Estimating healthcare outcomes for individuals
A method and apparatus for predicting a health benefit for an individual is provided. Outcomes from a first simulation on a set of simulated individuals reflecting a population are stored and used to determine a first risk function and corresponding cost values. Outcomes from a second simulation on a set of simulated individuals reflecting having a healthcare intervention are stored and used to determine a second risk function reflecting the intervention and corresponding cost values of the intervention. A benefit function is derived from the difference of the first and second risk functions. A cost function that describes the cost of the intervention is derived from the respective cost values. The derived benefit function and cost function are used to predict the corresponding benefit and cost of the healthcare intervention for a given individual. Individuals can be ranked by degree of expected benefit.
1. A computer-implemented method comprising:
at a computer, receiving input data comprising healthcare costs, processes, and a population;
receiving, using the computer, for a first simulation on a first set of simulated subjects reflecting the population, a first set of outcome data, comprising, in part, for each of the simulated subjects, a plurality of potential healthcare outcomes and associated first risk outcome values representing probabilities that the potential healthcare outcomes will occur;
storing a first set of initial variables and, for each simulated subject, corresponding first outcome data from the first set of outcome data;
receiving, using the computer, for a second simulation reflecting an effect of an additional healthcare intervention on a second set of simulated subjects, a second set of outcome data, comprising, in part, for each of the simulated subjects, a plurality of second potential healthcare outcomes and associated second risk outcome values and representing probabilities that the second potential healthcare outcomes will occur with the intervention;
storing a second set of initial variables and, for each simulated subject, corresponding second outcome data from the second set of outcome data;
determining, using the computer, from the stored first set of initial variables and from the corresponding first outcome data for each simulated subject, a first risk function that describes a first risk under the first set of processes;
determining, using the computer, from the stored second set of initial variables and from the corresponding second outcome data for each simulated subject, a second risk function that describes a second risk under the second set of processes containing data reflecting at least one intervention; and
deriving, using the computer, a benefit function that represents a benefit of the intervention by subtracting the second risk function from the first risk function.
2. A computer-implemented method as recited in claim 1 ,
wherein receiving a first set of outcome data, comprising, in part, first risk outcome values, further comprises receiving a first set of outcome data, comprising, in part, first risk outcome values and first net cost outcomes values;
wherein receiving a second set of outcome data, comprising, in part, second risk outcome values associated with the intervention, further comprises receiving a second set of outcome data, comprising, in part, second risk outcome values and second net cost outcome values associated with the intervention; and
determining, for each subject, a difference value of the corresponding first net cost outcome values and the corresponding second net cost outcome values, and deriving, from the difference value for each subject, a net cost function that describes the net cost of the intervention.
3. A computer-implemented method as recited in claim 2 , further comprising:
applying, for an input subject having associated initial values, the benefit function using the associated initial variables and generating a predicted benefit value of the intervention for the input subject;
optionally applying, for the input subject, the net cost function using the associated initial variables and generating a predicted net cost value of the intervention for the input subject; and
storing the predicted benefit value for the intervention and, optionally, the predicted net cost value for the intervention for post-processing.
4. A computer-implemented method as recited in claim 2 , wherein determining the first risk function and determining the second risk function or determining the first net cost function and determining the second net cost function comprise applying one or more regression methods selected from: generalized linear and nonlinear regressions, logistic and Poisson regressions, supervised machine learning algorithms, neural networks, support vector machines, response surface modeling, and multivariate adaptive regression splines.
5. A computer-implemented method as recited in claim 3 , wherein post-processing comprises determining, based in part on the predicted benefit value whether the input subject receives the intervention.
6. A computer-implemented method as recited in claim 1 , further comprising:
applying the benefit function and one or more intervention rules to a plurality of input subjects; and
outputting a list of input subjects from the plurality of input subjects to receive the intervention.
7. A computer-implemented method as recited in claim 1 , further comprising:
applying the benefit function and one or more ranking rules to a plurality of input subjects; and
outputting a ranked list of input subjects from the plurality of input subjects to receive the intervention.
