Palliative Care Is Not Hospice. I Had to Learn That Twice.

The short version

• Palliative care is not hospice. It is specialist care for the symptoms and stress of a serious illness, given alongside treatment meant to cure or control the disease. You can receive it on day one, during chemotherapy, for years.
• A landmark trial changed the standard of care. Patients with newly diagnosed metastatic lung cancer who saw palliative care early reported better quality of life, had fewer depressive symptoms, received less aggressive treatment at the end of life, and lived longer.
• Guidelines now say to refer early. Oncology clinicians should refer people with advanced cancer to palliative care teams early in the disease course, alongside active cancer treatment.
• Most people have never heard of it. An estimated 71% of US adults had never heard the term. Worldwide, only about 14% of people who need it receive it.
• It is not only for cancer. Most adults who need palliative care have heart disease, lung disease, kidney failure, dementia, or another chronic illness.
• You can ask for it yourself. Skip to what to say to your doctor.

A question I did not think to ask changed my mother's life more than any scan I have ever read. It came from an AI tool, and it was five words long: have you considered palliative care?

I want to tell you about two moments, fifteen years apart, that taught me the same lesson. I needed both of them, which tells you something about how hard the lesson is for doctors to learn.

The first moment: a hospice night shift

I trained in urology before I switched to radiology. In the gap between the two, I moonlighted at a hospice and palliative care center. I took the shifts for practical reasons. What happened there rearranged how I think about medicine.

Up to that point my training had one shape. Find the disease early. Intervene. Chase the abnormal lab. Stay on top of the plan. I spent my intern and resident years in a low hum of anxiety, tracking numbers, terrified of missing something. It was all directed at the illness. I am not sure I ever asked a patient what a good day would look like for them.

At the hospice, none of that applied. Nobody was waiting on a result. Nobody was deciding whether to intervene. Every conversation started somewhere else: are you in pain, can you sleep, what is making this day hard, what do you want to be able to do. The whole apparatus was pointed at the person instead of the pathology.

I had spent years treating the disease the patient had. That was the first time I watched a team treat the patient who had the disease.

I carried that with me into radiology and mostly filed it under "formative experience." Then it came back.

The second moment: my mother's cough

Before anyone said the word cancer, my mother had a cough.

Not a polite cough. A relentless one that ran her days and then took her nights. The coughing brought on severe tension headaches. The headaches and the cough together destroyed her sleep. She was exhausted, in pain, and shrinking out of her own life, and all of it was happening before a single treatment decision had been made. Eventually that cough, along with other symptoms, led to a diagnosis of lung cancer.

I used to think of a cough as a clue. A finding that points toward a diagnosis. It took watching my mother to understand that for the person coughing, it is not a clue. It is the whole day.

The research says she was far from alone. In a study of 202 patients with lung cancer, between 57% and 67% had a cough, depending on which validated instrument was used. About 15% reported that coughing often or always disrupted their sleep, and half felt their cough was bad enough to need treatment in its own right. Cough was not associated with cancer stage or cell type. It is not a marker of how advanced the disease is. It is simply a symptom that makes life worse, and it goes underaddressed.

The question I did not know to ask

I am a physician, and I still found the diagnosis disorienting. Lung cancer is not my field. I was reading pathology reports as a daughter, not a radiologist.

So I used AI to prepare. I worked through OpenEvidence to understand what her specific pathology meant, which treatments were on the table, and what I should raise when we met the oncology team. I wanted to walk into that room able to ask real questions instead of nodding.

Among the things it surfaced was a referral to palliative care, upfront, at diagnosis.

My honest first reaction was a flinch. Palliative care sat in my mind where it sits in most people's minds: at the end, after the treatments stop working, next to the word hospice. I went looking for why the recommendation existed, and found a trial I should have known about already.

The trial that moved palliative care to the front of the line

In 2010, researchers randomized 151 patients with newly diagnosed metastatic non-small-cell lung cancer to one of two arms. One group received standard oncology care. The other received the same standard oncology care plus a palliative care team from the start. Same chemotherapy decisions, same oncologists. The only variable was an extra team focused on symptoms, coping, and what mattered to the patient.

The results, published in the New England Journal of Medicine, were not what most people expect.

