We thank Mr. Stewart for his interest in our article1 and for his comments. We agree that accurate and precise data are always desirable for testing of hypotheses. However, we believe that it is not “relatively simple” to acquire accurate time data for the response to in-hospital cardiac arrest. Indeed, Mr. Stewart also acknowledges that no one else is reporting data better than the Get With The Guidelines–Resuscitation registry. The reality is that time data are difficult to capture accurately given the chaotic nature of an emergency response to in-hospital cardiac arrest. Therefore, some degree of inaccuracy likely exists. Furthermore, recorded clock times are impossible to validate as there is no gold standard with which to make comparisons.
Part of Mr. Stewart’s comments arises from our full disclosure of the distribution of times to cardiopulmonary resuscitation and subsequent times to either defibrillation or epinephrine. Our intention was clear: to fully inform the readers so that they may judge the validity of the conclusions themselves based on the best available, although imperfect, time data. Importantly, in performing our analysis, we excluded values that were implausible at face value, and meticulously documented exactly what exclusions had been made at each stage of analysis. Therefore, we believe our findings are unlikely to be invalidated by a relatively small subset of aberrant data. In addition, we posit that any misclassification of time delays in our data would be nondifferential and would bias the results toward the null hypothesis, (i.e., toward no effect of delay on survival). Therefore, our results represent a conservative estimate of the harm caused by the delays. The actual impact of delays in care on survival is likely to be larger.
The value of the Get With The Guidelines–Resuscitation database is to allow compilation of a large number of cases from multiple centers to ensure generalizability of study inferences. Despite some imperfections, this allows for the development of testable hypotheses as well as identifying limitations in the current standard of care. Such a database is arguably better and more clinically informative than expert opinion alone. The database and its analyses also provide a framework for refinement of the data gathering mechanism. Only by publication of analyses based on this set of data, and candid discussions such as this one, can improvements in data quality be made. At present, however, the data represent our best opportunity to understand the impact of delays in key processes on resuscitation outcomes and therefore provides valuable information. That information not only tells us what we know (that delays can lead to adverse outcomes for patients), but also quantifies to what degree delays lower the likelihood of survival.
The authors declare no competing interests.