Methods and systems for identifyin patients for particiption in a clinical study

Abstract

Computer implemented method for identifying patients suitable for participation in a clinical study. A study definition, comprising user selected patient criteria 310 for the study, is obtained. The p

Description

Field of the Invention The present application relates to methods and systems for identifying patients suitable for participation in a clinical study. Background to the invention In order to conduct a clinical trial it is typically necessary to identify and recruit a number of patients who satisfy certain criteria, such as having certain medical conditions, taking certain medications, or meeting certain demographics requirements. Potential patients may be recorded in a variety of different databases which may use different formats. This can make searching for suitable candidate patients for a medical trial difficult and can lead to delays and increased costs. Around 80% of trials fail to meet their initial enrolment target and timeline. An aim of the present invention is to provide improved methods and systems for searching patient records for suitable patients for recruitment into clinical studies. Summary of the invention According to a first aspect of the invention, there is provided a computer implemented method for identifying patients suitable for participation in a clinical study, the method comprising: obtaining a study definition comprising user-selected patient criteria for the clinical study, wherein the patient criteria comprise one or more selected clinical concepts each of which is associated with one of more subtypes thereof in a clinical concept database; deriving a database query from the study definition in which each of the selected clinical concepts is replaced with a set comprising that selected clinical concept and the one or more subtypes thereof with which it is associated in the clinical concept database; performing a search of a patient records database using the database query; and outputting information on patients identified in the patient records database which match the database query. While searching for potential participants for a clinical study, it is typically necessary for a researcher to search for patients whose records include certain clinical concepts relevant to the study. Clinical concepts are unique clinical ideas that can include clinical findings (such as diseases), pharmaceutical products, symptoms, and medical procedures as well as classes thereof. However, many clinical concepts are subtypes of other more general clinical concepts. For example, the clinical concepts "Wolfram syndrome" and "Insulin dependent diabetes mellitus type IA" are both subtypes of "Diabetes mellitus type 1". Different patient records and or databases are often inconsistent as to the precision with which clinical concepts are recorded. For example, in some circumstances a patient might be recorded as having been diagnosed with "Wolfram syndrome" and in others with "Diabetes mellitus type 1". This can make searching for patients associated with a given clinical concept difficult, as the searcher typically must search for all possible subtypes (including subtypes of subtypes, etc.) when searching patient records, in order to ensure all suitable possible participants are identified. Embodiments described herein can advantageously automatically transform an initial set of patient criteria including one or more user-selected clinical concepts into a database query including both the original user-selected clinical concepts and also all subtypes thereof, before performing a search using said query. This allows a user to only specify the exact range of clinical concepts they are looking for and have the search automatically identify all relevant patients, including those whose records only refer to specific subtypes of the concept in question. A user of the embodiments described herein may be a researcher who wishes to identify whether enough suitable patients are potentially available for participation in a potential clinical study or who wishes to have suitable patients identified for invitation to participate in the clinical study. Such a researcher may search a centralized patients records database, which may be anonymized, such that the researcher does not have access to private patient medical information. A user may also be direct care giver such as a general practitioner (GP) who may search records of their own patients to identify those who are suitable for a study that they have been informed of. Such a direct care giver may search a patients records database including only records of their own patients. Such direct care givers may be informed that they have patients suitable for a study by a researcher who has used a first instance of the invention to search a centralized anonymized patients records database as described above, and who has discovered that a number of patients of the direct care giver are suitable. The direct care giver may then use a second instance of the invention to search their own non-anonymized patient records and identify the suitable patients to potentially be invited for participation in the study. This approach may allow suitable patients to be invited for participation in the study without giving the researcher any access to nonanonymized patient records in advance. In some embodiments, the one or more selected clinical concepts, the one or more subtypes thereof, and the associations between the one or more clinical concepts and the subtypes thereof are comprised by a hierarchical or polyhierarchical dataset of clinical concepts, such as the polyhierarchical dataset SNOMED Clinical Terms. Hierarchical and polyhierarchical are different architectures for datasets. A hierarchical dataset is a dataset in which each entry (except one root entry) is a child of one other entry therein. A polyhierarchical dataset is one in which each entry (except the root entry) is a child of one or more other entries therein. Both hierarchical and polyhierarchical datasets define subtype-supertype relationships between their entries. In some such embodiments, the one or more selected clinical concepts, the one or more subtypes thereof, and the associations between the one or more clinical concepts and the subtypes thereof may be comprised by SNOMED Clinical Terms. Systematized Nomenclature of Medicine (SNOMED) Clinical Terms (CT) is the most comprehensive clinical healthcare terminology in the world and is widely used in electronic health records worldwide. SNOMED CT consists of a comprehensive polyhierarchical dataset of clinical terms, each of which is provided with a unique numerical identifier code, a fully specified name (FSN), a preferred term (PT), optionally one or more synonyms for the concept, and at least one relationship with another concept. SNOMED CT uses a variety of relationships between concepts, including "is-a" relationships which each define one clinical concept as a child of another parent subtype. Each clinical concept in the terminology (except the most general root concept "SNOMED CT Concept") has at least one parent. SNOMED CT defines a class hierarchy, in which the "descendants" of a concept (any child concepts, and any children of children etc.) are subtypes of that concept and the "ancestors" of a concept (its one or more parents, the parents of its parents etc.) are supertypes of that concept. Utilizing