Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
7 Cards in this Set
- Front
- Back
Goals of this study
|
To see how the following factors affect claim frequency and claim size:
Age/Sex Plan type Diagnosis Type of claim Enrollee type (Employee, spouse, dependent) To see which factors cause the largest claims. To see the relationship between diagnosis and procedure. |
|
Requested data format
|
Separate tables for (a) claims; (b) exposure data; (c) block of business (plan type)
Record “keys” allowing researcher to match exposure records with claims records Patient confidentiality: no names or SSNs. Each claim record was supposed to have: Date of birth Sex Claim date Diagnosis code Procedure code Dollar amount |
|
Data problems encountered
|
All different media and softwares (Excel, Access, Paradox…)
Incorrect format Incomplete data Missing data (e.g. sex was blank) Unreasonable data (e.g. more claimants than members; ages > 108) Useless record keys (couldn’t match claims with exposures; differentiate employee from spouse) Labor-intensive and time-consuming to edit, reformat, and clean all the data. |
|
Things you can do when bad data is encountered
|
Remove it
Fix it (danger! Must be sure you’re right) Reject the entire submission Ask insurer for clarification |
|
Ways in which the submitted data had to be edited
|
Check data for reasonability
Rearrange files into a standard format (dates; diagnosis codes; etc.) Fill in missing data Remove bad data Remove extraneous data (e.g. claims outside of the study period) Judgment was required. |
|
Other steps taken prior to analysis
|
Anonymize the insurers:
Shuffle the data records Only publish data combining at least 3 insurers Manually insert “derived fields”: Diagnosis Category, for each claim Primary diagnosis code and procedure, for each patient |
|
Recommendations for future data collection
|
Find insurers committed to ongoing submissions.
Improves efficiency and standardization of the study Develop standard methods for manipulating each insurer’s data into the correct format. Or, develop data specs easier for the insurers to comply with. Better communication and feedback Give insurers incentives to participate (e.g a free analysis of their data) Done |