COMPUTER-ASSSITED TRIAGE OF ABDOMINAL PAIN IN CHILDHOOD: A SUGGESTED CLINICAL ALGORITHM Steven Rubin, Children’s Hospital of Eastern Ontario, Ottawa, Ont., Canada (rubin@cheo.on.ca) Wojtek Michalowski, Faculty of Administration, University of Ottawa, Ottawa, Ont. Canada (wojtek@admin.uottawa.ca) Roman Slowinski and Szymon Wilk, Institute of Informatics, Poznan Technical University, Poznan, Poland Keywords: triage of abdominal pain, emergency room, Rough Sets, clinical algorithm, prospective evaluation. INTRODUCTION Abdominal pain is a common clinical emergency. The challenge at the triage of such patients is to differentiate those who may safely be discharged home from those that require urgent medical or surgical consultation. Notwithstanding the technological advances in diagnosis, the patient with abdominal pain is to a large extent reliant on the clinical expertise of the caregiver for specific management The process of reaching the final diagnosis may necessitate in-hospital observation. As abdominal pain is especially prevalent as an emergency in childhood, delay in diagnosis may result in anxiety to the child, the family and the medical staff. The more experienced the clinician in the assessment of children with abdominal pain the more rapid and the more reliable the management. Evidence from both clinical and psychological studies [Fiorovanti et al., 1988] indicates an obvious advantage in rapid triage 1 of patients with abdominal pain. The central difficulty of the triage is the choice of clinical symptoms and signs (attributes) that in combination contribute the most to the diagnosis and management of these children. A pertinent reduced set of attributes should assist the triage nurse and help the emergency room physician. Despite the large number of publications that review the management of acute appendicitis, there seems to be little information concerning validity of the clinical procedures in the triage of this condition. Most papers focus either on the physician’s ability to estimate the probability of the disease using clinical data [Hallan et al., 1997], or on the development of a scoring systems to aid in the diagnosis of this condition [Anatol and Holder, 1995]. Moreover, the 1 The term triage used throughout the paper refers to the initial assessment of the patient. attempts to develop a computer-based assistant rely heavily on the physician’s ability to make accurate probabilistic evaluations [Todd et al., 1994]. In this paper we discuss how the identification of a minimal set of clinical symptoms and signs, followed by the development of a clinical algorithm can facilitate the triage of the child with abdominal pain. We also present the results of a limited prospective evaluation of such an algorithm. The attendant goals of parental satisfaction, improved patient compliance and a reduction in overall cost of medical services may also be achieved. METHODS AND APPROACH A retrospective analysis of the emergency records of children seen at the Children's Hospital of Eastern Ontario in Ottawa, Ontario, Canada for the period 1997-1999, was conducted using a knowledge discovery methodology called Rough Sets analysis [Pawlak, 1991]. On the basis of the discharge diagnosis (i.e. final diagnosis), the patients were classified into two distinct categories: surgical consult and resolution. All patients in the surgical consult group had confirmed acute appendicitis at varying pathological stages, while the resolution group included patients where all symptoms and physical findings resolved spontaneously without specific medical or surgical treatment. The cases not specifically covered by the above groupings were classified as NYD (not yet determined) and were not included in the construction of the clinical algorithm. Universally acknowledged clinical symptoms and signs used to evaluate abdominal pain in the emergency room were recorded for each patient. They included: Published in J. Anderson, M. Katzper (eds): Simulation in the Health and Medical Sciences, San Diego, 2001 Downloaded from http://www.mobiledss.uottawa.ca