Loading...
Please wait, while we are loading the content...
Similar Documents
Everything should be made as simple as possible, but not simpler.
| Content Provider | Semantic Scholar |
|---|---|
| Author | Prengel, Andreas W. |
| Copyright Year | 2013 |
| Abstract | December 2013 • Volume 41 • Number 12 This saying, attributed to Albert Einstein, is suitable to describe an essential step in Classification and Regression Tree (CART) analysis (1), which was performed by Ebell et al (2) on data from the National Registry for Cardiopulmonary Resuscitation, to predict survival to hospital discharge after in-hospital cardiac arrest. The average survival to discharge after unexpected cardiac death in hospital has been quoted to be 17% (3). Although there are some known-risk factors for mortality after cardiac arrest, such as location of the arrest, whether the arrest has been witnessed or not, initial cardiac rhythm, age, gender, and race (4), only limited information is available regarding the effects of comorbidities predicting survival after in-hospital cardiac arrest, such as sepsis or multiple organ failure (5). Considering assistance in do-not-resuscitate orders in advance or during a patient’s hospitalization, it would be even more important to predict not only survival but the chances to return home to a meaningful quality of life after a cardiac arrest (6). Although cardiopulmonary resuscitation (CPR) is often frustrated by failure of neurologic recovery (7), most people would not desire to continue living in a severely disabled state. Even when cardiac arrest happens in the ICU, despite additional monitoring and qualified personal being available immediately, outcome in general is poor. It has been reported recently that 2% of patients admitted to the ICU suffer from cardiac arrest, with 16% surviving, but with only 20% of survivors functionally independent on hospital discharge (8). In this issue of Critical Care Medicine, Ebell et al (2) address very meaningful aspects of patient care. They report on the results of analyzing data of 52,000 patients suffering from in-hospital cardiac arrest. By using the elegant and powerful technique of CART analysis, models in shape of branching algorithms were developed to predict the likelihood of surviving with good neurologic function after inhospital cardiac arrest. One particular strength of this study is that it shows the importance of the preexisting neurologic status, leading to different outcome in the case of CPR. CART analysis seems to be especially helpful for the investigation of extensive data sources with many potential predictor variables exhibiting other than normal distribution, different degrees of variance, and complex patterns of interaction. Beyond that, CART trees are simple to interpret, when compared with more traditional methods, such as multiple regression analysis (9). There are also some drawbacks with this method to consider, as has been stated carefully by the authors. In addition, finding the optimal tree requires to compromise between a maximal tree, containing many branches and excellent accuracy, and a clinically useful tree, containing less nodes and branches at the expense of accuracy—everything should be made as simple as possible, but not simpler. Area under the receiver-operating characteristic curve values of almost 0.8, as reported in the study by Ebell et al (2), represent good, but not excellent predictive quality (10), suggesting that other than the variables investigated may have contributed to the outcome. However, although for instance the underlying form of cardiac arrest, a shockable versus a nonshockable rhythm, is highly predictive for survival in adults (11), this certainly would not help to decide in favor or against a do-not-resuscitate order. The possibility to predict who is likely to survive neurologically intact and who is not may help healthcare providers to advise patients and relatives on the preparation of advance directives, and in this respect, the article by Ebell et al (2) is inspiring to read. However, the potential decision for a do-not-resuscitation order is an individual decision, and thus, predictive models should never be applied as exclusive guidance. Quality of life certainly is not determined by the CPC score alone, but it is defined individually, based on personal, social, and religious history. Predictive models such as those presented in the exciting article by Ebell et al (2) will be helpful to classify the potential effects of outcome improving shifts, based on hospital quality management and research. It will be fascinating to see whether future developments in resuscitation medicine will change those models in favor of more patients surviving cardiac arrest and CPR with good neurologic function and allowing them to return to a meaningful life. |
| File Format | PDF HTM / HTML |
| DOI | 10.1097/CCM.0b013e31829ec84d |
| PubMed reference number | 24275390 |
| Journal | Medline |
| Volume Number | 41 |
| Issue Number | 12 |
| Alternate Webpage(s) | http://www.merit.unu.edu/publications/uploads/1301493190.pdf |
| Alternate Webpage(s) | http://www.merit.unu.edu/training/theses/Gyllensporre.pdf |
| Alternate Webpage(s) | https://doi.org/10.1097/CCM.0b013e31829ec84d |
| Journal | Critical care medicine |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
