Clinical decision-making artificial intelligence object oriented system and method
View Patent ↗The present invention involves a system and method of providing decision support for assisting medical treatment decision-making. A patient agent software module processes information about a particular patient. A doctor agent software module processes information about a health status of a particular patient, beliefs relating to patient treatments, and the actual effects of treatment decisions. By filtering information over time from the patient agent into the doctor agent, a plurality of decision-outcome nodes are created and formed into a patient-specific outcome tree with the plurality of decision-outcome nodes. An optimal treatment is determined by evaluating the plurality of decision-outcome nodes with a cost per unit change function to output the optimal treatment. When additional information is available from at least one of the patient agent and the doctor agent, the filtering, creating, and determining steps are repeated thus allowing for the system to “reason over time”, continuously updating and learning as new information is received.
1. A method of providing decision support for assisting medical treatment decision-making comprising:
receiving, by a health care device and from a first device operating an agent software module, evidence based information corresponding to a health status of a particular patient and patient treatment decisions;
filtering, by the health care device, the evidence based information to create a plurality of decision-outcome nodes comprising a plurality of belief state nodes reflecting potential future effects of different treatment choices;
determining an optimal treatment for the particular patient by evaluating the plurality of decision-outcome nodes with a scoring function, wherein the scoring function includes a cost per unit change function that calculates the cost of obtaining one unit of outcome change (delta) on a given outcome, wherein the calculation involves the following equation:
π
*
=
arg
min
π
CPUC
(
π
)
=
arg
min
π
E
{
C
T
(
π
)
Δ
CDOI
(
π
)
,
if
Δ
CDOI
(
π
)
≥
1
C
T
(
π
)
1
+
(
1
-
Δ
CDOI
(
π
)
)
*
CPS
,
else
where CPS=cost per service for a given session, CT is a random variable denoting accumulated cost at the end of treatment, E indicates expected value, and ΔCDOI(π) is calculated as:
ΔCDOI(π)=CDOI T (π)−CDOI 0
where CDOI T is a random variable denoting the CDOI-ORS value at the end of treatment;
transmitting, by the health computing device and to the first device, one or more first commands to cause the first device to physically administer the determined optimal treatment; receiving, by the health care computing device, observation data corresponding to the particular patent
updating at least one of the plurality of belief state nodes using the observation data;
determining, based on the scoring function and the updating the at least one of the plurality of belief state nodes, a new optimal treatment; and
transmitting, by the health care device and to the first device, one or more second commands to cause the first device to physically administer the new optimal treatment.
2. The method of claim 1 wherein the evidence based information includes a plurality of health status information at a plurality of times.
3. The method of claim 1 wherein the creating the plurality of decision-outcome nodes comprises:
receiving rewards/utilities; and
selecting patient treatments in order to maximize overall utilities.
4. The method of claim 1 wherein the decision-outcome nodes are updated according to a transition model.
5. The method of claim 1 wherein the plurality of decision-output nodes are configured as a multi-level tree.
6. The method of claim 1 wherein the evidence based information includes at least one of clinical data, electronic health record information, and genetic data.
7. The method of claim 1 wherein upon receiving additional evidence based information, each of the plurality of decision-output nodes is recalculated.
8. The method of claim 1 , wherein the plurality of belief nodes further indicate future health statuses of the particular patient.
