вход по аккаунту


Using Birth Certificate Data for Research on Fertility

код для вставкиСкачать
Using Birth Certificate Data for
Research on Fertility Assistance:
the Massachusetts Experience
Bruce B. Cohen, Ph.D.
Co-director, Center for Health Information and Statistics
Massachusetts Department of Public Health
NAPHSIS Conference
June 2004
пЃє Massachusetts has the third highest number of
Assisted Reproductive Techonology procedures
(ART) performed in US and highest rate (MMWR,
пЃє Massachusetts has the highest multiple birth rate in
пЃє On the MA certificate of live birth, there are check
boxes for use of ART and fertility drugs (FD)
пЃє Massachusetts provides full access to confidential
data for perinatal research purposes
Two projects underway
пЃєJoint research project with the CDC Division
of Reproductive Health to link birth
certificate data with all Massachusetts
fertility clinic data to create a populationbased cohort to examine risks and outcomes
of ART use
пЃєJoint research project with NICHD and
UMass Amherst to examine the quality of
birth certificate data on FD/ART
Objectives of CDC Project
To link information from CDC’s assisted
reproductive technology (ART) surveillance
data files with Massachusetts state birth
II. To create the first US comprehensive,
population-based dataset of ART motherinfant pairs which will include data on:
circumstances surrounding conception and
maternal and infant outcomes
Advantages of linked data file
Linked dataset would allow for evaluation of ART
mother-infant pairs in context of a US
 ability to analyze various maternal and infant
outcomes, including infant mortality
 proper control for confounding factors
 comparison group of non-ART births from MA birth
 potential for comparison group of births conceived
with infertility drugs and no ART
 potential for further file linkage and trend analyses
 Receive permission for project after CDC and MDPH
review: July 2001
 Develop, test and implement computer linkage
algorithms: summer 2001- spring 2002
 Contact clinics for information to increase linkage rate
and validate algorithm: fall 2001-summer 2002
 Create and finalize linked file for 1997-1998 births:
winter, 2002
 Develop detailed analytic plans: winter 2002-fall, 2003
 Began analysis: Winter 2004
 Completing linkage of 1999-2000 data: Summer 2004
Data sources
пЃ¶ ART Registry data base: patient history, ART
treatment information, ART outcome information
пЃ¶ MA birth file: demographic characteristics, prenatal
care use, maternal history, pregnancy complications,
delivery outcomes and complications
 ART clinics – name data: increased linkage and
validation of computer algorithm using secondary
Methods – Study Population
The study population was drawn from MA birth
certificate data base:
пѓ� infants born in 1997 and 1998
пѓ� maternal residency listed as MA.
пѓ� birth occurred in MA, RI, CT, NH
N=161,539 total infants;
158,229 deliveries
Methods – Study Population
Data from live births reported to ART Registry
were linked to the study population if:
пѓ� ART treatment was in a MA or RI clinic (11 clinics
пѓ� Infant(s) born in 1997 or 1998
пѓ� Maternal residency at time of ART listed as MA or
unknown (because a large proportion , 45%, were
missing residency data).
N=3704 total infants;
2703 deliveries
Methods – Study Population
пЃ±For purposes of linkage, both data files -MA birth certificate data base and ART
Registry -- were organized as live-birth
deliveries (i.e. infant-sets)
пЃ±Upon completion of linkage, the final
ART-MA linked dataset was reorganized
to include each infant as a unique
How were states chosen for inclusion
in target study population?
Aside from MA, only 3 states reported a relevant number
of births to MA resident mothers in 1997-98
пѓ� Rhode Island (N=2083)
пѓ� New Hampshire (N=429)
пѓ� Connecticut (N=191)
пѓ� No other states made a significant contribution
Together, MA and these additional three states accounted
for 99.7% of all births to MA resident mothers.
Data Linkage Strategy
пѓ� link ART record to birth record using mothers
date of birth and infant’s date of birth
пѓ� further evaluate records that link using other
variables in common to the data sets -- plurality,
parity and birth weight
� further linkage by maternal name – name data for
a portion of records obtained through follow-up
with ART clinics
Data linkage strategy, Stage 1 – expectation
of success using two dates of birth variables
пЃ¶ Review of 1997-1998 MA BC data showed that:
пѓј 97% of singleton births had unique MDOB/IDOB
пѓј 99.8% of multiple births had unique combinations
пЃ¶ Thus, expected duplication rate was 1-3% during
Stage 1 linkage
Stage 1 Linkage Results by Plurality
(from ART record)
2119 (78%)
ART record matched on more
than one birth record
46 (2%)
ART record did not match on
any birth record
538 (20%)
2703 (100%)
Linkage Results
ART record matched on one
birth record on MDOB and
Evaluation of Records that Matched* in Stage 1:
Concordance Between Data Sources on Parity and Birth Weight
Concordant on neither
Concordant on one variable
Concordant on both variables
* Matched defined as ART record matched on MDOB/IDOB with only one birth record
Stage 2 Linkage, Selection of Records
The following were initially selected for inclusion in
Stage 2:
пѓ� All records that were unmatched after Stage 1
пѓ� All records that matched with more than one birth record in
Stage 1
пѓ� All records that matched with one birth record but were
discordant on plurality
пѓ� All records that matched with one birth record but were
discordant on both parity and birth weight
пѓ� A 3% random sample of records that matched with one birth
record and were concordant on plurality and either parity or
birth weight, or both parity and birth weight
Methods: Stage 2
Algorithm for name data linkage among ART
records that were unmatched after stage 1
пѓ� 1. Search for exact mother name match (first
and last)
пѓ�If yes, check to ensure 5 of 6 DOB variables*
пѓ�If yes, consider record matched
пѓ�If no, put record back in pool for inclusion
in #2
* The 6 DOB variables are mother’s birth day, birth month, and
birth year and infant’s birth day, birth month, and birth year
Methods: Stage 2
Algorithm for name data linkage among ART
records that were unmatched after stage 1
пѓ� 2. Search for all BC records that matched ART record on 5 out of 6
DOB variables
пѓ� Examine 5/6 matches manually for concordance on name
пѓ� If clear name match identified, consider record matched
пѓ� If no clear name match, search for all BC records that
matched ART record on 4 out of 6 DOB variables and
examine all 4/6 matches manually for concordance on
пѓ� If no clear name match after 4/6 record search,
search for BC records that matched ART record on
all 3 MDOB or all 3 IDOB variables and examine
matches manually for concordance on name
Methods: Stage 2
Algorithm for name data linkage among ART
records that were unmatched after stage 1
For all potential name matches with less than 5/6 concordance on
DOB variables, a more extensive review was undertaken in
deciding whether the records matched.
