U.S. patent application number 16/429492 was filed with the patent office on 2019-10-17 for adaptive medical documentation system.
The applicant listed for this patent is CERNER INNOVATION, INC.. Invention is credited to Vikram Anand, Jan DeHaan, Faisal Farooq, Glenn Fung, John Haley, Balaji Krishnapuram, Joseph Marcus Overhage, Shipeng Yu.
Application Number | 20190318829 16/429492 |
Document ID | / |
Family ID | 50386041 |
Filed Date | 2019-10-17 |
![](/patent/app/20190318829/US20190318829A1-20191017-D00000.png)
![](/patent/app/20190318829/US20190318829A1-20191017-D00001.png)
![](/patent/app/20190318829/US20190318829A1-20191017-D00002.png)
![](/patent/app/20190318829/US20190318829A1-20191017-D00003.png)
![](/patent/app/20190318829/US20190318829A1-20191017-D00004.png)
United States Patent
Application |
20190318829 |
Kind Code |
A1 |
Fung; Glenn ; et
al. |
October 17, 2019 |
ADAPTIVE MEDICAL DOCUMENTATION SYSTEM
Abstract
Adaptive medical data collection for medical entities may
involve triggering an analysis of electronic records in response to
information input into an Electronic Medical Record (EMR) of a
patient. Determining a potential condition for the patient based on
the analysis. Identifying additional information indicated as
relevant to the potential condition of the patient, and generating
a request for the identified additional information.
Inventors: |
Fung; Glenn; (Madison,
WI) ; Krishnapuram; Balaji; (King of Prussia, PA)
; Farooq; Faisal; (Norristown, PA) ; Yu;
Shipeng; (Exton, PA) ; Overhage; Joseph Marcus;
(Zionsville, IN) ; Haley; John; (Chester Springs,
PA) ; DeHaan; Jan; (Hawley, PA) ; Anand;
Vikram; (Downingtown, PA) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
CERNER INNOVATION, INC. |
Kansas City |
KS |
US |
|
|
Family ID: |
50386041 |
Appl. No.: |
16/429492 |
Filed: |
June 3, 2019 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
14039125 |
Sep 27, 2013 |
10403403 |
|
|
16429492 |
|
|
|
|
61707147 |
Sep 28, 2012 |
|
|
|
Current U.S.
Class: |
1/1 |
Current CPC
Class: |
G16H 50/30 20180101;
G16H 50/20 20180101; G06F 19/00 20130101; G16H 10/60 20180101 |
International
Class: |
G16H 50/30 20060101
G16H050/30; G16H 10/60 20060101 G16H010/60; G16H 50/20 20060101
G16H050/20 |
Claims
1. A computerized method carried out by at least one server having
at least one processor for adaptive medical data collection, the
method comprising: triggering a probabilistic analysis of
electronic records in response to a first set of information input
into an Electronic Medical Record (EMR) of a patient; determining,
with the at least one processor, a first potential condition for
the patient based on the probabilistic analysis; identifying, with
the at least one processor, a second set of information indicated
as relevant to the first potential condition of the patient;
generating, with the at least one processor, a first request for
the second set of information by identifying an established
collection of information associated with the first potential
condition of the patient and providing the first request for the
second set of information in a preformatted medical form
established for the collection of the second set of information,
wherein the preformatted medical form is populated using the
established collection of information associated with the first
potential condition of the patient, and wherein the preformatted
medical form is automatically adapted based on the first set of
information input into the EMR; receiving at least a portion of the
second set of information; repeating the triggering of the
probabilistic analysis; and re-performing, with the at least one
processor, the probabilistic analysis using the second set of
information to identify a second potential condition.
2. The method of claim 1, further comprising identifying a third
set of information relevant to the second potential condition of
the patient based on the re-performed probabilistic analysis; and
generating, with the at least one processor, a second request for
the third set of information wherein the preformatted medical form
is automatically adapted based on the second set of information
received subsequent to the first request.
3. The method of claim 1, wherein triggering the probabilistic
analysis of the EMR comprises triggering an analysis of ontologies
of arbitrary contexts, clinical data records, practice data
records, clinical guidelines, or EMRs of prior patients of a
medical entity.
4. The method of claim 1, wherein triggering comprises recognizing
an input of data into the EMR of the patient.
5. The method of claim 1, wherein triggering the probabilistic
analysis comprises triggering an application of a machine learned
model to the EMR.
6. The method of claim 5, wherein triggering the application of the
machine learned model comprises a triggering an application of a
Bayes Net model trained using a Markov Chain Monte Carlo (MCMC)
method, and an Expectation Maximization method based model.
7. The method of claim 1, wherein determining the first potential
condition for the patient comprises determining a probability the
first potential condition applies to the patient, and comparing the
probability to a predetermined probability threshold.
8. The method of claim 1, wherein triggering the probabilistic
analysis comprises triggering an application of a first machine
learned model to the EMR, and identifying other information
regarding the first potential condition of the patient comprises
applying a second machine learned model to the second set of
information.
