Patent Application
20110258231
This application claims priority under 35 U.S.C. §119 from Chinese Patent Application No. 201010138982.2, filed Mar. 31, 2010, the entire contents of which are incorporated herein by reference.
The present invention relates to the collection and provision of pharmaceutical information, and more particularly, to a method and apparatus for providing the information concerning adverse drug effects.
An Adverse Drug Reaction (ADR) is a response to a drug which is noxious and unintended and which occurs at doses normally used for prophylaxis, diagnosis, or therapy of diseases, or for the modification of physiologic function. An Adverse Drug Event (ADE) is an adverse clinical event which occurs during the use of drugs, and usually the causal link between the drug use and the event is indeterminate. ADR and ADE can be resulted from a side effect (side reaction) or a toxic effect of a drug, or a drug-drug interaction.
For the sake of convenience, we will refer to the adverse symptoms such as ADR, ADE, etc. as Adverse Drug Effects (ADR/E) hereinafter. With the dramatic increase of the species of drugs, the Adverse Drug Effects are becoming more and more harmful to public health. Statistics show that ADE is one of the leading causes of death, ahead of lung disease, diabetes, AIDS, and automobile traffic accidents. ADE/ADR cause 1 out of 5 injuries or deaths per year to hospitalized patients. In China, reports of ADE/ADR cases reached at least 170,000 in 2005. In the United States, over 2 million serious ADEs occur yearly, causing 100,000 deaths.
This severe problem is a result of the inadequacy in the acquaintance and utilization of the information of Adverse Drug Effects. On the one hand, the information of Adverse Drug Effects are mainly described on drug labels, instructions, or research materials in pharmaceutical institutions, thus making it difficult to query comprehensively. Although some institutions are already involved in collecting drug information, the information thus provided has many problems in the use of systematic query. This is because, in the pharmaceutical industry, expression differences often exist. For example, one drug usually has different trade names and medical names, and one clinical symptom can have different descriptive languages; this descriptive inconsistency brings many difficulties to the provision and query of the ADE/ADR information. For example, the terms heart attack, myocardial infarction, and MI can refer to the same thing to a cardiologist, but, to a computer, they are all different. Therefore, currently, it is time-consumptive for a doctor to check the ADE/ADR information systematically and accurately.
Under these circumstances, the doctor has to prescribe for a patient based on his/her practical knowledge on drugs without checking the ADE/ADR information. Furthermore, the doctor has no idea of what other doctors prescribed for the patient, or the detailed physical quality of the patient, and therefore, the doctor has to recommend drugs according to the general symptoms without considering the individual condition of the patient.
In addition, the inconsistency in describing the ADE/ADR information by various information sources and various institutions makes it difficult to combine and process the information provided by different institutions. As such, drug-related institutions cannot obtain and utilize effectively the ADE/ADR information, and thus cannot apply this information to the drug-related research.
Therefore, a system is needed, which can automatically collect the information concerning the Adverse Drug Effects, and make it standardized and normalized, in order to facilitate the provision and update of the Adverse Drug Effects information by drug-related institutions and to expedite the query conducted by doctors and related personnel.
The present invention was made in view of the problems and disadvantages set forth above. The invention is proposed so as to provide a method and apparatus for providing the information of Adverse Drug Effects, which can provide comprehensively the normalized information of Adverse Drug Effects, thus overcoming the defects of the prior art.
Accordingly, one aspect of the present invention provides a method for providing information of adverse drug effects including: extracting at least a first information and a second information in basic information of a drug from a drug information source; matching the drug with a particular drug-related concept in a structured and normalized terminology system according to the first and the second information; extracting, from the drug information source, the information of Adverse Drug Effects associated with the drug; and matching the information of Adverse Drug Effects with a particular disorder-related concept in the structured and normalized terminology system; wherein the matching is along different paths in at least two disorder-related classified hierarchies.
