Back to results
US 11,699,507 B2Granted
Method and process for predicting and analyzing patient cohort response, progression, and survival
AssigneeTempus Labs, Inc.
InventorsHailey Lefkofsky, Ashraf Hafez, Julian Habib, Carin Fishel, Caroline Epstein
Life of this patent
Estimating the term…
Abstract
A system and method for analyzing a data store of de-identified patient data to generate one or more dynamic user interfaces usable to predict an expected response of a particular patient population or cohort when provided with a certain treatment. The automated analysis of patterns occurring in patient clinical, molecular, phenotypic, and response data, as facilitated by the various user interfaces, provides an efficient, intuitive way for clinicians to evaluate large data sets to aid in the potential discovery of insights of therapeutic significance.
Claims27
A method for identifying and separating groups of patients from a cohort of patients based on one or more targets, comprising:
- receiving patient information for the cohort of patients;
- identifying one or more interactions for each of one or more patients based at least in part on the received patient information;
- generating, for the one or more targets at each of the one or more interactions, one or more timeline metrics identifying whether each of the one or more targets occurs within a time period of an occurrence of the interaction;
- identifying, for each timeline metric of the one or more timeline metrics, whether a patient may incur one or more status characteristics within the time period;
- training a machine learning algorithm or a neural network implemented by a computer system to produce a target prediction model for each of the one or more targets based at least in part on the one or more status characteristics;
- generating, by the target prediction model, predictions for each of the one or more status characteristics for the one or more of patients;
- associating the predictions with a respective one or more timeline metrics of the one or more timeline metrics;
- generating, by the prediction model, a decision tree based on determining a greatest difference between each of the predictions and the timeline metrics, wherein the decision tree comprises a plurality of leaf nodes, each leaf node associated with a respective group of patients of the cohort of patients, each of the groups comprising similarities based at least in part on at least one status characteristic identified from respective branches leading to the leaf node, and wherein fewer than all of the patients in a leaf node satisfy the at least one status characteristic;
- providing importance ranks for the at least one status characteristic based on the fewer than all of the patients, wherein the importance ranks are provided without needing to retrain the target prediction model;
- and
- identifying at least one status characteristic for a patient not included in the cohort of patients based on matching the patient with one of the groups of patients associated with one of the plurality of leaf nodes.
8 more elements
A system for identifying and separating groups of patients from a cohort of patients based on one or more targets, comprising:
- at least one computer system configured to:
- receive patient information for the cohort of patients;
- identify one or more interactions for each of one or more patients based at least in part on the received patient information;
- generate, for the one or more targets at each of the one or more interactions, one or more timeline metrics identifying whether each of the one or more targets occurs within a time period of an occurrence of the interaction;
- identify, for each timeline metric of the one or more timeline metrics, whether a patient may incur one or more status characteristics within the time period;
- train a machine learning algorithm or a neural network to produce a target prediction model for each of the one or more targets based at least in part on the one or more status characteristics;
- generate, by the target prediction model, predictions for each of the one or more status characteristics for the one or more of patients;
- associate the predictions with a respective one or more timeline metrics of the one or more timeline metrics;
- generate, by the prediction model, a decision tree based on determining a greatest difference between each of the predictions and the timeline metrics, wherein the decision tree comprises a plurality of leaf nodes, each leaf node associated with a respective group of patients of the cohort of patients, each of the groups comprising similarities based at least in part on at least one status characteristic identified from respective branches leading to the leaf node, and wherein fewer than all of the patients in a leaf node satisfy the at least one status characteristic;
- provide importance ranks for the at least one status characteristic based on the fewer than all of the patients, wherein the importance ranks are provided without needing to retrain the target prediction model;
- and
- identify at least one status characteristic for a patient not included in the cohort of patients based on matching the patient with one of the groups of patients associated with one of the plurality of leaf nodes.
9 more elements
A computer program product for identifying and separating groups of patients from a cohort of patients based on one or more targets, comprising instructions stored on a non-transitory computer readable medium to cause at least one computer system to:
- receive patient information for the cohort of patients;
- identify one or more interactions for each of one or more patients based at least in part on the received patient information;
- generate, for the one or more targets at each of the one or more interactions, one or more timeline metrics identifying whether each of the one or more targets occurs within a time period of an occurrence of the interaction;
- identify, for each timeline metric of the one or more timeline metrics, whether a patient may incur one or more status characteristics within the time period;
- train a machine learning algorithm or a neural network to produce a target prediction model for each of the one or more targets based at least in part on the one or more status characteristics;
- generate, by the target prediction model, predictions for each of the one or more status characteristics for the one or more of patients;
- associate the predictions with a respective one or more timeline metrics of the one or more timeline metrics;
- generate, by the prediction model, a decision tree based on determining a greatest difference between each of the predictions and the timeline metrics, wherein the decision tree comprises a plurality of leaf nodes, each leaf node associated with a respective group of patients of the cohort of patients, each of the groups comprising similarities based at least in part on at least one status characteristic identified from respective branches leading to the leaf node, and wherein fewer than all of the patients in a leaf node satisfy the at least one status characteristic;
- provide importance ranks for the at least one status characteristic based on the fewer than all of the patients, wherein the importance ranks are provided without needing to retrain the target prediction model;
- and
- identify at least one status characteristic for a patient not included in the cohort of patients based on matching the patient with one of the groups of patients associated with one of the plurality of leaf nodes.
8 more elements
Dependencies are inferred from the claim text.