Standard oncology care alone Plus early palliative care

One trial, 151 patients, metastatic non-small-cell lung cancer (Temel 2010). Compare the two bars within a panel, never across panels, since each panel uses its own scale. Every difference shown reached statistical significance, though the quality-of-life gap is modest in absolute terms: 6.5 points on a 136-point scale. This was a single-center study, and the survival finding in particular has not held up consistently in later pooled analyses. More on that below.

Less aggressive treatment at the end, and yet a longer median survival. Better mood. Better quality of life. Adding a team whose entire job was comfort did not trade away time. In this trial it came with more of it.

That finding reshaped practice. Fourteen years later, the 2024 ASCO guideline update puts it plainly: clinicians should refer patients with advanced solid tumors and blood cancers to specialized interdisciplinary palliative care teams beginning early in the disease course, alongside active treatment of the cancer. Not after. Alongside.

The picture in most people's heads is the old one

Here is the shift, drawn out. The top row is what almost everyone pictures when they hear the word. The bottom row is what the evidence supports and what guidelines now recommend.

The difference is the word "alongside." Palliative care is not what happens when treatment stops. It is a second team working in parallel with the first, from diagnosis onward. Hospice is a specific kind of palliative care for the last months of life. It is one part of the picture, not the whole of it.

The gap between those two rows is not a technicality. It is the reason people say no to something that would have helped them.

71%

Of US adults reported they had never heard of palliative care

Nationally representative survey, 3,445 adults (Trivedi 2019)

14%

Of people worldwide who need palliative care actually receive it

World Health Organization

34%

Of adults needing palliative care have cancer. Most have heart, lung, kidney or neurologic disease.

World Health Organization

What actually happened when we went

We had excellent oncologists. I want to be precise about this, because what follows is not a complaint. They were deep subspecialists doing genuinely cutting-edge work, and they were exactly who I wanted deciding what to do about the tumor.

The palliative care visit was a different kind of appointment entirely.

Nobody opened by talking about the cancer. They asked about the cough. They asked about the headaches. They asked what was happening at two in the morning when she could not stop coughing long enough to fall asleep. They asked what she used to do in a normal week that she had stopped doing.

Then they treated those things. Directly, as problems worth solving on their own terms.

The tension headaches improved. The cough at night improved. She started sleeping. And here is the part I keep turning over: she ended up with a better quality of life than she had in the months before the diagnosis, when the cough was running unchecked and nobody had named it as a target. Nothing about that came from shrinking a tumor. All of it came from treating symptoms as if they mattered.

It made a tremendous difference for her, and for all of us around her. I had read about this. I had even lived it once, at a hospice, fifteen years earlier. I still did not see it coming.

Now the part where I read the evidence critically

I am a radiologist. I look at data for a living and I am suspicious of a single glowing trial, including one that confirms something I now believe. So here is the honest state of the evidence, including where it is weaker than the enthusiasm suggests.

Quality of life and symptoms: this holds up. A meta-analysis in JAMA pooled 43 randomized trials covering 12,731 patients and found that palliative care was associated with meaningful improvements in quality of life and symptom burden at one to three months, plus consistent gains in advance care planning, patient and caregiver satisfaction, and lower health care use. When the authors restricted the analysis to the five trials at lowest risk of bias, the quality-of-life benefit shrank but stayed significant, while the symptom-burden benefit lost significance. A Cochrane review of seven trials in 1,614 people with advanced cancer reached a similar conclusion: real improvements in quality of life and symptom intensity, small in size, rated low certainty.

Survival: genuinely unsettled. The 2010 trial found longer median survival. A later trial comparing early versus delayed palliative care found one-year survival of 63% versus 48%, favoring the early group. But the JAMA meta-analysis found no significant association between palliative care and survival across all trials, and Cochrane rated the survival evidence very low certainty. My read: the quality-of-life case is settled and the longevity case is a plausible bonus that nobody should promise you. The reason to go is not that it might buy you months. The reason to go is that it will very likely make the months better.

It works beyond cancer. In a randomized trial of 150 patients with advanced heart failure, adding an interdisciplinary palliative care team to usual cardiology care produced clinically significant improvements in quality of life, depression, anxiety, and spiritual well-being at six months. Hospitalization and mortality did not change. Quality of life did.

Depression: mixed. The 2010 lung cancer trial found substantially fewer depressive symptoms. The Cochrane pooled analysis of five trials did not find a significant difference. Both things are in the literature and I am not going to pick the flattering one.

Why most people still do not get it

If the evidence is this consistent, the obvious question is why palliative care remains something families stumble into rather than something offered by default. Three reasons, all fixable.

