WO2024033201A1 - Systèmes et procédés de prédiction par apprentissage automatique dans le contexte de données de traitements historiques - Google Patents

Systèmes et procédés de prédiction par apprentissage automatique dans le contexte de données de traitements historiques Download PDF

Info

Publication number
WO2024033201A1
WO2024033201A1 PCT/EP2023/071495 EP2023071495W WO2024033201A1 WO 2024033201 A1 WO2024033201 A1 WO 2024033201A1 EP 2023071495 W EP2023071495 W EP 2023071495W WO 2024033201 A1 WO2024033201 A1 WO 2024033201A1
Authority
WO
WIPO (PCT)
Prior art keywords
historical
patient
prediction
subject
historical patient
Prior art date
Application number
PCT/EP2023/071495
Other languages
English (en)
Inventor
Declan Patrick Kelly
Original Assignee
Koninklijke Philips N.V.
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Koninklijke Philips N.V. filed Critical Koninklijke Philips N.V.
Publication of WO2024033201A1 publication Critical patent/WO2024033201A1/fr

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the present disclosure is directed generally to machine- learning predictions, and more specifically, to systems and methods of providing healthcare-related machine-learning predictions with enhanced reliability and contextualization.
  • Machine-learning models have been used to identify high risk patients and/or to support decisions of healthcare providers based on historical data linking patient conditions with select outcomes. For example, a patient’s status in the intensive care unit can be correlated with an outcome of death within the next 48 hours.
  • conventional methods are often inaccurate or unreliable because the final outcome depends not only on the patient’s status but also on additional factors, such as the actions taken by the healthcare providers between the status evaluation and the final outcome.
  • the present disclosure is directed generally to systems and methods of providing healthcare-related machine-learning predictions. More specifically, the present disclosure is directed to systems and methods of providing healthcare-related machine-learning predictions with enhanced reliability and contextualization.
  • a system for providing a machine-learning prediction for a subject patient may comprise: a database storing historical medical records for a plurality of patients; and one or more processors in communication with the database.
  • the one or more processors may be configured to perform the following: select a subject patient profile for a subject patient, wherein the subject patient profile comprises medical information associated with the subject patient; using a machinelearning algorithm, generate a first prediction of a first outcome parameter for the subject patient; identify one or more historical patient profiles from the historical medical records stored in the database by comparing the medical information of the subject patient profile with the historical medical records, wherein each historical patient profile of the one or more historical patient profiles correspond to a historical patient and comprise (i) medical information for the corresponding historical patient, and (ii) one or more actual outcomes for the corresponding historical patient; for each of the one or more historical patient profiles identified, generate a second prediction of the first outcome parameter for the corresponding historical patient using the machine-learning algorithm; generate an output comprising the first prediction generated for the subject patient, the one or more actual outcomes of each historical patient profile, and the second prediction generated for each of the historical patient profiles; and report, via a user interface, the output to a healthcare professional.
  • the one or more actual outcomes of each historical patient profile may include at least an actual outcome corresponding to the first outcome parameter.
  • the one or more processors may further be configured to perform the following: for each of the one or more historical patient profiles identified, extract one or more interventions from the historical medical records, wherein the one or more extracted interventions were administered to the corresponding historical patient between a first time period and a second time period.
  • the output may further comprise the one or more interventions extracted for each of the one or more historical patient profiles.
  • the second time period may correspond to a time of the one or more actual outcomes of the historical patient profile
  • the first time period may correspond to a time prior to the second time period
  • the one or more interventions may be extracted from the historical medical records using a natural language processor.
  • the first prediction for the first outcome parameter may be generated for the subject patient based on the medical information of the subject patient profile.
  • the second prediction of the first outcome parameter may be generated for each of the historical patient profiles based on at least the medical information of the corresponding historical patient profile.
  • the one or more processors may be further configured to perform the following: determine at least a first reliability level associated with the first prediction of the first outcome parameter generated for the subject patient by comparing the first prediction of the first outcome parameter with at least one of the following: the second prediction of the first outcome parameter generated for one or more of the historical patient profiles; and the one or more actual outcomes associated with one or more of the historical patient profiles; and based on at least the first reliability level, generate a recommendation for the subject patient comprising at least one recommended treatment, the at least one recommended treatment including an intervention selected from the one or more interventions extracted from the historical medical records.
  • the output further comprises at least the first reliability level associated with the first prediction of the first outcome parameter and the generated recommendation.
  • system may further comprise a memory storing instructions to be executed by the one or more processors and the user interface, wherein the user interface includes at least one of an audio output component and/or a visual output component configured to report the output to the healthcare professional.
