A meta-analysis of the diagnostic test accuracy of artificial intelligence predicting emergency department dispositions

Kuo, Kuang-Ming; Chang, Chao Sheng

doi:10.1186/s12911-025-03010-x

BMC Medical Informatics and Decision Making

Table 7 Summary estimates for sensitivity and specificity of critical care studies

From: A meta-analysis of the diagnostic test accuracy of artificial intelligence predicting emergency department dispositions

Characteristic	Covariate	Metric	n	Estimate	95%	C.I.	p value
Data	Public dataset	Sens	5	0.76	0.64	0.85	0.316
	Public dataset	Specs	5	0.73	0.62	0.82	0.103
	Private dataset	Sens	40	0.87	0.79	0.92	[Reference]
	Private dataset	Specs	40	0.90	0.85	0.94	[Reference]
	Combined	Sens	12	0.86	0.77	0.92	0.865
	Combined	Specs	12	0.87	0.80	0.92	0.626
	Structured	Sens	32	0.86	0.77	0.92	[Reference]
	Structured	Specs	32	0.90	0.83	0.95	[Reference]
	Image	Sens	5	0.87	0.76	0.94	0.530
	Image	Specs	5	0.86	0.80	0.90	0.861
	Free text	Sens	8	0.83	0.67	0.92	[Reference]
	Free text	Specs	8	0.87	0.74	0.94	[Reference]
Sample	Adult	Sens	38	0.86	0.78	0.91	0.664
	Adult	Specs	38	0.90	0.83	0.94	0.616
	Youth	Sens	2	0.85	0.73	0.92	0.762
	Youth	Specs	2	0.72	0.51	0.86	0.204
	Mixed	Sens	3	0.90	0.74	0.96	[Reference]
	Mixed	Specs	3	0.84	0.75	0.90	[Reference]
Artificial intelligence technique	Machine learning	Sens	31	0.88	0.80	0.93	0.205
	Machine learning	Specs	31	0.91	0.84	0.95	0.171
	Deep learning	Sens	14	0.78	0.69	0.85	[Reference]
	Deep learning	Specs	14	0.83	0.78	0.87	[Reference]
	Random forest	Sens	9	0.91	0.77	0.97	0.465
	Random forest	Specs	9	0.96	0.83	0.99	0.297
	eXtreme gradient boosting	Sens	11	0.95	0.85	0.99	0.302
	eXtreme gradient boosting	Specs	11	0.84	0.71	0.92	0.965
	LightGBM	Sens	3	0.69	0.29	0.92	0.439
	LightGBM	Specs	3	0.98	0.59	1.00	0.259
	Logistic regression	Sens	2	0.78	0.71	0.84	0.673
	Logistic regression	Specs	2	0.84	0.80	0.87	0.922
	Deep neural network	Sens	10	0.75	0.64	0.83	0.442
	Deep neural network	Specs	10	0.83	0.77	0.88	0.932
	Convolutional neural network	Sens	3	0.84	0.63	0.94	[Reference]
	Convolutional neural network	Specs	3	0.84	0.74	0.90	[Reference]
	Ensemble	Sens	25	0.91	0.82	0.95	0.032
	Ensemble	Specs	25	0.91	0.82	0.96	0.339
	No ensemble	Sens	20	0.76	0.69	0.82	[Reference]
	No ensemble	Specs	20	0.86	0.81	0.90	[Reference]
	Cross validation	Sens	27	0.86	0.76	0.92	0.952
	Cross validation	Specs	27	0.87	0.79	0.92	0.288
	No cross validation	Sens	18	0.87	0.75	0.93	[Reference]
	No cross validation	Specs	18	0.92	0.83	0.96	[Reference]

Note: C.I. = Confidence Interval, Sens = Sensitivity, and Specs = Specificity

Back to article page

ISSN: 1472-6947

Contact us

General enquiries: journalsubmissions@springernature.com