US20230010216A1 - Diagnostic Effectiveness Tool - Google Patents
Diagnostic Effectiveness Tool Download PDFInfo
- Publication number
- US20230010216A1 US20230010216A1 US17/934,714 US202217934714A US2023010216A1 US 20230010216 A1 US20230010216 A1 US 20230010216A1 US 202217934714 A US202217934714 A US 202217934714A US 2023010216 A1 US2023010216 A1 US 2023010216A1
- Authority
- US
- United States
- Prior art keywords
- findings
- finding
- medical
- patient
- computer
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
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
- G16H50/20—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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/045—Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
-
- G06N7/005—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/22—Social work
-
- 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
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
-
- 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
-
- 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
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/20—ICT specially adapted for the handling or processing of medical references relating to practices or guidelines
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Definitions
- This disclosure is directed to systems and methods for improving the efficiency and effectiveness of diagnostic and treatment decision making.
- the diagnostic process begins when a patient arrives with a chief complaint. Before seeing the patient, the physician first searches through the patient's medical record for pertinent findings in the patient's history. From these findings, he creates a differential diagnosis, a broad list of diagnoses that is ranked based on the prevalence of each disease. The decision to test for a finding (ask a question, do an examination procedure, order a diagnostic test, etc.) is based on things such as:
- the physician interprets every newly acquired finding in the context of the patient's history, which changes the probability of each possible diagnosis. This process occurs after every clinical decision. Once there are enough findings to support a definitive diagnosis (high probability diagnosis with low probability alternatives), a diagnosis is made and treatment begins.
- the above process can be inefficient and not cost-effective, for a variety of reasons.
- a physician may not have time to search through the medical record for all relevant findings in the patient's history, particularly where the patient has an extensive or complex medical history.
- it may be difficult for the physician to objectively interpret the relevance of a finding for every diagnosis.
- It is difficult to rapidly make cost-effective, evidence-based decisions, especially where the patient has a complex medical history or information on costs of different potential tests are not readily available.
- physicians do not have time to document a high-quality note about the diagnostic process that occurred, especially in a busy clinical practice or hospital situation.
- Providers and payers often disagree about a patient's care because physicians are ill informed about the market, costs, and availability of resources, while payers lack the physician's experience and details about the patient's presentation.
- Communication is a slow and complex process, usually requiring a doctor to request authorization for a test from the patient's insurance company. This system creates an adversarial relationship between providers and payers, while excluding patients from the discussion entirely.
- Coded vocabularies are used for storing, retrieving, researching, and sharing patient data (i.e. HL7/FHIR, ICD, UMLS, SNOMED, and many others). Some databases have used these vocabularies to describe relationships between symptoms and diseases (Disease Database), which power online triage and symptom checkers. Symcat and Isabel Healthcare use probabilistic models to determine what question to ask next, but the critical components for clinical decision-making (physical exam findings, lab values, imaging reports, etc.) are excluded, making these tools not useful outside of patient education.
- CDS Clinical decision support
- Cancer screening reminders The physician's dashboard has a list of patients due for screening, based on age and gender.
- Alerts for possible drug interactions e.g., an alert appears when the physician orders a quinolone for a patient on warfarin.
- Clinical pathways A guideline for asthma management appears in the chart of a patient who has a history of asthma.
- CDS tools are usually implemented as point-fixes for specific metrics, like compliance rates or number of near-miss events. Therefore, they appear like regulatory tasks because they are not integrated with the physician's workflow/the diagnostic process. While CDS tools may improve metrics, their effectiveness is limited because they simply enforce rules based on static evidence.
- Medication reconciliation is a process for creating and maintaining an accurate list of medications by comparing what the patient reports to be taking with what the physician has ordered. Medication reconciliation is integrated in most electronic medical records (EMRs). The technology is limited to medications, it is not predictive, and it is not integrated with the diagnostic process.
- EMRs use macros to expand text for frequently typed phrases or paragraphs. Forms can also be inserted to free text notes or customized into templates.
- Canvas Medical has tools like word-autocompletion and keyword-to-ICD10 conversion. Documentation tools have been criticized for perpetuating “note bloat”. These tools are meant to improve the quality and efficiency of documentation, but bury the pertinent data instead. While efforts have been made to extract pertinent findings from free text, the results have been variable and are not in widespread use today.
- a system and method for evaluating, in advance, the medical effectiveness of diagnostic tests, new findings or interventions (i.e., treatments) from a set of known findings as to a patient and presenting the results to healthcare providers, healthcare payers (e.g., insurance companies), and/or patients.
- This disclosure is also directed to component aspects and combinations of the system.
- the system generates data for a computing device containing a software application which is used by a healthcare provider to assist in reviewing the patient's medical history and entering new findings as to the patient's condition or symptoms.
- the system includes a validated probabilistic model that is informed from aggregated electronic medical records and other sources of medical knowledge, such as medical journal articles, or expert user input.
- the system further includes a medical knowledge-based inference engine (processing unit and associated programming) that uses the patient's known findings (i.e., medical history and new findings, if any) and the model to determine a set of the most probable diseases.
- the engine also suggests a set of one or more tests or new findings that, it present, differentiate the set of most probable diseases, and generates indicia (e.g., text, scores, probability or statistical data, etc.) indicating diagnostic effectiveness of the one or more findings or tests, e.g., using Bayesian inference.
- This set of one or more tests or findings and associated indicia that are suggested by the engine can be transmitted to the computing device and presented to a user, e.g., healthcare provider, patient or payer.
- a user e.g., healthcare provider, patient or payer.
- results and indicia could be presented on the computing device for aiding the healthcare provider in selecting additional tests or findings to pursue.
- the results could be presented on a payer interface, e.g., a computer or workstation used by payer.
- the payer can make reimbursement (e.g., coverage and/or authorization) decisions based on the indicia.
- FIG. 1 is a block diagram of a system which includes features of the present disclosure.
- FIG. 2 is an illustration of a health-care provider facing interface of an electronic device (workstation, tablet computer, smartphone, etc.) showing a display of current findings, and suggested additional findings or tests for the provider to make to work towards a diagnosis.
- an electronic device workstation, tablet computer, smartphone, etc.
- FIG. 3 is an illustration of a Diagnostics Model Explorer which provides visualizations of suggestions of additional findings or related tests produced by the inference engine of FIG. 1 , indicia in the form of relevancy scores, and suggested diagnoses with probability scores.
- FIG. 4 is an illustration of the Diagnostic Model Explorer after the physician has selected an additional finding, showing a new ranking of suggested additional findings and a new ranking of suggested diagnoses.
- FIG. 5 is an illustration of a health-care provider facing interface of an electronic device showing a list of current known and pertinent findings for a patient, a list of most probable diseases and likelihood scores, and suggested new findings or tests with associated indicia.
- FIG. 6 shows the interface of FIG. 5 after the physician has selected one of the suggested new tests listed in FIG. 6 ; the display shows details of the suggested new tests and indicia in the form of delta probabilities and cost data.
- FIG. 7 shows an interface of a healthcare payer workstation showing a display of patient data, findings, and proposed tests including cost and delta probabilities to assist in making reimbursement decisions.
- a similar display could be provided on a web-based interface for a patient to assist them in making decisions.
- FIG. 8 shows the display on an electronic device showing known findings, a diagnosis, and proposed treatments for a patient along with indicia for each of the proposed treatments, including things such as survival data, cost data, side effects, etc. to assist a provider and patient in making treatment decisions; such information could also be displayed on workstation for a payer to assist in making reimbursement and coverage decisions.
- FIG. 9 is an illustration of an example of how the inference engine of FIG. 1 works.
- FIG. 10 is an illustration of an example of how the inference engine of FIG. 1 prioritizes findings that differentiates between two different diagnoses.
- FIG. 11 is an illustration of how the inference engine accounts for disease severity and prioritizes findings that rule in or out severe and urgent diseases over findings for non-urgent diseases.
- FIG. 12 an illustration of how the inference engine considers prevalence of diseases for ranking or scoring potential diagnoses; it seeks findings that differentiate more common diseases as compared to very rare diseases.
- FIG. 13 is an illustration of how the inference engine determines a likelihood of disease from multiple findings.
- FIG. 1 is directed to a system 10 for evaluating diagnostic effectiveness of one or more tests or additional findings from a set of known findings as to a patient.
- the system is part of or embedded in a larger clinical decision making system including device and software components aiding healthcare providers in arriving at a diagnosis and treatment options for a patient.
- the system generates data for a computing device 12 (which may take the form of a tablet, smartphone or desktop computer or workstation) containing a software application which is used by a healthcare provider to review the patient's medical history and enter findings as to the patient's condition or symptoms.
- a computing device 12 which may take the form of a tablet, smartphone or desktop computer or workstation
- This device 12 and the application will be described in more detail in FIGS. 2 - 8 and the discussion which follows.
- a system 14 which may be configured as a computer or complex of computers with ancillary memory, stores a validated probabilistic health model 16 and a database 18 of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge.
- the aggregated electronic medical records could come from an institution or multiple institutions, as well as electronic medical records fed back into the system by means of the flow of data indicated by the arrows 20 .
- data is patient de-identified such that compliance with all requirements for disclosure and use of a limited data set under HIPAA is performed.
