Data Mining

1. Predicting Risk of Patients

   • Segmenting into High, Medium and Low Risk groups
   • More priority is given to High Risk patients

2. Data Mining to Prevent Hospital Readmissions

   • Hospital readmissions cost $17 billion each year
   • Use predictive analytics to reduce readmissions

Heathcare analytics could potentially reduce 30% of total costs

1. Factors that have most impact on classification of patients

   • ICD-9 codes
   • gender
   • age
   • ethnicity

2. Selection of variables for our prediction models.

   • Good Selection

1. Identify a business problem that we want to solve and translate it into a data mining problem.
2. Select appropriate data to analyze.
3. Data pre-processing (explore and clean the data).
4. Data transformation (reduce data or create new variables as required).
5. Determine the data mining task (prediction, classification, clustering, etc.)
6. Choose the data mining techniques to be used.
7. Build the data mining models.
8. Assess the results of the models
9. Select the best model.
10. Perform classification using the selected model.

Admitting Diagnosis Code

Length of Stay v/s Patient Age

Relationships between Risk Mortality and Illness Severity

Nested tables

Diagnosis and Procedure Codes

Diagnosis and Procedure Codes Lookup

Patient_Diagnosis table

Patient_Procedure table

Logistic Regression

The model shows that illness severity, length of stay, patient age, type of admission are most important factors

Decision Trees

Naive Bayes Model

Decision Tree

Thank You!