Understanding the social determinants of child mortality in three latin american countries: an approach with machine learning
 
More details
Hide details
1
Center for Health Economics, University of York, UK United Kingdom
 
2
University of York, U United Kingdom
 
3
Health Research Consortium - CISIDAT
 
4
Institute of Collective Health, at the Federal University of Bahia (ISC-UFBA) Brazil
 
5
Universidad de los Andes, School of Government
 
6
Universidad de los Andes, School of Government, Colombia
 
7
Centro de Investigación para la Salud en América Latina (CISeAL), Pontificia Universidad Católica del Ecuador Ecuador
 
8
Instituto Nacional de Salud Publica Mexico
 
9
Swiss Tropical and Public Health Institute, Department of Public Health and Epidemiology Swaziland
 
10
Institute of Global Health (ISGlobal), Barcelona, Spain Spain
 
 
Publication date: 2023-04-26
 
 
Popul. Med. 2023;5(Supplement):A813
 
ABSTRACT
Objective:
Evaluate the relationship between the social determinants of health (sociodemographic and health system resources) and the under-five mortality rate (TMM5).

Methods:
Municipal-level data was obtained from 2000 to 2019 from 9,142 municipalities in three Latin American countries: Brazil, Ecuador, and Mexico. To explore the relationship between social determinants and U5MR, we trained a Random Forest (RF) algorithm, and to assess model robustness, we also trained a Gradient Boosting Machine and a Model Generalized Additive. Finally, we present the mean square error (MSE), root mean square error (RMSE), and mean absolute deviation (MAD) and r-squared to compare the performance of the trained algorithms.

Results:
The most important variables to predict the MMR5 according to the RF were illiteracy, poverty, and the Gini index according to the random forest algorithm. We found positive relationships between illiteracy and poverty with U5MR. Nonlinear relationships were also observed between the Gini index and the U5MR. The RF results were MSE = 60626.96, RMSE = 246.22, MAD = 125.61, r-squared = .14, from the Gradient Boosting Machine were MSE = 61956.91, RMSE = 248.91, MAD = 129.49, r-squared = .12 and, from the Model Generalized Additive were MSE = 65813.03, RMSE = 256.54, MAD = 135.87, r- squared = .07.

Conclusions:
According to the results obtained, long-term public policies to reduce the MMR5 should focus on reducing illiteracy, poverty, and inequality. Information on modifiable social factors can be useful in planning intervention programs to promote child survival in Latin America and other low-income countries.

ISSN:2654-1459
Journals System - logo
Scroll to top