- Ethics review and institutional review board exemption is obtained from each institution as necessary.
- Patient data is not linked to any Google user data.
- aggregated electronic health records are stored in a sandboxing infrastructure that keeps each EHR dataset separated from each other, in accordance with regulation, data license and/or data use agreements.
- the data in each sandbox is encrypted; all data access is controlled on an individual level, logged, and audited.
- the validated probabilistic health model and medical knowledge base includes a search and query application programming interface (API) 22 as will be explained below.
- API application programming interface
- the system 10 also includes a medical knowledge-based inference engine 24 (processor and associated programming), which operates on the patient's medical history and findings (either provided directly by the device or from a medical records database 26 ) and the validated probabilistic health model 16 .
- the engine is programmed to performs several tasks, including (1) determine a set of the most probable diseases of the patient, (2) suggest a set of one or more tests or additional findings that differentiate the set of most probable diseases, and (3) generate indicia, such as statistical data, text, or otherwise, indicating the effectiveness or relevancy of the one or more tests or additional findings. These aspects will be described in greater detail below. Data reflecting the results of these tasks are transmitted from the engine to the electronic device 12 to assist the healthcare provider in managing and planning the patient's care more efficiently.
- the electronic device 12 includes an interface 40 for presenting to the healthcare provider information as to at least one of (i) the set of the most probable diseases of the patient, (ii) a set of one or more tests or additional findings that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness or relevancy of the one or more tests or additional findings, or (iv) visualizations of the suggestions made by the inference engine.
- some or all of the above are displayed on the interface, e.g., in a series of screen displays with suitable controls or tools for toggling between the various features which are displayed. Examples will be described below.
- a computer workstation 30 residing on a payer network (not shown) which includes an application receiving the data from the engine as to the proposed additional finings or tests and the indicia, e.g. cost and relevancy scores.
- the application is configured to facilitate reimbursement or authorization decisions regarding the one or more tests. For example, an employee of the payer at the workstation views the proposed additional test(s) and associated indicia, and consults cost data either residing on the payer network, included in the indicia, or obtained from a database 32 , and makes reimbursement or coverage decisions on the proposed additional test(s).
- the system further supports a web-based patient interface 50 providing patients with diagnostic effectiveness information which they can access from their own electronic device, such as smartphone or PC.
- the interface provides a display of at least one of (i) the set of the most probable diseases of the patient and (ii) a set of one or more tests or additional findings that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness or relevancy of the one or more tests, or (iv) visualizations of the suggestions made by the inference engine.
- the inference engine uses the knowledge base 18 and the validated probabilistic health model 16 .
- the knowledge base 18 includes validated (i.e., proven or established) medical knowledge in the form of data indicating relationships between medical findings (i.e., facts as to a patient condition, such as test results, symptoms, etc.) and diagnoses.
- the medical knowledge may be stored in a data structure format which preserves these relationships as indicated in the box 100 .
- a finding, or a treatment has one or more attributes 104 .
- a finding attribute such as “abdominal pain” may have an attribute of “cost” of zero, since the finding can be made by interview of the patient and no expensive test is needed.
- a finding such as “loss of bone mineral density” may have an attribute of $2,000, as in order to make such a finding a DEXA (dual energy X-ray absorptiometry) scan is required which has, say, an average cost of $2,000.
- the medical knowledge includes an effectiveness relationship ( 106 ) between the finding 102 and a diagnosis 108 .
- a finding of abdominal pain has a sensitivity of 0.7 and a specificity 0.6 towards a diagnosis of “gastroenteritis.”
- the diagnosis 108 has attributes 110 , which may for example be severity, frequency (i.e., how common or rare the disease in a given population), or relationships to other diseases.
- the system of FIG. 1 is preferably configured to augment or increase the knowledge base 18 over time as more and more patients are treated by the system of FIG. 1 .
- a computer 120 functioning as an interpreter which looks at raw information and turns it into a format useful to the knowledge base.
- the interpreter generates useful data and updates which can be applied to the knowledge base 18 .
- the interpreter accesses document sources 112 , for example scientific papers, journals, clinical guidelines, Centers for Disease Control bulletins, etc. and formats the validated knowledge or information contained therein into the format consistent with the format 100 described above and supplies the additional medical knowledge to the knowledge base 18 .
- the interpreter 120 constructs relationships between medical findings and diagnoses, effectiveness scores for the relationships, and attributes of the findings and the diagnoses.
- the flow of data through the system indicated by the arrows 20 improves the knowledge base 18 and the performance of the engine 24 over time in recommending additional findings, additional tests, suggested diagnoses, and generation of useful and accurate indicia accompanying the proposed additional tests or findings.
- FIG. 2 is an illustration of the provider interface 40 of an electronic device 12 of FIG. 1 , for example a tablet computer.
- the interface 40 is designed to assists physicians through the diagnostic process and documents the process as it occurs, and provide a tool for evaluating medical effectiveness of proposed additional findings or tests.
- the interface 40 allows the physician to review a patient's medical history and enter new findings through assistance powered by a medical knowledge-base 18 and the inference engine 24 .
- the interface 40 has a region for display of three types of information that a physician can collect about a patient: 1) known findings, shown in region 200 , 2) proposed new findings in region 202 , including proposed new tests 204 , and 3) a differential diagnosis, not shown in FIG. 2 but available through appropriate menu selection and shown in FIGS. 3 , 4 , 5 and 6 .
- the system suggests relevant facts from the patient's history that the physician confirms or denies with one tap.
- the new findings section 202 the system suggests findings that are unknown and can be confirmed or denied by the physician (through questioning, examining, or placing an order and getting a result).
- the differential diagnosis section the system shows a differential diagnosis with the given known and new findings, ranked by probability. See FIGS. 3 , 4 , 5 , and 6 .
- the differential diagnosis is refreshed. Suggestions may change depending on the probabilities of diagnoses in the differential.
- the physician may continue to confirm/deny findings, or manually add new findings to arrive closer to a definitive diagnosis.
- a physician sees a patient with a chief complaint of “breast lump”. He would see relevant findings 202 that have been previously recorded, such as “family history of breast cancer”. The physician can confirm this finding and document it with one tap. This would increase the likelihood of breast cancer, ranking it higher in the differential diagnosis.
- the physician sees suggested facts to confirm or deny: “recurrent” and “immobile”. He asks the patient if her lump is recurrent, which she denies. He then performs a physical exam and finds an that the mass is immobile. Both findings are documented with a single tap, which increases the likelihood of breast cancer, putting it at the top in the differential diagnosis section.
- the system suggests “diagnostic mammogram” in the new findings section devoted to proposed new tests, region 204 in FIG. 2 . Tapping it places the order.
- FIG. 3 shows one example of how the interface 40 of the device 12 could display proposed diagnoses.
- the device includes a tool or icon on the display of FIG. 1 which, when activated displays to the healthcare provider what we have termed a “Diagnostics Model Explorer.”
- the Diagnostics Model Explorer visualizes the suggestions made by the inference engine, including proposed new findings and suggested diagnosis.
- the Diagnostics Model Explorer shows known findings in the leftmost column 302 , suggestions for new findings in the second column 304 , relevancy scores (i.e., “indicia”) for the proposed new findings in the third column 306 , and the differential diagnosis in the fourth column 308 .
- the fifth column shows a likelihood score (odds) that the suggested diagnosis in column 308 is correct.
- a tree 312 showing a sequence of related findings or symptoms along with probability data linking the sequences to a diagnosis (the lowermost level of the tree).
- the most cost effective finding is pharyngitis, listed at the top of the tree.
- the physician is not required to obtain this finding to progress. For example, the physician can select “dysuria” (first branch, right), and it becomes a “known finding.”
- the Diagnostic Model Explorer is updated. See FIG. 4 .
- “Dysuria” is now listed in the first column 302 as a known new finding.
- the differential diagnosis is refreshed, putting urinary tract infection at the top of the column 308 of suggested diagnoses.
- the decision tree is also recalculated, placing “lower abdominal pain” as the best next finding to obtain as indicated at the top of column 304 , since it has the highest relevancy score as indicated by the third column 306 .
- the diagnostic tree 312 is updated.
- FIG. 5 shows another arrangement of the display of the interface 40 of the electronic device in which the device includes in one screen a display of known findings 500 , a display of most probable diagnosis 502 , likelihood data for these diagnoses 504 , and a display of suggested new findings or tests in one column 506 and associated indicia 508 for each of proposed additional tests or findings, such as for example relevancy scores, costs, or other attributes for the findings.
- the provider can click on the “suggested new findings or tests” icon 510 and the display of FIG. 6 appears.
- the indicia include a “delta probability” and a cost.
- the delta probability reflects the degree of confidence in the accuracy of the diagnosis increases depending on the results of the particular test.
- the details of the design of the user interface of the device 12 is not particularly important and the examples of this disclosure are offered for purposes of illustration and not limitation.
- a patient may have a set of medical findings associated with a chief complaint, i.e., abdominal pain.
- the inference engine 24 determines a set of the most probable diseases (appendicitis and gall bladder disease) and infers a set of two tests that differentiate the set of most probable diseases: 1) an abdominal CT scan and 2) an abdominal ultrasound.
