Prediction Models for Return of Spontaneous Circulation in Patients with Cardiac Arrest: A Systematic Review and Critical Appraisal

Table 1

Basic characteristics of the ROSC prediction model included.

Authors and years

Country

Object

Sample (missing data)

Data sources

ROSC rate (%)

Candidate variables’ number

Variable selection

Method

Validation

Model performance

Predictors in the final model

AUC/C-index

Calibration

Gräsner et al. [14] 2011

Germany

OHCA

7689 (1112)

GRR

43.8

9

Multivariate analysis was performed after univariate analysis

LR

Internal

0.710^1

None

Sex, age, etiology, first witness identity, location, initial heart rhythm, bystander CPR, EMS arrival time

0.731^2

Harrison et al. [15] 2014

UK

IHCA

24136 (<0.3%)

NCAA

45.0

7

Backward stepwise selection method

LR

Internal, external

0.733^1

HL test, calibration plot, Cox calibrated regression, Brier scores

Age, sex, length of hospital stay, medical reasons, location, initial heart rhythm

0.720^2

0.725^3

Kim et al. [16] 2017

Korea

OHCA

727 (97)

ED

55.7

11

Multivariate analysis was performed after univariate analysis

Cox

Internal

None

Grønnesby and Borgan goodness of fit tests

Age, location, etiology, initial heart rhythm, time of BLS, prehospital low blood-flow time, PO2, ACLS duration, BE

Amnuaypattanapon et al. [17] 2020

Thailand

OHCA

347 (37)

ED

36.3

14

Backward stepwise selection method

LR

Internal

0.775^1

None

Witnessed at arrest, time of BLS, CPR duration

Baldi et al. [18] 2020

Italy and Switzerland

OHCA

2709 (1224)

Pavia CARE and TiReCa

P: 16.9; T: 33.5

7

Multivariate analysis was performed after univariate analysis

LR

Internal, external

0.83^1

None

Age, location, witnessed at arrest, bystander CPR, EMS arrival time, initial heart rhythm

0.82^2

0.77^3

Ji et al. [19] 2021

UK

OHCA

32789 (1817)

NHS

25.0

8

Multivariate analysis was performed after univariate analysis

MEM

Internal

0.70^1

HL test, Cox calibration, Brier scores

Sex, bystander CPR, etiology, initial heart rhythm

0.67^2

Kuo et al. [20] 2022

China

TOHCA

580 (10)

A trauma center in Taiwan

21.4

13

Multivariate analysis was performed after univariate analysis

LR

Internal, external

0.662^1

None

Obvious chest injury, body temperature below 33°C, and possible etiology of OHCA

Liu et al. [21] 2022

Singapore

OHCA

170678 (9146)

PAROS

8.26

13

The importance of the candidate predictors was ranked and the top five variables were included

RF

Internal

0.715^1

None

Age, witnessed at arrest, EMS arrival time, initial heart rhythm, prehospital administration

0.806^2

ROSC: recovery of spontaneous circulation; OHCA: out-of-hospital cardiac arrest; IHCA: hospital cardiac arrest; TOHCA: traumatic out-of-hospital cardiac arrest; ED: emergency department; GRR: German recovery registration system; NCAA: The UK’s National Cardiac Arrest Audit; Pavia CARE: Pavia department cardiac arrest registration system; TiReCa: Ticino Regional Cardiac Arrest Registry; NHS: National Health Service; PAROS: the Pan-Asian recovery outcome registry. ¹: development result, ²: internal validation result, ³: external validation result; AUC: area under the curve; C-index: concordance index; : consistency index result; HL test: Hosmer–Lemeshow test; LR: logistic regression; Cox: Cox regression model; MEM: mixed-effect model; RFs: random forests; CPR: cardiopulmonary resuscitation; EMS: emergency medical services; BLS: basic life support; ACLS: advanced cardiac life support; PO₂: partial pressure of blood oxygen; BE: base excess.