Скачать презентацию Improving the linkage of deliveries over time using Скачать презентацию Improving the linkage of deliveries over time using

5e13e4eabcbc96f877c647ec207c8135.ppt

  • Количество слайдов: 19

Improving the linkage of deliveries over time using vital records and hospital discharge data Improving the linkage of deliveries over time using vital records and hospital discharge data Mark Mc. Laughlin, Judy Weiss, Sc. D, Milton Kotelchuck, MPH, Ph. D, MA, Stephen Evans, MPH Boston University School of Public Health Dept. of Maternal and Child Health Boston, MA, November 6, 2006 Contact info: Mark. E. [email protected] ma. us APHA, Nov. 2006

Why link successive deliveries? • Create a reproductive history of individual women to study, Why link successive deliveries? • Create a reproductive history of individual women to study, for example: – Risk of maternal and infant morbidity – Risk & consequences of premature birth – Impact of OB practices (e. g. , method of delivery) on subsequent birth outcomes • Past linkage efforts in the US have been limited and/or have ended – Birth certificates +/- fetal death certificates • Need for improved ‘case-finding’ methods to capture most complete maternal linkage possible APHA, Nov. 2006

Pregnancy to Early Life Longitudinal (PELL) Data System • CDC-funded collaborative project • MCH Pregnancy to Early Life Longitudinal (PELL) Data System • CDC-funded collaborative project • MCH Dept. Boston Univ. School of Public Health • MA Department of Public Health (MDPH) • A population-based relational data system connected by unique identifiers for infants, deliveries and mothers • Consists of a Core plus additional program and other surveillance data • Is longitudinal and expandable APHA, Nov. 2006

PELL Data System Vital and Health Status Data Program Participation Data CORE WIC (2006) PELL Data System Vital and Health Status Data Program Participation Data CORE WIC (2006) Newborn Hearing Screening (2006) Birth Defects Registry Early Intervention Other DPH programs (future) Birth Certificate Child and Mother deaths Hospital Discharge (HD) Pregnancy-associated deaths Mother’s delivery Fetal Death Hospital Discharge (HD) Child’s birth Linked birth-infant deaths Health Services Utilization Data Non-birth hospital discharge Contextual Data Geocoded birth data (2006) Emergency Department 2000 Census (2006) Observational Stays Other contextual data (e. g. , environmental data) APHA, Nov. 2006 Other Future Data Sets (e. g. , School, ‘ART’, NICU, Medicaid)

PELL Core Data Sets Birth and Fetal Death Certificate with Hospital Discharge Maternal Delivery PELL Core Data Sets Birth and Fetal Death Certificate with Hospital Discharge Maternal Delivery Record and Child Birth Record CORE Birth Certificate Hospital Discharge (HD) Mother’s delivery Fetal Death Hospital Discharge (HD) Child’s birth APHA, Nov. 2006

Creation of PELL Core Linkages Birth/fetal death to Hospital records • Deterministic and probabilistic Creation of PELL Core Linkages Birth/fetal death to Hospital records • Deterministic and probabilistic linkage methodology using Link. Pro 2. 0 (SASbased) software • Core Linkage Variables: – Facility code – Maternal Date of Birth – Date of Delivery – Zipcode No Unique Identifier APHA, Nov. 2006

PELL Core Data Linkage Rates 1998 -2004 • 574, 442 occurrence live births and PELL Core Data Linkage Rates 1998 -2004 • 574, 442 occurrence live births and fetal death records – Overall, 98. 6% linked to maternal hospital delivery record – 77. 8% of fetal deaths linked to maternal hospital delivery record Now that the data are linked… APHA, Nov. 2006

…can we create a better maternal linkage? Is there a way to combine Vitals …can we create a better maternal linkage? Is there a way to combine Vitals (both births and fetal deaths) and Hospital Discharge data to create a more robust maternal linkage? APHA, Nov. 2006

Creation of longitudinally-linked delivery data • 3 -step process using vital records and HD Creation of longitudinally-linked delivery data • 3 -step process using vital records and HD data individually and jointly • Step 1: identify unique women within each data source and count the number of deliveries per woman – Vitals linkage based on names, DOB – HD linkage based on encrypted SSN and/or facility code plus medical record number APHA, Nov. 2006

Example Hospital Discharge Vitals First name Last name DOB Judy Miller 11/16/70 # of Example Hospital Discharge Vitals First name Last name DOB Judy Miller 11/16/70 # of delivs DOB # of delivs UHIN* MRN** 2 XZ 9432 0123456 11/16/70 3 2 3 * Encrypted SSN APHA, Nov. 2006 ** Medical record number

