The Effect of Using Rules Technology with Computerized Provider Order Entry on Medication Errors in an Outpatient Setting
Andy Steele, MD, MPH June 6th, 2004
Co-Investigators
Sheri Eisert, Ph.D
– Denver Health
Eduardo Ortiz, M.D., M.P.H.
– Washington DC VA (Previously AHRQ)
Patricia Gabow, M.D.
– Denver Health
Joel Witter, MD
– Denver Health
Pat Lyons
– Siemens Medical Solutions USA, Inc.
Mike Jones PharmD
– Micromedex
Research Objective
Assess the impact of rules technology on medication errors related to druglaboratory interactions in an outpatient primary care setting
Background
700,000 people die or are injured by adverse drug events (ADE) >50% of preventable ADE’s associated with prescribing ~ 25% of outpatients have an ADE
Inpatient POE systems can decrease errors by ~80%
Background
Medication Errors
– 45% related to drug-laboratory issues – 29% related to drug-drug interactions
Information Overload
– For top 40 prescribed medication, each one has an average of almost seven drug-laboratory interactions
Monitoring of relevant laboratory values is inadequate
– Troglitazone-after four FDA warnings, liver function testing compliance was less than five percent at 3 months
Study Design
Non-randomized, non-blinded, prospective intervention Drug-laboratory interaction rules for:
-Drug-induced Hypokalemia -Drug-induced Hyperkalemia -Drug-induced Thrombocytopenia -Drug-induced Nephrotoxicity -Drug-induced Hepatotoxicity
Alerts were displayed to providers for above interactions
– “No lab” in last six months – “Lab abnormal”
Evaluation for 4 months pre- and 5 months post-intervention Outcome Measures:
– # orders not completed – # rule-associated laboratory test ordered after alert
Population Studied
Four primary care clinics (~80,000 annual visits) All orders (~6,000/week) are entered on an Ambulatory Provider Order Entry (ACPOE) All registered patients were eligible for the intervention
Principal Findings
54,206 patient visits 17,444 (32%) visits: medications were ordered Rule processed 16,291 times(~49% of med orders)
– 7,017 during the pre-intervention period – 9,274 during the post-intervention period
Principle Findings
4% 13% 42%
4% 3% 11%
41%
82%
African American Caucasian Hispanic Other
Medicaid Uninsured Medicare/Private Other
Alert Output for Drug-Lab Interaction Rules
6% 6%
No Alert Alert: "Abnormal Labs" Alert: "No Labs"
88%
Order Cessation Rate in Response to an “Abnormal Lab” Alert
12% 10% 8% 6% 4% 2% 0% Pre Post 5.6% 10.9%
P value=0.03
Recommended Laboratory Test Ordering Rate in Response to an Alert
100% 80% 60% 40% 20% 0% Pre Post 39.0% 51.0%
P value<.001
Effect on “Definite & Probable Adverse Drug Events
(n=173, Naranjo Scoring)
20% 15% 10.3% 10% 5% 0% Pre Post 4.3%
P value=0.023
Limitations
Intervention focused on a select group of druglaboratory interactions Setting was in a safety-net primary care clinic Unique patient population
– lower to middle income – minority dominated-80% Hispanic – medically underserved population
Intervention was not randomized Short-term study (5 months)
Conclusions
Providers will change ordering behavior in response to computer-based alerts
Although this may lead to less medication errors it is uncertain if adverse drug events will decrease
Implications for Policy, Delivery or Practice
Computerized clinical decision support systems have potential for decreasing errors and improving many aspects of patient safety and quality of care
Implications for Policy, Delivery or Practice
Computerized clinical decision support systems have potential for decreasing errors and improving many aspects of patient safety and quality of care
Contact Information Andy Steele, MD, MPH Director, Medical Informatics Denver Health (1932) 660 Bannock St. Denver, CO 80218 Email: asteele@dhha.org
This study was supported by the Agency for Healthcare Research and Quality (AHRQ) Contract No. 290-00-0014, Task Order No. 3
8. Was the reaction more severe when the dose was increased, or less severe when the dose was decreased?
+1
9. Did the patient have a similar reaction to the same or similar drugs in any previous exposure?
+1
10. Was the adverse event confirmed by objective evidence?
+1
Total
Total score determines probability category: Definite>=9, probable, 8-5, possible 1-4, doubtful<=0. Naranjo et al, Clinical Pharmacology and Therapeutics, ADR Probability Scale: Drug Withdrawal Table, August 1981;30-239-245.
Other Questions (some removed since covered above-1,5)
Provider
Medication Order Entered
Trigger Rules to evaluate laboratory values
If appropriate laboratory value is not available, or if appropriate laboratory value is abnormal, an alert message is provided to the provider, suggesting that they should consider changing the medication order, ordering a laboratory test, or monitoring patient more closely. Provider then continues order entry session.
Rules Engine Registration
Laboratory values pulled from clinical repository
LCR/EAD Clinical Repository
Rule Example: Arden Syntax
Effect on Adverse Drug Events
(Naranjo Scoring)
120 100 80 60 40
10.3%
Possible & Doubtful
104
Definite & Probable
20 0 12 Pre
P=0.23
45 2 Post
4.3%