The name. Most people have never heard the term, and many who have heard it think it means giving up. Awareness is lower in rural areas and lower among Hispanic respondents regardless of geography, which makes this an access problem, not just a vocabulary problem.

The workforce. There are not enough palliative care clinicians to see every eligible patient every month, which is the schedule most of the landmark trials used. This is stated openly as a barrier in the trial literature.

Distance and logistics. Monthly specialist visits are a real burden on a sick person and a working family.

The good news is that researchers went after the last two directly, and the results are encouraging. A trial of 1,250 patients across 22 cancer centers found that delivering early palliative care by secure video visits produced equivalent quality of life to delivering it in person. A separate trial of 507 patients tested a stepped model, where visits happen at key transition points rather than monthly, and patients step up to more intensive contact if their quality-of-life score drops. That model roughly halved the number of visits without giving up the quality-of-life benefit. Worth noting honestly: the stepped group spent fewer days in hospice, which the authors flagged rather than buried.

Between video delivery and stepped scheduling, the "we do not have the capacity" objection is losing its force.

If someone you love has a serious illness

This is the practical part. You do not have to wait to be offered palliative care. In most systems you can ask for it, and asking works.

What to say, and what to ask

• Ask directly: "Can we get a palliative care consult alongside treatment?" The word "alongside" does a lot of work. It signals you are not asking to stop treatment.
• Name the symptoms, not the diagnosis. Cough, pain, nausea, breathlessness, insomnia, anxiety, appetite. These are the targets. List the ones wrecking the day.
• Say what a good day would look like. Palliative teams build the plan around this, and most of us never get asked.
• Ask if it can be virtual. The evidence says video visits work as well as in person for this.
• Ask what the team covers: symptom management, mood, coordination between specialists, help for the caregiver, and planning conversations. It is usually a physician, nurse, social worker, and chaplain, not one person.
• Ask about coverage. Palliative care is a medical subspecialty billed like any other. Medicare, Medicaid, and most private plans cover it.
• If nobody has an answer, the Center to Advance Palliative Care runs a public provider directory searchable by ZIP code and setting.

One more note, from the daughter rather than the doctor. Bring someone to the appointment, and make sure the person who is sick gets asked the questions directly. My mother does not speak English as her first language. In every room we sat in, the risk was that the conversation would happen over her head and around her, between clinicians and the family member who could keep up. The palliative team was the one that consistently aimed the questions at her.

What I keep coming back to

The thing that took me two decades to absorb is that treating the disease and treating the person are not the same project, and doing one well does not accomplish the other. Our oncologists were excellent at the first. Nothing about their excellence produced a night of sleep.

Both projects are real. Both need someone assigned to them. The evidence says running them in parallel from day one makes people feel better, and may do more than that.

I learned this once in a hospice fifteen years ago and then let it fade into a nice story about my training. It took my mother's cough, and a question from a machine, to make me actually use it.

---

How this piece was built

I started with a brain dump. I talked out the whole story, unstructured and unedited: the hospice moonlighting, my mother's cough, the AI query, the palliative care visit, what changed. Then I handed that raw transcript to Claude (Anthropic) and worked with it to turn the material into this piece.

AI disclosure. The experience, the argument, and the point of view are mine. Claude searched PubMed, Consensus, and the WHO for the peer-reviewed evidence cited here and verified every trial number against the source abstract, produced both figures, structured the narrative from my spoken brain dump, and helped me tighten the draft. I directed the framing, supplied the experience, and reviewed every claim, number, and citation before publishing. Separately, I used OpenEvidence during my mother's care to prepare questions for her oncology team, which is how the palliative care referral came up in the first place.