In addition to DOB variables and mother’s first and last name, the
following were examined:
parity, birth weight and plurality concordance between datasets
mother’s state of residence in ART dataset – MA or unknown
whether ART use and/or fertility drug use listed on BC
The Steering Committee examined all data and rendered a group
decision on whether a record matched. (In nearly all cases, the
decision was unanimous)
Maiden name review and results
 Initially, name data consisted only of mother’s first
and last name. Maiden names were subsequently
obtained from BC dataset to check for cases of:
• Possible mis-recordings of last and maiden names
• Marriage between ART treatment and birth
пЃ¶ The name algorithm was repeated for all records that
remained unmatched after stages 1-3. Maiden name
was substituted for last name
пЃ¶ Only 1 additional match was found
Final match rate by clinic size
# of live –
eligible for
# of records
after Stages
Final percent (range
across clinics) of
84% (62-100%)
(100-250 ART live-birth
89% (76-96%)
(250-999 ART live-birth
90% (87-94%)
ART clinic size
(1<100 ART live-birth
Final match rate by residency status
Total eligible live-birth deliveries
Final match %
% that remained unmatched
ART record
lists patient as MA
residents at time
ART record -Residency
Possible reasons for non-match
1) Patient received treatment in MA but actually lived in another
state. This is likely to have been the case with patients who
reported state of residency as unknown.
2) Patient was MA resident at the time of ART treatment but
moved before delivery –i.e. migration.
3) Misclassification/data entry error of state of residency by
clinics – default entry of MA.
4) Residency may have been recorded as MA for some couples who
worked in MA and received MA benefits but lived in a
surrounding state.
5) For some records, name data were not available from clinics, and
therefore linkage status could not be verified or remained
unknown (I.e. true matches that could not be reconciled).
Sensitivity Analyses of Linkage Rate
пЃ¶ Final match rate: 2413/2703 = 89.3%
пЃ¶ If we assume:
• Residency unknown – 87.75% MA residents; 12.25% non-residents
• Migration rate between ART and birth was 2.5% for MA residents
 Projected “true” match rate is changed to 2413/2614 = 92.3%
(may be higher based on assumptions of about migration of
residents between receiving fertility treatments and time of
 Other issues such as data entry error on state of residence –
i.e. by default mis-coding residency as MA for non-residents –
might have resulted in additional sources of over-reporting in
the linkage denominator. This even the projected 92% match
rate might be conservative.
Future Plans
пЃ¶Data analysis of specific topics using linked
data file.
пЃ¶Expand project to include more recent years
пЃ¶Explore linkages with other data sources such
as hospital discharge data base
Objectives of Joint NICHD Project
пЃєTo evaluate the quality of the FD/ART data
provided on the hospital work sheet
пЃєTo review hospital procedures for
ascertainment, completion, and submission of
FD/ART variables to the RVRS
пЃєTo evaluate: 1) if these variables provide a
useful frame work to pursue FD/ART
research; 2) if not, recommend how to collect
this information to improve quality of the
NICHD Project--Study 1, Methods
пЃєidentify cohort of births for whom use of
ART/FD was indicated on birth certificate
пЃєidentify cohort of births where FD/ART use
likely, but not reported (higher order multiple
births to older women)
пЃєdevelop sampling strategy for ART only, FD
only, both; stratify by hospital
пЃєdevelop survey instrument to collect
information on fertility treatment
пЃєperform survey
NICHD Project--Study 1:
Timeline and Issues
пЃєCurrently in final stages of MDPH approval of
пЃєSurvey instrument developed; CATI
пЃєDesigning sample
пЃєField work scheduled: Summer 2004-Winter
пЃєAnalysis: early 2005
пЃєData collected from providers on hospital
work sheet vs. from parents in survey
NICHD Project--Study 2
пЃєReview process of the completion of
hospital work sheet prenatal care
procedures check boxes for FD and
пЃ№document process of completion
пЃ№contact sample of hospitals and interview
appropriate staff about issues--where are
the data found in record? Who records?
пЃєMassachusetts is currently involved in two
projects using birth certificate data for
research and evaluation of FD/ART
пЃєThis is an important and rapidly expanding use
of birth data
пЃєIf we can accurately identify FD/ART use from
bc, then many avenues of cross-sectional and
longitudinal research can be implemented that
will enable us to better understand the
outcomes of fertility treatment assisted births
Размер файла
114 Кб
Пожаловаться на содержимое документа