Every year, more than 300,000 U.S. residents experience an out-of-hospital cardiac arrest (OHCA) [1, 2]. Corresponding rates of survival to hospital discharge are very low, ranging from 1.1 % to 8.1 % across regions, and survivors often experience severe neurological impairment [3].

National guidelines have called for the use of therapeutic hypothermia (TH), which involves reducing the body’s core temperature for an extended period of time (e.g., 12-24 h), on initial survivors of OHCA [4, 5]. These guidelines are based on the results of three clinical trials demonstrating the efficacy of TH for improving rates of neurologically intact survival [6–8]. These trials were small (with a combined population of 375), conducted outside of the United States, and focused on a narrowly targeted group of patients receiving treatment for OHCA. In contrast, the real-world application of TH involves a much broader range of patient selection, cooling techniques, and other implementation details [9]. In addition, a recent trial raised questions about which temperature should be targeted after cardiac arrest [10]. Nevertheless, targeted temperature management (TTM) has come to be viewed as one of many crucial elements in the post cardiac arrest “bundle of care”, which also includes management of shock, hemodynamic management, seizure suppression, and glucose control [5, 11].

In addition to hospital-based care, OHCA patients rely on rapid intervention by prehospital emergency medical services (EMS). This intervention includes in-the-field procedures such as cardiopulmonary resuscitation (CPR) as well as the choice of destination hospital. Moreover, the choice of destination hospital is potentially endogenous in the sense that prehospital EMS personnel may decide on where to transport patients based on clinical characteristics that are unobservable to researchers but related to survival outcomes. Thus, it is not well established whether and how prehospital transportation to TH centers affects survival outcomes for OHCA patients.

To address these issues, we used a unique database of linked prehospital, hospital, and mortality records for the state of New Jersey where TH adoption by hospitals increased dramatically from 2004 to 2011 [9]. Using data from 2009-2010, we compared neurologically intact survival outcomes for patients treated at TH centers versus those treated at other hospitals. To account for potential endogeneity in prehospital transportation of patients, we used an instrumental variable (IV) approach based on distances from the scene of the OHCA to the nearest TH and non-TH facilities.

We identified 6887 adults treated by EMS for OHCA in 2009-2010 (Fig. 1). Of these, 1133 cases had resuscitation efforts terminated in the field and were removed from the analysis. Among the 5754 treated and transported cases, 5017 (87.2 %) could be linked reliably to hospital and mortality records, while the remaining 737 cases contained insufficient information for reliable linkage. Among the 5107 linked cases, 166 were transferred from the initial destination hospital to other hospitals that were not identified on the NJDDCS record. Since these cases could not be followed throughout the entire course of OHCA treatment, they were excluded from the analysis. (Among these 166 excluded cases, 46.4 % were initially transported to a TH center, a percentage that is very similar to the corresponding percentage in the final study sample documented below.) After excluding 9 additional cases with missing values of covariates, we retained 4842 observations in the analysis below.

The Stock-Yogo test confirmed the use of differential distance as an appropriate IV (F = 2349.91, p < 0.0001). The stage 1 residual was statistically insignificant in both stage 2 models (In-hospital survival: Z = -0.08, p = 0.934; 30-day survival: Z = 0.94, p = 0.349). As a result, our data provided no evidence of endogenous transportation to TH centers (see Tables, Additional file 1, for full stage 1 and stage 2 estimation results). Thus, we focus on estimates from non-IV based logistic regression models. In both of these models, likelihood ratio tests rejected the ordinary logistic regression specification in favor of the mixed effects model for both study outcomes (Survival to discharge: Chi-square = 143.67, p < 0.01; 30-day survival: Chi-square = 11.86, p < 0.01).

After adjusting for covariates, treatment at a TH center was not significantly associated with neurologically intact survival to hospital discharge. However, after adjusting for confounders, the odds of 30-day neurologically intact survival were 70 % greater for patients taken to TH centers relative to other patients (Table 2). Findings were very similar in sensitivity analyses that examined raw survival outcomes without accounting for neurological status (see Appendix for details).

In this study, we found a strong positive association between transportation to a TH center and neurologically intact 30-day survival among OHCA patients. We did not, however, observe a similar relationship between TH centers and neurologically intact survival to hospital discharge. In interpreting these results, one must bear in mind that our designation of a “TH center” was an indicator of each hospital’s capabilities, not a measure of whether TH was provided to individual patients.

This distinction is important, since some TH centers provide the procedure to only a very narrowly selected group of patients on the basis of ECG rhythms, age, and other selection criteria [9]. Thus, our analysis evaluates the effectiveness of care rendered to all OHCA patients at hospitals that have adopted TH, regardless of whether any TTM was actually performed on each individual.

Recently established consensus guidelines emphasize that optimal post-arrest care entails a broad bundle of key elements in addition to TTM, including appropriate management of shock, hemodynamic management, seizure suppression, and glucose control [5]. Our findings suggest that hospitals providing TH may be more attuned to this broad set of care elements for all of their OHCA patients.

