Impact of Decision-Support On Clinicians' Assessment of Febrile Neutropenia Risk.

Abstract

Abstract 1409Poster Board I-431 Background:Febrile neutropenia (FN) remains one of the most clinically significant adverse events associated with chemotherapy administration. Myeloid growth factors can prevent febrile neutropenia; however, administration patterns are highly variable and, as a result, are less than optimally effective, both clinically and economically. Here, we investigate the role that an electronic clinical decision support system (CDSS) based on a validated predictive model can play in standardizing administration of myeloid growth factor. Methods:Medical oncology clinicians (physicians, physician's assistants, nurse practitioners) were recruited from the Atlanta, GA and Research Triangle Park, NC areas, according to the primary cancer type they treated. After Institutional Review Board approval, a total of 39 clinicians were recruited, 22 of whom were medical oncologists, 2 were medical oncology fellows, 3 were physician assistants, and 12 were oncology nurse practitioners. Each clinician was presented with a series of eight simulated patient cases (in randomized order) relevant to the cancer type they primarily treat in practice. Each simulated patient case contained FN risk factor information for the patient, including demographics, baseline labs, medical history, planned regimen and other treatment information. Clinicians were asked to provide their estimation of the risk of febrile neutropenia as high, intermediate or low probability using their typical heuristic and process. The patient's risk for FN as determined by a validated logistic regression model was then presented to the participant through the CDSS. Afterwards, the clinicians were again asked their assessment of the patient's risk. Fixed-marginal kappa (κ) and Kendall's Coefficient of Concordance (W) were used to determine the concordance between the participants' assessments both before and after the CDSS intervention. Results:304 of the 312 (97.4%) case reviews were completed with 8 participants failing to complete the final case review in the time allotted. These data, along with the distribution of providers by cancer type, are presented in Table 1. As shown in Table 2, the overall inter-observer agreement as expressed by kappa significantly improved from 0.105 (slight agreement) to 0.888 (almost perfect agreement) after use of the CDSS (p < 0.001). Similarly, the strength of relationship as expressed by the overall Kendall's Coefficient of Concordance also significantly improved with use of the CDSS from 0.426 to 0.962 (p < 0.001). The clinicians changed their assessment of risk in 174 of the 304 (57.2%) cases after using the CDSS.Table 1Provider Type and Cases Reviewed by Cancer TypeCancer TypeNumber of CliniciansCases ReviewedMDFellowPANPBreast400246 of 48NSCLC301246 of 48SCLC411264 of 64Ovarian301138 of 40Lymphoma510264 of 64Colorectal300346 of 48Table 2Kappa and Kendall's Correlation Coefficient by Cancer TypeCancer TypePre-InterventionPost-InterventionκWκWBreast0.0890.5200.6990.877NSCLC0.0020.1910.9240.979SCLC0.1110.3840.9530.985Ovarian0.0970.6060.9030.975Lymphoma0.1550.4461.0001.000Colorectal0.1730.4080.8500.956 Conclusions:Use of the clinical decision support system dramatically influenced clinicians' assessment of risk and resulted in a significant increase in the concordance between the assessments of medical oncology clinicians. This effect was observed across several cancer types. Disclosures:Webb: Proventys: Employment, Equity Ownership. Head: Proventys: Employment, Equity Ownership. Leithe: Proventys: Employment, Equity Ownership. Wu: Proventys: Employment, Equity Ownership. Singh: Proventys: Employment, Equity Ownership. Lyman: Proventys: Consultancy, Equity Ownership, Membership on an entity's Board of Directors or advisory committees.