29
Calculating Measures of Calculating Measures of Comorbidity Using Comorbidity Using Administrative Data Administrative Data Vicki Stagg Vicki Stagg Statistical Programmer Statistical Programmer Department of Community Health Sciences Department of Community Health Sciences Dr. Robert Hilsden Dr. Robert Hilsden Associate Professor Associate Professor Departments of Medicine and Community Health Departments of Medicine and Community Health Sciences Sciences Dr. Hude Quan Dr. Hude Quan Associate Professor Associate Professor Centre for Health and Policy Studies (CHAPS) Centre for Health and Policy Studies (CHAPS) Department of Community Health Sciences Department of Community Health Sciences University of Calgary University of Calgary Calgary, Alberta, Canada Calgary, Alberta, Canada

Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Embed Size (px)

Citation preview

Page 1: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Calculating Measures of Calculating Measures of Comorbidity Using Comorbidity Using

Administrative DataAdministrative DataVicki StaggVicki Stagg

Statistical ProgrammerStatistical ProgrammerDepartment of Community Health SciencesDepartment of Community Health Sciences

Dr. Robert HilsdenDr. Robert HilsdenAssociate ProfessorAssociate Professor

Departments of Medicine and Community Health SciencesDepartments of Medicine and Community Health Sciences

Dr. Hude QuanDr. Hude QuanAssociate ProfessorAssociate Professor

Centre for Health and Policy Studies (CHAPS)Centre for Health and Policy Studies (CHAPS)Department of Community Health SciencesDepartment of Community Health Sciences

University of CalgaryUniversity of CalgaryCalgary, Alberta, CanadaCalgary, Alberta, Canada

Page 2: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

BackgroundBackground

• Medical Administrative Data Medical Administrative Data • Inpatient hospital visit informationInpatient hospital visit information

• ComorbidityComorbidity• Pre-existing diagnosis / additional complication of admitted patientPre-existing diagnosis / additional complication of admitted patient

• Comorbidity IndexComorbidity Index• For measurement of burden of disease and case-mix adjustmentFor measurement of burden of disease and case-mix adjustment• Allows for stratification or adjustment by severity of illnessAllows for stratification or adjustment by severity of illness• Two common tools – Two common tools – CharlsonCharlson and and ElixhauserElixhauser

• Clinical ConditionsClinical Conditions • International Classification of DiseaseInternational Classification of Disease

– 99thth Revision, Clinical Modification ( Revision, Clinical Modification (ICD-9-CM ICD-9-CM codes)codes)– 1010thth Revision ( Revision (ICD-10ICD-10 codes) codes)

((International statistical classification of disease and related health problems, International statistical classification of disease and related health problems, 1992)1992)

Page 3: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Algorithms included in ado Algorithms included in ado programsprograms• Charlson (17 comorbidity definitions)Charlson (17 comorbidity definitions)

• Presence/absence, weighted sum (Charlson index)Presence/absence, weighted sum (Charlson index)• Charlson index developed to predict risk of one-year mortality Charlson index developed to predict risk of one-year mortality

from comorbid illnessfrom comorbid illness((J Chron Dis, J Chron Dis, 1987;40(5):373-383)1987;40(5):373-383)

• Deyo modification for Deyo modification for ICD-9-CMICD-9-CM((J Clin Epi, J Clin Epi, 1992;45(6):613-619)1992;45(6):613-619)

• Quan’s Quan’s EnhancedEnhanced ICD-9-CMICD-9-CM• Quan’sQuan’s ICD-10 ICD-10

((Medical Care, Medical Care, 2005;43(11):1130-1139)2005;43(11):1130-1139)

• Elixhauser (30 comorbidity definitions)Elixhauser (30 comorbidity definitions)• Presence/absence, sumPresence/absence, sum

((Medical Care, Medical Care, 1998:36(1):8-27)1998:36(1):8-27)

• Quan’s Quan’s Enhanced ICD-9-CMEnhanced ICD-9-CM• Quan’s Quan’s ICD-10ICD-10

((Medical Care, Medical Care, 2005;43(11):1130-11392005;43(11):1130-1139))

