# US 2018/0322251 A1 · Identifying Drug-to-Drug Interactions in Medical Content and Applying Interactions to Treatment Recommendations > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2018/0322251 A1 - Document ID: US20180322251A1 - Status: Published application - Kind: A1 - Application number: 15/587,846 - Filed: 2017-05-05 - Published: 2018-11-08 - Earliest claimed: 2017-05-05 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Corville O. Allen; Timothy A. Bishop; Michael T. Payne; Sue S. Schmidt; Leah R. Smutzer - Web page: https://patentgrep.com/patent/US20180322251A1 ## CPC - G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems - G16H 15/00: ICT specially adapted for medical reports, e.g. generation or transmission thereof - G16H 70/40: relating to drugs, e.g. their side effects or intended usage - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA ## Abstract Mechanisms are provided for collecting and applying drug-to-drug interaction information from natural language documents. The mechanisms perform natural language processing of natural language documents in a corpus of natural language documents, to identify content in the natural language documents that describe a drug-to-drug interaction. The mechanisms categorize the drug-to-drug interaction and calculate, for a first drug in the drug-to-drug interaction, an exclusion weighting factor for use in evaluating treatments for patients that include a second drug of the drug-to-drug interaction based on the categorization of the drug-to-drug interaction. The mechanisms store the drug-to-drug interaction as a drug-to-drug interaction data structure that specifies the first drug, the second drug, the categorization of the drug-to-drug interaction, and the exclusion weighting factor associated with the drug-to-drug interaction. The mechanisms generate a treatment recommendation for treating a medical condition of a patient based on one or more drug-to-drug interaction data structures. ## Claims ### Claim 1 A method, in a data processing system comprising at least one processor and at least one memory, the at least one memory comprising instructions executed by the at least one processor to cause the at least one processor to perform the method, comprising: performing, by the data processing system, natural language processing of natural language documents in a corpus of natural language documents, to identify content in the natural language documents that describe a drug-to-drug interaction; categorizing, by the data processing system, the drug-to-drug interaction into a category of drug interaction based on at least one of severity type of the drug-to-drug interaction, medical concepts involved in the drug-to-drug interaction, or possible conditions that may occur due to the drug-to-drug interaction, as determined from features extracted from the identified content by the natural language processing; calculating, by the data processing system, for a first drug in the drug-to-drug interaction, an exclusion weighting factor for use in evaluating treatments for patients that include a second drug of the drug-to-drug interaction based on the categorization of the drug-to-drug interaction; storing, by the data processing system, the drug-to-drug interaction as a drug-to-drug interaction data structure in a drug-to-drug interaction data storage, wherein the drug-to-drug interaction data structure specifies the first drug, the second drug, the categorization of the drug-to-drug interaction, and the exclusion weighting factor associated with the drug-to-drug interaction; and generating, by the data processing system, a treatment recommendation for treating a medical condition of a patient based on one or more drug-to-drug interaction data structures stored in the drug-to-drug interaction data storage. ### Claim 2 (depends on 1) The method of claim 1, wherein generating the treatment recommendation comprises: receiving, by the data processing system, an indicator of a medical condition of a patient; evaluating, by the data processing system, a plurality of candidate treatments for treating the medical condition of the patient; and generating, by the data processing system, the treatment recommendation for treating the medical condition of the patient based on results of evaluating the candidate treatments, wherein candidate treatments evaluated that include at least one of the first drug or the second drug, are evaluated against patient information using the exclusion weighting factor specified in the drug-to-drug interaction data structure stored in the drug-to-drug interaction data storage. ### Claim 3 (depends on 1) The method of claim 1, wherein the corpus of natural language documents comprises at least one of drug label information comprising drug-to-drug interaction information, clinical statement documents comprising clinical statements from medical personnel describing drug-to-drug interactions, or patient statement documents comprising patient statements from patients describing drug-to-drug interactions. ### Claim 4 (depends on 2) The method of claim 2, wherein the patient information comprises patient electronic medical record information, and wherein evaluating the candidate treatments that include at least one of the first drug or the second drug against the patient information comprises, for each candidate treatment that includes at least one of the first drug or the second drug: determining whether the patient information indicates that the patient has been prescribed the other of the first drug or the second drug; and applying the exclusion weighting factor to the candidate treatment in response to a determination that the patient information indicates that the patient has been prescribed the other of the first drug or the second drug. ### Claim 5 (depends on 1) The method of claim 1, wherein performing natural language processing of natural language documents in a corpus of natural language documents, to identify content in the natural language documents that describe a drug-to-drug interaction further comprises