References

1. Temel JS, et al. Early palliative care for patients with metastatic non-small-cell lung cancer. N Engl J Med. 2010;363(8):733-742. doi:10.1056/NEJMoa1000678
2. Sanders JJ, et al. Palliative care for patients with cancer: ASCO guideline update. J Clin Oncol. 2024;42(19):2336-2357. doi:10.1200/JCO.24.00542
3. Kavalieratos D, et al. Association between palliative care and patient and caregiver outcomes: a systematic review and meta-analysis. JAMA. 2016;316(20):2104-2114. doi:10.1001/jama.2016.16840
4. Haun MW, et al. Early palliative care for adults with advanced cancer. Cochrane Database Syst Rev. 2017;6(6):CD011129. doi:10.1002/14651858.CD011129.pub2
5. Bakitas MA, et al. Early versus delayed initiation of concurrent palliative oncology care: patient outcomes in the ENABLE III randomized controlled trial. J Clin Oncol. 2015;33(13):1438-1445. doi:10.1200/JCO.2014.58.6362
6. Zimmermann C, et al. Early palliative care for patients with advanced cancer: a cluster-randomised controlled trial. Lancet. 2014;383(9930):1721-1730. doi:10.1016/S0140-6736(13)62416-2
7. Rogers JG, et al. Palliative care in heart failure: the PAL-HF randomized, controlled clinical trial. J Am Coll Cardiol. 2017;70(3):331-341. doi:10.1016/j.jacc.2017.05.030
8. Greer JA, et al. Telehealth vs in-person early palliative care for patients with advanced lung cancer: a multisite randomized clinical trial. JAMA. 2024;332(14):1153-1164. doi:10.1001/jama.2024.13964
9. Temel JS, et al. Stepped palliative care for patients with advanced lung cancer: a randomized clinical trial. JAMA. 2024;332(6):471-481. doi:10.1001/jama.2024.10398
10. Harle A, et al. A cross sectional study to determine the prevalence of cough and its impact in patients with lung cancer: a patient unmet need. BMC Cancer. 2020;20(1):9. doi:10.1186/s12885-019-6451-1
11. Trivedi N, et al. Awareness of palliative care among a nationally representative sample of U.S. adults. J Palliat Med. 2019;22(12):1578-1582. doi:10.1089/jpm.2018.0656
12. Langan E, et al. Comparing palliative care knowledge in metropolitan and nonmetropolitan areas of the United States: results from a national survey. J Palliat Med. 2021;24(12):1833-1839. doi:10.1089/jpm.2021.0114
13. World Health Organization. Palliative care fact sheet. 2020. who.int

Peer-reviewed sources were located through PubMed and Consensus. Nothing in this post describes anyone's protected health information beyond what my family has chosen to share.

Moving Is the New Standing: Why I Traded My Chair for a Treadmill

Life as a radiologist means living in front of a screen. A regular day means hours at a reading station, scrolling through CT and MRI. A weekend on call can mean twelve straight hours in the chair. Sitting seems harmless because everyone does it. It isn't.

The wake-up call

In my mid-thirties, I considered myself pretty athletic. I had run a marathon during fellowship and stayed active. Then I spent two weeks writing a grant, leaning into my computer in the same seated position, hour after hour, day after day. Those two weeks ended with low back pain so sudden and severe that it landed me in the hospital.

I had a standing desk available at the time. I sat anyway, because the chair always offers the path of least resistance. I have lived with low back pain and some disc disease in my lower spine ever since. A marathon a year earlier didn't protect me from what two weeks in a chair did.

I'm far from the only one. When researchers surveyed 123 radiologists, 38% reported a work-related musculoskeletal injury, and low back discomfort topped the list (Rodrigues, J Digit Imaging 2014). And it isn't just us. The average US adult now sits about 6.4 hours a day, nearly an hour more than in 2007 (Yang, JAMA 2019).

What sitting does to a body

My turning point came when I read Deskbound: Standing Up to a Sitting World by physical therapist Kelly Starrett, written with Juliet Starrett and Glen Cordoza. Its core premise: your body adapts to whatever positions you hold most. Sit most of the day, and your body gets very good at sitting. Starrett describes the lower back flattening and the upper back and shoulders rounding forward. It reminds me of playing guitar. Press the strings long enough and your fingertips grow calluses. Your body protects itself by taking the shape of your habits.

As a radiologist, I can show you part of this on imaging. When researchers compared sitting and standing X-rays of the lumbar spine in 109 patients with low back pain, the lumbar curve averaged 49 degrees standing and only 34 degrees sitting (Lord, Spine 1997).

Figure 1. Same people, same spines, different posture. Curves are schematic; the angles are the study's measured averages. Source: Lord et al., Spine 1997.

Your discs care about movement too. In a classic experiment, researchers placed a pressure sensor inside a volunteer's lumbar disc and recorded a full day of ordinary life. They concluded that constantly changing position helps move fluid, and with it nutrition, into the disc (Wilke, Spine 1999). The same study also challenged a number you may have heard: that sitting loads your discs far more than standing. In their measurements, relaxed standing and unsupported sitting produced similar pressures, with sitting slightly lower. That study followed one person, so I treat it as a clue rather than a verdict. But both findings point the same way: position changes matter more than any one "correct" posture.