8. A computer-implemented method as recited in claim 7 , further comprising:
applying the benefit function and the net cost function to a plurality of input subjects selected according to the ranked list;
obtaining, for the plurality of input subjects in the ranked list, a cumulative benefit value and a cumulative net cost value;
displaying, on a graph having one axis represent net cost values and the other axis represent benefit values, the cumulative net cost values and the cumulative benefit values for each of the input subjects; and
using, at a point on the graph, a slope at the point to set a net cost-effectiveness threshold for determining which input subjects receive the intervention.
9. A computer-implemented method as recited in claim 1 , wherein the intervention is configured to represent a combination of two or more different interventions.
10. A computer-implemented method as recited in claim 1 , wherein determining the first and the second risk functions, comprises:
fitting a function to a density of the population over a space of biomarker data to determine a weighted risk function; and
integrating the weighted risk function over a range of unknown variables with a subspace defined by known variables.
11. A computer-implemented method as recited in claim 10 , wherein the weighted risk function represents the likelihood of an outcome given specific biomarker data.
12. A computer-implemented method as recited in claim 3 , wherein the predicted net cost value is adjusted by a likelihood that the input subject will change behavior towards compliance.
13. A computer-implemented method as recited in claim 1 , further comprising:
creating copies of starting variables for an input subject;
generating random data to represent any missing variables of the starting variables;
applying the benefit function to the starting variables and to any generated random data to compute one or more benefit values;
storing the computed one or more benefit values;
repeating the generating random data step through the storing the computed one or more benefit values step for a predetermined number of times; and
computing an average benefit value using the stored computed one or more benefit values.
14. A computer-implemented method as recited in claim 2 , further comprising:
creating copies of starting variables for an input subject;
generating random data to represent any missing variables of the starting variables;
applying the net cost function to the starting variables and to any generated random data to compute one or more net cost values;
storing the computed one or more net cost values;
repeating the generating random data step through the storing the computed one or more net cost values step for a predetermined number of times; and
computing an average net cost value using the stored computed one or more net cost values.
15. A computer-implemented method as recited in claim 1 , wherein results relevant to an individual member are presented directly to that member in tabular or graphical form.
16. A non-transitory computer-readable medium storing one or more sequences of instructions for estimating healthcare costs and benefits for individuals, which instructions, when executed by one or more processors, cause the one or more processors to carry out:
at a computer, receiving input data comprising healthcare costs, processes, and a population;
receiving, using the computer, for a first simulation on a first set of simulated subjects reflecting the population, a first set of outcome data, comprising, in part, for each of the simulated subjects, a plurality of potential healthcare outcomes and associated first risk outcome values representing probabilities that the potential healthcare outcomes will occur;
storing a first set of initial variables and, for each simulated subject, corresponding first outcome data from the first set of outcome data;
receiving, using the computer, for a second simulation reflecting an effect of an additional healthcare intervention on a second set of simulated subjects, a second set of outcome data, comprising, in part, for each of the simulated subjects, a plurality of second potential healthcare outcomes and associated second risk outcome values and representing probabilities that the second potential healthcare outcomes will occur with the intervention;
storing a second set of initial variables and, for each simulated subject, corresponding second outcome data from the second set of outcome data;
determining, using the computer, from the stored first set of initial variables and from the corresponding first outcome data for each simulated subject, a first risk function that describes a first risk under the first set of processes;
determining, using the computer, from the stored second set of initial variables and from the corresponding second outcome data for each simulated subject, a second risk function that describes a second risk under the second set of processes containing data reflecting at least one intervention; and
deriving, using the computer, a benefit function that represents a benefit of the intervention by subtracting the second risk function from the first risk function.
17. A non-transitory computer-readable medium as recited in claim 16 , further causing the one or more processors to carry out:
wherein receiving a first set of outcome data, comprising, in part, first risk outcome values, further comprises receiving a first set of outcome data, comprising, in part, first risk outcome values and first net cost outcomes values;
wherein receiving a second set of outcome data, comprising, in part, second risk outcome values associated with the intervention, further comprises receiving a second set of outcome data, comprising, in part, second risk outcome values and second net cost outcome values associated with the intervention; and
determining, for each subject, a difference value of the corresponding first net cost outcome values and the corresponding second net cost outcome values, and deriving, from the difference value for each subject, a net cost function that describes the net cost of the intervention.