Another aspect of the present invention provides an apparatus for providing information of Adverse Drug Effects of a drug including: a drug information extracting unit configured to extract at least a first information and a second information in basic information of a drug from a drug information source; a drug information matching unit configured to match the drug with a particular drug-related concept in a structured and normalized terminology system according to the first and the second information; an adverse effects information extracting unit, configured to extract from the drug information source the information of Adverse Drug Effects; and an adverse effects information matching unit, configured to match the information of Adverse Drug Effects with a particular disorder-related concept in the structured and normalized terminology system, wherein the matching is along different paths in at least two disorder-related classified hierarchies.
By using the method and apparatus of the invention, one can comprehensively extract the information concerning Adverse Drug Effects, and make it standardized and normalized, so as to facilitate the collection, integration, search, calculation, and propagation of the information, and thereby bring convenience to medicine-related organizations and individuals.
Next, detailed embodiments of the invention will be described in conjunction with detailed examples. It should be appreciated that the description of the following detailed embodiments are merely to explain the invention, rather than to impose any limitation on scope of the invention.
As described above, the present invention proposes a method and system which can automatically and comprehensively provide the information of Adverse Drug Effects in a standardized and normalized way. However, providing such a system faces challenges of several aspects. The first problem is with respect to the information sources of drugs. As drugs are present in a great variety and a large number, and change frequently, the information sources are supposed to be comprehensive, accurate, and updated. In addition, it is desired that the information sources are organized in a structured or semi-structured way so as to facilitate the extraction and analysis of the information. Another problem is with respect to term unification, which is very important in the standardization and normalization of the information of Adverse Drug Effects. To this end, it is necessary to refer to the standard terminology system that is commonly used in the industry, and it is also desired that the system is organized in a hierarchy form, in order to indicate the classification and subordination relationship between various terms.
As to the selection of the information sources of drugs, the most easily accessible and accurate information sources are drug labels. Drug labels include a comprehensive, concise and accurate description to the characteristics, efficacy and safety of drugs. Usually, drug labels have the following main content: chief description, clinical pharmacology, route of administration and dosage, contraindication, warning information, etc. In order to collect the information on drug labels, a structured pharmaceutical/product labeling (SPL) system has been developed to promote the summarizing and publishing of drug information. SPL was initially developed by HL7 (Health Level Seven) in the U.S., and then was adopted by the U.S. Food and Drug Administration (FDA) as a standard system for exchanging drug information. The FDA requires all drug companies which produce prescription drugs, OTC drugs, biological drugs or animal medicine to register and submit all drug labels in SPL standard format.
Particularly, according to SPL, the contents in drug labels are defined in XML format, and displayed in a web browser. A SPL file includes the contents in a drug label (all the texts, tables and pictures) as well as additional machine-readable information. Usually, SPL includes in its first level (level-1) structure the description relating to the basic information of a drug, such as the drug name, the active ingredient, the dosage form, the appearance, etc. Furthermore, as a structured file, SPL includes in its second level (level-2) a section relating to the Adverse Drug Effects, which section generally includes a start tag such as “adverse reaction” or “warning”. The SPL for some drugs also includes in its third level (level-3) more detailed information relating to the Adverse Drug Effects. Therefore, it can be seen that the SPL structured files, which are adopted by the FDA as authoritative and accurate drug information, are very suitable to be the information sources for extracting the information of Adverse Drug Effects. However, it can be understood that the drug information of other sources can be used as information sources, such as the summary reports on drug information made in other countries or by other institutions (for example, an institution of studying and analyzing drugs).
As to the selection of the standard terminology system, SNOMED CT (Systematized Nomenclature of Medicine—Clinical Terms) is a terminology system which is currently widely used. SNOMED CT is a systematically organized computer processable collection of medical terminology covering most areas of clinical information such as diseases, findings, procedures, microorganisms, pharmaceuticals etc. It allows a consistent way to index, store, retrieve, and aggregate clinical data across specialties and sites of care. It also helps organizing the content of medical records, reducing the variability in the way data is captured, encoded and used for clinical care of patients and research.
