Rapid reconciliation of errors and bottlenecks in data-driven workflows
A unified framework for healthcare workflows to introduce multiple integrated approaches to error analysis. A first approach uses machine learning to extend probabilistic record linkage and apply it to the task of reconciliation, classifying changes between datasets as intentional or unintentional. A second approach uses process mining to extract maximum information about process diagrams and process bottlenecks.
1. An electronic system comprising:
at least one non-transitory computer-readable memory storing machine-readable instructions; and
a control system including one or more processors configured to execute the machine readable instructions to:
receive as input an event log, the event log comprising entries describing activities performed in a healthcare-administration workflow, the start time of each of the activities, the end time of each of the activities and the place in the workflow at which each of the activities occurs;
determine from the event log a data set characterizing all of the activities performed in the workflow;
determine from the event log a data set characterizing the start time of each of the activities and the end time of each of the activities;
determine from the event log a data set characterizing the places in the workflow at which each of the activities occurs;
determine from the event log a starting place in the workflow and an ending place in the workflow;
based on the data sets, generate a diagram illustrating the places in the workflow and the activities of the workflow occurring in the places;
based on the data sets, generate a set of arrows in the diagram connecting the illustrated places in the workflow with each other according to the sequence of the activities performed in the workflow, the set of arrows beginning at the illustrated starting place in the workflow and terminating at the illustrated ending place in the workflow;
receive first and second datasets respectively including first and second record sets of a predetermined number, each of the first and second record sets including one or more associated record identifiers and one or more associated numerically represented dataset features, the second dataset resulting from one or more transformations of the first dataset by an entity, the one or more transformations each being one of an intentional transformation or an unintentional transformation;
generate first and second matrices based respectively on the first and second datasets, rows of the first and second matrices including the respective one or more record identifiers and columns of the first and second matrices including the one or more dataset features;
calculate a first checksum on the first matrix and a second checksum on the second matrix;
in response to determining that the first checksum does not equal the second checksum, calculate a third checksum on all rows of the first matrix and a fourth checksum on all rows of the second matrix;
in response to determining that the third checksum does not equal the fourth checksum, calculate, based on all rows of the first and second matrices, a first set of scores indicating that one or more of the transformations was intentional and a second set of scores indicating that one or more of the transformations was unintentional;
determine from the first and second set of scores which transformations were intentional and unintentional;
create the event log; and
record only the intentional transformations in the event log.
2. The electronic system of claim 1 , wherein the record identifiers comprise an identifier of a medical patient.
3. The electronic system of claim 1 , wherein the record identifiers comprise an identifier of a medical insurance claim.
4. The electronic system of claim 1 , wherein the data set features comprise a similarity function.
5. The electronic system of claim 1 , wherein the event log comprises entries describing the activity, a subject of the activity and the time at which the activity occurred.
6. At least one non-transitory computer-readable medium on which are stored instructions that, when executed by at least one processing device, enable the at least one processing device to perform a method, the method comprising the steps of:
receiving as input an event log, the event log comprising entries describing activities performed in a healthcare-administration workflow, the start time of each of the activities, the end time of each of the activities and the place in the workflow at which each of the activities occurs, the activities being performed by one or more of health insurance entities, medical patients and physician offices;
determining from the event log a data set characterizing all of the activities performed in the workflow;
determining from the event log a data set characterizing the start time of each of the activities and the end time of each of the activities;
determining from the event log a data set characterizing the places in the workflow at which each of the activities occurs;
determining from the event log a starting place in the workflow and an ending place in the workflow;
based on the data sets, generating a diagram illustrating the places in the workflow and the activities of the workflow occurring in the places; and
based on the data sets, generating a set of arrows in the diagram connecting the illustrated places in the workflow with each other according to the sequence of the activities performed in the workflow, the set of arrows beginning at the illustrated starting place in the workflow and terminating at the illustrated ending place in the workflow;
receiving first and second datasets respectively including first and second record sets of a predetermined number, each of the first and second record sets including one or more associated record identifiers and one or more associated numerically represented dataset features, the second dataset resulting from one or more transformations of the first dataset by an entity, the one or more transformations each being one of an intentional transformation or an unintentional transformation;
generating first and second matrices based respectively on the first and second datasets, rows of the first and second matrices including the respective one or more record identifiers and columns of the first and second matrices including the one or more dataset features;
calculating a first checksum on the first matrix and a second checksum on the second matrix;
in response to determining that the first checksum does not equal the second checksum, calculating a third checksum on all rows of the first matrix and a fourth checksum on all rows of the second matrix;
in response to determining that the third checksum does not equal the fourth checksum, calculating, based on all rows of the first and second matrices, a first set of scores indicating that one or more of the transformations was intentional and a second set of scores indicating that one or more of the transformations was unintentional;
determining from the first and second set of scores which transformations were intentional and unintentional;
creating the event log; and
recording only the intentional transformations in the event log.
7. The at least one medium of claim 6 , wherein the record identifiers comprise an identifier of a medical patient.
8. The at least one medium of claim 6 , wherein the record identifiers comprise an identifier of a medical insurance claim.
9. The at least one medium of claim 6 , wherein the data set features comprise a similarity function.
10. The at least one medium of claim 6 , wherein the event log comprises entries describing the activity, a subject of the activity and the time at which the activity occurred.