Upload
di-wu
View
27
Download
1
Embed Size (px)
Citation preview
Prediction of Medical Malpractice Payment Claims
Gopher 6 Consulting Group
Abstract The goal was to create a model predicting medical
malpractice payments. All of our models were created using the Cognalysis
MultiRate software provided by Gross Consulting. In our process of creating the model, using data from the
National Practitioner Data Bank, we modified our variables, adjusted data, and created new fields.
Cognalysis MultiRateGross Consulting’s inhouse predictive
modeling softwareEasy to useComplex Math Concepts
Simple and visible
How to Run AnalysisImport DataSelect FieldsChoose Credibility and IterationFilter
Train and testRun!
Result Inspection PanelRaw Factor
factor without accounting for other variablesAdjusted Factor
factor after accounting for other variablesModel Factor
Takes adjusted factor and credibility
Grouped vs. Generic CharacteristicGrouped
groups numeric data into bins
i.e, Age, Years
Generic CharacteristicEach distinct value will
be treated independently
i.e., Male and Female, Field of License
New Fields“paymentperperson”
The payment divided by the number of people who were paid
“diffyear”The number of years from when the malpractice took
place and when the claim was filed“yearexperience”
The number of years between the practitioner’s graduation and the time the malpractice took place.
“diffyear”
Why?“malyear” and “origyear” both had low significance
How?Functions: IF, ISBLANK, AVERAGE, MROUND
SummaryBest submission
Used the fields with larger R^2Used Fields with higher significance and high effective
ratioRemoved fields with large average absolute error
(result inspection panel)Use exposure 10, 100 iterations
Future Investigation interaction effectrandom variablesfields we created but did not useunsubmitted model