Somebody Pulled the Cord and the Chairman of Radiology Showed Up

At Allegheny General Hospital in Pittsburgh, a nurse named Candice Bena decided a 76-year-old patient needed a new intravenous line. Radiology couldn't install it right away. She was worried about infection, so she called Richard Shannon, who was chairman of medicine.

Shannon called the chairman of radiology. Within two hours, the chairman of radiology came down and installed the line himself.

The Wall Street Journal reported it at the time. Shannon described the nurse's call as the equivalent of pulling the andon cord, and said that's what the Toyota Production System is. Shannon said more specifically:

“That's the Toyota production system. No problem should be left unsolved.”

Recently, I wrote about a recovery room nurse with no way to ask for help, and I ended by asking who would come. At Allegheny General, the answer, in that case at least, was a department chairman, within two hours.

What They Were Actually Doing

Shannon's two ICUs were averaging around 5.5 central line infections per 1,000 patient days, which Shannon described as a bit above the Pittsburgh average and a bit below the national one. Over twelve months, 37 patients had 49 infections. Slightly more than half of those patients died.

The team investigated every new infection immediately. They found that femoral lines, inserted near the groin, carried a much higher infection rate than other sites. So they moved to sites in the arm or near the collarbone.

Infections went to zero within ninety days. The 26-bed unit later went 35 months without one. Every person on the team got a daily report on whether their unit had an infection.

Side question: Has every hospital learned from that standard, adopted it, and improved it?

Allegheny responded by tightening the process. They narrowed the choice of insertion site, they reviewed every case, and they gave any staff member a way to escalate when the standard didn't fit the patient in front of them.

A standard with no way to raise an exception is just a rule. A way to raise exceptions with no standard gives nobody anything to compare against.

Twenty Years Later

Shannon is now chief medical officer and chief quality officer at Duke Health. In 2021, he founded something called CACHE, the Collaborative to Advance Clinical Health Equity, now directed by Michael Pignone.

Duke describes CACHE as looking “in the mirror of our own clinical outcomes,” analyzing that data through the lens of race and ethnicity to address root causes of disparities. The scope covers cancer screening, hypertension, diabetes, maternal morbidity, and gun violence. Duke frames CACHE as an extension of their Duke Quality System, which is built upon Lean practices.

The sequence is: pull outcomes from across the whole health system, check them for racial, ethnic, and socioeconomic differences, work with patients to understand what's driving those differences, then intervene on the causes, including causes that sit outside the hospital.

Recent CACHE projects include finding and treating chronic kidney disease, transitions of care for patients with opioid use disorder, and lung cancer screening.

The andon cord lets one nurse escalate one patient's risk. CACHE is aimed at risks that no single nurse is in a position to see or speak up about.

You Can't Find a Disparity Without a Standard

The nurse whose post started all this made a specific argument about colonoscopy. Standardized bowel preparation instructions assume an average regimen produces an average result, and that assumption fails some patients much more than others.

She's right, and it isn't close. A 2024 review in the American Journal of Gastroenterology pooled 154 studies covering more than 358,000 patients. Black race and male sex both came out as predictors of inadequate bowel preparation. So did Medicaid insurance, low education level, age over 65, and being unmarried.

Most of that list describes people with less help at home and less room in their week.

Now ask how anyone knows it.

Somebody had a standard protocol. Somebody applied it to a large number of people. Somebody measured the results and broke them out by race and sex. That's the only reason the failure has a number attached to it instead of being a vague sense that some patients don't follow directions.

Take away the protocol and there's nothing to measure against. Take away the measurement and the failure doesn't disappear, it just stops being visible. It looks like a series of unrelated individual patients who showed up unprepared, and the explanation that costs a health system the least is that those patients didn't try hard enough.

Without a standard and without measurement, the variation is still there. It just isn't recorded anywhere. And the patients who disappear into unrecorded variation are usually the ones with the least capacity to come back for a second attempt, take another day off work, or find another ride.

What This Doesn't Do

I want to be careful about how much I'm claiming.

Finding a disparity doesn't close it. But identifying and defining a problem (and it's importance and context) is the start of the problem-solving process.

I've described what CACHE says it does and what projects it has funded. I don't have published outcome data showing gaps that narrowed, and I'd want that before treating this as proof of anything. If somebody at Duke has those numbers, I'd like to see them.

The interventions are also the hard part, and most of them aren't clinical. Data science doesn't get anyone a ride to an endoscopy center or a paid day off. Duke says CACHE works on causes beyond the health system. That's the right ambition, and it's also where I'd expect the work to get hardest, since those causes sit outside anyone's authority at the hospital.

Examining your own results by race also means finding out that your organization produced worse outcomes for some patients. Whether leaders want that in front of them is a separate question from whether the method works.

Nobody Can Pull This Cord

Bena could pull the cord because the problem was in front of her. A patient, a line, a risk she could name in the moment.

A disparity isn't in front of anybody. The nurse doesn't watch the patient who never comes back for a repeat colonoscopy. The physician doesn't see the one who never scheduled the first. Nothing happens in a room where a person could raise a hand. The pattern spreads across thousands of encounters over years, and only shows up in aggregate.

So there's no cord for this one. Somebody has to go looking on purpose, and then be willing to read what comes back.

Most health systems already have the data.