Patent Application
20230153663
Various embodiments of the present invention address technical challenges related to performing predictive data analysis operations and address the efficiency and reliability shortcomings of various existing predictive data analysis solutions.
In general, embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for performing predictive data analysis operations. For example, certain embodiments of the present invention utilize systems, methods, and computer program products that perform predictive data analysis operations by an end-to-end machine learning framework that performs at least the following steps/operations: (i) a service request data object is processed by a diagnosis prediction machine learning model to generate a probabilistic diagnosis data object, (ii) the probabilistic diagnosis data object is processed by the hybrid diagnosis-provider classification machine learning model to generate a variable-length classification for the service request data object, and (iii) the variable-length classification is processed by a recommendation scoring machine learning model to generate a consultation recommendation score for the service request data object.
In accordance with one aspect, a method is provided. In one embodiment, the method comprises: identifying a probabilistic diagnosis data object associated with the service request data object, wherein the probabilistic diagnosis data object is generated by a diagnosis prediction machine learning model based at least in part on diagnosis input data associated with the service request data object; determining, using a hybrid diagnosis-provider classification machine learning model and based at least in part on the probabilistic diagnosis data object and the provider data object, a variable-length classification for the service request data object, wherein: (i) the variable-length classification maps the service request data object to a variable-length subset of a plurality of candidate classes, and (ii) the plurality of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes; determining, based at least in part on a recommendation scoring machine learning model and based at least in part on the variable-length classification, a consultation recommendation score for the service request data object, wherein: (i) each candidate class is associated with a recommendation score, and (ii) the consultation recommendation score is determined based at least in part on each recommendation score for the variable-length subset; performing one or more prediction-based actions based at least in part on the noted consultation recommendation score.
In accordance with another aspect, a computer program product is provided. The computer program product may comprise at least one computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions comprising executable portions configured to: identify a probabilistic diagnosis data object associated with the service request data object, wherein the probabilistic diagnosis data object is generated by a diagnosis prediction machine learning model based at least in part on diagnosis input data associated with the service request data object; determine, using a hybrid diagnosis-provider classification machine learning model and based at least in part on the probabilistic diagnosis data object and the provider data object, a variable-length classification for the service request data object, wherein: (i) the variable-length classification maps the service request data object to a variable-length subset of a plurality of candidate classes, and (ii) the plurality of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes; determine, based at least in part on a recommendation scoring machine learning model and based at least in part on the variable-length classification, a consultation recommendation score for the service request data object, wherein: (i) each candidate class is associated with a recommendation score, and (ii) the consultation recommendation score is determined based at least in part on each recommendation score for the variable-length subset; perform one or more prediction-based actions based at least in part on the consultation recommendation score.
In accordance with yet another aspect, an apparatus comprising at least one processor and at least one memory including computer program code is provided. In one embodiment, the at least one memory and the computer program code may be configured to, with the processor, cause the apparatus to: identify a probabilistic diagnosis data object associated with the service request data object, wherein the probabilistic diagnosis data object is generated by a diagnosis prediction machine learning model based at least in part on diagnosis input data associated with the service request data object; determine, using a hybrid diagnosis-provider classification machine learning model and based at least in part on the probabilistic diagnosis data object and the provider data object, a variable-length classification for the service request data object, wherein: (i) the variable-length classification maps the service request data object to a variable-length subset of a plurality of candidate classes, and (ii) the plurality of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes; determine, based at least in part on a recommendation scoring machine learning model and based at least in part on the variable-length classification, a consultation recommendation score for the service request data object, wherein: (i) each candidate class is associated with a recommendation score, and (ii) the consultation recommendation score is determined based at least in part on each recommendation score for the variable-length subset; perform one or more prediction-based actions based at least in part on the consultation recommendation score.
Having thus described the invention in general terms, reference will now be made to the accompanying drawings, which are not necessarily drawn to scale, and wherein:
Various embodiments of the present invention now will be described more fully hereinafter with reference to the accompanying drawings, in which some, but not all, embodiments of the inventions are shown. Indeed, these inventions may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will satisfy applicable legal requirements. The term “or” is used herein in both the alternative and conjunctive sense, unless otherwise indicated. The terms “illustrative” and “exemplary” are used to be examples with no indication of quality level. Like numbers refer to like elements throughout. Moreover, while certain embodiments of the present invention are described with reference to predictive data analysis, one of ordinary skill in the art will recognize that the disclosed concepts can be used to perform other types of data analysis tasks.
Various embodiments of the present invention introduce techniques for using the output of a diagnosis prediction machine learning model to generate a recommendation score. By using the noted techniques, various embodiments of the present invention enable using pretrained diagnosis prediction machine learning models to generate a recommendation score, thus reducing or avoiding the need for computationally expensive operations that are performed to generate diagnosis prediction machine learning models. In this way, various embodiments of the present invention reduce the number of computational operations performed to generate recommendation scores, thus improving the computational efficiency of predictive data analysis systems that are configured to generate recommendation scores and make important technical contributions to the field of predictive data analysis.
For example, various embodiments of the present invention utilize systems, methods, and computer program products that perform predictive data analysis operations by an end-to-end machine learning framework that performs at least the following steps/operations: (i) a service request data object is processed by a diagnosis prediction machine learning model to generate a probabilistic diagnosis data object, (ii) the probabilistic diagnosis data object is processed by the hybrid diagnosis-provider classification machine learning model to generate a variable-length classification for the service request data object, and (iii) the variable-length classification is processed by a recommendation scoring machine learning model to generate a consultation recommendation score for the service request data object. By using the noted techniques, various embodiments of the present invention use a hybrid diagnosis-provider classification machine learning model and a recommendation scoring machine learning model to map the output of a diagnosis prediction machine learning model (e.g., a pretrained diagnosis prediction machine learning model) to a consultation recommendation score, thus in some embodiments performing transfer learning to use a diagnosis prediction machine learning model to generate recommendation scores that are different from diagnosis scores generated by the diagnosis prediction machine learning model.
In some embodiments, the diagnosis prediction machine learning model is a pretrained machine learning model whose output with respect to a service request data object is used to generate a dynamic-length classification for the service request data object. In other words, various embodiments of the present invention perform transfer learning on a diagnosis prediction machine learning model by mapping the outputs of the diagnosis prediction machine learning model to a variable-length subset of a plurality of candidate classes. In some embodiments, a diagnosis prediction machine learning model (e.g., a pretrained diagnosis prediction machine learning model or an untrained diagnosis prediction machine learning model that is trained from scratch) is trained as part of an end-to-end machine learning framework that comprises the diagnosis prediction machine learning model, a hybrid diagnosis-provider classification machine learning model, and a recommendation scoring machine learning model.
An exemplary application of various embodiments of the present invention relates to determining when physicians may need a consultation arrangement. In some embodiments, the following operations are performed: receiving data describing user need for medical care; receiving data describing user selection of provider and/or clinical facility; receiving data describing user descriptions of symptoms; receiving data describing user-provided multimedia (e.g., videos, photographs, etc.); loading the above-described received data into a diagnostic agent; generating a probabilistic diagnosis using an existing/external probabilistic diagnostic model; providing a combination of the received data and the probabilistic diagnosis (e.g., probabilistic diagnosis plus patient location; probabilistic diagnosis plus patient health provider data, probabilistic diagnosis plus raw data provided by patient, etc.); using the consultation engine to determine a peer/expert consult prediction, prompting the attending physician on need for peer/expert consult; matching other physicians for a potential consult; and scheduling the consult and/or notifying remote physician of consult need.
