TABLE 1.
Contraception and pregnancy model parameters
| Parameter | Description | Value | Calculated from |
|---|---|---|---|
| Method_Use_In_2010 | Proportion of women using each method in 2010, by age. | Table A1.2 in online Appendix 1 | From DHS 2010 contraception calendar data (The Demographic and Health Surveys Program 2018), for each year of age of the woman from 15 to 49, the proportion of women who are using each of the 10 contraception methods (co_contraception, Table 2). |
| Pregnancy_NotUsing_In_2010 | Probability per year of a woman not on contraceptive becoming pregnant, by age. | Table A1.2 in online Appendix 1 | From Malawi DHS 2010 data (National Statistical Office (NSO) and ICF Macro 2011) using the data on births in the last year for each woman, and estimates of the relative risk of pregnancy given each contraceptive method and the proportion of women using each contraceptive method, as explained in online Appendix 1. |
| Pregnancy_NotUsing_HIVeffect | Relative probability of becoming pregnant while not using a contraceptive for HIV‐positive women compared to HIV‐negative women. |
15–19: 1.4 20–24: 0.9 25–29: 0.8 30–34: 0.7 35–39: 0.5 40–44: 0.4 45–49: 0.3 |
From Marston, Zaba, and Eaton 2017 (Marston, Zaba, and Eaton 2017) Figure 1(a), age‐specific fertility rates (by five‐year age group: 15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49. |
| Age_specific_fertility_rates | Age‐specific fertility rates used for scheduling births in the first 9 months of the simulation. This is necessary because at the initiation of the simulation no women are pregnant. |
15–19: 0.144 20–24: 0.239 25–29: 0.213 30–34: 0.174 35–39: 0.123 40–44: 0.061 45–49: 0.022 |
Data table from official source (United Nations World Population Prospects (United Nations Department of Economic and Social Affairs Population Division 2019)) for age‐specific fertility rates (by five‐year age group: 15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49) and calendar period (2010‐2014). |
| Scaling_factor_on_monthly_risk_of_pregnancy | Scaling factor (by five‐year age group: 15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49) on the monthly risk of pregnancy and contraceptive failure rate. |
15–19: 1.227 20–24: 0.799 25–29: 0.829 30–34: 0.809 35–39: 0.749 40–44: 0.645 45–49: 0.941 |
Calibrated so that, at the beginning of the simulation, the age‐specific monthly probability of a woman having a live birth matches the United Nations World Population Prospects age‐specific fertility rate value for the same year. Calibration is done in two stages (explained in detail in online Appendix 1): 1. The scaling_factor_on_monthly_risk_of_pregnancy is used to induce the correct number of age‐specific births initially, given the initial pattern of contraceptive use. 2. Trends over time in the risk of starting (time_age_trend_in_initiation) and stopping (time_age_trend_in_stopping) contraception, adjusted for older women, are used to induce the correct trend in the number of live births over time. |
| Initiation_ByMethod | Probability per month of a woman who is not using any contraceptive method of starting use of a method, by method. | Table A2.1 in online Appendix 2 | This was calculated from an analysis of DHS contraception calendar data (The Demographic and Health Surveys Program 2018) for Malawi DHS 2010 (National Statistical Office (NSO) and ICF Macro 2011), in Stata, as explained in online Appendix 2. |
| Initiation_AfterBirth | The probability of a woman starting a contraceptive immediately after birth, by method. | Table A2.2 in online Appendix 2 | This was calculated from an analysis of DHS contraception calendar data (The Demographic and Health Surveys Program 2018) for Malawi DHS 2010 (National Statistical Office (NSO) and ICF Macro 2011), in Stata, as explained in online Appendix 2. |
| Prob_Switch_From | The probability per month that a woman switches from one method of contraceptive to another is conditional that she will not discontinue use of the method. | Table A3.1 in online Appendix 3 | This was calculated from an analysis of Malawi DHS 2016 (National Statistical Office (NSO) [Malawi] and ICF 2017) contraception calendar data (The Demographic and Health Surveys Program 2018), in Stata, via competing risks regression, as explained in online Appendix 3. |
