From the 2018 HVPAA National Conference
Christopher Petrilli (Michigan Medicine), Lauren Heidemann (Michigan Medicine), Matthew Johnson (Michigan Medicine), Katie Schwalm (Michigan Medicine), Lee Schroeder (Michigan Medicine), Scott Owens (Michigan Medicine), Alice Cusick (Michigan Medicine)
Background
Approximately 0.2% of patients treated with heparin will experience HIT, a serious condition with an estimated 20% mortality rate. The diagnostic algorithm of HIT is an initial PF4-ELISA followed by confirmatory SRA testing – both are expensive and resource intensive. In practice, standard testing procedures for HIT are highly variable.
Objectives
1) To retrospectively quantify the amount of inappropriate PF4 and SRA testing performed at our institution.
2) To reduce the amount of inappropriate HIT tests via pathology driven reflexive testing algorithm in the EHR.
Methods
A 6-month retrospective review of unique patient encounters that had documented use of PF4 ELISA serologic assay or serotonin release assays (SRA) ordered during their hospital stay were queried using structured query language (SQL) of the Epic® Clarity databased and analyzed. Using an automated, algorithmic approach to chart review, an assessment of 4T score (a validated HIT risk calculator which incorporates platelet count drop, timing, thrombosis, and alternative etiologies for thrombocytopenia) for first two components was conducted to assess the appropriateness of ordering the PF4 ELISA serologic assay. Additionally, downstream ordering of the confirmatory SRA test was analyzed for appropriateness based on results of the initial PF4 ELISA. Based on these results, an EHR algorithm was created in order to reduce inappropriate testing.
Results
A total of 294 patient encounters were identified. Analytical review of 4T scoring demonstrated high variability in risk stratification and provider ordering. Specifically, the timing of platelet count fall and nadir had no association with HIT testing or empiric argatroban use In addition, providers simultaneously ordered both an SRA and PF4 ELISA 83 times (62.4% of all HIT tests) under the same laboratory draw instead of waiting to order the SRA only when the PF4 ELSA is elevated. Of the SRA tests ordered in this manner, 46 tests (55.5%) were unnecessary as no confirmatory test was required based on low PF4. An EHR change was instituted to hold order fulfillment of SRA testing prior to results of PF4 ELISA with no additional needle sticks required. A 3-month review of ordering post change implementation showed a 75% reduction in SRA testing.
Conclusion
We successfully reduced the amount of inappropriate SRA testing for HIT at our institution using an EHR algorithm. The reduction of SRA testing will directly save health system resources and may impact downstream reduction of further unnecessary testing and expensive argatroban use.
Implications for the Patient
This EHR algorithm to reduce inappropriate testing using automated reflexive ordering serves as a model pilot that can potentially be expanded to other inpatient tests.