The term “diagnosis prediction machine learning model” may refer to a data construct that is configured to describe parameters, hyper-parameters, and/or defined operations of a model that is configured to process diagnosis input data for a service request data object in order to generate a probabilistic diagnosis data object for the service request data object. In some embodiments, the diagnosis prediction machine learning model is a pretrained machine learning model whose output with respect to a service request data object is used to generate a dynamic-length classification for the service request data object. In other words, various embodiments of the present invention perform transfer learning on a diagnosis prediction machine learning model by mapping the outputs of the diagnosis prediction machine learning model to a variable-length subset of a plurality of candidate classes. In some embodiments, a diagnosis prediction machine learning model (e.g., a pretrained diagnosis prediction machine learning model or an untrained diagnosis prediction machine learning model that is trained from scratch) is trained as part of an end-to-end machine learning framework that comprises the diagnosis prediction machine learning model, a hybrid diagnosis-provider classification machine learning model, and a recommendation scoring machine learning model, wherein the end-to-end machine learning framework may operate using at least the following steps/operations: (i) a service request data object is processed by the diagnosis prediction machine learning model to generate a probabilistic diagnosis data object, (ii) the probabilistic diagnosis data object is processed by the hybrid diagnosis-provider classification machine learning model to generate a variable-length classification for the service request data object, and (iii) the variable-length classification is processed by the recommendation scoring machine learning model to generate a consultation recommendation score for the service request data object. In some embodiments, during training of the end-to-end machine learning framework, any trainable parameters of the diagnosis prediction machine learning model, the hybrid diagnosis-provider classification machine learning model, and the recommendation scoring machine learning model are trained based at least in part on inferred consultation recommendation scores for service request data objects and ground-truth consultation outcomes for the service request data objects as determined based at least in part on historical data associated with the service request data object. For example, if an inferred consultation recommendation score for a service request data object is 0.4 but the ground-truth consultation outcome for the service request data object describes an affirmative consultation outcome (e.g., describes that the patient was scheduled for a consult in response to the service request data object), then a deviation measure of 1−0.4=0.6 may be used as part of an error measure that is used to train the end-to-end machine learning framework by updating any trainable parameters of the diagnosis prediction machine learning model, the hybrid diagnosis-provider classification machine learning model, and the recommendation scoring machine learning model in a manner that is configured to optimize the error measure. In some embodiments, inputs to a diagnosis prediction machine learning model include one or more vectors describing diagnostic input data for a service request data object, while outputs of a diagnosis prediction machine learning model include a vector describing n diagnostic probabilities for m candidate conditions, where those candidate conditions having a threshold-satisfying diagnostic probability may be deemed to be a diagnosed condition described by the vector.
The term “probabilistic diagnosis data object” may refer to a data construct that is configured to describe, for a particular service request data object that is associated with particular diagnosis input data, a set of diagnosed conditions along with a diagnosis probability for each diagnosed condition. For example, a probabilistic diagnosis data object may describe that a service request data object is associated with a diagnosis D1 with the probability P1, a diagnosis D2 with the probability P2, and so on. In some embodiments, the probabilistic diagnosis data object is generated by: (i) obtaining a set of raw diagnosis probabilities for a set of conditions from a diagnosis prediction machine learning model, (ii) normalizing the set of raw diagnosis probabilities to generate the set of diagnosis probabilities for the set of conditions, and (iii) generating the probabilistic diagnosis data object based at least in part on the set of diagnosis probabilities. For example, in some embodiments, the diagnosis prediction machine learning model may include n machine learning components, where each machine learning component is configured to generate a diagnosis probability for a condition of n conditions. In the noted example, the output of each machine learning component may have a different range. Thus, to generate the diagnosis probabilities described by the probabilistic diagnosis data object, the n diagnosis probabilities generated by the n machine learning components may be normalized to have a unified range (e.g., a unified range of [0, 1]), and then the normalized diagnosis probabilities may be used to generate the diagnosis probability data object. For example, the diagnosis probability data object may describe all n normalized diagnosis probabilities, the top m of the normalized diagnosis probabilities (where m may be a predefined value that is smaller than n), and/or those normalized diagnosis probabilities that satisfy (e.g., exceed) a normalized diagnosis probability threshold. In some embodiments, the diagnosis probabilities described by a probabilistic diagnosis data object are determined based at least in part on diagnosis input data associated with a corresponding service request data object, e.g., one or more service fields associated with the service request data object, one or more medical history fields associated with a patient identifier that is associated with the service request data object, and/or the like.
The term “hybrid diagnosis-provider classification machine learning model” may refer to a data construct that is configured to describe parameters, hyper-parameters, and/or defined operations of a model that is configured to process a probabilistic diagnosis data object for a service request data object and a provider data object for a service request data object to generate a variable-length classification for the service request data object. In some embodiments, the hybrid diagnosis-provider classification machine learning model may be configured to: (i) map the probabilistic diagnosis data object for a service request data object to a variable-length subset of one or more diagnosis-based classes, (ii) map the provider data object for a service request data object to a variable-length subset of one or more provider-based classes, and/or (iii) map a combination of the probabilistic diagnosis data object for a service request data object and a provider data object for a service request data object to a variable-length subset of one or more hybrid classes. In some of the noted embodiments, given a service request data object S1 that is associated with a probabilistic data object PD1 and a provider data object P1, the hybrid diagnosis-provider classification machine learning model may determine: (i) of a set of available classes for the P1, e of the available classes that P1 actually corresponds to (where e is a variable number), (ii) of a set of available classes for the PD1, f of the available classes that PD1 actually corresponds to (where f is a variable number), and (iii) of a set of available classes for the combination P1-PD1, g of the available classes that P1-PD1 actually corresponds to (where g is a variable number). In other words, the hybrid diagnosis-provider classification machine learning model may be configured to determine, based at least in part on P1 and PD1, a variable-length classification for S1, where the variable-length classification maps S1 to a variable-length subset of a plurality of candidate classes, where the plurality of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes. In some embodiments, inputs to a hybrid diagnosis-provider classification machine learning model include a vector describing a probabilistic diagnosis data object and one or more vectors describing data extracted from a provider data object, while outputs of a hybrid diagnosis-provider classification machine learning model include a vector that describes, for each candidate classification, whether the variable-length classification generated by the hybrid diagnosis-provider classification machine learning model indicates mapping of the candidate classification to a particular service request data object.
The term “variable-length classification” may refer to a data construct that is configured to describe whether an input data object (e.g., a service request data object) maps to any of a set of candidate classes. In some embodiments, the variable-length classification maps a service request data object to h candidate classes of a set of candidate classes, where h>=0. In some embodiments, the set of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes, which are described in greater detail below.
The term “diagnosis-based class” may refer to a data construct that is configured to describe an available label for a service request data object, where the service request data object may be mapped to the available label if a probabilistic diagnosis data object for the service request data object satisfies one or more conditions associated with the available label. Examples of diagnosis-based classes include a complexity class to which a service request data object is mapped if the probabilistic diagnosis data object for the service request data object describes a threshold-satisfying diagnosis probability for a diagnosed condition that is listed as being complex, unusual, and/or urgent; and an urgency class to which a service request data object is mapped if the probabilistic diagnosis data object for the service request data object describes a threshold-satisfying diagnosis probability for a diagnosed condition that is listed as being in need of time-sensitive-treatment. In some embodiments, when a probabilistic diagnosis data object for a service request data object includes n diagnosed probabilities for n conditions, then mapping the service request data object to a diagnosis-based class comprises: (i) determining a related subset of the n diagnosed conditions that are listed as being related to the diagnosis-based class (e.g., for a complexity class, determining a related subset of the n diagnosed conditions that are listed as being complex, unusual, and/or urgent), (ii) combining (e.g., summing, averaging, and/or the like) each diagnosis probability for a diagnosis condition in the related subset for the diagnosis-based class to determine a classification score for the diagnosis-based class, and (iii) mapping the service request data object to the diagnosis-based class if the classification score for the diagnosis-based class satisfies (e.g., exceeds) a classification score threshold.