  • a method for providing a machine-learning prediction for a subject patient may include: selecting a subject patient profile for a subject patient, wherein the subject patient profile comprises medical information associated with the subject patient; generating, using a machine-learning algorithm, a first prediction of a first outcome parameter for the subject patient; identifying one or more historical patient profiles using the historical medical records stored in the database by comparing the medical information of the subject patient profile with the historical medical records, wherein each historical patient profile of the one or more historical patient profiles correspond to a historical patient and comprise: medical information for the corresponding historical patient; and [0013] one or more actual outcomes for the corresponding historical patient; for each of the one or more historical patient profiles identified, generating a second prediction of the first outcome parameter using the machine-learning algorithm; generating an output comprising the first prediction generated for the subject patient, the one or more actual outcomes of each historical patient profile, the second prediction generated for each of the historical patient profiles; and reporting, via a user interface, the output to
  • the one or more actual outcomes of each historical patient profile may include at least an actual outcome corresponding to the first outcome parameter.
  • method may further include: for each of the one or more historical patient profiles identified, extract one or more interventions from the historical medical records, wherein the one or more extracted interventions were administered to the corresponding historical patient between a first time period and a second time period.
  • the output may further comprise the one or more interventions extracted for each of the one or more historical patient profiles.
  • the second time period may correspond to a time of the one or more actual outcomes of the historical patient profile
  • the first time period may correspond to a time prior to the second time period
  • the one or more interventions may be extracted from the historical medical records using a natural language processor.
  • the first prediction for the first outcome parameter may be generated for the subject patient based on the medical information of the subject patient profile.
  • the second prediction of the first outcome parameter may be generated for each of the historical patient profiles based on at least the medical information of the corresponding historical patient profile.
  • the method may further include: determine at least a first reliability level associated with the first prediction of the first outcome parameter generated for the subject patient by comparing the first prediction of the first outcome parameter with at least one of the following: the second prediction of the first outcome parameter generated for one or more of the historical patient profiles; and the one or more actual outcomes associated with one or more of the historical patient profiles; and based on at least the first reliability level, generate a recommendation for the subject patient comprising at least one recommended treatment, the at least one recommended treatment including an intervention selected from the one or more interventions extracted from the historical medical records.
  • the output further comprises at least the first reliability level associated with the first prediction of the first outcome parameter and the generated recommendation.
  • FIG. 1 is a system diagram illustrating the use of the disclosed prediction systems according to aspects of the present disclosure.
  • FIG. 2 is a diagram illustrating a timeline and patient profile for a subject patient according to aspects of the present disclosure.
  • FIG. 3 is a block diagram illustrating a subject patient profile and outcome parameter predictions according to aspects of the present disclosure.
  • FIG. 4 is a diagram illustrating a timeline and patient profile for a historical patient according to aspects of the present disclosure.
  • FIG. 5 is a block diagram illustrating a historical patient profile and outcome parameter predictions according to aspects of the present disclosure.
  • FIG. 6 is another diagram illustrating a timeline and patient profile for a historical patient according to aspects of the present disclosure.
  • FIG. 7 is a flowchart illustrating the identification of historical patient profiles according to aspects of the present disclosure.
  • FIG. 8 is a sample output that may be generated and reported to a healthcare professional according to aspects of the present disclosure.
  • FIG. 9 is another sample output that may be generated and reported to a healthcare professional according to aspects of the present disclosure.
  • FIG. 10 is still another sample output that may be generated and reported to a healthcare professional according to aspects of the present disclosure.
  • FIG. 11 is a systemized block diagram of a system used to provide a healthcare-related prediction for a subject patient illustrated according to aspects of the present disclosure.
  • FIG. 12 is a flowchart illustrating a method of providing a healthcare- related prediction for a subject patient illustrated according to aspects of the present disclosure.
  • FIG. 13 is another flowchart illustrating a method of providing a healthcare-related prediction for a subject patient illustrated according to aspects of the present disclosure.
  • FIG. 14 is a flowchart illustrating an additional portion of a method of providing a healthcare-related prediction for a subject patient illustrated according to aspects of the present disclosure.
  • the present disclosure is directed to systems and methods of providing healthcare- related machine-learning predictions with enhanced reliability and contextualization.
  • Machine- learning models are currently being used in healthcare environments to support healthcare provider decision making, not only for diagnostic purposes but also to identify patients with the highest risk and to predict the likelihood of deterioration.
  • one issue with conventional methods of machine-learning prediction in the medical field is that the final outcome depends not only on the patient’s status at the time of the prediction but also on the actions taken by the healthcare provider between the status evaluation and the final outcome.
