Table 3 Test results for predictive models of neonatal risk of death with the addition of new predictors.

From: Development and evaluation of machine learning training strategies for neonatal mortality prediction using multicountry data

Country	Model	Algorithm	Support	Positive Outcome	CI AUC-ROC	Comparison to neonatal baseline model
General¹	General	LGBM Tuned	134,007	2985	0.821 [0.812, 0.830]	Superior
General¹	Biggest train size	Adaboost Tuned	134,007	2985	0.788 [0.779, 0.798]	Superior
DRC	General	LGBM Tuned	18,568	445	0.794 [0.767, 0.821]	Inferior
DRC	Country-specific	XGB Tuned	18,568	445	0.788 [0.759, 0.816]	Inferior
DRC	Biggest train size	Adaboost Tuned	18,568	445	0.782 [0.753, 0.810]	Superior
Guatemala	General	LGBM Tuned	25,167	539	0.802 [0.782, 0.825]	Superior
Guatemala	Country-specific	XGB Tuned	25,167	539	0.790 [0.768, 0.813]	Inferior
Guatemala	Biggest train size	Adaboost Tuned	25,167	539	0.736 [0.708, 0.763]	Inferior
Zambia	General	LGBM Tuned	18,677	215	0.788 [0.747, 0.828]	Superior
Zambia	Country-specific	XGB Tuned	18,677	215	0.809 [0.771, 0.846]	Superior
Zambia	Biggest train size	Adaboost Tuned	18,677	215	0.760 [0.716, 0.802]	Inferior
India-Belagavi	General	LGBM Tuned	16,070	313	0.787 [0.756, 0.816]	Superior
India-Belagavi	Country-specific	Adaboost Tuned	16,070	313	0.785 [0.754, 0.816]	Superior
India-Belagavi	Biggest train size	Adaboost Tuned	16,070	313	0.785 [0.754, 0.816]	Superior
India-Nagpur	General	LGBM Tuned	18,278	323	0.823 [0.791, 0.850]	Superior
India-Nagpur	Country-specific	Adaboost Tuned	18,278	323	0.811 [0.78, 0.84]	Superior
India-Nagpur	Biggest train size	Adaboost Tuned	18,278	323	0.819 [0.789, 0.848]	Superior
Pakistan	General	LGBM Tuned	16,521	838	0.777 [0.758, 0.795]	Superior
Pakistan	Country-specific	LGBM Tuned	16,521	838	0.777 [0.756, 0.794]	Superior
Pakistan	Biggest train size	Adaboost Tuned	16,521	838	0.750 [0.727, 0.77]	Inferior
Kenya	General	LGBM Tuned	19,637	280	0.814 [0.783, 0.844]	Superior
Kenya	Country-specific	LGBM Tuned	19,637	280	0.809 [0.777, 0.838]	Superior
Kenya	Biggest train size	Adaboost Tuned	19,637	280	0.777 [0.745, 0.812]	Inferior
Bangladesh²	General	LGBM Tuned	1,089	32	0.860 [0.781, 0.926]	Superior
Bangladesh²	Biggest train size	Adaboost Tuned	1,089	32	0.822 [0.739, 0.898]	Superior

¹The general model does not have a country-specific approach.
²Due to the lack of data on training set, Bangladesh does not present a country-specific approach, since it had no data collected in the training period.

Back to article page

Table 3 Test results for predictive models of neonatal risk of death with the addition of new predictors.

Search

Quick links