8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 1/24

Agency for Healthcare Research and Quality Agency for Healthcare Research and Quality 

Advancing Excellence in Health Care Advancing Excellence in Health Care • • www.ahrq.govwww.ahrq.gov

Enhancing Claims Data to Improve Risk Adjustment Enhancing Claims Data to Improve Risk Adjustment of Mortality and Patient Safety Indicators of Mortality and Patient Safety Indicators 

Anne Elixhauser, Ph.D.Anne Elixhauser, Ph.D.

National Committee on Vital and Health StatisticsNational Committee on Vital and Health Statistics

June 19, 2007June 19, 2007

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 2/24

2

Advancing Advancing Excellence in Excellence in Health Care Health Care  Outline Outline 

BackgroundBackground

Summary of study: Adding clinical dataSummary of study: Adding clinical dataelements to administrative dataelements to administrative data

Supporting the enhancement of administrativeSupporting the enhancement of administrativeclaims data: Next stepsclaims data: Next steps

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 3/24

3

Advancing Advancing Excellence in Excellence in Health Care Health Care  Background Background 

Hospital administrative claims dataHospital administrative claims data

 –  –  Statewide data readily available from most statesStatewide data readily available from most states AHRQ Quality IndicatorsAHRQ Quality Indicators

 –  –  Prevention Quality IndicatorsPrevention Quality Indicators

 –  –  Inpatient Quality Indicators (mortality, utilization,Inpatient Quality Indicators (mortality, utilization,volume)volume)

 –  –  Patient Safety IndicatorsPatient Safety Indicators

 –  –  Pediatric Quality IndicatorsPediatric Quality Indicators

Administrative data currently used for publicAdministrative data currently used for publicreporting on quality of carereporting on quality of care

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 4/24

4

Advancing Advancing Excellence in Excellence in Health Care Health Care  Limitations of Current Data Limitations of Current Data 

Lack clinically important informationLack clinically important information

 –  –  Limited to ICDLimited to ICD--99--CM diagnosis codesCM diagnosis codes

Do not distinguish between diagnoses presentDo not distinguish between diagnoses presenton admission (POA) and those that originateon admission (POA) and those that originateduring the hospital stayduring the hospital stay

Questions regarding use of only administrativeQuestions regarding use of only administrativedata for hospitaldata for hospital--specific reportingspecific reporting –  –  Inadequate risk adjustmentInadequate risk adjustment –  – additional dataadditional data

needed to predict individual patientneeded to predict individual patient’’s risk ofs risk ofmortalitymortality

 –  –  Concern about penalizing providers with theConcern about penalizing providers with thesickest patientssickest patients

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 5/24

5

Advancing Advancing Excellence in Excellence in Health Care Health Care 

Tension Between Value of Data Tension Between Value of Data 

and Cost of Obtaining the Data and Cost of Obtaining the Data 

New York and California provide POA codingNew York and California provide POA coding

for diagnosesfor diagnoses –  – more states adding thismore states adding this Pennsylvania hospitals provide chartPennsylvania hospitals provide chart--

abstracted clinical detailabstracted clinical detail

 –  –  Hospital concern about costs of medical recordHospital concern about costs of medical recordabstractionabstraction

Electronic medical records not yet poised toElectronic medical records not yet poised to

provide data efficientlyprovide data efficiently –  –  Exception: Lab dataException: Lab data

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 6/24

6

Advancing Advancing Excellence in Excellence in Health Care Health Care  Study Objective Study Objective 

Assess impact of incrementally adding:Assess impact of incrementally adding:

 –  –  POA codes for diagnosesPOA codes for diagnoses –  –  Lab values on admissionLab values on admission

 –  –  Increased number of diagnosis fieldsIncreased number of diagnosis fields

 –  –  Improved documentation (ICDImproved documentation (ICD--99--CM codes)CM codes)

 –  –  Vital signsVital signs

 –  –  More difficult to obtain clinical dataMore difficult to obtain clinical data

Identify costIdentify cost--effective enhancements toeffective enhancements toadministrative dataadministrative data

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 7/24

7

Advancing Advancing Excellence in Excellence in Health Care Health Care  Study Reported in Study Reported in ……

Pine M, Jordan HS, Elixhauser A, et al. EnhancementPine M, Jordan HS, Elixhauser A, et al. Enhancement

of claims data to improve risk adjustment of hospitalof claims data to improve risk adjustment of hospitalmortality.mortality. JAMAJAMA 2007; 267(1):712007; 267(1):71--76.76.

