Technical Field

The present invention relates to Clinical Trial Recommendations for a medical patient, made available within the Electronic Health Record (EHR), an electronic version of a patient’s medical history maintained by the healthcare provider. This makes it possible for the provider to recommend the right Clinical Trial(s) to the patient.

Background

According to Science Daily, over 95% of the world’s population suffers from health problems, with one third suffering from more than five ailments. Several thousand diseases affect humans, of which only 500 have proven, approved treatments, according to the NIH. Various groups including hospitals, independent biomedical scientists and researchers, and pharmaceutical companies are working to develop treatments for the remaining currently untreatable diseases.

Before a potential treatment reaches patients, it goes through several phases of clinical trials testing the treatment for both safety and effectiveness. This process is typically both long and costly. According to Phrma.org clinical trial phases take six to seven years on average. According to Biopharmadive, about 80% of clinical trials are delayed or closed because they cannot enroll patients. 9 out of 10 trials take twice as much time as originally anticipated. 66% of the research sites fail to meet enrollment goals, while 11% of them cannot enroll a single patient. According to Pharmafile, delays in trials can cost sponsors between $600,000 and $8 million for each day that a trial delays a drug’s development and launch. According to Clinical leader, clinical trials account for 40% of the US pharma research budget ~$7 billion per year. Enrolling patients in clinical trials who fit inclusion and exclusion criteria (a set of conditions based on which a patient is either included or excluded from a trial) is difficult because rare diseases affect very few people. Also, these patient populations are often widely dispersed and heterogeneous in disease subtype, symptoms, stages and exposure to prior treatment. Similarly, it is difficult for these patients to find the right clinical trials for their diseases: clinical trials are listed on many different medical websites, hospitals, government databases that are not easily accessible.

According to research conducted by Center for Information and Study on Clinical Research Participation (CISCRP), majority of nonprofit organizations, patient advocacy groups, and pharmaceutical companies are working to solve patient recruitment for clinical trials via social media and advertising. Unfortunately, only 5% of clinical trial participants have heard about clinical trials through online patient communities, and 16% have heard about it from advertisements. Furthermore, 73% of patients would like to hear about clinical trial opportunities majority of nonprofit organizations, patient advocacy groups, and pharmaceutical companies are working to solve patient recruitment for clinical trials via social media and advertising. Unfortunately, only 5% of clinical trial participants have heard about clinical trials through online patient communities, and 16% have heard about it from advertisements. Furthermore, 73% of patients would like to hear about clinical trial opportunities from their physicians. At the same time, while patients may prefer to hear about clinical trial opportunities from their physicians, this doesn't typically happen. Only 32% of patients reported that their physicians had ever shared information with them about clinical trials.

The reason for the presented novel approach in this patent -- Systems and Methods for Clinical Trial Recommendations is to automatically and regularly put the right clinical trials for each patient in the knowledge of physicians, so they can share the information with their patients.

Summary

Systems and methods that enable physicians to recommend Clinical Trials to patients. The Clinical Trials are collected, using APIs (Application Programming Interface) and Web Crawlers, from several Clinical Trial Sources including medical organizations, hospitals, and government websites. Patient Records, excluding PII (Personally Identifiable Information), are retrieved automatically from Electronic Health Records (EHRs) through secure integrations using authorized login credentials. Insights are extracted from both Clinical Trials and Patient Records via the Insights Engine based on natural language processing (NLP). The GetWellSoon Match Engine, powered by machine learning, compares the Patient Records with Clinical Trials to determine the right Clinical Trials for each patient and displays them directly within the EHR. Physicians can now recommend Clinical Trials to their patients without searching hundreds of Clinical Trial Sources and spending hours researching eligibility criteria and details for each Clinical Trial. Physicians and patients will get access to recommended Clinical Trials for free.

Brief description of drawings

The present application can be best understood by reference to the figures described below taken along with the accompanying drawing figures, in which like parts may be referred to by like numerals.

  • FIG. 1 illustrates the Clinical Trial Recommendation Platform.
  • FIG. 2 illustrates the method for Clinical Trial Aggregation.
  • FIG. 3 illustrates the Insights Engine for Clinical Trials.
  • FIG. 4 illustrates the process for EHR Integration.
  • FIG. 5 illustrates the Insights Engine for Patient Records.
  • FIG. 6 illustrates the Match Engine. FIG. 6A illustrates the Match Engines Details.
  • FIG. 7 illustrates the Trial Recommender.
  • FIG. 17 is a block diagram of a special-purpose computer system.

