AI applications to reduce medical cost

PROVIDER NETWORK · LATIN AMERICA

Pathway gaps found one patient at a time before the payer measures coverage.

64%
of patients due pathway care had an open activity
28%
of open activities had evidence already on file
1 in 9
orders repeated a still-current test or procedure

Background

A network that answers for the cost and the coverage of its population.

This provider network delivers primary and preventive care under a prospective global payment with a single payer. The cost of care sits on the network’s side of the contract.

The payer scores the network on its compliance with the national preventive care pathway, and that score sets a payment discount and the network’s ranking against peer networks.

Challenge

Aggregate dashboards left the per-patient question open.

The preventive pathway assigns each patient a set of activities by age, sex, and condition: consultations, laboratory tests, vaccinations, and procedures, each at a defined interval. Clinical guidelines add a second requirement: episodes of common conditions follow a stepped protocol that places conservative management ahead of imaging and referral. The network tracked the first requirement in spreadsheets and dashboards built on encounter data from its legacy systems, with laboratory results held in a separate system. The second requirement lived inside each clinician’s judgment. The finest grain the team could see was a percentage per activity.

That view leaked in five ways. Patients with pending activities surfaced only when the payer’s validator measured coverage. Activities performed outside protocol parameters, a screening repeated ahead of schedule or done outside its age range, brought zero credit after the network had paid for them. Tests and procedures ordered for a chronic condition could have credited the pathway and were ordered again, because the records sat in separate systems. Clinicians with 20 minutes per consult reordered what the patient already had. And episodes such as an uncomplicated back pain went to imaging ahead of the conservative phase, at a cost the guideline would have deferred. Two of the five leaks hit the payer score. The other three landed on medical cost.

Solution

The full history, read continuously against both rule sets.

Falcon’s agents read the visit records, laboratory results, and procedure reports the network’s systems already produce. Each night they rebuild one ledger per patient and evaluate it on two axes: coverage, which pathway activities are due and what evidence exists for each; and sequence, whether each episode followed its stepped protocol. By morning the worklists refresh: a credit queue for the reporting team, a reschedule queue for activities performed outside parameters, an outreach queue per site that leads with patients who already have a visit scheduled, and a pre-visit summary of what each patient already has. Episodes that escalated ahead of the guideline route to medical direction. When the clinical team corrects a finding, the next run carries the correction into every ledger.

The engagement began with a retrospective: a sample of active patients, three to five years of history each, evaluated end to end and reviewed by the network’s clinical team before the engine ran across the full population.

Mechanism

How the nightly run turns records into worklists.

VISIT RECORDS LAB RESULTS PROCEDURE RECORDS PATIENT HISTORY PATHWAY RULES CLINICAL GUIDELINES LEDGER CHECK coverage · sequence PATIENT LEDGER rebuilt nightly PRE-VISIT SUMMARY CORRECTIONS CLINICAL REVIEW CREDIT ON FILE RESCHEDULE OUTREACH EPISODE REVIEW VISIT RECORDS LAB RESULTS PROCEDURES HISTORY PATHWAY RULES GUIDELINES CHECK LEDGER REVIEW CREDIT RESCHEDULE OUTREACH EPISODES

Fig. 01 · One ledger per patient, checked for coverage and for sequence

Results

01
Care delivered off schedule and out of range
8% of executed pathway activities fell outside protocol parameters and earned zero credit. Scheduling inside the parameters earns credit for spend the network was making anyway.
02
Episodes escalated ahead of the guideline
1 in 4 episodes of conditions with a stepped protocol skipped the conservative phase and went straight to imaging or a specialist. Medical direction sees the pattern per clinician and per site.
03
A sample reviewed line by line
The network’s clinical team confirmed 94% of the findings on the retrospective sample, the gate set before the engine ran across the full population.

Next steps

Start reducing medical cost through contracts,
authorizations and bills.

Get a demo

See how Falcon helps healthcare organizations reduce medical cost.

Thank you. We will be in touch at the email you shared.