We all are aware that data mining holds great potential for healthcare providers to use data mining and data analysis in such a way that the physicians identify effective treatments and best practices, and patients receive better and more affordable healthcare services.
Here for the example, I have taken the Apollo Health organization, India. And would like to discuss about implementing the data Mining opportunity: Get 2nd opinion.
Suppose take the scenario, where I have consulted a doctor in a local hospital. Based on the criticality/other factors, I would like to take 2nd opinion from a famous hospital. Here we will have the problems like cost for the 2nd opinion, time taken for the result of the 2nd opinion would be very less like 30 minutes like they will be seeing the results of our previous diagnosis, and for this I may have to travel for more than 5 – 6 hours which is also added cost.
Describing the data that is useful for this opportunity: Based on the health problems which the organization have diagnosed in the past they know on what kind of information they have to collected from the new users who are seeking 2nd opinion. Because of various problem’s there would be different data that needs to be collected. Please find the sample data for reference.
|
User ID |
Gender/ Age |
Suffering from |
Symptoms |
Frequency |
Previous consultation date |
Drugs practised before |
Resultof previous drugs |
Drugs prescribed by Apollo |
Result |
Diagnosis prescribed before |
Diagnosis preferred by apollo |
|
123 |
F/25 |
Fever, Cough |
X, Y, Z |
Daily |
05/12/2016 |
Dola, etc. |
Not solved |
Q, W, E |
Solved |
||
|
56 |
M/23 |
Shoulder pain |
A, B, C |
Once/twice |
05/04/2015 |
G, H, J |
Solved, but occurring again |
D, E, F |
Solved |
Data mining model built: Now the organization can build a model using their past experience on how they have solved/categorized the problem. For example, take the model built below for the problem diabetes using the data mining technique decision tree to find out how certain variables are associated with the onset of diabetes. Further based on the information submitted by the user seeking 2nd opinion, the model predicts the output which is nothing but the prescription suggested by the 1st doctor is valid or not.
Evaluation:
- We can split the training data into 2 parts and calculate the Sensitivity/Specificity/True positive/True/negative rates.
- Also we can use the data given by the new user for training the curent model/testing the model. If the model gives any different output, then we can contact the expert doctors for further evaluation.
Monetary value:
- By allowing other users to upload their prescriptions/reports, the organization would be getting the knowledge on what practises followed by the other doctors and update theirs.
- More importantly they would be new user’s information who have not visited the hospital. Now the organizations can run the model built and predict the outcome whether the proposed suggestion by the 1st doctor is valid or not. Further they can impress the new users by following up continuously, not let them slip away in future or making them refer our organization to their friends.
Integration of data mining model built:
This opportunity can be integrated into either the patient care -> Value Added Service or they can integrate into any their data collection mechanisms or marketing strategy for attracting new users.