SNOMED CT clinical concepts provides the method with an exhaustive set of subtype-supertype concept relationships and allows the search to be performed on a large body of suitable patient records. The one or more selected clinical concepts and subtypes thereof may comprise clinical findings ("findings"), pharmaceutical / biologic products ("products"), procedures ("procedures"), or observations ("observable entities") as defined within the SNOMED CT terminology. In some embodiments, the one or more selected clinical concepts, the one or more subtypes thereof, and the associations between the one or more clinical concepts and the subtypes thereof may be comprise by a specific edition of the SNOMED CT terminology, such as the SNOMED CT UK edition that is maintained by the United Kingdom National Health Service. The one or more selected clinical concepts that are each associated with one of more subtypes may be all the clinical concepts comprised by the patient criteria, or the patient criteria may further comprise one or more clinical concepts that are not associated with any subtypes in the clinical database (for example, because said concepts are "leaf" concepts with no child concepts themselves). The method comprises obtaining a study definition comprising user-selected patient criteria for the clinical study. In some embodiments, the study definition and / or individual patient criteria thereof may be obtained by being input by a user, by being received from another system, and / or by being loaded from a memory. For example, an instance of the method may initially be performed in which the study definition is input by a user and before or after a search is performed the study definition may be saved into a computer memory, and a further instance of the method may then subsequently be performed in which the study definition is loaded from the computer memory and is re-converted into a database query and used to re-run the search. This may allow the search results to be updated to account for updates to the patient records being searched and / or to the relationships between the clinical terms defined in the clinical database. In some embodiments, the method may further comprise saving the study definition to a computer memory or exporting the study definition to some external system. This may enable the study definition to be reused in future to rederive the database query, potentially with a more up-to-date clinical concept database. In some embodiments, the study definition is obtained by loading it from a computer memory, or the study definition is obtained from user input and the method comprises saving the study definition to the computer memory. The user selected patient criteria may comprise inclusion and / or exclusion criteria. For example, the study definition may be configured to include patients diagnosed with a given clinical finding but to exclude those prescribed a given pharmaceutical product. A patient may match an inclusion criteria comprising a clinical concept if they are associated with that clinical concept in patient records. A patient may be excluded by an exclusion criteria comprising a clinical concept if they are associated with said clinical concept in the patient records. Therefore, a patient may match patient criteria including an exclusion criteria comprising a clinical concept only if they are not associated with said clinical concept in patient records. In some embodiments the study definition may comprise logical relationships between different patient criteria, such as AND and / or OR. relationships. In alternative embodiments, the study definition may comprise AND, OR, NOR, NAND, XOR, and / or XNOR relationships between different patient criteria. Alternatively, NOR, NAND, XOR, or XNOR relationships can be achieved by combining AND, OR and NOT logical operators. In some embodiments, the study definition may comprise NOT logical relationships associated with one or more patient criteria. In some embodiments, one, some or all of the user selected patient criteria comprising a clinical concept may further comprise requirements for one or more parameters associated with that clinical concept, such as an associated date or amount of that clinical concept. For example, the criteria may be to include or exclude patients associated with a given clinical concept in a given date range, or patients associated with a given amount of a given clinical concept (for example, an observable entity clinical concept which may be a measurement performed on a patient). In some embodiments, the some or all of the clinical concept criteria may be in one or more pre-constructed groups of related clinical criteria. Such groups may stored by a system for creating a search query and / or may be selectable by a user creating a search query. Such groups may facilitate the creation of search queries involving common groups of related clinical concepts. In some embodiments, in addition to the one or more selected clinical concepts, the user selected patient criteria may comprise demographic criteria. In some embodiments demographic criteria may include patient age, location, ethnicity, sex, whether they are and / or have been a smoker, and a relative deprivation score (such as an index of multiple deprivation (IMD)). The demographic criteria may be inclusive or exclusive criteria. In some embodiments, one or more of clinical concept criteria may be individually associated with one or more corresponding demographic criteria, for example via an AND logical relationship. These clinical concepts together with their associated demographic criteria may have logical relationships with other clinical concepts, with or without associated demographic criteria. For example, this may allow the study definition to search for patients who have been associated with a first clinical concept OR patients over a threshold age who have been associated with a second clinical concept. The various optional features of the user-selected criteria may enable a researcher using the method to perform a more appropriate search for their intended clinical study. The one or more selected clinical concepts are each associated with one or more subtypes thereof in a clinical concept database. The clinical concept database may comprise formatted clinical concepts and relationships therebetween. In some embodiments, the clinical concept database may directly and / or separately associate each clinical concept with each of its subtypes, not just with its immediate child concepts. In some such embodiments, the clinical concept database comprises a transitive closure table that associates the one or more subtypes with each of the selected clinical concepts. Such a transitive closure table may store hierarchical relationships between clinical concepts and their subtypes in a flat database. In some embodiments, the clinical concept database may be configured to obtain updates from a parent terminology database, such as a database hosting an up-to-date edition of a clinical terminology such as SNOMED CT. Such updates may include updates to clinical concepts in the terminology and / or relationships therebetween. The updates may be received from the parent terminology database and / or requested by the clinical concept database. In some embodiments, the updates may be obtained periodically, for example, monthly. In some such embodiments, updated clinical concepts and relationships therebetween