Inside the Loop — Dan Pettus

Data

US healthcare spend: $5.3T — 18% of GDP US per capita: $14,885 · peer avg: $7,860 Japan: $5,300 per capita — life expectancy 84.1 yrs US life expectancy: 79.0 yrs — lowest among wealthy nations 250,000–440,000 Americans die annually from preventable hospital errors EMR adoption 12%→84% (2009–2015) — adverse event reduction: 7% ML diversion detection false positive rate: 77% Reasoning-layer false positive rate: under 8% US healthcare spend: $5.3T — 18% of GDP US per capita: $14,885 · peer avg: $7,860 Japan: $5,300 per capita — life expectancy 84.1 yrs US life expectancy: 79.0 yrs — lowest among wealthy nations 250,000–440,000 Americans die annually from preventable hospital errors EMR adoption 12%→84% (2009–2015) — adverse event reduction: 7% ML diversion detection false positive rate: 77% Reasoning-layer false positive rate: under 8%

Where it started

Duke University Medical Center · June 1989

It was a Saturday morning. I was in the cafeteria when a senior anesthesia resident sat down looking like he hadn't slept. His job every weekend was to manually calculate how much time every third-year resident had spent in each surgical specialty over the past week. Pull the paper records. Find the start and stop times. Identify which residents were on which cases. Tally the hours by specialty. Do it again next Saturday. All year.

I told him I thought that information was already sitting in our ARKIVE database — an anesthesia information management system I'd helped build that captured data automatically from OR monitors, ventilators, and IV pumps. We wrote a query together in Paradox. It ran in about ten seconds.

He looked at the screen. Then he looked at me.

"You gave me my weekends back. I truly owe you."

Senior Anesthesia Resident, Duke University Medical Center — 1989

He didn't owe me anything. I'd spent maybe an hour. But that hour taught me something I've spent the next 36 years trying to apply at scale: the most valuable use of clinical data is rarely the one it was collected for. The data was already there. It just needed someone to ask it the right question.

That's what this newsletter is about. The questions nobody thought to ask. The tools that finally exist to answer them. And what gets in the way — the financial incentives, the regulatory friction, the organizational inertia — when the answer is sitting right there in the database.

Dan Pettus

50 years in medical device & health IT · Alaris, CareFusion, BD · iMetrikus co-founder · IHE PCD contributor · 2× patent holder

440,000

Americans die annually from preventable hospital errors — Johns Hopkins School of Medicine

The FDA would shut down any medical device that killed 440,000 Americans a year. Preventable hospital errors do exactly that. We give hospitals a quality improvement budget.

The number isn't new. What's new is that the architectural answer finally exists — a reasoning layer that reads the full clinical story in real time, across every data source simultaneously, instead of scoring patterns in a single silo.

The investment case, the clinical case, and the operational case all point the same direction. Inside the Loop is where that argument gets made — issue by issue, with the data to back it.

Editorial pillars

What you'll read. What you won't find anywhere else.

The healthcare AI conversation is dominated by vendors selling platforms and academics publishing studies. Neither group has spent fifty years watching the same problems not get solved. Inside the Loop comes from inside that experience.

Pattern

The Gap Between Promise and Delivery

EMR, BCMA, interoperability, Meaningful Use — what each promised, what each delivered, and what was structurally impossible from the start.

Application

AI Applied to Real Clinical Problems

Not AI in the abstract. What happens when a reasoning model is pointed at medication diversion, readmission prediction, or OR scheduling — specifically.

Decision

The Executive Decision Problem

How hospital CEOs and med-tech boards are evaluating AI investments right now. The right questions. The costly wrong ones. The financial incentives in opposition.

Standards

Regulation, Standards, and the AI Layer

IHE, FDA, DEA scheduling, HIPAA in the AI context — written by someone who contributed to the standards, not someone who read about them.

Investment

The Capital Allocation Question

Where the smart money is moving in healthcare AI — and where it's being wasted. Written for the investor who wants an insider read, not a pitch deck.

Story

From the Inside — Personal Accounts

First-person accounts from fifty years of building, deploying, and watching healthcare technology succeed and fail in ways the press never covered.

The author

The view from inside the room.

I spent nearly fifty years in medical device and health IT leadership — long enough to have built some of the systems I'm now critiquing, and to understand exactly why they fell short.

I'm not writing from a think tank or a VC portfolio. I'm writing because I've watched $5.3 trillion in annual healthcare spend fail to close the gap with countries that spend half as much — and because I believe the architectural answer finally exists.

A former colleague said I gave away a $250,000 investment strategy for free on LinkedIn. He's probably right. Inside the Loop is where I stop doing that.

VP-level roles at Alaris, CareFusion, and BD (Becton Dickinson)
Co-founder, iMetrikus — connected medical devices to clinical networks, 1998
Lead architect, ARKIVE anesthesia information management system, 1989
Contributor, IHE Patient Care Device interoperability standards
Two patents held; third provisional filed
AAMI BI&T Best Article of the Year, 2014
Inventor, AI MedAgent — proof-of-concept closed-loop medication architecture

Years inside medical device & health IT

1989

First clinical data query that changed how I saw everything

$5.3T

Annual US healthcare spend at 18% of GDP

−5 yrs

US life expectancy gap vs. Japan at half the per-capita cost

77%

False positive rate on best-in-class ML diversion detection

<8%

Target rate with full clinical story reasoning

Coming issues

Six issues already written in my head. Here's what's coming.

Issue 01

The Saturday morning a ten-second query changed everything

Duke, 1989. A Paradox database, a frustrated resident, and the insight that clinical data almost always contains the answer to a question nobody thought to ask.

Issue 02

Why the EMR didn't move the needle — and what we should have known

$35 billion in federal incentives. 84% hospital adoption. A 7% reduction in adverse events. The math was always going to disappoint.

Issue 03

The diversion detection problem nobody wants to say out loud

ML platforms flag 847 transactions. 653 are false positives. Nursing leadership is furious. Here's why the architecture guarantees this result.

Issue 04

iMetrikus: the right vision, the wrong decade

We connected medical devices to clinical networks in 1998 before the infrastructure existed to act on the data. What that timing teaches us about deploying AI now.

Issue 05

The one question every hospital CEO should ask before signing an AI contract

Not "does it work." Not "what's the ROI." The question that separates a point solution from a structural change — and why vendors don't want you asking it.

Issue 06

What Japan's healthcare system tells us about the AI argument

$5,300 per capita. 84.1 years of life expectancy. The gap with the US isn't genetic. It's structural. And structure is what AI can change.

What you'll read. What you won't find anywhere else.

The view from inside the room.

Six issues already written in my head. Here's what's coming.

Fifty years of pattern recognition. Yours for $10 a month.