Normalization of Phenotypic Data from a Clinical Data Warehouse: Case Study of Heterogeneous Blood Type Data with Surprising Results James J. Cimino, MD Formerly: Chief of the Laboratory for Informatics Development NIH Clinical Center, National Institutes of Health Bethesda, Maryland, USA Now: Director, Informatics Institute University of Alabama at Birmingham Birmingham, Alabama, USA INFORMATICS INSTITUTE

Lecture Overview

INFORMATICS INSTITUTE



Started out trying to normalize laboratory data



Mapped different tests to common findings



Identified need to “atomize” data



Real data set found some surprises



Speculation on causes



Take-home messages

Biomedical Translational Research Information System (BTRIS) Institute System Old EHR

Curent EHR

BTRIS

INFORMATICS INSTITUTE

Personal System

Lab System

INFORMATICS INSTITUTE

ABO Blood Typing

- from Wikipedia INFORMATICS INSTITUTE

ABO and Rh Blood Typing Rh Negative

-

AntiRh

-

AntiRh

-

AntiRh

-

AntiRh

- from Wikipedia INFORMATICS INSTITUTE

ABO and Rh Blood Typing Rh Positive

+ R h

R h

Rh antigen

+

R h

Rh antigen

+ R h

R h

Rh antigen

+ R h

R h

Rh antigen

R h

- from Wikipedia INFORMATICS INSTITUTE

Panels Reporting Multiple Antigens and Interpretations

Panel

Tests

Result

ABO GRP-RH TYPE

ABO GRP-RH TYPE

O POSITIVE

ABO GRP-RH TYPE

ABO GRP-RH TYPE

A POSITIVE

ABO GRP-RH TYPE

ABO GRP-RH TYPE

A NEGATIVE

ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - A

0

ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - B ABO Group and Rh Type [ABORH] - Rh

4+ POS

ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - A

0

ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - B ABO Group and Rh Type [ABORH] - Rh

0 NEG

INFORMATICS INSTITUTE

B positive

O negative

What are the Underlying Atomic Findings? Rh Positive

+ R h

R h

Rh antigen

+

R h

Absence of B ag

Rh antigen

+ R h

R h

Absence of A ag

Rh antigen

+ R h

R h

Rh antigen

R h

Absence of A ag Absence of B ag

- from Wikipedia INFORMATICS INSTITUTE

Variants and Typographical Errors

Panel

Tests

Result

ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO Group and Rh Type [ABORH]

ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO Group and Rh Type [ABORH] - A

O POSITIVE 0 POS A POSITIVE A NEG A NEGATIVE AB NEG B POS 0

ABO ABO ABO ABO ABO

ABO ABO ABO ABO ABO

Group and Rh Type Group and Rh Type Group and Rh Type Group and Rh Type Group and Rh Type

[ABORH] [ABORH] [ABORH] [ABORH] [ABORH]

Group and Rh Type Group and Rh Type Group and Rh Type Group and Rh Type Group and Rh Type

[ABORH] [ABORH] [ABORH] [ABORH] [ABORH]

-A -A -A -A -A

1+ 2+ 3+ 4+ M4

ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - B

0

ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - B ABO Group and Rh Type [ABORH] - Rh ABO Group and Rh Type [ABORH] - Rh

4+ NEG POS

INFORMATICS INSTITUTE

Interpretation of Presence or Absence of Antigens

Panel

Tests

Result

Antigens

ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE

ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE ABO GRP-RH TYPE

O POSITIVE 0 POS A POSITIVE A NEG A NEGATIVE AB NEG B POS

abR abR AbR Abr Abr ABr aBR

ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - A

0

a

ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - B ABO Group and Rh Type [ABORH] – Rh

1+ Pos

b R

ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - A ABO Group and Rh Type [ABORH] - B

4+ 4+

A B

ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] - Rh

NEG

r

ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH] ABO Group and Rh Type [ABORH]

ABO Group and Rh Type [ABORH] – A ABO Group and Rh Type [ABORH] – B ABO Group and Rh Type [ABORH] - Rh

0 0 POS

a b R

INFORMATICS INSTITUTE

Summary

abR

ABr

abR

Does the Atomic Approach Support Data Integration?

Hypothesis: Different tests of the same blood type should produce the same atomic results. Experiment: Different tests on the same patient should produce the same atomic results.