9. A system for adaptive medical data collection, the system
comprising: at least one memory operable to store Electronic
Medical Records (EMRs) and other medical data relating to
conditions of patients of a medical facility; and a processor
configured to: trigger a probabilistic analysis of electronic
records in response to a first set of information input into an EMR
of a patient, the probabilistic analysis comprising a probabilistic
network of associated terms derived from the other medical data;
determine a first potential condition for the patient based on the
probabilistic analysis; identify a second set of information
indicated as relevant to the first potential condition of the
patient from the probabilistic network; and generate a first
request for the second set of information by identifying an
established collection of information associated with the first
potential condition of the patient and providing the first request
for the second set of information in a preformatted medical form
established for the collection of the second set of information,
wherein the preformatted medical form is populated using the
established collection of information associated with the first
potential condition of the patient, and wherein the preformatted
medical form is automatically adapted based on the first set of
information input into the EMR and data input into the EMR
subsequent to the generation of the first request; receive at least
a portion of the second set of information repeat the triggering of
the probabilistic analysis; and re-perform the probabilistic
analysis using the second set of information to identify a second
potential condition.
10. The system of claim 9, further comprising: identifying a third
set of information relevant to the second potential condition of
the patient based on the re-performed probabilistic analysis; and
generate a second request for the third set of information wherein
the preformatted medical form is automatically adapted based on at
least a portion of the second set of information received
subsequent to the first request.
11. The system of claim 9, wherein the other medical data comprises
ontologies of arbitrary contexts, clinical data records, practice
data records, clinical guidelines, or EMRs of prior patients of a
medical entity.
12. The system of claim 9, wherein the probabilistic analysis is
triggered by a recognition of data input into the EMR of the
patient.
13. The system of claim 9, wherein the probabilistic analysis
comprises the application of a machine learned model.
14. The system of claim 13, wherein the machine learned model
comprises a generative probabilistic model.
15. A non-transitory computer readable storage medium having stored
therein data representing instructions executable by a programmed
processor for adaptive medical data collection, the storage medium
comprising the instructions for: triggering a probabilistic network
analysis of electronic records in response to a first set of
information input into an Electronic Medical Record (EMR) of a
patient; determining a first potential condition for the patient
based on the probabilistic analysis; identifying a second set of
information indicated as relevant to the first potential condition
of the patient; generating a first request for the second set of
information by selecting an established collection of information
associated with the first potential condition of the patient and
providing the first request for the second set of information in a
preformatted medical form established for the collection of the
second set of information, wherein the preformatted medical form is
populated using the established collection of information
associated with the first potential condition of the patient, and
wherein the preformatted medical form is automatically adapted
based on the first set of information input into the EMR; receiving
at least a portion of the second set of information; repeating the
triggering of the probabilistic analysis; and re-performing the
probabilistic network analysis of the EMR using the at least the
portion of the second set of information to update at least one
probability of the probabilistic network.
16. The medium of claim 15, further comprising: identifying a third
set of information relevant to a second potential condition of the
patient; and generating, a second request for the third set of
information wherein the preformatted medical form is automatically
adapted based on at least a portion of the second set of
information received subsequent to the first request.
17. The medium of claim 16, wherein when adequate information is
not available the instructions further comprise: identifying a
fourth set of information regarding the second potential condition
of the patient based on the third set of information; updating at
least one probability of the probabilistic network; and generating
a third request for the fourth set of information based on the
updated probability.
18. The medium of claim 15, wherein the instructions further
comprise performing the instructions iteratively until it is
determined that adequate information has been requested such that
each probability of the probabilistic network is either below a
first threshold or above a second threshold.
19. The medium of claim 15, wherein information is requested using
electronic forms.
20. The medium of claim 15, wherein triggering comprises
recognizing an input of data into the EMR of the patient.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application, having attorney docket number
27098.224525, is a Continuation application of copending U.S.
patent application Ser. No. 14/039,125, filed Sep. 27, 2013,
entitled "Adaptive Medical Documentation System" which claims the
benefit of the filing date under 35 U.S.C. .sctn. 119(e) of
Provisional U.S. Patent Application Ser. No. 61/707,147, filed Sep.
28, 2012. Both of these applications are incorporated by reference
in their entirety.
FIELD OF THE INVENTION
[0002] The present embodiments relate to medical documentation
customization. Specifically, the present embodiments relate to
automatic medical documentation adaptation for predicted or
probable patient medical conditions.
BACKGROUND
[0003] Medical entities acquire and store significant amounts of
patient medical information for diagnosis and tracking purposes.
Historically, this information was acquired using paper forms,
filled out by patients or medical entity personnel. Also, medical
entity personnel would need to know specifically which forms to
provide to patients depending on the specific medical history,
current condition of the patient, and any other information that
may be relevant to medical care of the patient. Often, the
multitude of forms actually used for a given patient would request
the same information multiple times. These forms may then be stored
in a paper file, for future references by medical entity
personnel.
