Psychology in Primary Care An Evaluation of Best Practices BASU
Evaluation of Gestational Length Algorithm
1. Evaluation of the Validity of a Gestational
Length Algorithm Based upon Electronic Health
Plan Data
Qian Li, Susan. E. Andrade, William O. Cooper, Robert L.
Davis, Sascha Dublin, Tarek A. Hammad, Pamala. A.
Pawloski, Simone P. Pinheiro, Marsha A. Raebel, Pamela
E. Scott, David. H. Smith, Inna Dashevsky, Katie
Haffenreffer, Karin E. Johnson, Darren Toh
18th Annual HMO Research Network Conference, Seattle
WA
May 2, 2012
2. Funding Source & Conflict of Interest
Contracts
HHSF223200510012C, HHSF223200510009C, an
d HHSF223200510008C from the U.S. FDA
Dr. Dublin funded by Paul Beeson Career
Development Award from the National Institute on
Aging, grant K23AG028954, and by Group Health
Research Institute internal funds
Abstract not necessarily represent official views or
endorsement of the FDA or the National Institute
on Aging or the NIH
None of the other authors have conflict of interest
3. Background
Medication effects often specific to
particular gestational period
Electronic health plan databases are
increasingly used in pregnancy research
Valid prenatal exposure status
Pharmacy dispensing data
Pregnancy beginning & gestational length
Computerized algorithm (delivery date +
preterm birth ICD-9-CM)
4. Objectives
To examine the validity of a common
algorithm by comparing
algorithm-derived gestational length &
prevalence of medication exposures during
pregnancy
“gold standard” measures in birth certificates
5. Data Source
Medication Exposure in Pregnancy Risk Evaluation
Program (MEPREP)
- U.S. Food and Drug Administration
- HMO Research Network (8 health plans)
- Kaiser Permanente California
- Vanderbilt School of Medicine/Tenn Medicaid
MEPREP
- Enrollment
- Socio-demographic (race/ethnicity)
- Demographics
- Medical
- Outpatient pharmacy dispensing
- Reproductive (parity, gestational age)
- Outpatient and inpatient encounter
Administrative and Birth
Claims Certificate
6. Study Population
Live born deliveries among women aged
15-45 years between Jan 1, 2001 and Dec
31, 2007
Availability of valid gestational length in
linked birth certificate
Continuous enrollment and pharmacy
benefit, 100 days before pregnancy
through delivery
7. Gestational Length Algorithm based on Health Plan Data
ICD-9-CM code Definition Algorithm-derived gestational length
Weeks Days
765.21 Less than 24 completed weeks of gestation
24 168
765.22 24 weeks of gestation
765.23 25-26 weeks of gestation 26 182
765.24 27-28 weeks of gestation
28 196
765.0-765.09 Extreme immaturity
765.25 29-30 weeks of gestation 30 210
765.26 31-32 weeks of gestation 32 224
765.27 33-34 weeks of gestation 34 238
765.28 35-36 weeks of gestation 36 252
765.1-765.19 Other preterm infants
765.20 Preterm with unspecified weeks of gestation
35 245
644.21 Onset of delivery before 37 completed weeks of
gestation
Gestational length for deliveries without an ICD-9-CM code for preterm birth in the table was assumed to be
270 days.
8. “Gold Standard” Gestational Length
Birth certificate
last menstrual period (LMP)
clinical estimate (CE) / obstetric estimate (OE)
CDC’s National Center for Health
Statistics (NCHS) approach
LMP primarily
CE/OE, when LMP not
available
20-45 weeks (adapted from NCHS)
compatible with birth weight
9. Medication Exposure
Dispensing dates + days supplied ; 14-day grace
period
Long term
Short term
Chronic
Acute use
basis
Antidepressant
Antibiotics
s
Amoxicillin
Fluoxetine
Azithromyci
Sertraline
n
10. Statistical Analysis
Mean, range, proportion of term/preterm
deliveries
Deliveries with two gestational lengths
differ within 0, ±1, ±2, ±3, ±4, or greater
than ±4 weeks (stratified by plurality)
Prenatal medication exposure
Sensitivity, specificity, PPV, NPV
Any time in pregnancy or by trimester
Stratified by term/preterm determined by the
algorithm
11. Study Results
Infants’ birth certificate files linked to
health plan data in 92% deliveries
Gestational age missing/invalid in linked
birth certificates in 0.4% deliveries
Final study population included 225,384
deliveries
12.
13.
14.
15.
16.
17. Discussion
Algorithm underestimated gestational length
by average 5.5 days
Restricted to singleton deliveries (86% term)
270-day upper bound
Not in multiple-gestation deliveries (36% term)
ICD-9-CM codes for preterm births
Algorithm underestimated prevalence of
preterm deliveries
15% in study population > 12% nationally
More women aged >35 years (21% vs. 14%)
18. Discussion
Algorithm correctly classified the
antidepressants and antibiotics exposure
status in most women
Specificity and NPV close to 100%
Poorer sensitivity and PPV for antibiotics
(sporadic) vs. antidepressants (chronic)
Overestimate on antibiotics due to 14-day
grace period for dispensings
19. Discussion
Strengths
Study population geographically and
demographically diverse, increasing
generalizability
Reasonable gold standard of gestational length
for majority of study population
Limitations
Only evaluated 1 algorithm
Only evaluated 2 antidepressants and 2
antibiotics, unknown for other medications
Medication dispensed =?= medication use
20. Conclusion
Gestational length algorithm based on
health plan data (delivery date + preterm
birth diagnosis) classified prenatal
medication use well
Performance slightly poorer for short-term
drugs (e.g. antibiotic)