 | 0.84–0.97 | 0.005 |
Other | 0.52 | 0.45–0.61 | < 0.001 |
Place of delivery | | | |
Health center | — | — | |
Home | 0.53 | 0.46–0.63 | < 0.001 |
Other | 0.75 | 0.58–0.98 | 0.033 |
Private clinic/polyclinic | 0.83 | 0.66–1.05 | 0.12 |
Provincial/district hospital | 1.06 | 0.99–1.13 | 0.12 |
Referral hospital | 1.02 | 0.89–1.16 | 0.8 |
Religion | | | |
Adventist | — | — | |
Catholic | 1.15 | 1.04–1.27 | 0.005 |
Muslim | 1.26 | 1.00–1.58 | 0.047 |
Other | 0.89 | 0.70–1.14 | 0.4 |
Protestant | 0.99 | 0.90–1.09 | 0.8 |
Current marital status | | | |
Married/Living with partner | — | — | |
Never in union | 0.54 | 0.48–0.60 | < 0.001 |
Widowed/Divorced/Separated | 0.46 | 0.41–0.52 | < 0.001 |
Highest Level of education | | | |
No Education | — | — | |
Primary | 1.38 | 1.24–1.53 | < 0.001 |
Secondary | 1.42 | 1.26–1.60 | < 0.001 |
Higher | 1.32 | 1.11–1.58 | 0.002 |
Wealth Category | | | |
Poorest | — | — | |
Poorer | 1.11 | 1.02–1.22 | 0.020 |
Middle | 1.15 | 1.05–1.25 | 0.003 |
Richer | 1.06 | 0.97–1.17 | 0.2 |
Richest | 0.97 | 0.89–1.07 | 0.6 |
Type pf Residence | | | |
Rural | — | — | |
Urban | 0.90 | 0.84–0.97 | 0.007 |
Had terminated pregnancy | | | |
No | — | — | |
Yes | 0.87 | 0.80–0.95 | 0.001 |
Delivered by caesarean section | | | |
No | — | — | |
Yes | 1.27 | 1.17–1.38 | < 0.001 |
Have medical insurance | | | |
No | — | — | |
Yes | 1.24 | 1.14–1.34 | < 0.001 |
Problem getting money for medical | | | |
No | — | — | |
Yes | 0.89 | 0.84–0.94 | < 0.001 |
Problem distance to health facility | | | |
No | — | — | |
Yes | 0.89 | 0.83–0.96 | 0.002 |
Respondent currently working | | | |
No | — | — | |
Yes | 1.02 | 0.95–1.10 | 0.6 |
Province & Region | | | |
Kigali | — | — | |
South | 1.17 | 1.05–1.31 | 0.005 |
West | 0.89 | 0.79–0.99 | 0.035 |
North | 1.34 | 1.19–1.50 | < 0.001 |
East | 1.22 | 1.10–1.36 | < 0.001 |
Sex of household head | | | |
Male | — | — | |
Female | 0.64 | 0.59–0.68 | < 0.001 |
1 HR = Hazard Ratio - CI = Confidence Interval |
Multivariable analysis revealed that women who delivered via cesarean section had an increased chance of starting family planning early compared to women who delivered naturally (hazard ratio:1.18, CI:1.07–1.3, P-Value < 0.001). Women who delivered by cesarean section had an 18% increased chance of using modern contraceptives after delivery compared with women with natural birth. Compared to women with no health insurance, women who had access to health insurance had 15% increased odds of using modern contraceptives after delivery (hazard ratio:1.15, CI: 1.05–1.26, P-Value < 0.01). The increase in the number of under 5-years children in households increased the use of postpartum family planning (hazard ratio:1.17, CI:1.11–1.25, P-value < 0.001). An increase of one under 5-year child, increases the odds of utilizing postpartum contraceptives by 17%.
In multivariable analysis, the total number of children born was associated with an increased use of postpartum family planning (hazard ratio:1.04, CI:1.01–1.08, P-value < 0.05). This is different from the bivariate analysis, in which the number of children reduced the odds of using postpartum contraceptives. The use of postpartum contraceptives varies across the Provinces and Regions. Compared to women who lived in Kigali, those who lived in the South, East, and Northern provinces had increased use of postpartum family planning. The effect of religion was statistically significant in Catholic Women who were more likely to use contraceptives than in Adventists.
