This study provides the first U.S. estimates of ED visits for acute pancreatitis. From 1993 through 2003, there has been a steady increase in the absolute number and population rates of ED visits for acute pancreatitis. The number of estimated ED visits in 2003 (318,000) is approximately four times a recent U.S. government estimate of 80,000 annual cases [26]. Several European studies have described an increase in the incidence of acute pancreatitis in a variety of locations throughout Western Europe [1-3,27-30]. While these studies purport to describe incidence rates rather than ED visits – some use hospital admission data only, some include outpatient visits, and others include autopsy data – due to the lack of data from the United States and from EDs in general, they provide the most relevant context for interpreting our data. The overall rate of ED visits of 6.8/10,000/year that we describe here is higher than the incidence rates identified in most European studies, which report incidence rates from 1.5/10,000/year to 7.3/10,000/year in recent decades [1-3,27-34]. The 2003 peak of 10.9 ED visits per 10,000 U.S. population was higher than the peak annual rates in all the studies cited above.
There are several potential methodological explanations for these results. Our study is ED-based while some of the European studies cited above were based only on hospital admissions [2,3,27-30,32,34]. Extrapolating from our data, in which only 65% of ED visits ended in hospital admission, our calculated hospital admission rate would be significantly lower. Nevertheless, other European studies were prospective and included all healthcare facilities in the study area including outpatient clinics and the incidence rates reported by these investigators were notably lower than ours [33,35].
Our study did not impose strict criteria for the diagnosis of acute pancreatitis. The possibility that some of our sampled and selected ED visits did not have pancreatitis at all, or suffered exacerbations of chronic pancreatitis rather than episodes of acute pancreatitis is real. As noted above, though the NHAMCS reports a less than 2% error for items involving medical coding generally, the exact accuracy of pancreatitis diagnoses in the database is unknown. This problem of assuring accurate diagnosis was encountered in three large registry-based European studies with the capacity for chart review and validation (which our study lacks due to the anonymous nature of the NHAMCS). These showed from 82% to 90% rates of accurate diagnosis and coding for cases entered as acute pancreatitis [2,3,28]. All of these registries contained inpatient admissions, and so benefited from a sometimes lengthy hospital stay to clarify diagnosis – a luxury not afforded by the ED. In this respect, it is reassuring for our data that the National Hospital Discharge Summary, the largest national database for inpatient admissions in the U.S. has similarly documented a rise in hospital admissions for acute pancreatitis from 108,000 in 1987 to 224,000 in 2003 and that the 2003 figure for hospital admissions correlates closely with the number which would be computed from our ED visit and hospital admission rates [36,37]. Because of its anonymous nature, the NHAMCS database does not allow identification of repeat ED visits by the same patient for the same episode of acute pancreatitis. This may be mitigated somewhat by the inclusion of each hospital in the sampling frame for only one month out of approximately every 15, and is likely to have a small impact at most. While the inclusion of sample records without true acute pancreatitis or the capture of multiple visits for the same episode of acute pancreatitis would artificially elevate the observed caseload, one of the European studies cited documented capture of only 76% of a cohort of known pancreatitis patients by the national database, a source of error that would offset the over counting in terms of total numbers of cases [3]. Another study has shown that acute pancreatitis case selection based on reliance on diagnostic lists as in our study misses a large quantity of cases [38]. Despite these known limitations of registry-based studies and the difficulty of precisely validating NHAMCS for this particular cohort of cases, our selection criteria fell well within the standard methodology used to establish the literature in this field, and indeed were quite conservative.
Thus, we believe that methodologic reasons alone are unlikely to explain the high ED visit rates we describe, which are significantly higher than most of the incidence rates reported by European investigators. While NHAMCS itself does not allow calculation of incidence rates, we consider a higher incidence rate of acute pancreatitis in the U.S. to be the simplest explanation for the relatively high ED visit rate in this country. Methodological differences between our study and the numerous European studies cited cannot explain the observed upward trend in the overall number and rate of ED visits for acute pancreatitis in the U.S., nor can they explain differences in visit rates between subgroups within our study. Our selection criteria did not change for the different years of the study period, were applied uniformly, and no significant changes occurred in diagnostic modalities used in acute pancreatitis during the study period.
The rate of ED visits for acute pancreatitis among U.S. blacks was significantly higher than that among whites throughout the study period. To our knowledge, this is the first time that this racial disparity has been described in a large-scale nationwide study. This study is unable to further investigate the cause of this race disparity since the etiology of each case and whether or not it was a recurrent episode is not known. Because we cannot track the etiology of acute pancreatitis in any of our study subjects and do not know how many of the cases were recurrent, neither an explanation nor a means of rectifying this important racial disparity is possible. However, we believe that these results provide a compelling argument for future research on this topic.
