PILOT SURVEY OF PHYSICAN PREFERENCES FOR TEST THRESHOLDS FOR PEDIATRIC NEUROIMAGING GUIDELINES

Purpose: To determine physician preferences for test thresholds for pediatric neuroimaging guidelines, and to identify beliefs associated with threshold preference.

Method: The following types of attending physicians at our institution in Canada were surveyed: family physicians, general pediatricians, child neurologists, pediatric neurosurgeons, and pediatric radiologists. Survey participants were presented with results from a hypothetical clinical prediction rule that determines the risk of treatable pathology in children with recurrent headache. Participants were asked to state whether, as part of a guideline development committee, they would recommend routine neuroimaging in addition to follow-up for children with 0.1%, 0.4%, 1%, 10%, 25%, or 50% risk of treatable pathology.  Levels of risk were based on a prior unpublished pilot. Participants were asked if they would be willing to endorse guidelines with the recommendation opposite their preference for the 1% category, and to rate their agreement with 13 beliefs about pediatric neuroimaging guidelines using a 7-point Likert scale.     Participants were categorized as having a high (would not recommend neuroimaging for 0.4% or lower categories) or low threshold preference. Belief items were also summarized using binary variables. Associations between threshold preference and beliefs were evaluated using Fisher’s exact test.

Result: At the time of submission, 12 eligible participants completed the threshold questions and 11 completed the entire survey. The proportion of subjects who would recommend routine neuroimaging for each risk category was: 50% (12/12); 25% (12/12); 10% (12/12); 1% (9/12); 0.4% (5/12); and 0.1% (3/12).  All participants’ answers were internally consistent. Of five subjects with low thresholds, two were willing to endorse guidelines recommending against neuroimaging for the 1% category.    Agreement with several belief items about neuroimaging showed nonsignificant associations with a high threshold preference in this small sample: that it is appropriate to consider the monetary cost to society (p=0.061); that patient comfort should be considered (p=0.182); and that it would be possible to create a clinical prediction rule that accurately determines risk for children with recurrent headaches (p=0.182).

Conclusion: These preliminary results suggest that there is substantial variability in physician preferences for test thresholds for pediatric neuroimaging guidelines, which could impact both guideline development and adherence. Specific beliefs may be associated with threshold preference. We plan to perform a similar survey in a national sample.