| Prob_Switch_From_And_To | The probability of switching to a new method, by method, conditional that the woman will switch to a new method. | Table A3.2 in online Appendix 3 | This was calculated from an analysis of Malawi DHS 2016 (National Statistical Office (NSO) [Malawi] and ICF 2017) contraception calendar data (The Demographic and Health Surveys Program 2018), in Stata, as explained in online Appendix 3. |
| Failure_ByMethod | Probability per month of a woman on a contraceptive becoming pregnant, by method. | Table A3.3 in online Appendix 3 | These parameters were calculated from an analysis of Malawi DHS 2016 (National Statistical Office (NSO) [Malawi] and ICF 2017) a contraception calendar data (The Demographic and Health Surveys Program 2018), in Stata, via a competing risks regression model (separate to that for switching), as explained in online Appendix 3. |
| Discontinuation_ByMethod | The probability per month of discontinuing use of a method, by method. | Table A3.4 in online Appendix 3 | |
| rr_fail_under25 | The relative risk of becoming pregnant while using a contraceptive for women younger than 25 years compared to older women. | 2.2 | From Guttmacher analysis (Polis et al. 2016) Table 9, page 52; see online Appendix 4 Table A4.1. b |
| Initiation_ByAge | The effect of age on the probability of starting use of contraceptive (add one for multiplicative effect). | Table A4.2 in online Appendix 4 | This was calculated from an analysis of Malawi DHS 2016 (National Statistical Office (NSO) [Malawi] and ICF 2017) contraception calendar data (The Demographic and Health Surveys Program 2018), in Stata, using fractional polynomial regression (better fitting model, higher F statistic), as explained in online Appendix 4. The results of this model are plotted in online Figure A4.1 and are used to calculate the proportionate difference in Initiation_ByMethod probability from the average probability for each age in years. c |
| Discontinuation_ByAge | The effect of age on the probability of discontinuing use of contraceptives (add one for multiplicative effect). | Table A4.3 in online Appendix 4 | This was calculated from an analysis of Malawi DHS 2016 (National Statistical Office (NSO) [Malawi] and ICF 2017) contraception calendar data (The Demographic and Health Surveys Program 2018), in Stata, using fractional polynomial regression, as explained in online Appendix 4. The results of this model are plotted in online Figure A4.2 and are used to calculate the proportionate difference in Discontinuation_ByMethod probability from the average probability for each age in years. |
| Interventions_Pop | Pop (population scale contraception intervention) intervention multiplier of Initiation_ByMethod. Representing the method‐specific proportional increases due to Pop intervention. | Table A5.1 in online Appendix 5 | Calibrated to meet the expected changes in the percentage of women using each method from our 2010 baseline to 2020 as a result of the Malawi Costed Implementation Plan For Family Planning 2016–2020 (CIP), estimated on pages 37–38 (Figure 31) of the CIP report (Government of Malawi 2015). Explained in detail in online Appendix 5. |
| Interventions_PPFP | PPFP (postpartum family planning) intervention multiplier of Initiation_AfterBirth. Representing the method‐specific proportional increases due to PPFP intervention. | Table A5.1 in online Appendix 5 |
We use DHS 2016 data only because the reason for discontinuation (failure) is not in the DHS 2010 contraception calendar data.
Other “lifestyle” variables potentially associated with increased probabilities of failure (marital status, parity, wealth, urban‐rural, education) were not included because the effect estimates for these were not statistically significant for >50 percent of those using contraception and only significant for one or two minor contraception categories–see Table A4.1— in Appendix 4.
We note a similar analysis was done for Initiation_AfterBirth though the model did not result in statistically significant or large enough effect estimates to be considered important enough to include. A simpler model with age and age‐squared was also not significant, and a very simple model with just age resulted in the Initiation_AfterBirth probabilities only changing by ∼±10–15 percent throughout the 15–49 age range (p = 0.03), that is, not an important enough change to make it worth adding an additional parameter (Initiation_AfterBirth_ByAge), especially given Initiation_AfterBirth is much rarer than initiation (Initation_ByMethod).