Linking Data from a Population-Based Assisted Reproductive Technology Registry to Vital
Statistics Data:
Encouraging Results from a CDC-Massachusetts Department of
Public Health Collaborative Project
NAPHSIS/NCHS Joint MeetingBruce Cohen, PhD
June 2007
Presentation Outline
• Background• Evaluation of ART reporting on the
birth certificate• Characteristics of mothers using ART• Associations between ART and
maternal and perinatal outcomes• Future direction for linkage of
fertility clinic data and birth certificates
Background
Use of ART has steadily increased over the past decade, and accounts for 1% of all live-born children in the USA
Safety concerns demand large-scale studies of adverse outcomes of ART
Whereas the ART registry maintained by CDC can be used to address certain outcomes (e.g., multiple gestation, pre-term delivery), its scope is limited (e.g., poor data on birth defects, no data on infant mortality)
Massachusetts state law mandates health insurance coverage of ART, and fertility assistance now accounts for about 3% of all live-born children in that state (based on birth certificate data)
Massachusetts has considerable experience with research based on record linkage
Objectives of the Linkage Project
I. To link information from CDC’s assisted reproductive technology (ART) surveillance data files with Massachusetts state birth records
II. To create a population-based dataset of ART mother-infant pairs including data on circumstances surrounding conception and delivery and data on maternal and infant outcomes
III. To compare infants conceived using ART and their mothers with other Massachusetts infants and their mothers
Study Population
The study population was drawn from MA birth certificate data base: infants born in 1997-2000 maternal residency listed as MA. birth occurred in MA, RI, CT, NH
N=158,229 deliveries and 161,539 total infants in 1997-1998N=158,771 deliveries and 162,260 total infants in 1999-2000
Information abstracted: demographic characteristics, prenatal care use, maternal history, pregnancy complications, delivery outcomes and complications
Study Population
Data from live births reported to ART Registry were linked to the study population if: ART treatment was in a MA or RI clinic (11 clinics total) Infant(s) born in 1997-1998, 1999-2000 Maternal residency at time of ART listed as MA or unknown.
N=2703 deliveries and 3704 total infants in 1997-1998; N=3330 deliveries and 4494 total infants in 1999-2000;
Information abstracted: patient history, ART treatment information, ART outcome information
Data Linkage Strategy
STAGE 1: link ART record to birth record using mother’s
date of birth and infant’s date of birth further evaluate potential matches using
ancillary variables -- plurality, parity and birth weight
STAGE 2: further linkage by maternal name for a portion
of records -- ART record names obtained through follow-up with ART clinics
Stage 1 Linkage Results
Linkage Results1997-1998 1999-2000
ART record matched on one birth record on MDOB and IDOB
2119 (78%) 2691 (82%)
ART record matched on more than one birth record
46 (2%) 77 (2%)
ART record did not match on any birth record
538 (20%) 562 (17%)
Total 2703 (100%) 3300 (100%)
Final Linkage Results
Linkage Results1997-1998
Linkage rate (%)
1999-2000Linkage rate (%)
ART record matched on one birth record on MDOB and IDOB
2117/2119 (100%)
2686/2691 (100%)
ART record matched on more than one birth record
33/46 (72%) 44/77 (57%)
ART record did not match on any birth record
263/538 (57%) 139/562 (25%)
Total 2413/2703 (89%)
2869/3300 (86%)
For additional details on linkage methods and 1997-1998 results:
Sunderam S, Schieve LA, Cohen B, et al.: Linking Birth and Infant Death Records With Assisted Reproductive Technology Data: Massachusetts, 1997–1998.Maternal and Child Health Journal, 2005: 9, 1 – 11
Evaluation of ART Reporting on the MA Birth Certificates: Background
Massachusetts is one of the first states that collect information on assisted reproductive technology (ART) and fertility during use via the birth certificate, started in 1996
ART information was extracted by hospital staff from medical records, recorded on the birth certificates, then sent to the state Registry of Vital Records as part of the routine data transmission.
The completeness and quality of reporting has not been systematically evaluated, thus limiting the potential of use in our surveillance and research work related to the assisted reproductive technology.
Research Questions
How consistent is the ART information reported on the Massachusetts birth certificate with ART reported in ART Registry?