A radiologist's caveat. How permanent these adaptations become, and how directly posture causes back pain, remains unsettled. A review of 41 systematic reviews linked sitting, standing, and awkward postures to low back pain but found no consensus that any single posture causes it (Swain, J Biomech 2019). The better-supported lesson is that holding any one position for hours is the problem.

A drop at a time

Even a drop of water will eventually fill a bucket. The hours in a chair add up quietly over a career. The largest analysis of this question pooled data from more than one million adults followed for up to 18 years (Ekelund, Lancet 2016). People who sat more than 8 hours a day and moved the least had a 59% higher risk of dying during follow-up than people who sat less than 4 hours and moved the most. Now the hopeful part: people who sat more than 8 hours a day but also got roughly 60 to 75 minutes of moderate activity daily showed no significant increase in risk.

Figure 2. Four cells from the harmonised meta-analysis. These are observational data with self-reported sitting and activity, so they show association, not proof of cause. Source: Ekelund et al., Lancet 2016.

Moving is the new standing

After reading the book, I switched to standing. It took time, but standing became my default. Then, after a few hours on my feet, a familiar ache would creep back into my lower back. Sitting down for a while eased it. I had simply traded one static posture for another.

Research explains why. In lab studies, people with no history of back pain stood for longer than 42 minutes, and about half of them (528 of 1,070) developed low back pain (Khoshroo, Sci Rep 2023). A separate meta-analysis of desk work found that prolonged standing caused no less low back pain than prolonged sitting (De Carvalho, Work 2020). Standing isn't a cure. Moving is the point.

So I added a small under-desk treadmill. At about 1 mile per hour, the standing ache doesn't show up, my hands stay steady on the mouse and keyboard, and my body keeps moving. You don't need a workout. You need motion.

Can a radiologist really read while walking? One study tested exactly that. Three radiologists read 55 lung cancer screening CTs twice, once seated and once walking on a treadmill workstation. They found about as many nodules walking as seated (no statistically significant difference), made consistent follow-up recommendations, and finished each exam faster while walking (Johnson, JACR 2019). It was a small study, but a reassuring one.

Expect a learning curve. In a lab study of first-time treadmill users with no practice, typing, mouse clicking, and math problem solving slipped by 6% to 11%, while attention, processing speed, and reading comprehension held steady (John, J Phys Act Health 2009). Start slow and give yourself a few weeks.

Your metabolism notices the difference, too. Walking and working at about 1.1 mph burned roughly 120 more calories per hour than seated work in a study of 15 office workers with obesity (Levine and Miller, Br J Sports Med 2007). And in a randomized crossover trial of 19 adults with overweight or obesity, a 2-minute light walk every 20 minutes lowered the blood sugar rise after a meal by about 25% and the insulin rise by about 24%, compared with sitting straight through (Dunstan, Diabetes Care 2012).

Figure 3. Both walking conditions differed significantly from uninterrupted sitting (P < 0.01). Small trial, one day per condition, adults aged 45 to 65. Source: Dunstan et al., Diabetes Care 2012.

The psychological toll

This past weekend I was on call. Both days ran twelve hours, nonstop and busy. I skipped my run and my walks and barely moved. By the end, I felt terrible, and not only in my back. I felt flat, foggy, and low.

Movement matters for the mind as much as the body. We often credit endorphins for that lift. Whatever the mechanism, the data are striking. Across 15 studies and more than 191,000 adults, people who got just half the recommended amount of activity had an 18% lower risk of depression than inactive people, and those who met the recommendation had a 25% lower risk (Pearce, JAMA Psychiatry 2022). In one workplace study, sit-stand desks cut sitting by 66 minutes a day, reduced upper back and neck pain by 54%, and improved mood. When researchers took the desks away, most of those gains disappeared within two weeks (Pronk, Prev Chronic Dis 2012).

That finding explains my call weekend. You can't bank movement. It works more like a daily dose.