  • a conventional machine-learning model may conclude that asthma patients presenting with pneumonia are low risk because this group of patients have a lower death rate. Although the historical data supported this conclusion, it was ultimately incorrect. In fact, asthma patients with pneumonia are a high-risk group and are traditionally given the most intensive treatment. As a result, this group of high-risk patients had a lower death rate. Although the machine-learning model in this scenario may be corrected manually, more subtle mistakes like this can occur in other situations and are difficult to spot. Additionally, any machine-learning model trained on historical data will obviously not include newly emerging diseases in the training data. As the SARS-CoV-2 pandemic has shown, a new disease may emerge quickly, which may make previously trained machine-learning models obsolete.
  • the system 100 can include a database 104 that stores medical records for a plurality of patients.
  • the stored medical records can be medical records for a plurality of past and/or current patients.
  • the stored medical records can be historical medical records corresponding to a plurality of past patients.
  • the system 100 can further include one or more processors 106 in communication with the database 104 and configured to perform one or more steps of the methods described herein (e.g., method 1200 illustrated in FIGS. 12-14).
  • the one or more processors 106 may be configured to perform at least the following: select a subject patient profile for a subject patient 102; generate a first prediction of a first outcome parameter for the subject patient 102 using a machine-learning algorithm; identify one or more historical patient profiles from the historical medical records stored in the database 104; for each of the one or more historical patient profiles identified, generate a second prediction of the first outcome parameter for the corresponding historical patient using the machine-learning algorithm; generate an output; and report the output to a healthcare professional 108 via a user interface 110.
  • the one or more processors 106 can be configured to select a subject patient profile for a subject patient 102, wherein the subject patient profile comprises medical information associated with the subject patient 102.
  • a chronological view 200 of a subject patient 102 is illustrated showing the selection of a subject patient profile 202 for the subject patient 102.
  • the history of the subject patient 102 may be taken at a first point in time 204, such as upon admission to a hospital, and then the subject patient 102 may be monitored over time. More specifically, the hospital or healthcare provider may measure and record one or more medical parameters (e.g., vital signs, conditions, physiological signals, etc.) for the subject patient 102.
  • medical parameters e.g., vital signs, conditions, physiological signals, etc.
  • the system 100 and/or a healthcare professional 108 may select a subject patient profile 202 that includes a subset of the recorded medical information 206.
  • the duration and medical parameters included within the selected subject patient profile 202 may be performed automatically or may be customized based on a determination by the healthcare professional 108.
  • the system 100 can be used to provide a prediction as to an outcome parameter 208 (e.g., P OUTCOME PARAM(O) to P OUTCOME PARAM(n)) to occur at some point in the future (as indicated by the broken lines 210).
  • an outcome parameter 208 e.g., P OUTCOME PARAM(O) to P OUTCOME PARAM(n)
  • the medical information 206 included within the subject patient profile 206 may depend on the outcome parameter 208 to be predicted.
  • the subject patient profile 202 can include medical information 206 that comprises one or more medical parameters 302 (e.g., MED_ PARAM(O) to MED PARAM(n)).
  • the medical parameters 302 may include the patient’s age, weight, height, BMI, and the like.
  • the medical parameters 302 may include one or more vital signs and/or physiological measurements, such as blood pressure, heart rate, respiratory rate, oxygen saturation, blood composition, and the like.
  • the medical information 206 of the subject patient profile 202 may include one or more interventions 304 (e.g., INTERVENTION(O) to INTERVENTION(n)), such as activities or exercises performed by the patient, treatments administered to the patient, and the like.
  • an existing machine-learning model 306 can be used to make a prediction 308 for one or more outcome parameters 208.
  • the one or more outcome parameters 208 can include, for example and without limitation, a discharge risk score, a risk of readmission score, risk of death score, and the like.
  • the machine-learning model 306 can be used to make a first prediction 308 for each of one or more such outcome parameters 208.
  • one or more historical patient profiles from the medical records stored in the database 104 may be identified using the medical information 206 of the subject patient profile 202.
  • the one or more historical patient profiles may be identified prior to making any predictions 308 for the subject patient 102, may be identified after making one or more predictions 308, or may be identified concurrently with the prediction process.
  • FIGS. 4-6 historical patient profiles are illustrated according to various aspects of the present disclosure.
  • a historical patient profile 400 is illustrated on a chronological timeline for a historical patient 402.
  • the historical patient profile 400 may correspond to a particular individual patient 402 for whom historical medical records are stored in the database 104.
  • Each historical patient profile 400 can include medical information 404 for the corresponding historical patient 402, one or more interventions 406, 408 taken with respect to the historical patient 402, and one or more actual outcomes 410.
  • the medical information 404 of each historical patient profile 400 can comprise one or more medical parameters 502 (e.g., MED PARAM(O) to MED PARAM(n)) corresponding to the historical patient 402.
  • the medical parameters 502 may include the patient’s age, weight, height, BMI, and the like.
  • the medical parameters 502 may include one or more vital signs and/or physiological measurements, such as blood pressure, heart rate, respiratory rate, oxygen saturation, blood composition, and the like.
  • the medical information 404 of a historical patient profile 400 may include one or more interventions 504 (e.g., INTERVENTION(O) to INTERVENTION(n)), such as activities or exercises performed by the patient 402, treatments administered to the patient 402, and the like.
  • interventions 504 e.g., INTERVENTION(O) to INTERVENTION(n)