- Indicia are also generated by the engine: based on the prevalences of the two diseases and the known medical findings the patient's probability of having gallbladder disease is 60% and appendicitis is 40%. This information is displayed in regions 502 and 504 of FIG. 5 .
- a CT scan that shows an inflamed gall bladder would increase the probability of gallbladder disease being the correct diagnosis to 90%, and would decrease the probability of appendicitis to 10%, increasing the delta probability by 80%.
- An abdominal ultrasound that shows an inflamed gallbladder has the same delta probability.
- a cost comparison indicates that the CT scan is 10 times more expensive than the abdominal ultrasound, but both tests have the same delta probability, i.e., have the same diagnostic effectiveness in this particular situation. This information is displayed on the device 12 as shown in FIG. 6 . Accordingly, the healthcare provider may then pursue the abdominal ultrasound, since it has the same medical effectiveness as a CT scan but costs much less.
- FIG. 7 shows an example of the payer interface on the computer terminal 30 of FIG. 1 .
- the payer interface includes patient data, findings and proposed tests (identification code, cost and delta probability).
- the payer viewing the cost and delta probability in this situation may not authorize a physician's order of the CT scan and instead authorizes the abdominal ultrasound since they both have the same effectiveness but the abdominal ultrasound has a cost of only 1/10 that of the CT scan.
- FIG. 8 is an illustration of the web interface 50 for a patient.
- the patient logs into the medical records system for the hospital or provider they are using and navigate to a page having the interface 50 , which includes a display of the pertinent aspects of their medical records, including findings, a diagnosis and proposed additional tests or treatments.
- Each of the proposed additional tests or treatments is associated with indicia, such as cost, delta probability, survivability data, relevancy, side effects, risks, etc.,
- inference engine 24 makes suggestions of diagnosis and additional findings making use of a stored knowledge base 18 and validated probabilistic model 16 .
- the data in the knowledge base 18 can be curated manually (e.g., from aggregated electronic medical records, medical literature, textbooks, expert user input, etc.) or gathered from the flow of information through the system as describe above.
- the system includes an interpreter 120 to augment the knowledge base using information extraction from medical texts, records, or user input.
- the probabilistic model 14 uses two medical concepts: diagnoses and findings. These concepts have attributes, like severity, informer, and method, that are useful for ranking the concept's relevance. For example, a finding with a costly method can be ranked lower than one with a less costly method.
- the model API 22 FIG. 1 ) supports queries like: “find all concepts with an effectiveness relationship to concept A”, “find all attributes of concept B”, etc., taking advantage of the format of the data indicated by the box 100 in FIG. 1 , namely the linking of findings to attributes, linking of diagnoses to attributes, and providing effectiveness attributes (sensitivity, specificity) between findings, or sets of findings, and diagnoses.
- the diagnosis-finding relationship has two values, sensitivity (true positive rate) and specificity (true negative rate), modeled after diagnostic tests. From this, we can derive two synthetic relationships, one for a positive finding and one for a negative finding.
- Diagnosis finding:sensitivity, specificity
- Diagnosis neg. finding:LR ⁇ or (1 ⁇ sensitivity)/specificity
- Diagnoses are ranked based on their posttest probability/odds given a set of known findings.
- the likelihood ratios can be derived from concept relationships and are consequently used to compute the posttest odds.
- the engine 24 works essentially as follows.
- the engine queries the model 16 and applies an algorithm in order to draw inferences from the model 16 .
- the engine given a set of known findings for a patient, seeks to answer the question: “what tests should be performed next?” To answer the question, the engine could apply the following algorithm:
- FIGS. 9 - 13 will illustrate further examples of how the model 16 and inference engine 24 work to generate diagnoses and findings/tests.
- FIG. 9 is an example of how the decision-making engine 24 would work.
- the engine must infer the most cost-effective test between the top two diagnoses, appendicitis and diverticulitis.
- cost is weighed over effectiveness.
- the test for each finding is sorted by cost first, then by likelihood ratio.
- RLQ Light Lower abdominal Quarter
- RLQ can be determined by the interview, which is inexpensive as compared to a lab test required to determine leukocytosis. Therefore, the engine infers that RLQ pain is the best finding to obtain.
- FIG. 10 illustrates that the engine 24 will prioritize findings that differentiate between diseases.
- a CT scan and ultrasound US are equally costly.
- a CT scan that shows appendicolith strongly suggests appendicitis (LR 9.0) and strongly suggests against diverticulitis (LR ⁇ 8.0).
- An abscess on ultrasound has an LR of 3.0 for appendicitis and 3.5 for diverticulitis, which does not differentiate between the diseases. Therefore the engine would recommend a CT scan over US.
- FIG. 11 illustrates that the engine 24 accounts for disease severity and prioritizes findings that rule in or out severe and urgent diseases over findings for non-urgent diseases.
- myocardial infarction is more threatening compared to costochondritis, therefore, the engine would recommend obtaining an EKG over palpating the sternum for point tenderness.
- FIG. 12 illustrates that the engine 24 considers prevalence (from the validated probabilistic model 16 ) of the diseases in the differential. While it considers all diseases, it seeks findings that differentiate the more common diseases. For example, Creutzfeldt Jakob disease is extremely rare compared to generalized anxiety and essential tremor, and therefore would prioritize findings or tests that differentiate the more common diseases.
- FIG. 13 illustrates that as findings are collected, their individual LRs disease are multiplied for each disease. If RLQ pain is the only known finding, the likelihood ratio of appendicitis is the prevalence of appendicitis multiplied by 1.5. The additional finding of rebound tenderness will multiply 1.5 by 3.0. Similarly, an appendicolith will multiply 4.5 by 8.0. These three findings increase this patient's likelihood of having appendicitis 36 times greater than the general population.
- a computer processor 24 configured as a medical knowledge-based inference engine operating on the patient's medical history and findings and the validated probabilistic health model:
- the method can include the step of transmitting the indicia from the engine to an application residing on a healthcare payer network configured to facilitate reimbursement or authorization decisions regarding the one or more tests, as for example described above in the context of FIGS. 1 and 7 .
- the method may further include the step of transmitting the indicia from the engine 24 to the software application residing on the computing device 12 or 30 .
- the computing device 12 further includes an interface ( FIG. 2 ), and wherein the method further comprises the step of presenting on the interface information as to at least one of (i) the set of the most probable diseases of the patient and (ii) a set of one or more tests that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness of the one or more tests or (iv) visualizations of the suggestions made by the inference engine, see FIGS. 2 - 8 and the above description.
- the method may further include a step of transmitting to a web-based patient interface 50 providing patients with medical effectiveness information data for generating a display of at least one of (i) the set of the most probable diseases of the patient and (ii) a set of one or more tests or additional findings that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness or relevancy of the one or more tests, or (iv) visualizations of the suggestions made by the inference engine, such as shown in FIG. 8 .
- the format in the illustrated embodiment is in the form of relationships between medical findings and diagnoses, effectiveness scores for the relationships, and attributes of the findings and the diagnoses.
- the attributes of the findings could include a cost attribute.
- the attributes of the diagnoses could include a severity attribute, or a frequency attribute, or a treatment attribute or attributes.
- a method for assisting in making reimbursement or coverage decisions for medical care of a patient having a diagnosis and a medical history including findings.
- the method includes the steps of a) providing a validated probabilistic health model 14 including a database 18 of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge; b) providing a medical knowledge-based inference engine 24 operating on the patient's medical history and findings and the validated probabilistic health model to (1) determine a set of potential treatments for the patient and (2) generate indicia indicating effectiveness or relevancy of the potential treatments in the set; c) providing an interpreter 120 converting raw medical information from at least one of document sources, medical records, payment data or input from healthcare providers into a format useful to the database of medical knowledge and supplying the information in the format to the database; and d) generating data for a payer-facing interface (see FIG. 1 , 30 ; FIG. 7 ) for presenting the set of potential treatments for the patient and the indicia to assist a user of the payer-facing interface
Abstract
A system is disclosed for evaluating diagnostic effectiveness of one or more diagnostic tests or additional findings from a set of known findings as to a patient. The system includes a computing device containing a software application which is used by a healthcare provider to review the patient's medical history and enter findings as to the patient's condition or symptoms, a system storing a validated probabilistic health model including a database of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge; and a medical knowledge-based inference engine operating on the patient's medical history and findings and the validated probabilistic health model. The engine determines a set of the most probable diseases of the patient, suggests a set of one or more tests or additional findings that differentiate the set of most probable diseases, and generates indicia indicating diagnostic effectiveness or relevancy of the one or more tests or additional findings.
Description
- This application is a continuation of U.S. patent application Ser. No. 15/692,550, filed on Aug. 31, 2017, which is incorporated herein by reference in its entirety.
- This disclosure is directed to systems and methods for improving the efficiency and effectiveness of diagnostic and treatment decision making.
- The diagnostic process begins when a patient arrives with a chief complaint. Before seeing the patient, the physician first searches through the patient's medical record for pertinent findings in the patient's history. From these findings, he creates a differential diagnosis, a broad list of diagnoses that is ranked based on the prevalence of each disease. The decision to test for a finding (ask a question, do an examination procedure, order a diagnostic test, etc.) is based on things such as:
-
- the finding's effectiveness in differentiating between top ranked diagnoses (depends on the finding's positive and negative likelihood ratios (LR) for each disease);
- the finding's ability to rule out emergent diseases (e.g., while rare, myocardial infarction should be ruled out with an ECG in patients with chest pain); and
- the cost to obtain the finding (iatrogenic risk, resource availability, timeliness, etc.).