INFORMATICS INSTITUTE

Experimenting with BTRIS



Queried BTRIS for all ABO and Rh test results



Identify unique panel/test combinations



Identify unique results of panel/tests combinations



Create atomic maps for each unique result



Identify each patient’s phenotype (union of atoms)



Examine phenotypes for discrepant results

INFORMATICS INSTITUTE

INFORMATICS INSTITUTE

Summary of Results 43,760 Patients 176,676 Panels 593,637 Tests

Summarization

66 unique Panels 139 unique Tests 334 unique Panel-Test combinations 3949 unique results

Manual Review to Select Relevant Tests 43,486 patients 165,981 panels 307,884 Tests

Filtering

23,903 patients with multiple panels

479 discrepant phenotypes (2.00%) INFORMATICS INSTITUTE

21 unique Panels 32 unique tests 59 unique Panel-Test combinations 1452 unique results 19,583 patients with single panel

Expected Phenotypes

Antigenic Evidence abR AbR aBR abr Abr ABR aBr ABr

INFORMATICS INSTITUTE

Phenotype O+ A+ B+ OAAB+ BAB-

# Patients 17132 13925 4710 2538 2316 1441 645 214

Incomplete Phenotypes

Antigenic Evidence r R ab Ab AB aB bR

INFORMATICS INSTITUTE

Phenotype + O A AB B +

# Patients 10 8 7 5 1 1 1

Discrepant Phenotypes Antigenic Evidence AabR abRr AbRr aBbR AaBbR AabRr ABbR aBRr AaBR Aabr aBbr ABRr AaBbRr aBbRr ABbr AaBbr ABbRr INFORMATICS INSTITUTE

Phenotype (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant) (discrepant)

# Patients 132 89 67 51 50 28 24 19 17 13 11 7 6 6 6 3 2

Examples of Same-Patient Discrepant Results Subj

Date

Test

Result

Ags

Interp.

59

1/31/1989

ABO & RH

O POSIT.

abR

abR (O+)

59

1/31/1989

ABO & RH

A POSIT.

AbR

AbR (O-)

724

1/24/1989

ABO & RH

O NEG

abr

abr (O-)

724

2/13/1989

ABO & RH

O POS

abR

abR (O+)

986

1/2/1999

ABO Group and Rh - Rh

POS

R

986

1/2/1999

ABO Group and Rh - A

4+

A

986

1/2/1999

ABO Group and Rh - B

4+

B

986

1/18/2000

ABO Group and Rh - Rh

POS

R

986

1/18/2000

ABO Group and Rh - A

4+

A

986

1/18/2000

ABO Group and Rh – B

0

b

INFORMATICS INSTITUTE

ABR (AB+)

AbR (A+)

Examples of Same-Patient Discrepant Results Subj

Date

Test

Result

Ags

Interp.

59

1/31/1989

ABO & RH

O POSIT.

abR

abR (O+)

Phen.

AabRr 59

1/31/1989

ABO & RH

A POSIT.

AbR

AbR (O-)

724

1/24/1989

ABO & RH

O NEG

abr

abr (O-)

724

2/13/1989

ABO & RH

O POS

abR

abR (O+)

986

1/2/1999

ABO Group and Rh - Rh

POS

R

986

1/2/1999

ABO Group and Rh - A

4+

A

986

1/2/1999

ABO Group and Rh - B

4+

B

986

1/18/2000

ABO Group and Rh - Rh

POS

R

986

1/18/2000

ABO Group and Rh - A

4+

A

986

1/18/2000

ABO Group and Rh – B

0

b

INFORMATICS INSTITUTE

abRr

ABR (AB+) AbBR AbR (A+)

More Examples of Same-Patient Discrepant Results

Subj

Date

Test

Result

Ags

1090

1/2/2002

ABO Group and Rh - ABO

A

Ab

1090

1/2/2002

ABO Group and Rh - Rh

POS

R

1090

1/2/2002

ABO Group and Rh - A

4+

A

1090

1/2/2002

ABO Group and Rh - B

0

b

1090 1/28/2003

ABO Group and Rh - ABO

B

aB

1090 1/28/2003

ABO Group and Rh - Rh

POS

R

1090 1/28/2003

ABO Group and Rh - A

0

a

1090 1/28/2003

ABO Group and Rh - B

4+

B

INFORMATICS INSTITUTE

Interp.

AbR (A+)

aBR (B+)

More Examples of Same-Patient Discrepant Results

Subj

Date

Test

Result

Ags

1090

1/2/2002

ABO Group and Rh - ABO

A

Ab

1090

1/2/2002

ABO Group and Rh - Rh

POS

R

1090

1/2/2002

ABO Group and Rh - A

4+

A

1090

1/2/2002

ABO Group and Rh - B

0

b

1090 1/28/2003

ABO Group and Rh - ABO

B

aB

1090 1/28/2003

ABO Group and Rh - Rh

POS

R

1090 1/28/2003

ABO Group and Rh - A

0

a

1090 1/28/2003

ABO Group and Rh - B

4+

B

INFORMATICS INSTITUTE

Interp.

Phen.

AbR (A+) AaBbR

aBR (B+)

Possible Explanation: Random Laboratory Error

INFORMATICS INSTITUTE



Doubling the tests for a patient should double the chance of random error



bCorrelation was 0.7127 (P