Multivariable Cox Proportional Hazard Model |
---|
Variable Name | Coefficient | HR1 | 95% CI HR2 | Z | P value |
Age in 5-year groups | -0.18498 | 0.83 | 0.8 | 0.87 | -8.63 | 0.00000 |
Number of antenatal visits during pregnancy | 0.04681 | 1.05 | 0.99 | 1.11 | 1.699 | 0.08919 |
Timing of 1st antenatal check months | -0.00896 | 0.99 | 0.96 | 1.02 | -0.543 | 0.58702 |
Number of household members | -0.03794 | 0.96 | 0.94 | 0.99 | -2.717 | 0.00657 |
Current Marital Status (Married as reference) |
Never in Union | -0.9688 | 0.38 | 0.33 | 0.44 | -13.25 | 0.00000 |
Divorced, Separated, Widowed | -0.7788 | 0.46 | 0.4 | 0.53 | -10.66 | 0.00000 |
Had terminated pregnancy | -0.12186 | 0.89 | 0.8 | 0.98 | -2.45 | 0.01415 |
Delivered by Cesarean delivery | 0.16492 | 1.18 | 1.07 | 1.3 | 3.449 | 0.00056 |
Have insurance | 0.1382 | 1.15 | 1.05 | 1.26 | 2.963 | 0.00304 |
Region (reference as Kigali) |
South | 0.16233 | 1.18 | 1.02 | 1.36 | 2.217 | 0.02657 |
West | -0.12637 | 0.88 | 0.76 | 1.02 | -1.698 | 0.08941 |
North | 0.25628 | 1.29 | 1.1 | 1.51 | 3.191 | 0.00141 |
East | 0.27882 | 1.32 | 1.14 | 1.53 | 3.743 | 0.00018 |
Religion (reference as Adventists) |
Catholic | 0.12607 | 1.13 | 1.01 | 1.27 | 2.194 | 0.02823 |
Muslim | 0.2052 | 1.23 | 0.97 | 1.56 | 1.673 | 0.09416 |
Others | -0.03406 | 0.97 | 0.77 | 1.22 | -0.291 | 0.77068 |
Protestant | -0.02338 | 0.98 | 0.88 | 1.09 | -0.426 | 0.66952 |
Highest Education Level (reference: No education) |
Primary | 0.14646 | 1.16 | 1.02 | 1.31 | 2.287 | 0.02215 |
Secondary | 0.06157 | 1.06 | 0.92 | 1.23 | 0.810 | 0.41766 |
Higher | -0.09602 | 0.91 | 0.73 | 1.13 | -0.864 | 0.38729 |
Total children ever born | 0.04291 | 1.04 | 1.01 | 1.08 | 2.463 | 0.01375 |
Ideal number of children | -0.09978 | 0.91 | 0.88 | 0.93 | -7.026 | 0.00000 |
Problem getting money for medical | 0.02814 | 1.03 | 0.95 | 1.11 | 0.733 | 0.46331 |
Number of children under 5-years | 0.16 | 1.17 | 1.11 | 1.25 | 5.254 | 0.00000 |
Problem distance to health facility | -0.11647 | 0.89 | 0.82 | 0.97 | -2.703 | 0.00685 |
1: Hazard ratio, 2: 95% Confidence Interval for Hazard ratio |
Compared to women with no education, those with primary education had increased odds of using contraceptives after delivery. There were no significant differences between women with secondary and higher education and women with no education.
Women’s Age was associated with reduced use of postpartum family planning (hazard ratio:0.83. CI: 0.8–0.87, P-Value < 0.001). An increase from one five-year category age to another five-year category of age, was associated with reduction in use of postpartum contraceptives by 17%. There were significant differences in the use of postpartum family planning among women in different marital status categories. Women who had never been in union were associated with reduced odds of utilizing postpartum family planning (hazard ratio:0.38, CI:0.33–0.44, P-value < 0.001). Women who were divorced, separated, or widowed also had a reduction in utilizing postpartum family planning (hazard ratio:0.46, CI:0.4–0.53, P-value < 0.001).
In addition, women with a history of terminated pregnancies have reduced odds of using postpartum contraceptives. The increase in the number of household members was also associated with reduced odds of utilization of postpartum family planning (hazard ratio:0.96, CI:0.94–0.99, P-value < 0.01). An increase in the ideal number of children reported by women was also associated with reduced odds of using family planning after delivery. Similarly, women who reported challenges in accessing health facilities due to long distances were also associated with a reduced use of family planning after delivery.