Another important observation is that although 30-day neurologically intact survival was associated with transportation to a TH center, there was no difference in survival to hospital discharge. This finding is similar to prior research using data from the Resuscitation Outcomes Consortium (ROC), which demonstrated that 24-h OHCA survival in the community often differs from conditional survival to hospital discharge [21]. Both of these related findings suggest that immediate and longer-term outcomes may be influenced by different mechanisms. For example, our findings may reflect superior care by TH centers in later stages of post-arrest treatment such as care provided in the intensive care unit, which has greater potential to affect longer term outcomes than initial treatment in the emergency department. In addition, longer-term survival may also be influenced by well-coordinated post-discharge follow-up care. Unfortunately, we are unable to ascertain from our database whether the improved longer-term survival outcomes are driven by interventions taking place within TH centers or during the post-discharge follow-up stage.

Our study provides unique and important insights into the “real-world” consequences of TH adoption. Previous work has been limited to small-scale clinical trials outside of the U.S. (where acute and post-acute care delivery systems are often organized and operated very differently) and observational studies using before-after designs to describe care at single TH centers [6–8, 22–24]. In contrast, our large-scale study of NJ includes a much larger number of patients and a heterogeneous range of TH and non-TH hospitals. In addition, our analysis is the first to our knowledge that uses an IV approach to account for the potential endogeneity of patient transportation to TH centers by prehospital EMS. Our finding that transportation to a TH center is not endogenous suggests that local EMS units do not alter the transport destination of OHCA patients on the basis of survival probability or likely benefit from care at a TH center.

Some experts have argued in favor of regional systems of care for OHCA, similar to those established for trauma, burn, stroke, and STEMI centers [25]. Specifically, if a patient in cardiac arrest is resuscitated in the field by EMS, then the arrest outcome may be improved if the patient is taken directly to a TH center, even if this means bypassing a closer non-TH hospital. Our finding of more favorable outcomes at TH centers provides support for regionalization. Moreover, even though longer EMS response time was independently associated with worse outcomes, we found that EMS transport time to the destination hospital was not a significant predictor of neurologically intact survival. Our analysis is also consistent with prior work showing better OHCA outcomes in larger hospitals but no significant difference between teaching and non-teaching hospitals [26].

Our findings should be viewed in light of some limitations. Since we were unable to identify individual patients who received TTM, our analysis is an evaluation of the care provided by TH centers and not an evaluation of TH itself. Similarly, we did not have information on parallel resuscitation interventions such as use of percutaneous coronary intervention (PCI), which is known to affect OHCA outcomes but is not used uniformly across hospitals [27–31]. We were, therefore, unable to determine whether favorable 30-day outcomes at TH centers were influenced by use of PCI or other processes of care. Our data also did not allow us to identify the elapsed time between arrest and return of spontaneous circulation, which is strongly associated with brain injury and overall survival.

Because detailed neurological assessments were are not available in our hospital data, we had to proxy neurological status among OHCA survivors using hospital discharge codes. More detailed neurological assessments would likely change the outcome classification for some patients in the study but we do not expect the classification would change differently for patients transported to TH centers versus other hospitals. Similarly, some OHCA patients may have originally been nursing home residents who would have returned to their nursing home if they survived to hospital discharge. Such individuals would be classified as having a poor outcome even though they may have returned to their original neurological status after hospital discharge. Although our data do not allow us to identify how often these situations occur, the imprecision in our measure of neurological status does not create any systematic bias for or against TH centers.

In addition, sensitivity analyses confirm our main findings based on the proxy measure for neurological status. First, although post-discharge placement may be influenced by socioeconomic and geographic concerns, our findings were not influenced by patient insurance status or poverty rates in hospital service areas. Second, conclusions were identical in models that examined survival outcomes regardless of neurological status.

Although participation in New Jersey’s EMS data collection system is nearly universal among ALS units, responses to OHCA calls from a number of BLS units are likely to be missing. Nevertheless, we expect these missing cases to be minimal, since most responses to OHCA will involve ALS response at some point of the episode.

Finally, despite our use of validated data linkage software [32], the records for 737 patients treated and transported by EMS were lost due to insufficient identifying information for reliable linkage. Records for an additional 166 patients were lost after these individuals were transferred to other hospitals. It is also possible that some records may have been linked erroneously. While the existence of missing or erroneous links adds uncertainty to our conclusions, we do not expect these issues to add any systematic bias for or against the outcomes measured at TH centers. In addition, confidence in the validity of our study data is enhanced by its comparability to the previously published national rate of survival to hospital discharge among adults treated by EMS for OHCA: 13.7 % for NJ in 2009-2010 versus 9.8 % nationwide in 2010 [2].

Despite these caveats, our study provides evidence that post-arrest outcomes are more favorable when patients are transported to TH centers. These improved outcomes are not apparent, however, until after hospital discharge. This finding may reflect superior care by TH centers in later stages of post-arrest treatment such as care provided in the intensive care unit, which has greater potential to affect longer term outcomes than initial treatment in the emergency department. More research is needed to delineate specifically the mechanisms that lead to these improved outcomes at later stages of post-arrest treatment.