Start small

Don't wait for a hospital visit to change your habits. Here is what works for me:

• Make standing the starting position. Set up your desk standing, so sitting becomes a deliberate choice instead of the default.
• Move before it hurts. At the first twinge from standing, walk for a while or sit for a stretch. Rotate through positions all day.
• Walk slowly. About 1 mph keeps your hands steady and your body moving.
• Break up long stretches. A 2-minute walk every 20 to 30 minutes, or between cases, adds up over a shift.
• Protect movement on your hardest days. Those are the days you need it most.
• Ask for the equipment. Sit-stand desks cut workplace sitting by about 100 minutes a day in the short term and about an hour a day at 3 to 12 months (Shrestha, Cochrane 2018). That gives you a solid case to request one.

My next goal is more time on the treadmill, adding minutes slowly the same way I once added standing time. It took two weeks of stillness to put me in the hospital. Small daily choices are what keep me out of it.

This post shares my own experience and my reading of the research. It is not medical advice. If you have new, severe, or worsening back pain, please see your clinician.

References

1. Starrett K, Starrett J, Cordoza G. Deskbound: Standing Up to a Sitting World. 2016.
2. Rodrigues JCL, et al. Musculoskeletal symptoms amongst clinical radiologists and the implications of reporting environment ergonomics. J Digit Imaging. 2014;27(2):255-261. doi:10.1007/s10278-013-9642-3
3. Yang L, et al. Trends in sedentary behavior among the US population, 2001-2016. JAMA. 2019;321(16):1587-1597. doi:10.1001/jama.2019.3636
4. Lord MJ, et al. Lumbar lordosis: effects of sitting and standing. Spine. 1997;22(21):2571-2574. doi:10.1097/00007632-199711010-00020
5. Wilke HJ, et al. New in vivo measurements of pressures in the intervertebral disc in daily life. Spine. 1999;24(8):755-762. doi:10.1097/00007632-199904150-00005
6. Swain CTV, et al. No consensus on causality of spine postures or physical exposure and low back pain. J Biomech. 2019;102:109312. doi:10.1016/j.jbiomech.2019.08.006
7. Ekelund U, et al. Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality? Lancet. 2016;388(10051):1302-1310. doi:10.1016/S0140-6736(16)30370-1
8. Khoshroo F, et al. Distinctive characteristics of prolonged standing low back pain developers. Sci Rep. 2023;13(1):6392. doi:10.1038/s41598-023-33590-5
9. De Carvalho D, et al. Does objectively measured prolonged standing for desk work result in lower ratings of perceived low back pain than sitting? Work. 2020;67(2):431-440. doi:10.3233/WOR-203292
10. Johnson CR, et al. Effect of dynamic workstation use on radiologist detection of pulmonary nodules on CT. J Am Coll Radiol. 2019;16(4 Pt A):451-457. doi:10.1016/j.jacr.2018.10.017
11. John D, et al. Effect of using a treadmill workstation on performance of simulated office work tasks. J Phys Act Health. 2009;6(5):617-624. doi:10.1123/jpah.6.5.617
12. Levine JA, Miller JM. The energy expenditure of using a "walk-and-work" desk for office workers with obesity. Br J Sports Med. 2007;41(9):558-561. doi:10.1136/bjsm.2006.032755
13. Dunstan DW, et al. Breaking up prolonged sitting reduces postprandial glucose and insulin responses. Diabetes Care. 2012;35(5):976-983. doi:10.2337/dc11-1931
14. Pearce M, et al. Association between physical activity and risk of depression. JAMA Psychiatry. 2022;79(6):550-559. doi:10.1001/jamapsychiatry.2022.0609
15. Pronk NP, et al. Reducing occupational sitting time and improving worker health: the Take-a-Stand Project, 2011. Prev Chronic Dis. 2012;9:E154. doi:10.5888/pcd9.110323
16. Shrestha N, et al. Workplace interventions for reducing sitting at work. Cochrane Database Syst Rev. 2018;12:CD010912. doi:10.1002/14651858.CD010912.pub5

How I made this post. I dictated my story and asked Claude, Anthropic's AI assistant, to co-draft it. Claude searched PubMed for the studies cited here, fact-checked and merged two earlier drafts into this version, and built all three figures from numbers reported in those papers. I reviewed and edited the text, checked the figures and citations against the original studies, and decided what to keep. The story, the back, and the opinions are mine.

The Good Trials Are Better Than I Expected. The Failure Modes Are Worse.