Detailed description of drawings

FIG. 1 illustrates the entire GetWellSoon Clinical Trial Recommendations platform.

FIG. 2 illustrates the method for Clinical Trial Aggregation. This process retrieves Clinical Trials and their corresponding fields from Clinical Trial Sources including medical organizations, hospitals, and government websites and stores them in the Clinical Trial Database.

Trial Fields: Each Clinical Trial Source, like cancer.gov, clinicaltrials.gov, etc. has thousands of clinical trials. Each Clinical Trial Source defines and maintains different Clinical Trial fields. The nomenclature for corresponding fields differs across Clinical Trial Sources. Clinical Trial fields are defined and normalized by GetWellSoon across all Clinical Trial Sources.

URL Aggregation: The URLs for all the Clinical Trials found on each Clinical Trial Source are retrieved and stored in the GetWellSoon Clinical Trial Database. A daily cron job updates the URLs in the Clinical Trial Database based on changes in Clinical Trial Sources.

Field Aggregation: The defined fields for all the Clinical Trials found on each Clinical Trial Source are retrieved and stored in the GetWellSoon Clinical Trial Database. A daily cron job updates the fields in the Clinical Trial Database based on changes in Clinical Trial Sources.

Trial Deduplication: This is a process to ensure duplicate clinical trials are not stored in the GetWellSoon database. Each Clinical Trial has a unique identifier provided by the government called the NCT ID (National Clinical Trial Identifier). This field is used by GetWellSoon for Clinical Trial deduplication. If multiple Clinical Trials with the same NCT ID exist in the Clinical Trial Database, the Clinical Trial from the Clinical Trial Source with the greatest monthly website traffic or usage is prioritized and kept, while the others are discarded to ensure the most relevant Clinical Trial Fields are stored.

Trial Status: Each Clinical Trial in the Clinical Trial Database contains various fields on trial enrollment and completion. Based on these fields, status is determined for each Clinical Trial. Only “Active” Clinical Trials are displayed by the Clinical Trial Recommender.

Trial Locations: Each Clinical Trial in the Clinical Trial Database may contain fields on one or more sites where the Clinical Trial is conducted. Based on these fields, locations can be determined for each Clinical Trial. Trial locations along with the patient’s location affect the order of Clinical Trials displayed by the Clinical Trial Recommender.

FIG. 3 illustrates the Insights Engine for Clinical Trials. This engine inputs the Clinical Trial Database, converts unstructured text from the stored Clinical Trials into structured Entity Fields, and updates the Clinical Trial Database accordingly.

Entity Database: A GetWellSoon Entity Database is created using APIs and Web Crawlers; it contains a catalog of Entities (drugs, procedures, measurements, observations, etc) relevant to each Clinical Trial. A comprehensive list of possible values for each Entity is retrieved from several Entity Sources including medical organizations and government websites and stored in the Entity Database.

Sentence Partitioning: Eligibility criteria consists of inclusion and exclusion criteria each of which contains multiple paragraphs and sentences. Rules and heuristic methods utilizing Stanford’s CoreNLP29 library are employed to parse and separate paragraphs and sentences.

Named Entity Recognition: Named Entity Recognition and other machine learning and natural language processes are performed based on the GetWellSoon Entity Database and using Stanford’s CoreNLP toolkit and libraries. The Unified Medical Language System synonym dictionary enables conversion between abbreviations and full forms (e.g., Human Immunodeficiency Syndrome to HIV and vice versa). The web service Usagi and Observational Health Data Sciences and Informatics (OHDSI) APIs map Entities found in sentences to standardized Observational Medical Outcomes Partnership (OMOP) vocabulary, allowing for Entity normalization. See following steps for further detail.

Negation Detection: Negation detection is achieved using NegEx such that negation status for each Entity identified can be ascertained. For example, if keywords like “no”, “never”, or “not” appear before or after an Entity in a snippet, then that Entity is declared to be negated.

Relation Detection: Relation detection handles the Entities with snippets that contain relational or temporal attributes (e.g “platelet count of <10000”, “within 1 month”, “age 13-15”).

