Estimating the ROI for Computerized Clinical Decision Support Systems: Pieces of the Puzzle

1. Estimating the ROI for Computerized Clinical Decision Support Systems:Pieces of the Puzzle Terry Field, D.Sc. Meyers Primary Care InstituteUniversity of Massachusetts Medical School, Fallon Community Health Plan, Fallon Clinic

2. Pieces of the PuzzleAdding CDSS to Existing EHR • Development and implementation costs • Immediate, direct costs and savings • Potential additional savings

3. Development CostExample 1 • Long-term care setting • CDSS to provide prescribers with patient-specific maximum dosing recommendations based on renal function • Added to a commercial EHR with integrated CPOE (Meditech) • Included 62 drugs; 94 alerts specific to the level of renal insufficiency

4. Tracking Personnel Time & Costs • Internal physicians, pharmacists, informatics project manager, project coordinator, health services researcherweekly reports from each participant with hours by category • Externalspecialized programmertracked through bills submitted

5. Cost Analysis • Reported hours combined with US national average hourly wages for the appropriate personnel categories • Submitted bills from external programmer

6. Results – Personnel Time & Costs

7. Results – Activities & Costs

8. Alternative Scenarios • CPOE system does not require specialized programmerHours: 924, Cost: $43,268 • Database for renal dosing existsHours: 657, Cost: $34,201 • CDSS Product existsHours: 475, Cost: $23,695

9. Development CostExample 2 • Ambulatory setting – large group practice • Automated alert system to provide PCPs with:- notification of hospital and SNF discharges- new drugs added during hospital stay- recommendations related to dosing and monitoring - reminders to support staff to schedule follow- up visit • Added to a commercial EHR with CPOE (EpicCare Ambulatory EMR)

10. Results – Personnel Time & Costs

11. Results – Activities & Costs

12. Special Issues in Development Costs • Substantial time required from clinical personnel!- determining contents (or reviewing if purchased)- extensive time spent testing

13. Potential Costs & SavingsImmediate, Direct Impact • Long-term care setting • Within an RCT of the renal dosing CDSS described earlier • Randomized by unit within a large long-term care facility • Costs and savings related to drugs and laboratory tests

14. Tracking Costs & Savings • Drugs that triggered an alert as prescriber began the order vs. drugs actually ordered • All drug orders for the day of an alert reviewed to identify potential substitutes • Drug costs based on US wholesale price at the time • Serum creatinine tests ordered within 24 hrs of alert of missing lab information – costs based on Medicare allowable payments at the time of the order

15. Analysis • Within both intervention and control units, we compared costs for initial vs. final submitted drug orders • Adjusted findings from the intervention units by findings in the control units Note: even in the control units, prescribers changed their minds during an order!

16. Results • Estimated savings for drug orders: $2,160 • Estimated additional costs for lab orders: $769 • Total estimated savings: $1,391

17. Potential Additional Savings:Reduced ADEs • Setting: large, multispecialty group practice providing care to >30,000 Medicare enrollees • Case-control study nested in a cohort study that identified adverse drug events from 7/1/1999 to 6/30/2000 • Control group – for each subject with an event, we randomly selected a control matched by having an encounter and dispensing in the month prior to the event

18. Determination of Costs • Outcome measure: costs of medical care from 6 weeks prior to the event through 6 weeks after • In-patient stays, ED visits – national average of cost-to-charge ratios • MD visits, dx tests, therapy, lab, ambulance use, home health, DME – Medicare fee schedules • Pharmaceuticals – average wholesale cost on day dispensed

19. Analysis • Average total costs for cases and controls calculated and plotted • Estimated surge in costs calculated by subtracting pre-event costs from post-event costs for each individual • MVA with cost surge as outcome and case status as exposure, controlling for confounders • Analyzed for 1225 case/control pairs and 325 pairs for preventable ADEs

20. Results

21. Results *Controlling for age, gender, Charlsoncomorbidity index, # scheduled meds, hospitalization in pre-period

22. Extrapolation • 1,000 enrollees age 65+ for 1 year13.8 preventable ADEs$27,365 (CI $2,663, $52,067) in 2000 dollars • All Medicare enrollees age 65+ in 2000$887 million for preventable adverse drug events

23. Summary • Development costs are significant • Development (or even implementation) requires extensive time from clinicians • Immediate, direct cost savings may be minor • Savings from reductions in adverse events are likely to be substantial • Complete, detailed tracking of adverse events and their associated costs is a large and expensive task!