The term “provider-based class” may refer to a data construct that is configured to describe an available label for a service request data object, where the service request data object may be mapped to the available label if a provider data object for the service request data object satisfies one or more conditions associated with the available label. Examples of provider-based classes include an availability class to which a service request data object is mapped if a provider schedule described by the provider data object for the service request data object includes available time slots for scheduling a service appointment associated with the service request data object, and a past interaction class to which a service request data object is mapped if a patient/service recipient identifier associated with the service request data object has had past interactions with the provider identifier associated with the provider data object for the service request data object as described by provider history data in the provider data object.
The term “hybrid diagnosis class” may refer to a data construct that is configured to describe an available label for a service request data object, where the service request data object may be mapped to the available label if a combination of the provider data object associated with the service request data object and the probabilistic diagnosis data object associated with the service request data object satisfies one or more conditions associated with the available label. Examples of hybrid diagnosis classes include an individual past service class that is mapped to a service request data object if provider history data described by the provider data object for the service request data object describes past consultation by the provider identifier associated with the provider data object in relation to one or more diagnosed conditions described by the probabilistic diagnosis data object for the service request data object, a facility-based past service class that is mapped to a service request data object if provider facility history data described by the provider data object for the service request data object describes past consultation by a facility identifier associated with the provider data object in relation to one or more diagnosed conditions described by the probabilistic diagnosis data object for the service request data object, and a suboptimal past service class that is mapped to a service request data object if provider history data described by the provider data object for the service request data object describes suboptimal past service delivery by the provider identifier associated with the provider data object in relation to one or more diagnosed conditions described by the probabilistic diagnosis data object for the service request data object.
The term “recommendation scoring machine learning model” may refer to a data construct that is configured to describe parameters, hyper-parameters, and/or defined operations of a model that is configured to process a variable-length classification for a service request data object to generate a consultation recommendation score for the service request data object. In some embodiments, the recommendation scoring machine learning model combines each recommendation score for a variable-length subset of candidate classifications that are associated with the variable length classification to generate the consultation recommendation score for the service request data object. For example, consider a service request data object that is associated with a variable-length classification that describes that the service request data object is associated with candidate classifications C1-C3, where the candidate classifications C1-C3 are associated with recommendation scores R1-R3 respectively. In some embodiments, the consultation recommendation score for the service request data object may be determined based at least in part on a combination of C1-C3 (e.g., based at least in part on
In some embodiments, the recommendation score for a candidate classification may describe a degree of correlation between a mapping of the candidate classification to a service request data object and whether a provider consult should be scheduled with respect to the service request data object. In some embodiments, each recommendation score for a candidate classification is a trained parameter of the recommendation scoring machine learning model. In some embodiments, inputs to a recommendation scoring machine learning model comprise a vector describing a variable-length classification, while outputs of a recommendation scoring machine learning model comprise a vector and/or an atomic value describing a consultation recommendation score.
Embodiments of the present invention may be implemented in various ways, including as computer program products that comprise articles of manufacture. Such computer program products may include one or more software components including, for example, software objects, methods, data structures, or the like. A software component may be coded in any of a variety of programming languages. An illustrative programming language may be a lower-level programming language such as an assembly language associated with a particular hardware architecture and/or operating system platform. A software component comprising assembly language instructions may require conversion into executable machine code by an assembler prior to execution by the hardware architecture and/or platform. Another example programming language may be a higher-level programming language that may be portable across multiple architectures. A software component comprising higher-level programming language instructions may require conversion to an intermediate representation by an interpreter or a compiler prior to execution.
Other examples of programming languages include, but are not limited to, a macro language, a shell or command language, a job control language, a scripting language, a database query or search language, and/or a report writing language. In one or more example embodiments, a software component comprising instructions in one of the foregoing examples of programming languages may be executed directly by an operating system or other software components without having to be first transformed into another form. A software component may be stored as a file or other data storage construct. Software components of a similar type or functionally related may be stored together such as, for example, in a particular directory, folder, or library. Software components may be static (e.g., pre-established or fixed) or dynamic (e.g., created or modified at the time of execution).
A computer program product may include a non-transitory computer-readable storage medium storing applications, programs, program modules, scripts, source code, program code, object code, byte code, compiled code, interpreted code, machine code, executable instructions, and/or the like (also referred to herein as executable instructions, instructions for execution, computer program products, program code, and/or similar terms used herein interchangeably). Such non-transitory computer-readable storage media include all computer-readable media (including volatile and non-volatile media).
In one embodiment, a non-volatile computer-readable storage medium may include a floppy disk, flexible disk, hard disk, solid-state storage (SSS) (e.g., a solid-state drive (SSD), solid-state card (SSC), solid-state module (SSM), enterprise flash drive, magnetic tape, or any other non-transitory magnetic medium, and/or the like. A non-volatile computer-readable storage medium may also include a punch card, paper tape, optical mark sheet (or any other physical medium with patterns of holes or other optically recognizable indicia), compact disc read-only memory (CD-ROM), compact disc-rewritable (CD-RW), digital versatile disc (DVD), Blu-ray disc (BD), any other non-transitory optical medium, and/or the like. Such a non-volatile computer-readable storage medium may also include read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory (e.g., Serial, NAND, NOR, and/or the like), multimedia memory cards (MMC), secure digital (SD) memory cards, SmartMedia cards, CompactFlash (CF) cards, Memory Sticks, and/or the like. Further, a non-volatile computer-readable storage medium may also include conductive-bridging random access memory (CBRAM), phase-change random access memory (PRAM), ferroelectric random-access memory (FeRAM), non-volatile random-access memory (NVRAM), magnetoresistive random-access memory (MRAM), resistive random-access memory (RRAM), Silicon-Oxide-Nitride-Oxide-Silicon memory (SONOS), floating junction gate random access memory (FJG RAM), Millipede memory, racetrack memory, and/or the like.
In one embodiment, a volatile computer-readable storage medium may include random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), fast page mode dynamic random access memory (FPM DRAM), extended data-out dynamic random access memory (EDO DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), double data rate type two synchronous dynamic random access memory (DDR2 SDRAM), double data rate type three synchronous dynamic random access memory (DDR3 SDRAM), Rambus dynamic random access memory (RDRAM), Twin Transistor RAM (TTRAM), Thyristor RAM (T-RAM), Zero-capacitor (Z-RAM), Rambus in-line memory module (RIMM), dual in-line memory module (DIMM), single in-line memory module (SIMM), video random access memory (VRAM), cache memory (including various levels), flash memory, register memory, and/or the like. It will be appreciated that where embodiments are described to use a computer-readable storage medium, other types of computer-readable storage media may be substituted for or used in addition to the computer-readable storage media described above.
As should be appreciated, various embodiments of the present invention may also be implemented as methods, apparatus, systems, computing devices, computing entities, and/or the like. As such, embodiments of the present invention may take the form of an apparatus, system, computing device, computing entity, and/or the like executing instructions stored on a computer-readable storage medium to perform certain steps or operations. Thus, embodiments of the present invention may also take the form of an entirely hardware embodiment, an entirely computer program product embodiment, and/or an embodiment that comprises a combination of computer program products and hardware performing certain steps or operations.
Embodiments of the present invention are described below with reference to block diagrams and flowchart illustrations. Thus, it should be understood that each block of the block diagrams and flowchart illustrations may be implemented in the form of a computer program product, an entirely hardware embodiment, a combination of hardware and computer program products, and/or apparatus, systems, computing devices, computing entities, and/or the like carrying out instructions, operations, steps, and similar words used interchangeably (e.g., the executable instructions, instructions for execution, program code, and/or the like) on a computer-readable storage medium for execution. For example, retrieval, loading, and execution of code may be performed sequentially such that one instruction is retrieved, loaded, and executed at a time. In some exemplary embodiments, retrieval, loading, and/or execution may be performed in parallel such that multiple instructions are retrieved, loaded, and/or executed together. Thus, such embodiments can produce specifically-configured machines performing the steps or operations specified in the block diagrams and flowchart illustrations. Accordingly, the block diagrams and flowchart illustrations support various combinations of embodiments for performing the specified instructions, operations, or steps.