- The physician interprets every newly acquired finding in the context of the patient's history, which changes the probability of each possible diagnosis. This process occurs after every clinical decision. Once there are enough findings to support a definitive diagnosis (high probability diagnosis with low probability alternatives), a diagnosis is made and treatment begins.
- The above process can be inefficient and not cost-effective, for a variety of reasons. For one thing, a physician may not have time to search through the medical record for all relevant findings in the patient's history, particularly where the patient has an extensive or complex medical history. Furthermore, it may be difficult for the physician to objectively interpret the relevance of a finding for every diagnosis. It is difficult to rapidly make cost-effective, evidence-based decisions, especially where the patient has a complex medical history or information on costs of different potential tests are not readily available. Furthermore, often physicians do not have time to document a high-quality note about the diagnostic process that occurred, especially in a busy clinical practice or hospital situation.
- Physicians and insurance companies struggle to find a balance between high quality healthcare and economical use of resources when delivering care to patients. Providers and payers often disagree about a patient's care because physicians are ill informed about the market, costs, and availability of resources, while payers lack the physician's experience and details about the patient's presentation. Communication is a slow and complex process, usually requiring a doctor to request authorization for a test from the patient's insurance company. This system creates an adversarial relationship between providers and payers, while excluding patients from the discussion entirely.
- It is believed that no single technology adequately addresses all the above problems. Decision-making models (using Bayes' theorem) are used in clinical medicine, but in practice, they only serve as a mental framework for objectivity. Where there is objective data, the decision to order a test for a patient can be supported by test's sensitivity and specificity, as long as the patient falls in the population that was studied in the clinical trial. Health insurance companies have an adjudication process for authorizing reimbursement. They determine whether a study ordered by a physician is justified by its cost. Decisions are based on static rules or guidelines created by a committee.
- Coded vocabularies are used for storing, retrieving, researching, and sharing patient data (i.e. HL7/FHIR, ICD, UMLS, SNOMED, and many others). Some databases have used these vocabularies to describe relationships between symptoms and diseases (Disease Database), which power online triage and symptom checkers. Symcat and Isabel Healthcare use probabilistic models to determine what question to ask next, but the critical components for clinical decision-making (physical exam findings, lab values, imaging reports, etc.) are excluded, making these tools not useful outside of patient education.
- Clinical decision support (CDS) systems is a broad category of simple tools available to physicians, usually integrated with the EMR. Some of these tools use information from the patient's medical record. Some examples include:
- Cancer screening reminders: The physician's dashboard has a list of patients due for screening, based on age and gender.
- Alerts for possible drug interactions: e.g., an alert appears when the physician orders a quinolone for a patient on warfarin.
- Clinical pathways: A guideline for asthma management appears in the chart of a patient who has a history of asthma.
- CDS tools are usually implemented as point-fixes for specific metrics, like compliance rates or number of near-miss events. Therefore, they appear like regulatory tasks because they are not integrated with the physician's workflow/the diagnostic process. While CDS tools may improve metrics, their effectiveness is limited because they simply enforce rules based on static evidence.
- Medication reconciliation is a process for creating and maintaining an accurate list of medications by comparing what the patient reports to be taking with what the physician has ordered. Medication reconciliation is integrated in most electronic medical records (EMRs). The technology is limited to medications, it is not predictive, and it is not integrated with the diagnostic process.
- Assistance with documentation is available in the form of software, tools, and personnel. EMRs use macros to expand text for frequently typed phrases or paragraphs. Forms can also be inserted to free text notes or customized into templates. Canvas Medical has tools like word-autocompletion and keyword-to-ICD10 conversion. Documentation tools have been criticized for perpetuating “note bloat”. These tools are meant to improve the quality and efficiency of documentation, but bury the pertinent data instead. While efforts have been made to extract pertinent findings from free text, the results have been variable and are not in widespread use today.
- A system and method is disclosed for evaluating, in advance, the medical effectiveness of diagnostic tests, new findings or interventions (i.e., treatments) from a set of known findings as to a patient and presenting the results to healthcare providers, healthcare payers (e.g., insurance companies), and/or patients. This disclosure is also directed to component aspects and combinations of the system.
- In one embodiment the system generates data for a computing device containing a software application which is used by a healthcare provider to assist in reviewing the patient's medical history and entering new findings as to the patient's condition or symptoms.
- The system includes a validated probabilistic model that is informed from aggregated electronic medical records and other sources of medical knowledge, such as medical journal articles, or expert user input. The system further includes a medical knowledge-based inference engine (processing unit and associated programming) that uses the patient's known findings (i.e., medical history and new findings, if any) and the model to determine a set of the most probable diseases. The engine also suggests a set of one or more tests or new findings that, it present, differentiate the set of most probable diseases, and generates indicia (e.g., text, scores, probability or statistical data, etc.) indicating diagnostic effectiveness of the one or more findings or tests, e.g., using Bayesian inference.
- This set of one or more tests or findings and associated indicia that are suggested by the engine can be transmitted to the computing device and presented to a user, e.g., healthcare provider, patient or payer. These results and indicia could be presented on the computing device for aiding the healthcare provider in selecting additional tests or findings to pursue. Alternatively the results could be presented on a payer interface, e.g., a computer or workstation used by payer. The payer can make reimbursement (e.g., coverage and/or authorization) decisions based on the indicia.
-
FIG. 1 is a block diagram of a system which includes features of the present disclosure. -
FIG. 2 is an illustration of a health-care provider facing interface of an electronic device (workstation, tablet computer, smartphone, etc.) showing a display of current findings, and suggested additional findings or tests for the provider to make to work towards a diagnosis. -
FIG. 3 is an illustration of a Diagnostics Model Explorer which provides visualizations of suggestions of additional findings or related tests produced by the inference engine ofFIG. 1 , indicia in the form of relevancy scores, and suggested diagnoses with probability scores. -
FIG. 4 is an illustration of the Diagnostic Model Explorer after the physician has selected an additional finding, showing a new ranking of suggested additional findings and a new ranking of suggested diagnoses. -
FIG. 5 is an illustration of a health-care provider facing interface of an electronic device showing a list of current known and pertinent findings for a patient, a list of most probable diseases and likelihood scores, and suggested new findings or tests with associated indicia. -
FIG. 6 shows the interface ofFIG. 5 after the physician has selected one of the suggested new tests listed inFIG. 6 ; the display shows details of the suggested new tests and indicia in the form of delta probabilities and cost data. -
FIG. 7 shows an interface of a healthcare payer workstation showing a display of patient data, findings, and proposed tests including cost and delta probabilities to assist in making reimbursement decisions. A similar display could be provided on a web-based interface for a patient to assist them in making decisions. -
FIG. 8 shows the display on an electronic device showing known findings, a diagnosis, and proposed treatments for a patient along with indicia for each of the proposed treatments, including things such as survival data, cost data, side effects, etc. to assist a provider and patient in making treatment decisions; such information could also be displayed on workstation for a payer to assist in making reimbursement and coverage decisions. -
FIG. 9 is an illustration of an example of how the inference engine ofFIG. 1 works. -
FIG. 10 is an illustration of an example of how the inference engine ofFIG. 1 prioritizes findings that differentiates between two different diagnoses. -
FIG. 11 is an illustration of how the inference engine accounts for disease severity and prioritizes findings that rule in or out severe and urgent diseases over findings for non-urgent diseases. -
FIG. 12 an illustration of how the inference engine considers prevalence of diseases for ranking or scoring potential diagnoses; it seeks findings that differentiate more common diseases as compared to very rare diseases. -
FIG. 13 is an illustration of how the inference engine determines a likelihood of disease from multiple findings. -
FIG. 1 is directed to a system 10 for evaluating diagnostic effectiveness of one or more tests or additional findings from a set of known findings as to a patient. In the illustrated embodiment the system is part of or embedded in a larger clinical decision making system including device and software components aiding healthcare providers in arriving at a diagnosis and treatment options for a patient. - The system generates data for a computing device 12 (which may take the form of a tablet, smartphone or desktop computer or workstation) containing a software application which is used by a healthcare provider to review the patient's medical history and enter findings as to the patient's condition or symptoms. This
device 12 and the application will be described in more detail inFIGS. 2-8 and the discussion which follows. - A
system 14, which may be configured as a computer or complex of computers with ancillary memory, stores a validatedprobabilistic health model 16 and adatabase 18 of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge. The aggregated electronic medical records could come from an institution or multiple institutions, as well as electronic medical records fed back into the system by means of the flow of data indicated by thearrows 20. In a situation where the knowledge base is based on aggregated health records, data is patient de-identified such that compliance with all requirements for disclosure and use of a limited data set under HIPAA is performed. Ethics review and institutional review board exemption is obtained from each institution as necessary. Patient data is not linked to any Google user data. Furthermore, aggregated electronic health records are stored in a sandboxing infrastructure that keeps each EHR dataset separated from each other, in accordance with regulation, data license and/or data use agreements. The data in each sandbox is encrypted; all data access is controlled on an individual level, logged, and audited. - The validated probabilistic health model and medical knowledge base includes a search and query application programming interface (API) 22 as will be explained below.
