AAN Poster - Sponsored Whitepaper

Deviations from physiologic targets in this passively acquired hospital data repository matched general expectations from prior prospective studies and permitted efficient quality performance surveillance. Aggregation and mining of automatically acquired electronic patient data is a new and powerful technique of performance improvement in acute stroke care.

We analyzed 809 consecutive ischemic stroke and TIA patients admitted to a Medical Center from 01/2005 to 03/2009. An HL7 database was interrogated that pooled in a single repository data arising from multiple sources within the health enterprise, including transduced physiologic parameters, nursing-entered vital signs, bedside point of care blood tests, central laboratory results, radiology reports, financial transactions, and the GWTG-S database. This dataset was mined to perform surveillance for deviations from target ranges for key physiologic variables, including glucose levels, temperature, and systolic blood pressure. In addition, each patient was cross-referenced to their location in the hospital for the first seven days after stroke onset.

Objective: To evaluate the feasibility of data mining as a component of performance improvement in acute stroke care.

Background/Purpose: Quality improvement programs in stroke have focused on review of individual cases and laborious prospective data abstraction of select variables. Modern health enterprises accumulate vast amounts of physiologic and clinical data in diverse computer systems in the course of routine care. Data mining permits this data to be mobilized and interrogated.

Among 809 acute cerebral ischemia patients, median admission NIHSS was 7, range 0-40. During the first 72 hours post-onset, rates of physiologic deviation from target for glucose (target < 200) were highest on post-onset day 1 and declined steadily: day 1 – 15.0%, day 2 – 12.2%, day 3 – 7.4%. Rates of physiologic deviation from target for temperature (target < 38.3) were high on day 1 (6.3%), and equally lower on days 2 and 3 (4.7% and 4.3%). Deviations from target SBP (>220) were extremely infrequent on all days (<1%). During days 4-7, the proportion of glucose and temperature values outside target rose mildly, likely reflecting early discharge of healthier patients.