[0004] Electronic Medical Records (EMR) have become a standard
storage technique for medical and health records for patients of
medical practitioners and medical entities. EMRs contain a
considerable amount of medical data for specific patients, from
various sources and in various formats. Collections of EMRs for
medical facilities provide medical records and history for most, if
not all, patients in a medical entity.
[0005] The entry of data into an EMR, however, may still be a very
complex issue involving the manual selection of proper electronic
forms for particular patients. Medical facilities and medical
entities face challenges in improving the quality of care for
patients, as well as reducing costs and increasing revenue.
Efficient and effective entry of information into medical systems
and EMRs may aid in the pursuit of these goals by increasing
availability of data relevant to the care of patients.
BRIEF SUMMARY
[0006] By way of introduction, the preferred embodiments described
below include methods, computer readable media, and systems for
adaptively requesting and presenting medical information.
Indication of a condition in an electronic medical record may be
used to solicit input of additional information relevant to the
indicated condition or multiple indicated conditions. In this way,
a singular integrated or consolidated form may be presented to a
user rather than requiring entry of the same information into
different forms.
[0007] In a first aspect, medical data may be collected using an
electronic medical documentation system coupled with a collection
of Electronic Medical Records (EMRs) for patients of a medical
entity, as well as other sources and repositories of medical data.
An analysis may be triggered in response to information input into
an EMR of a patient. A potential condition for the patient may be
determined based on the analysis. Additional information indicated
as relevant to the potential condition of the patient may be
identified. A request for the identified additional information may
be generated.
[0008] In a second aspect, a system is presented for adaptive
medical data collection. The system may involve at least one memory
operable to store EMRs and other medical data relating to
conditions of patients of a medical facility. The system may also
involve a processor configured to trigger an analysis of electronic
records in response to information input into an EMR of a patient,
identify additional information indicated as relevant to the
potential condition of the patient from the other medical data, and
generate a request for the identified additional information.
[0009] In a third aspect, a non-transitory computer readable
storage medium has stored therein data representing instructions
executable by a programmed processor for adaptive medical data
collection. The storage medium includes instructions for triggering
a probabilistic network analysis of electronic records in response
to information input into an Electronic Medical Record (EMR) of a
patient, determining a potential condition for the patient based on
the probabilistic analysis, identifying additional information
indicated as relevant to the potential condition of the patient,
generating a request for the identified additional information,
receiving at least a portion of the identified additional
information, re-performing the probabilistic network analysis of
electronic records using the received at least a portion of the
identified additional information to update at least one
probability of the probabilistic network.
[0010] The present invention is defined by the following claims,
and nothing in this section should be taken as a limitation on
those claims. Further aspects and advantages of the invention are
discussed below in conjunction with the preferred embodiments and
may be later claimed independently or in combination.
BRIEF DESCRIPTION OF THE DRAWINGS
[0011] The components and the figures are not necessarily to scale,
emphasis instead being placed upon illustrating the principles of
the invention. Moreover, in the figures, like reference numerals
designate corresponding parts throughout the different views.
[0012] FIG. 1 is a flow chart diagram of one embodiment of a method
for adaptive medical data collection;
[0013] FIG. 2 is a flow chart diagram of another embodiment of a
method for adaptive medical data collection;
[0014] FIG. 3 is a block diagram of one embodiment of a system for
adaptive medical data collection; and
[0015] FIG. 4 is a representation of an electronic medical
record.
DETAILED DESCRIPTION
[0016] The collection of medical data for a patient may adapt to
information input into an electronic medical record (EMR) of a
patient. The collection adapts based on inferences and conclusions
that can be made using existing knowledge of the patient and other
clinical information sources. Information currently being input may
be combined with prior patient information and the other clinical
information sources to suggest information related to a patient
pertinent to current patient medical conditions. For example,
previous patients having data similar to a current patient may be
associated based on a distance metric relating to data of the
current and previous patients. Data pertaining to associated
previous patients may provide inferences and conclusions related to
the current patient. The inferences and conclusions indicate
suggested information specific to the current patient. A request
for the suggested information specific to the patient may be made.
The suggested information may further aid in the diagnosis or
treatment of the patient. The suggested information may also be
information related to a specific diagnosis or treatment determined
to be relevant to a patient.
[0017] Adaptive medical information intake may take the form of a
clinical documentation system. The clinical documentation system
may be able to merge predefined form sections or templates such
that information requests and presentation is not duplicated.
Information being provided in real-time by a user and patient
information extracted or accessed from previous patient EMRs may be
used to determine which templates, or what parts of templates, are
to be presented. The end result may be a real-time adaptable form
constructed of templates, or template sections, specifically
selected to suit a particular patient such that the constructed
form contains all the relevant information needed to document the
medical care of the patient. Adaptive medical information intake
may also involve adapting to a user type or role by providing
specific information related to specific roles of users accessing a
clinical documentation system. For example, a nurse may be
presented with different information than a physician. Physicians,
nurses, and patients may be associated with different types of form
sections and templates, containing specialized information relating
to the role of the user.
