FARROKH ALEMI, PH.D. Measurement of Severity from Claims Data
In Case of Poor Outcomes First blame patient’s illness Then examine patient’s preferences Last attribute it to poor quality
Not Accurate Enough 319,740 articles in PubMed Maybe insufficient Large data make it worst
Example of Lack of Accuracy Severity of MI explains 76% of outcomes of care
Severity of Single Diagnosis Case 1st Diagnosis 2nd Diagnosis 3rd Diagnosis 4th Diagnosis 5th Diagnosis Length of stay 1MICHF5 2MI2 3 CHFDiabetes6 4MICHF7 5 Diabetes4 6MIHypertension2 7ACL surgery1 8MIDiabetesCHFHypertensionAlcohol use6 9MIAlcohol use3 10CHF3
Severity of Single Diagnosis CaseMICHFDiabetes Alcohol use Hypertensio n Length of stay
Severity of Single Diagnosis Coefficients Intercept0 MI2.38 CHF3.40 Diabetes0.33 Alcohol use0.49 Hypertension-0.50
Severity of Single Diagnosis Coefficients Standardized score Intercept0 MI CHF Diabetes Alcohol use Hypertension
Severity of Multiple Diagnoses
Overall severity = 1 - ( ) ( ) = 0.42
Accuracy Alemi and Walters TypicalNew 53% More than doubled accuracy 9.8%- 16.9%
TAKE HOME LESSON A simple Procedure for More Accurate Measurement of Severity