Logic Detection: Logic detection is implemented using heuristic methods, enabling the identification of logical expressions like AND and OR that can be found in sentences or snippets.

FIG. 4 illustrates Electronic Health Record (EHR) Integration. This process accesses EHRs given the authorized credentials, extracts the Patient Records inside, excluding Personal Identifiable Information (PII), and then stores them in the GetWellSoon Patient Record Database.

Patient Record ID: A unique Patient Record ID, encoded for patient privacy under Health Insurance Portability and Accountability Act (HIPAA), is determined for each Patient Record based on the EHR identifier and record number. The Patient Record ID allows for writing the Clinical Trial Recommendations URL to the EHR.

Fig. 5 illustrates the Insights Engine for Patient Records. This engine inputs the Patient Record Database, converts unstructured text from specific fields of the stored Patient Records (physician notes, laboratory or test results, treatment plans, etc) into structured Entity Fields, and updates the Patient Record Database accordingly. See steps following steps for further detail.

FIG. 6 illustrates the Match Engine. This process takes as inputs: the GetWellSoon Patient Record Database and Clinical Trial Database. The Match Engine updates the Patient Record Database with a Clinical Trial Recommendations URL for each Patient Record.

Match Criteria: Fields used to match Clinical Trials with Patient Records are defined. Broadly, these are the structured fields plus Entity Fields of Clinical Trials and Patient Records.

Match Eligibility: For each Patient Record, a list of Eligible Clinical Trials is created. By default, a patient is eligible for all Clinical Trials. Eligibility fields for each Eligible Clinical Trial are parsed and compared with the Patient Record fields. Clinical Trials that do not “match” the Patient Record are eliminated. If a Clinical Trial’s age field restricts age between 20 and 45, and if the Patient Record says the given patient is 50 years old, then that Clinical Trial is eliminated from the Eligible Clinical Trials for that patient.

Match Score: For every Patient Record, a Match Score is calculated for each Eligible Clinical Trial. This Match Score is derived based on the number of Eligible Clinical Trial and Patient Record fields that “match”. For example, “matching” location fields are assigned a low weight because a patient with a severe disease may be willing to travel to participate in a Clinical Trial they are eligible for. See FIG. 6A for details on Match Score.

Clinical Trial Recommendations: For each Patient Record, a list of Clinical Trial Recommendations is created from the Eligible Clinical Trials, sorted by their Match Score.

Clinical Trials Recommendations URL: For each Patient Record, a GetWellSoon hosted web page is created, “getwellsoon.ai/recommendedtrials/X”, where X is the unique Patient Record ID, displaying Clinical Trial Recommendations for that patient.

FIG. 7 illustrates the Clinical Trial Recommender. This process inputs the GetWellSoon Patient Record Database and for each Patient Record, pushes the corresponding Clinical Trials Recommendations URL to the relevant Patient Record in the EHR. At this point, a physician can access any Patient Record in the EHR, click on the Clinical Trial Recommendations URL, and be redirected to a GetWellSoon page displaying relevant Clinical Trials for that particular patient.

FIG. 17 is a block diagram of a special-purpose computer system 1700. For example, the software platform in FIG 1, or any of the methods in FIG 2-7 may be implemented as a special-purpose computer system 1700.

Special-purpose computer system 1700 comprises a computer 1702, a monitor 104 coupled to computer 1702, one or more additional user output devices 1706 (optional) coupled to computer 1702, one or more user input devices 1708 (e.g., keyboard, mouse, track ball, touch screen) coupled to computer 1702, an optional communications interface 1710 coupled to computer 1702, and a computer-program product including a tangible computer-readable storage medium 1712 in or accessible to computer 1702. Instructions stored on computer-readable storage medium 1712 may direct system 1700 to perform the methods and processes described herein. Computer 1702 may include one or more processors 1714 that communicate with a number of peripheral devices via a bus subsystem 1716. These peripheral devices may include user output device(s) 1706, user input device(s) 1708, communications interface 1710, and a storage subsystem, such as random access memory (RAM) 1718 and non-volatile storage drive 1720 (e.g., disk drive, optical drive, solid state drive), which are forms of tangible computer-readable memory.

What is claimed

The following claims are unique to this patent.