Page 4: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Algorithms –Algorithms –development of ICD-10 & enhanced ICD-development of ICD-10 & enhanced ICD-

9-CM9-CM

• ICD-10 comorbidity coding algorithmICD-10 comorbidity coding algorithm• Based on Charlson indexBased on Charlson index

• Swiss, Australian, Canadian collaborative groupsSwiss, Australian, Canadian collaborative groups

• ICD-10 Canadian version (ICD-10-CA)ICD-10 Canadian version (ICD-10-CA)

• Enhanced ICD-9-CM coding algorithmEnhanced ICD-9-CM coding algorithm• Back-translated from new ICD-10 coding algorithmBack-translated from new ICD-10 coding algorithm

• To improve original Deyo (Charlson) and Elixhauser To improve original Deyo (Charlson) and Elixhauser comorbidity classificationscomorbidity classifications

Page 5: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Charlson Comorbidities with Corresponding ICD-9-CM and ICD-Charlson Comorbidities with Corresponding ICD-9-CM and ICD-10 Codes10 Codes

(Medical Care. (Medical Care. 2005;43:1130-11392005;43:1130-1139))

Page 6: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Example coding algorithmExample coding algorithm

ComorbidityComorbidity DeyoDeyo EnhancedEnhanced ICD-10ICD-10

Myocardial Myocardial infarctioninfarction

410.x, 412.x410.x, 412.x 410.x, 412.x410.x, 412.x I21.x, I22.x, I21.x, I22.x, I25.2I25.2

Congestive Congestive heart failureheart failure

428.x428.x 398.91, 398.91, 402.01, 402.01, 402.11, 402.11, 402.91, 402.91, 404.01, 404.01, 404.03, 404.03, 404.11, 404.11, 404.13, 404.13, 404.91, 404.91, 404.93, 425.4-404.93, 425.4-425.9, 428.x425.9, 428.x

I09.9, I11.0, I09.9, I11.0, I13.0, I13.2, I13.0, I13.2, I12.5, I42.0, I12.5, I42.0, I42.5-I42.9, I42.5-I42.9, I43.x, I40.x, I43.x, I40.x, P29.0P29.0

((Medical Care, Medical Care, 2005;43(11):1130-1139)2005;43(11):1130-1139)

Page 7: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Charlson Comorbidities & WeightsCharlson Comorbidities & Weights

CHARLSON COMORBIDITYCHARLSON COMORBIDITY ASSIGNED WEIGHTSASSIGNED WEIGHTS

1. Myocardial infarction1. Myocardial infarction 11

2. Congestive heart failure2. Congestive heart failure 11

3. Peripheral vascular disease3. Peripheral vascular disease 11

4. Cerebrovascular disease4. Cerebrovascular disease 11

5. Dementia5. Dementia 11

6. Chronic pulmonary disease6. Chronic pulmonary disease 11

7. Rheumatic disease7. Rheumatic disease 11

8. Peptic ulcer disease8. Peptic ulcer disease 11

9. Mild liver disease9. Mild liver disease 11

10. Diabetes without chronic 10. Diabetes without chronic complicationcomplication

11

11. Diabetes with end organ damage11. Diabetes with end organ damage 22

12. Hemiplegia / paraplegia12. Hemiplegia / paraplegia 22

13. Renal disease13. Renal disease 22

14. Any 14. Any malignancy/lymphoma/leukemiamalignancy/lymphoma/leukemia

22

15. Moderate or severe liver disease15. Moderate or severe liver disease 33

16. Metastatic solid tumor16. Metastatic solid tumor 66

17. AIDS/HIV17. AIDS/HIV 66

Page 8: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Input dataInput data

• Patient demographic dataPatient demographic data• ID variable (string) required if multiple visitsID variable (string) required if multiple visits

• Comorbidity diagnoses codes (strings)Comorbidity diagnoses codes (strings)•Charlson Charlson

– ICD-9-CM / ICD-10ICD-9-CM / ICD-10

•ElixhauserElixhauser– ICD-9-CM / ICD-10ICD-9-CM / ICD-10

• Additional medical informationAdditional medical information• For subsequent modeling, if desiredFor subsequent modeling, if desired