Jordan HS, Pine M,Jordan HS, Pine M, ElixhauserElixhauser A, et al. CostA, et al. Cost--effectiveeffectiveenhancement of claims data to improve comparisonsenhancement of claims data to improve comparisons

of patient safety.of patient safety. Journal of Patient Safety Journal of Patient Safety 2007; 3(2)2007; 3(2)8282--90.90.

Fry DR, Pine M, Jordan HS, et al. CombiningFry DR, Pine M, Jordan HS, et al. Combiningadministrative and clinical data to stratify surgical risk.administrative and clinical data to stratify surgical risk.

Annals of Surgery Annals of Surgery (forthcoming).(forthcoming). Pine M, Jordan HS, Elixhauser A, et al. ModifyingPine M, Jordan HS, Elixhauser A, et al. Modifying

claims data to improve riskclaims data to improve risk--adjustment of inpatientadjustment of inpatientmortality rates. (Submitted for publication)mortality rates. (Submitted for publication)

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 8/24

8

Advancing Advancing Excellence in Excellence in Health Care Health Care  Sources of Data Sources of Data 

188 Pennsylvania hospitals188 Pennsylvania hospitals

 –  –  Claims data from July 2000 to June 2003Claims data from July 2000 to June 2003

 –  –  Corresponding Atlas clinical dataCorresponding Atlas clinical data

 –  –  Hospital day recorded for each data elementHospital day recorded for each data element

New York and California claims dataNew York and California claims data –  –  Distinguish which conditions were comorbiditiesDistinguish which conditions were comorbidities

versus complicationsversus complications

 –  –  Identify potential risk factorsIdentify potential risk factors

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 9/24

9

Advancing Advancing Excellence in Excellence in Health Care Health Care  Indicators Studied Indicators Studied 

Mortality Mortality 

Indicators Indicators  AAA repairAAA repair

CABG surgeryCABG surgery

CraniotomyCraniotomy AMIAMI

CHFCHF

CerebrovascularCerebrovascular

accidentaccident GI hemorrhageGI hemorrhage

PneumoniaPneumonia

Post Post - - operative patient operative patient 

safety events safety events  Pulmonary embolism/deepPulmonary embolism/deep

vein thrombosisvein thrombosis

Physiologic/metabolicPhysiologic/metabolicabnormalitiesabnormalities

Respiratory failureRespiratory failure

SepsisSepsis

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 10/24

10

Advancing Advancing Excellence in Excellence in Health Care Health Care 

Data Used in Incrementally More Data Used in Incrementally More 

Complex Models Complex Models 

Same as POASame as POA--8 with up to 24 secondary8 with up to 24 secondary

diagnosesdiagnoses

POAPOA--2424

POAPOA--24 + secondary diagnoses not included in24 + secondary diagnoses not included inPOAPOA--24 because they were underreported in24 because they were underreported inadministrative database but were established asadministrative database but were established as

present on admission in clinical databasepresent on admission in clinical database

POAPOA--ICDICD

ADMADM

--8 + secondary diagnoses not included in8 + secondary diagnoses not included in

ADMADM--8, when clinical data establish that they were8, when clinical data establish that they werepresent on admissionpresent on admission

POAPOA--88

Age, sex, principal diagnosis, up to 8 secondaryAge, sex, principal diagnosis, up to 8 secondarydiagnoses only infrequently acquired duringdiagnoses only infrequently acquired duringhospitalization, selected surgical procedureshospitalization, selected surgical procedures

