Which patients will survive to discharge after an in-hospital cardiac arrest?

Data indicate that physicians are not able to determine who will survive after cardiac arrest based on clinical judgment alone. In one study, when physicians were given clinical summaries and asked to predict which patients would or would not survive, their predictions were no better than chance [15].

Factors that have been associated with survival include [16]:

Very poor survival has been noted in patients with both age over 60 years and CPR efforts lasting more than 10 min [17]. In a comprehensive meta-analysis, pre-arrest factors predictive of poor survival included older age, inability to perform activities of daily living (ADLs) prior to hospitalization, abnormal mental status, abnormal renal function, hypotension, and malignancy [18]. The likelihood of survival declines progressively with age, and there is a prominent drop off in survival after the age of 75 years.

Three scores have been developed to predict survival following in-hospital cardiac arrest. The Pre-Arrest Morbidity (PAM) Score was derived from a prospective series of in-hospital cardiac arrests at a teaching hospital in the United States [19]. The investigators found that pre-arrest hypotension, renal dysfunction and age > 65 were strong univariate predictors of poor outcome. Other factors included in the PAM score are listed in Table 7.2. Patients with a PAM score of 7 or greater had a very low likelihood of long term survival (<15 %), and no patient with a score of 8 or higher survived. For example, a patient with a serum creatinine >2.5 mg/dl, homebound lifestyle, NYHA class III heart failure, and an S3 gallop at admission would not be expected to survive an in-hospital cardiac arrest. The PAM score has been validated in three different patient populations and in each of these studies, no patients with scores >8 survived [19–21].

The modified PAM index (MPI) added dementia (2 points), reduced cancer from 3 to 2 points, assigned 1 point for myocardial infarction (MI) only if the arrest occurred more than 48 h later, and removed cirrhosis from the score [22]. While no direct validation of the PAM vs. the MPI was performed, the authors based the modifications on a robust data pool derived from 32 studies.

The Prognosis After Resuscitation (PAR) score was developed from a meta-analysis of 14 studies of survival following in hospital post cardiac arrest and assigns a score based on 8 variables, each of which is assigned a score ranging from −2 to 10 points as indicated below [23].

When the PAR score was applied to a cohort of 218 patients, 37 (20.1 %) had a score greater than 8 and none of these patients survived.

The utility of the PAM, MPI, and PAR scores has been assessed in independent populations. In one series from a hospital in England, each score was able to reliably predict non-survivors with 100 % specificity [24]. However, the sensitivity was low and varied from 20 to 30 %. Thus, the majority of patients who do not survive are not reliably identified by the scores. The authors noted that if either the PAR or MPI score was greater than 6, the sensitivity for predicting death increased to 41 % with no loss of specificity (i.e., all patients with a score >6 died). However, this has not been independently validated. Since each of the scores selected slightly different populations, use of all 3 scores could maximize identification of patients for whom resuscitation attempts are likely to be futile.

Another large meta-analysis evaluated these scores along with the APACHE-II (Acute Physiology and Chronic Health Evaluation) score [25] and found similar results; i.e., there was high specificity at the expense of poor sensitivity [18]. Patients with a PAM >8, a PAR >8, or an MPI >6 were not likely to survive. APACHE-II had less robust specificity, and an APACHE-II score >20 was associated with a 4.8 % chance of survival.

Prediction of a successful neurological outcome after ROSC is challenging in the era of therapeutic hypothermia. As with physicians’ ability to predict cardiac arrest survival, neurologists’ clinical judgment for predicting neurological outcomes lacks reliability. One case series demonstrated that in several patients for whom board certified neurologists had predicted grave prognoses, full recovery was achieved [26]. Many of the standard rules to predict adverse prognosis do not apply in the era of therapeutic hypothermia. For example, the absence of an extensor response to pain on day 3 after cardiac arrest is considered a grave finding [27]. In patients treated with therapeutic hypothermia, however, the extensor pain response was absent in 10 % of patients with satisfactory neurological recovery, suggesting that this sign alone cannot be used to decide when to withdraw support [28]. Generalized myoclonus on the first day after a cardiac arrest has also been strongly predictive of non-recovery of neurological function [29]. There are now, however, reports of patients treated with therapeutic hypothermia who survived with good neurological outcome despite early generalized myoclonus [30]. Neurological examination at 72 h after cardiac arrest has often been the standard to determine if neurological function will return. In patients receiving therapeutic hypothermia, this is no longer considered reliable [31–33]. In one series, 6 of 34 patients with persistent coma 4–5 days after arrest later regained consciousness and were alive 6 months later. Neuron-specific enolase (NSE) is a serologic biomarker and a value >33 μmol/L has been considered a reliable predictor of poor outcome [34]. However, in a more recent prospective trial of patients treated with hypothermia, 10 of 99 patients with an NSE level >33 μmol/L had a good neurological outcome [28].

To provide insight into how best to assess neurological prognosis in patients receiving therapeutic hypothermia, Friberg et al. summarized the available data and suggested a multimodality approach with continuous evaluation including daily neurological examinations and simplified electroencephalographic (EEG) recordings [33]. Continuous amplitude integrated EEGs (aEEGs) can provide prognostic information, especially in the presence of 2 distinct patterns [35]. A continuous pattern on aEEG is defined as continuous cortical activity within the delta, theta, and/or alpha bands of the patient’s standard EEG. A suppression burst pattern on aEEG is seen as high voltage bursts (>50 μV) of slow waves interrupted by suppression (low amplitude <10 μV lasting >1 s). In patients whose aEEGs showed a continuous pattern, 90 % regained consciousness. Conversely, patients with a suppression burst pattern did not survive. EEGs are also important for detecting electrographic status epilepticus (ESE) [36]. Somatosensory Evoked Potentials (SSEP) can also be helpful. The N20 potential is an electrical signal measured over the somatosensory cortex during contralateral wrist stimulation, thus suggesting intact cortical sensory activity. Bilateral loss of N20 potentials is associated with very poor neurological outcome in cardiac arrest patients who underwent treatment with hypothermia and rewarming [27]. NSE levels measured over time can provide additional information, especially if they remain low [33]. If NSE levels are consistently <33 μmol/L, an investigation for other causes of persistent coma, such as ESE or prolonged sedation effect, is advised. Computed tomographic (CT) imaging is valuable for detecting massive cerebral edema and herniation. Quantitative CT within the first 24 h after arrest to measure the gray to white matter attenuation ratio (GWR) can be helpful. In one study, only 2 of 58 patients (3.4 %) with a GWR <1.2 survived, and no patients (0 of 20) with a GWR <1.1 survived [37]. Friberg et al. suggested that withdrawal of care could appropriately be performed under the following circumstances: