Methods for automatic cohort selection in epidemiologic studies and clinical trials

1. A method for selecting a cohort for a clinical study, comprising:

accessing from an electronic health record database, an unstructured electronic health record for each patient in a population of patients;

for each patient in the population of patients:

transforming the associated unstructured electronic health records of the patient into a patient graph, the patient graph representing a set of clinical concepts and corresponding semantic roles for the patient; and

appending the patient graph to a set of patient graphs;

at a clinical research interface, recording a first set of inclusion and exclusion criteria;

transforming the first set of inclusion and exclusion criteria into a first target graph representing a first set of target clinical concepts for the clinical study;

for each patient in the population of patients, in response to alignment between the patient graph of the patient and the first target graph, appending the patient to a first subpopulation of patients;

rendering a first representation of the first subpopulation of patients within the clinical research interface;

at the clinical research interface, recording a second set of inclusion and exclusion criteria;

transforming the first set of inclusion and exclusion criteria and the second set of inclusion and exclusion criteria into a second target graph for the clinical study;

for each patient in the population of patients, in response to alignment between the patient graph of the patient and the second target graph, appending the patient to a second subpopulation of patients;

rendering a second representation of the second subpopulation of patients within the clinical research interface;

recording a selection of a potential clinical cohort from the second subpopulation of patients;

in response to access of a particular electronic health record of a particular patient included in the potential clinical cohort at a point-of-care interface, rendering a visual indicator of the particular patient's eligibility for the clinical study within the point-of-care interface;

at the point-of-care interface, recording a first interaction with the visual indicator; and

in response to recording the first interaction with the visual indicator, enrolling the particular patient in the clinical study,

wherein the method further comprises:

ingesting a corpus of medical data; and

generating a universal graph based on the corpus of medical data;

wherein transforming the associated unstructured electronic health records of the patient into the patient graph, comprises transforming the associated unstructured electronic health records of the patient into the patient graph based on the universal graph, the patient graph representing the set of clinical concepts and the corresponding semantic roles for the patient;

wherein transforming the first set of inclusion and exclusion criteria into the first target graph comprises transforming the first set of inclusion and exclusion criteria into the first target graph, based on the universal graph, the first target graph representing the first set of target clinical concepts for the clinical study; and

wherein transforming the first set of inclusion and exclusion criteria and the second set of inclusion and exclusion criteria into the second target graph for the clinical study comprises transforming the first set of inclusion and exclusion criteria and the second set of inclusion and exclusion criteria into the second target graph for the clinical study based on the universal graph.

2. The method of claim 1 , wherein generating the universal graph based on the corpus of medical data comprises generating a set of context-specific universal graphs, each context-specific universal graph in the set of context-specific universal graphs representing a distinct clinical context.

3. The method of claim 1 further comprising:

for each patient in the population of patients, indexing the patient graph of the patient within the universal graph;

for each patient in the population of patients, detecting alignment between the first target graph and the patient graph of the patient by accessing a first index corresponding to the first target graph within the universal graph; and

for each patient in the population of patients, detecting alignment between the second target graph and the patient graph of the patient by accessing a second index corresponding to the second target graph.

4. The method of claim 3 , wherein indexing the patient graph of the patient within the universal graph comprises:

identifying an isomorphic subgraph of the universal graph isomorphic with the patient graph; and

listing the patient of the patient graph in an index corresponding to the isomorphic subgraph.

5. The method of claim 3 , wherein indexing the patient graph of the patient within the universal graph comprises:

identifying a set of morphologically similar subgraphs of the universal graphs morphologically similar to the patient graph; and

listing the patient of the patient graph in a set of indices corresponding to the set of morphologically similar subgraphs.

6. The method of claim 1 , wherein recording the first set of inclusion and exclusion criteria comprises:

receiving a selection of a file representing the clinical study; and

extracting the first set of inclusion and exclusion criteria based on the file representing the clinical study.

7. The method of claim 1 , for each inclusion criterion and exclusion criterion in the first set of inclusion and exclusion criteria and in the second set of inclusion and exclusion criteria:

generating a modified target graph without the criterion;

for each patient in the population of patients, in response to alignment between the patient graph of the patient and the modified target graph, appending the patient to a modified subpopulation of patients;

calculating a difference in population between the second subpopulation of patients and the modified subpopulation of patients; and

rendering a second visual indicator of the difference in population within the clinical research interface.

8. The method of claim 1 , wherein rendering the visual indicator comprises rendering the visual indicator within a point-of-care application displaying an electronic health record of the particular patient.

9. The method of claim 1 , wherein, in response to recording the interaction with the visual indicator, enrolling the particular patient in the clinical study comprises, in response to recording the interaction with the visual indicator:

rendering the first set of inclusion and exclusion criteria and the second set of inclusion and exclusion criteria applicable to the particular patient;

rendering an option icon for enrolling the particular patient in the clinical study; and

in response to recording an interaction with the option icon, enrolling the particular patient in the clinical study.

10. The method of claim 1 , further comprising:

monitoring an unstructured electronic health record of the particular patient;

in response to detecting a change in the unstructured electronic health record of the patient, generating a second patient graph for the particular patient; and

in response to detecting that the second patient graph does not match the second target graph, removing the particular patient from the second subpopulation of patients.