9. A decision support system for assisting medical treatment decision-making, said system comprising:
a processor and associated memory, said memory including program memory configured to store instructions for enabling said processor to perform operations, and said memory including storage memory configured to store data upon which said processor performs operations;
said storage memory including data relating to a particular patient;
said program memory including a plurality of instructions that when executed by said processor enables said processor to execute the following steps:
receive, from a first device operating a patient agent software module, information about the particular patient;
receive, from a second device operating a doctor agent software module, information about a health status of the particular patient, doctor beliefs relating to at least one of patient treatments and treatment effects, and evidence based information relating to effects of patient treatments;
filter the information from the patient agent and the information from the doctor agent to create a plurality of decision-outcome nodes comprising a plurality of belief state nodes reflecting potential future effects of different treatment choices;
create a patient-specific outcome based on the plurality of decision-outcome nodes;
determine an optimal treatment by evaluating the plurality of decision-outcome nodes with a scoring function and outputting the optimal treatment, wherein the scoring function includes a cost per unit change function that calculates the cost of obtaining one unit of outcome change (delta) on a given outcome, wherein the calculation involves the following equation:
π
*
=
arg
min
π
CPUC
(
π
)
=
arg
min
π
E
{
C
T
(
π
)
Δ
CDOI
(
π
)
,
if
Δ
CDOI
(
π
)
≥
1
C
T
(
π
)
1
+
(
1
-
Δ
CDOI
(
π
)
)
*
CPS
,
else
where CPS=cost per service for a given session, C T is a random variable denoting accumulated cost at the end of treatment, E indicates expected value, and ΔCDOI(π) is calculated as:
ΔCDOI(π)=CDOI T (π)−CDOI 0
where CDOI T is a random variable denoting the CDOI-ORS value at the end of treatment; transmit, to the second device, one or more first commands to cause the second device to physically administer the determined optimal treatment; receive observation data corresponding to the particular patent;
update at least one of the plurality of belief state nodes using the observation data;
determine, based on the scoring function and the updating the at least one of the plurality of belief state nodes, a new optimal treatment; and
transmit to the second device, one or more second commands to cause the second device to physically administer the new optimal treatment.
10. The decision support system of claim 9 , wherein the patient agent software module includes a plurality of health status information at a plurality of times.
11. The decision support system of claim 9 , wherein the doctor agent software module includes a module that receives rewards/utilities, and a module to select patient treatments in order to maximize overall utilities.
12. The decision support system of claim 9 , wherein the plurality of decision-outcome nodes are updated according to a transition model.
13. The decision support system of claim 9 further including a learning software module with a knowledge base wherein when additional information is available, such information is included in the knowledge base used by at least one of the patient agent software module, the doctor agent software module, and the determining optimal treatment step.
14. The decision support system of claim 9 wherein the doctor agent software has evidence based information from at least one of clinical information, electronic health record information, and genetic information.
15. The decision support system of claim 9 , wherein said plurality of decision-outcome nodes are configured as a multi-level tree.
16. A server for providing decision support for medical treatment decision-making, said system comprising:
a processor and associated memory, said memory including program memory configured to store instructions for enabling said processor to perform operations, and said memory including storage memory configured to store data upon which said processor performs operations;
said storage memory including data relating to a particular patient;
said program memory including a plurality of instructions that when executed by said processor enables said processor to execute the following steps:
receive, from a first device, evidence based information about a health status of the particular patient, doctor beliefs relating to at least one of patient treatments and treatment effects, and patient treatment decisions;
filter the evidence based information to create a plurality of decision-outcome nodes;
determine an optimal treatment by evaluating the plurality of decision-outcome nodes with a scoring function and outputting a message including the optimal treatment, wherein the scoring function includes a cost per unit change function that calculates the cost of obtaining one unit of outcome change (delta) on a given outcome, wherein the calculation involves the following equation:
π
*
=
arg
min
π
CPUC
(
π
)
=
arg
min
π
E
{
C
T
(
π
)
Δ
CDOI
(
π
)
,
if
Δ
CDOI
(
π
)
≥
1
C
T
(
π
)
1
+
(
1
-
Δ
CDOI
(
π
)
)
*
CPS
,
else
where CPS=cost per service for a given session, C T is a random variable denoting accumulated cost at the end of treatment, E indicates expected value, and ΔCDOI(π) is calculated as:
ΔCDOI(π)=CDOI T (π)−CDOI 0
where CDOI T is a random variable denoting the CDOI-ORS value at the end of treatment; transmit, to the first device, one or more first commands to cause the first device to physically administer the determined optimal treatment; receive observation data corresponding to the particular patent;
update at least one of the plurality of belief state nodes using the observation data;
determine, based on the scoring function and the updating the at least one of the plurality of belief state nodes, a new optimal treatment; and
transmit to the first device, one or more second commands to cause the first device to physically administer the new optimal treatment.