The Challenge
The pairing algorithm caused much frustration with providers, patients, and our internal teams.
- Providers are frustrated they only have a brief window to claim a patient before said patient is seemingly passed to another provider.
- Providers are penalized if they don't respond to a patient request within a brief window, even if they were in an appointment when the request came through.
- Matching for provider location only accounts for distance, which doesn't work well for high-traffic areas like LA.
- Providers are frustrated that patients are not being evenly distributed amongst them.
- Providers are frustrated there isn't a way to customize their boundary(ies).
- Some states require state licensure to provide care within them.
- Patients who request a new provider can be repaired with the same provider within the existing method.
- Internal teams are struggling with manually matching patients when the automated process doesn't succeed in a timely manner.
The goal? Develop a pairing algorithm that accounts for inputs specific to visit modalities and insurance eligibility while fairly balancing the needs of providers and patients.
My Role
- Synthesized an investigation of the current state pairing algorithm to share it out with stakeholders.
- Led focus groups with providers to understand provider pain points with current state.
- Documented the product requirements and developed the logic for the proposed algorithm.
- Presented the proposal using mixed media to executives and other stakeholders to obtain feedback and buy-in.
- Groomed and prioritized the work with the engineering team using an agile framework.
The Outcome
Due to financial termoil at the company, this work was never delivered. It was my brain child and I will forever wonder the impact it had. :(
The Solution
Pairing Algorithm Business Service
- When a patient match request comes in, the network of providers and their respective practices should be filtered for only providers/practices that meet a patient's specific conditions outlined. [REDACTED] Additionally, providers who have previously been matched with a patient should be black-listed upon the patient request a rematch.
- Each provider/practice combo should be assigned a composite score based on the inputs outlined. [REDACTED]
For instance, if an is part of multiple practices, each provider/practice combo should be assigned a separate composite score.
- The contending practices should be queued or pooled for the patient, given the opportunity to accept/reject the patient.
- Pooling for emergent patients: [REDACTED]
- Queue for non-emergent patients: [REDACTED]
- Pooling for idled patients: [REDACTED]
- The conditions should be revisited and composite scores adjusted to account for any changed patient or provider preferences, as well as newly contracted providers/practices. [REDACTED]
UI for patients to provide preferences
- Patients can filter on provider-set distinguishers per practice.
- A patient can update their visit modality preferences and their urgency preference.
- Patients who have indicated an emergent preference are unable to view a list of potential providers.
- Patients who have indicated a non-emergent preference can view and choose from a list of potential providers.
UI for Providers to provide in-home boundary
- Practice owners will have the ability to set a single boundary for their entire acie
- Providers within a practice will have the ability to provide their state credentialing, which will likely be completed as part of a separate initiative prior to the implementation of this work.
UX for practices to accept/reject patients
- Practices will be notified to accepted / reject a patient when they are first in line for a patient's queue.
- Practices will be made aware when they are no longer in a patient's queue due to their allotted window passing without excessive communications.
- Practices will be notified to accepted / reject a patient when they are in a patient's pool, with transparency behind the reason for the pool / urgency.
- Practices will be made aware when a patient's pool has closed without excessive communications.
- Practices will be required to log into their account (thusly completing MFA) to accept/reject patients.
Internal UI for internal teams to manually match a patient
Within the internal tool, support agents will have:
- The ability to see patients who have idled.
- The ability to see patients who have had a delayed pairing.
- The ability to see all contending practices for a patient based on inputs.
- The ability to manually assign a contending practice to a patient.
- The ability to manually archive a patient (due to no contending provider based on preferences).
- The ability to filter practices for a patient.
- The ability to see prior filters submitted by a patient.
- The ability to see practices that have been notified of, accepted, or rejected a given patient.
- The ability to set filters / see a list of contending practices for patients regardless of their submission request status.
- The ability to see prior filters submitted by a patient.
- The ability to see practices that have been notified of, accepted, or rejected a given patient.
- The ability to set filters / see a list of contending practices for patients regardless of their submission request status.
Patient Comms
- Patients who have experienced an eligibility delay will be notified of the delay and how it impacts the pairing timeline.
- Patients who have been in a pool for 2 hours without a match will be alerted and guided to update their previously-provided preferences, specifically asking if they are interested in telehealth.
- Patients who do not have a full pool of contending practices will be alerted and guided to update their previously-provided preferences, specifically asking if they are interested in telehealth.
- Patients who do not have a match will be alerted and guided to update their previously-provided preferences, specifically asking if they are interested in telehealth.
- Patients who have been matched to a practice will be alerted via a communication + within myTLN.Patients who are archived will be alerted.
Pairing Reporting
[REDACTED]
The Output
.png)