In some embodiments, predictive data analysis system 101 may communicate with at least one of the client computing entities 102 using one or more communication networks. Examples of communication networks include any wired or wireless communication network including, for example, a wired or wireless local area network (LAN), personal area network (PAN), metropolitan area network (MAN), wide area network (WAN), or the like, as well as any hardware, software and/or firmware required to implement it (such as, e.g., network routers, and/or the like).
The predictive data analysis system 101 may include a predictive data analysis computing entity 106 and a storage subsystem 108. The predictive data analysis computing entity 106 may be configured to receive predictive data analysis requests from one or more client computing entities 102, process the predictive data analysis requests to generate predictions corresponding to the predictive data analysis requests, provide the generated predictions to the client computing entities 102, and automatically perform prediction-based actions based at least in part on the generated predictions.
The storage subsystem 108 may be configured to store input data used by the predictive data analysis computing entity 106 to perform predictive data analysis as well as model definition data used by the predictive data analysis computing entity 106 to perform various predictive data analysis tasks. The storage subsystem 108 may include one or more storage units, such as multiple distributed storage units that are connected through a computer network. Each storage unit in the storage subsystem 108 may store at least one of one or more data assets and/or one or more data about the computed properties of one or more data assets. Moreover, each storage unit in the storage subsystem 108 may include one or more non-volatile storage or memory media including, but not limited to, hard disks, ROM, PROM, EPROM, EEPROM, flash memory, MMCs, SD memory cards, Memory Sticks, CBRAM, PRAM, FeRAM, NVRAM, MRAM, RRAM, SONOS, FJG RAM, Millipede memory, racetrack memory, and/or the like.
As indicated, in one embodiment, the predictive data analysis computing entity 106 may also include one or more communications interfaces 220 for communicating with various computing entities, such as by communicating data, content, information, and/or similar terms used herein interchangeably that can be transmitted, received, operated on, processed, displayed, stored, and/or the like.
As shown in
For example, the processing element 205 may be embodied as one or more complex programmable logic devices (CPLDs), microprocessors, multi-core processors, coprocessing entities, application-specific instruction-set processors (ASIPs), microcontrollers, and/or controllers. Further, the processing element 205 may be embodied as one or more other processing devices or circuitry. The term circuitry may refer to an entirely hardware embodiment or a combination of hardware and computer program products. Thus, the processing element 205 may be embodied as integrated circuits, application-specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), hardware accelerators, other circuitry, and/or the like.
As will therefore be understood, the processing element 205 may be configured for a particular use or configured to execute instructions stored in volatile or non-volatile media or otherwise accessible to the processing element 205. As such, whether configured by hardware or computer program products, or by a combination thereof, the processing element 205 may be capable of performing steps or operations according to embodiments of the present invention when configured accordingly.
In one embodiment, the predictive data analysis computing entity 106 may further include, or be in communication with, non-volatile media (also referred to as non-volatile storage, memory, memory storage, memory circuitry and/or similar terms used herein interchangeably). In one embodiment, the non-volatile storage or memory may include one or more non-volatile storage or memory media 210, including, but not limited to, hard disks, ROM, PROM, EPROM, EEPROM, flash memory, MMCs, SD memory cards, Memory Sticks, CBRAM, PRAM, FeRAM, NVRAM, MRAM, RRAM, SONOS, FJG RAM, Millipede memory, racetrack memory, and/or the like.
As will be recognized, the non-volatile storage or memory media may store databases, database instances, database management systems, data, applications, programs, program modules, scripts, source code, object code, byte code, compiled code, interpreted code, machine code, executable instructions, and/or the like. The term database, database instance, database management system, and/or similar terms used herein interchangeably may refer to a collection of records or data that is stored in a computer-readable storage medium using one or more database models, such as a hierarchical database model, network model, relational model, entity-relationship model, object model, document model, semantic model, graph model, and/or the like.
In one embodiment, the predictive data analysis computing entity 106 may further include, or be in communication with, volatile media (also referred to as volatile storage, memory, memory storage, memory circuitry and/or similar terms used herein interchangeably). In one embodiment, the volatile storage or memory may also include one or more volatile storage or memory media 215, including, but not limited to, RAM, DRAM, SRAM, FPM DRAM, EDO DRAM, SDRAM, DDR SDRAM, DDR2 SDRAM, DDR3 SDRAM, RDRAM, TTRAM, T-RAM, Z-RAM, RIMM, DIMM, SIMM, VRAM, cache memory, register memory, and/or the like.
As will be recognized, the volatile storage or memory media may be used to store at least portions of the databases, database instances, database management systems, data, applications, programs, program modules, scripts, source code, object code, byte code, compiled code, interpreted code, machine code, executable instructions, and/or the like being executed by, for example, the processing element 205. Thus, the databases, database instances, database management systems, data, applications, programs, program modules, scripts, source code, object code, byte code, compiled code, interpreted code, machine code, executable instructions, and/or the like may be used to control certain aspects of the operation of the predictive data analysis computing entity 106 with the assistance of the processing element 205 and operating system.
As indicated, in one embodiment, the predictive data analysis computing entity 106 may also include one or more communications interfaces 220 for communicating with various computing entities, such as by communicating data, content, information, and/or similar terms used herein interchangeably that can be transmitted, received, operated on, processed, displayed, stored, and/or the like. Such communication may be executed using a wired data transmission protocol, such as fiber distributed data interface (FDDI), digital subscriber line (DSL), Ethernet, asynchronous transfer mode (ATM), frame relay, data over cable service interface specification (DOCSIS), or any other wired transmission protocol. Similarly, the predictive data analysis computing entity 106 may be configured to communicate via wireless external communication networks using any of a variety of protocols, such as general packet radio service (GPRS), Universal Mobile Telecommunications System (UMTS), Code Division Multiple Access 2000 (CDMA2000), CDMA2000 1× (1×RTT), Wideband Code Division Multiple Access (WCDMA), Global System for Mobile Communications (GSM), Enhanced Data rates for GSM Evolution (EDGE), Time Division-Synchronous Code Division Multiple Access (TD-SCDMA), Long Term Evolution (LTE), Evolved Universal Terrestrial Radio Access Network (E-UTRAN), Evolution-Data Optimized (EVDO), High Speed Packet Access (HSPA), High-Speed Downlink Packet Access (HSDPA), IEEE 802.11 (Wi-Fi), Wi-Fi Direct, 802.16 (WiMAX), ultra-wideband (UWB), infrared (IR) protocols, near field communication (NFC) protocols, Wibree, Bluetooth protocols, wireless universal serial bus (USB) protocols, and/or any other wireless protocol.
Although not shown, the predictive data analysis computing entity 106 may include, or be in communication with, one or more input elements, such as a keyboard input, a mouse input, a touch screen/display input, motion input, movement input, audio input, pointing device input, joystick input, keypad input, and/or the like. The predictive data analysis computing entity 106 may also include, or be in communication with, one or more output elements (not shown), such as audio output, video output, screen/display output, motion output, movement output, and/or the like.
The signals provided to and received from the transmitter 304 and the receiver 306, correspondingly, may include signaling information/data in accordance with air interface standards of applicable wireless systems. In this regard, the client computing entity 102 may be capable of operating with one or more air interface standards, communication protocols, modulation types, and access types. More particularly, the client computing entity 102 may operate in accordance with any of a number of wireless communication standards and protocols, such as those described above with regard to the predictive data analysis computing entity 106. In a particular embodiment, the client computing entity 102 may operate in accordance with multiple wireless communication standards and protocols, such as UMTS, CDMA2000, 1×RTT, WCDMA, GSM, EDGE, TD-SCDMA, LTE, E-UTRAN, EVDO, HSPA, HSDPA, Wi-Fi, Wi-Fi Direct, WiMAX, UWB, IR, NFC, Bluetooth, USB, and/or the like. Similarly, the client computing entity 102 may operate in accordance with multiple wired communication standards and protocols, such as those described above with regard to the predictive data analysis computing entity 106 via a network interface 320.
