CORRESPONDENCE
Because we evaluated small samples from populations that were not necessarily normally distributed, we used nonparametric statistical techniques 2 that necessitated the reporting of average values as medians. The use of nonparametric methods likely resulted in more conservative estimates of statistical significance, for they generally are less sensitive than their parametric counterparts. During the second half of the study, triage criteria for trauma patients were limited to clinical shock, cardiorespiratory distress, and cardiac arrest. Medical direction of the treatment and triage of each patient is accomplished thr0ugh one of six base stations by on-line emergency phy-
sicians using county-wide protocols. The trauma score was not used for prehospital care until 1983, or for hospital care until 1984.
Richard H Cales, MD, Chairman Department of Emergency Medicine Portland Adventist Medical Center Portland, Oregon 1. West JG, Trunkey DD, Lira RC: Systems of trauma care: A study of two counties. Arch Surg 1979;114:455-460. 2. Siegel S: Nonparametric Statistics for the Behavioral Sciences. New York, McGraw-Hill, 1956.
Resuscitation Time and Ventricular Fibrillation To the Editor: I would like to comment on the article, "Resuscitation Time and Ventricular Fibrillation - - A Prognostic Indicator," b y Pionkowski et al in the December 1983 issue of Annals (pages 733-738). The main point of the article is that the duration of resuscitation for patients who are ultimately admitted to the hospital following out-of-hospital cardiac arrest is an important predictor for subsequent hospital mortality. This findingis to be expected; intuitively one would predict that an h0ur-long resuscitation would result in a patient in poor condition at admission, whereas a patient resuscitated a few minutes after collapse should be closer to precoilapse condition. As a consequence of this observed relationship between duration of resuscitation and hospital mortality, the authors conclude that "efforts must be directed at decreasing the total arrest time." Although I agree with this statement, I do not think such a conclusion follows from the facts presented. Mor e importantly, however, the authors also reached other conclusions that may be misleading. In the last paragraph they state, "Evaluation of the effects of bystander CPR and witnessing of cardiac arrest on resuscitation time in our rapid response paramedic system indicates that neither factor was a prognostic indicator of survival of resuscitated prehospital ventricular fibrillation patients." Presumably this statement is based on the fact that after conditioning on duration of resuscitation time, they found no effect on hospital mortality of either bystander CPR or witnessing the arrest. In my view, the problem here is that the cart has gone before the horse. The relevant question is whether the duration of resuscitation is predictive of hospital mortality after adjusting for fundamental factors that likely affect the duration of the resuscitative effort. In particular, these are whether the arrest was witnessed or unwitnessed; the delay to initiation of definitive therapy (especially defibrillation); the delay from collapse to initiation of CPR (bystander CPR can be considered as a surrogate for a shortened delay to initiation of CPR); and, finally, the functional status of the patient prior to cardiac arrest (a factor that is difficult to quantify and, hence, to adjust for). Reanalysis of the data to 158/375
provide a quantitative measure of the effect of resuscitation time after adjusting for the more fundamental factors would be an important contribution. In an unpublished analysis of 242 witnessed cardiac arrests in which the victims were found in VF by paramedical personnel, we noted that time from collapse to initiation of CPR, fire department response time, age, and length of time required to obtain a stabilized rhythm were linearly and inversely related to survival. Age did not appear t ° be a factor in resuscitation, but was a factor in hospital mortality (P < .04). We found that even after adjustment for fundamental factors, the duration of resuscitation was a significant predictor of hospital mortality (P < .003). Possibly, with better information on the functional status of the patient prior to collapse, the duration of resuscitation would have been less significant. Never'theless, these data (and probably Pionkowski's) suggest that methods that either shorten the duration of resuscitation or protect the patient during resuscitation should be sought and evaluated. We found, however, that even after adjustment for duration of resuscitation and other factors, the delay to initiation of CPR was a significant predictor of hospital mortality (P < .004). Because the delay to initiation of CPR is a factor upon which community involvement can have a major impact, it is important that it not be denigrated by a misleading analysis. Alfred Hallstrom, PhD Professor of Biostatistics Warren G Magnuson Health Sciences Center University of Washington Seattle
In Reply: We thank Dr Hallstrom for his interest in our article and are pleased to learn that he will soon report the results of a study that will confirm the association of resuscitation time and survival. To replace assumed curves, i we have defined an actual survival curve for our rapid response paramedic system
Annals of Emergency Medicine
14:4 April 1985