Creation of longitudinally-linked delivery data • Step 2: Merge vital records and HD data Creation of longitudinally-linked delivery data • Step 2: Merge vital records and HD data by delivery and compare the number of deliveries – Three scenarios • Both classify equal number of deliveries (92. 8%) • Vitals classifies more deliveries than HD (5. 5%) • HD classifies more deliveries than Vitals (1. 7%) • Accept whichever data (Vitals or HD) classified more deliveries APHA, Nov. 2006

Example Hospital Discharge data Vitals data First name Last name DOB Delivery date UHIN Example Hospital Discharge data Vitals data First name Last name DOB Delivery date UHIN MRN DOB Delivery date Judy Miller 11/16/70 6/1/98 XZ 9432 0123456 11/16/70 6/1/98 Judy Miller 11/16/70 12/1/00 XZ 9432 0123456 11/16/70 12/1/00 Judi Millar 11/16/70 9/1/02 XZ 9432 0123456 11/16/70 APHA, Nov. 2006 9/1/02

Creation of longitudinally-linked delivery data • Step 3: Assess validity of linkages – Validate Creation of longitudinally-linked delivery data • Step 3: Assess validity of linkages – Validate names • Compare names from first delivery to names from all subsequent deliveries using SPEDIS function in SAS • Women with at least one name that appears significantly different (score>30) are flagged for manual review • When reviewing names, judgment of validity is made based on names, dates, parity, and other demographic variables • Review performed by two people and compared APHA, Nov. 2006

Examples Correct maternal link First name Score Last name Score Delivery date Parity Judy Examples Correct maternal link First name Score Last name Score Delivery date Parity Judy 0 Miller 0 6/1/98 1 Judy 0 Miller 0 12/1/00 2 Judi 25 Millar 16 9/1/02 3 Incorrect maternal link First name Score Last name Score Delivery date Parity Jennifer 0 Smith 0 3/1/98 1 Jennifer 0 Smyth 20 9/1/99 2 Barbara 104 Jones 120 9/5/99 5 APHA, Nov. 2006

Creation of longitudinally-linked delivery data • Step 3, cont’d: Assess validity of links – Creation of longitudinally-linked delivery data • Step 3, cont’d: Assess validity of links – Validate short inter-pregnancy intervals • Calculate the time between deliveries • Women with short delivery intervals (<250 days) are output for review • Review performed by two people and compared – After review, maternal ID is assigned APHA, Nov. 2006

Creation of longitudinally-linked delivery data RESULTS (1998 -2004): • 427, 817 unique women identified Creation of longitudinally-linked delivery data RESULTS (1998 -2004): • 427, 817 unique women identified • 122, 094 women found with >1 delivery – 118, 860 women found with >1 delivery in Vitals alone – 114, 014 women found with >1 delivery in HD alone • 2. 7% more women with >1 delivery than Vitals alone • ~500 women with >1 delivery saw additional increase in deliveries APHA, Nov. 2006

Practical Challenges • Incomplete information – Early fetal losses (prior to 20 weeks) – Practical Challenges • Incomplete information – Early fetal losses (prior to 20 weeks) – Non-occurrence fetal deaths – Births to women who move out of state • Initial linkage can be time consuming (if spanning several years) – Subsequent years require less time – 350 -400 records reviewed each year APHA, Nov. 2006

Conclusions • Using 3 data sources (births, fetal deaths, hospital discharge data) provides a Conclusions • Using 3 data sources (births, fetal deaths, hospital discharge data) provides a more complete reproductive history than births and fetals alone • Validation of linkages makes final data linkage more robust • Longitudinally-linked delivery data provide a rich source for exploration of MCH issues APHA, Nov. 2006

Caveats • Vitals data – Name changes (marriage, divorce, etc. ) – Spelling errors, Caveats • Vitals data – Name changes (marriage, divorce, etc. ) – Spelling errors, typographical errors • Hospital Discharge data – Only includes births in MA hospitals (no home births, out of state births, most births at birthing centers – Encrypted SSN (UHIN) sometimes missing (n=32, 963 [5. 6%]) so reliance on medical record number and assumption that woman goes to the same hospital – No Hospital Discharge data on unlinked deliveries (n=14, 021 [2. 5%]) APHA, Nov. 2006