Presentation is loading. Please wait.

Presentation is loading. Please wait.

Calculating Measures of Comorbidity Using Administrative Data

Similar presentations


Presentation on theme: "Calculating Measures of Comorbidity Using Administrative Data"— Presentation transcript:

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

2 Background Medical Administrative Data Comorbidity Comorbidity Index
Inpatient hospital visit information Comorbidity Pre-existing diagnosis / additional complication of admitted patient Comorbidity Index For measurement of burden of disease and case-mix adjustment Allows for stratification or adjustment by severity of illness Two common tools – Charlson and Elixhauser Clinical Conditions International Classification of Disease 9th Revision, Clinical Modification (ICD-9-CM codes) 10th Revision (ICD-10 codes) (International statistical classification of disease and related health problems, 1992) Medical administrative data contain anonymized records of inpatient hospital visit information for use by provincial health care departments A comorbidity is a pre-existing diagnosis of an admitted patient, which could potentially negatively effect the hospital outcome Burden of disease – impact of illnesses on society; can be measured in years of life lost due to the disease Case-mix adjustment – process of adjusting for the severity of disease Stratification Two common tools – different classes/ categories of comorbidites and definitions Clinical conditions are summarised in the International Classification of Disease, 9th Revision, Clinical Modification which contains the ICD-9-CM codes; the 10th Revision give the ICD-10 codes Various comorbidity coding algorithms have been developed to effectively define and summarize comorbidities. These are based on the different disease codes. The Enhanced ICD-9-CM was developed to resolve discrepancies between coding algorithms for some condittions, by redefining those comorbidities. SAS to Stata

3 Algorithms included in ado programs
Charlson (17 comorbidity definitions) Presence/absence, weighted sum (Charlson index) Charlson index developed to predict risk of one-year mortality from comorbid illness (J Chron Dis, 1987;40(5): ) Deyo modification for ICD-9-CM (J Clin Epi, 1992;45(6): ) Quan’s Enhanced ICD-9-CM Quan’s ICD-10 (Medical Care, 2005;43(11): ) Elixhauser (30 comorbidity definitions) Presence/absence, sum (Medical Care, 1998:36(1):8-27)

4 Algorithms – development of ICD-10 & enhanced ICD-9-CM
ICD-10 comorbidity coding algorithm Based on Charlson index Swiss, Australian, Canadian collaborative groups ICD-10 Canadian version (ICD-10-CA) Enhanced ICD-9-CM coding algorithm Back-translated from new ICD-10 coding algorithm To improve original Deyo (Charlson) and Elixhauser comorbidity classifications 3 international research groups

5 Charlson Comorbidities with Corresponding ICD-9-CM and ICD-10 Codes (Medical Care. 2005;43: )

6 Example coding algorithm
Comorbidity Deyo Enhanced ICD-10 Myocardial infarction 410.x, 412.x I21.x, I22.x, I25.2 Congestive heart failure 428.x 398.91, , , , , , , , , , , 428.x I09.9, I11.0, I13.0, I13.2, I12.5, I42.0, I42.5-I42.9, I43.x, I40.x, P29.0 (Medical Care, 2005;43(11): )

7 Charlson Comorbidities & Weights
CHARLSON COMORBIDITY ASSIGNED WEIGHTS Myocardial infarction 1 Congestive heart failure Peripheral vascular disease Cerebrovascular disease Dementia Chronic pulmonary disease Rheumatic disease Peptic ulcer disease Mild liver disease Diabetes without chronic complication Diabetes with end organ damage 2 Hemiplegia / paraplegia Renal disease Any malignancy/lymphoma/leukemia Moderate or severe liver disease 3 Metastatic solid tumor 6 AIDS/HIV Weighted scores are related to relative risks determined for each illness (<1/2 were deleted) and reflect the seriousness of the disease

8 Input data Patient demographic data
ID variable (string) required if multiple visits Comorbidity diagnoses codes (strings) Charlson ICD-9-CM / ICD-10 Elixhauser Additional medical information For subsequent modeling, if desired Input database can contain patient demographic data as well as the comorbidity codes or simply the codes for each patient. If some patients have multiple records, a patient ID variable is required. Additional

9 Syntax Charlson Elixhauser
charlson varlist [if exp] [in range], index(string) [idvar(varname) diagprfx(string) assign0 wtchrl cmorb noshow] by may be used with charlson Elixhauser elixhauser varlist [if exp] [in range], index(string) [idvar(varname) diagprfx(string) smelix cmorb noshow] by may be used with elixhauser

10 Input options index (string) idvar(varname) diagprfx(string) assign0
ICD-9-CM (charlson) c Enhanced ICD-9-CM (charlson/elixhauser) e ICD-10 (charlson/elixhauser) 10 idvar(varname) Required when multiple records per patient diagprfx(string) Gives common root of the comorbidity variables Necessary only when varlist not used assign0 Only applicable to charlson Flag to apply hierarchical method Index identifies algorithm to be used. Hierarchical method is applied when it is desired to exclude from the calculations a mild comorbidity when a patient also exhibits a more severe form of the diagnosis.