Particularly, SNOMED CT is a thesaurus of more than 365,000 clinical concepts, and each concept is defined by a unique numeric code, a unique name (Fully Specified Name) and a “description”. It contains more than 993,420 descriptions or synonyms for flexibility in expressing clinical concepts. These concepts are organized into 19 upper level hierarchies, including the hierarchy for medical procedure-related concepts, the hierarchy for drug-related concepts, the hierarchy for clinical disorder-related concepts, and the like. Each upper level hierarchy has several classified children hierarchies, for example, the drug-related concepts can be classified based on the drug name, the dosage form, etc, thus obtaining the further classified hierarchies; the clinical disorder-related concepts can be classified based on the body sites, the causes (induced by drugs), etc, thus obtaining the further classified hierarchies. The different concepts within a hierarchy or across hierarchies are linked by using about 1,460,000 “relationships.” Thus, SNOMED CT forms a compositional concept system on the basis of description logic. As SNOMED CT has characteristics set forth above, it is preferred to take it as the standard terminology system to standardize the description of drug ADE/ADR information. However, it can be understood that the terminology system is not limited to SNOMED CT, and any normalized and structured terminology system, which has been already developed or will be developed in future, can be used, such as MedDRA terminology system.
For the purpose of detailed description, the embodiments of the invention will be described in conjunction with exemplary SPL information sources and SNOMED CT terminology system.
The exemplary SPL code:
The above code contains the structured definitions and descriptions to pieces of information involved in a drug label, wherein the first half is a description to the basic information of the drug. The basic information substantially includes the main features of the drug, such as drug name, dosage form, ingredients, etc. Pieces of basic information of the drug that the code is directed to can be obtained by recognizing tags in the code. For example, by recognizing the code “<manufacturedMedicine> . . . <name>Fludarabine Phosphate </name>, it can be known that the manufactured medicine name of the drug is Fludarabine Phosphate; by recognizing the code “formCode code=“C42946””, which stands for “dosage form” in the SPL system, and recognizing that the tag value is Injection (“displayName=“INJECTION”), it can be known that the dosage form of the drug is injection. Similarly, at least the following information can be obtained from the above code:
Manufactured medicine name: Fludarabine Phosphate;
Generic drug name: Fludarabine Phosphate;
Dosage form: Injection;
Active ingredient substance: Fludarabine Phosphate;
Active moiety: Fludarabine.
The examples of the basic information are not limited to the information enumerated above. In other examples, the basic information can comprise different or additional pieces of information, such as drug property, chemical name, etc. From the basic information extracted, at least two pieces of the basic information can be selected for subsequent use in matching the drug into the standard terminology system, wherein the selected two pieces of the basic information cross with each other in two corresponding classified hierarchies relating to drugs. In one example, generic drug name is selected as the first information (Generic drug name: Fludarabine Phosphate), and dosage form is selected as the second information (Dosage form: Injection). By using the selected first information and second information, the drug can be matched with a particular drug-related concept in SNOMED CT. In particular, firstly, the first information is used to carry out preliminary matching, thus obtaining at least one candidate concept; then, the second information is used to carry out further matching for the at least one candidate concept, thus obtaining the matched particular concept.
Hence, step 202 is to judge whether the current concept obtained is a leaf node in the first classified hierarchy. If it is, the method jumps to step 207, in which the current concept is considered as the concept matched with the drug in the structured and normalized terminology system. If the current concept is not a leaf node, but has one or more child nodes, the method advances to step 203 to search for child nodes of the concept in the first classified hierarchy. In step 204, the child nodes thus found are in turn set as the current concept. For each child node set as the current concept, in step 205, the method searches for parent nodes of the child node (i.e. the current concept) in the second classified hierarchy, wherein the second classified hierarchy is a hierarchy in which the drug-related concepts are classified based on the second information. Then in step 206, the method judges whether the parent node described above matches with the second basic information of the drug; if it does not, it can be deemed that the corresponding child node is not the desired concept, and the method goes back to step 204 to set the next child node as the current concept and continue the judgment. If the result of judgment in step 206 is “matching”, the concept of this child node is considered as the selected concept. Then, the method goes back to step 202 to continue judging whether the selected concept is a leaf node; the method continues until the selected concept is a leaf node and at the same matches with the second information.