ADMADM--88

Types of Data ElementsTypes of Data ElementsModelModel

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 11/24

11

Advancing Advancing 

Excellence in Excellence in Health Care Health Care 

Data Used in Incrementally More Data Used in Incrementally More 

Complex Models Complex Models 

LAB + vital signs and lab data not in LAB (e.g.,LAB + vital signs and lab data not in LAB (e.g.,

blood culture results) + key clinical findingsblood culture results) + key clinical findingsabstracted from medical records (e.g.,abstracted from medical records (e.g.,immunocompromised) + composite clinicalimmunocompromised) + composite clinicalscores (i.e., ASA Classification)scores (i.e., ASA Classification)

FULLFULL

LAB + secondary diagnoses not included inLAB + secondary diagnoses not included in

POAPOA--24 because they were underreported in24 because they were underreported inadministrative database but were established asadministrative database but were established aspresent on admission in clinical databasepresent on admission in clinical database

LABLAB--ICDICD

POAPOA--24 + numerical laboratory data on24 + numerical laboratory data onadmission (e.g., creatinine, white blood celladmission (e.g., creatinine, white blood cellcount) generally available in electronic formcount) generally available in electronic form

LABLAB

Types of Data ElementsTypes of Data ElementsModelModel

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 12/24

12

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  C C - - Statistics for Mortality Models Statistics for Mortality Models 

0.76

0.78

0.80

0.82

0.84

0.86

0.88

0.90

ADM-8 POA-8 POA-24 POA-

ICD

LAB LAB-

ICD

FULL

   A  v  e  r  a  g  e   C

  -   S   t  a   t   i  s   t   i  c

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 13/24

13

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Bias Due to Suboptimal Data Bias Due to Suboptimal Data 

+ 2 Std Dev

Good Average Poor

- 2 Std Dev

Measured Performance

+ 0.5 Std Dev

Problematic Problematic

- 0.5 Std Dev

Bias

OK

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 14/24

14

Advancing Advancing 

Excellence in Excellence in Health Care Health Care 

Hospital Bias Due to Suboptimal Hospital Bias Due to Suboptimal 

Data Data – – Mortality Models Mortality Models 

0%

10%

20%

30%

40%

50%

60%

70%

0.5 1.0 1.5 2.0

Upper Threshold for Bias in Standard Deviations

   P  e  r  c

  e  n   t   E  x  c  e  e   d   i  n  g

   U  p  p  e  r   T   h  r  e  s   h  o   l   d

RAW ADM-8 POA-8 POA-24 POA-IC LAB LAB-IC

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 15/24

15

Advancing Advancing 

Excellence in Excellence in Health Care Health Care 

Hospital Bias Due to Suboptimal Hospital Bias Due to Suboptimal 

Data Data – – Patient Safety Models Patient Safety Models 

0%

10%

20%

30%

40%

50%

60%

70%

0.5 1.0 1.5 2.0

Upper Threshold for Bias in Standard Deviations

   P  e  r

  c  e  n   t   E  x  c  e  e   d   i  n  g

   U  p  p  e  r   T   h  r  e  s   h  o

   l   d

RAW ADM POA LAB

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 16/24

16

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Numerical Lab Data Numerical Lab Data 

pH (11)pH (11)

PTT (10)PTT (10)

Na (9)Na (9)

WBC (9)WBC (9)

BUN (8)BUN (8) pOpO22 (8)(8)

K (7)K (7)

SGOT (7)SGOT (7)

Platelets (7)Platelets (7)

Albumin (5)Albumin (5)

pCOpCO22 (4)(4)

Glucose (4)Glucose (4) Creatinine (4)Creatinine (4)

CPKCPK--MB (4)MB (4)

Results of 22 lab tests entered at least onemodel

Results of 14 of these tests entered four orResults of 14 of these tests entered four ormore models:more models:

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 17/24

17

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Vital Signs and Other Clinical Data Vital Signs and Other Clinical Data 

All vital signs entered four or more modelsAll vital signs entered four or more models

 –  –  Pulse (8)Pulse (8) –  –  Temp (6)Temp (6)