may be pre-processed and / or reformatted, manually and / or automatically upon being received before being incorporated into the clinical concept database. The clinical concept database may be comprised by the same computer system that performs the method or may be comprised by some separate or remote system or server with which the method-performing system communicates. It will be appreciated that in addition to the associations between each of the one or more selected clinical concepts and the one or more subtypes thereof, the clinical concept database may store associations between additional clinical concepts. The clinical concept database may store all the clinical concept relationships of a medical terminology dataset, such as SNOMED CT. The method comprises deriving a database query from the study definition in which each of the selected clinical concepts is replaced with a set comprising that selected clinical concept and the one or more subtypes thereof with which it is associated in the clinical database. In some embodiments, deriving the database query may comprise providing each of the selected clinical concepts to the clinical concept database, and receiving in response the one or more subtypes associated therewith. Each set of a selected clinical concept and its subtypes may comprise an OR logical relationship between the concepts therein. Any logical relationship or additional requirements applied to a selected clinical concept in the user-selected patient criteria may automatically also be applied to each subtype thereof. In some embodiments, in the event that a selected clinical concept in the user-selected patient criteria is a subtype of another of the selected clinical concepts, said clinical concepts may be combined in the same set along with all subtypes of the more general clinical concept. The method comprises performing a search of a patient records database using the database query. The patient records database may define associations between individual patients and one or more clinical concepts. The patient records database may be anonymised, for example where the patient records database is a centralized database combining records of patients from a plurality of direct care givers. Patients in the database may be specified in the database by anonymous patient ID numbers. Alternatively, the patient records database may not be anonymised, for example where the patient records database only includes records of patients of a direct care giver who has access to search the database. Non-anonymised patient records may be provided by a combination of an anonymised patient records database in which patients are specified by anonymous patient ID numbers, and a separate database storing associations between the anonymous patient ID numbers and non-anonymous patient identities. The patient records database may comprise a plurality of entries, each comprising a patient identifier, such as a patient ID number, and clinical concept, which may be denoted using a clinical concept ID number, such as the concept ID number associated with a clinical concept in the SNOMED CT terminology. A plurality of such entries may exist for a single patient, separately associating them with different clinical concepts, which may facilitate searching the database. In some embodiments, each entry may further comprise a date, which may indicate when the patient in question was associated with the clinical concept, such as when they were diagnosed with a given condition, prescribed a given product, or had a given procedure performed. In such embodiments, a patient may be associated with the same clinical concept multiple times in multiple separate entries with different dates. In some embodiments, some of the entries may further comprise an amount of the clinical concept, for example entries for which the clinical concept is an observable entity (for such as 'standing height', 'Concentration of 1,5-anhydroglucitol' or other measurements performed on a patient) may further comprise an amount of said clinical concept, and / or some entries for which the clinical concept is a pharmaceutical / biologic product (such as a prescribed drug) may further comprise a dosage amount of said clinical concept. The amount may be specified by a value and an associated unit. For example, the patient records database may comprise a plurality of entries, each comprising a patient ID number, a clinical concept ID number, and a date, and optionally further comprising an amount of the clinical concept and an associated unit. The patient records database may also include demographic data associated with individual patients. Such demographic criteria may be searched using demographic criteria comprised by the user-selected criteria, for example to include or exclude certain patients. The contents of the patient records database may be compiled from a plurality of external databases, such as general practitioners' and / or hospitals' patient databases. The patient records database may record which separate database each patient's records were obtained from. In some embodiments, the patient records may be configured to obtain updates from said separate databases. The updates may be received from the separate database and / or requested by the patient records database. In some embodiments, the updates may be obtained periodically, for example, monthly. In some such embodiments, updated patient records may be pre-processed and / or reformatted, manually and / or automatically upon being received before being incorporated into the patient records database. Constructing the patient records database may comprise receiving patient data from a plurality of other databases, such as general practitioner databases, combining and anonymising said data, and formatting the anonymised data for searching using the database query. The method comprises outputting information on patients identified in the patient records database which match the database query. Outputting information on the identified patients may comprise displaying the information, saving the information to a computer memory, exporting a file comprising the information and / or transmitting the information to a separate device. Patients may be identified as matching the database query if their records in the patient records database satisfy the combination of criteria in the database query, said criteria comprising the user-selected criteria with the selected clinical concepts replaced with sets comprising said concepts as well as their subtypes. In some embodiments, the output information may comprise a list of the patients who were found to match the database query, for example, when the method is used by a direct care giver to search their own non-anonymised patient records. Alternatively, or additionally, the output information may comprise the identities of one or more external databases containing patient records of at least a threshold number of patients who were found to match the database query (such as external databases from which their records were obtained as described above) and / or of one or more medical practitioners or institutions (such as general practitioners) associated with at least a threshold number of patients who were found to match the database query. For example, when the method is used by a researcher to search a centralized, potentially anonymised patient records database. The threshold number may be