Page 9: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

SyntaxSyntax• CharlsonCharlson

•charlson charlson varlist varlist [[ifif exp exp]] [[in in rangerange], ], indexindex((stringstring) [) [idvaridvar((varnamevarname) ) diagprfxdiagprfx((stringstring) ) assign0 wtchrl cmorb assign0 wtchrl cmorb noshownoshow]]

by by may be used with may be used with charlsoncharlson

• ElixhauserElixhauser

•elixhauser elixhauser varlist varlist [[ifif exp exp]] [[in in rangerange], ], indexindex((stringstring) [) [idvaridvar((varnamevarname) ) diagprfxdiagprfx((stringstring) ) smelix cmorb noshowsmelix cmorb noshow]]

by by may be used with may be used with elixhauserelixhauser

Page 10: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Input optionsInput options

• index (index (stringstring))• ICD-9-CM (charlson)ICD-9-CM (charlson) cc

• Enhanced ICD-9-CM (charlson/elixhauser)Enhanced ICD-9-CM (charlson/elixhauser) ee

• ICD-10 (charlson/elixhauser)ICD-10 (charlson/elixhauser) 1010

• idvar(idvar(varnamevarname))• Required when multiple records per patientRequired when multiple records per patient

• diagprfx(diagprfx(stringstring))• Gives common root of the comorbidity variablesGives common root of the comorbidity variables

• Necessary only when Necessary only when varlist varlist not usednot used

• assign0 assign0 • Only applicable to charlsonOnly applicable to charlson

• Flag to apply hierarchical methodFlag to apply hierarchical method

Page 11: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Output optionsOutput options• wtchrl (charlson command)wtchrl (charlson command)

• Presents summary of Charlson Index (frequencies of Presents summary of Charlson Index (frequencies of weighted sums)weighted sums)

• wtelix (elixhauser command)wtelix (elixhauser command)• Displays frequencies of sum of elixhauser Displays frequencies of sum of elixhauser

comorbiditiescomorbidities

• cmorbcmorb• Displays frequencies of individual comorbiditiesDisplays frequencies of individual comorbidities

• noshownoshow• Controls display of chosen optionsControls display of chosen options

Page 12: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #1 – charlsonSample program #1 – charlson (Enhanced ICD-9-CM Algorithm) (Enhanced ICD-9-CM Algorithm)

• Input data (ICD-9-CM codes) -Input data (ICD-9-CM codes) -

Small Sample DataSmall Sample Datapatientipatienti

dddiag1diag1 diag2diag2 diag3diag3

id1id1 3989139891 58345834 342342

id2id2 09300930

id3id3 25002500 25072507

id4id4 342342 25002500 34413441

id5id5 34413441 342342

id6id6 25002500 57225722

id7id7 57225722 196196

id8id8 042042

id9id9 V427V427 45614561

id10id10 176176 197197 V427V427

Page 13: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #1 – charlsonSample program #1 – charlson

• Command –Command –

. charlson, index(e) diagprfx(diag) . charlson, index(e) diagprfx(diag) wtchrl cmorbwtchrl cmorb

Page 14: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #1 – charlson – Output (part Sample program #1 – charlson – Output (part 1)1)

• (Option noshow omitted)(Option noshow omitted)

(0 observations deleted)(0 observations deleted)COMORBIDITY INDEX MACROCOMORBIDITY INDEX MACROProviding COMORBIDITY INDEX SummaryProviding COMORBIDITY INDEX SummaryOPTIONS SELECTED: OPTIONS SELECTED: INPUT DATA: Enhanced ICD-9INPUT DATA: Enhanced ICD-9OBSERVATIONAL UNIT: VisitsOBSERVATIONAL UNIT: VisitsID VARIABLE NAME (Given only if Unit is Patients): ID VARIABLE NAME (Given only if Unit is Patients): PREFIX of COMORBIDITY VARIABLES: diagPREFIX of COMORBIDITY VARIABLES: diagHIERARCHY METHOD APPLIED: NOHIERARCHY METHOD APPLIED: NOSUMMARIZE CHARLSON INDEX and WEIGHTS: YESSUMMARIZE CHARLSON INDEX and WEIGHTS: YESSUMMARIZE INDIVIDUAL COMORBIDITIES: YESSUMMARIZE INDIVIDUAL COMORBIDITIES: YESPlease wait. Thank you!Please wait. Thank you!Program takes a few minutes - there are up to 3 ICD codes per Program takes a few minutes - there are up to 3 ICD codes per subject.subject.Iteration 1 of 3 - Program is running - Please waitIteration 1 of 3 - Program is running - Please waitIteration 2 of 3 - Program is running - Please waitIteration 2 of 3 - Program is running - Please waitIteration 3 of 3 - Program is running - Please waitIteration 3 of 3 - Program is running - Please waitTotal Number of Observational Units (Visits OR Patients): 10Total Number of Observational Units (Visits OR Patients): 10

