Table 3. Performance of NAFLD screening scores: development and external validation datasets.

Method, by Dataset	High risk(%)	Sensitivity	Specificity	PPV	NPV	Positive	Negative	Youden	AUC*	RequiresClinician’sInput?
						LR	LR	Index
Development dataset (N = 15676)
Comprehensive score (male) ≥40	43	69	84	82	72	4.46	0.36	54	0.86	Yes
Comprehensive score (female) ≥40	29	72	89	73	88	6.50	0.31	61	0.90	Yes
Simple score (male) ≥8	58	80	67	72	76	2.41	0.29	47	0.82	No
Simple score (female) ≥8	36	77	81	63	89	4.09	0.28	58	0.87	No
Park’s index*	41	72	81	73	80	3.84	0.35	53	0.83	Yes
Hepatic steatosis index*	20	41	95	85	70	7.91	0.62	36	0.85	Yes
Fatty liver index*	12	26	98	89	65	11.09	0.76	23	0.86	Yes
NAFLD liver fat score* , ‡	19	37	94	81	68	6.26	0.67	31	0.77	Yes
External validation dataset (N = 66868)
Comprehensive score (male)† ≥40	37	67	85	76	77	4.32	0.39	52	0.85	Yes
Comprehensive score (female)† ≥40	20	71	88	49	95	6.13	0.33	59	0.89	Yes
Simple score (male)† ≥8	49	75	71	66	79	2.53	0.36	45	0.80	No
Simple score (female)† ≥8	22	68	85	42	94	4.57	0.38	54	0.85	No
Park’s index*	38	76	78	59	89	3.44	0.30	54	0.84	Yes
Hepatic steatosis index*	21	52	92	74	82	6.56	0.52	44	0.86	Yes
Fatty liver index*	11	30	97	82	76	10.77	0.72	27	0.87	Yes
NAFLD liver fat score* , ‡	16	40	94	75	79	6.96	0.63	35	0.82	Yes

*The definition of the above risk models for NAFLD.

Park’s index for NAFLD: index ≥3, in which Park’s index = (ALT/AST ratio >1.5)×1+(γ-glutamyl-transferase >50)×1+(triglycerides >150)×1+(23≤ BMI <25)×2+(25≤ BMI)×3.

Hepatic steatosis index (HSI): HSI >36 for high-risk and <30 for no NAFLD, in which HSI = 8×ALT/AST ratio+BMI (+2, if diabetes;+2, if female).

Fatty liver index (FLI): FLI ≥60 for high-risk and <30 for no NAFLD, in which FLI = 1/(1+exp(−x))×100, where x = 0.953×log_e (triglycerides)+0.139×BMI+0.718×log_e (γ-glutamyl-transferase)+0.053×WC −15.745.

NAFLD liver fat score: score ≥−0.640, in which NAFLD liver fat score = −2.89+1.18×metabolic syndrome+0.45×diabetes+0.15×fasting insulin+0.04×AST+0.94×AST/ALT ratio.

^†After imputing the missing data for alcohol consumption and menopausal status.

^‡After imputing the missing data of fasting insulin levels.