- The system 10 also includes a medical knowledge-based inference engine 24 (processor and associated programming), which operates on the patient's medical history and findings (either provided directly by the device or from a medical records database 26) and the validated
probabilistic health model 16. The engine is programmed to performs several tasks, including (1) determine a set of the most probable diseases of the patient, (2) suggest a set of one or more tests or additional findings that differentiate the set of most probable diseases, and (3) generate indicia, such as statistical data, text, or otherwise, indicating the effectiveness or relevancy of the one or more tests or additional findings. These aspects will be described in greater detail below. Data reflecting the results of these tasks are transmitted from the engine to theelectronic device 12 to assist the healthcare provider in managing and planning the patient's care more efficiently. - For example, the
electronic device 12 includes aninterface 40 for presenting to the healthcare provider information as to at least one of (i) the set of the most probable diseases of the patient, (ii) a set of one or more tests or additional findings that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness or relevancy of the one or more tests or additional findings, or (iv) visualizations of the suggestions made by the inference engine. In one embodiment, some or all of the above are displayed on the interface, e.g., in a series of screen displays with suitable controls or tools for toggling between the various features which are displayed. Examples will be described below. - In one embodiment, there is a
computer workstation 30 residing on a payer network (not shown) which includes an application receiving the data from the engine as to the proposed additional finings or tests and the indicia, e.g. cost and relevancy scores. The application is configured to facilitate reimbursement or authorization decisions regarding the one or more tests. For example, an employee of the payer at the workstation views the proposed additional test(s) and associated indicia, and consults cost data either residing on the payer network, included in the indicia, or obtained from adatabase 32, and makes reimbursement or coverage decisions on the proposed additional test(s). - Additionally, the system further supports a web-based
patient interface 50 providing patients with diagnostic effectiveness information which they can access from their own electronic device, such as smartphone or PC. As with the case of theprovider device 12, the interface provides a display of at least one of (i) the set of the most probable diseases of the patient and (ii) a set of one or more tests or additional findings that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness or relevancy of the one or more tests, or (iv) visualizations of the suggestions made by the inference engine. These features allow the patient to make informed decisions about proposed tests which the provider may be recommending. - As noted above, in order for the inference engine to perform its tasks, including determine a set of the most probable diseases of the patient, suggest a set of one or more tests or additional findings that differentiate the set of most probable diseases, and generate indicia, such as statistical data, text, or otherwise, indicating the effectiveness or relevancy of the one or more tests or additional findings, it uses the
knowledge base 18 and the validatedprobabilistic health model 16. Theknowledge base 18 includes validated (i.e., proven or established) medical knowledge in the form of data indicating relationships between medical findings (i.e., facts as to a patient condition, such as test results, symptoms, etc.) and diagnoses. The medical knowledge may be stored in a data structure format which preserves these relationships as indicated in thebox 100. In particular, a finding, or a treatment, has one or more attributes 104. For example, a finding attribute such as “abdominal pain” may have an attribute of “cost” of zero, since the finding can be made by interview of the patient and no expensive test is needed. A finding such as “loss of bone mineral density” may have an attribute of $2,000, as in order to make such a finding a DEXA (dual energy X-ray absorptiometry) scan is required which has, say, an average cost of $2,000. - The medical knowledge includes an effectiveness relationship (106) between the finding 102 and a
diagnosis 108. For example, a finding of abdominal pain has a sensitivity of 0.7 and a specificity 0.6 towards a diagnosis of “gastroenteritis.” Thediagnosis 108 hasattributes 110, which may for example be severity, frequency (i.e., how common or rare the disease in a given population), or relationships to other diseases. For example the “gastroenteritis” diagnosis has a severity=2 on a scale of 1-10. - The system of
FIG. 1 is preferably configured to augment or increase theknowledge base 18 over time as more and more patients are treated by the system ofFIG. 1 . In this regard, there is acomputer 120 functioning as an interpreter which looks at raw information and turns it into a format useful to the knowledge base. The interpreter generates useful data and updates which can be applied to theknowledge base 18. For example, the interpreter accesses document sources 112, for example scientific papers, journals, clinical guidelines, Centers for Disease Control bulletins, etc. and formats the validated knowledge or information contained therein into the format consistent with theformat 100 described above and supplies the additional medical knowledge to theknowledge base 18. In particular, theinterpreter 120 constructs relationships between medical findings and diagnoses, effectiveness scores for the relationships, and attributes of the findings and the diagnoses. This can be done manually (by experts reviewing de-identified patient health records, medical journals, etc.) or using machine-learning models. Additionally, as indicated by theexpert user input 124, qualified trained medical personnel may be provided with simple tools to construct relationships between findings and diagnoses, and assign attributes to findings and diagnoses, and such data is formatted by theinterpreter 120 and supplied to theknowledge base 18. Additionally, application data, e.g.,medical records 26 andpayment data 32 may be provided to theinterpreter 120 and added to the knowledge base, e.g., as attributes of findings or diagnoses or new relationships and attributes of findings and diagnoses. In essence, the flow of data through the system indicated by thearrows 20 improves theknowledge base 18 and the performance of theengine 24 over time in recommending additional findings, additional tests, suggested diagnoses, and generation of useful and accurate indicia accompanying the proposed additional tests or findings. -
User Interface 40 of Electronic Device 10 (FIG. 2 ) - With the above description in mind, attention is directed to
FIG. 2 which is an illustration of theprovider interface 40 of anelectronic device 12 ofFIG. 1 , for example a tablet computer. Theinterface 40 is designed to assists physicians through the diagnostic process and documents the process as it occurs, and provide a tool for evaluating medical effectiveness of proposed additional findings or tests. Theinterface 40 allows the physician to review a patient's medical history and enter new findings through assistance powered by a medical knowledge-base 18 and theinference engine 24. - The
interface 40 has a region for display of three types of information that a physician can collect about a patient: 1) known findings, shown inregion 200, 2) proposed new findings inregion 202, including proposednew tests 204, and 3) a differential diagnosis, not shown inFIG. 2 but available through appropriate menu selection and shown inFIGS. 3, 4, 5 and 6 . In the knownfindings section 200, the system suggests relevant facts from the patient's history that the physician confirms or denies with one tap. In thenew findings section 202, the system suggests findings that are unknown and can be confirmed or denied by the physician (through questioning, examining, or placing an order and getting a result). In the differential diagnosis section, the system shows a differential diagnosis with the given known and new findings, ranked by probability. SeeFIGS. 3, 4, 5, and 6 . - After confirming or denying a finding in
region - For example, a physician sees a patient with a chief complaint of “breast lump”. He would see
relevant findings 202 that have been previously recorded, such as “family history of breast cancer”. The physician can confirm this finding and document it with one tap. This would increase the likelihood of breast cancer, ranking it higher in the differential diagnosis. In thenew findings section 202, the physician sees suggested facts to confirm or deny: “recurrent” and “immobile”. He asks the patient if her lump is recurrent, which she denies. He then performs a physical exam and finds an that the mass is immobile. Both findings are documented with a single tap, which increases the likelihood of breast cancer, putting it at the top in the differential diagnosis section. Next, the system suggests “diagnostic mammogram” in the new findings section devoted to proposed new tests,region 204 inFIG. 2 . Tapping it places the order. -
FIG. 3 shows one example of how theinterface 40 of thedevice 12 could display proposed diagnoses. In one embodiment, the device includes a tool or icon on the display ofFIG. 1 which, when activated displays to the healthcare provider what we have termed a “Diagnostics Model Explorer.” The Diagnostics Model Explorer visualizes the suggestions made by the inference engine, including proposed new findings and suggested diagnosis. The Diagnostics Model Explorer shows known findings in theleftmost column 302, suggestions for new findings in thesecond column 304, relevancy scores (i.e., “indicia”) for the proposed new findings in thethird column 306, and the differential diagnosis in thefourth column 308. The fifth column shows a likelihood score (odds) that the suggested diagnosis incolumn 308 is correct. - Below the textual and numerical data at the top of the display is a
tree 312 showing a sequence of related findings or symptoms along with probability data linking the sequences to a diagnosis (the lowermost level of the tree). In this scenario, the most cost effective finding is pharyngitis, listed at the top of the tree. However, the physician is not required to obtain this finding to progress. For example, the physician can select “dysuria” (first branch, right), and it becomes a “known finding.” - As indicated previously, when new findings are made, as in this case a new finding of “dysuria,” the Diagnostic Model Explorer is updated. See