Page 15: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #1 – charlson – Output Sample program #1 – charlson – Output (part 2)(part 2)

CHARLSON |CHARLSON | INDEX | Freq. Percent Cum.INDEX | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- 1 | 1 10.00 10.001 | 1 10.00 10.00 2 | 1 10.00 20.002 | 1 10.00 20.00 3 | 2 20.00 40.003 | 2 20.00 40.00 4 | 2 20.00 60.004 | 2 20.00 60.00 5 | 1 10.00 70.005 | 1 10.00 70.00 6 | 1 10.00 80.006 | 1 10.00 80.00 9 | 2 20.00 100.009 | 2 20.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

GROUPED |GROUPED | CHARLSON |CHARLSON | INDEX | Freq. Percent Cum.INDEX | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- 1 | 1 10.00 10.001 | 1 10.00 10.00 2 | 9 90.00 100.002 | 9 90.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

Variable | Obs Mean Std. Dev. Min MaxVariable | Obs Mean Std. Dev. Min Max-------------+---------------------------------------------------------------------+-------------------------------------------------------- charlindex | 10 4.6 2.716207 1 9charlindex | 10 4.6 2.716207 1 9

• (option (option wtchrl) wtchrl)

Page 16: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #1 – charlson – Output Sample program #1 – charlson – Output (part 3)(part 3)

Diabetes | Freq. Percent Cum.Diabetes | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- Absent | 7 70.00 70.00Absent | 7 70.00 70.00 Present | 3 30.00 100.00Present | 3 30.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

Diabetes + |Diabetes + |Complicatio |Complicatio | ns | Freq. Percent Cum.ns | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- Absent | 9 90.00 90.00Absent | 9 90.00 90.00 Present | 1 10.00 100.00Present | 1 10.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

• (option cmorb)(option cmorb)• selected comorbidities displayedselected comorbidities displayed

Page 17: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Output dataset – describe Output dataset – describe obs: 10 obs: 10 vars: 41 17 Oct 2007 10:08vars: 41 17 Oct 2007 10:08 size: 1,880 (99.9% of memory free)size: 1,880 (99.9% of memory free)-------------------------------------------------------------------------------------------------------------------------------------------------------------- storage display valuestorage display valuevariable name type format label variable labelvariable name type format label variable label--------------------------------------------------------------------------------------------------------------------------------------------------------------id str23 %23s id str23 %23s diag1 str5 %9s diag1 str5 %9s diag2 str4 %9s diag2 str4 %9s diag3 str4 %9s diag3 str4 %9s ynch1 float %9.0g ynlab AMI (Acute Myocardial)ynch1 float %9.0g ynlab AMI (Acute Myocardial)ynch2 float %9.0g ynlab CHF (Congestive Heart)ynch2 float %9.0g ynlab CHF (Congestive Heart)ynch3 float %9.0g ynlab PVD (Peripheral Vascular)ynch3 float %9.0g ynlab PVD (Peripheral Vascular)ynch4 float %9.0g ynlab CEVD (Cerebrovascularynch4 float %9.0g ynlab CEVD (Cerebrovascularynch5 float %9.0g ynlab Dementiaynch5 float %9.0g ynlab Dementiaynch6 float %9.0g ynlab COPD (Chronic Obstructiveynch6 float %9.0g ynlab COPD (Chronic Obstructive Pulmonary)Pulmonary)ynch7 float %9.0g ynlab Rheumatoid Diseaseynch7 float %9.0g ynlab Rheumatoid Diseaseynch8 float %9.0g ynlab PUD (Peptic Ulcer)ynch8 float %9.0g ynlab PUD (Peptic Ulcer)ynch9 float %9.0g ynlab Mild LD (Liver)ynch9 float %9.0g ynlab Mild LD (Liver)ynch10 float %9.0g ynlab Diabetesynch10 float %9.0g ynlab Diabetesynch11 float %9.0g ynlab Diabetes + Complicationsynch11 float %9.0g ynlab Diabetes + Complicationsynch12 float %9.0g ynlab HP/PAPL (Hemiplegia orynch12 float %9.0g ynlab HP/PAPL (Hemiplegia or Paraplegia)Paraplegia)