11 Output options wtchrl (charlson command) wtelix (elixhauser command)
Presents summary of Charlson Index (frequencies of weighted sums) wtelix (elixhauser command) Displays frequencies of sum of elixhauser comorbidities cmorb Displays frequencies of individual comorbidities noshow Controls display of chosen options

12 Sample program #1 – charlson (Enhanced ICD-9-CM Algorithm)
Input data (ICD-9-CM codes) - Small Sample Data patientid diag1 diag2 diag3 id1 39891 5834 342 id2 0930 id3 2500 2507 id4 3441 id5 id6 5722 id7 196 id8 042 id9 V427 4561 id10 176 197

13 Sample program #1 – charlson
Command – . charlson, index(e) diagprfx(diag) wtchrl cmorb

14 Sample program #1 – charlson – Output (part 1)
(Option noshow omitted) (0 observations deleted) COMORBIDITY INDEX MACRO Providing COMORBIDITY INDEX Summary OPTIONS SELECTED: INPUT DATA: Enhanced ICD-9 OBSERVATIONAL UNIT: Visits ID VARIABLE NAME (Given only if Unit is Patients): PREFIX of COMORBIDITY VARIABLES: diag HIERARCHY METHOD APPLIED: NO SUMMARIZE CHARLSON INDEX and WEIGHTS: YES SUMMARIZE INDIVIDUAL COMORBIDITIES: YES Please wait. Thank you! Program takes a few minutes - there are up to 3 ICD codes per subject. Iteration 1 of 3 - Program is running - Please wait Iteration 2 of 3 - Program is running - Please wait Iteration 3 of 3 - Program is running - Please wait Total Number of Observational Units (Visits OR Patients): 10

15 Sample program #1 – charlson – Output (part 2)
INDEX | Freq. Percent Cum. 1 | 2 | 3 | 4 | 5 | 6 | 9 | Total | GROUPED | 2 | Variable | Obs Mean Std. Dev Min Max charlindex | (option wtchrl)

16 Sample program #1 – charlson – Output (part 3)
(option cmorb) selected comorbidities displayed Diabetes | Freq. Percent Cum. Absent | Present | Total | Diabetes + | Complicatio | ns | Freq. Percent Cum. Absent | Present |

17 Output dataset – describe
obs: vars: Oct :08 size: ,880 (99.9% of memory free) storage display value variable name type format label variable label id str23 %23s diag str5 %9s diag str4 %9s diag str4 %9s ynch float %9.0g ynlab AMI (Acute Myocardial) ynch float %9.0g ynlab CHF (Congestive Heart) ynch float %9.0g ynlab PVD (Peripheral Vascular) ynch float %9.0g ynlab CEVD (Cerebrovascular ynch float %9.0g ynlab Dementia ynch float %9.0g ynlab COPD (Chronic Obstructive Pulmonary) ynch float %9.0g ynlab Rheumatoid Disease ynch float %9.0g ynlab PUD (Peptic Ulcer) ynch float %9.0g ynlab Mild LD (Liver) ynch float %9.0g ynlab Diabetes ynch float %9.0g ynlab Diabetes + Complications ynch float %9.0g ynlab HP/PAPL (Hemiplegia or Paraplegia) When patient is the observational unit – only last visit data retained (warning – sort by visit before running program)

18 Output dataset – describe continued
ynch float %9.0g ynlab RD (Renal) ynch float %9.0g ynlab Cancer ynch float %9.0g ynlab Moderate/Severe LD (Liver) ynch float %9.0g ynlab Metastic Cancer ynch float %9.0g ynlab AIDS weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g weightch float %9.0g charlindex float %9.0g CHARLSON INDEX grpci float %9.0g GROUPED CHARLSON INDEX Sorted by: Note: dataset has changed since last saved

19 Output dataset – selected variables
. list id ynch10 ynch11 ynch15 | id ynch10 ynch11 ynch15 | | | 1. | id1 Absent Absent Absent | 2. | id2 Absent Absent Absent | 3. | id3 Present Present Absent | 4. | id4 Present Absent Absent | 5. | id5 Absent Absent Absent | 6. | id6 Present Absent Present | 7. | id7 Absent Absent Present | 8. | id8 Absent Absent Absent | 9. | id9 Absent Absent Present | 10. | id10 Absent Absent Absent | . list id weightch10 weightch11 weightch15, c | id we~10 we~11 we~15 | | | 1. | id | 2. | id | 3. | id | 4. | id | 5. | id | 6. | id | 7. | id | 8. | id | 9. | id | 10. | id |