What are the predictors of checking off the ART variable on the Massachusetts birth certificate ?
Methods
Data sources: live-birth deliveries to MA resident mothers occurring in 1997-2000 in MA, NH, RI, and CT
Unit of analysis: live-birth deliveries Use linked SART-BC records as “gold
standard” for “true” ART births Validity measures : sensitivity and
specificity
Results – Overall Sensitivity and Specificity
0
20
40
60
80
100
Overall
Pe
rce
nt
(%)
Sensitivity Spec ific ity
Concordance of ART Information between the Massachusetts Birth Certificate and the SART
Database, 1997-2000
Birth Certificate Records
ART
No ART
Linked with
SART
(a)
Not Linked
(b)
Linked with
SART
(c)
Not Linked
(d)
Sensitivity
(e)
= a /
(a + c)
Specificity
(f)
= d /
(b + d)
PPV*
(g)
= a /
(a + b)
1997 315 216 772 76,077 29.0% 99.7% 59.3%
1998 418 270 862 76,671 32.7% 99.6% 60.8%
1999 364 237 1,013 76,176 26.4% 99.7% 60.6%
2000 312 246 1,134 76,871 21.6% 99.7% 55.9%
Total 1,409 969 3,781 305,795 27.1% 99.7% 59.3% * positive predictive value
Sensitivity of Reporting ART on Birth Certificate by Plurality
0
10
20
30
40
50
Singletons Multiples
Pe
rce
nt
(%)
Sensitivity of Reporting ART on Birth Certificate by Kotelchuck Index
0
10
20
30
40
50
Inadequate Intermediate Adequate Adequate +
Pe
rce
nt
(%)
Variation in Sensitivity by Hospital Characteristics
Range of sensitivity by facility : 0-60+ %Teaching hospitals: sensitivity > 60% in 3
out of 7Non-teaching hospitals: sensitivity > 60%
in 2 out of 49Variation also observed by number of ART
deliveries/year and hospital level
Summary/Comments
While specificity of reporting ART on the birth certificates is high, the sensitivity is low
Results probably reflect experience of early stage of collecting ART information on the birth certificate in Massachusetts
Definitional differences could affect linkage rates: birth certificate response category includes non-invasive procedures--artificial insemination (AI) or intrauterine insemination (IUI) whereas CDC ART definition is “all treatments or procedures that include the handling of human oocytes and sperm for the purpose of establishing a pregnancy. This includes, but is not limited to in vitro fertilization and transcervical embryo transfer, gamete intrafallopian transfer, zygote intrafallopian transfer, tubal embryo transfer, embryo cryopreservation, oocyte or embryo donation, and gestational surrogacy. ART does not include assisted insemination using sperm from either a woman's partner or sperm donor.
Summary/Comments
Consider alternative reporting methods for ART on the birth certificate (PRAMS pilot testing indicated much higher response from MA mothers than recorded on birth certificate)
What are the implications of these results for the new revised certificate items? For research using these items?
Comparison Groups
The following slides use three comparison groups:1. ART Births/Deliveries: Massachusetts
births/deliveries linked to the CDC ART Registry (1.7%)
2. FA on BC: Evidence of fertility assistance on the birth certificate but not linked to the CDC ART registry (0.6%)
3. NO FA on BC: All other births or deliveries (97.8%)
Percentage of Deliveries, Age ≥35, Massachusetts 1997-2000
0%
10%
20%
30%
40%
50%
60%
ART Deliveries FA on BC No FA on BC
Percentage of Deliveries that are Multiple Births, Massachusetts 1997-
2000
0%
5%
10%
15%
20%
25%
30%
35%
ART Deliveries FA on BC No FA on BC
Percentage of Deliveries that are first Deliveries, Massachusetts 1997-2000
0%
10%
20%
30%
40%
50%
60%
70%
80%
ART Deliveries FA on BC No FA on BC
Percentage of Non-Hispanic White Mothers, Massachusetts 1997-2000
0%10%20%30%40%50%60%70%80%90%
100%
ART Deliveries FA on BC No FA on BC
Percentage of Mothers with College Education, Massachusetts 1997-2000
0%
10%
20%
30%
40%
50%
60%
70%
ART Deliveries FA on BC No FA on BC
Percentage of Mothers Privately Insured, Massachusetts 1997-2000
0%10%20%30%40%50%60%70%80%90%
100%
ART Deliveries FA on BC No FA on BC
Percentage of Deliveries in Boston Hospitals, Massachusetts 1997-2000
0%
10%
20%
30%
40%
50%
ART Deliveries FA on BC No FA on BC
Percentage of Deliveries with Kotelchuck Index = Adequate +, Massachusetts 1997-
2000
0%
10%
20%
30%
40%
50%
60%
ART Deliveries FA on BC No FA on BC
Percentage of C-Section Deliveries, Massachusetts 1997-2000
0%
10%
20%
30%
40%
50%
ART Deliveries FA on BC No FA on BC
Percentage of Low Birthweight Infants, Massachusetts 1997-2000
0%
5%
10%
15%
20%
25%
30%
35%
ART Deliveries FA on BC No FA on BC
Percentage of Multiple Deliveries where Plurality = 3+, Massachusetts 1997-2000
0%2%4%6%8%
10%12%14%16%18%20%
ART Deliveries FA on BC No FA on BC
?