FIG. 4 . “Dysuria” is now listed in thefirst column 302 as a known new finding. The differential diagnosis is refreshed, putting urinary tract infection at the top of thecolumn 308 of suggested diagnoses. The decision tree is also recalculated, placing “lower abdominal pain” as the best next finding to obtain as indicated at the top ofcolumn 304, since it has the highest relevancy score as indicated by thethird column 306. Likewise, thediagnostic tree 312 is updated. -
FIG. 5 shows another arrangement of the display of theinterface 40 of the electronic device in which the device includes in one screen a display of known findings 500, a display of mostprobable diagnosis 502, likelihood data for thesediagnoses 504, and a display of suggested new findings or tests in onecolumn 506 and associatedindicia 508 for each of proposed additional tests or findings, such as for example relevancy scores, costs, or other attributes for the findings. InFIG. 5 , the provider can click on the “suggested new findings or tests”icon 510 and the display ofFIG. 6 appears. In this case there are two new tests which are suggested—a CT scan and an ultrasound. For each new test, the indicia include a “delta probability” and a cost. The delta probability reflects the degree of confidence in the accuracy of the diagnosis increases depending on the results of the particular test. Of course, the details of the design of the user interface of thedevice 12 is not particularly important and the examples of this disclosure are offered for purposes of illustration and not limitation. - Consider the following example: a patient may have a set of medical findings associated with a chief complaint, i.e., abdominal pain. From the medical findings, the
inference engine 24, with the aid of theprobabilistic model 16 and knowledge base, determine a set of the most probable diseases (appendicitis and gall bladder disease) and infers a set of two tests that differentiate the set of most probable diseases: 1) an abdominal CT scan and 2) an abdominal ultrasound. Indicia are also generated by the engine: based on the prevalences of the two diseases and the known medical findings the patient's probability of having gallbladder disease is 60% and appendicitis is 40%. This information is displayed inregions FIG. 5 . A CT scan that shows an inflamed gall bladder would increase the probability of gallbladder disease being the correct diagnosis to 90%, and would decrease the probability of appendicitis to 10%, increasing the delta probability by 80%. An abdominal ultrasound that shows an inflamed gallbladder has the same delta probability. A cost comparison indicates that the CT scan is 10 times more expensive than the abdominal ultrasound, but both tests have the same delta probability, i.e., have the same diagnostic effectiveness in this particular situation. This information is displayed on thedevice 12 as shown inFIG. 6 . Accordingly, the healthcare provider may then pursue the abdominal ultrasound, since it has the same medical effectiveness as a CT scan but costs much less. -
FIG. 7 shows an example of the payer interface on thecomputer terminal 30 ofFIG. 1 . The payer interface includes patient data, findings and proposed tests (identification code, cost and delta probability). The payer viewing the cost and delta probability in this situation may not authorize a physician's order of the CT scan and instead authorizes the abdominal ultrasound since they both have the same effectiveness but the abdominal ultrasound has a cost of only 1/10 that of the CT scan. -
FIG. 8 is an illustration of theweb interface 50 for a patient. The patient logs into the medical records system for the hospital or provider they are using and navigate to a page having theinterface 50, which includes a display of the pertinent aspects of their medical records, including findings, a diagnosis and proposed additional tests or treatments. Each of the proposed additional tests or treatments is associated with indicia, such as cost, delta probability, survivability data, relevancy, side effects, risks, etc., -
Inference Engine 24 and ValidatedProbabilistic Model 16 - As indicated above,
inference engine 24 makes suggestions of diagnosis and additional findings making use of a storedknowledge base 18 and validatedprobabilistic model 16. The data in theknowledge base 18 can be curated manually (e.g., from aggregated electronic medical records, medical literature, textbooks, expert user input, etc.) or gathered from the flow of information through the system as describe above. The system includes aninterpreter 120 to augment the knowledge base using information extraction from medical texts, records, or user input. - The
probabilistic model 14 uses two medical concepts: diagnoses and findings. These concepts have attributes, like severity, informer, and method, that are useful for ranking the concept's relevance. For example, a finding with a costly method can be ranked lower than one with a less costly method. The model API 22 (FIG. 1 ) supports queries like: “find all concepts with an effectiveness relationship to concept A”, “find all attributes of concept B”, etc., taking advantage of the format of the data indicated by thebox 100 inFIG. 1 , namely the linking of findings to attributes, linking of diagnoses to attributes, and providing effectiveness attributes (sensitivity, specificity) between findings, or sets of findings, and diagnoses. - Relationships exist between these concepts, such as between diagnoses, between findings, as indicated in the
box 100 ofFIG. 1 . The diagnosis-finding relationship has two values, sensitivity (true positive rate) and specificity (true negative rate), modeled after diagnostic tests. From this, we can derive two synthetic relationships, one for a positive finding and one for a negative finding. -
Diagnosis—finding:sensitivity, specificity -
Diagnosis—pos. finding:LR+ or sensitivity/(1−specificity) -
Diagnosis—neg. finding:LR− or (1−sensitivity)/specificity - (where LR is Likelihood Ratio).
- With this knowledge representation, we can guide the user towards a differential diagnosis. Unlike traditional decision-tree-based instruction, our method does not require any specific input order, simply the known set of findings. Eventually the “known findings” input can be expanded to include anything and everything we might know about the patient e.g., f(patient).
-
f(known findings)=potential diagnosis - Diagnoses are ranked based on their posttest probability/odds given a set of known findings. The likelihood ratios can be derived from concept relationships and are consequently used to compute the posttest odds.
-
posttest odds=likelihood ratio*pr-test odds - This is derived from Bayes' theorem, which states:
-
p(D|F)=p(F|D)/p(F)*p(D) - Using Bayesian inference, we can update the diagnosis probability as we add findings; the posttest odds becomes the pretest odds of the next input. Thus, given a set of independent, equally prevalent findings F1 . . . Fx (i.e. p(F)=1), the posttest odds of a diagnosis D is:
-
p(D|F1 . . . Fx)=p(F1|D)p(F2|D) . . . p(Fx|D)p(D). -
g(known findings)=relevant findings - Referring again to
FIG. 1 , theengine 24 works essentially as follows. The engine queries themodel 16 and applies an algorithm in order to draw inferences from themodel 16. For example, the engine, given a set of known findings for a patient, seeks to answer the question: “what tests should be performed next?” To answer the question, the engine could apply the following algorithm: - 1. Query the model for disease concepts related to the known findings.
2. Score the disease concepts by computing the pre-test probability given the disease prevalence and known test results.
3. Query the model for related tests.
4. Score the tests based on factors like the post-test probability (effectiveness), cost, urgency, frequency/rarity and/or other factors. Data as to the results of the inference engine are then presented to theelectronic device 12 of the healthcare provider, theinterface 50 for the patient, and/or thepayer computer terminal 30. -
FIGS. 9-13 will illustrate further examples of how themodel 16 andinference engine 24 work to generate diagnoses and findings/tests. - In
FIG. 9 is an example of how the decision-makingengine 24 would work. The engine must infer the most cost-effective test between the top two diagnoses, appendicitis and diverticulitis. In this implementation, cost is weighed over effectiveness. The test for each finding is sorted by cost first, then by likelihood ratio. RLQ (Right Lower abdominal Quarter) can be determined by the interview, which is inexpensive as compared to a lab test required to determine leukocytosis. Therefore, the engine infers that RLQ pain is the best finding to obtain. -
FIG. 10 illustrates that theengine 24 will prioritize findings that differentiate between diseases. For example, assume here that a CT scan and ultrasound (US) are equally costly. A CT scan that shows appendicolith strongly suggests appendicitis (LR 9.0) and strongly suggests against diverticulitis (LR −8.0). An abscess on ultrasound has an LR of 3.0 for appendicitis and 3.5 for diverticulitis, which does not differentiate between the diseases. Therefore the engine would recommend a CT scan over US. -
FIG. 11 illustrates that theengine 24 accounts for disease severity and prioritizes findings that rule in or out severe and urgent diseases over findings for non-urgent diseases. In this example, myocardial infarction is more threatening compared to costochondritis, therefore, the engine would recommend obtaining an EKG over palpating the sternum for point tenderness. -
FIG. 12 illustrates that theengine 24 considers prevalence (from the validated probabilistic model 16) of the diseases in the differential. While it considers all diseases, it seeks findings that differentiate the more common diseases. For example, Creutzfeldt Jakob disease is extremely rare compared to generalized anxiety and essential tremor, and therefore would prioritize findings or tests that differentiate the more common diseases. -
FIG. 13 illustrates that as findings are collected, their individual LRs disease are multiplied for each disease. If RLQ pain is the only known finding, the likelihood ratio of appendicitis is the prevalence of appendicitis multiplied by 1.5. The additional finding of rebound tenderness will multiply 1.5 by 3.0. Similarly, an appendicolith will multiply 4.5 by 8.0. These three findings increase this patient's likelihood of having appendicitis 36 times greater than the general population. - While the illustrated embodiment illustrates a simple Bayesian approach for the probabilistic model, other types of models could be used which are known in the literature, for example the patent literature described in the Background section of this document.