If the procedure described above fails to locate a particular concept based on the first information and the second information, the probable reason can be that the classified hierarchies corresponding to the selected first information and second information do not cross with each other. In this case, the first information and the second information can be reselected or changed, and the above procedure can be conducted once again and does not stop until a particular concept is located.
Now the above procedure will be described in combination with a given example. In step 201, the concept “Fludarabine” has been found as the result of fuzzy matching with the drug “Fludarabine Phosphate”, wherein the concept “Fludarabine” is explained and described in SNOMED CT as shown in
In particular,
At this time, the description to the current concept is shown in
In an alternative embodiment, after a child node is obtained as the current concept in step 205, a person skilled in the art can extract directly from the description to the current concept the description relating to the second information, and judge whether the description matches with the second information. For example, in the descriptions to the first child node “Fludarabine phosphate 10 mg tablet (product)” as shown in
The particular concepts exemplified above are only for exemplary purpose. In cases that the current concept is not a leaf node, the process can recursively carry out the steps of searching for child nodes and analyzing the child nodes by using the second information until the finally obtained concept is a leaf node and matches with the second information. The above implementation mode performs preliminary matching from top to bottom in the first classified hierarchy by using the first information as chief information, thus obtaining a generic concept; then screens and selects the successor nodes of the generic concept by using the second information, thus obtaining the matched specific concept. By combining the first information with the second information, accuracy can be guaranteed for matching drugs with concepts in SNOMED CT system. Additionally, the above process does not stop until the obtained concept is a leaf node, which ensures the accuracy and enough fineness when matching drugs with concepts.
Furthermore, although the above example selects the generic drug name as the first information, and the dosage form as the second information, the selection of the first information and the second information is not limited thereto. Other items in the basic information of drugs can be selected for use in matching drugs with concepts. For example, in one embodiment, the active ingredients and the dosage form of drugs can be selected as the first and second information, or the chemical names and properties of drugs can be selected as the first and second information. It should be understood that any two pieces of basic information of drugs, as long as they have corresponding hierarchies and explanations respectively in a structured terminology system and the two hierarchies cross with each other, can be selected for use in matching drugs with specific concepts in the terminology system. Additionally, the method can select more than two pieces of information, which can include the third information, the fourth information, and the like, in order to serve as a reference for further refining the concept or to verify the accuracy of the matched concept.
After obtaining the concept matched with the drug in the structured terminology system, the method advances to process the information of Adverse Drug Effects associated with the drug. First, it needs to extract from the drug information source the information of Adverse Drug Effects associated with the adverse effects of the drug, that is, to perform step 104 in
In order to perform the extraction mentioned above, in one embodiment, the content of a section is labeled with three tokens, including terms for adverse effects, related key words, and clinical conditions. The labeling of the contents of sections can be realized by defining a list of probable related key words (for example, including adverse action, adverse event, include, occur, report, and the like), and considering the grammar of the language. Many well-established algorithms are already present in the prior art for the labeling and extraction of such key information.
Subsequently, the method goes on to analyze terms for adverse effects which directly describe the symptoms of adverse effects, in order to match them precisely with the corresponding concepts in the structured and normalized terminology system. This matching process will be illustrated by taking the term for adverse effects “nausea” for example as shown in
If we simply search for nausea in the SNOMED CT system, we will find many fuzzy matched concepts, as shown in
According to one example of the invention, the combination of two paths is employed to find the most appropriate concept. In one embodiment, the first path is a path in the hierarchy in which disorder-related concepts are classified based on body sites, and the second path is a path in the hierarchy in which disorder-related concepts are classified based on drug-induced symptoms.