  • the historical patient profile 400 may further include one or more actual outcomes 410 for a variety of outcome parameters 506 (e.g., OUTCOME PARAM(O) to OUTCOME_ PARAM(n)).
  • the one or more outcome parameters 506 can include, for example and without limitation, a discharge risk score, a risk of readmission score, risk of death score, and the like.
  • one or more of the outcome parameters 506 are the same as the outcome parameters 208 that are of interest with respect to the subject patient 102.
  • the historical patient profile 400 may include actual outcomes 410 for one or more outcome parameters 506 that were (or will be) predicted with respect to the subject patient 102 (i.e., outcome parameters 208).
  • each of the one or more historical patient profiles 400 identified include at least one actual outcome 410 for at least one outcome parameter 506 that was (or will be) predicted with respect to the subject patient 102 (i.e., outcome parameters 208).
  • each historical patient profile 400 corresponds to the condition of a historical patient 402 within a period of time.
  • the period of time may only cover a subset of the total amount of information stored in the historical medical records of the database 104.
  • this period of time is scalable depending on various factors, including but not limited to, the subject patient profile 202.
  • the timespan of the subject patient profile 202 may determine the period of time used to examine and define the window for the historical patient profile 400.
  • multiple historical patient profiles 400 may be identified for a single historical patient 402.
  • each of the historical patient profiles 400A, 400B can include medical information 404 A, 404B (e.g., medical parameters 502) corresponding to a particular historical patient 402, one or more interventions 406A, 406B, 408B (e.g., interventions 504), and one or more actual outcomes 410A, 410B for outcome parameters associated with the historical patient 402 (e.g., outcome parameters 506).
  • medical information 404 A, 404B e.g., medical parameters 502
  • interventions 406A, 406B, 408B e.g., interventions 504
  • actual outcomes 410A, 410B for outcome parameters associated with the historical patient 402 e.g., outcome parameters 506
  • the one or more historical profiles 400 may be identified by the system 100 in a variety of ways. In some embodiments, only the parameters that are (or will be) used by the machine-learning model 306 are used to identify the one or more historical patient profiles 400. In particular embodiments, the one or more historical patient profiles 400 may be identified based on feature similarity (i.e., the state of the patient at a point in time). In other embodiments, the one or more historical patient profiles 400 may be identified based on outcome similarity (i.e., using specific temporal start and/or end points). In still further embodiments, the one or more historical patient profiles 400 may be identified based on exposure similarity (i.e., using the presence and/or absence of certain therapeutic interventions).
  • feature similarity i.e., the state of the patient at a point in time
  • outcome similarity i.e., using specific temporal start and/or end points
  • exposure similarity i.e., using the presence and/or absence of certain therapeutic interventions.
  • the one or more historical patient profiles 400 may be identified based on some combination of the factors described above. However, in some embodiments, the one or more historical patient profiles 400 may be identified based whether the historical patient profile 400 contains at least one actual outcome 410 for at least one outcome parameter 506 that matches the hypothetical outcome 308 that was predicted with respect to the subject patient 102 (i.e., outcome parameters 208).
  • a method 700 for identifying one or more historical patient profiles 400 can include: in a step 710, searching a patient database (e.g., database 104) for patients 402 with similarity (e.g., feature similarity, outcome similarity, exposure similarity, etc.) using parameters considered by the machine- learning model 306; in a step 720, for each patient, calculating a distance metric from the current patient that quantifies the similarities and differences between the historical patient profile 400 and the subject patient profile 202; and in a step 730, classifying the identified patient profiles 400 based on the distance metrics and one or more thresholds to stratify the identified patient profiles 400 into sets 740A, 740B, 740C (e.g., indicating varying levels of relevant historical patients 402 and historical patient profiles 400).
  • a patient database e.g., database 104
  • similarity e.g., feature similarity, outcome similarity, exposure similarity, etc.
  • the step 710 includes applying one or more high-level filters to the patient database 104, such as limiting the search of patient records by department (e.g., ICU, Emergency, etc.), by physician, by primary diagnosis, by time period (e.g., last 3 years, etc.), and the like.
  • department e.g., ICU, Emergency, etc.
  • primary diagnosis e.g., primary diagnosis, by time period (e.g., last 3 years, etc.), and the like.
  • the step 720 includes calculating a distance metric for different points in time for a given historical patient 402. For example, if a historical patient — Patient A — has a first historical patient profile 400 corresponding to a hospital visit five years ago and a second historical patient profile 400 corresponding to a hospital visit one year ago, a distance metric may be calculated for both profiles 400 to determine which visit is most similar and/or instructive. In certain embodiments, the distance metric may be calculated upon the occurrence of certain events, such as when particular measurements or test results become available, when there is a dramatic change in continuous measurement values, or at fixed periods of time (e.g., once per day, or once every 6 hours, etc.).
  • the distance metric may be calculated by treating all parameters equally (e.g., all parameters used in the machine- learning model equally), by using a weighted combination of all parameters (e.g., based on odds ratios for the target outcome), or by using a predefined customized approach.
  • one or more parameters may be required to match exactly (such as disease diagnosis), one or more parameters may be required to fall within a particular range (e.g., age is greater than 65), one or more parameters are ignored or deemed irrelevant (e.g., gender), and one or more parameters are matched based on degree of closeness.