18. A non-transitory computer-readable medium as recited in claim 17 , further comprising:
applying, for an input subject having associated initial values, the benefit function using the associated initial variables and generating a predicted benefit value of the intervention for the input subject;
applying, for the input subject, the net cost function using the associated initial variables and generating a predicted net cost value of the intervention for the input subject; and
storing the predicted benefit value for the intervention and the predicted net cost value for the intervention for post-processing.
19. A non-transitory computer-readable medium as recited in claim 17 , wherein determining the first risk function and determining the second risk function or determining the first net cost function and determining the second net cost function comprise applying one or more regression methods selected from: generalized linear and nonlinear regressions, logistic and Poisson regressions, supervised machine learning algorithms, neural networks, support vector machines, response surface modeling, and multivariate adaptive regression splines.
20. A non-transitory computer-readable medium as recited in claim 18 , wherein post-processing comprises determining, based in part on the predicted benefit value whether the input subject receives the intervention.
21. A non-transitory computer-readable medium as recited in claim 17 , further causing the one or more processors to carry out:
applying the benefit function and one or more intervention rules to a plurality of input subjects; and
outputting a list of input subjects from the plurality of input subjects to receive the intervention.
22. A non-transitory computer-readable medium as recited in claim 16 , further causing the one or more processors to carry out:
applying the benefit function and one or more ranking rules to a plurality of input subjects; and
outputting a ranked list of input subjects from the plurality of input subjects to receive the intervention.
23. A non-transitory computer-readable medium as recited in claim 22 , further causing the one or more processors to carry out:
applying the benefit function and the net cost function to a plurality of input subjects selected according to the ranked list;
obtaining, for each of the plurality of input subjects, a benefit value and a net cost value;
displaying, on a graph having one axis represent net cost values and the other axis represent benefit values, the cumulative net cost values and the cumulative benefit values for each of the input subjects; and
using, at a point on the graph, a slope at the point to set a net cost-effectiveness threshold for determining which input subjects receive the intervention.
24. A non-transitory computer-readable medium as recited in claim 16 , wherein the intervention is configured to represent a combination of two or more different interventions.
25. A non-transitory computer-readable medium as recited in claim 16 , wherein determining the first and the second risk functions, comprises:
fitting a function to a density of the population over a space of biomarker data to determine a weighted risk function; and
integrating the weighted risk function over a range of unknown variables with a subspace defined by known variables.
26. A non-transitory computer-readable medium as recited in claim 25 , wherein the weighted risk function represents the likelihood of an outcome given specific biomarker data.
27. A non-transitory computer-readable medium as recited in claim 18 , wherein the predicted net cost value is adjusted by a likelihood that the input subject will change behavior towards compliance.
28. A non-transitory computer-readable medium as recited in claim 16 , further causing the one or more processors to carry out:
creating copies of starting variables for an input subject;
generating random data to represent any missing variables of the starting variables;
applying the benefit function to the starting variables and to any generated random data to compute one or more benefit values;
storing the computed one or more benefit values;
repeating the generating random data step through the storing the computed one or more benefit values step for a predetermined number of times; and
computing an average benefit value using the stored computed one or more benefit values.
29. A non-transitory computer-readable medium as recited in claim 17 , further causing the one or more processors to carry out:
creating copies of starting variables for an input subject;
generating random data to represent any missing variables of the starting variables;
applying the net cost function to the starting variables and to any generated random data to compute one or more net cost values;
storing the computed one or more net cost values;
repeating the generating random data step through the storing the computed one or more net cost values step for a predetermined number of times; and
computing an average net cost value using the stored computed one or more net cost values.
30. A non-transitory computer-implemented method as recited in claim 16 , wherein results relevant to an individual member are presented directly to that member in tabular or graphical form.