Randomized trial of a clinical decision support system: impact on the management of children with fever without apparent source
Introduction: This study examines whether a clinical decision support system could reduce the length of stay in the emergency department and the amount of lab ordered in children who had fever without apparent source.
Subjects: Children 0-16 years old with temperature equal to or greater than 38.0 Celsius who presented to Sophia Children’s Hospital in Rotterdam from July 1, 2003 until December 31, 2005. Patients who were excluded were those with chronic disease and acutely ill children (those who were treated immediately). About 50% of the children were not appropriately registered for the study and were not included. Children were excluded if there was a clear source of fever and after screening, if they would considered at low risk for serious bacterial illness (SBI).
Method: The Clinical Decision Support System (CDSS) calculated a risk score for all children who met eligibility criteria. Those considered at high risk of serious bacterial illnesses were randomly assigned to a CDSS group and a control group. Serious bacterial illness was defined as culture or radiologically proven bacterial infection including meningitis, sepsis, bacteremia, pneumonia, urinary tract infection, bacterial gastroenteritis, osteomyelitis, or ethmoiditis. In the intervention group, the CDSS determined what laboratory to order prior to being seen by a physician. These tests consisted of a complete blood count and a c-reactive protein. The control group was seen by physicians who determined what laboratory test to order.
Results: The clinical decision rules were unable to predict SBI based on those with low risk versus those with high risk scores. There also was no significant difference in time spent in the ER for those in the CDSS as compared to the control group. There was substantial difference in frequency of laboratory test order, but opposite what had been hypothesized. Children in the CDSS group had lab tests ordered 82% of the time as compared to 44% of the time in the control group. The use of the CDSS for fever without an apparent source was discontinued at the conclusion of the study.
Comments: There are some substantial possibilities for bias in this study design. Only about 50% of the initial pool of patients was registered. We have no evidence to determine whether this resulted in any skewing of the study population. Secondly, we do know that the most seriously ill patients and those with chronic illness were excluded from the study. It would be reasonable to assume that both of those groups may be at higher risk of serious bacterial illness and may be aided by CDSS.
The evaluation of a child with fever may be in part non-heuristic. The appearance of the child—how sick they appear—has a strong influence on how rigorously an evaluation is pursued. The two tests that were ordered were screening tests and could not make the diagnosis of SBI. Even with the CDSS ordered test, after the physician examined the patient additional test would have to be ordered. It is also likely that in children who did not appear ill, the physician would not have ordered any laboratory test. Thus the appearance of the child could have had a strong influence of test ordered and length of stay and may have not been adequately captured in the clinical prediction rules used in this CDSS.