Who follows up when AI flags a result in a longevity clinic?
Harvey Castro, MD, contributed to Nozomi’s 2026 longevity report. His Friday afternoon example shows how a clinic can test who reviews an AI alert and who contacts the patient.
It’s Friday evening. Imagine opening your patient portal and seeing that your estimated biological age has gone up four years. You feel the same as you did yesterday. The portal gives you a number, but it leaves you guessing whether your physician has seen it and whether you should do anything this weekend.
I contributed to five expert commentaries in Nozomi’s 2026 report, The State of Longevity. Two of them deal with that moment between an AI alert and a conversation with a patient. My view is that a longevity clinic should name the clinician responsible for reviewing a flagged result before it puts that result in anyone’s portal.
More than half saw their results before hearing from a clinician
How U.S. patients received test results
Viewed test results in their patient portal70%
Saw results before hearing from a clinician58%
Were offered a choice about seeing results before clinician contact28%
0%50%100%
Weighted estimates from the 2024 U.S. Health Information National Trends Survey. The study included 6,045 patients across care settings. The same person could report several experiences. Nozomi cites the study on printed page 144. Read the original study by Richwine, Steitz and Everson.
When patients saw the result first, how many understood it?
66 in 100reported understanding their result and what it meant for their care.
This question applied to the 3,432 people who saw their result before hearing from a clinician. The filled dots show the survey’s weighted estimate of 65.9%, rounded to the nearest whole percent. See Table 5 in the original study. Nozomi discusses it on printed page 144.
Trace one alert through a Friday afternoon
If I were reviewing a proposed AI tool with a clinic, I’d ask the people who answer patient messages to show me how they would handle this example.
Follow the Friday alert until someone closes it
The result arrives
Clinical team
Review the measurements behind the estimate and decide whether the change needs attention.
Patient experience
The new estimate appears in the portal. It may arrive before an explanation.
The shift ends
Clinical team
The outgoing clinician names the colleague taking over and shows them the open alert.
Patient experience
The patient may still be waiting. The handoff should preserve the promised time for an update.
The clinician calls
Clinical team
Explain what the estimate means, what you recommend and when the patient will hear from you again.
Patient experience
The patient can ask questions and knows whom to contact if the update is late.
The team closes the alert
Clinical team
Check whether the agreed test or referral happened. If no action is needed, document why.
Patient experience
The patient hears the outcome, including why further testing may be unnecessary.
Illustrative clinic workflow adapted from the distinction between a result received, a patient informed and a finding closed on printed page 136 of the Nozomi report.
I’d use that walk through before buying a tool, then repeat it with the team after launch. If the alert can disappear at a shift change, the clinic has found something it can fix before the next patient spends a weekend wondering.
Four studies behind the patient journey
Adoption · Report page 133
Portal use looked different after rollout
Routine deployments · 16 studies
23% estimate · 95% confidence interval 13% to 33%
Actively recruited participants · 24 studies
71% estimate · 95% confidence interval 64% to 79%
0%50%100%
A review of 40 studies found a 48 percentage point difference in average adoption. The study settings and recruitment methods differed. For a clinic planning a rollout, the useful comparison is with patients who were offered the portal in ordinary practice.
Dots show pooled estimates. Lines show their 95% confidence intervals. These historical estimates come from a 2017 review. Read the portal adoption review.
Enrollment · Report page 133
A timely invitation brought more patients into the portal
3.9%Enrolled within a week with no email
10.8%Enrolled within a week after a results email
That is about seven additional enrollments per 100 people offered an email. Patients had a reason to sign in that day. Their lab result was ready. Whether a patient understood the result was a separate question.
Randomized trial of 5,009 adults in one health system. The email figure combines the two invitation groups. Read the randomized trial.
Screening · Report page 136
An unexpected MRI finding can open a second care journey
Critical or indeterminate incidental findings
32.1% estimate · 95% confidence interval 18.3% to 50.1%
0%50%100%
Across 12 studies of 5,373 people without symptoms, the pooled estimate was 32.1%. The wide interval reflects uncertainty, and findings varied substantially across studies. A flagged finding still needs interpretation. Who will explain it and arrange follow up if needed?
The review measured findings on whole body MRI. It did not establish how many people benefited from screening. Read the MRI review.
Continuity · Report page 155
A six month check can miss who is still engaged a year later
68.3%Still enrolled and active at six months
45.9%Still enrolled and active at twelve months
In this employer sponsored weight management program, retention meant staying enrolled and taking part in coaching calls. The same measure fell by 22.4 percentage points. For a longevity clinic, the useful lesson is to define participation and check it again over time.
Retrospective analysis of one commercial program from 2013 to 2016. These rates describe that program. Read the retention study.
Find your way through the report
Choose a starting point in the full report
For a clinician or patient advocate
Start with the patient journey on page 129. Turn to pages 140 to 145 for dashboard design, complex results and information overload.
For a clinic operator
Start with business models on page 51 and unit economics on page 74. Page 222 takes up the staffing and workflow questions that come with growth.
For a product team
Start with technology on page 195. Pages 199 to 205 connect a fragmented patient record, wearable measurements and what a patient sees in the portal.
For a reader following the industry
Start with the market on page 8, company comparisons on page 262 and the report’s outlook on page 279.
All references above use the report’s printed pages. Add two when entering a page number in your PDF viewer.
Five places to find my commentary
Select a question to see the idea and its page in the PDF viewer.
What makes a repeat visit worthwhile? Printed page 63
Use changes in a patient’s results to explain why another test or visit would help.
PDF viewer page 65What is a forecast worth to a patient? Printed page 83
A forecast earns its place when a patient and clinician can use it to make a decision together.
PDF viewer page 85How many alerts can a team actually review? Printed page 230
Have the staff who manage the inbox help decide which alerts deserve their attention.
PDF viewer page 232What happens between annual visits? Printed page 286
Repeated measurements can connect visits if someone is responsible for reviewing a change.
PDF viewer page 288Who owns a decision informed by AI? Printed page 291
Check the evidence behind the estimate, explain uncertainty to patients and name the person responsible for the decision.