Page 18: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Output dataset – describe continuedOutput dataset – describe continuedynch13 float %9.0g ynlab RD (Renal)ynch13 float %9.0g ynlab RD (Renal)ynch14 float %9.0g ynlab Cancerynch14 float %9.0g ynlab Cancerynch15 float %9.0g ynlab Moderate/Severe LD (Liver)ynch15 float %9.0g ynlab Moderate/Severe LD (Liver)ynch16 float %9.0g ynlab Metastic Cancerynch16 float %9.0g ynlab Metastic Cancerynch17 float %9.0g ynlab AIDSynch17 float %9.0g ynlab AIDSweightch1 float %9.0g weightch1 float %9.0g weightch2 float %9.0g weightch2 float %9.0g weightch3 float %9.0g weightch3 float %9.0g weightch4 float %9.0g weightch4 float %9.0g weightch5 float %9.0g weightch5 float %9.0g weightch6 float %9.0g weightch6 float %9.0g weightch7 float %9.0g weightch7 float %9.0g weightch8 float %9.0g weightch8 float %9.0g weightch9 float %9.0g weightch9 float %9.0g weightch10 float %9.0g weightch10 float %9.0g weightch11 float %9.0g weightch11 float %9.0g weightch12 float %9.0g weightch12 float %9.0g weightch13 float %9.0g weightch13 float %9.0g weightch14 float %9.0g weightch14 float %9.0g weightch15 float %9.0g weightch15 float %9.0g weightch16 float %9.0g weightch16 float %9.0g weightch17 float %9.0g weightch17 float %9.0g charlindex float %9.0g CHARLSON INDEXcharlindex float %9.0g CHARLSON INDEXgrpci float %9.0g GROUPED CHARLSON INDEXgrpci float %9.0g GROUPED CHARLSON INDEX--------------------------------------------------------------------------------------------------------------------------------------------------------------Sorted by: Sorted by: Note: dataset has changed since last savedNote: dataset has changed since last saved

Page 19: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Output dataset – selected variablesOutput dataset – selected variables

. list id ynch10 ynch11 ynch15. list id ynch10 ynch11 ynch15

+------------------------------------++------------------------------------+ | id ynch10 ynch11 ynch15 || id ynch10 ynch11 ynch15 | |------------------------------------||------------------------------------| 1. | id1 Absent Absent Absent |1. | id1 Absent Absent Absent | 2. | id2 Absent Absent Absent |2. | id2 Absent Absent Absent | 3. | 3. | id3 Present Presentid3 Present Present Absent | Absent | 4. | id4 Present Absent Absent |4. | id4 Present Absent Absent | 5. | id5 Absent Absent Absent |5. | id5 Absent Absent Absent | |------------------------------------||------------------------------------| 6. | id6 Present Absent Present |6. | id6 Present Absent Present | 7. | id7 Absent Absent Present |7. | id7 Absent Absent Present | 8. | id8 Absent Absent Absent |8. | id8 Absent Absent Absent | 9. | id9 Absent Absent Present |9. | id9 Absent Absent Present | 10. | id10 Absent Absent Absent |10. | id10 Absent Absent Absent | +------------------------------------++------------------------------------+

. list id weightch10 weightch11 . list id weightch10 weightch11 weightch15, cweightch15, c