  • the step 730 includes determining one or more classes of patient profiles 400, which may be dynamically generated based on the calculated distance metrics, one or more parameters of interest, and/or one or more thresholds.
  • each class can be generated based on different parameters and/or thresholds.
  • the one or more classes may be predefined by the system and/or by a user.
  • a healthcare provider 108 may provide an input to the system 100 that defines a first class as patients over 75 years old and a second class as patients of any age with highly similar set of parameters (based on the calculated distance metrics). As shown in FIG.
  • three classes of patient profiles may be created based on the distance metrics and one or more thresholds, but it is contemplated that more than three or less than three classes may be generated. Further, as discussed with respect to FIG. 10 below, one or more of these classes may be empty (i.e., not include any identified patient profiles), highlighting the case when there are few similar historical patients 402.
  • the system 100 may then generate a prediction for one or more outcome parameters based on each of the identified historical patient profiles 400.
  • the historical patient profile 400 corresponding to a particular patient 402 can be used to generate one or more predictions 508 for one or more outcome parameters 510 (e.g., P OUTCOME PARAM(O) TO P OUTCOME PARAM(n)) using an existing machine-learning model 306.
  • the machine-learning model 306 used to make predictions 308 is also used to make predictions 508.
  • the machine- learning models 306 are different.
  • the one or more predicted outcome parameters 510 can include, for example and without limitation, a discharge risk score, a risk of readmission score, risk of death score, and the like.
  • the one or more predicted outcome parameters 510 can include at least one outcome parameter 208 predicted with respect to the subject patient 102. In some embodiments, the one or more predicted outcome parameters 510 can include only outcome parameters 208 that were the subject of prediction for the subject patient 102. Thus, a second prediction 508 of one or more outcome parameters 510 may be made using the machine- learning model 306 for each of the identified historical patient profiles 400.
  • the one or more historical patient profiles 400 may include one or more interventions 406, 504 provided to the historical patient 402.
  • these interventions 406, 504 may be extracted from the records of the medical records database 104 using a natural language processor and natural language processing.
  • each of the extracted interventions 504 may have been administered to a corresponding historical patient 402 between a first time period and a second time period. More specifically, the first time period may correspond to a time prior to the second time period, and the second time period corresponds to a time of at least one of the actual outcomes 410 for the corresponding historical patient 402.
  • an actual outcome 410 such as risk of death after discharge or readmission may be recorded for 72 hours after that seven-day period.
  • the one or more extracted interventions 504 for that historical patient profile 400 would have been administered to the corresponding historical patient 402 between the occurrence of the actual outcome (i.e., a second time period) and an earlier time (i.e., a first time period).
  • the “earlier time” may be as early as the start of the historical patient profile 400 or anytime thereafter.
  • the system 100 may then generate an output 112 (shown in FIG. 1) and report the output 112 to a healthcare professional 108 via a user interface 110.
  • the output 112 may include at least the first predictions 308 made for the subject patient profile 202, the second predictions 508 made for each of the historical patient profiles 400, and the actual outcomes 410 of each of the historical patient profiles 400.
  • certain outputs 112 that may be reported via a user interface 110 are illustrated according to various aspects of the present disclosure.
  • the name “John Doe” refers to one or more hypothetical subject patients.
  • FIGS. 8-10 may reference one or more hypothetical subject patients named “John Doe,” and do not necessarily refer to the same individual.
  • the output 112 can include a table 800 comprising medical information for a subject patient (i.e., John Doe, PID: 000000001) as well as medical information for various other patients, predictions for certain parameters, and interventions given to similar patients.
  • the output 112 can include a table 900 comprising medical information for a subject patient (i.e., John Doe, PID: 000000001) as well as medical information for various other patients, predictions, and actual outcomes.
  • the output 112 can include a table 1000 comprising medical information for a subject patient (i.e., John Doe, PID: 000000001) as well as medical information for various other patients, predictions, and actual outcomes.
  • FIG 10 no highly relevant or similar historical patients 402 were identified and therefore this section is blank, whereas at least two relevant historical patients 402 were identified in the output generated and illustrated in FIG. 9.
  • the system 100 can comprise one or more processors 106 (also referred to as central processing units or CPUs), machine-readable memory 1102, an interface bus 1104, all of which may be interconnected and/or communicate through a system bus 1106 containing conductive circuit pathways through which instructions (e.g., machine-readable signals) may travel to effectuate communications, tasks, storage, and the like.
  • the system 100 may be connected to a power source 1108, which can include an internal power source and/or an external power source.
  • the one or more processors 106 can comprise a high-speed data processor adequate to execute program components, which may include various specialized processing units as may be known in the art.
  • the general processor may be a microprocessor, or may also be any traditional processor, controller, microcontroller, or state machine.
  • one or more of the features described herein may be implemented on components such as an Application-Specific Integrated Circuit (“ASIC”), a Digital Signal Processor (“DSP”), a Field Programmable Gate Array (“FPGA”), or similar electronics.
  • ASIC Application-Specific Integrated Circuit
  • DSP Digital Signal Processor
  • FPGA Field Programmable Gate Array