one. Alternatively, or additionally, the output information may comprise the number of patients who were found to match the database query with patient records in each of one or more external databases (such as the numbers of patients whose patient records were obtained from each of one or more external patient databases) and / or who are associated with each of one or more medical practitioners or institutions (such as the number of patients who are associated with each of one or more direct care givers, such as general practitioners). Such embodiments may enable a researcher using the invention to contact medical practitioners or institutions associated with patients that match the criteria to inform them that some of - or a specific number of - their patients match the clinical study's requirements. In some embodiments, the output information comprises the number of patients who were found to match the database query with patient records in each of one or more external databases and / or who are associated with each of one or more medical practitioners or institutions and does not comprise any information identifying individual patients, for example wherein the searched patient records database is anonymised as described above. This may allow a researcher using the invention to inform medical practitioners or institutions that they have suitable patients, without the researcher being provided with the identities of the matching patients, thereby providing enhanced privacy and anonymity. After a researcher contacts a medical professional or institution to inform them that a number of their patients have been identified as suitable for participation in the study, the medical professional or institution may then search records of their patients to identify their own patients who satisfy the criteria. The medical professional or institution may perform this search using a further instance of the invention, wherein the patient records database containing non-anonymised records of their own patients. The medical practitioners and / or institutions may then selectively contact their matching patients to inform them of the clinical study, to invite them to participate, and / or to request permission to provide their details to the researcher. In some embodiments, the method further comprises communicating one or more outside parties associated with some or all of the patients who were found to match the database query, such as medical practitioners and / or institutions as described above. For example, the method may comprise automatically sending a communication to each of one or more medical practitioners or institutions associated with patients who match the database query and / or to the medical practitioners, institutions or other persons associated with each of one or more external databases containing records of patients who match the database query (or from which records of said patients were obtained). Such communications may comprise the number of patients in said databases and / or associated with said medical practitioners and / or institutions who satisfied the database query. Alternatively, or additionally, such communications may comprise the identities of the patients in said databases and / or associated with said medical practitioners and / or institutions who satisfied the database query. In some embodiments, such communications may comprise study definition and or user-selected criteria thereof. This may enable the recipient of the communication such as a medical practitioner or institution to conduct their own searches, for example by using the invention themselves, deriving their own database query and searching their own set of patient records, which may be a separate patient records database, which may be nonanonymised. In some embodiments, communications as described above may comprise one or more clinical study invitations (such as the text of a letter, email or other message). The invitations may be pre-generated for specific identified patients, or may be a template for completion with individual patient details. The recipients of the communications, such as medical practitioners and / or institutions, may use the one or more invitations to contact identified patients to invite them to participate in the study. The recipients may be provided with means to automatically transmit digital such invitations, or to print invitations for posting. Alternatively, or additionally, the method may comprise automatically sending clinical study invitations directly to some or all of the identified patients, or enabling a user to selectively send such study invitations, for example, wherein the method is used by a medical practitioner searching their own non-anonymised patient records after being informed that a number of their patients are suitable for [participation in the study. The method may comprise receiving or generating clinical study invitations as described above. The method may be performed by a system or server which may comprise or be in communication with the clinical database and / or the patient records database. According to a further aspect of the invention, there is provided a method of recruiting patients for a clinical study, wherein the patients are recruited on the basis of information on patients that has been output using a method as described above. The method may comprise obtaining such information, such as by performing a method as described above. The method with which the information on patients is output may comprise any of the optional features described herein. According to a further aspect of the invention there is provided a method of conducting a clinical study, the method comprising recruiting patients according to a method as described above and further comprising administering a pharmaceutical product to one or more of the recruited patients. According to a further aspect of the invention there is provided a system for identifying patients suitable for participation in a clinical study, the system comprising: one or more processors, and one or more computer readable storage devices, the one or more storage devices storing computer instructions which when implemented cause the one or more processors to: obtain a study definition comprising user-selected patient criteria for the clinical study, wherein the patient criteria comprise one or more selected clinical concepts each of which is associated with one of more subtypes thereof in a clinical concept database; derive a database query from the study definition in which each of the selected clinical concepts is replaced with a set comprising that selected clinical concept and the one or more subtypes thereof with which it is associated in a clinical concept database; perform a search of a patient records database using the database query; and output information on patients identified in a patient records database which match the database query. In some embodiments, the one or more computer readable storage devices further store the clinical concept database and / or the patients records database. Alternatively, the system may be configured to communicate with devices storing said databases. The computer instructions may cause the processors to perform any of the optional steps described above with reference to the first aspect of the invention, and the system may comprise any optional features described above with reference to the first aspect of the invention. Additionally, or