+------------------------------++------------------------------+ | id we~10 we~11 we~15 || id we~10 we~11 we~15 | |------------------------------||------------------------------| 1. | id1 0 0 0 |1. | id1 0 0 0 | 2. | id2 0 0 0 |2. | id2 0 0 0 | 3. | 3. | id3 1 2id3 1 2 0 | 0 | 4. | id4 1 0 0 |4. | id4 1 0 0 | 5. | id5 0 0 0 |5. | id5 0 0 0 | |------------------------------||------------------------------| 6. | id6 1 0 3 |6. | id6 1 0 3 | 7. | id7 0 0 3 |7. | id7 0 0 3 | 8. | id8 0 0 0 |8. | id8 0 0 0 | 9. | id9 0 0 3 |9. | id9 0 0 3 | 10. | id10 0 0 0 |10. | id10 0 0 0 | +------------------------------++------------------------------+

Page 20: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Output dataset –Output dataset – Charlson index & grouped Charlson indexCharlson index & grouped Charlson index

. list id charlindex grpci. list id charlindex grpci

+-------------------------++-------------------------+ | id charli~x grpci || id charli~x grpci | |-------------------------||-------------------------| 1. | id1 5 2 |1. | id1 5 2 | 2. | id2 1 1 |2. | id2 1 1 | 3. | 3. | id3 3 2id3 3 2 | | 4. | id4 3 2 |4. | id4 3 2 | 5. | id5 2 2 |5. | id5 2 2 | |-------------------------||-------------------------| 6. | id6 4 2 |6. | id6 4 2 | 7. | id7 9 2 |7. | id7 9 2 | 8. | id8 6 2 |8. | id8 6 2 | 9. | 9. | id9 4 2id9 4 2 | | 10. | 10. | id10 9 2id10 9 2 | | +-------------------------++-------------------------+

Page 21: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Program rerun with assign0 option –Program rerun with assign0 option –(changes frequencies)(changes frequencies)

. comorbid, index(e) diagprfx(diag) wtchrl cmorb assign0. comorbid, index(e) diagprfx(diag) wtchrl cmorb assign0

CHARLSON |CHARLSON | INDEX | Freq. Percent Cum.INDEX | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- 1 | 1 10.00 10.001 | 1 10.00 10.00 2 | 2 20.00 30.002 | 2 20.00 30.00 3 | 2 20.00 50.003 | 2 20.00 50.00 4 | 1 10.00 60.004 | 1 10.00 60.00 5 | 1 10.00 70.005 | 1 10.00 70.00 6 | 1 10.00 80.006 | 1 10.00 80.00 7 | 1 10.00 90.007 | 1 10.00 90.00 9 | 1 10.00 100.009 | 1 10.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

GROUPED |GROUPED | CHARLSON |CHARLSON | INDEX | Freq. Percent Cum.INDEX | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- 1 | 1 10.00 10.001 | 1 10.00 10.00 2 | 9 90.00 100.002 | 9 90.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

+-------------------------++-------------------------+ | id charli~x grpci || id charli~x grpci | |-------------------------||-------------------------| 1. | id1 5 2 |1. | id1 5 2 | 2. | id2 1 1 |2. | id2 1 1 | 3. | 3. | id3 2 2id3 2 2 | | 4. | id4 3 2 |4. | id4 3 2 | 5. | id5 2 2 |5. | id5 2 2 | |-------------------------||-------------------------| 6. | id6 4 2 |6. | id6 4 2 | 7. | id7 9 2 |7. | id7 9 2 | 8. | id8 6 2 |8. | id8 6 2 | 9. | 9. | id9 3 2id9 3 2 | | 10. | 10. | id10 7 2id10 7 2 | | +-------------------------++-------------------------+

Page 22: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Selected comorbidities revisited-Selected comorbidities revisited-

Diabetes | Freq. Percent Cum.Diabetes | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- Absent | 8 80.00 80.00Absent | 8 80.00 80.00 Present | 2 20.00 100.00Present | 2 20.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

Diabetes + |Diabetes + |Complicatio |Complicatio | ns | Freq. Percent Cum.ns | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- Absent | 9 90.00 90.00Absent | 9 90.00 90.00 Present | 1 10.00 100.00Present | 1 10.00 100.00------------+-----------------------------------------------+----------------------------------- Total | 10 100.00Total | 10 100.00