identifying terms or phrases in the content that are specific to drug-to-drug interactions, and wherein categorizing, by the data processing system, the drug-to-drug interaction into a category of drug interaction comprises categorizing the terms or phrases in the content into one or more severity categories based on at least one of a statement type associated with a statement in which the terms or phrases are found, indications by the terms or phrases of possible conditions that may occur due to the drug-to-drug interactions, or medical concepts specified in the terms or phrases. ### Claim 6 (depends on 5) The method of claim 5, wherein the terms or phrases that are specific to drug-to-drug interactions are learned by the data processing system using automated machine learning. ### Claim 7 (depends on 2) The method of claim 2, wherein evaluating, by the data processing system, a plurality of candidate treatments for treating the medical condition of the patient comprises, for each candidate treatment: identifying at least one drug that is part of the candidate treatment; identifying, for the at least one drug, at least one drug-to-drug interaction data structure in the drug-to-drug interaction data storage associated with the at least one drug; identifying, for each drug-to-drug interaction data structure in the at least one drug-to-drug interaction data structure, a delta exclusion factor specified in the associated drug-to-drug interaction data structure; determining, for each drug-to-drug interaction data structure associated with the drug, whether a drug-to-drug interaction represented by the drug-to-drug interaction data structure applies to the patient based on the patient information and criteria specified in the drug-to-drug interaction data structure; and accumulating the delta exclusion factors for each drug-to-drug interaction, represented by each of the drug-to-drug interaction data structures, that apply to the patient to generate the exclusion weighting factor for the candidate treatment. ### Claim 8 (depends on 7) The method of claim 7, wherein determining whether the drug-to-drug interaction applies to the patient based on the patient information comprises modifying the delta exclusion factor associated with a drug-to-drug interaction associated with the drug, based on a matching of portions of the patient information to criteria of the drug-to-drug interaction. ### Claim 9 (depends on 8) The method of claim 8, wherein each criteria of the drug-to-drug interaction has an associated modification value for modifying the delta exclusion factor, and wherein the criteria of the drug-to-drug interaction comprises at least one of particular medical concepts involved in the drug-to-drug interaction, the possible medical conditions that may result from the drug-to-drug interaction, co-morbidity information associated with the drug-to-drug interaction, or cross referenced patient medical conditions. ### Claim 10 (depends on 7) The method of claim 7, wherein the delta exclusion factors specified in the drug-to-drug interaction data structures are learned through automated machine learning and natural language processing of the corpus of natural language documents. ### Claim 11 (depends on 2) The method of claim 2, wherein generating a treatment recommendation for treating the medical condition of the patient comprises: for the plurality of candidate treatments for treating the medical condition of the patient, calculating a score for each candidate treatment based on the exclusion weighting factors corresponding to drugs specified in the candidate treatments; ranking the candidate treatments in the plurality of candidate treatments based on their corresponding scores relative to one another, to generate a ranked listing; and selecting a candidate treatment from the ranked listing for output as the treatment recommendation. ### Claim 12 (depends on 1) The method of claim 1, further comprising outputting an output specifying the treatment recommendation to a client computing device associated with one of a medical professional treating the patient or a client computing device associated with the patient. ### Claim 13 (depends on 2) The method of claim 2, further comprising outputting an output identifying the treatment recommendation, wherein the output further specifies the plurality of candidate treatments, an exclusion factor associated with each candidate treatment that was used to calculate the score for the associated candidate treatment, and a reason as to why each exclusion weighting factor associated with each candidate treatment was applied to the associated score calculated for the candidate treatment. ### Claim 14 (depends on 1) The method of claim 1, wherein generating the treatment recommendation for treating the patient comprises processing a request to generate the treatment recommendation via a cognitive system pipeline, wherein the request specifies a patient and a medical condition of the patient, and wherein the cognitive system pipeline generates a plurality of candidate treatments, scores the plurality of candidate treatments, ranks the candidate treatments based on the scores associated with the candidate treatments, and selects at least one candidate treatment to be the treatment recommendation based on the ranking of the candidate treatments, and wherein the score of each candidate treatment in the plurality of candidate treatments is based on at least one exclusion weighting factor associated with drug-to-drug interactions associated with drugs that are part of the candidate treatment. ### Claim 15 A computer program product comprising a computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to: perform natural language processing of natural language documents in a corpus of natural language documents, to identify content in the natural language documents that describe a drug-to-drug interaction; categorize the drug-to-drug interaction into a category of drug interaction based on at least one of severity type of the drug-to-drug interaction, medical concepts involved