Table 3.

AUTOMATED MACHINE LEARNING (AUTOML) MODEL-PREDICTED VERSUS ACTUAL EARLY CHILDHOOD CARIES (ECC) PROBABILITIES ACCORDING TO PARENTS’ REPORTS AND CHILDREN’S AGES

Age group		Parent-reported child oral health status
	ECC* probability	Poor	Fair	Good	Very good	Excellent
3-year-old	Predicted	0.88	0.74	0.54	0.40	0.31
	Actual	1.00	0.83	0.55	0.36	0.24
4-year-old	Predicted	0.90	0.79	0.63	0.47	0.35
	Actual	0.96	0.86	0.66	0.46	0.29
5-year-old	Predicted	0.91	0.82	0.69	0.51	0.38
	Actual	0.90	0.87	0.73	0.50	0.36

^*ECC was defined as one or more decayed, restored, or extracted primary tooth surfaces due to caries (i.e., dmfs greater than zero) where caries lesions were considered at the International Caries Detection and Assessment System (ICDAS) threshold of greater than or equal to three.