Table 4

Performance of admission prediction models in the adult population

Study	Model name	Admission, N (%)	Derivation AUC (95% CI)	Calibration method	Calibration derivation	Validation method	Validation AUC (95% CI)	Calibration validation
Alam et al 28	NEWS	130 (47.4)				External	t0: 0.664 (0.599 to 0.728) t1: 0.687 (0.620 to 0.754) t2: 0.697 (0.609 to 0.786)
Cameron et al 19	GAPS	NS	0.8778 (0.8764 to 0.8793)	HL GOF test		Split sample	0.8774 (0.8752 to 0.8796)	p=0.524
Cameron et al 18	GAPS	745 (40.7)		Wilcoxon Signed Rank test		External	0.876 (0.860 to 0.892)	1.20%
Cameron et al 18	VAS	745 (40.7)		Wilcoxon Signed Rank test		External	0.875 (0.859 to 0.891)	9.20%
Kraaijvanger et al 24	Own model	400 (31.7)	NS	Calibration plot		External	0.88 (0.85 to 0.90), 0.87 (0.85 to 0.89), 0.76 (0.72 to 0.80)	α: 0.023, β: 0.974 α: 0.05, β: 0.98
Lucke et al 25	Own model adults	4044 (23.6)	0.85 (0.84 to 0.86)	Calibration plot, HL GOF test		External	0.86 (0.85 to 0.87)	p>0.05
Noel et al 22	TNP	2313 (23.5)	0.815 (0.805 to 0.826)
Noel et al 22	Own model	2313 (23.5)	0.815 (0.805 to 0.825)
Noel et al 22	TNP+own model	2313 (23.5)	0.857 (0.848 to 0.865)
Zlotnik et al 23	Own model LR	34 694 (13.6)	0.8611 (0.8568 to 0.8615)	Calibration plot, HL GOF test	χ²= 85.18	Split sample	0.8568 (0.8508 to 0.8583)	χ²= 65.32
Zlotnik et al 23	Own model ANN	34 694 (13.6)	0.8631 (0.8605 to 0.8656)	Calibration plot, HL GOF test	χ²= 16.01	Split sample	0.8575 (0.8540 to 0.8610)	χ²= 17.28

Empty cells mean that specific characteristics were not tested.
α, calibration intercept; β, calibration slope; ANN, artificial neural network; AUC, area under the curve; GAPS, Glasgow Admission Prediction Score; HL GOF, Hosmer-Lemeshow goodness of fit; LR, logistic regression; N, number; NEWS, National Early Warning Score; NS, not specified; t, timepoint; TNP, triage nurse prediction; VAS, Visual Analogue Scale.;