20 Output dataset – Charlson index & grouped Charlson index
. list id charlindex grpci | id charli~x grpci | | | 1. | id | 2. | id | 3. | id | 4. | id | 5. | id | 6. | id | 7. | id | 8. | id | 9. | id | 10. | id |

21 Program rerun with assign0 option – (changes frequencies)
. comorbid, index(e) diagprfx(diag) wtchrl cmorb assign0 CHARLSON | INDEX | Freq. Percent Cum. 1 | 2 | 3 | 4 | 5 | 6 | 7 | 9 | Total | GROUPED | 2 | | id charli~x grpci | | | 1. | id | 2. | id | 3. | id | 4. | id | 5. | id | 6. | id | 7. | id | 8. | id | 9. | id | 10. | id |

22 Selected comorbidities revisited-
Diabetes | Freq. Percent Cum. Absent | Present | Total | Diabetes + | Complicatio | ns | Freq. Percent Cum. Absent | Present |

23 Sample program #2 – elixhauser (ICD-10 Algorithm)
Input - real inpatient data - obs: ,987 vars: Oct :18 size: 1,000,645 (90.5% of memory free) storage display value variable name type format label variable label dx str6 %9s DIAG1 dx str6 %9s DIAG2 . . . dx str6 %9s DIAG24 dx str6 %9s DIAG25 cdr_keyforqsh~e long %12.0g CDR_KEY (for QSHI use) admitdate str20 %20s Admit Date dischargedate str20 %20s Discharge Date acutelosdays int %8.0g ACUTE LOS (days) birthdate str11 %11s Birth Date age int %8.0g AGE pc str6 %9s PC residence str7 %9s RESIDENCE entrycodetoho~l str61 %61s ENTRY CODE to hospital strokediagtyp~a str25 %25s Stroke Diag Type when Stroke not the Main Diag gender long %8.0g gender gender site long %8.0g site site stroketype long %13.0g stroke Stroke type disposition long %60.0g disp discharge disposition cohort float %9.0g cohort cohort

24 Sample program #2 – elixhauser
Command – . elixhauser dx1-dx25, index(10) smelix cmorb

25 Sample program #2 – elixhauser Output
COMORBIDITY | SUM | Freq. Percent Cum. 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | Total | , Variable | Obs Mean Std. Dev Min Max elixsum |

26 ELIXHAUSER COMORBIDITY
PERCENT Congestive Heart Failure 8.20 Cardiac Arrhythmias 22.23 Valvular Disease 4.49 Pulmonary Circulation Disorders 1.77 Peripheral Vascular Disorders 5.49 Hypertension, Uncomplicated 58.49 Paralysis 25.68 Other Neurological Disorders 23.84 Chronic Pulmonary Disease 7.20 Diabetes, Uncomplicated 15.80 Diabetes, Complicated 4.08 Hypothyroidism 3.85 Renal Failure 4.69 Liver Disease 0.77 Peptic Ulcer Disease Excluding Bleeding 0.40 AIDS/HIV 0.13

27 ELIXHAUSER COMORBIDITY
…continued ELIXHAUSER COMORBIDITY PERCENT Lymphoma 0.57 Metastatic Cancer 1.61 Solid Tumor Without Metastasis 2.98 Rheumatoid Arthritis/Collagen Vascular 1.47 Coagulopathy 2.68 Obesity 3.31 Weight Loss 0.67 Fluid and Electrolyte Disorders 6.46 Blood Loss Anemia 0.50 Deficiency Anemia 1.31 Alcohol Abuse 3.35 Drug Abuse 0.74 Psychoses Depression 4.59 Hypertension, Complicated 3.82

28 Acknowledgements Dr. Robert Hilsden Dr. Hude Quan
I would like to express sincere gratitude to: Dr. Robert Hilsden Depts. of Medicine/ Community Health Sciences, U of Calgary For supervising this work and for all his advice and support. Dr. Hude Quan Centre for Health and Policy Studies Dept. of Community Health Sciences, U of Calgary For providing the SAS code and databases and for his support. Haifeng Zhu MSc Graduate Student Dept. of Community Health Sciences For her assistance with converting the Elixhauser algorithms to Stata. Malcolm Stagg Student, Vista Virtual School, Calgary AB My son, for his help with preparing this PowerPoint presentation and all his encouragement. Andrew Stagg Intern, Google Inc., Mountain View CA My son, for his encouragement.

29 SUGGESTIONS / COMMENTS WELCOME
Thank you!


Download ppt "Calculating Measures of Comorbidity Using Administrative Data"

Similar presentations


Ads by Google