Via these communication standards and protocols, the client computing entity 102 can communicate with various other entities using concepts such as Unstructured Supplementary Service Data (USSD), Short Message Service (SMS), Multimedia Messaging Service (MMS), Dual-Tone Multi-Frequency Signaling (DTMF), and/or Subscriber Identity Module Dialer (SIM dialer). The client computing entity 102 can also download changes, add-ons, and updates, for instance, to its firmware, software (e.g., including executable instructions, applications, program modules), and operating system.
According to one embodiment, the client computing entity 102 may include location determining aspects, devices, modules, functionalities, and/or similar words used herein interchangeably. For example, the client computing entity 102 may include outdoor positioning aspects, such as a location module adapted to acquire, for example, latitude, longitude, altitude, geocode, course, direction, heading, speed, universal time (UTC), date, and/or various other information/data. In one embodiment, the location module can acquire data, sometimes known as ephemeris data, by identifying the number of satellites in view and the relative positions of those satellites (e.g., using global positioning systems (GPS)). The satellites may be a variety of different satellites, including Low Earth Orbit (LEO) satellite systems, Department of Defense (DOD) satellite systems, the European Union Galileo positioning systems, the Chinese Compass navigation systems, Indian Regional Navigational satellite systems, and/or the like. This data can be collected using a variety of coordinate systems, such as the Decimal Degrees (DD); Degrees, Minutes, Seconds (DMS); Universal Transverse Mercator (UTM); Universal Polar Stereographic (UPS) coordinate systems; and/or the like. Alternatively, the location information/data can be determined by triangulating the client computing entity's 102 position in connection with a variety of other systems, including cellular towers, Wi-Fi access points, and/or the like. Similarly, the client computing entity 102 may include indoor positioning aspects, such as a location module adapted to acquire, for example, latitude, longitude, altitude, geocode, course, direction, heading, speed, time, date, and/or various other information/data. Some of the indoor systems may use various position or location technologies including RFID tags, indoor beacons or transmitters, Wi-Fi access points, cellular towers, nearby computing devices (e.g., smartphones, laptops) and/or the like. For instance, such technologies may include the iBeacons, Gimbal proximity beacons, Bluetooth Low Energy (BLE) transmitters, NFC transmitters, and/or the like. These indoor positioning aspects can be used in a variety of settings to determine the location of someone or something to within inches or centimeters.
The client computing entity 102 may also comprise a user interface (that can include a display 316 coupled to a processing element 308) and/or a user input interface (coupled to a processing element 308). For example, the user interface may be a user application, browser, user interface, and/or similar words used herein interchangeably executing on and/or accessible via the client computing entity 102 to interact with and/or cause display of information/data from the predictive data analysis computing entity 106, as described herein. The user input interface can comprise any of a number of devices or interfaces allowing the client computing entity 102 to receive data, such as a keypad 318 (hard or soft), a touch display, voice/speech or motion interfaces, or other input device. In embodiments including a keypad 318, the keypad 318 can include (or cause display of) the conventional numeric (0-9) and related keys (#, *), and other keys used for operating the client computing entity 102 and may include a full set of alphabetic keys or set of keys that may be activated to provide a full set of alphanumeric keys. In addition to providing input, the user input interface can be used, for example, to activate or deactivate certain functions, such as screen savers and/or sleep modes.
The client computing entity 102 can also include volatile storage or memory 322 and/or non-volatile storage or memory 324, which can be embedded and/or may be removable. For example, the non-volatile memory may be ROM, PROM, EPROM, EEPROM, flash memory, MMCs, SD memory cards, Memory Sticks, CBRAM, PRAM, FeRAM, NVRAM, MRAM, RRAM, SONOS, FJG RAM, Millipede memory, racetrack memory, and/or the like. The volatile memory may be RAM, DRAM, SRAM, FPM DRAM, EDO DRAM, SDRAM, DDR SDRAM, DDR2 SDRAM, DDR3 SDRAM, RDRAM, TTRAM, T-RAM, Z-RAM, RIMM, DIMM, SIMM, VRAM, cache memory, register memory, and/or the like. The volatile and non-volatile storage or memory can store databases, database instances, database management systems, data, applications, programs, program modules, scripts, source code, object code, byte code, compiled code, interpreted code, machine code, executable instructions, and/or the like to implement the functions of the client computing entity 102. As indicated, this may include a user application that is resident on the entity or accessible through a browser or other user interface for communicating with the predictive data analysis computing entity 106 and/or various other computing entities.
In another embodiment, the client computing entity 102 may include one or more components or functionality that are the same or similar to those of the predictive data analysis computing entity 106, as described in greater detail above. As will be recognized, these architectures and descriptions are provided for exemplary purposes only and are not limiting to the various embodiments.
In various embodiments, the client computing entity 102 may be embodied as an artificial intelligence (AI) computing entity, such as an Amazon Echo, Amazon Echo Dot, Amazon Show, Google Home, and/or the like. Accordingly, the client computing entity 102 may be configured to provide and/or receive information/data from a user via an input/output mechanism, such as a display, a camera, a speaker, a voice-activated input, and/or the like. In certain embodiments, an AI computing entity may comprise one or more predefined and executable program algorithms stored within an onboard memory storage module, and/or accessible over a network. In various embodiments, the AI computing entity may be configured to retrieve and/or execute one or more of the predefined program algorithms upon the occurrence of a predefined trigger event.
As described below, various embodiments of the present invention introduce techniques for using the output of a diagnosis prediction machine learning model to generate a recommendation score. By using the noted techniques, various embodiments of the present invention enable using pretrained diagnosis prediction machine learning models to generate a recommendation score, thus reducing or avoiding the need for computationally expensive operations that are performed to generate diagnosis prediction machine learning models. In this way, various embodiments of the present invention reduce the number of computational operations performed to generate recommendation scores, thus improving the computational efficiency of predictive data analysis systems that are configured to generate recommendation scores and make important technical contributions to the field of predictive data analysis.
The process 400 begins at step/operation 401 when the predictive data analysis computing entity 106 identifies the service request data object. In some embodiments, the service request data object describes diagnosis input data (e.g., symptom data, medical history data, and/or the like) associated with a service request (e.g., a request for medical services), a patient/service recipient identifier for the service request, a provider (e.g., medical provider) identifier for the service request, a medical facility identifier for the service request, one or more multimedia objects associated with the service request.
In some embodiments, step/operation 401 may be performed in accordance with the process that is depicted in
At step/operation 502, the predictive data analysis computing entity 106 identifies a provider identifier and/or a facility identifier for the service request. In some embodiments, the provider identifier and/or the facility identifier may be explicitly provided as part of the service request. In some embodiments, the provider identifier and/or the facility identifier may be inferred based at least in part on patient preference data maintained by the predictive data analysis computing entity 106.
At step/operation 503, the predictive data analysis computing entity 106 identifies one or more symptom fields associated with a patient identifier for the service request. A patient may use external third-party application provider applications or payer applications to enter symptoms (e.g., by using drop-down selection, free-form text, question-and-answer decision trees, and/or in other manners). In some embodiments, if the service request includes free-form text, the service request is processed in accordance with a natural language processing machine learning model in order to detect one or more diagnosis codes,/ one or more symptom codes, and/or one or more procedure codes in the free form text. In some embodiments, the one or more symptom fields are identified by a process that includes at least the following steps/operations: (i) for each symptom of a plurality of candidate symptoms, generating a symptom presence score that is generated by the natural language processing machine learning model via processing the free-form text using the natural language processing machine learning model and a symptom historical score that is generated by a relative occurrence frequency of the symptom within electronic health record (EHR) data of the patient identifier, (ii) for each symptom, generating a combined symptom score based at least in part on the symptom presence score for the symptom and historical symptom score for the symptom, and (iii) determining the one or more symptom fields for the patient identifier based at least in part on each symptom having a threshold-satisfying combined symptom score.