Page 23: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #2 – Sample program #2 – elixhauserelixhauser

(ICD-10 Algorithm)(ICD-10 Algorithm)• Input - real inpatient data -Input - real inpatient data - obs: 2,987 obs: 2,987 vars: 43 18 Oct 2007 10:18vars: 43 18 Oct 2007 10:18 size: 1,000,645 (90.5% of memory free)size: 1,000,645 (90.5% of memory free)-------------------------------------------------------------------------------------------------------------------------------------------------------------- storage display valuestorage display valuevariable name type format label variable labelvariable name type format label variable label--------------------------------------------------------------------------------------------------------------------------------------------------------------dx1 str6 %9s DIAG1dx1 str6 %9s DIAG1dx2 str6 %9s DIAG2dx2 str6 %9s DIAG2. . . . . . dx24 str6 %9s DIAG24dx24 str6 %9s DIAG24dx25 str6 %9s DIAG25dx25 str6 %9s DIAG25cdr_keyforqsh~e long %12.0g CDR_KEY (for QSHI use)cdr_keyforqsh~e long %12.0g CDR_KEY (for QSHI use)admitdate str20 %20s Admit Dateadmitdate str20 %20s Admit Datedischargedate str20 %20s Discharge Datedischargedate str20 %20s Discharge Dateacutelosdays int %8.0g ACUTE LOS (days)acutelosdays int %8.0g ACUTE LOS (days)birthdate str11 %11s Birth Datebirthdate str11 %11s Birth Dateage int %8.0g AGEage int %8.0g AGEpc str6 %9s PCpc str6 %9s PCresidence str7 %9s RESIDENCEresidence str7 %9s RESIDENCEentrycodetoho~l str61 %61s ENTRY CODE to hospitalentrycodetoho~l str61 %61s ENTRY CODE to hospitalstrokediagtyp~a str25 %25s Stroke Diag Type when Strokestrokediagtyp~a str25 %25s Stroke Diag Type when Stroke not the Main Diagnot the Main Diaggender long %8.0g gender gendergender long %8.0g gender gendersite long %8.0g site sitesite long %8.0g site sitestroketype long %13.0g stroke Stroke typestroketype long %13.0g stroke Stroke typedisposition long %60.0g disp discharge dispositiondisposition long %60.0g disp discharge dispositioncohort float %9.0g cohort cohortcohort float %9.0g cohort cohort--------------------------------------------------------------------------------------------------------------------------------------------------------------

Page 24: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #2 – Sample program #2 – elixhauserelixhauser

• Command –Command –

. elixhauser dx1-dx25, index(10) . elixhauser dx1-dx25, index(10) smelix cmorbsmelix cmorb

Page 25: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

Sample program #2 – elixhauserSample program #2 – elixhauserOutput Output

ELIX |ELIX |COMORBIDITY |COMORBIDITY | SUM | Freq. Percent Cum.SUM | Freq. Percent Cum.------------+-----------------------------------------------+----------------------------------- 0 | 402 13.46 13.460 | 402 13.46 13.46 1 | 715 23.94 37.401 | 715 23.94 37.40 2 | 751 25.14 62.542 | 751 25.14 62.54 3 | 529 17.71 80.253 | 529 17.71 80.25 4 | 303 10.14 90.394 | 303 10.14 90.39 5 | 174 5.83 96.225 | 174 5.83 96.22 6 | 71 2.38 98.596 | 71 2.38 98.59 7 | 30 1.00 99.607 | 30 1.00 99.60 8 | 10 0.33 99.938 | 10 0.33 99.93 9 | 1 0.03 99.979 | 1 0.03 99.97 10 | 1 0.03 100.0010 | 1 0.03 100.00------------+-----------------------------------------------+----------------------------------- Total | 2,987 100.00Total | 2,987 100.00

Variable | Obs Mean Std. Dev. Min MaxVariable | Obs Mean Std. Dev. Min Max-------------+---------------------------------------------------------------------+-------------------------------------------------------- elixsum | 2987 2.216605 1.621281 0 10elixsum | 2987 2.216605 1.621281 0 10