[0018] A clinical documentation system may be configured to react
to user input with suggestions that are sensitive to contexts
specific to the patient. For example, causes of shortness of breath
in an elderly patient may be provided based on information
identified from an EMR for the patient. The contexts involved would
indicate that causes of shortness of breath in a child would not
apply because of an age identified for the patient. Similarly, an
elderly patient presenting a problem of back pain may cause a
clinical documentations system to prompt a user to ask four
selected questions relevant to elderly patients amongst a total of
16-20 risk assessment questions relating to back pain for all
possible types of patients.
[0019] A clinical documentation system may rely on prior clinical
knowledge to support context sensitive suggestions for integrations
of sections into a form. The different sources of prior knowledge
may include ontologies to describe arbitrary contexts, clinical
information and practice settings, clinical guidelines and
workflows, prior patient EMRs, or any other source of clinical
knowledge that may be useful in determining inferences for form
assembly and creation. The use of both general and site-specific
prior knowledge and information in a clinical documentation system
increases the ability of the system to adapt and customize
functionality for different users and different medical
facilities.
[0020] In an embodiment two combined graphical models may be used
in a clinical documentation system. The first graphical model may
infer new sections to be added to a medical form document given a
set of medical terms determined from an EMR of a patient. Another
graphical model may infer new relevant medical terms or concepts
associated with sections for a form of a clinical documentation
system. The two models may also be modeled as a singular unified
graphical model.
[0021] In an embodiment, a document model, a mapping model, and a
domain model may be used. The document model outlines the structure
a document may take. For example, multiple forms, form sections,
subsections, elements, or questions may have rules associated with
their respective presentation, content, and structure, and a
document model may contain and enforce those rules. The domain
model may be operable to link terms or concepts with other terms
and concepts. The mapping model may be operable to link particular
terms and concepts with the document model. For example, a weighted
probabilistic network model may indicate links between specific
terms and document elements such as form sections. The whole of the
weighted probabilistic model may contain all possible connections
between terms and form sections. When data is input into the three
model system, inferences may be made using the three model system
such that iterative analysis of input data may present acceptable
levels of probability that proper form sections are included. For
example, data including an age and gender may be input. Some
connections in probabilistic network may be reduced to zero
probability, or dropped, based on the data. In this example, the
entry of Male may reduce the probability the patient is pregnant to
0%, and thus no sections relating to pregnancy will be included in
a final document or form. As indicated above, as more data is input
into the system, the process may work iteratively based on the
additional data. In this example, the patient may be requested for
an age, and provide information indicating that the patient is 10
years old, which in turn may reduce other probabilities in the
model such as the probability the patient has Alzheimer's disease.
The combined data may also be used to cumulatively reduce
probabilities. In this example, the information indicating the
patient is male may reduce the probability of the patient have
breast cancer to 20%, and the information indicating the patient is
10 years old may further reduce the probability connection of the
patient to breast cancer to 2%. Also, the domain model may analyze
the input information to find associated information to provide the
probabilistic network model to further update the probability
connections. The iterative nature of providing and requesting
information may continue until all probabilities are found
acceptable, or found to be stable such that the entry of further
information may not significantly affect the probabilities of the
model. When a model reaches a steady probability state, a final
document having form sections related to the remaining probable
connections may be produced based on the rules of the document
model. In an embodiment, the process may be continually iterative,
and update probabilities and/or request further information
continually as new information is input.
[0022] In an embodiment, a clinical documentation system may start
with an initial user, which may be a medical practitioner or a
patient, inputting data into a record. The data may be related to a
presenting problem of a patient. The data may be as simple as a
gender, age, or chief complaint. A set of relevant medical terms or
concepts may be extracted from a record or group of records and fed
as evidence to a first graphical model. The first graphical model
may then output the most probable sections that may be included in
a final form document given the set of relevant medical terms. Once
information is provided for the most probable sections, a set of
new relevant medical terms or concepts may be inferred by the
second graphical model. The new set of relevant medical terms may
be included with the initial set of relevant terms for use by the
first model. Form sections may be generated all at once, or as
determined applicable in real-time. If the document contains
adequate sections based on predefined criteria, a final form
document is produced. Also, in an embodiment, the document may be
continually updated with new sections as new probable sections are
identified based on input data.
[0023] FIG. 1 shows a flow chart diagram of an embodiment of a
method for adaptive medical data collection. The method is
implemented by a computerized physician order entry (CPOE) system,
an automated workflow system, a review station, a workstation, a
computer, a picture archiving and communication system (PACS)
station, a server, combinations thereof, or other system in a
medical facility. For example, the system or computer readable
media shown in FIG. 3 implements the method, but other systems may
be used.
