Conflict of Interest Declaration
The ILCOR Continuous Evidence Evaluation process is guided by a rigorous ILCOR Conflict of Interest policy. The following Task Force members and other authors were recused from the discussion as they declared a conflict of interest: none. The following Task Force members and other authors declared an intellectual conflict of interest and this was acknowledged and managed by the Task Force Chairs and Conflict of Interest committees: Content expert Dr. Theresa Djärv has published studies on pre-arrest prediction scores and was excluded from bias assessment.
CoSTR Citation
Lauridsen KG, Djärv T, Couper K, Tjissen J, Breckwoldt J, Greif R on behalf of the International Liaison Committee on Resuscitation Education, Implementation, and Teams Task Force. Pre-arrest Prediction of Survival following In-hospital Cardiac Arrest. Consensus on Science with Treatment Recommendations [Internet] Brussels, Belgium: International Liaison Committee on Resuscitation (ILCOR) Advanced Life Support Task Force, 2021 December 7. Available from: http://ilcor.org
Methodological Preamble and Link to Published Systematic Review
The continuous evidence evaluation process for the production of Consensus on Science with Treatment Recommendations (CoSTR) started with a systematic review conducted by a working group from the ILCOR Education, Implementation and Team (EIT) Task Force under the lead of KG Lauridsen, with involvement of clinical content experts (T. Djärv, K. Couper, J. Tjissen, J. Breckwoldt, R. Greif). The task force agreed on the search strategy and PICOST. The task force working group conducted the screening, extraction, and assessment of the available data.
We searched for studies using pre-arrest clinical prediction rules aiming to predict the chance of surviving (or not surviving) an in-hospital cardiac arrest (IHCA), with or without favourable neurological outcome. We defined IHCA as any cardiac arrest with clinical indication for cardiopulmonary resuscitation (CPR) occurring inside the hospital regardless of the underlying cause of the arrest. Patients with out-of-hospital cardiac arrest being transported to the hospital with ongoing CPR were excluded. Pre-arrest clinical prediction rules was defined as a set of clinical variables used to predict the chance of surviving a cardiac arrest (± favourable neurological outcome) in patients who experienced an IHCA.
Measures of prediction were negative predictive value (NPV), sensitivity, specificity, positive predictive value (PPV). Whenever possible, we extracted all prediction endpoints from the included studies. We streamlined reporting of outcomes as true positives being patients surviving who were predicted to survive and true negatives as patients dying who were predicted to die.
PICOST
The PICOST (Population, Intervention, Comparator, Outcome, Study Designs and Timeframe)
Population: Hospitalized adults and children experiencing an in-hospital cardiac arrest.
Intervention: Does any pre-arrest clinical prediction rule.
Comparators: Compared to no clinical prediction rule.
Outcomes: Predict return of spontaneous circulation, survival to hospital discharge/ 30-days or survival with favorable neurological outcome.
Study Designs: Randomized controlled trials (RCTs) and non-randomized studies (non-randomized controlled trials, interrupted time series, controlled before-and-after studies, cohort studies, case series where n ≥ 5) were included.
Excluded studies: unpublished results (e.g. trial protocols), commentaries, editorials, reviews, conference abstracts.
Timeframe: All years and all languages were included as long as there was an English abstract up to January 13 2022.
PROSPERO Registration CRD42021268005
Risk of bias was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. We assessed the risk of bias per study rather than per score or per outcome, since there were no meaningful differences in bias across outcomes. In cases where differences in risk of bias existed between outcomes, this was noted. We did not conduct any meta-analysis as the included studies were judged to have high risk of bias and evidence was considered very low certainty.