in the drug-to-drug interaction, or possible conditions that may occur due to the drug-to-drug interaction, as determined from features extracted from the identified content by the natural language processing; calculate, for a first drug in the drug-to-drug interaction, an exclusion weighting factor for use in evaluating treatments for patients that include a second drug of the drug-to-drug interaction based on the categorization of the drug-to-drug interaction; store the drug-to-drug interaction as a drug-to-drug interaction data structure in a drug-to-drug interaction data storage, wherein the drug-to-drug interaction data structure specifies the first drug, the second drug, the categorization of the drug-to-drug interaction, and the exclusion weighting factor associated with the drug-to-drug interaction; and generate a treatment recommendation for treating a medical condition of a patient based on one or more drug-to-drug interaction data structures stored in the drug-to-drug interaction data storage. ### Claim 16 (depends on 15) The computer program product of claim 15, wherein the computer readable program further causes the computing device to generate the treatment recommendation at least by: receiving an indicator of a medical condition of a patient; evaluating a plurality of candidate treatments for treating the medical condition of the patient; and generating the treatment recommendation for treating the medical condition of the patient based on results of evaluating the candidate treatments, wherein candidate treatments evaluated that include at least one of the first drug or the second drug, are evaluated against patient information using the exclusion weighting factor specified in the drug-to-drug interaction data structure stored in the drug-to-drug interaction data storage. ### Claim 17 (depends on 16) The computer program product of claim 16, wherein the patient information comprises patient electronic medical record information, and wherein the computer readable program further causes the computing device to evaluate the candidate treatments that include at least one of the first drug or the second drug against the patient information at least by, for each candidate treatment that includes at least one of the first drug or the second drug: determining whether the patient information indicates that the patient has been prescribed the other of the first drug or the second drug; and applying the exclusion weighting factor to the candidate treatment in response to a determination that the patient information indicates that the patient has been prescribed the other of the first drug or the second drug. ### Claim 18 (depends on 16) The computer program product of claim 16, wherein the computer readable program further causes the computing device to evaluate the plurality of candidate treatments for treating the medical condition of the patient at least by, for each candidate treatment: identifying at least one drug that is part of the candidate treatment; identifying, for the at least one drug, at least one drug-to-drug interaction data structure in the drug-to-drug interaction data storage associated with the at least one drug; identifying, for each drug-to-drug interaction data structure in the at least one drug-to-drug interaction data structure, a delta exclusion factor specified in the associated drug-to-drug interaction data structure; determining, for each drug-to-drug interaction data structure associated with the drug, whether a drug-to-drug interaction represented by the drug-to-drug interaction data structure applies to the patient based on the patient information and criteria specified in the drug-to-drug interaction data structure; and accumulating the delta exclusion factors for each drug-to-drug interaction, represented by each of the drug-to-drug interaction data structures, that apply to the patient to generate the exclusion weighting factor for the candidate treatment. ### Claim 19 (depends on 17) The computer program product of claim 17, wherein the computer readable program further causes the computing device to determine whether the drug-to-drug interaction applies to the patient based on the patient information at least by modifying the delta exclusion factor associated with a drug-to-drug interaction associated with the drug, based on a matching of portions of the patient information to criteria of the drug-to-drug interaction. ### Claim 20 An apparatus comprising: at least one processor; and at least one memory coupled to the at least one processor, wherein the at least one memory comprises instructions which, when executed by the at least one processor, cause the at least one processor to: perform natural language processing of natural language documents in a corpus of natural language documents, to identify content in the natural language documents that describe a drug-to-drug interaction; categorize the drug-to-drug interaction into a category of drug interaction based on at least one of severity type of the drug-to-drug interaction, medical concepts involved in the drug-to-drug interaction, or possible conditions that may occur due to the drug-to-drug interaction, as determined from features extracted from the identified content by the natural language processing; calculate, for a first drug in the drug-to-drug interaction, an exclusion weighting factor for use in evaluating treatments for patients that include a second drug of the drug-to-drug interaction based on the categorization of the drug-to-drug interaction; store the drug-to-drug interaction as a drug-to-drug interaction data structure in a drug-to-drug interaction data storage, wherein the drug-to-drug interaction data structure specifies the first drug, the second drug, the categorization of the drug-to-drug interaction, and the exclusion weighting factor associated with the drug-to-drug interaction; and generate a treatment recommendation for treating a medical condition of a patient based on one or more drug-to-drug interaction data structures stored in the drug-to-drug interaction data storage.