Applying Inverse Probability Weighting to Measure Contraceptive Prevalence Using Data from a Community-Based Reproductive Health Intervention in Pakistan

Ece Amber Özçelik, Harvard T.H. Chan School of Public Health Julia Rohr, Harvard T.H. Chan School of Public Health Kristy Hackett, Harvard T.H. Chan School of Public Health Iqbal Shah, Harvard T.H. Chan School of Public Health David Canning, Harvard T.H. Chan School of Public Health

First published online:

| DOI: https://doi.org/10.1363/46e8520
Abstract / Summary

Many community-based reproductive health programs use their program data to monitor progress toward goals. However, using such data to assess programmatic impact on outcomes such as contraceptive use poses methodological challenges. Inverse probability weighting (IPW) may help overcome these issues.


Data on 33,162 women collected in 2013–2015 as part of a large-scale community-based reproductive health initiative were used to produce population-level estimates of the contraceptive prevalence rate (CPR) and modern contraceptive prevalence rate (mCPR) among married women aged 15–49 in Pakistan's Korangi District. To account for the nonrandom inclusion of women in the sample, estimates of contraceptive prevalence during the study's four seven-month intervention periods were made using IPW; these estimates were compared with estimates made using complete case analysis (CCA) and the last observation carried forward (LOCF) method—two approaches for which modeling assumptions are less flexible.


In accordance with intervention protocols, the likelihood that women were visited by intervention personnel and thus included in the sample differed according to their past and current contraceptive use. Estimates made using IPW suggest that the CPR increased from 51% to 64%, and the mCPR increased from 34% to 53%, during the study. For both outcomes, IPW estimates were higher than CCA estimates, were generally similar to LOCF estimates and yielded the widest confidence intervals.


IPW offers a powerful methodology for overcoming estimation challenges when using program data that are not representative of the population in settings where cost impedes collection of outcome data for an appropriate control group.

Author's Affiliations

Ece Amber Özçelik is a doctoral candidate, Julia Rohr and Kristy Hackett are research associates, Iqbal Shah is principal research scientist and David Canning is Richard Saltonstall Professor of Population Sciences, Deputy Director of the Program on the Global Demography of Aging and professor of economics and international health—all at the Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, MA, USA.


The views expressed in this publication do not necessarily reflect those of the Guttmacher Institute.