- In view of the above, a method has been disclosed for evaluating medical effectiveness of one or more diagnostic tests from a set of known findings as to a patient, comprising:
- a) providing a computing device 12 (
FIG. 1 ) containing a software application which is used by a healthcare provider to review the patient's medical history and enter findings as to the patient's condition or symptoms (seeFIG. 2 ); - b) providing a validated
probabilistic health model 16 informed from aggregated electronic medical records or other sources of medical knowledge; and - c) using a
computer processor 24 configured as a medical knowledge-based inference engine operating on the patient's medical history and findings and the validated probabilistic health model: - (1) determining a set of the most probable diseases of the patient,
- (2) suggest a set of one or more tests or additional findings that differentiate the set of most probable diseases, and
- (3) generating indicia indicating effectiveness or relevancy of the set of one or more tests or additional findings. (See
FIGS. 2-8 ) - Furthermore, the method can include the step of transmitting the indicia from the engine to an application residing on a healthcare payer network configured to facilitate reimbursement or authorization decisions regarding the one or more tests, as for example described above in the context of
FIGS. 1 and 7 . - The method may further include the step of transmitting the indicia from the
engine 24 to the software application residing on thecomputing device - In one embodiment, the
computing device 12 further includes an interface (FIG. 2 ), and wherein the method further comprises the step of presenting on the interface information as to at least one of (i) the set of the most probable diseases of the patient and (ii) a set of one or more tests that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness of the one or more tests or (iv) visualizations of the suggestions made by the inference engine, seeFIGS. 2-8 and the above description. - As explained above the method may further include a step of transmitting to a web-based
patient interface 50 providing patients with medical effectiveness information data for generating a display of at least one of (i) the set of the most probable diseases of the patient and (ii) a set of one or more tests or additional findings that differentiate the set of most probable diseases, (iii) the indicia indicating effectiveness or relevancy of the one or more tests, or (iv) visualizations of the suggestions made by the inference engine, such as shown inFIG. 8 . - In another aspect, we have described a system in the form of a) a validated
probabilistic health model 14 including adatabase 18 of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge; and b) a medical knowledge-basedinference engine 24 operating a patient's medical history and findings and the validated probabilistic health model to (1) determine a set of the most probable diseases of the patient, (2) suggest a set of one or more tests or additional findings that differentiate the set of most probable diseases, and (3) generate indicia indicating effectiveness or relevancy of the one or more tests or additional findings; and c) aninterpreter 120 converting raw medical information from at least one of documents sources, medical records, payment data or expert input from healthcare providers into a format useful to the database of medical knowledge and supplying the information in the format to thedatabase 18. The format in the illustrated embodiment is in the form of relationships between medical findings and diagnoses, effectiveness scores for the relationships, and attributes of the findings and the diagnoses. As an example the attributes of the findings could include a cost attribute. The attributes of the diagnoses could include a severity attribute, or a frequency attribute, or a treatment attribute or attributes. - In another aspect, a method has been described for assisting in making reimbursement or coverage decisions for medical care of a patient having a diagnosis and a medical history including findings. The method includes the steps of a) providing a validated
probabilistic health model 14 including adatabase 18 of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge; b) providing a medical knowledge-basedinference engine 24 operating on the patient's medical history and findings and the validated probabilistic health model to (1) determine a set of potential treatments for the patient and (2) generate indicia indicating effectiveness or relevancy of the potential treatments in the set; c) providing aninterpreter 120 converting raw medical information from at least one of document sources, medical records, payment data or input from healthcare providers into a format useful to the database of medical knowledge and supplying the information in the format to the database; and d) generating data for a payer-facing interface (seeFIG. 1, 30 ;FIG. 7 ) for presenting the set of potential treatments for the patient and the indicia to assist a user of the payer-facing interface in making a reimbursement or coverage decision.
Claims (21)
1. A computer-implemented method, comprising:
providing, by a graphical user interface, an interactive relational tree, wherein the relational tree represents one or more sequences of diagnosis-finding relationships between findings and diagnoses of medical conditions with respective likelihood scores indicative of a relevance of a medical condition to a patient;
detecting, a selection of a node in the interactive relational tree, wherein the selection of the node indicates a selection by a healthcare provider of an additional finding indicative of one or more most probable medical conditions of the patient;
updating a set of known findings by adding the selected additional finding to the set of known findings; and
displaying, by the graphical user interface, an updated interactive relational tree based on the updated set of known findings.
2. The computer-implemented method of claim 1 , further comprising:
retrieving a stored validated probabilistic health model comprising a database of medical knowledge informed from aggregated electronic medical records or other sources of medical knowledge; and
wherein the set of known findings is determined based on the validated probabilistic health model.
3. The computer-implemented method of claim 1 , further comprising:
determining the set of known findings by applying a medical knowledge-based machine learning model operating on medical history and findings associated with the patient.
4. The computer-implemented method of claim 3 , further comprising:
applying the machine learning model to generate one or more additional diagnosis-finding relationships between the additional finding and associated additional diagnoses, and
wherein the updated interactive relational tree is based on the one or more additional diagnosis-finding relationships.
5. The computer-implemented method of claim 1 , further comprising:
providing, by the graphical user interface and to the healthcare provider, a set of one or more diagnostic tests that differentiate the one or more most medical conditions, and indicia indicating effectiveness or relevancy of the one or more diagnostic tests.
6. The computer-implemented method of claim 5 , further comprising an application residing on a healthcare payer network receiving the indicia from a machine learning model, the application configured to facilitate reimbursement or authorization decisions regarding the one or more diagnostic tests.
7. The computer-implemented method of claim 5 , wherein the indicia comprise at least one of delta probabilities, likelihood ratios, or costs associated with the set of one or more diagnostic tests.
8. The computer-implemented method of claim 5 , further comprising:
receiving, by the graphical user interface, a selection of a diagnostic test of the one or more diagnostic tests; and
updating, based on the selection, the indicia indicating diagnostic effectiveness or relevancy of the one or more diagnostic tests.
9. The computer-implemented method of claim 1 , further comprising:
updating, based on a confirmation or a denial of a finding from the set of known findings by the healthcare provider, one or more of the set of known findings, or the one or more sequences of diagnosis-finding relationships.
10. The computer-implemented method of claim 1 , further comprising:
storing raw medical information from at least one of document sources, medical records, payment data or input from healthcare providers in a medical database, the medical database having been configured to preserve one or more relationships between one or more of: a diagnosis, a diagnosis attribute, a finding, a finding attribute, a treatment, or a treatment attribute.
11. The computer-implemented method of claim 10 , wherein the storing comprises storing effectiveness scores for the relationships.
12. The computer-implemented method of claim 10 , wherein the finding attribute includes a cost attribute, and wherein the diagnosis attribute includes a severity attribute.
13. The computer-implemented method of claim 1 , wherein the one or more sequences of diagnosis-finding relationships is based on a positive likelihood ratio and a negative likelihood ratio, wherein each likelihood ratio is based on (i) a sensitivity indicative of a true positive rate of a diagnosis, and (ii) a specificity indicative of a true negative rate for the diagnosis, and the operations further comprising:
determining the positive likelihood ratio as a ratio of the sensitivity and a relative specificity, wherein the relative specificity is a difference of the specificity from 1; and
determining a negative likelihood ratio as a ratio of a relative sensitivity and the specificity, wherein the relative sensitivity is a difference of the sensitivity from 1.
14. The computer-implemented method of claim 13 , further comprising:
ranking one or more diagnostic tests based on respective costs and respective likelihood ratios.
15. The computer-implemented method of claim 1 , further comprising:
receiving, by the interactive relational tree, a denial of a second finding from the set of additional findings;
updating the set of additional findings by removing, based on the denial of the second finding, the second finding from the set of additional findings.
16. The computer-implemented method of claim 1 , further comprising:
providing, by a diagnostics model explorer and by the interactive graphical user interface, the set of known findings, wherein the diagnostics model explorer is a tabular representation of the set of known findings, the set of additional findings with associated relevance scores, and the associated diagnoses with associated likelihood ratios.
17. The computer-implemented method of claim 1 , further comprising:
providing, by the graphical user interface, a display comprising one or more suggested new diagnostic tests, with associated likelihood ratios;
receiving, from the healthcare provider, a selection of a new diagnostic test of the one or more suggested new diagnostic tests; and
providing, by the graphical user interface, a delta probability indicating a degree of confidence that a result of the selected new diagnostic test increases an accuracy of a diagnosis for the patient.
18. The computer-implemented method of claim 17 , further comprising:
providing, by the graphical user interface, a cost associated with the selected new diagnostic test.
19. The computer-implemented method of claim 17 , further comprising:
providing, by the graphical user interface, a new medical finding related to the patient;
receiving, from the healthcare provider, a confirmation or denial of the new medical finding; and
updating the one or more suggested new diagnostic tests based on the confirmation or the denial of the new medical finding.
20. A server to facilitate medical diagnosis for a patient, comprising:
one or more processors; and
memory storing computer-executable instructions that, when executed by the one or more processors, cause the server to perform operations comprising:
providing, by a graphical user interface, an interactive relational tree, wherein the relational tree represents one or more sequences of diagnosis-finding relationships between findings and diagnoses of medical conditions with respective likelihood scores indicative of a relevance of a medical condition to the patient;
detecting, a selection of a node in the interactive relational tree, wherein the selection of the node indicates a selection by a healthcare provider of an additional finding indicative of one or more most probable medical conditions of the patient;
updating a set of known findings by adding the selected additional finding to the set of known findings; and
displaying, by the graphical user interface, an updated interactive relational tree based on the updated set of known findings.