  • the interface bus 1104 may include an input/output interface 1110 configured to connect the system 100 to one or more peripheral devices, a network interface 1112 configured to connect the system 100 to a communications network 1116 (e.g., using a network protocols such as IEEE 802.3 and/or 802.11), and/or a storage interface 1114 configured to accept, communicate, and/or connect to a number of machine-readable memory devices (e.g., storage device 1118, removable storage devices, etc.).
  • a network protocols such as IEEE 802.3 and/or 802.11
  • a storage interface 1114 configured to accept, communicate, and/or connect to a number of machine-readable memory devices (e.g., storage device 1118, removable storage devices, etc.).
  • the network interface 1112 operatively connects the system 100 to a communications network 1116, which can include a direct interconnection, the Internet, a local area network (“LAN”), a metropolitan area network (“MAN”), a wide area network (“WAN”), a wired or Ethernet connection, a wireless connection, and similar types of communications networks, including combinations thereof.
  • a communications network 1116 can include a direct interconnection, the Internet, a local area network (“LAN”), a metropolitan area network (“MAN”), a wide area network (“WAN”), a wired or Ethernet connection, a wireless connection, and similar types of communications networks, including combinations thereof.
  • LAN local area network
  • MAN metropolitan area network
  • WAN wide area network
  • wired or Ethernet connection a wireless connection
  • the memory 1102 can be variously embodied in one or more forms of machine-accessible and machine-readable memory, including a various types of storage devices 1118, random access memory 1124, and read-only memory 1126.
  • the storage device 1118 can include a non-transitory storage medium, a magnetic disk storage, an optical disk storage, an array of storage devices, a solid-state memory device, and the like, including combinations thereof.
  • memory 1102 can include an improved prediction component 1128 that includes a collection of program and/or database components and/or data.
  • the improved prediction component 1128 may include software components, hardware components, and/or some combination of both hardware and software components.
  • the improved prediction component 1128 can include, but is not limited to, instructions 1130 having an identification component 1132, a machine- learning model 1134, an output component 1136. These components may be incorporated into, loaded from, loaded onto, or otherwise operatively available to and from the system 100. Similarly, the system 100 can be incorporated into, loaded from, loaded onto, or otherwise operatively available to and from the system 100.
  • program components may be stored in a local storage device 1118, they may also be loaded and/or stored in other memory, such as a remote cloud storage facility accessible through a communications network (e.g., communications network 1116).
  • the operating system component 1138 can be an executable program component facilitating the operation of the system 100.
  • the operation system component 1138 may facilitate access of the I/O, network, and storage interfaces, and may communicate with other components of the system.
  • the identification component 1132 can be a stored program component that is executed by at least one processor, such as the one or more processors 106 of the system 100.
  • the identification component 1132 can be configured to identify one or more historical patient profiles from the historical medical records, as described herein.
  • the machine-learning model 1134 can be a stored program component that is executed by at least one processor, such as the one or more processors 106 of the system 100.
  • the machine-learning model 1134 can be an artificial intelligence algorithm configured to receive medical information 1140 and generate predictions 1142 related to healthcare-related outcomes, as described herein.
  • the output component 1136 can be a stored program component that is executed by at least one processor, such as the one or more processors 106 of the system 100.
  • the output component 1136 can be configured to receive the various medical information 1140, predictions 1142, interventions 1144, and outcome data 1146 built and/or generated as described herein and report that output to a healthcare professional via a user interface.
  • the improved prediction component 1128 can also include instructions 1130 having a reliability component 1148, which can be a stored program component that is executed by at least one processor, such as the one or more processors 106 of the system 100.
  • the reliability component 1148 can be configured to receive one or more of the predictions 1142, medical information 1140, interventions 1144, and outcomes 1146, to determine one or more reliability levels associated with one or more of the predictions 1142 generated for the subject patient 102 (e.g., first predictions 308), and to generate a recommendation 1150 for the subject patient 102 based on at least the reliability level.
  • an independent reliability level may be determined for each outcome parameter 208 that was predicted as part of the first predictions 308 for the subject patient 102.
  • each of the reliability levels may be determined by comparing the predicted outcome and/or outcomes for the subject patient 102 with at least one of the following: (1) the predicted outcomes 508 for the one or more historical patient profiles 400; and (2) the one or more actual outcomes 410 associated with one or more of the historical patient profiles 400.
  • the one or more recommendations 1150 generated can include one or more recommended treatments or interventions.
  • the recommended treatments or interventions can be automatically selected from a group consisting of all the interventions 504 for each of the identified historical patient profiles 400.
  • the recommendation 1150 may include at least one intervention 504 selected from at least one of the identified historical patient profiles 400.