Consensus on Science
We identified 23 studies investigating 13 different pre-arrest prediction rules of survival following in-hospital cardiac arrest. For the outcome of predicting survival to hospital discharge, we identified very low certainty evidence from seven historical cohort studies {Ebell 1997 171, O’Keeffe 1994 21, Bowker 1999 89, Ohlsson 2014 294, Limpawattana 2018 1231, George 1989 28, Cohn 1993 347} investigating the pre-arrest morbidity (PAM) score (downgraded for risk of bias, indirectness, imprecision, and inconsistency) and four of these studies investigated the prognosis after resuscitation (PAR) score (downgraded for risk of bias, indirectness, imprecision, and inconsistency), Table 2. The studies identified various cut-off values for the score to predict no chance of survival to hospital discharge. Due to clinical heterogeneity in study cohorts, no meta-analysis was conducted. The outcomes of the PAM score and PAR score are presented in Table 1 and Table 2 respectively. Limpawattana et al., {Limpawattana 2018 1231} did not report data to calculate sensitivity, specificity, NPV, and PPV with 95% confidence intervals (CI). However, they reported an area under the curve (AUC) of 0.65 (95% CI: 0.56-0.74) for the PAM score and self-calculated outcome measures without confidence intervals for the prediction of death (as opposed to survival) with a PPV of 92.2, a specificity of 87.8, a sensitivity of 39.2, and a NPV of 28.1. For the PAR score, they reported an AUC of PAR 0.6 (95% CI: 0.52-0.70).
Table 1: Predictive values of historical cohort studies using the pre-arrest morbidity (PAM) score to predict survival to hospital discharge (presented with 95% CI). NPV negative predictive value; PPV positive predictive value.
Table 2: Predictive values of historical cohort studies using the prognosis after resuscitation (PAR) score to predict survival to hospital discharge (presented with 95% CI) For the outcome of predicting survival to hospital discharge, we identified very low certainty evidence from two historical cohort studies {Bowker 1999 89, Limpawattana 2018 1231} investigating the modified pre-arrest morbidity (MPI) score (downgraded for risk of bias, indirectness, imprecision, and inconcistency). Bowker et al. showed a sensitivity of 100 (95% CI: 87.7-100), a specificity 22.5 (95% CI: 17.3-28.3), a NPV of 100 (95% CI: 93.3-100), and a PPV of 13.3 (95% CI: 9.0-18.6) for a MPI score >6. Limpawattana et al. did not report data to calculate sensitivity, specificity, NPV, and PPV with 95% CIs. However, they reported self-calculated outcome measures without confidence intervals for the prediction of death (as opposed to survival) with a PPV of 92.2, a specificity of 87.8, a sensitivity of 39.2, and a NPV of 28.1 for a MPI score >5. For the outcome of predicting survival to hospital discharge, we identified very low certainty evidence from one historical cohort study investigating the modified early warning score (MEWS) {Stark 2015 916}, two historical cohort studies investigating the National Early Warning Score (NEWS) {Haegdorens 2020 4594, Roberts 2017 1601}, one historical cohort study investigating the Clinical Frailty Scale {Ibitoye 2021 147}, and one historical cohort study investigating the APACHE III score {Ebell 1997 171}. The level of evidence for all scores was downgraded for downgraded for risk of bias, indirectness, imprecision, and inconsistency. Ibitoye et al. showed a sensitivity of 100 (95% CI: 75.3-100), a specificity of 51.9 (95% CI: 40.3-63.5), a NPV of 100 (95% CI: 91.2-100), and a PPV of 26.0 (95% CI: 14.6-40.3) for a Clinical Frailty Scale >4. Haegdorens et al. showed a sensitivity of 57.9 (95% CI: 33.5-79.7), a specificity of 71.4 (95% CI: 41.9-91.6), a NPV of 55.6 (95% CI: 30.8-78.5), and a PPV of 73.3 (95% CI: 44.9-92.2) for a NEWS ≥5 and Roberts et al. showed a sensitivity of 89.3 (95%CI: 80.1-95.3), a specificity of 31.7 (95% CI: 25.6-38.2), a NPV of 89.7 (95% CI: 80.8-95.5), and a PPV of 30.7 (95%CI: 24.7-37.3) for a NEWS ≥7. Stark et al. did not report data to calculate sensitivity, specificity, NPV, and PPV with 95% CIs. However, they reported self-calculated outcome measures without confidence intervals for the prediction of death (as opposed to survival) with a PPV of 76, a specificity of 80, a sensitivity of 47, and a NPV of 53 for a Modified Early Warning Score of 7. Ebell et al. did not report data to calculate sensitivity, specificity, NPV, and PPV with 95% CIs. However, they reported an area under the curve of 0.59 for the APACHE III score to predict survival to hospital discharge. For the outcome of predicting survival to hospital discharge with favorable neurological outcome, we identified low certainty evidence from seven historical cohort studies {Ebell 2013 1872, Piscator 2018 63, Rubins 2019 2530, Cho 2020 36, Thai 2019 140, Ohlsson 2016 294, Hong 2021 10631} investigating the Good Outcome Following Attempted Resuscitation (GO-FAR) score to predict survival with a cerebral performance category (CPC) of 1 (downgraded for risk of bias, indirectness, and imprecision). The outcomes are presented in Table 3. Hong et al. did not report data on survival with CPC of 1 but the authors provided data showing a sensitivity of 94.1 (95% CI: 87.6-97.8), a specificity of 11.7 (95% CI: 8.5-15.6), a NPV of 87.0 (95% CI: 73.7-95.1), and a PPV of 24.1 (95% CI: 20.0-28.6) for the GO-FAR score to predict survival to hospital discharge.
Table 3: Predictive values of historical cohort studies using the good outcome following attempted resuscitation (GO-FAR) score to predict survival to hospital discharge with a cerebral performance category (CPC) of 1 (presented with 95% CIs). NPV negative predictive value; PPV positive predictive value. For the outcome of predicting survival to hospital discharge with favorable neurological outcome, we identified low certainty evidence from one historical cohort study {George 2020 162} investigating the Good Outcome Following Attempted Resuscitation 2 (GO-FAR 2) score, one historical cohort study {Piscator 2019 92} investigating the Prediction of Outcome for In-hospital Cardiac Arrest (PIHCA) score, and two classification and regression tree models (CART 1, CART 2) {Ebell 2013 2688 , Guilbault 2017 333}. The CART models {Ebell 2013 2688, Guilbault 2017 333} aimed to predict survival with a CPC=1 whereas the GO-FAR 2 score and the PIHCA score investigated survival with CPC ≤2. The outcomes are summarized in Table 4. All scores were downgraded for risk of bias and imprecision.
Table 4: Predictive values of historical cohort studies using different scores than the GO-FAR score to predict survival to hospital discharge with favorable neurological outcome (presented with 95% CIs). |
Treatment Recommendations
We recommend against using any currently available pre-arrest prediction rule as a sole reason to not resuscitate an adult with in-hospital cardiac arrest (strong recommendation, very low certainty evidence).
We are unable to recommend for or against any available pre-arrest prediction rule to facilitate do-not-attempt cardiopulmonary resuscitation discussions with adult patients or their next of kin as there are no studies investigating the effect of clinical implementation of such score.
We are unable to provide any recommendation for pediatric patients as no studies on children were identified.
Justification and Evidence to Decision Framework Highlights
In making this recommendation, the task force valued a perfect negative predictive value (i.e. no chance of classifying a survivor as a non-survivor). None of the existing pre-arrest prediction rules were able to reliably predict no chance of survival to hospital discharge or survival with favorable functional outcome. The task force also noted that most studies on the PAM, PAR, APACHE III and MPI scores were based on cohorts before 2000, when survival rates were lower. The PAM score and the PAR scores did not perform consistently across cohorts.