At step/operation 504, the predictive data analysis computing entity 106 optionally receives one or more multimedia data objects (e.g., videos, photographs, Digital Imaging and Communications in Medicine (DICOM) files and/or the like) for the service request. Examples of multi-media data objects include a magnetic resonance imaging (MRI) report, a computed tomography (CT) scan, a colonoscopy image, one or more DICOM files, or other professionally-derived images. The multimedia data objects may be uploaded by an end-user that is generating the service request.
At step/operation 505, the predictive data analysis computing entity 106 determines the service request data object. In some embodiments, the predictive data analysis computing entity 106 combines at least one of the provider identifiers, the facility identifier, the symptom fields, and the multimedia data objects in order to generate the service request data object. In some embodiments, the service request data object describes diagnostic input data associated with the service request data object, such as one or more symptom fields associated with the service request data object, one or more medical history fields associated with the service request data object, and/or the like.
Returning to
In some embodiments, a probabilistic diagnosis data object describes, for a particular service request data object that is associated with a particular diagnosis input data, a set of diagnosed conditions along with a diagnosis probability for each diagnosed condition. For example, a probabilistic diagnosis data object may describe that a service request data object is associated with a diagnosis D1 with the probability P1, a diagnosis D2 with the probability P2, and so on. In some embodiments, the probabilistic diagnosis data object is generated by: (i) obtaining a set of raw diagnosis probabilities for a set of conditions from a diagnosis prediction machine learning model, (ii) normalizing the set of raw diagnosis probabilities to generate the set of diagnosis probabilities for the set of conditions, and (iii) generating the probabilistic diagnosis data object based at least in part on the set of diagnosis probabilities.
In some embodiments, the probabilistic diagnosis data object for a service request data object is generated by a diagnosis prediction machine learning model based at least in part on diagnosis input data associated with the service request data object. The diagnosis prediction machine learning model may be configured to process diagnosis input data for a service request data object in order to generate a probabilistic diagnosis data object for the service request data object. In some embodiments, the diagnosis prediction machine learning model is a pretrained machine learning model whose output with respect to a service request data object is used to generate a dynamic-length classification for the service request data object. In other words, various embodiments of the present invention perform transfer learning on a diagnosis prediction machine learning model by mapping the outputs of the diagnosis prediction machine learning model to a variable-length subset of a plurality of candidate classes.
In some embodiments, the diagnosis prediction machine learning model may include n machine learning components, where each machine learning component is configured to generate a diagnosis probability for each condition of n conditions. In the noted example, the output of each machine learning component may have a different range. Thus, to generate the diagnosis probabilities described by the probabilistic diagnosis data object, the n diagnosis probabilities generated by the n machine learning components may be normalized to have a unified range (e.g., a unified range of [0, 1]), and then the normalized diagnosis probabilities may be used to generate the diagnosis probability data object. For example, the diagnosis probability data object may describe all n normalized diagnosis probabilities, the top m of the normalized diagnosis probabilities (where m may be a predefined value that is smaller than n), and/or those normalized diagnosis probabilities that satisfy (e.g., exceed) a normalized diagnosis probability threshold. In some embodiments, the diagnosis probabilities described by a probabilistic diagnosis data object are determined based at least in part on diagnosis input data associated with a corresponding service request data object, e.g., one or more service fields associated with the service request data object, one or more medical history fields associated with a patient identifier that is associated with the service request data object, and/or the like.
In some embodiments, a diagnosis prediction machine learning model (e.g., a pretrained diagnosis prediction machine learning model or an untrained diagnosis prediction machine learning model that is trained from scratch) is trained as part of an end-to-end machine learning framework that comprises the diagnosis prediction machine learning model, a hybrid diagnosis-provider classification machine learning model, and a recommendation scoring machine learning model, wherein the end-to-end machine learning framework may operate using at least the following steps/operations: (i) a service request data object is processed by the diagnosis prediction machine learning model to generate a probabilistic diagnosis data object, (ii) the probabilistic diagnosis data object is processed by the hybrid diagnosis-provider classification machine learning model to generate a variable-length classification for the service request data object, and (iii) the variable-length classification is processed by the recommendation scoring machine learning model to generate a consultation recommendation score for the service request data object.
In some embodiments, during training of the end-to-end machine learning framework, any trainable parameters of the diagnosis prediction machine learning model, the hybrid diagnosis-provider classification machine learning model, and the recommendation scoring machine learning model are trained based at least in part on inferred consultation recommendation scores for service request data objects and ground-truth consultation outcomes for the service request data objects as determined based at least in part on historical data associated with the service request data object. For example, if an inferred consultation recommendation score for a service request data object is 0.4 but the ground-truth consultation outcome for the service request data object describes an affirmative consultation outcome (e.g., describes that the patient was scheduled for a consult in response to the service request data object), then a deviation measure of 1−0.4=0.6 may be used as part of an error measure that is used to train the end-to-end machine learning framework by updating any trainable parameters of the diagnosis prediction machine learning model, the hybrid diagnosis-provider classification machine learning model, and the recommendation scoring machine learning model in a manner that is configured to optimize the error measure.
At step/operation 403, the predictive data analysis computing entity 106 determines a variable-length classification for the service request data object based at least in part on the probabilistic diagnosis data object for the service request data object and the provider data object for the service request data object. In some embodiments, step/operation 403 may be performed by a consultation agent (e.g., a cloud-based consultation agent).
In some embodiments, the hybrid diagnosis-provider classification machine learning model may be configured to: (i) map the probabilistic diagnosis data object for a service request data object to a variable-length subset of one or more diagnosis-based classes, (ii) map the provider data object for a service request data object to a variable-length subset of one or more provider-based classes, and/or (iii) map a combination of the probabilistic diagnosis data object for a service request data object and a provider data object for a service request data object to a variable-length subset of one or more hybrid classes. In some of the noted embodiments, given a service request data object S1 that is associated with a probabilistic data object PD1 and a provider data object P1, the hybrid diagnosis-provider classification machine learning model may determine: (i) of a set of available classes for the P1, e of the available classes that P1 actually corresponds to (where e is a variable number), (ii) of a set of available classes for the P1, f of the available classes that PD1 actually corresponds to (where f is a variable number), and (iii) of a set of available classes for the combination P1-PD1, g of the available classes that P1-PD1 actually corresponds to (where g is a variable number). In some embodiments, the hybrid diagnosis-provider classification machine learning model is trained using training data determined based at least in part on historical consultation arrangements (e.g., such that, if a proposed consultation has placed in the past, the training entry describing features of the proposed consultation is assigned a training label of one and otherwise the training entry is assigned a training label of zero.