[0024] Additional, different, or fewer acts may be performed. For
example, an act for optimizing performance of a task of a workflow
is provided. The method is implemented in the order shown or a
different order. For example, acts 102, 104, 106, and 108 may be
performed in parallel or repeated.
[0025] In act 102, an analysis of electronic records is triggered
in response to information input into an EMR of a patient. The
information may be input into electronic format through any method.
In an embodiment, the information may be input into an electronic
form of an EMR for a patient. In another embodiment, the
information may be freely input into a generic EMR. For example, a
user may speak into a microphone indicating a symptom. The patient
may input the language "I have a headache" into the microphone
where it is electronically transcribed to indicate a symptom. Any
automated speech recognition method, such as Hidden Markov Models,
Dynamic Time Warping, or Neural Networks, may be used.
[0026] The analysis may be triggered by any act. In an embodiment,
the analysis is triggered based on recognition of an input of data
into an EMR. Recognition of an updated field in an electronic form
or database of an EMR of a patient may also trigger the analysis.
The analysis may be performed using the information input into the
EMR that triggered the analysis.
[0027] The electronic records may be any electronic record from
which a potential diagnosis or treatment for a patient may be
inferred independently, or in combination with other electronic
records. For example, electronic records may include ontologies of
arbitrary contexts, clinical data records, practice data records,
clinical guidelines, EMRs of prior patients of a medical
entity.
[0028] In act 104, a potential condition for the patient is
determined based on the analysis. The analysis may be any analysis
of electronic records capable of determining a potential condition
for a patient. In an embodiment, the analysis involves the
application of a machine learned model to the electronic records.
For example, a Bayesian Network model trained using a Markov Chain
Monte Carlo (MCMC) method may be used. In another example, an
Expectation Maximization method based model may be used. The
machine learned model may be trained using knowledge of an expert,
or documented prior knowledge such as ontologies or medical
databases. The model relates possible inputs as a feature vector
for a patient to conditions.
[0029] A condition may involve any condition for a patient. In an
embodiment, a determined condition involves a medical condition of
the patient. For example, a condition may be heart disease,
diabetes, epilepsy, hepatitis B, an allergy, or any condition
related to the health or status of a patient. The condition is a
possible diagnosis. A given input feature vector may indicate only
one or more than one possible diagnoses and corresponding
conditions.
[0030] In an embodiment, a probability that a patient has a
condition is determined. The determined probability may be compared
to a probability threshold, and if the probability meets a
threshold it is determined that the condition applies to a patient.
For example, a probability threshold may be 75% probability that a
condition applies to a patient. Any determined probability larger
than 75% may indicate that the condition applies to a patient. Any
probability scoring system may be used. The machine learnt model
may be probabilistic, so outputs a probability associated with one
or more conditions.
[0031] In an embodiment, a probabilistic network may be used to
determine possible conditions for a patient. The probabilistic
network may have connections between terms or concepts and
associated conditions represented by probabilities indicating that
a current patient may have a condition. The probabilities may
represent a current state of knowledge or data for a patient, and
may be updated with inputs of additional information.
[0032] Multiple conditions may be determined to apply to a patient.
A given model may provide more than one condition. Alternatively,
different models, such as models specific to one or more conditions
are applied to the data for the patient. Different models test for
different conditions.
[0033] In act 106, additional information indicated as relevant to
the potential condition of the patient is identified. The
additional information may be associated with a condition
determined for a patient, and identified upon determining that a
condition applies to a patient. The additional information may
involve any information relevant to determine a diagnosis for a
patient. In an embodiment, the information may be determined to
provide further indication that a condition applies to a patient.
For example, if a patient indicates that they have a headache,
other information such as history of headaches or recent physical
injuries may be identified as relevant to a potential condition of
persistent migraines, or post-concussion syndrome may be determined
as potential conditions for a patient. All possible additional
information may be stored in a collection, and relevant information
may be identified from the collection. Additional information may
be indicated as relevant through the application of a model, such
as a machine learned graphical probabilistic model, as described
herein.
[0034] Information may be configured as individual fields in a
database associated with conditions, or collections of fields
associated with conditions. When a condition is determined, or a
possible condition is identified, fields associated with the
condition may be identified. For example, a possible condition may
be indicated with a probability of 73%, and all fields associated
with the condition may be identified. Contrarily, if a probability
of a condition is below a certain probability, such as below 35%,
the fields associated with the condition may not be selected.
[0035] In an embodiment, additional information may be grouped as
related to various conditions. For example, fields for the
additional information may be assembled or contained in
preformatted form templates. The form templates may each be
associated with at least one condition. Once a condition is
indicated, an associated form template may be identified. Multiple
form templates may also be identified.
[0036] In an embodiment, additional information may be identified
as individual fields for the additional information, wherein the
individual fields are associated with a condition. All fields
determined to be associated with a condition above a certain
probability may be identified.
