AM Spehar, Patient Safety Center, JAHVA; NA Coblio, JAHVA; MH Walton, VISN 8; RR Campbell, Patient Safety Center, JAHVA; J Wolfson, College of Public Health, University of South Florida

Objectives: Health care managers and researchers need rapid access to critical data to measure outcomes, improve performance, and increase patient safety. Achieving these goals can be impeded by slow and limited data files, lack of cross-linking to other files permitting outcome evaluation, and lack validation for meaningful research. This study is an evaluation of the usefulness of a pharmacy data mart for management decision-making, monitoring outcomes of clinical interventions, and identifying areas of increased cost-effectiveness while protecting patient safety.

Methods: A data mart, a relational data base on a SQL server with a 2 terabyte design size, was constructed for the pharmacy department of a large VA Hospital. Our data mart, a "self service" collection of data extracted from the healthcare organization’s operational data, currently holds over 2.5 million outpatient prescription records for over 63,000 patients, over 400,000 laboratory records for patients on select drugs, as well as ICD-9 and demographic data. Data scrubbing and validation is routinely performed allowing salvage of incomplete or incorrect data.

Specific questions asked in our evaluation included: 1) Was patient compliance significantly affected by prescribing lansoprazole at reduced dosage twice-a-day instead of once-a-day? 2) What areas of pharmacy offer the greatest opportunity for cost savings without adversely affecting patient health and safety? 3) Does tablet splitting of simvastatin, which could result in substantial cost savings, change patient compliance and clinical outcomes? 4) Will non-insulin dependent diabetics on oral hypoglycemic agents be adversely affected by reducing the number of glucose strips dispensed from 3/day to 1/day (still exceeding recommendations)?

Results: Using the data mart, a rapid (<5 minutes) pareto analysis showed that actual b.i.d. dosing of lansoprazole was minimal, despite clinical impressions to the contrary. Thus, an expensive and time-consuming Medication Utilization Review (MUE) was avoided, allowing managers to focus on real improvement areas. Formulary managers, using the data mart, performed a pareto analysis to identify areas of greatest cost and maximum potential cost-benefit to the VISN. Two areas were targeted for intervention, the use of split tablets of simvastatin, and reduction in the use of blood glucose strips by patients on oral anti-diabetic agents. By linking laboratory with pharmacy data, patient compliance and clinical outcomes were evaluated by monitoring average LDL for all patients on simvastatin, after 70% were trained in tablet-splitting. Patients splitting tablets had decreased average LDL levels, while controls showed no significant change. Thus patient outcomes were improved at significant cost savings. Similarly, hemoglobin A1c levels, an indicator of metabolic control in diabetes, did not change after reducing glucose strips dispensed.

Conclusions: Using the data mart, clinicians and managers were able to rapidly answer questions about patient outcomes and cost-effectiveness of pharmaceutical drug usage. Currently, we are expanding beyond outpatient data, and evaluating the data mart in other studies, such as correlating pharmacological markers of drug reactions with diagnostic entry codes to identify trends in medication error reporting.

Impact: Using the data mart to rapidly identify and target potentially cost-effective interventions resulted in facility savings of over $3 million and more than $20 for the VISN. With the data mart we documented greater cost-benefit with improved patient outcomes.