Using ICD Codes and Birth Records to Prevent Mismatches of Multiple Births in Linked Hospital Readmi - PowerPoint PPT Presentation

1 / 1
About This Presentation
Title:

Using ICD Codes and Birth Records to Prevent Mismatches of Multiple Births in Linked Hospital Readmi

Description:

... the remaining mismatches with names were primarily 'Baby A' and 'Baby B' ... Figure 1. Percentage of Inpatient Records with Full Names by Age and Sex ... – PowerPoint PPT presentation

Number of Views:120
Avg rating:3.0/5.0
Slides: 2
Provided by: alison
Category:

less

Transcript and Presenter's Notes

Title: Using ICD Codes and Birth Records to Prevent Mismatches of Multiple Births in Linked Hospital Readmi


1
Using ICD Codes and Birth Records to Prevent
Mismatches of Multiple Births in Linked Hospital
Readmission Data
Alison Fraser1, MSPH, Zhiwei Liu2, MS, Emily
Smith1, MA, Geraldine Mineau1, PhD and Wu Xu2,
PhD 1University of Utah, Salt Lake City, Utah
2Utah Department of Health, Salt Lake City, Utah
Linkage Methods
Abstract
Linkage was conducted separately by the Utah
Department of Health (UDOH) using LinkSolv 7.0
(MatchWare Technologies) and then at the
University of Utah (UofU) using QualityStage 7.5,
a part of IBMs Websphere Information Integration
Solution. Both products assign a numeric weight
which is used to determine threshold values of
accepted and rejected links. The UDOH used
social security number (SSN), hospital medical
record number, name, birth date, gender, zip
code, and hospital as linking variables with nine
independent blocking and matching strategies.
The UofU linked the inpatient records
stratified by gender using names, dates, zip
codes and encrypted SSN in 22 different
combinations for blocking and matching. In
addition, UofU linked 49,464 (83.1) inpatient
records to the UPDB. The UofU identified 1,640
infant twins on inpatient records via ICD-9-CM
diagnosis codes V3100-V372. After linking the
inpatient records to the UPDB, the multiplicity
flag on their birth certificates was used to
identify infants admitted to the hospital after
delivery. There were an additional 114 infants
identified via linking to birth certificates for
a total of 1,754 inpatient events involving
infant twins. UDOH did not flag multiple births
in their linkage.
Multiple births present a challenge for
accurately identifying newborn readmissions when
using probabilistic linkage methods and the given
name is missing. We discovered that using ICD
newborn codes and linking to birth certificates
to identify multiple births in hospital discharge
data improves the linking accuracy for newborn
readmissions by 37.6
Table 1. Differences in Availability of Linking
Variables and Strategies
Introduction
Using probabilistic linkage methods for
identifying hospital inpatient record
readmissions can lead to incorrect matches for
newborns and infants from multiple births as many
hospitals do not record the babys name, but all
other significant linking variables match
including parents social security number (SSN).
As part of a larger study linking hospital
discharge records, this study examined the impact
of identifying multiple births on newborn and
infant (
Data
  • Gold Standard Linking Strategy for Infant
    Multiple Birth Re-Admissions
  • Utilize diagnosis codes and linked birth
    certificates to flag multiple births
  • Identify different gender twins
  • Use diagnosis codes to differentiate twins, eg.
    one twin is low birth weight, other twin is
    normal weight
  • Use diagnosis codes to flag delivery records to
    avoid mis-matching
  • Use medical record number from same facility to
    identify readmissions where name is missing
  • Link multiple birth only if
  • first name is available and matches
  • other twin is different gender
  • medical record number matches

Utah Population Database (UPDB)
Findings from Linkage Comparison
  • research resource at the University of Utah
    (UofU) includes over 12 million records or
    documents representing over 8 million
    individuals.
  • central component of the UPDB is a vast set of
    family histories or genealogies, in which family
    members are linked to demographic and medical
    information.
  • individuals in this database are linked to a
    number of state-wide data sets, especially birth
    certificates. These birth certificates and
    genealogies provide a source for identifying
    multiple births.
  • The results of the independent linkage projects
    were compared for 59,975 infant hospital
    discharge records. The medical record number was
    used as an independent source of verification.
  • the UDOH linkage method identified 1,826 (3.2)
    infants with 3,856 multiple admissions
  • the UofU method identified only 646 (1.1)
    infants with 1,386 multiple admissions
  • the readmissions from UDOH which did not match
    on medical record number and were not matched by
    UofU were manually scanned and determined to be
    mismatches
  • there were 670 infants with 1,450 (37.6)
    mismatched admissions, of which 609 (90.1) were
    from multiple births
  • the majority of mismatches (1,193) were newborn
    delivery records linked to their siblings
    newborn delivery records
  • the major source of the mismatches was absent
    first names with 629 (93.8) mismatches where one
    or both of the records did not have a given name,
    but had matched on other pertinent linking
    variables the remaining mismatches with names
    were primarily Baby A and Baby B
  • UofU did not match 753 infants to 1,563
    admissions due to missing key linking variables,
    especially medical record number
  • A combination of links via medical record number
    and links by UofU was considered the Gold
    Standard with 1,567 infants identified with
    3,522 multiple admissions.

Utah Department of Health Inpatient Discharge
Summaries (UDOH)
  • contains statewide, population-based healthcare
    information associated with 56 hospitals in Utah
    from 1996-2006.
  • developed a comparative study on 290,537
    discharge records from 12/1/2004 to 12/31/2005,
    specifically 59,488 infants
  • restricts transfer of SSN and Medical Record
    Number must be encrypted
  • has incomplete linking information - 42 of
    patients 1 year old have full names cf. 22 of
    Infants

Table 2. Distribution of Infants with multiple
Admissions
Figure 1. Percentage of Inpatient Records with
Full Names by Age and Sex
Conclusion
Linkage methods for hospital discharge record
must address multiple births. The majority of
incorrectly linked infants are twins and can be
flagged through the delivery diagnosis code on
inpatient records. Furthermore, linking to an
external database that contains genealogical
records and birth certificates means that
multiple births can be identified, even if they
dont link to their delivery record or if they
are born out of state. Access to medical record
number is key to linking infants when names are
missing from the records.
Acknowledgement This project is sponsored by
NIH grant R01 RR021746, P.I. Mineau, Sharing
statewide health data for genetic research CDC
grant P01 CD000284, P.I. Samore, Utah Research
Center for Excellence in Public Health
Informatics and the Office of Health Care
Statistics, Utah Department of Health. Partial
support for datasets within the UPDB is provided
by the Huntsman Cancer Institute. UPDB web site
http//www.hci.utah.edu/groups/ppr/
Write a Comment
User Comments (0)
About PowerShow.com