Reference outcomes
Published results from other teams in bio, pharma, and health. These aren’t our engagements: they’re the kinds of outcomes we aim our work at, and the constraints we build inside.
Time to therapy from 12.2 days to 3.7
For: VP, Patient Services
The reference
A specialty-drug support program moved prescriptions straight from the EHR into one digital intake: e-consent, benefits verification, prior authorization, affordability support, and specialty-pharmacy referral. The hub vendor (AssistRx) reports time to therapy fell from 12.2 to 3.7 days. Vendor-reported; study design and sample size not published.
AssistRx progress report, 2024
What made it hard
Every step waits on someone else: payers, specialty pharmacies, prescribers. All of it carries protected patient data across organizations.
Where we aim
Shorten prescription-to-therapy time across consent, coverage, assistance, and pharmacy referral, with every handoff instrumented so delays have an owner.
Prior-auth denials from 7.6% to 2.6%
For: VP, Market Access or revenue cycle leader
The reference
At Northwell Health, software read radiation-oncology treatment orders, checked them against each payer’s criteria, flagged missing documentation before submission, and tracked status. Across 6,551 cases, denials fell from 7.6% to 2.6% and authorization time from 4.2 to 2.8 business days. Nonrandomized quality-improvement study at one health system, with vendor coauthors.
JAMA Network Open, 2025 · CMS-0057-F prior authorization rule
What made it hard
Payer rules differ and change, the criteria live in PDFs, and the work has to sit inside the ordering workflow in the EHR. CMS’s new rule is now forcing payer-side APIs.
Where we aim
Cut avoidable denials and days waiting on decisions by putting payer requirements into the ordering workflow, not into a fax queue.
Five more months of median survival
For: Head of Patient Services or oncology service-line leader
The reference
At Memorial Sloan Kettering, patients on chemotherapy reported symptoms online between visits. Severe or worsening reports alerted nurses, and summaries went to oncologists. In a randomized trial of 766 patients, median overall survival was 31.2 months with monitoring versus 26.0 with usual care. Single center; survival was not the trial’s prespecified primary outcome.
What made it hard
An alert is only worth the response behind it: someone has to see it, act on it, and document it, around the clock, with patient-reported data protected.
Where we aim
Turn patient-reported symptoms into timely, reviewable clinical action between visits, with response times you can measure.
23% less time per adverse-event follow-up
For: Head of Drug Safety
The reference
Bayer France built an internal generative-AI tool that drafts follow-up requests for missing information on adverse-event reports. Safety staff review every letter before it goes out and record the case in the safety database. In a 122-case crossover study, handling time fell from 22.3 to 17.0 minutes per case, about 23% after adjustment, with no change in reply rates. One team; cases not randomized.
Journal of Medical Internet Research, 2025
What made it hard
Pharmacovigilance is audited: good-practice rules, reporter privacy, and a traceable record of who wrote and approved what. The model drafts; people decide.
Where we aim
Reduce case-handling effort with private AI that keeps a human reviewer and an audit trail on every case.
98.5% monthly survey completion, fully remote
For: VP, Clinical Operations
The reference
Cooper University Health Care ran a randomized trial almost entirely through automated workflows: remote invitations, e-consent, study-product shipping, monthly surveys, reminders, and safety alerts. They screened 406 people, enrolled 299, and averaged 98.5% monthly survey completion over nine months. An operational account, with manual follow-up and no comparison against a traditional trial.
Journal of Clinical and Translational Science, 2023 · FDA e-consent guidance
What made it hard
Consent records, participant privacy, and, in FDA-regulated trials, 21 CFR Part 11 rules for electronic records and signatures.
Where we aim
Make remote consent, follow-up, and safety monitoring easy to run, with complete, traceable data and less load on sites.
241 new diagnoses from data already on file
For: Head of Precision Medicine or CIO
The reference
Australian Genomics researchers built software (Talos) that repeatedly re-analyzes inconclusive genomic results as gene and variant evidence changes, then routes candidates to experts and an accredited lab for confirmation. Across 4,735 undiagnosed people it surfaced 241 additional diagnoses, a 5.1% yield. Observational; yield depends on prior workup, and the software never issues a diagnosis itself.
What made it hard
Genomic data and consent are sensitive, and every finding needs validated clinical interpretation before it reaches a patient.
Where we aim
Make new evidence actionable on data you already hold, through repeatable reanalysis that ends in accredited human review.