Abstract:
Target population estimation for immunization coverage calculations from census data is often inaccurate. This study aimed to evaluate the accuracy of the traditional census extrapolation method in comparison with three alternative approaches: the Cohort-Component Population Projections Method (CCPPM), using the Expanded Program on Immunisation (EPI) numerators – BCG and DTP1 doses as denominators, and estimates derived from first antenatal care clinic (ANC1) visits. We obtained target population estimates in Kenya from 1999 - 2023 using all 4 methods with data for ANC1 available only for 2020–2023. We assessed the accuracy of the estimates for 2003–2018 by computing the Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE) and the Pearson Correlation Coefficient (r), excluding outliers. A sub-analysis for the period 2020–2023 included ANC1 data. The CCPPM method had the largest population estimates while the census-based method had pronounced discontinuities at the census years. The CCPPM method compared to the DTP1 doses was associated with the greatest error magnitude (MAE = 212917.19 and MAPE = 18.18) while the DTP1 doses and census-based methods showed the smallest error (MAE = 44317.16 and MAPE = 3.77). Sub-analysis of target populations for the period 2020–2023 showed similar upward trends except for the census-based method which exhibited a significantly divergent trajectory. Comparison between the ANC1 and DTP1 doses showed the strongest linear correlation (r = 1.00). Although sub-national analysis was not done and there was the significant challenge of missing data, the results nevertheless reveal significant inaccuracies in the current target population estimation methods which may have serious implications on immunisation coverage assessments. Immunisation programs should utilise diverse sources of data and triangulate results as a more pragmatic approach for approximating the target populations for vaccination in the absence of well-established civil registration systems. Additionally, more research is warranted to address this gap.