Summary
These linked data provide the first population-based assessment of the prevalence and characteristics of ART mothers in the US
Women who use other methods of fertility assistance are more similar to those who use ART than the general population, with some notable differences
These unadjusted data provide a preliminary, baseline snapshot and need to be examined more closely to fully understand the implications for public health practice
Associations Between Assisted Reproductive
Technology and Maternal and Perinatal Outcomes
Schieve LA, Cohen BB, Naninni A, Ferre C, Reynolds, MA, Zhang Z, Macaluso, M, and Wright V. “A Population-Based Study
of Maternal and Perinatal Outcomes Associated with Assisted Reproductive
Technology in Massachusetts.” Maternal Child Health J. March 8, 2007.
Summary
After considering numerous potential confounding factors through sample restriction and matching, ART was associated with:o Pre-existing diabeteso Incompetent cervixo Pregnancy induced hypertensiono Uterine bleedingo Placental abruptiono Placenta previao Preterm deliveryo Very preterm deliveryo Low birth weighto Infant not discharged home
Although confidence intervals overlapped 1.0, there was also a suggestion of increased risk among ART births for:o Gestational diabeteso Very low birth weight
Summary
Findings of associations with pre-existing diabetes and incompetent cervix are supported by the literature. Both are conditions associated with infertility disorders treated with ART.
Findings of associations with pregnancy-induced hypertension, uterine bleeding placenta abruption, placenta previa (and marginal association with gestational diabetes) confirm and expand on previous studies by more fully considering potential confounders.
Negative finding for cesarean section contradicts several previous studies reporting association.
o While C-section rates were high in women who conceived via ART in comparison to the general population of births in MA, ART does not appear to be the reason.
o C-section rates for ART births were comparable to those observed in women who did not conceive with ART after matching on birth place and time, age, parity, and race/ethnicity.
Findings of associations with preterm and low birth weight also support previous studies.
o Supplemental analysis in a subgroup without maternal health or labor and delivery complications suggests the increase in perinatal risk among ART births is not solely explained by maternal health factors.
Future Directions
Update linkages for currently available data
Continue research using these data: extend etiologic and descriptive analyses
Explore linkages with other data sources such as hospital discharge data bases, birth defects registries…perhaps through PELL
Explore Linking CDC- ART Data with Pregnancy to Early Life Longitudinal Study (PELL)
Early I ntervention
Linked birth- infant deaths
Child and Mother deaths
Birth Defects Registry
WI C (2007)
Birth Certificate
(HD) Birth Mothers
Fetal Death (HD) Birth Child
Emergency DepartmentCensus 2000 Data (2007)
CORE
Other MDPH programs (future)
Geocoded birth data (2007)Observational Stays
CDC- ART data
Contextual Data
Pregnancy- associated deaths
Non- birth Hospital Discharge
Newborn Hearing Screening (2007)
Vital and Health Status DataProgram Participation Data
Health Services Utilization Data
Other Contextual Data (future)
Future Directions
Investigate limitations of linkage-based analyses (sensitivity analyses, potential for information/detection bias)
Explore the feasibility of augmented studies (nested follow-up, case-control studies)
Explore application of the linkage methods to other states with high ART prevalence