Method and system for outcome based referral using healthcare data of patient and physician populations
A recommendation system and method includes extracting patient features for a current patient to generate representation of the current patient. The patient features for the current patient are compared to physician features of one or more physicians and patient-to-physician features of a group of patients from medically related records. Outcome measures associated with physicians are compared related to a current query. A future outcome for patient, physician pairs are predicted for the current patient based upon at least one predictive model constructed from the features and outcome measures to output.
1. A recommendation method, comprising:
extracting patient features for a current patient to generate a representation of the current patient, wherein the patient features include data from at least one of patient profile records and medical records of the current patient;
comparing the patient features for the current patient to physician features of one or more physicians and patient-to-physician features of a group of patients from medically related records;
comparing outcome measures associated with physicians related to a current query; and
predicting a future outcome, using a computer device, for patient, physician pairs for the current patient based upon at least one predictive model constructed from the features and outcome measures to output.
2. The method as recited in claim 1 , wherein extracting patient features includes extracting patient features of a current patient from a patient profile including biographical patient data.
3. The method as recited in claim 1 , wherein extracting patient features includes selecting relevant features, and the patient features are represented by one or more N-dimensional feature vectors.
4. The method as recited in claim 1 , further comprising extracting physician features from physician profiles including medical practice information and patient treatment information, wherein the physician features are represented by one or more N-dimensional feature vectors.
5. The method as recited in claim 1 , further comprising extracting patient-to-physician features including outcomes of similar patients treated by the one or more physicians.
6. The method as recited in claim 1 , wherein comparing outcome measures includes collecting statistics from medical literature for similar patients.
7. The method as recited in claim 1 , further comprising: outputting patient, physician pair scores for a current patient and a plurality of physicians to select a physician based upon the pair scores.
8. The method as recited in claim 1 , further comprising: posing a conditional query to output patient, physician pair scores for a current patient and a plurality of physicians to select a physician based upon the pair scores.
9. The method as recited in claim 1 , further comprising: adding a current patient's information to update the at least one predictive model.
10. A non-transitory computer readable storage medium comprising a computer readable program, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:
extracting patient features for a current patient to generate a representation of the current patient, wherein the patient features include data from at least one of patient profile records and medical records of the current patient;
comparing the patient features for the current patient to physician features of one or more physicians and patient-to-physician features of a group of patients from medically related records;
comparing outcome measures associated with physicians related to a current query; and
predicting a future outcome for patient, physician pairs for the current patient based upon at least one predictive model constructed from the features and outcome measures to output.
11. The computer readable storage medium as recited in claim 10 , wherein extracting patient features includes extracting patient features of a current patient from a patient profile including biographical patient data.
12. The computer readable storage medium as recited in claim 10 , wherein extracting patient features includes selecting relevant features, and the patient features are represented by one or more N-dimensional feature vectors.
13. The computer readable storage medium as recited in claim 10 , further comprising extracting physician features from physician profiles including medical practice information and patient treatment information, wherein the physician features are represented by one or more N-dimensional feature vectors.
14. The computer readable storage medium as recited in claim 10 , further comprising extracting patient-to-physician features including outcomes of similar patients treated by the one or more physicians.
15. The computer readable storage medium as recited in claim 10 , wherein comparing outcome measures includes collecting statistics from medical literature for similar patients.
16. The computer readable storage medium as recited in claim 10 , further comprising: outputting patient, physician pair scores for a current patient and a plurality of physicians to select a physician based upon the pair scores.
17. The computer readable storage medium as recited in claim 10 , further comprising: posing a conditional query to output patient, physician pair scores for a current patient and a plurality of physicians to select a physician based upon the pair scores.
18. The computer readable storage medium as recited in claim 10 , further comprising: adding a current patient's information to update the at least one predictive model.
19. A system, comprising:
a processor,
memory coupled to the processor, the memory storing a patient, physician pair outcome prediction tool, the tool comprising:
an extraction module configured to store patient features, physician features and patient-to-physician features from medically related records for a current patient, a group of patients and one or more physicians, wherein the patient features include data from at least one of patient profile records and medical records of the current patient;
outcome measures stored from health care data related to a current query; and
at least one predictive model constructed from the features and outcome measures to predict a future outcome for a patient, physician pair.
20. The system as recited in claim 19 , wherein the extraction module extracts patient features of a current patient from a patient profile including biographical patient data; extracts physician features from physician profiles including medical practice information and patient treatment information, wherein the physician features are represented by one or more N-dimensional feature vectors; and extracts patient-to-physician features including outcomes of similar patients treated by the one or more physicians.
21. The system as recited in claim 19 , wherein the outcome measures include statistics from medical literature for similar patients.
22. The system as recited in claim 19 , wherein the at least one predictive model outputs patient, physician pair scores for a current patient and a plurality of physicians to select a physician based upon the pair scores.
23. The system as recited in claim 19 , wherein the at least one predictive model outputs patient, physician pair scores, responsive to a conditional query, for a current patient and a plurality of physicians to select a physician based upon the pair scores.
24. The system as recited in claim 19 , wherein the at least one predictive model is updated by adding a current patient's information.
25. The system as recited in claim 19 , further comprising a user interface configured to pose a query for determining a physician.