Notes:
There is overlap with the interoperability side
Can we say that a Patient is better off at another care center?
How often are similar referrals accepted or rejected?
How often is this type of patient accepted or rejected?
"you typically accept referrals like this"
regionally the avg reimbursement for a patient like this is...
a patient like this typically has a 75 day length of stay
requires pt and ot
at least 3 visits a week.
Financial and/or Demographic/Diagnosis based
Can we get reimbursement data?
historical dump?
We could focus on the financial data
avg length of stay?
Branch Outlier detection
turnaround time - looking at what branches to improve.
what would bringing those branches to the AVG do to the overall integration.
Patient Medication or disease Timeline, Word clouds, other visualizations.
Predicting time it will take a specific fax to get through various phases (Inline)
Mobile - providing optimal route planning. - big project.
Mobile - Outlier detection on length of time at a patient's home.
James is taking 25% longer than Jan for similar types of appointments for similar patients
Utilize Patient2Vec
This would be hard to pin down if we didn't factor in the patient. One patient could be older and have more comorbidities.
James could be taking 25% longer to treat wounds on a patient with 10 of them who is also immobile while Jan's patients happen to be more mobile.
James could also be taking 25% longer on patients that are almost exactly the same as Jan's
Intelligently routing different types of orders/documents/referrals to a specific User.
Secure Messenger - sentiment analysis, diseases, medications, etc mentioned in messages.
For patient conversations we could aggregate all of the various diseases, medications, and therapies that are found.
They could flow into the patient chart.
Could be a way for them to capture notes without needing to answer specific questions.
Automated testing
Semi-automated model deployment pipelines
Predicting how long it'll take a physician to respond
more than just a trend analysis. We would use time of week, month, year and number currently in their queue as factors. Maybe more.
IQ: Text Extraction - More fields
payer information
race, ethnicity, religion.
physician information
visit date
timestamp vitals and what not
IQ: Patient Risk Detection inline and as a report Long Term FOCUS
Patient Self Harm Score
Patient Readmission Score
Risk of emergency transport in the next 90 days
Fall Risk
Post surgical complications
Requires a lot of UI work.
Would need some sort of dashboard eventually (or report) that organized patients by various risks or the need to visit sooner.
We have odt and adt files for patients that contain a lot of pertinent patient information.
Help physician's and payers intervene earlier.
Would this help the agency's? They could notify the physician's and payers...
Requires a lot of UI work.
UI could list what we've pulled and allow the user to tab through the results down to their locations
Documents
Provide Analytics on
top Referral providers
top Lab Reports
Allows us to target training using clustering algorithms that tell us what documents to train on without direct client input.
Patient
We can build a deep understanding of patients that is "longitudinal"
Being longitudinal refers to the ability to create an embedding that captures the patient's changes over time.
Can be built from an EHR
E-Referral Documents come from Hospitals mainly, but they require the user to go to a separate website to download and then attach within forcura.
Healing Time
Healing Risks
Types of tissue.
Infections
Connect with patients after discharge automatically via text or phone call
Use predefined questions that we convert to text and run NLP based sentiment analysis on.
Audio could also be analyzed for signs of pain, worry, depression, happiness, etc.