First, searching process along the first path will be described. In some particular examples, terms for adverse effects appear in subsections of SPL which correspond to particular body systems, for example, as shown in
In other particular examples, terms for adverse effects do not appear in subsections corresponding to particular body systems, for example, as shown in
During the process of searching along the first path as described above, it can also combine the searching from top to bottom with the searching from bottom to top in order to improve the efficiency of searching and enhance its performance.
After obtaining some candidate concepts via the first path, the process further locks on the final target concept via the second path. The second path is a path in which searching is conducted based on drug-induced symptoms in a disorder-related concepts hierarchy. In the hierarchy in which classification is based on drug-induced symptoms, the root node is “drug-related disorder”, and all drug-related disorders are the successor nodes of this root node.
As symptoms induced by adverse effects belong to drug-induced symptoms, therefore, terms for adverse effects should have corresponding concepts in the drug-induced symptoms hierarchy. Based on that, the common nodes shared by the first path and the second path are considered to be the concepts corresponding to terms for adverse effects. In order to find such common nodes, the process can analyze the candidate concepts obtained by searching along the first path, to determine whether the candidate concepts are present in the second path. In particular, in one example, beginning from the root node “drug-related disorder”, it traverses all paths along the drug-induced symptom hierarchy, to check whether the nodes involved in the paths belong to the candidate concepts.
Alternatively, in another example, for each candidate concept, it backtracks from bottom to top along the drug-induced symptom hierarchy, to check whether it can reach the root node “drug-related disorder”. Of the common concepts shared by the first path and the second path, the finest grained concept is considered to be the concept most appropriate for the term for adverse effects.
The matching process along two paths will be illustrated by taking the term for adverse effects “nausea” for example, as shown in
More particularly, the path taken to search for the particular concept based on body sites in the disorder-related concept hierarchy, i.e. the first path, is, from top to bottom, Disorder by body site (disorder)->Disorder of body system (disorder)->Disorder of digestive system (disorder)->Disorder of digestive tract (disorder)->Disorder of gastrointestinal tract (disorder)->Disorder of upper gastrointestinal tract (disorder)->Nausea and vomiting (disorder)->Drug-induced nausea and vomiting (disorder). The path taken to search for the particular concept based on drug-induced symptoms in the disorder-related concept hierarchy, i.e. the second path, is, from top to bottom, Drug-related disorder (disorder)->Drug-induced gastrointestinal disturbance (disorder)->Drug-induced nausea and vomiting (disorder). Thus, we match a single term for adverse effects with a particular concept in SNOMED CT.
For compound words and phrases in terms for adverse effects, if we fail to locate a matched concept in the SNOMED CT system, we can split the compound words or phrases and perform the matching process described above to the split terms separately. For example, for the phrase “pulmonary toxity” which is a term for adverse effects, we fail to locate a concept matched with it in SNOMED CT system. Therefore, we can split the phrase into two parts, i.e. pulmonary and toxity. For each part, we perform the above mentioned matching process separately. Finally, “Pulmonary” is matched with the concept “poisoning (disorder)”, and “toxity” is matched with the concept “disorder of lung (disorder)”. Thus, the phrase can be matched with a set of concepts—poisoning (disorder) and disorder of lung (disorder).
By the process described above, the terms or phrases for adverse effects in the ADR/ADE information can be matched with particular concepts in SNOMED CT system respectively, so that the key information in the ADR/ADE information can be normalized into the SNOMED CT system.
Considering the characteristics of SNOMED CT system, we select two paths for locating terms for adverse effects, i.e. the path classified by body sites and the path classified by drug-induced symptoms, and consider the common nodes shared by the two paths as the most appropriate nodes.
For the SNOMED CT system, such two paths are the most convenient for locating an appropriate disorder-related concept. However, for other structured and normalized terminology system, there can be different classification for various terms and concepts, and therefore there can be different paths that are suitable to locate terms for adverse effects. Generally speaking, it needs two or more paths to accurately locate the terms, and the finally matched concepts are the common nodes shared by the two or more paths.