 –  –  Blood pressure (6)Blood pressure (6)

 –  – Respirations (5)Respirations (5)

Ejection fraction and culture results enteredEjection fraction and culture results enteredtwo modelstwo models

Composite scores entered four or moreComposite scores entered four or more

modelsmodels –  –  ASA classification (6)ASA classification (6)

 –  –  Glasgow Coma Score (4)Glasgow Coma Score (4)

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 18/24

18

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Abstracted Key Clinical Findings Abstracted Key Clinical Findings 

35 clinical findings entered at least one model35 clinical findings entered at least one model

Only three findings entered more than two modelsOnly three findings entered more than two models –  –  Coma (6)Coma (6)

 –  –  Severe malnutrition (4)Severe malnutrition (4)

 –  –  Immunosuppressed (4)Immunosuppressed (4)

14 of these clinical findings have corresponding ICD14 of these clinical findings have corresponding ICD--99--CM codes (e.g., coma, malnutrition)CM codes (e.g., coma, malnutrition)

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 19/24

19

Advancing Advancing 

Excellence in Excellence in Health Care Health Care 

Marginal Cost of Improved Risk Marginal Cost of Improved Risk 

Adjustment Adjustment 

0%

20%

40%

60%

80%

100%

ADM POA LAB LAB-IC FULL

   P

  r  e  c  e  n   t  a  g  e

  o   f   H  o  s  p   i   t  a

   l  s

$0

$20

$40

$60

$80

$100

   C  o  s   t   i  n

   $

Bias <0.5 Std Dev (%)

Added Cost per 10 Cases ($)Marginal Cost per Correction ($)

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 20/24

20

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Summary of Analyses Summary of Analyses 

Administrative data can be improved atAdministrative data can be improved at

relatively low cost by:relatively low cost by: –  –  Adding POA modifiersAdding POA modifiers

 –  –  Adding numerical lab data on admissionAdding numerical lab data on admission

 –  –  Improved codingImproved coding

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 21/24

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 22/24

22

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Objectives of Pilots Objectives of Pilots 

Establish feasibility of linking clinical andEstablish feasibility of linking clinical and

administrative dataadministrative data Develop reproducible approachDevelop reproducible approach

Set the stage for integrating clinical andSet the stage for integrating clinical and

administrative data streams in the futureadministrative data streams in the future

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 23/24

23

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Specific Activities Specific Activities 

Identify and select clinical data elements to add toIdentify and select clinical data elements to add toadministrative dataadministrative data

Translate clinical data from electronic formatTranslate clinical data from electronic format Electronically transfer data from at least five hospitalsElectronically transfer data from at least five hospitals

to the data organizationto the data organization

Process data into a multiProcess data into a multi--hospital databasehospital database

Collaborate with stakeholders, e.g.Collaborate with stakeholders, e.g. –  –  Hospital representativesHospital representatives

 –  –  State government agenciesState government agencies

 –  – 

Researchers, quality measurement professionalsResearchers, quality measurement professionals

 –  –  Regional or state health care quality organizationsRegional or state health care quality organizations

 –  –  Regional health information exchangeRegional health information exchange

Engage in peerEngage in peer--toto--peer learning, information sharing,peer learning, information sharing,

disseminationdissemination

8/14/2019 Health and Human Services: 070619p4

http://slidepdf.com/reader/full/health-and-human-services-070619p4 24/24

24

Advancing Advancing 

Excellence in Excellence in Health Care Health Care  Conclusion Conclusion 

Judicious addition of a few clinical dataJudicious addition of a few clinical data

elements can significantly improve ability to doelements can significantly improve ability to doquality assessment using administrative dataquality assessment using administrative data

 –  –  POAPOA

 –  –  Labs on admissionLabs on admission –  –  (Potentially) vital signs(Potentially) vital signs

 –  –  Improved ICDImproved ICD--99--CM codingCM coding

Pilots and planning contracts will jumpstart thePilots and planning contracts will jumpstart theenhancement of administrative data byenhancement of administrative data bystatewide data organizationsstatewide data organizations