21. An article manufacture comprising one or more computer readable media having computer-readable instructions stored thereon that, when executed by one or more processors of a computing device, cause the computing device to perform operations comprising:
providing, by a graphical user interface of the computing device, an interactive relational tree, wherein the relational tree represents one or more sequences of diagnosis-finding relationships between findings and diagnoses of medical conditions with respective likelihood scores indicative of a relevance of a medical condition to a patient;
detecting, a selection of a node in the interactive relational tree, wherein the selection of the node indicates a selection by a healthcare provider of an additional finding indicative of one or more most probable medical conditions of the patient;
updating a set of known findings by adding the selected additional finding to the set of known findings; and
displaying, by the graphical user interface, an updated interactive relational tree based on the updated set of known findings.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US17/934,714 US20230010216A1 (en) | 2017-08-31 | 2022-09-23 | Diagnostic Effectiveness Tool |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US15/692,550 US11488712B2 (en) | 2017-08-31 | 2017-08-31 | Diagnostic effectiveness tool |
US17/934,714 US20230010216A1 (en) | 2017-08-31 | 2022-09-23 | Diagnostic Effectiveness Tool |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/692,550 Continuation US11488712B2 (en) | 2017-08-31 | 2017-08-31 | Diagnostic effectiveness tool |
Publications (1)
Publication Number | Publication Date |
---|---|
US20230010216A1 true US20230010216A1 (en) | 2023-01-12 |
Family
ID=65436198
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/692,550 Active 2041-04-18 US11488712B2 (en) | 2017-08-31 | 2017-08-31 | Diagnostic effectiveness tool |
US17/934,714 Pending US20230010216A1 (en) | 2017-08-31 | 2022-09-23 | Diagnostic Effectiveness Tool |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/692,550 Active 2041-04-18 US11488712B2 (en) | 2017-08-31 | 2017-08-31 | Diagnostic effectiveness tool |
Country Status (2)
Country | Link |
---|---|
US (2) | US11488712B2 (en) |
CN (2) | CN109427420B (en) |
Families Citing this family (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
IL263465A (en) * | 2018-12-04 | 2020-06-30 | Mae Wein Leila | Robotic emergency room having human collaborative modes |
JP2020126284A (en) * | 2019-02-01 | 2020-08-20 | コニカミノルタ株式会社 | Information processing apparatus, diagnosis support system, and program |
US20210241910A1 (en) * | 2020-01-30 | 2021-08-05 | Canon Medical Systems Corporation | Learning assistance apparatus and learning assistance method |
EP3910645A1 (en) * | 2020-05-13 | 2021-11-17 | Siemens Healthcare GmbH | Image retrieval |
CN112735542B (en) * | 2021-01-18 | 2023-08-22 | 北京大学 | Data processing method and system based on clinical test data |
CN113990518B (en) * | 2021-10-25 | 2023-05-16 | 北京百度网讯科技有限公司 | Information generation method, device, electronic equipment and storage medium |
WO2023181009A1 (en) * | 2022-03-24 | 2023-09-28 | Quai.MD, Ltd. | Method and system for generating a patient-specific clinical meta-pathway using machine learning |
Family Cites Families (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6678669B2 (en) | 1996-02-09 | 2004-01-13 | Adeza Biomedical Corporation | Method for selecting medical and biochemical diagnostic tests using neural network-related applications |
US6988088B1 (en) | 2000-10-17 | 2006-01-17 | Recare, Inc. | Systems and methods for adaptive medical decision support |
US6450956B1 (en) | 2000-11-06 | 2002-09-17 | Siemens Corporate Research, Inc. | System and method for treatment and outcome measurement analysis |
US7593913B2 (en) | 2006-01-11 | 2009-09-22 | Siemens Medical Solutions Usa, Inc. | Systems and method for integrative medical decision support |
JP5038671B2 (en) | 2006-09-25 | 2012-10-03 | 株式会社東芝 | Inspection item selection device, inspection item selection method, and inspection item selection program |
EP2225681A2 (en) | 2007-12-18 | 2010-09-08 | Koninklijke Philips Electronics N.V. | Integration of physiological models in medical decision support systems |
US20100174555A1 (en) * | 2009-01-05 | 2010-07-08 | Klaus Abraham-Fuchs | System for automatic clinical pathway optimization |
US8645165B2 (en) | 2010-06-03 | 2014-02-04 | General Electric Company | Systems and methods for value-based decision support |
US20120290310A1 (en) * | 2011-05-12 | 2012-11-15 | Onics Inc | Dynamic decision tree system for clinical information acquisition |
US20130268203A1 (en) * | 2012-04-09 | 2013-10-10 | Vincent Thekkethala Pyloth | System and method for disease diagnosis through iterative discovery of symptoms using matrix based correlation engine |
US10559377B2 (en) * | 2013-01-09 | 2020-02-11 | Biomed Concepts Inc. | Graphical user interface for identifying diagnostic and therapeutic options for medical conditions using electronic health records |
EP2775412A1 (en) | 2013-03-07 | 2014-09-10 | Medesso GmbH | Method of generating a medical suggestion as a support in medical decision making |
US20140358570A1 (en) * | 2013-06-04 | 2014-12-04 | Koninklijke Philips N.V. | Healthcare support system and method |
US10282512B2 (en) | 2013-07-09 | 2019-05-07 | Indiana University Research And Technology Corporation | Clinical decision-making artificial intelligence object oriented system and method |
JP2015207111A (en) * | 2014-04-18 | 2015-11-19 | ソニー株式会社 | Examination server, examination method, and examination system |
US20160004829A1 (en) * | 2014-07-03 | 2016-01-07 | Clinical Lab Consulting, Llp | Clinical laboratory decision support system |
ITMI20142149A1 (en) * | 2014-12-16 | 2016-06-16 | L I A Di Giuseppe Capasso | DEVICE FOR DIFFERENTIAL DIAGNOSIS IN MEDICINE ADAPTED TO DETERMINE THE OPTIMAL TEST SEQUENCE SUITABLE TO IDENTIFY A PATHOLOGY ACCORDING TO DIAGNOSTIC APPROPRIATENESS CRITERIA |
CN106778039A (en) | 2017-01-20 | 2017-05-31 | 深圳第医学健康管理有限公司 | Diagnosis and treatment accessory system and method |
-
2017
- 2017-08-31 US US15/692,550 patent/US11488712B2/en active Active
-
2018
- 2018-08-23 CN CN201810965569.XA patent/CN109427420B/en active Active
- 2018-08-23 CN CN202211258695.4A patent/CN115565670A/en active Pending
-
2022
- 2022-09-23 US US17/934,714 patent/US20230010216A1/en active Pending
Also Published As
Publication number | Publication date |
---|---|
CN109427420B (en) | 2022-10-14 |
CN115565670A (en) | 2023-01-03 |
CN109427420A (en) | 2019-03-05 |
US11488712B2 (en) | 2022-11-01 |
US20190065688A1 (en) | 2019-02-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US20230010216A1 (en) | Diagnostic Effectiveness Tool | |
US11163763B2 (en) | Decision-support application and system for medical differential-diagnosis and treatment using a question-answering system | |
US20230026624A1 (en) | Structured support of clinical healthcare professionals | |
US10937551B2 (en) | Medical concept sorting based on machine learning of attribute value differentiation | |
US8856188B2 (en) | Electronic linkage of associated data within the electronic medical record | |
US20180211726A1 (en) | Patient Communication Priority By Compliance Dates, Risk Scores, and Organizational Goals | |
Halamka | Early experiences with big data at an academic medical center | |
US20100145720A1 (en) | Method of extracting real-time structured data and performing data analysis and decision support in medical reporting | |
KR20200003407A (en) | Systems and methods for predicting and summarizing medical events from electronic health records | |
US20170132371A1 (en) | Automated Patient Chart Review System and Method | |
WO2020243732A1 (en) | Systems and methods of clinical trial evaluation | |
US20180082030A1 (en) | Automatic Adjustment of Treatment Recommendations Based on Economic Status of Patients | |
US20180211008A1 (en) | Assist Selection of Provider/Facility for Surgical Procedures Based on Frequency of Procedure, History of Complications, and Cost | |
WO2010062830A2 (en) | Systems and methods for clinical element extraction, holding, and transmission in a widget-based application | |
US11031109B2 (en) | Contextual EMR based dashboard graphical user interface elements | |
Mohammed et al. | Developing a semantic web model for medical differential diagnosis recommendation | |
KR20240008838A (en) | Systems and methods for artificial intelligence-assisted image analysis | |
Ahmadi et al. | Radiology reporting system data exchange with the electronic health record system: a case study in Iran | |
US11688510B2 (en) | Healthcare workflows that bridge healthcare venues | |
Vanderby et al. | Finding the truth in medical imaging: painting the picture of appropriateness for magnetic resonance imaging in Canada | |
WO2023019253A2 (en) | Methods and systems for longitudinal patient information presentation | |
Collen et al. | Medical informatics: past and future | |
Dewaswala et al. | Natural language processing for identification of hypertrophic cardiomyopathy patients from cardiac magnetic resonance reports | |
GM et al. | Healthcare Data Analytics Using Artificial Intelligence | |
US11748361B1 (en) | Systems and methods for multi-dimensional ranking of experts |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: GOOGLE INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:LEE, RYAN;CEDILNIK, ANDREJ;SO, CHRISTOPHER;AND OTHERS;SIGNING DATES FROM 20170913 TO 20171023;REEL/FRAME:061209/0282 Owner name: GOOGLE LLC, CALIFORNIA Free format text: CHANGE OF NAME;ASSIGNOR:GOOGLE INC.;REEL/FRAME:061537/0276 Effective date: 20170929 |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION |