System and method for medical treatment hypothesis testing

• US 8,719,210 B2
• Filed: 08/13/2013
• Issued: 05/06/2014
• Est. Priority Date: 06/12/2009
• Status: Active Grant

• Current Assignee: Consolidated Research Incorporated
• Original Assignee: Consolidated Research Incorporated
• Inventors: Ridgeway, Gregory K.
• Primary Examiner(s): CHANG, LI WU
• Application Number:
  US13/966,141
• Publication Number:
  US 20140025398A1
• Time in Patent Office: 266 Days
• Field of Search: None
• US Class Current: 706/50
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

A computer-implemented system and method of evaluating the effects of medical treatments, the method including receiving patient record data; identifying relevant characteristics for evaluation; identifying a first treatment; identifying a second treatment; assigning a weight to each patient case; determining the relative likelihood, using the assigned weights, that an identified treatment will result in an identified effect when contrasted with a second identified treatment; and, outputting this estimated relative likelihood.

Claims

1. An apparatus for evaluating effects of treatments comprising:

a processor; and

a memory operably coupled to the processor and having program instructions stored therein, the processor being configured to execute the program instructions, the program instructions comprising;

receiving record data of a plurality of subjects;

identifying relevant characteristics for evaluation;

identifying a first treatment;

identifying a second treatment;

identifying an effect;

identifying, from among the subjects, a first group of subjects that received the first treatment;

identifying, from among the subjects, a second group of subjects who did not receive the first treatment;

assigning a weight to the record data of each subject of the second group of subjects corresponding to a likelihood that the corresponding subject would have received the first treatment to generate weighted record data of the second group of subjects;

determining whether the weighted record data of the second group of subjects is balanced with the record data of the first group of subjects;

adjusting the weight assigned to the record data of each subject of the second group of subjects to balance the weighted record data of the second group of subjects with the record data of the first group of subjects;

determining an estimated relative likelihood, using the weighted record data of the second group of subjects to perform a regression, that the first treatment, as compared to the second treatment, will result in the effect; and

outputting the estimated relative likelihood.

2. The apparatus of claim 1, wherein determining the estimated relative likelihood that the first treatment, as compared to the second treatment, will result in the effect further comprises using the record data of the first group of subjects, which is unweighted, in the regression.

3. The apparatus of claim 1, wherein the likelihood that the corresponding subject would have received the first treatment is based on the relevant characteristics of the record data of the corresponding subject.

4. The apparatus of claim 1, wherein assigning the weight to the record data of each subject of the second group of subjects comprises setting a propensity score and a residual value for the record data of each of the subjects.

5. The apparatus of claim 4,

wherein the record data comprises columns of indicator functions based on the relevant characteristics, each of the indicator functions being associated with a corresponding one of the subjects,

wherein the residual value for the record data of each of the subjects is arranged in a residual value column, each place in the column corresponding to one of the subjects, and

wherein assigning the weight to the record data of each subject of the second group of subjects further comprises:

identifying which column of the indicator functions is most correlated to the residual value column;

adjusting the propensity score for the record data of each of the subjects based on the identified column; and

determining the weight to assign to the record data of each subject of the second group of subjects based on the propensity score.

6. The apparatus of claim 5, wherein when identifying which column of the indicator functions is most correlated to the residual value column, the following equation is used:

r_j=Sum_i((t_i−

p_i)(I_ji−

mean(I_j)))/((n−

1)sd(t−

p)sd(I_j)).

7. The apparatus of claim 5, wherein when adjusting the propensity score of the record data of each of the subjects based on the identified column, the following equation is used:

p_i/(1−

p_i)=p_i/(1−

p_i)exp(δ

sign(r_j)(2I_ji−

1)).

8. The apparatus of claim 5, wherein when determining the weight to assign to the record data of each subject of the second group of subjects based on the propensity score, the following formula is used:

w_i=p_i/(1−

p_i).

9. The apparatus of claim 5, wherein the adjusting the weight assigned to the record data of each subject of the second group of subjects to balance the weighted record data of the second group of subjects with the record data of the first group of subjects comprises:

adjusting the residual value of the record data of each of the subjects based on the propensity score;

identifying which column of the indicator functions is most correlated to the residual value column;

adjusting the propensity score for the record data of each of the subjects based on the identified column; and

determining the weight to assign to the record data of each subject of the second group of subjects based on the propensity score.