Claim 1

A method for collecting information to store the information in a database, the method comprising:

  1. defining sources for collection of information
  2. collecting the information using different processes
  3. discarding duplicate information collected
  4. creating a database to store the collected information
  5. storing the collected information in the database

The method according to Claim 1, wherein the information is information on medical clinical trials.

The method according to Claim 1, wherein the process is a web crawler.

Claim 2

A method for generating insights from information, the method comprising:

  1. collecting information from a database
  2. parsing unstructured data in that information
  3. dividing that unstructured data into sentences
  4. defining attributes common to that information
  5. identifying those attributes in the sentences using different processes
  6. storing the attributes and corresponding values as insights in the database

The method according to Claim 2, wherein the information is information on Clinical Trials and Patient Records.

The method according to Claim 2, wherein the unstructured data in Clinical Trials include inclusion criteria, exclusion criteria, and detailed descriptions; wherein the unstructured data in Patient Records include physician notes and laboratory results.

The method according to Claim 2, wherein the attributes are Entities like conditions, drugs, procedures, measurements, observations, etc.

The method according to Claim 2, wherein the processes include machine learning and natural language processing utilizing such tools as the Stanford CoreNLP Library.

Claim 3

A method for filtering information based on a record, the method comprising:

  1. collecting the record and information from a database
  2. removing a subset of information that does not match the record using different fields
  3. sorting the remaining subset of information based on a score derived using from different processes
  4. storing that sorted subset of information in the original database of that record

The method according to Claim 3, wherein the information is information and insights on Clinical Trials; wherein the record is a single Patient Record.

The method according to Claim 3, wherein the different fields include conditions, minimum and maximum age, gender, trial status, trial availability, etc.

The method according to Claim 3, wherein the different processes include machine learning algorithms.

Claim 4

A method for displaying information to a person accessing a record, the method comprising:

  1. collecting the record and corresponding information from a database
  2. showing that information to that person each time the person accesses that record
  3. showing details of that information when that person clicks on the record using a process

The method according to Claim 4, wherein the information is information on Clinical Trials; wherein the record is a single Patient Record stored in the Electronic Health Record (EHR); wherein the person is a medical provider.

The method according to Claim 4, wherein the process is the process that takes the medical provider to a GetWellSoon hosted webpage that displays the recommended Clinical Trials for the Patient Record that the medical provider clicked.

Claim 5

Computer program comprising instructions which when executed on at least one processor of an apparatus causes the apparatus to perform, or execute, a method according to claim 1, 2, 3 and 4.

Abstract

Systems and methods that enable physicians to recommend Clinical Trials to patients. The Clinical Trials are collected, using APIs (Application Programming Interface) and Web Crawlers, from several Clinical Trial Sources including medical organizations, hospitals, and government websites. Patient Records, excluding PII (Personally Identifiable Information), are retrieved automatically from Electronic Health Records (EHRs) through secure integrations using authorized login credentials. Insights are extracted from both Clinical Trials and Patient Records via the Insights Engine based on natural language processing (NLP). The GetWellSoon Match Engine, powered by machine learning, compares the Patient Records with Clinical Trials to determine the right Clinical Trials for each patient and displays them directly within the EHR. Physicians can now recommend Clinical Trials to their patients without searching hundreds of Clinical Trial Sources and spending hours researching eligibility criteria and details for each Clinical Trial. Physicians and patients will get access to recommended Clinical Trials for free.

FIG 1. Clinical trial recommendation platformFIG 1. Clinical trial recommendation platform
FIG 2. Clinical trial aggregationFIG 2. Clinical trial aggregation
FIG 3. Insights engine for clinical trialsFIG 3. Insights engine for clinical trials
FIG 4. Electronic Health Record (EHR) integrationFIG 4. Electronic Health Record (EHR) integration
FIG 5. Insights engine for patient recordsFIG 5. Insights engine for patient records
FIG 6. Match engineFIG 6. Match engine
FIG 6a. Match engine detailsFIG 6a. Match engine details
FIG 7. Clinical trial recommenderFIG 7. Clinical trial recommender
FIG 17. Special purpose computer systemFIG 17. Special purpose computer system

IN THE UNITED STATES PATENT AND TRADEMARK OFFICE
Utility Patent Application (Provisional)
October 1st, 2021