[0037] In act 108, a request for the identified additional
information is generated. The request may involve selecting an
established collection of information associated with the
determined potential condition of the patient.
[0038] In an embodiment, the established collection of information
is a preformatted medical form or form section. The form is
electronic and may be a collection of fields associated with an EMR
of a patient. The fields may be used to record and store the
information associated with a condition. The fields may be clear of
data, and configured for inputting information into an EMR, or
other database. In an embodiment, the fields may have standardized
information contained in them relating to a condition. For example,
the standardized information may indicate that a certain dosage of
acetaminophen is a typical treatment for a patient with a headache
condition.
[0039] In an embodiment, some of the fields may be pre-populated
with information for the patient pulled or mined from other
portions of the EMR. For example, an age of the patient may be
determined from other areas or individual records of an EMR, and
the age may be pre-populated in the form. In another embodiment,
the fields having information mined from an EMR will be withheld
from presentation with the form. The information, however, may
still be noted and associated with the condition for the patient.
For example, a form may not indicate an age for a patient, but an
age is determined, and a dosage for a medication may be determined
based on the condition and the mined age of the patient. Any data
mining may be used, such as is disclosed in U.S. Pat. No.
7,617,078, the disclosure of which is incorporated herein by
reference.
[0040] In an embodiment, a collection of form templates or sections
may be stored, each associated with a particular condition and
containing fields for information relating to the condition. Form
templates relating to conditions determined for a patient may be
selected and presented on a display as part of generating a request
for identified information relating to the conditions. The form
templates may be presented in a pre-formatted whole, containing
fields for all identified information relating to a condition. The
form templates may also be presented in part, displaying only
fields relating to information identified as most critical to
diagnosis or treatment of the condition.
[0041] Multiple form templates for different conditions may be
displayed at once as a singular form, or separately. If different
forms include requests for the same information, the forms may be
merged to provide one request rather than duplicative entry.
[0042] In an embodiment, condition determination may be iterative.
A portion of the identified additional information may be received.
Other information regarding the potential condition of the patient
may be identified based on the received identified additional
information. The analysis may be re-performed using the other
information to identify further information relevant to the
potential condition of the patient, and a second request for the
further information may be generated. Numerous inputs of
information and requests may be generated during an iterative
process of adaptive medical information input to determine a
diagnosis or treatment of a condition.
[0043] In an embodiment involving machine learned models, the
performing and re-performing an analysis may be undertaken by
applying a first machine learned model and the identifying other
information may involve the application of a second machine learned
model. The second machine learned model may receive the identified
additional information and generate the other information. The
other information may be additional terms or medical
characteristics that may be added to the analysis of the first
machine learned model to provide higher accuracy for determining
information, or forms, pertinent to the patient. As such, there may
exist a form hierarchy wherein a general condition, such as head
pain, has a general form template designed to gather further
information to further diagnose and narrow the head pain condition.
For example, the request for information may involve a query for
information to determine if a patient has a concussion, or if the
patient suffers from chronic migraines, each of which may have
respective forms in the hierarchy, with fields for specific
information relating to each condition. The second machine learned
model may be designed to take the input from the initial request,
and combine the information with other information to further
output a potential diagnosis for a condition that may be input into
the first model to identify further forms.
[0044] FIG. 2 shows a flow chart diagram of an embodiment of
adaptive medical data collection. The diagram may describe the
operation of a system, such as that described with respect to FIG.
3 below, or another structure operably consistent with the
diagramed components.
[0045] An analysis of electronic records using a model 202 may be
triggered in response to information 212 input into an EMR of a
patient. The model 202 may determine a potential condition for the
patient based on the analysis. The model 202 may identify
additional information contained in form sections 204 indicated as
relevant to the potential condition of the patient. A request for
the identified additional information may be generated by selecting
a medical form or form section 204. The form or form section 204
may be determined relevant and selected based on a probability a
patient has a condition associated with the form determined by the
model 202. Form sections 204 may be entire forms, or sections of
forms designated for certain conditions. The relevant form sections
204 may be displayed for data presentation or data input.
[0046] Adaptive medical data collection may be iterative. An
iterative embodiment may involve receiving at least a portion of
the identified information. For example, additional patient
information 214 may be input into some, or all, of the fields of
relevant form sections 204. Using a second model 208, other
information or additional terms 210 may be identified or determined
regarding a potential condition of a patient based on the received
identified information. The analysis may be re-performed by a first
model 202 using the additional terms 210 to identify further
relevant form sections 204.
[0047] Iterations may continue until is determined in act 206 that
there is enough, or adequate, requested or presented information,
or whether there have been adequate form sections identified,
generated, and/or presented. The determination may be made based on
predefined criteria. In an embodiment, a minimum number of form
sections may be required. In an embodiment, an adequate section
determination 206 may be made based on probabilities of a condition
for a patient. For example, form templates associated with
conditions determined beyond a threshold may be provided. Further,
conditions within a range of probabilities may require iterations
to better establish a likelihood the patient has the condition. The
model 208 may refine the information contained by providing
additional terms to use with the model 202 to select further
relevant form sections 204. For example, iterations may be provided
for conditions determined with a probability of 45% to 75%, where
75% may be the probability threshold. Iterations may continue until
all probabilities determined for all conditions are either below
45%, or above 75%.