In some embodiments, the probabilistic diagnosis data object comprises a sequence of diagnosis probabilities for a sequence of conditions, and the hybrid diagnosis-provider classification machine learning model comprises a recurrent neural network machine learning model (e.g., a long-short term memory neural network machine learning model, a gated recurrent unit machine learning model, and/or the like) that is configured to, at each timestep of a sequence of timesteps, process a corresponding diagnosis probability in the sequence of diagnosis probabilities in accordance with trained parameters of the recurrent neural network machine learning model to generate a hidden state for the timestep. In some embodiments, during each non-initial timestamp, generating the hidden state of the non-initial timestamp comprises processing a corresponding diagnosis probability in the sequence of diagnosis probabilities and a hidden state of an immediately preceding hidden state in accordance with trained parameters of the recurrent neural network machine learning model to generate a hidden state for the non-initial timestep. In some embodiments, during an initial timestamp, generating the hidden state of the initial timestamp comprises processing a corresponding diagnosis probability in the sequence of diagnosis probabilities and a null hidden state of an immediately preceding hidden state in accordance with trained parameters of the recurrent neural network machine learning model to generate a hidden state for the initial timestep. In some embodiments, the variable-length subset of a plurality of candidate classes is determined based at least in part on a hidden state of a final timestamp of the sequence of timestamps. In some embodiments, given a provider identifier that is associated with a provider specialty condition, the sequence of conditions is determined such that the provider specialty condition is the first condition in the sequence and each n+1th condition in the sequence is deemed more similar to the provider specialty condition than a preceding nth condition. Accordingly, in some embodiments, a set of n conditions may be ordered based at least in part on similarity to the provider specialty condition to generate the sequence of conditions.
In some embodiments, the hybrid diagnosis-provider classification machine learning model comprises n recurrent neural network machine learning models each associated with a diagnosis-based class. In some embodiments, the probabilistic diagnosis data object is used to determine, for each diagnosis-based class, a sequence of diagnosis probabilities for a sequence of conditions arranged in a sequence ordering associated with the diagnosis-based class. In some embodiments, for each diagnosis-based class, the sequence of diagnosis probabilities arranged in a sequence ordering associated with the diagnosis-based class is processed by via the recurrent neural network machine learning model via a sequence of timestamps that is associated with the candidate class to generate a class probability for the candidate class, where the class probability is determined based an output of a final timestep. In some embodiments, the top m candidate classes having the highest classes probabilities are selected as the variable-length subset of the diagnosis-based classes.
For example, given a particular recurrent neural network machine learning model (e.g., a long-short term memory neural network machine learning model, a gated recurrent unit machine learning model, and/or the like) that is associated with a particular diagnosis-based class, the diagnosis probabilities of the probabilistic diagnosis data object as arranged based at least in part on a sequence ordering associated with the diagnosis-based class may be processed by the recurrent neural network machine learning model to generate the class probabilities for the diagnosis-based class. In some embodiments, during each non-initial timestamp, generating the hidden state of the non-initial timestamp comprises processing a corresponding diagnosis probability in the sequence of diagnosis probabilities and a hidden state of an immediately preceding hidden state in accordance with trained parameters of the recurrent neural network machine learning model to generate a hidden state for the non-initial timestep. In some embodiments, during an initial timestamp, generating the hidden state of the initial timestamp comprises processing a corresponding diagnosis probability in the sequence of diagnosis probabilities and a null hidden state of an immediately preceding hidden state in accordance with trained parameters of the recurrent neural network machine learning model to generate a hidden state for the initial timestep. In some embodiments, the class probability for the diagnosis-based class is determined based at least in part on a hidden state of a final timestamp of the sequence of timestamps for a recurrent neural network machine learning model that is associated with the noted diagnosis-based class.
In some embodiments, a service request data object is mapped to a variable-length classification by using a hybrid diagnosis-provider classification machine learning model that is configured to process a probabilistic diagnosis data object for a service request data object and a provider data object for a service request data object to generate a variable-length classification for the service request data object. In some embodiments, the hybrid diagnosis-provider classification machine learning model may be configured to: (i) map the probabilistic diagnosis data object for a service request data object to a variable-length subset of one or more diagnosis-based classes, (ii) map the provider data object for a service request data object to a variable-length subset of one or more provider-based classes, and/or (iii) map a combination of the probabilistic diagnosis data object for a service request data object and a provider data object for a service request data object to a variable-length subset of one or more hybrid classes. In some embodiments, the hybrid diagnosis-provider classification machine learning model may be configured to determine, based at least in part on a provider identifier P1 and a probabilistic diagnosis PD1 for a service request data object S1, a variable-length classification for S1, where the variable-length classification maps S1 to a variable-length subset of a plurality of candidate classes, where the plurality of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes.
In some embodiments, a variable-length classification maps a service request data object to h candidate classes of a set of candidate classes, where h>=0. In some embodiments, the set of candidate classes comprise one or more diagnosis-based classes, one or more provider-based classes, and one or more hybrid classes, as further described below.
A diagnosis-based class may be an available label for a service request data object, where the service request data object may be mapped to the available label if a probabilistic diagnosis data object for the service request data object satisfies one or more conditions associated with the available label. Examples of diagnosis-based classes include a complexity class to which a service request data object is mapped if the probabilistic diagnosis data object for the service request data object describes a threshold-satisfying diagnosis probability for a diagnosed condition that is listed as being complex, unusual, and/or urgent; and an urgency class to which a service request data object is mapped if the probabilistic diagnosis data object for the service request data object describes a threshold-satisfying diagnosis probability for a diagnosed condition that is listed as being in need of time-sensitive-treatment. In some embodiments, when a probabilistic diagnosis data object for a service request data object includes n diagnosed probabilities for n conditions, then mapping the service request data object to a diagnosis-based class comprises: (i) determining a related subset of the n diagnosed conditions that are listed as being related to the diagnosis-based class (e.g., for a complexity class, determining a related subset of the n diagnosed conditions that are listed as being complex, unusual, and/or urgent), (ii) combining (e.g., summing, averaging, and/or the like) each diagnosis probability for a diagnosis condition in the related subset for the diagnosis-based class to determine a classification score for the diagnosis-based class, and (iii) mapping the service request data object to the diagnosis-based class if the classification score for the diagnosis-based class satisfies (e.g., exceeds) a classification score threshold.
A provider-based class may be an available label for a service request data object, where the service request data object may be mapped to the available label if a provider data object for the service request data object satisfies one or more conditions associated with the available label. Examples of provider-based classes include an availability class to which a service request data object is mapped if a provider schedule described by the provider data object for the service request data object includes available time slots for scheduling a service appointment associated with the service request data object, and a past interaction class to which a service request data object is mapped if a patient/service recipient identifier associated with the service request data object has had past interactions with the provider identifier associated with the provider data object for the service request data object as described by provider history data in the provider data object.
A hybrid diagnosis class may be an available label for a service request data object, where the service request data object may be mapped to the available label if a combination of the provider data objects associated with the service request data object and the probabilistic diagnosis data object associated with the service request data object satisfies one or more conditions associated with the available label. Examples of hybrid diagnosis classes include an individual past service class that is mapped to a service request data object if provider history data described by the provider data object for the service request data object describes past consultation by the provider identifier associated with the provider data object in relation to one or more diagnosed conditions described by the probabilistic diagnosis data object for the service request data object, a facility-based past service class that is mapped to a service request data object if provider facility history data described by the provider data object for the service request data object describes past consultation by a facility identifier associated with the provider data object in relation to one or more diagnosed conditions described by the probabilistic diagnosis data object for the service request data object, and a suboptimal past service class that is mapped to a service request data object if provider history data described by the provider data object for the service request data object describes suboptimal past service delivery by the provider identifier associated with the provider data object in relation to one or more diagnosed conditions described by the probabilistic diagnosis data object for the service request data object.
At step/operation 404, the predictive data analysis computing entity 106 determines the consultation recommendation score for the service request data object based at least in part on the variable-length classification for the service request data object. In some embodiments, the predictive data analysis computing entity 106 combines (e.g., using a recommendation scoring machine learning model) each recommendation score for a variable-length subset of candidate classifications that are associated with the variable length classification to generate the consultation recommendation score for the service request data object. For example, consider a service request data object that is associated with a variable-length classification that describes that the service request data object is associated with candidate classifications C1-C3, where the candidate classifications C1-C3 are associated with recommendation scores R1-R3 respectively. In some embodiments, the consultation recommendation score for the service request data object may be determined based at least in part on a combination of C1-C3 (e.g., based at least in part on
In some embodiments, the recommendation score for a candidate classification may describe a degree of correlation between mapping of the candidate classification to a service request data object and whether a provider consult should be scheduled with respect to the service request data object. For example, in some embodiments, the complexity class may be associated with a positive recommendation score, such that service request data objects having more complex conditions are more likely to need a provider consult.
As another example, in some embodiments, the urgency class may be associated with a positive recommendation score, such that service request data objects having more urgent conditions are more likely to need a provider consult.
As another example, in some embodiments, an availability class may be associated with a positive recommendation score, such that service request data objects that are associated with provider availability are more likely to lead to a provider consult.
As another example, in some embodiments, a past interaction class has a positive recommendation score, such that service request data objects that are associated with past interactions between the patient identifier and the provider identifier are more likely to lead to a provider consult.
As another example, in some embodiments, an individual past service class has a positive recommendation score, such that service request data objects that are associated with past interactions between the provider identifier and the diagnosed conditions of the service request data object are more likely to lead to a provider consult.
As another example, in some embodiments, a facility-based past service class has a positive recommendation score, such that service request data objects that are associated with past interactions between the facility identifier and the diagnosed conditions of the service request data object are more likely to lead to a provider consult.
As another example, in some embodiments, a suboptimal past service class has a positive recommendation score, such that service request data objects that are associated with substandard past interactions between the facility identifier and the diagnosed conditions of the service request data object are more likely to lead to a provider consult.
For example, consider a scenario in which a 14-year-old girl in Isabella, Minn., experiences right side abdominal pain, nausea and vomiting, among other symptoms. Her parents use an online agent to check symptoms. Appendicitis is a possible diagnosis. Worsening weather conditions make the normal 90-minute drive to Duluth ill-advised. Her parents work through a client-side application to confirm that they will instead head to a Silver Bay clinic. The complexity of this case, and fact that the clinic seldom treats this condition, may lead to a recommendation that a peer-to-peer video consultation with a Duluth surgeon is offered to the Silver Bay provider, in advance of the girl's arrival.
As another example, consider a scenario in which a 55-year-old man in Grand Forks, N. Dak. has experienced several days of fever and severe joint pain, along with headache, fatigue, and other symptoms. Through his provider application, he indicates to his doctor an intent to seek medical care. Because he has given permission for the provider application to access location-based services (LBS), the application reports to the clinician's proxy service or electronic medical record system that the patient spent two recent weeks in southeast Asia. Thus, both Dengue and Chikungunya are listed as possible diagnoses. Because of the criteria set in the provider services, the attending physician in Grand Forks is connected to a tropical disease expert in Rochester, Minn., with consultation scheduled shortly after the patient's arrival.
In this way, various embodiments of the present invention introduce techniques for using the output of a diagnosis prediction machine learning model to generate a recommendation score. By using the noted techniques, various embodiments of the present invention enable using pretrained diagnosis prediction machine learning models to generate a recommendation score, thus reducing or avoiding the need for computationally expensive operations that are performed to generate diagnosis prediction machine learning models. In this way, various embodiments of the present invention reduce the number of computational operations performed to generate recommendation scores, thus improving the computational efficiency of predictive data analysis systems that are configured to generate recommendation scores and make important technical contributions to the field of predictive data analysis.
At step/operation 405, the predictive data analysis computing entity 106 performs one or more prediction-based actions based at least in part on the consultation recommendation score. In some embodiments, performing the one or more prediction-based actions comprises, in response to determining that the consultation recommendation score satisfies a consultation recommendation score threshold, scheduling a consultation entry on a calendar data object associated with the provider data object via interacting with an integrated calendar application programming interface (API).
In some embodiments, step/operation 405 may be performed in accordance with the process that is depicted in
At step/operation 602, the predictive data analysis computing entity 106 determines, in response to determining that the provider is available for performing the consultation associated with the consultation recommendation score, whether the provider is interested in performing the consultation. In some embodiments, the predictive data analysis computing entity 106 presents a prompt to the provider about confirming whether the physician is interested to perform the consultation and determines the provider interest confirmation based at least in part on the provider selection of the input options provided by the prompt. An example of such a prompt is depicted in the exemplary user interface 700 of
At step/operation 603, the predictive data analysis computing entity 106 schedules, in response to determining that the provider is available for performing the consultation associated with the consultation recommendation score and that the provider is interested in performing the consultation associated with the consultation recommendation score, a provider consult with the provider in a location and time determined by the predictive data analysis computing entity 106. In some embodiments, scheduling the provider consult comprises scheduling a consultation entry on a calendar data object associated with the available provider data object via interacting with an integrated calendar application programming interface (API). In addition to electronic integration methods, a scheduling notification can be sent to the call center for manual communication to clinical systems. In some embodiments, pager systems (e.g., secure text messaging systems) can be integrated electronically of necessary information for consultation needs (e.g., patient demographics, timing/scheduling, and additional pertinent clinical data).
At step/operation 604, the predictive data analysis computing entity 106 determines, based at least in part on a provider availability indicator describing that the provider is unavailable for performing the consultation or a provider interest indicator describing that the provider is not interested in performing the consultation, one or more alternative provider prompts to one or more alternative provider identifiers associated with one or more alternative providers. In some embodiments, the provider availability indicator is determined based at least in part on a response to the prompt that is generated and displayed at step/operation 601. In some embodiments, to identify an alternative provider identifier, the predictive data analysis computing entity 106: (i) identifies a set of candidate alternative provider identifiers, and (ii) for each candidate alternative provider identifier: (a) generate a service request data object that includes the diagnosis input data of the original service data object but provider data extracted from the provider data object of the candidate alternative provider identifier, (b) process the generated service request data object using the diagnosis prediction machine learning model to generate a probabilistic diagnosis data object for the candidate alternative provider identifier, (c) process the probabilistic diagnosis data object for the candidate alternative provider identifier to generate a variable-length classification for the candidate alternative provider identifier, (d) determine a recommendation score for the candidate alternative provider identifier based at least in part on the variable-length classification for the candidate alternative provider identifier and using a recommendation scoring machine learning model, a classification recommendation score for the candidate alternative provider identifier, and (e) identify the candidate alternative provider identifier as an alternative provider identifier if the classification recommendation score for the candidate alternative provider identifier satisfies a classification recommendation score threshold.
At step/operation 605, the predictive data analysis computing entity 106 identifies an alternative available and interested provider identifier for a provider consult and schedules the provider consult by interacting with a calendar data object for the provider identifier. In some embodiments, scheduling the provider consult comprises scheduling a consultation entry on a calendar data object associated with the available provider data object via interacting with an integrated calendar application programming interface (API). In addition to electronic integration methods, a scheduling notification can be sent to the call center for manual communication to clinical systems. In some embodiments, pager systems (e.g., secure text messaging systems) can be integrated electronically of necessary information for consultation needs (e.g., patient demographics, timing/scheduling, and additional pertinent clinical data).
Accordingly, as described above, various embodiments of the present invention introduce techniques for using the output of a diagnosis prediction machine learning model to generate a recommendation score. By using the noted techniques, various embodiments of the present invention enable using pretrained diagnosis prediction machine learning models to generate a recommendation score, thus reducing or avoiding the need for computationally expensive operations that are performed to generate diagnosis prediction machine learning models. In this way, various embodiments of the present invention reduce the number of computational operations performed to generate recommendation scores, thus improving the computational efficiency of predictive data analysis systems that are configured to generate recommendation scores and make important technical contributions to the field of predictive data analysis.
Many modifications and other embodiments will come to mind to one skilled in the art to which this disclosure pertains having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. Therefore, it is to be understood that the disclosure is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims. Although specific terms are employed herein, they are used in a generic and descriptive sense only and not for purposes of limitation.