  • the method 1200 can include: in a step 1205, selecting a subject patient profile for a subject patient, wherein the subject patient profile comprises medical information associated with the subject patient; in a step 1210, generating a first prediction of a first outcome parameter for the subject patient using a machine-learning algorithm; in a step 1215, identifying one or more historical patient profiles using historical medical records stored in a database by comparing the medical information of the subject patient profile with the historical medical records; in a step 1220, generating a second prediction of the first outcome parameter using the machine- learning algorithm for each of the one or more historical patient profiles identified; in a step 1225, generating an output comprising the first prediction generated for the subject patient, one or more actual outcomes of each historical patient profile identified, and the second prediction generated for each of the historical patient profiles; and in a step 1230, reporting the output to a healthcare professional via a
  • the method 1200 may further include a step 1222, where one or more interventions are extracted from the historical medical records for each of the identified historical patient profiles.
  • the method 1200 can also the following: in a step 1223, determining at least a reliability level associated with the first prediction of one or more outcome parameter that was generated for the subject patient; and in a step 1224, generating a recommendation for the subject patient based on at least the reliability level, wherein the recommendation comprises at least one recommended treatment or intervention.
  • a reliability level may be determined for each outcome parameter that was predicted for the subject patient.
  • the reliability level may be determined by comparing the predicted outcomes for the subject patient with at least one of the following: (1) the predicted outcomes for the one or more historical patient profiles; and (2) the one or more actual outcomes associated with one or more of the historical patient profiles.
  • the output generated in a step 1225 may thus further include any recommendations generated based on the determined reliability levels.
  • the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and not excluding any combinations of elements in the list of elements.
  • This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified.
  • phrases such as “OUTCOME PARAM(O)”, “OUTCOME_ PARAM(n)”, “MED PARAM(O)”, “MED PARAM(n)”, “INTERVENTION(O)”, “INTER- VENTION(n)”, “P OUTCOME PARAM(O)”, “P OUTCOME PARAM(n)”, and the like should be understood to refer to a grouping of one or more corresponding types of elements (e.g., medical parameters, outcome parameters, interventions, predictions, etc.). It should also be understood that the indices zero to n indicate a list or other type of data structure containing these elements, where n is an integer greater than zero. This definition allows that these parameters may optionally be a computer-readable data structure with individually accessible memory locations zero to n, where n is an integer greater than zero.
  • the present disclosure can be implemented as a system, a method, and/or a computer program product at any possible technical detail level of integration
  • the computer program product can include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.
  • the computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device.
  • the computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • a non-exhaustive list of more specific examples of the computer readable storage medium comprises the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing.
  • RAM random access memory
  • ROM read-only memory
  • EPROM or Flash memory erasable programmable read-only memory
  • SRAM static random access memory
  • CD-ROM compact disc read-only memory
  • DVD digital versatile disk
  • memory stick a floppy disk
  • a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon
  • a computer readable storage medium is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.
  • Computer readable program instructions described herein can be downloaded to respective computing/processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and/or a wireless network.
  • the network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers.
  • a network adapter card or network interface in each computing/processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing/processing device.
  • Computer readable program instructions for carrying out operations of the present disclosure can be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, configuration data for integrated circuitry, or either source code or object code written in any combination of one or more programming languages, comprising an object oriented programming language such as Smalltalk, C++, or the like, and procedural programming languages, such as the “C” programming language or similar programming languages.
  • the computer readable program instructions can execute entirely on the user’s computer, partly on the user’s computer, as a standalone software package, partly on the user’s computer and partly on a remote computer or entirely on the remote computer or server.
  • the remote computer can be connected to the user's computer through any type of network, comprising a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).
  • electronic circuitry comprising, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.
  • the computer readable program instructions can be provided to a processor of a, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
  • These computer readable program instructions can also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and/or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture comprising instructions which implement aspects of the function/act specified in the flowchart and/or block diagram or blocks.
  • the computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions/acts specified in the flowchart and/or block diagram block or blocks.
  • each block in the flowchart or block diagrams can represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s).
  • the functions noted in the blocks can occur out of the order noted in the Figures.
  • two blocks shown in succession can, in fact, be executed substantially concurrently, or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved.
  • inventive embodiments are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, inventive embodiments may be practiced otherwise than as specifically described and claimed.
  • inventive embodiments of the present disclosure are directed to each individual feature, system, article, material, kit, and/or method described herein.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Data Mining & Analysis (AREA)
  • Epidemiology (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