Some studies were based on selected patient cohorts or patients from a single center, raising concerns about generalizability. All studies were based on historical cohorts, and concern for bias was very high as use if historical cohorts may create self-fulfilling prophecies. As there were no prospective studies identified on clinical implementation of a pre-arrest prediction model to facilitate do-not-attempt cardiopulmonary resuscitation (DNACPR) discussions, it is unknown whether the clinical implementation of such a score would influence the rate of DNACPR discussions, the rate of DNACPR orders, survival outcomes, or patient perspectives.
- All scores predicting survival with favorable neurological outcome included variables such as hypotension, respiratory insufficiency, or sepsis before the arrest that may change during the hospital admission. Thus, there are concerns regarding applicability of these models.
- The GO-FAR score identifies the chance of survival with perfect neurological outcome (i.e. CPC of 1) although patients and relatives may value survival with a CPC > 1.
- Scores that can predict a very low chance of survival with favorable functional outcome may be used to facilitate DNACPR discussions with patients, although the score may not be able to predict no chance of survival or survival with favorable neurological outcome.
- None of the identified scores were developed to be used for pediatric patients and should not be used for this patient population. Specific scores for pediatric patients are warranted.
- All scores predicting favorable neurological outcome included variables such as e.g. hypotension, respiratory insufficiency, or sepsis before the arrest that may change during the hospital admission. Thus, there are concerns regarding applicability of these models.
Knowledge Gaps
We identified several knowledge gaps in the published literature.
- We identified several knowledge gaps in the published literature.
- There are no clinical decision tools to predict return of spontaneous circulation and several scores did not predict survival to hospital discharge.
- We found no studies assessing long term outcomes beyond hospital discharge or outcomes on quality of life.
- No studies were found on in-hospital pre-arrest prediction of survival for pediatric patients.
- No studies were found on in-hospital pre-arrest prediction of survival in low-resource settings.
- We did not identify any score predicting survival with favorable neurological outcome that did not include physiological deterioration before cardiac arrest.
- There is a lack of prospective clinical validation studies and randomized trials investigating the use of a in hospital pre-arrest clinical prediction rule to be used for do-not-attempt cardiopulmonary resuscitation discussions and/ or making DNACPR orders.
- It is unknown how the use of a clinical decision tool affects resuscitation practices, cost-benefit, or how it affects survival outcomes.1–6
Attachments
References
1. Thai TN, Ebell MH. Prospective validation of the Good Outcome Following Attempted Resuscitation (GO-FAR)score for in-hospital cardiac arrest prognosis.Resuscitation 2019; 140: 2–8.
2. Piscator E, Goransson K, Bruchfeld S, Hammar U, El Gharbi S, Ebell M, Herlitz J, Djarv T. Predicting neurologically intact survival after in-hospital cardiac arrest-external validation of the Good Outcome Following Attempted Resuscitation score.Resuscitation 2018; 128: 63–9.
3. Ebell MH, Afonso AM, Geocadin RG. Prediction of survival to discharge following cardiopulmonary resuscitation using classification and regression trees.Crit Care Med 2013; 41: 2688–97.
4. Ebell MH, Kruse JA, Smith M, Novak J, Drader-Wilcox J. Failure of three decision rules to predict the outcome of in-hospital cardiopulmonary resuscitation.Med Decis Making 1997; 17: 171–7.
5. Ebell MH, Jang W, Shen Y, Geocadin RG. Development and validation of the Good Outcome Following Attempted Resuscitation (GO-FAR) score to predict neurologically intact survival after in-hospital cardiopulmonary resuscitation.JAMA Intern Med 2013; 173: 1872–8.
6. Ohlsson MA, Kennedy LM, Ebell MH, Juhlin T, Melander O. Validation of the good outcome following attempted resuscitation score on in-hospital cardiac arrest in southern Sweden.Int J Cardiol 2016; 221: 294–7.
7. Piscator E, Goransson K, Forsberg S, Bottai M, Ebell M, Herlitz J, Djarv T. Prearrest prediction of favourable neurological survival following in-hospital cardiac arrest: The Prediction of outcome for In-Hospital Cardiac Arrest (PIHCA) score.Resuscitation 2019; 143: 92–9.