alternatively, the computer instructions may cause the processors to perform additional steps of methods of recruiting patients or conducting a clinical study as described above. Brief Description of the Drawings Exemplary embodiments of the invention will now be described in detail with reference to the accompanying figures, in which: Fig. 1 shows an example of a study definition comprising user-selected patient criteria; Fig. 2 shows a tree hierarchy showing examples of subtypes of the clinical concept 'Diabetes mellitus' in the SNOMED CT terminology; Fig. 3 shows the operation of a system configured to perform an embodiment of a method for identifying patients suitable for participation in a clinical study; Fig. 4 shows a database query derived from the study definition shown in Fig. 1; Fig. 5 shows an embodiment of identifying patients suitable for participation in a clinical study including demographic criteria, a combined anonymised patient database and SNOMED CT clinical concept relationships defined in a concept transitive closure table database; Fig. 6 shows entries of a transitive closure table in a clinical concept database defining the relationships between the clinical concepts shown in Fig. 5a; Fig. 7 shows an example of patient records in a patient records database; Fig. 8a shows results of a search, indicating the number of patients that satisfy the study definition of Fig. 1 and individual patient criteria thereof; Fig. 8b shows a user interface displaying the numbers of patients satisfying the study definition of Fig. 1 in each of a plurality of general practices; Fig. 9a shows a user interface displaying a list of identified patients; and Fig. 9b shows a user interface displaying means for printing invitations for identified patients. Detailed Description Referring to the figures generally, there are shown embodiments of systems and methods for identifying patients suitable for participation in a clinical study, in particular for searching records of a plurality of patients to identify all of those whose records satisfy meet a study definition specified in terms of clinical concepts. Clinical concepts are medical terms that identify individual ideas such as clinical findings, pharmaceutical products, medical procedures, observable entities or classes thereof. Collections of clinical concepts and the relationships between them make up specific medical terminologies such as the widely used 'Systematized Nomenclature of Medicine Clinical Terms' (SNOMED CT) terminology. A study definition may be defined in terms of clinical concepts in order to specify which patients are or are not suitable for participation in the study, for example to specify patients who have or have not been diagnosed with certain clinical findings and / or who have or have not been prescribed certain pharmaceutical products. Fig. 1 shows an example of a study definition 100 consisting of user-selected patient criteria specified in terms of clinical concepts, as presented in a user interface. The study definition consists of clinical concept-based inclusion criteria 110, clinical concept-based exclusion criteria 120, and demographic criteria 130. In the illustrated example study definition the inclusion criteria 110 consist of the clinical concepts 'Diabetes mellitus' and 'Obese', the exclusion criteria 120 consists of the group of related clinical concepts 'Diabetes Medications', and the demographic criteria are male patients of any age, ethnicity, smoker status and IMD score. The user interface enables a user to edit the study definition, to add, edit, or remove inclusion criteria 110, exclusion criteria 120, demographic criteria 130, and / or logical relationships between different criteria. The individual inclusion and exclusion criteria are arranged in 'inclusion groups' and 'exclusion groups' respectively to which additional clinical concept criteria and demographic criteria may be added using "add criteria" or "add demographics" options respectively. In the example configuration shown in Fig. 1 in the absence of additional qualifiers in the study definition the criteria are all interrelated by AND logical relationships such that all the criteria must be satisfied (with the exclusion criteria being qualified by a NOT logical relationship or equivalent NAND logical relationship). However, it will be appreciated that in alternative embodiments, the default logical relationships may differ, or explicit logical relationships between criteria may always be necessary to include. The patient criteria may therefore be specified as: (<Diabetes mellitus> AND <Obese>) AND (NOT<Diabetes Medications>) AND <Gender=Male> However, many clinical terms are medical terms that define classes of other more specific medical ideas. These other clinical concepts are 'subtypes' of the more general clinical concept. The relationships between clinical concepts which specify which concepts are subtypes of other clinical concepts are specified in medical terminologies such as SNOMED CT. Fig. 2 is a tree hierarchy showing an illustrative selection of the subtypes of the clinical concept 'Diabetes mellitus' as defined in the SNOMED CT UK Edition, which was part of the inclusion criteria 110 of the study definition shown in Fig. 1. The subtypes of a clinical concept such as 'Diabetes mellitus' include 'child' subtypes which are directly related to that 'parent' clinical concept directly by a single is-a relationship, as well as a number of other 'descendants' that are subtypes of other subtypes related back to that clinical concept by multiple generations of is-a relationships. The property of being a subtype is transitive, such that if first concept 'A' is a subtype of a second concept 'B' and 'B' is a subtype of a third concept 'C', then 'A' is also a subtype of 'C'. Diabetes mellitus has twenty direct child subtypes, of which two ('Diabetes mellitus type 1' and 'Diabetes mellitus type 2') are shown. The remaining unillustrated subtypes are 'Atherosclerosis, deafness, diabetes, epilepsy, nephropathy syndrome', 'Atypical diabetes mellitus', 'Brittle diabetes mellitus', 'Diabetes mellitus due to genetic defect in beta cell function', 'Diabetes mellitus due to genetic defect in insulin action', 'Diabetes mellitus due to pancreatic injury', 'Diabetes mellitus during pregnancy, childbirth and the puerperium', 'Diabetes mellitus in remission', 'Diabetes mellitus with multiple complications', 'Diabetes mellitus without complication', 'Gingival disease co-occurrent with diabetes mellitus', 'Houssay's syndrome', 'Ketosis-prone diabetes mellitus', 'Maternally inherited diabetes mellitus', 'Newly diagnosed diabetes', 'Pancreatic hypoplasia, diabetes mellitus, congenital heart disease syndrome', 'Pre-existing diabetes mellitus', and 'Secondary diabetes mellitus'. Of the twenty child subtypes, eleven have subtypes of their own, which are necessarily also subtypes of Diabetes mellitus. Fig. 2 shows two of Diabetes mellitus type 1's twenty own children ('Wolfram syndrome' and 'Insulin dependent diabetes mellitus type IA') and one of Diabetes mellitus type 2's twelve children ('Brittle type 2 diabetes mellitus'). In total, Diabetes mellitus has one hundred and fifty-eight subtypes, of which Fig. 2 only shows five for visual clarity. When searching patient records for potential participants in a clinical study using criteria that include or exclude a given clinical concept, individual patients' records may