[0048] In an embodiment, when it is determined that there are
adequate form sections, a document builder 216 may be used to
create a total collection of forms/form sections for generation or
presentation. The document builder 216 may also be included in a
non-iterative embodiment, after relevant form sections 204 have
been determined by the model 202. The collection of form sections
218 may be presented in an order according to a set of ordering or
ranking rules. In an embodiment, form sections may be ranked by
probability of a patient having the condition associated with the
form section, with the highest probabilities being placed most
prominently in a collection of form sections 218. Examples of
prominent placements of form sections 204 may include being placed
at the top of a collection 218, displayed with highlighted or more
noticeable text than other form sections of the collection 218, or
a form section may require input prior to inputs of other
sections.
[0049] In an embodiment, the models 202 and 208 may be one unified
model. In an embodiment, the models may be machine learning models.
The models 202 and 208 may be trained based on a collection of
prior medical knowledge to represent and efficiently manipulate a
probability distribution of conditions for a patient associated
with document sections 206. Document sections 206 associated with a
condition determined to a certain probability to apply to a patient
may be included in a personalized main document 218. The main
document 218 may group relevant subsections 206 that contain
information needed to be registered for a given patient in a given
clinical visit. The relevant subsections 206 included in a main
document 218 are associated with conditions having a probability of
relating to a patient. The probability of relating to a patient may
be modeled using the machine learned model 202, which may be a
generative probabilistic model such as a Bayesian Network model.
The generative probabilistic model may represent relationships
among medical concepts or terms and document sections such as
term-term relations, sections-term relations, sections-terms
relations, and section-section relations.
[0050] Probabilistic graphical models are graph-based
representations for encoding a distribution over multi-dimensional
space, wherein each node in a graph represents a random variable.
Links between nodes specify a direction or relevance of an
association. The edges of the graph each have an associated real
number usually referred to as an exponential family weight. A
positive link weight between two nodes means that an increase or
decrease in the value of node 1 causes an increase or decrease,
respectively, in a value for linked node 2. A negative link weight
indicates a decrease value for node 1 increases the value for node
2 or vice versa. The absolute value of the weights is a measure of
strength of influence by any parent node on a child, or linked,
node. A node in a graphical model may encode either discrete or
continuous probability distributions.
[0051] Graphical models for adaptive medical information input may
be trained in two steps. The first step involves learning or
designing the structure of the network. The first step may be
performed by an expert in the knowledge of the medical area and
form constructs being graphed, by a prior form knowledge structure,
or automatically through a learning algorithm such as a Markov
Chain Monte Carlo (MCMC) local search method. For example, an
expert may recognize that a particular form may be associated with
a particular condition. An expert may also recognize that the
information in one form is related to information in another form.
These associations may be recorded and integrated to form the
structure of the network. The first step may also involve a hybrid
creation which may consist of applying an automatic algorithm first
and later modifying the resulting network using known relations, an
expert, or prior knowledge.
[0052] A second step in training a graphical model for adaptive
medical information input may involve learning the parameters of a
network including unrealized relationships between conditions and
forms, or strength of associations between forms and conditions. In
an embodiment, an Expectation Maximization search algorithm may be
used. Such an embodiment may alternate between solving two
problems, an expectation and a maximization, to compute maximum
likelihood estimates of parameters for the model. The algorithm may
start with random initializations of model parameters, and converge
onto optimal point estimates, resulting in a network of nodes
relating to conditions and associated form sections.
[0053] Once a model is trained, a set of evidences may indicate a
probability that a given section should be included in a current
personalized collection of forms 218. The evidences may include
information such as patient characteristics, complaints, or
sections already included in a document. Probabilities may be
determined by a model using any method. In an embodiment, a
Junction Tree algorithm is used.
[0054] FIG. 3 shows a system for adaptive medical data collection.
The system is a server, network, workstation, computer, database,
or combinations thereof. The system 10 includes a processor 12, a
memory 14, and a display 16. Additional, different, or fewer
components may be provided. For example, the system includes a
scanner, a network connection, a wireless transceiver or other
device for receiving patient information and/or communicating
patient information to other systems.
[0055] The memory 14 is a buffer, cache, RAM, removable media, hard
drive, magnetic, optical, database, or other now known or later
developed memory. The memory 14 is a single device or group of two
or more devices. The memory 14 is shown within the system, but may
be outside or remote from other components of the system, such as a
database or PACS memory.
[0056] The memory 14 stores EMRs for patients and other medical
data relating to conditions of patients of a medical facility.