Blacks were significantly less likely to be admitted than whites when controlling for age, insurance type, hospital location, and urgency at triage. This finding deserves particular attention given the high population rate of ED visits for acute pancreatitis for blacks, and has several possible explanations. Blacks in our study may have presented with milder pancreatitis than whites. Some studies have suggested that minority patients are more likely to use the ED as a source of primary care and more likely to present with non-urgent conditions [39]. While the NHAMCS does not provide many objective measures of disease severity, our observed disparity in admission likelihood persisted when we controlled for urgency at triage. Hospital differences may account for the disparity – hospitals treating blacks may be overcrowded, understaffed, or poorly funded. These hospitals may not discriminate between the black and white patients they treat, but because their patient population is disproportionately black, this phenomenon may produce the kinds of disparity we observed [40,41]. Patient decisions cannot be assessed with our data, and blacks in our study population may have declined admission in equivocal circumstances, while whites were more prone to accept. Finally, racial discrimination cannot be excluded and race-based differences in treatment and disposition of a variety of types of patients have been reported [42-44]. Variations in ethnicity, insurance type, urban setting, or U.S. region did not correlate with decision to admit.
The overall admission rate in our study was 65%. We are not aware of prior studies that have examined acute pancreatitis from the perspective of the ED so there are little (if any) data with which to compare this figure. Some European investigators have implied that in their study populations essentially all patients diagnosed with pancreatitis, even those diagnosed as outpatients, were referred for hospital admission [35]. We do not know how many of the people discharged from the ED in our study failed outpatient management and returned for admission. The very fact that the number of cases admitted directly from the ED to intensive care units (ICUs) is too small for reliable statistical extrapolation, illustrates the relative infrequency of initial admission to the ICU. This is in spite of estimates that between 14 and 20% of acute pancreatitis cases are considered to be "severe" [26,30,34]. We do not know how many of our admitted patients went on to be admitted to the ICU during their hospital stay, and at what point in the stay that might have happened, nor if initial admission to the ICU would have had any positive impact on the care of patients with this type of hospital course. All of these questions merit study and our data provide a foundation on which to build such studies. In the meantime, however, our data suggest that our population suffered from mild pancreatitis at the time of presentation, that previous estimates of the percentage of acute pancreatitis cases becoming severe are overstated, or that patients with severe pancreatitis are frequently admitted to low acuity inpatient units, as indeed has been noted by others [27]. In light of this latter possibility, it is reassuring that the one piece of information that we have available to us that is associated with both disease severity and prognosis – age – was independently associated with an increased likelihood of admission to the hospital (as noted, ICU admission numbers were too small for this type of analysis) [45,46].
Overall, diagnostic imaging was rarely used during ED visits for acute pancreatitis. CT scanning can be helpful in establishing the diagnosis of pancreatitis when it is in doubt and may be prognostically useful, though it is more frequently reserved for analyzing complications of severe acute pancreatitis such as necrosis, abscess, or pseudocyst [47,48]. Ultrasound is less sensitive for confirming the diagnosis of acute pancreatitis, but by detecting gallstones may suggest a biliary etiology, though this distinction is seldom important to make in the ED [49]. Infrequent use of these tests in the ED is probably appropriate.
Our data yield a variety of interesting questions that are beyond the scope of this study. For example, NHAMCS data do not allow us to evaluate the etiology of each attack of acute pancreatitis. The development of the International Classification of Disease, 10th Revision, Clinical Modification, which features distinct codes for alcoholic and biliary pancreatitis will likely aid future investigators in this regard. We could not determine whether a case represented a first attack or a recurrent episode, a limitation of the anonymous nature of the NHAMCS and one that could be addressed by a prospective audit. Likewise, we could not evaluate mortality or any other outcomes beyond the ED, nor could we analyze care after admission to the hospital. Nationwide practice patterns regarding length and cost of hospital stay, rates of ICU admission, the use of antibiotics, the use of invasive therapies such as endoscopic retrograde cholangiopancreatography, percutaneous drainage, or surgery, and the use of enteral versus parenteral nutrition in inpatients with acute pancreatitis are essentially unknown. A large-scale study using nationwide inpatient data could address these large knowledge gaps in this important but relatively understudied condition.
Answers to all your Biology Questions