10. The apparatus of claim 4, further comprising setting a treatment identifier in the record data of each of the subjects,

wherein the treatment identifier of the record data of each of the first group of subjects is set to 1,

wherein the treatment identifier of the record of each of the second group of subjects is set to 0,

wherein the propensity score of the record data of each of the subjects is initially set to be equal to the number of subjects in the first group of subjects divided by the total number of subjects in the record, and

wherein the residual value of the record data of each of the subjects is initially set by subtracting the corresponding propensity score from the corresponding treatment identifier.

11. The apparatus of claim 1, wherein the determining whether the weighted record data of the second group of subjects is balanced with the record data of the first group of subjects comprises:

aggregating the record data of the first group of subjects to generate aggregate first data;

aggregating the weighted record data of the second group of subjects to generate aggregate second data; and

comparing the aggregate first data to the aggregate second data.

12. The apparatus of claim 11, wherein the determining whether the weighted record data of the second group of subjects is balanced with the record data of the first group of subjects further comprises:

determining that the weighted record data of the second group of subjects is balanced with the record data of the first group of subjects when the aggregate first data is substantially similar to the aggregate second data.

13. The apparatus of claim 12, wherein the aggregate first data is substantially similar to the aggregate second data when a maximum absolute difference between corresponding ones of the relevant characteristics of the first aggregate data and the second aggregate data is less than or equal to 0.2 percent.

14. The apparatus of claim 1, wherein the estimated relative likelihood that the first treatment, as compared to the second treatment, will result in the effect is determined by estimating a ratio of rate ratios.

15. The apparatus of claim 14, further comprising adjusting the estimated relative likelihood that the first treatment, as compared to the second treatment, will result in the effect is determined by estimating a weighted Poisson regression model.

17. The system of claim 1, wherein determining the estimated relative likelihood that the first treatment, as compared to the second treatment, will result in the effect further comprises using the record data of the first group of subjects, which is unweighted, in the regression.

18. The system of claim 17, wherein the estimated relative likelihood that the first treatment, as compared to the second treatment, will result in the effect is determined by estimating a ratio of rate ratios.

16. A system for evaluating effects of treatments comprising:

at least one record database coupled to a network, the at least one record database configured to store record data of a plurality of subjects; and

a server coupled to the network, wherein the server is configured to;

access, through the network, the record data of the plurality of subjects stored in the at least one record database;

identify relevant characteristics for evaluation;

identify a first treatment;

identify a second treatment;

identify an effect;

identify, from among the subjects, a first group of subjects that received the first treatment;

identify, from among the subjects, a second group of subjects who did not receive the first treatment;

assign a weight to the record data of each subject of the second group of subjects corresponding to a likelihood that the corresponding subject would have received the first treatment to generate weighted record data of the second group of subjects;

determine whether the weighted record data of the second group of subjects is balanced with the record data of the first group of subjects;

adjust the weight assigned to the record data of each subject of the second group of subjects to balance the weighted record data of the second group of subjects with the record data of the first group of subjects;

determine an estimated relative likelihood, using the weighted record data of the second group of subjects to perform a regression, that the first treatment, as compared to the second treatment, will result in the effect; and

output the estimated relative likelihood.

19. An apparatus for evaluating effects of treatments comprising:

a processor; and

a memory operably coupled to the processor and having program instructions stored therein, the processor being configured to execute the program instructions, the program instructions comprising;

receiving record data of a plurality of subjects;

receiving relevant characteristics for evaluation;

receiving an identification of a first treatment, a second treatment, and an effect;

identifying, from among the subjects, a first group of subjects that experienced the first treatment;

identifying, from among the subjects, a second group of subjects who did not experience the first treatment;

assigning a weight to the record data of each subject of the second group of subjects according to a propensity score and a residual value of the record data of each of the subjects to generate weighted record data of the second group of subjects;

determining whether the weighted record data of the second group of subjects is balanced with the record data of the first group of subjects;

adjusting the weight assigned to the record data of each subject of the second group of subjects to balance the weighted record data of the second group of subjects with the record data of the first group of subjects;

determining an estimated relative likelihood, using the weighted record data of the second group of subjects and unweighted data of the first group of subjects to perform a regression, that the first treatment, as compared to the second treatment, will result in the effect; and

outputting the estimated relative likelihood.