Models, such as probabilistic graphical models trained using
medical data may also be stored on the memory 14. Multiple EMRs of
other patients may also be stored on the memory 14. In an
embodiment, the memory 14 is operable to store a plurality of
electronic medical records of a plurality of patients of a medical
entity, specific electronic medical record of a patient as well as
ontologies, electronic clinical information, practice settings,
machine logs, clinical guidelines, and workflows.
[0057] The memory 14 is additionally or alternatively a
non-transitory computer readable storage medium with processing
instructions. The memory 14 stores data representing instructions
executable by the programmed processor 12 for adaptive medical
information input. The instructions for implementing the processes,
methods and/or techniques discussed herein are provided on
computer-readable storage media or memories, such as a cache,
buffer, RAM, removable media, hard drive or other computer readable
storage media. Computer readable storage media include various
types of volatile and nonvolatile storage media. The functions,
acts or tasks illustrated in the figures or described herein are
executed in response to one or more sets of instructions stored in
or on computer readable storage media. The functions, acts or tasks
are independent of the particular type of instructions set, storage
media, processor or processing strategy and may be performed by
software, hardware, integrated circuits, firmware, micro code and
the like, operating alone or in combination. Likewise, processing
strategies may include multiprocessing, multitasking, parallel
processing and the like. In one embodiment, the instructions are
stored on a removable media device for reading by local or remote
systems. In other embodiments, the instructions are stored in a
remote location for transfer through a computer network or over
telephone lines. In yet other embodiments, the instructions are
stored within a given computer, CPU, GPU, or system.
[0058] In an embodiment, the instructions may include triggering a
probabilistic network analysis of electronic records in response to
information input into an Electronic Medical Record (EMR) of a
patient, determining a potential condition for the patient based on
the probabilistic analysis, identifying additional information
indicated as relevant to the potential condition of the patient,
generating a request for the identified additional information,
receiving at least a portion of the identified additional
information, and re-performing the probabilistic network analysis
of electronic records using the received at least a portion of the
identified additional information to update at least one
probability of the probabilistic network.
[0059] The processor 12 is a server, general processor, digital
signal processor, graphics processing unit, application specific
integrated circuit, field programmable gate array, digital circuit,
analog circuit, combinations thereof, or other now known or later
developed device for medical category determination. The processor
12 is a single device, a plurality of devices, or a network. For
more than one device, parallel or sequential division of processing
may be used. Different devices making up the processor 12 may
perform different functions, such as a handwriting detector by one
device and a separate device for communicating or processing the
detected handwritten data. In one embodiment, the processor 12 is a
control processor or other processor of a computerized data entry
system for an EMR storage or database system. The processor 12
operates pursuant to stored instructions to perform various acts
described herein.
[0060] The processor 12 is configured by software or hardware
adaptive medical information input. The processor 12 may be
configured to trigger an analysis of electronic records stored on
the memory 14 in response to information input into an EMR of a
patient. The processor 12 may further be configured to identify
additional information indicated as relevant to the potential
condition of the patient from the other medical data. The processor
12 may also be configured to generate a request for the identified
additional information. The request may be presented on the display
16. A collection of requests may also be presented on the display
16.
[0061] The display 16 is a CRT, LCD, plasma, projector, printer, or
other output device for showing an image. The display 16 displays a
user interface with an image. The user interface may be for the
entry of information, such as information that may be used for
triggering an analysis of electronic records stored on the memory
14. The user interface may be for entering information into an EMR,
or displaying a graphical model.
[0062] FIG. 4 shows an exemplary EMR 200. Health care providers may
employ automated techniques for information storage and retrieval.
The use of an EMR to maintain patient information is one such
example. As shown in FIG. 4, an exemplary EMR 200 includes
information collected over the course of a patient's treatment or
use of an institution. The information may be collected using
forms, form templates, form sections, or combinations thereof. The
information may include, for example, computed tomography (CT)
images, X-ray images, laboratory test results, doctor progress
notes, details about medical procedures, prescription drug
information, radiological reports, other specialist reports,
demographic information, family history, patient information, and
billing (financial) information. Any of this information may
provide for information related to a potential condition for a
patient.
[0063] An EMR may include a plurality of data sources, each of
which typically reflects a different aspect of a patient's care.
Alternatively, the EMR is integrated into one data source.
Structured data sources, such as financial, laboratory, and
pharmacy databases, generally maintain patient information in
database tables. Information may also be stored in unstructured
data sources, such as, for example, free text, images, and
waveforms. Often, characteristics, such as key clinical findings,
are stored within unstructured physician reports, annotations on
images or other unstructured data source.
[0064] While the invention has been described above by reference to
various embodiments, it should be understood that many changes and
modifications can be made without departing from the scope of the
invention. It is therefore intended that the foregoing detailed
description be regarded as illustrative rather than limiting, and
that it be understood that it is the following claims, including
all equivalents, that are intended to define the spirit and scope
of this invention.
* * * * *