Table 3. Study Outcomes for Bedside Paediatric Early Warning System (BedsidePEWS) vs Usual Care.

	BedsidePEWS				Usual Care				Adjusted Between-Group Rate Difference (95% CI)a	Adjusted Ratio (95% CI)	P Value
	Baseline Period		Intervention Period		Baseline Period		Intervention Period
	No. of Events	Rate	No. of Events	Rate	No. of Events	Rate	No. of Events	Rate
Patient dischargesb	26 664		50 173		46 718		94 366
Patient-daysb	129 700		251 859		162 497		307 584
Primary Outcome
All-cause hospital mortalityc	52	1.95	97	1.93	61	1.31	147	1.56	0.01 (−0.80 to 0.81)	OR, 1.01 (0.61 to 1.69)d	.96
All-cause hospital mortality without a DNR orderc	26	0.98	42	0.84	16	0.34	47	0.50	0.36 (−0.53 to 1.25)	OR, 2.05 (0.64 to 6.61)d	.23
Secondary Outcome
Significant clinical deterioration evente	80	0.62	127	0.50	144	0.89	259	0.84	−0.34 (−0.73 to 0.05)	RR, 0.77 (0.61 to 0.97)f	.03
Post hoc Outcomesg
ICU mortalityh	33	17.75	56	16.92	34	13.08	91	17.90	−3.01 (−12.26 to 6.25)	OR, 0.89 (0.51 to 1.57)d	.69
ICU mortalityc	33	1.24	56	1.12	34	0.73	91	0.96	−0.11 (−0.73 to 0.51)	OR, 0.95 (0.48 to 1.86)d	.88
Cardiac arreste	15	0.12	27	0.11	18	0.11	32	0.10	0 (−0.06 to 0.07)	RR, 1.02 (0.65 to 1.62)f	.92
Potentially preventable cardiac arreste,i	11	0.08	21	0.08	12	0.07	29	0.09	−0.02 (−0.07 to 0.02)	RR, 0.87 (0.49 to 1.54)f	.62
Immediate call for resuscitation teame	64	0.49	126	0.50	97	0.60	179	0.58	0.02 (−0.07 to 0.10)	RR, 0.98 (0.82 to 1.17)f	.83
Immediate call for physiciane,j	1007	7.76	1727	6.86	844	5.19	1157	3.76	3.10 (−1.92 to 8.11)	RR, 1.17 (0.73 to 1.88)f	.52
Urgent (<15 min) ICU consultatione	478	3.69	1015	4.03	928	5.71	1694	5.51	0.16 (−0.57 to 0.89)	RR, 1.05 (0.85 to 1.30)f	.64
Urgent ICU admissione	469	3.62	828	3.29	652	4.01	1178	3.83	−0.18 (−0.67 to 0.30)	RR, 0.95 (0.82 to 1.09)f	.45
ICU readmission <48 hh	64	34.43	94	28.40	73	28.09	108	21.25	4.95 (−1.62 to 11.52)	OR, 1.11 (0.77 to 1.61)d	.58
Hospital readmission <48 hc	101	3.79	170	3.39	201	4.30	387	4.10	−0.71 (−4.92 to 3.49)	OR, 0.93 (0.61 to 1.41)d	.74

Abbreviations: DNR, do-not-resuscitate; ICU, intensive care unit; OR, odds ratio; RR, rate ratio.

^aCalculated using binomial and Poisson generalized estimating equation models with an identity link function and adjustment for baseline values. For significant clinical deterioration, cardiac arrest, immediate call for physician, and hospital readmission, the model was not adjusted for baseline values.

^bData were used for the rate calculations.

^cRates expressed per 1000 patient discharges from the hospital.

^dLogistic regression was used. Analyses included adjustment for baseline event rates and used the generalized estimating equation approach to group data by center to account for clustering.

^eRates expressed per 1000 patient-days.

^fPoisson regression was used. Analyses included adjustment for baseline event rates and used the generalized estimating equation approach to group data by center to account for clustering.

^gThe analyses of these outcomes should be regarded as exploratory. There was adjustment for multiple comparisons using the Holm method and yielded P values of >.99 for these 10 outcomes.

^hRates expressed per 1000 discharges from the ICU.

ⁱAssessed by 2 blinded expert reviewers using a 6-point scale. A rating of 4 indicates more than likely preventable. Weighted κ, 0.35 (95% CI, 0.15-0.51) for agreement of initial reviewer ratings. Discussion between the 2 reviewers increased agreement such that arbitration by the third reviewer was used for only 5 events. There were 21 (77.8%) cardiac arrest events rated as potentially preventable at the BedsidePEWS hospitals vs 29 (90.6%) events at the control hospitals.

^jDetermined from multiple potentially overlapping sources (eg, switchboard paging logs, ward clerk documentation, review of patients with events, and other hospital reports). The customized approach taken at each hospital was developed as part of site coordinator training provided by the coordinating center and was applied consistently throughout the study at each site.