Study objectives: We determine whether variation in the frequency of emergency department (ED) visits is beyond what would be expected by chance after adjusting for known confounders such as calendar and weather variables. Methods: The Poisson goodness-of-fit test was used to determine whether emergency visits in a Midwestern teaching hospital occurred randomly over time. Total visits, as well as selected diagnoses, were analyzed. Final diagnosis was determined using the physicians' final chart diagnosis, which was assigned an <i>International Classification of Diseases, Ninth Revision</i> code by one of the investigators. Data on appendicitis were obtained for 6 years to check for randomness of distribution; only cases of operatively confirmed appendicitis were included in the analysis. We also looked at the distribution of specific diagnoses that might be expected to occur randomly (eg, atrial fibrillation, urinary retention, headache). Multivariable Poisson regression was used to control for calendar and weather variables. Autocorrelation coefficients were used to detect "contagion" from one period to the next. Results: There were 24,100 ED visits in 1998, with 2,054 distinct final diagnoses. Forty-six visits were laboratory tests only, and 5 records were lacking a final diagnosis. The most common diagnoses were "abdominal pain" (643 visits), upper respiratory infection (548 visits), and open wound of finger (544 visits). Overall, visits did not occur randomly, even after controlling for time of day, day of week, academic calendar, season, and weather variables (<i>P</i><.001). Visits were more common on Sundays, in the fall, and on warm days. Of 32 diagnoses we predicted would occur randomly (eg, have a Poisson distribution), 29 did so, including appendicitis, congestive heart failure, chest pain, venous thrombosis, reflux esophagitis, intestinal obstruction, gastrointestinal foreign body, gastrointestinal hemorrhage, hypertension, right-lower-quadrant abdominal pain, subarachnoid hemorrhage, transient ischemic attack, atrial fibrillation, angina, cerebral hemorrhage, constipation, seizure, intermediate coronary syndrome, dizziness, hematuria, kidney stone, migraine, acute pancreatitis, rectal hemorrhage, blood in stool, syncope, tension headache, Tietze syndrome, and urticaria. The other 3, cerebrovascular accident, migraine headache, and back pain, were not randomly distributed. Conclusion: ED visits did not occur randomly over time. Most diagnoses that would be expected to occur randomly did so, but a few exhibited nonrandom variation. ED staff should be alert to nonrandom occurrence of disease as a mechanism for finding unsuspected causes or associations of different diseases.
Read full abstract