only refer to a subtype of that concept, preventing them from being identified by a search using the more general supertype clinical concept. Embodiments described herein provide improved searching in which an initial study definition comprising one or more user-selected patient criteria is transformed into a database query comprising both the original clinical concepts and also all their associated subtype. This enables a user to only specify the exact desired class of clinical concepts and to have the search automatically identify all relevant patients, including those whose records only refer to specific subtypes of the concept in question. Fig. 3 is a diagram showing the operation of a system configured to perform an embodiment of a method for identifying patients suitable for participation in a clinical study. The system obtains a study definition, for example from user input or from storage in a memory device. The study definition comprises user-selected patient criteria 310 for the clinical study, including one or more clinical concepts 315, such as a study definition as described above with reference to Fig. 1. The system queries a clinical concept database 320 that stores associations between clinical concepts and subtypes thereof, including between the one or more user-selected clinical concepts 315 and subtypes thereof. The clinical concept database returns each of subtypes 325 associated with the one or more user-selected clinical concepts therein and the system combines these subtypes 325 with the original patient criteria 310 to together define a database query 330. In the database query 330, each of the selected clinical concepts 315 is replaced with a set comprising that selected clinical concept 315 and the one or more subtypes 325 thereof. The system uses the database query 330 to search a patients records database 340 which stores patient records including clinical concepts and obtains search results 350 with information on patients that satisfy the database query 330. Fig. 4 shows an example of a database query 400 derived from the example study definition 100 shown in Fig. 1, using a method as described above with reference to Fig. 3. The database query 400 comprises inclusion criteria 410 and exclusion criteria 420 derived from the inclusion and exclusion criteria 110, 120 of the study definition 100 by introducing the subtypes of the clinical concepts therein, as well as demographics criteria 430 identical to the demographics criteria 130 of the study definition 110. The inclusion criteria 110 of the study definition 100 included two clinical concepts 'Diabetes mellitus' and 'Obese', and the inclusion criteria 410 of the database query 400 therefore include two sets 412, 414 of inclusion criteria, each consisting of one of the original clinical concepts and all the subtypes thereof recorded in the clinical concept database, interrelated by logical OR relationships. The two sets are interrelated by a logical AND relationship, in the same manner as the two original clinical concepts in the study definition 100. The exclusion criteria 420 of the database query consists of a set of clinical concepts, including all the members of the originally excluded group of clinical concepts 'Diabetes Medications' as well as all the subtypes of each clinical concept therein, the clinical concepts in the set being interrelated by logical OR. relationships. Fig. 5 is a diagram showing the operation of a further specific embodiment of a system as described above with reference to fig. 3 which is configured to perform an embodiment of a method for identifying patients suitable for participation in a clinical study. In the embodiment shown in Fig. 5 the obtained study definition comprises concept-based criteria 515 (such as the inclusion and exclusion criteria 110, 120 of the example study definition 100 shown in Fig. 1) and demographic criteria 510 (such as the demographics criteria 130 shown in Fig. 1). The clinical concepts in the concept-based criteria are provided to a clinical concept transitive closure table database 520 which stores relationships between the subtypes and supertypes of a set of clinical concepts in the SNOMED CT system. A transitive closure table is a method of storing hierarchical relationships between clinical concepts and their subtypes in a flat database in which each clinical concept is directly associated with each of its subtypes in a separate entry in the table and database, instead of only with its immediate child concepts. Fig. 6 shows an example of a portion of a transitive closure table string the subtype-supertype relationships between each of the clinical concepts shown in Fig. 2. In the example shown in Fig. 5, the clinical concept transitive closure table database 520 periodically obtains updated clinical concept data 522 from a SNOMED Terminology repository 524. This ensures that the contents of the clinical concept transitive closure table database 520 are kept up to date and can allow a saved study definition to be researched using more up to date clinical concept data in future. Each of the subtypes associated with each of the study definition's clinical concepts 515 in the clinical concept transitive closure table database 520 are obtained and original clinical concepts and their subtypes together 530 are used to construct a database query 540 as described above. The database query 540 is used to search a combined anonymised patient database 550 which contains patient records obtained from a plurality of external databases 555 - such as individual general practitioner's patient records databases - which have been combined, anonymised, and reformatted for inclusion in the combined anonymised patient database 550 and searching using the database query 540. Fig. 7 shows an example of patient records 700 as stored within a patient database such as the combined anonymised patient database 550 shown in Fig. 5. Each entry in the patient database 700 associates a single patient, specified by an anonymised patient ID number 720 with a single clinical concept 740 specified by that clinical concept's ID number in the SNOMED CT system. This facilitates searching the patient records for patients associated with specific clinical concepts. Additionally, each entry includes an associated entry number 710 for uniquely identifying that entry and an entry date 730 specifying when that clinical concept was associated with that patient (for example, when they were diagnosed with a disorder or when they were prescribed a pharmaceutical product). A single patient may have multiple entries, potentially associating them with a number of different clinical concepts over their lifetime, for example Fig. 7 shows an excerpt of ten entries, each relating to the patient specified by PatientID number 482763. A patient may be associated with the same clinical concept multiple times, for example on different dates. In addition to the entry number 710, patient ID number 720, entry date 730, and clinical concept ID number 740, each entry further comprises a concept amount 750 and an associated unit 760. The values of the concept amount 750 and the associated unit 760 are NULL for a number entries whose clinical concepts do not have associated amounts. However, entries with clinical concepts such as observable entities (i.e. measurements performed on patients) or pharmaceutical products may have amounts in their associated entries (such as measurement