8. O’Keeffe S, Ebell MH. Prediction of failure to survive following in-hospital cardiopulmonary resuscitation: comparison of two predictive instruments.Resuscitation 1994; 28: 21–5.
9. Guilbault RWR, Ohlsson MA, Afonso AM, Ebell MH. External Validation of Two Classification and Regression Tree Models to Predict the Outcome of Inpatient Cardiopulmonary Resuscitation.J Intensive Care Med 2017; 32: 333–8.
10. George N, Thai TN, Chan PS, Ebell MH. Predicting the probability of survival with mild or moderate neurological dysfunction after in-hospital cardiopulmonary arrest: The GO-FAR 2 score.Resuscitation 2020; 146: 162–9.
11. Rubins JB, Kinzie SD, Rubins DM. Predicting Outcomes of In-Hospital Cardiac Arrest: Retrospective US Validation of the Good Outcome Following Attempted Resuscitation Score.J Gen Intern Med 2019; 34: 2530–5.
12. Bowker L, Stewart K. Predicting unsuccessful cardiopulmonary resuscitation (CPR): a comparison of three morbidity scores.Resuscitation 1999; 40: 89–95.
13. Ibitoye SE, Rawlinson S, Cavanagh A, Phillips V, Shipway DJH. Frailty status predicts futility of cardiopulmonary resuscitation in older adults.Age Ageing 2020 Jun. doi: 10.1093/ageing/afaa104
14. Cho YJ, Kim YJ, Kim MY, Shin YJ, Lee J, Choi E, Hong S-B, Huh JW, Yang WS, Kim WY. Validation of the Good Outcome Following Attempted Resuscitation (GO-FAR) score in an East Asian population.Resuscitation 2020; 150: 36–40.
15. Haegdorens F, Monsieurs KG, De Meester K, Van Bogaert P. The optimal threshold for prompt clinical review: An external validation study of the national early warning score.J Clin Nurs 2020; 29: 4594–603.
16. Stark AP, Maciel RC, Sheppard W, Sacks G, Hines OJ. An early warning score predicts risk of death after inhospital cardiopulmonary arrest in surgical patients.Am Surg 2015; 81: 916–21.
17. Limpawattana P, Suraditnan C, Aungsakul W, Panitchote A, Patjanasoontorn B, Phunmanee A, Pittayawattanachai N. A comparison of the ability of morbidity scores to predict unsuccessful cardiopulmonary resuscitation in thailand.J Med Assoc Thail 2018; 101: 1231–6.
18. Ohlsson MA, Kennedy LM, Juhlin T, Melander O. Evaluation of pre-arrest morbidity score and prognosis after resuscitation score and other clinical variables associated with in-hospital cardiac arrest in southern Sweden.Resuscitation 2014; 85: 1370–4.
19. George ALJ, Folk BP 3rd, Crecelius PL, Campbell WB. Pre-arrest morbidity and other correlates of survival after in-hospital cardiopulmonary arrest.Am J Med 1989; 87: 28–34.
20. Ebell MH. Artificial neural networks for predicting failure to survive following in-hospital cardiopulmonary resuscitation.J Fam Pract 1993; 36: 297–303.
21. Roberts D, Djärv T. Preceding national early warnings scores among in-hospital cardiac arrests and their impact on survival.Am J Emerg Med 2017; 35: 1601–6.
22. Hong S-I, Kim Y-J, Cho YJ, Huh JW, Hong S-B, Kim WY. Predictive value of pre-arrest albumin level with GO-FAR score in patients with in-hospital cardiac arrest.Sci Rep 2021; 11: 10631.
23. George ALJ, Folk BP 3rd, Crecelius PL, Campbell WB. Pre-arrest morbidity and other correlates of survival after in-hospital cardiopulmonary arrest.Am J Med 1989; 87: 28–34.