Scientific Papers

Association of migration and family planning use among women in Malawi: Evidence from 2019/2020 Malawi Multiple Indicators Survey | Contraception and Reproductive Medicine

Study setting, data sources and population

Malawi is a landlocked country with an estimated area of 118,484 km2 and a 2023 projected population of 20.6 million [8]. Urban centers hold approximately 16% of the population [8].

We used a cross-sectional study design using the 2019/20 Malawi Multiple Indicator Cluster Survey (MICS). MICS is a nationally representative household survey that provides up-to-date information on sexual behavior, child health indicators, HIV, maternal health, fertility, childbearing as well as hygiene and sanitation. MICS are conducted every four to five years by the Malawi national statistics office (NSO) and Ministry of Health with technical support from the United Nations Children’s Funds (UNICEF). Details of the sampling strategy and methodology of the MICS have been described in detail elsewhere [6]. A total of 25,543 women aged 15 to 49 were asked to participate in the survey but 24,543 women participated in the survey.

Outcome, exposure and confounder variables

Outcome variable

The outcome variable was any modern contraceptive use among women of childbearing age (aged 15 to 49 years). Modern family planning use was coded 1 for currently using any modern contraceptive and 0 for not using. Modern contraceptive use was restricted to use of the following methods: female sterilization, male sterilization, IUD, injectable, implants, pills, male condom, Female condom, diaphragm, foam, jelly and lactation amenorrhea (LAM). This variable was created by recoding 1 for use of each of the above method and 0 for not using each of the above method. The overall CPR variable was generated by generating a variable using the egen rowmax command which created a variable coded 1 for any use of the above-mentioned methods and 0 for not using any method. Use of modern contraceptive was assessed by asking women if they were currently using any family planning method mentioned above. Those who responded with a yes were further asked about the method they were using to avoid pregnancy. Any method indicated above was considered modern contraceptive method.

Main exposure/ predictor variable

The main exposure was migration status. This variable described whether the participant changed residence within 12 months, coded 1 for women who recently changed residence within 12 months and 0 for women who did not change residence regardless of the type of migration involved (rural-rural, rural-urban, urban-rural as well as urban-urban). Migration status variable was generated from the duration of stay at current residence variable that ranged from 0 to 49 years by recoding 0 (representing women who changed residence for less than 12 months) into 1 representing migrants and ≥1 (representing women who did not change residence within 12 months) into 0 representing non-migrants. This definition of migration was in accordance with the United Nations definition of short-term migration [26].

Confounding variables

Covariates that were identified within the 2019/20 MICS dataset as potential confounders were participants age group (categorized into 15 to 19, 20 to 24, 25 to 29, 30 to 34, 35 to 39, 40 to 44 and 45 to 49), age at first sex (categorized into ≤15, 16 to 19 and ≥ 20), age at first marriage (categorized into ≤15, 16 to 19 and ≥ 20), marital status (categorised into married, formerly married and never married), residence (rural versus urban), region (northern, central and southern region), participants wealth index (categorized into lowest, second, middle, fourth and highest), education (categorized into pre-primary, primary, lower secondary, upper secondary, higher and vocational training), any functional disability (categorized as having any functional disability and not having) and children ever born (categorized into 0, 1 to 2, 3 to 5 and ≥ 6).

Statistical analysis

Data for this study was downloaded from MICS website ( in SPSS format and was exported into Stata 18.0 [32] for analyses. After data cleaning, we declared the data as survey data with complex design features using svyset command to allow use of sampling weights to correct unequal representation of participants at cluster, district, and regional level. All subsequent analyses utilised the svy prefix for survey data analysis. Results were stratified by married women (N = 14,934) and all women of childbearing age (N = 24,543). We grouped and recoded some variables that had more numeral values for ease of analysis and comparison such as age at first sex, age at first marriage, age and duration of residence. For variables on use of specific FP method, women who indicated they were using a method were coded 1 and those not using that particular method were recoded 0 indicating they were not using that particular method. These variables were grouped based on common reporting of these variables in literature.

To examine association between any modern contraceptive use and potential confounders one at a time, Pearson’s chi-square tests (X2) were conducted. Predictors were considered significant at p < 0.05. We investigated for possible multicollinearity among independent variables using pairwise correlation. Variables with correlation coefficient (r) ≥ 0.5 either direction were not included in the final multivariable model.

We estimated two separate binary logistic regression models for married women and for all women of childbearing age. Children ever born and marital status were not included in the final models due to their strong linear relationship with women’s age. The variables included in the final multivariable models had missing observation on age at first marriage, age at first sex and migration status. The missing values were only observed on all women of childbearing dataset. We performed multiple imputation using chained equations (MICE) to impute the missing observations [31]. Consequently, estimation of the final logistic regression model was performed using multiple imputation with Monte Carlo error estimates for odds ratios to assess association of migration status and modern contraceptive use among women of childbearing age, controlling for the independent effects of other confounders. The final model had participants age, age at first sex, age at first marriage, residence, region, levels of education, and wealth index as confounders.

Source link