amounts or dosages). For example, the second entry shown in Fig. 7 with clinical concept ID 997591000000109 ('Serum total bilirubin level') has an associated concept amount of 15 and associated concept units pmol / L, indicating that the patient had a serum total bilirubin level 15 micromoles per litre on the associated date of 13 January 1980. The database query 540 is used to search the combined anonymised patient database 550 for patients who match the database query and search results 560 including information on said patients are output. In some instances, the information on the patients output in the search results 560 may be a number of patients who satisfy the database query 540 and / or individual criteria of the database query or study definition. Such results may be used by a researcher to establish the feasibility of a study by reviewing how many patients might be available for participation therein. Fig. 8a shows an example of such search results being provided in a version of the user interface shown in Fig. 1. Such a user interface may be used by a user to configure or view a study definition and to then run a search based on said study definition and obtain a number of matching patients in the searched patient records database. The user interface shows a Recruitment estimate of thirty-nine patients who were found to satisfy all of inclusion, exclusion and demographics criteria of the study definition. Additionally, the user interface presents the "Group total" numbers of patients whose records are found to satisfy each individual inclusion or exclusion criteria, as well as a "Running total" numbers of patients whose records satisfy that criteria as well as all preceding criteria. For example, in the illustrated example, 19,404 patients are associated with "Diabetes Mellitus" or subtypes thereof so satisfy the first inclusion criteria, 2,344 patients are associated with "Obese" or subtypes thereof so satisfy the second inclusion criteria, and 2,156 patients are associated with both inclusion criteria. These totals may be updated every time the search query is run by a user. The user interface may also enable a user to 'expand' the user interface elements associated with individual clinical concept criteria to view a list of the subtypes thereof and the "Group total" and "Running total" numbers of patients for each individual subtype. In some embodiments, small numbers of patients below a threshold may be suppressed. For example, if fewer than five patients are in a given group or running total <5 may be displayed instead of an exact number. In some instances, the information on the patients output in the search results 560 may be a number of patients associated with each of one or more medical practitioners or institutions, such as the number of patients associated with each of a plurality of general practitioners or other direct care givers. These medical practitioners or institutions may be medical practitioners or institutions from whose external patient records databases 555 patient records were obtained. These numbers may be determined using a separate table that associates Patient ID numbers with corresponding medical practitioners or institutions. Outputting this information may enable a researcher to contact the medical practitioners or institutions, informing them that a number of their patients have been identified as suitable for participation in the clinical study. Fig. 8b shows an example of a user interface showing which general practitioner practices the thirty-nine patients satisfying the search query shown in fig. 8a are associated with. In the illustrated example, nineteen patients are associated with 'Example Practice 1', nine patients are associated with 'Example Practice 2', seven patients are associated with 'Example Practice 3', and four patients are associated with 'Example Practice 4'. The user interface displays additional information on each of the practices. A status column indicates whether each practice has signed up to be involved in a medical trial, those with ticks have signed up to be involved as active sites for recruitment. Those practices without a tick in the status column have not yet signed up as active recruitment signs and would need to be approached for an agreement before inviting patients from them. IMD decile and rank columns give information on the index of multiple deprivation for the practice, and the practices' locations can be viewed on a map. The user interface includes functionality for a user to invite a signed-up practice to participate in the medical trial. A recruitment type column allows a user to select what role the practice will play during recruitment for the trial, such as whether the practice is an Active site where patients are actively recruited from that site, or a PIC (Participant Identification Centres) or Hub site who coordinate recruitment across a number of other sites / practices. After a medical practitioner or institution has been invited to participate in a medical trial, they may then use the system shown in Fig. 5 to perform their own search of the patients records in the patient records database 550. The information on the patients output in the search results 560 comprises a list of patients of that medical practitioner or institution who satisfy the database query. The anonymous patientID numbers are used to lookup identifiable data for the patients associated with that medical practitioner or institution and the search results 560 presented to them are non-anonymized, allowing them to identify and invite individual patients to participate in the study. Fig. 9a shows a first view example of a user interface for an administrator, medical practitioner or institution showing results of a search on a non-anonymized patients records database derived from a single external database 555. The user interface shows the identified patients listed in an 'Identified Patients' tab. The user interface enables a user to select individual patients for invitation to the clinical study, to be opted out of invitation to the clinical study, or to be opted out of invitation to all future clinical studies. Fig. 9b shows a second view of the user interface, showing an 'invitation' in which patients selected for invitation in the 'Identified Patients' tab are listed. The user interface enables a user to print automatically generated invitation letters for individual selected suitable patients, or for all selected suitable patients. The clinical concept database returns each of subtypes 325 associated with the one or more user-selected clinical concepts therein and the system combines these subtypes 325 5 with the original patient criteria 310 to together define a database query 330. In the database query 330, each of the selected clinical concepts 315 is replaced with a set comprising that selected clinical concept 315 and the one or more subtypes 325 thereof. This user interface enables the user, such as a medical practitioner or institution to select patients who they are more familiar with who they believe are more suitable for invitation 10 to the clinical study, while keeping patient records anonymized from the perspective of a researcher arranging the clinical study until patients have agreed to participate. It will be appreciated that variation may be made to the illustrated embodiments described above without departing from the scope of the invention as defined by the claims.