La présente divulgation concerne des systèmes et des procédés de fourniture de prédictions liées aux soins de santé par apprentissage automatique, avec une fiabilité et une contextualisation améliorées. Selon certains aspects de la présente divulgation, les systèmes et les procédés ici décrits permettent à des fournisseurs de soins de santé d'utiliser des modèles d'apprentissage automatique de manière plus confortable et fiable en incorporant des informations pertinentes de patients historiques dans la sortie délivrée en retour aux fournisseurs. Par conséquent, les systèmes et les procédés décrits permettent de surmonter de nombreux inconvénients des prédictions classiques par modèle d'apprentissage automatique, qui ne peuvent pas tenir compte de nuances subtiles dans les dossiers médicaux et les interventions des patients du fait qu'elles se rapportent à des résultats de patient.
PCT/EP2023/071495 2022-08-11 2023-08-03 Systèmes et procédés de prédiction par apprentissage automatique dans le contexte de données de traitements historiques WO2024033201A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2022111820 2022-08-11
CNPCT/CN2022/111820 2022-08-11

Publications (1)

Publication Number Publication Date
WO2024033201A1 true WO2024033201A1 (fr) 2024-02-15

Family

ID=87571194

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2023/071495 WO2024033201A1 (fr) 2022-08-11 2023-08-03 Systèmes et procédés de prédiction par apprentissage automatique dans le contexte de données de traitements historiques

Country Status (1)

Country Link
WO (1) WO2024033201A1 (fr)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150112710A1 (en) * 2012-06-21 2015-04-23 Battelle Memorial Institute Clinical predictive analytics system
US20200402665A1 (en) * 2019-06-19 2020-12-24 GE Precision Healthcare LLC Unplanned readmission prediction using an interactive augmented intelligent (iai) system
WO2021222802A1 (fr) * 2020-04-30 2021-11-04 Arine, Inc. Recommandation de traitement

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150112710A1 (en) * 2012-06-21 2015-04-23 Battelle Memorial Institute Clinical predictive analytics system
US20200402665A1 (en) * 2019-06-19 2020-12-24 GE Precision Healthcare LLC Unplanned readmission prediction using an interactive augmented intelligent (iai) system
WO2021222802A1 (fr) * 2020-04-30 2021-11-04 Arine, Inc. Recommandation de traitement

Similar Documents

Publication Publication Date Title
US20220101967A1 (en) Methods for automatic cohort selection in epidemiologic studies and clinical trials
US7917377B2 (en) Patient data mining for automated compliance
US11923094B2 (en) Monitoring predictive models
US9257052B2 (en) Evaluating candidate answers to questions in a target knowledge domain
WO2016036831A1 (fr) Système de génération et de mise à jour de directives de traitement et d'estimation d'importance de l'effet d'étapes de traitement
CN111382275A (zh) 医疗知识图谱的构建方法、装置、介质及电子设备
US20200058408A1 (en) Systems, methods, and apparatus for linking family electronic medical records and prediction of medical conditions and health management
Xiang et al. Integrated Architectures for Predicting Hospital Readmissions Using Machine Learning
JPWO2016120955A1 (ja) 行動予測装置、行動予測装置の制御方法、および行動予測装置の制御プログラム
US11527312B2 (en) Clinical report retrieval and/or comparison
US20210375443A1 (en) System and Method Associated with Determining Physician Attribution Related to In-Patient Care Using Prediction-Based Analysis
WO2021103624A1 (fr) Appareil d'alerte précoce de septicémie, dispositif, et support de stockage
EP4131279A1 (fr) Procédé de moteur d'expérience et appareil d'apprentissage à partir de patients similaires
US20190179883A1 (en) Evaluating textual annotation model performance
US20190287675A1 (en) Systems and methods for determining healthcare quality measures by evalutating subject healthcare data in real-time
US20230060794A1 (en) Diagnostic Tool
WO2024033201A1 (fr) Systèmes et procédés de prédiction par apprentissage automatique dans le contexte de données de traitements historiques
US20230068453A1 (en) Methods and systems for determining and displaying dynamic patient readmission risk and intervention recommendation
US20190035502A1 (en) Systems and methods for managing care teams
Shade et al. COVID-HEART: development and validation of a multi-variable model for real-time prediction of cardiovascular complications in hospitalized patients with COVID-19
US11694801B2 (en) Identifying and extracting stimulus-response variables from electronic health records
US20200051673A1 (en) Generation of concept scores based on analysis of clinical data
US20230317222A1 (en) Machine learning-based electronic health record prediction
US20240186021A1 (en) Monitoring predictive models
US20230041051A1 (en) Methods and systems for predicting and preventing frequent patient readmission

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23754174

Country of ref document: EP

Kind code of ref document: A1