Using Administrative Data in Health Research
William Ghali, MD, MPH Faculty of Medicine University of Calgary
Overview
Historical perspective Conceptual model for studying outcomes and processes of care Administrative data Comorbidity measures for defining severity of illness in administrative data ICD-9 and the transition to ICD-10
A global methodological challenge
Overview of today’s symposium
Historical perspective
Florence Nightingale (1820-1910)
Observed that death rates were lower in hospitals with better sanitation “accurate hospital statistics are much more rare than is generally imagined” “because health is the ultimate object of hospital care, statistics should concentrate on recovery and its speed”
(clearly ahead of her time…)
Historical perspective
Ernest Codman (1869-1940) - Surgeon, MGH, Boston “so I am called eccentric for saying in public that hospitals, if they wish to be sure of improvement must find out what their results are must analyze their results to find their strong and weak points must compare their results with those of others must welcome publicity, not for their successes, but for their errors Such opinions will not be eccentric a few years hence” 1917
Historical perspective
Both Nightingale and Codman allude to value of studying outcomes of care Nightingale also links outcomes to process of care Both are indicators of quality of care
A conceptual model for studying quality of care
Outcome = fx {severity of illness + quality}
Structure
Process
A conceptual model for studying quality of care
Outcome = fx {severity of illness + quality}
Structure
Process
Health Care Financing Administration
U.S. Medicare program 1986: release of hospital-specific mortality rates several medical and surgical conditions studied clinical information extracted from ICD-9 codes and attempts at risk adjustment made Criticisms:
Incomplete measurement of severity of illness Administrative data may not be sufficiently detailed
Administrative data
Many types
Hospital discharge data Physician claims data Pharmacy claims data Hospital equipment purchasing data Other data
Widely used in health research Generation and storage of hospital discharge data in Canada…
Hospital discharge data
Administrative data
Hospital discharge data
Typical content:
– – – – – – – – Unique identifier Age Sex Diagnostic information (ICD-9 or ICD-10) Procedure information (either ICD or other) Length of stay Use of special care units (Costing information)
Administrative data
Hospital discharge data
Document health care contacts with hospital facilities (complete capture) Low cost source of clinical information Contain useful outcome information
In-hospital death Length of stay Special care unit use Readmission
As a result, used in many health “outcomes research” initiatives
Administrative data
Hospital discharge data But:
– Clinical information in administrative data may not be entirely valid nor complete – Sequencing of diagnoses may be difficult to determine – Can see undercoding of secondary diagnoses in patients who die – Lack of important prognostic variables (e.g., ventricular ejection fraction in heart surgery)
Extracting clinical information from administrative data
Important methodological work by
– Charlson et al. / Deyo et al. – Elixhauser et al.
Charlson Index Comorbidity (J Chron Dis 1987,40:337)
1. AIDS 2. Cerebrovascular disease 3. Chronic pulmonary disease 4. Congestive heart failure 5. Dementia 6. Diabetes 7. Diabetes with complication 8. Hemiplegia or paraplegia 9. Malignancy 10. Metastatic solid tumor 11. Mild liver disease 12. Moderate or severe liver disease 13. Myocardial infarction 14. Peptic ulcer disease 15. Peripheral vascular disease 16. Renal disease 17. Rheumatologic disease
Charlson Index Score (J Chron Dis 1987,40:337) Weight
6 1 1 1 1 1 2 2 2 6 1 3 1 1 1 2 1 AIDS Cerebrovascular disease Chronic pulmonary disease Congestive heart failure Dementia Diabetes Diabetes with complication Hemiplegia or paraplegia Malignancy Metastatic solid tumor Mild liver disease Moderate or severe liver disease Myocardial infarction Peptic ulcer disease Peripheral vascular disease Renal disease Rheumatologic disease
0 - 33
Charlson Comorbidity Index (J Chron Dis 1987,40:337)
1. AIDS 2. Cerebrovascular disease 3. Chronic pulmonary disease 4. Congestive heart failure 5. Dementia 6. Diabetes 7. Diabetes with complication 8. Hemiplegia or paraplegia 9. Malignancy 10. Metastatic solid tumor 11. Mild liver disease 12. Moderate or severe liver disease 13. Myocardial infarction 14. Peptic ulcer disease 15. Peripheral vascular disease 16. Renal disease 17. Rheumatologic disease
Deyo’s ICD9CM Codes (J Clin Epidemiol 1992,45:613)
1. 042-044.9 2. 430-438 3. 490-496, 500-505, 506.4 4. 428 5. 290 6. 250-250.3, 250.7Diabetes 7. 250.4-250.6 8. 344.1, 342-342.9 9. 140-172.9, 174-195.8,200-208.9 10. 196-199.1 11. 571.2, 571.5, 571.4-571.49 12. 572.2-572.8 13. 410, 412 14. 531-534, 531.4-531.7,534.4 15. 443.9,441,785.4,V43.4,38.48 16. 582-582.9,583-583.7 17. 710.0, 710.1, 710.4,714.0
Elixhauser’s 30 Comorbidities
Congestive heart failure Cardiac arrhythmias Valvular disease Pulmonary circulation disorders Peripheral vascular disorders Hypertension, uncomplicated Hypertension, complicated Paralysis Other neurological disorders Chronic pulmonary disease Diabetes, uncomplicated Diabetes, complicated Hypothyroidism Renal failure Liver disease
Peptic ulcer disease excluding bleeding AIDS Lymphoma Metastatic cancer Solid tumor without metastasis Rheumatoid arthritis/collagen vascular diseases Coagulopathy Obesity Weight loss Fluid and electrolyte disorders Blood loss anemia Deficiency anemias Alcohol abuse Drug abuse Psychoses Depression
CABG in Canada
Multivariate predictors of in-hospital death
age female sex (OR=1.4) urgent admission (1.5) cerebrovasc dis (1.5) PVD (1.8) vent aneurysm (1.8) prior CABG (1.9) failed PTCA (2.0) CHF (2.2) recent MI (2.3) hemiplegia (2.5) CABG + valve (2.6) kidney disease (3.2) neoplasm (3.7) severe liver dis (11.7)
Validity of ICD10 administrative data in recording comorbidity information
(study funded by the Canadian Institutes of Health Research)
Hude Quan, William Ghali, Duncan Saunders
1. Assess validity of ICD-10-CA administrative data 2. Develop ICD-10-CA coding algorithms for comorbidities - Charlson comorbidities - Elixhauser comorbidities
Many methodological issues…
Validity of ICD-9 and ICD-10 coded data Validity of comorbidity coding algorithms Validity of procedure codes in administrative data Validity of physician claims Comorbidity algorithms based on pharmacy claims data Longitudinal linkage of records Etc….
Today’s symposium
International experts in use of administrative data (all contributors of methodological developments) Representation from:
Canada (academic, CIHI, Statistics Canada) U.S. Australia Switzerland China U.K
Funding from CIHR (through the Institute for Health Services and Policy Research) Internationally collaborative research being planned
Warning…
You will hear a lot… You will see a lot… Some material will be quite technical… So…
– We expect some follow-up e-mails and phone calls: – wghali@ucalgary.ca (403 210 9317) – hquan@ucalgary.ca (403 944 8912)
Bring it on…!