Claims

1. A computer implemented method for identifying patients suitable for participation in a clinical study, the method comprising:obtaining a study definition comprising user-selected patient criteria for the clinical study, wherein the patient criteria comprise one or more selected clinical concepts each of which is associated with one of more subtypes thereof in a clinical concept database;deriving a database query from the study definition in which each of the selected clinical concepts is replaced with a set comprising that selected clinical concept and the one or more subtypes thereof with which it is associated in the clinical concept database;performing a search of a patient records database using the database query; and outputting information on patients identified in the patient records database which match the database query.

2. A method according to claim 1, wherein the one or more selected clinical concepts, the one or more subtypes thereof, and the associations between the one or more clinical concepts and the subtypes thereof are comprised by a hierarchical or polyhierarchical dataset of clinical concepts.

3. A method according to claim 1 or claim 2, wherein the one or more selected clinical concepts, the one or more subtypes thereof, and the associations between the one or more clinical concepts and the subtypes thereof are comprised by SNOMED Clinical Terms.

4. A method according to any preceding claim, further comprising updating the clinical concept database to reflect changes in a SNOMED database source used to populate the clinical concept database5. A method according to any preceding claim wherein the clinical concept database comprises a transitive closure table defining the relationship between the one or more selected clinical concepts and the one or more subtypes thereof.

6. A method according to any preceding claim, wherein the patient criteria further comprise one or more demographic criteria.

7. A method according to any preceding claim, wherein the patient records database is built from one or more sources of digitized patient health data, wherein each patient record comprises structured data which includes one or more clinical concepts associated with the patient, and preferably also includes patient demographic criteria associated with the patient.

8. A method according to any preceding claim, further comprising communicating the study definition to a clinician, and responsive to user input by the clinician, performing a search of a patient records database restricted to patients of that clinician based on the database search query, and outputting a list of patients.

9. A method according to any preceding claim, further comprising generating a clinical study invitation for sending to one or more patients based on selections from the list of patients made by the clinician.

10. A method according to any preceding claim wherein the study definition is obtained by loading it from a computer memory, or the method is obtained from user input and the method comprises saving the study definition to the computer memory.

11. A method of recruiting patients for a clinical study, wherein the patients are recruited on the basis of information on patients that has been output using a method according to any preceding claim.

12. A method of conducting a clinical study, the method comprising recruiting patients according to the method of claim 11, and further comprising administering a pharmaceutical product to one or more of the recruited patients.

13. A system for identifying patients suitable for participation in a clinical study, the system comprising:one or more processors, andone or more computer readable storage devices, the one or more storage devices storing computer instructions which when implemented cause the one or more processors to:obtain a study definition comprising user-selected patient criteria for the clinical study, wherein the patient criteria comprise one or more selected clinical concepts each of which is associated with one of more subtypes thereof in a clinical concept database;derive a database query from the study definition in which each of the selected clinical concepts is replaced with a set comprising that selected clinical concept and the one or more subtypes thereof with which it is associated in a clinical concept database;perform a search of a patient records database using the database query; and output information on patients identified in a patient records database which match the database query.