Page 26: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

ELIXHAUSER COMORBIDITYELIXHAUSER COMORBIDITY PERCENTPERCENT

Congestive Heart FailureCongestive Heart Failure 8.208.20

Cardiac ArrhythmiasCardiac Arrhythmias 22.2322.23

Valvular DiseaseValvular Disease 4.494.49

Pulmonary Circulation DisordersPulmonary Circulation Disorders 1.771.77

Peripheral Vascular DisordersPeripheral Vascular Disorders 5.495.49

Hypertension, UncomplicatedHypertension, Uncomplicated 58.4958.49

ParalysisParalysis 25.6825.68

Other Neurological DisordersOther Neurological Disorders 23.8423.84

Chronic Pulmonary DiseaseChronic Pulmonary Disease 7.207.20

Diabetes, UncomplicatedDiabetes, Uncomplicated 15.8015.80

Diabetes, ComplicatedDiabetes, Complicated 4.084.08

HypothyroidismHypothyroidism 3.853.85

Renal FailureRenal Failure 4.694.69

Liver DiseaseLiver Disease 0.770.77

Peptic Ulcer Disease Excluding BleedingPeptic Ulcer Disease Excluding Bleeding 0.400.40

AIDS/HIVAIDS/HIV 0.130.13

Page 27: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

ELIXHAUSER COMORBIDITYELIXHAUSER COMORBIDITY PERCENTPERCENT

LymphomaLymphoma 0.570.57

Metastatic CancerMetastatic Cancer 1.611.61

Solid Tumor Without MetastasisSolid Tumor Without Metastasis 2.982.98

Rheumatoid Arthritis/Collagen VascularRheumatoid Arthritis/Collagen Vascular 1.471.47

CoagulopathyCoagulopathy 2.682.68

ObesityObesity 3.313.31

Weight LossWeight Loss 0.670.67

Fluid and Electrolyte DisordersFluid and Electrolyte Disorders 6.466.46

Blood Loss AnemiaBlood Loss Anemia 0.500.50

Deficiency AnemiaDeficiency Anemia 1.311.31

Alcohol AbuseAlcohol Abuse 3.353.35

Drug AbuseDrug Abuse 0.740.74

PsychosesPsychoses 0.500.50

DepressionDepression 4.594.59

Hypertension, ComplicatedHypertension, Complicated 3.823.82

……continuedcontinued

Page 28: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

AcknowledgementsAcknowledgementsI would like to express sincere gratitude to:I would like to express sincere gratitude to:

• Dr. Robert Hilsden Dr. Robert Hilsden Depts. of Medicine/ Community Health Sciences, U of CalgaryDepts. of Medicine/ Community Health Sciences, U of CalgaryFor supervising this work and for all his advice and support.For supervising this work and for all his advice and support.

• Dr. Hude QuanDr. Hude QuanCentre for Health and Policy StudiesCentre for Health and Policy StudiesDept. of Community Health Sciences, U of CalgaryDept. of Community Health Sciences, U of CalgaryFor providing the SAS code and databases and for his support.For providing the SAS code and databases and for his support.

• Haifeng ZhuHaifeng ZhuMSc Graduate StudentMSc Graduate StudentDept. of Community Health SciencesDept. of Community Health SciencesFor her assistance with converting the Elixhauser algorithms to Stata.For her assistance with converting the Elixhauser algorithms to Stata.

• Malcolm StaggMalcolm StaggStudent, Vista Virtual School, Calgary AB Student, Vista Virtual School, Calgary AB My son, for his help with preparing this PowerPoint presentation and all his My son, for his help with preparing this PowerPoint presentation and all his encouragement.encouragement.

• Andrew StaggAndrew StaggIntern, Google Inc., Mountain View CAIntern, Google Inc., Mountain View CAMy son, for his encouragement.My son, for his encouragement.

Page 29: Calculating Measures of Comorbidity Using Administrative Data Vicki Stagg Statistical Programmer Department of Community Health Sciences Dr. Robert Hilsden

SUGGESTIONS / SUGGESTIONS / COMMENTSCOMMENTSWELCOMEWELCOME

[email protected]@ucalgary.ca

Thank you!Thank you!