Transcript Using Large Observational Data to Inform Pediatric Drug Safety

Using Large Observational
Data to Inform
Pediatric Drug Safety
James Feinstein, MD MPH
Outline
• A few case examples:
1. Evaluation of existing policy
2. Evidence to inform clinical practice
3. Evidence to shape new policy
• Illustrate ways to present data to stakeholders
The problem…
The problem…
• Just like adults, kids experience side effects,
drug-drug interactions, and adverse drug
events, too!
• We lack high-quality, pediatric-specific
evidence for many medications
• We have limited mechanisms to
monitor/reduce harm
Why use big data for drug safety research?
• Pediatric population exposed to these drugs is
small (compared to adult population)
• Side effects, drug-drug interactions, adverse
drug events are rare
• Necessary to ensure sample size requisite to
make valid conclusions
1. Evaluation of Existing Policy
1. Evaluation of Existing Policy
• In 2007, the FDA implemented Tall Man
lettering
– Improve safety for look-alike or sound-alike drugs
– cloNIDine and clonazePAM
• No small task to implement
– Impacts FDA labeling, drug compendiums,
electronic medical records
• But, does it do anything to reduce error?
Methods
• Pharmacy data for pediatric inpatients (<21 years old) from
42 children’s hospitals in 2004–2012
• Pre-specified set of 8 potential LA-SA drug error patterns
• Searched within each hospitalization for the occurrence of
one of these patterns for a total of 12 LA-SA drug pairs
• Assessed for potential change in error rates before and
after Tall Man lettering implementation using segmented
regression analyses for each drug pair
Results
• 1,676,700 hospitalizations
• No statistically significant change was detected
for either the intercept or the slope of LA-SA
error rate for each of the drug pairs
• No downward trend in potential LA-SA drug error
rates was evident over any time period 2004
onwards
Trends of potential look-alike sound-alike errors before and after Tall Man lettering
implementation.
Wenjun Zhong et al. BMJ Qual Saf 2016;25:233-240
Copyright © BMJ Publishing Group Ltd and the Health Foundation. All rights reserved.
2. Evidence to Inform Clinical Practice
2. Evidence to Inform Clinical Practice
• Polypharmacy increases the potential for drug
interactions
• Direct implications for prescribers,
pharmacists, and healthcare systems
• We aimed to assess the:
– Prevalence of PDDI exposures
– Specific problematic drug combinations
Methods
• PHIS data from 2011 for patients <21 yo
• For each patient day, we analyzed every
potential 2-way drug-drug interaction
– MicroMedex DRUG REAX compendium
– Severity (contraindicated, major, moderate)
– Evidence (excellent, good, fair)
– Mechanism of harm
Results
498,956 Hospitalizations  4,497,448 PDDI Exposures
Results
MAJOR DRUG DRUG INTERACTIONS
Drug Pair
Type of Interaction
Prevalence 95% CI
Fentanyl and Morphine
Additive respiratory
depression
13.17
(13.07,
13.27)
Fentanyl and Midazolam
Additive respiratory
depression
11.19
(11.09,
11.28)
Midazolam and Morphine
Additive respiratory
depression
9.20
(9.11,
9.28)
Results
A large proportion of contraindicated and major
interactions were due to less common PDDIs
Fentanyl and Morphine
2.8%
3. Evidence to Shape New Policy
3. Evidence to Shape New Policy
• Medicare pays for Medication Therapy
Management in the adult population,
supervised by a pharmacist
• We have no similar mechanism to periodically
review medications for complex children
• How would we identify patients requiring
focused pharmacy reviews?
Methods
• Colorado Medicaid medication claims
– 242,230 children <21 yo
– 1 year continuous enrollment
• We calculated the:
– Daily maximum count over 1 year
– The durations for each level
– Patient characteristics associated with high-degree, highduration polypharmacy
• Multinomial logistic regression to identify patient
characteristics associated with high-depth high-duration
Hotspotting High Risk Patients
Who are the high-degree subjects?
High-degree polypharmacy was associated with
increasing age, male gender, and presence of a CCC
Stay tuned…
• We have a forthcoming position piece that
“prescribes” the use of big population-level
data to:
– Investigate 5 major domains of drug safety
– Prioritize pediatric drug research and funding
Questions? Ideas?
• Thank you!
The Team
Allison Kempe, MD MPH
Chris Feudtner, MD PhD MPH
Robert Valuck, PhD RPH
Dingwei Dai, PhD
Wenjun Zhong, PhD
UC Denver/CHCO
UPenn/CHOP
UC Denver/SOP
UPenn/CHOP
UPenn/CHOP