In some cases, the information of adverse effects also includes the precondition information of the adverse effects, as shown by the rectangles in
By the process described above, the drug information and the information of adverse drug effects have been matched with particular concepts in the structured and normalized terminology system. Subsequently, we organize the obtained particular concepts, and establish the relationship between the concepts corresponding to the drug information and the concepts corresponding to the information of adverse drug effects, thereby obtaining complete information of adverse drug effects. Thus, all information relating to adverse drug effects extracted from the information source has been standardized and normalized. Since each concept in the structured and normalized terminology system has a unique code, such standardization and normalization convert the information of adverse drug effects extracted from various information sources, usually in text format, into definite concepts in code format. Such conversion is very advantageous to the collection, integration, search, calculation, propagation and further analysis of the information.
By standardizing and normalizing the information of adverse drug effects, doctors, patients, drug administrative institutes, and drug research and manufacture institutes can conveniently search, exchange and update the ADR/ADE-related information, therefore substantially avoiding unfortunate events associated with the adverse effects. In one example, the information of adverse effects provided by the above examples can be integrated into the existing Electronic Medicine Record (EMR) system. Since the EMR system has already employed similar, normalized terminology system to describe the medical history and drug-administration history of patients, and the information of adverse effects in the above examples is also provided in code format in the normalized terminology system, therefore, these two types of information can be easily integrated into each other. Thus, when a doctor prescribes, he/she can give suggestions that are more suitable for the individual conditions of a patient by referencing simultaneously the medical history and drug-administration history of the patient as well as the information of adverse drug effects. In another example, the information of adverse effects in standard code format is also very advantageous to help further treatment and analysis by computers.
For example, we suppose the following information of adverse effects is provided by the above examples: Drug A and Drug B have adverse interactions, and they have active ingredients A′ and B′ respectively. Given such information, the analyzing and treating system can infer that all the parent nodes of Drug A that comprises the ingredient A′ can probably have adverse reaction with Drug B. In addition, the information of adverse effects in code format is very advantageous to transmit across systems. The above mentioned advantages cannot be possessed by the information of adverse effects that is in general text format and is not standardized or normalized.
A method for providing the information of adverse drug effects according to the invention is described. Based on the same inventive concept, the present invention also relates to an apparatus for providing the information of adverse drug effects accordingly.
In an example, the drug information source 10 is SPL files, and the structured and normalized terminology system 20 is SNOMED CT system.
In the above mentioned apparatus for providing the information of Adverse Drug Effects, each unit is configured to perform a corresponding step of the method for providing the information of Adverse Drug Effects according to the present invention. Therefore, it is unnecessary to describe in detail the implementation and design of the apparatus.
Through the above description of the embodiments, those skilled in the art will recognize that the above-mentioned system and method for providing the information of Adverse Drug Effects can be practiced by executable instructions and/or controlling codes in the processors, e.g. codes in mediums like disc, CD or DVD-ROM; memories like ROM or EPROM; and carriers like optical or electronic signal carrier. The system, apparatus and its units in the embodiments can be realized using hardware like VLSI or Gates and Arrays, like semiconductors e.g. Logic Chip, transistors, etc., or like programmable hardware equipments e.g. FPGA, programmable logic equipments, etc.; or using software executed by different kinds of processors; or using the combination of said hardware and software
Although a method and apparatus of the present invention for providing the information of Adverse Drug Effects evaluating attention degree have been described in conjunction with detailed embodiments, the present invention is not limited thereto. Those skilled in the art can make various changes, substitutions and modifications in light of the teachings of the description without departing from the spirit and scope of the invention. It should be appreciated that, all such changes, substitutions and modifications still fall into protection scope of the invention which is defined by appended claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 201010138982.2 | Mar 2010 | CN | national |