The 30-second version

• When an AI handed radiologists the wrong answer, very experienced readers went from scoring 82.3% of mammograms correctly to 45.5%. The least experienced fell to 19.8%. Experience helped. It did not protect.
• And yet the best evidence is genuinely good: in a randomized trial of 105,934 women, AI-supported screening found more cancers with 44% less reading, with no rise in interval cancers.
• Lab performance does not transfer. The same class of tool that dazzles in a trial hit 35% sensitivity in a real primary-care population.
• So the answer is not "does AI work." It is does this model still work here, this month, which is a monitoring problem, and radiology has quietly started building the boring infrastructure to solve it.

The most important number I found while researching this series is not a sensitivity or an area under a curve. It is what happened to expert radiologists when the machine was confidently wrong.

This is the second of three posts. The first argued that the fear of AI eliminating jobs is aimed at the wrong target, because health care cannot staff the work it already has. That argument has a hole in it, and I want to put my finger in it before going any further: a staffing crisis is a reason to want a tool. It is not evidence that the tool works.

So this post is the evidence. All of it, including the parts I wish were different.

The short version is that the good trials are better than I expected and the failure modes are worse. Both of those things are true at once, and any version of this conversation that gives you only one of them is selling something.

What the good evidence actually shows

The strongest data we have comes from breast screening, because that is where somebody finally did the randomized trial.

The MASAI trial in Sweden randomized 105,934 women to either AI-supported screening or standard double reading by two radiologists. The AI triaged which exams needed a second reader and flagged suspicious findings. The final results, published in The Lancet in 2026, reported the primary outcome: the interval cancer rate, meaning cancers that surface between screening rounds because the screen missed them. That rate was 1.55 per 1,000 with AI versus 1.76 without, which met the trial's bar for non-inferiority. Sensitivity was higher with AI, 80.5% versus 73.8%. Specificity was identical at 98.5%.

Two radiologists, no AI AI-supported reading

One randomized trial, 105,934 women, Sweden. More cancers found, higher sensitivity, same specificity, and 44% less reading. The three panels use different scales, so only the within-panel comparison is meaningful. Detection and workload figures from the 2025 Lancet Digital Health report; sensitivity from the 2026 Lancet primary analysis of the same trial.

The workload number is the one I keep coming back to. AI-supported screening required 61,248 screen readings where standard double reading required 109,692. That is a 44% reduction in reading, with more cancers found and no significant increase in false positives.

Germany replicated the direction at enormous scale. The PRAIM study followed 463,094 women screened by 119 radiologists across 12 sites and found a cancer detection rate of 6.7 per 1,000 with AI support versus 5.7 without, a 17.6% relative increase, with a recall rate that was slightly lower rather than higher. One important caveat: PRAIM was observational and the radiologists chose for themselves whether to use the AI, so the groups were not randomly assigned and the comparison is weaker than MASAI's.

Notice what actually improved. Not the radiologist's eye. The radiologist's throughput, and the number of second reads that never needed a human at all.

The place this matters most is not Scottsdale

Everything above happened in wealthy countries with organized screening programs and plenty of radiologists. The larger prize is somewhere else entirely.

In Bangladesh, researchers ran 23,954 chest X-rays from three tuberculosis screening centers past five commercial AI algorithms and a panel of three registered radiologists. All five algorithms significantly outperformed the radiologists, and all five cut the number of confirmatory molecular tests needed by about half while holding sensitivity above 90%. The authors also reported that every algorithm performed worse in people over 60 and in people with a history of TB, which is exactly the kind of subgroup detail that gets dropped when these results are summarized.

In China, a study across 7 county-level and 32 township-level facilities reviewed 93,319 patients, of whom 273 had bacteriologically confirmed pulmonary TB. The AI flagged 83.9% of those confirmed cases; the radiologists reading at the time caught 25.6%. The AI's positive predictive value was much worse, 1.7% against 10.3%, meaning far more false alarms. But used as a triage filter with human review of flagged images, it cut the radiologist workload by 85.5% without missing any case the radiologists had found on their own.

A separate validation on more than one million chest X-rays reported an AUC of 98.51% and a false negative rate slightly better than the radiologists', with the potential to auto-report up to 80% of normal studies.

These are the numbers that make me hopeful, and they have almost nothing to do with whether AI is better than me. They are about places where there is no radiologist to be better than. That is where "raising all boats" stops being a slogan.

Four findings that should keep us honest

I do not want to write a brochure. Here is the evidence that cuts the other way, and some of it is genuinely alarming.

OR 1.20

Higher odds of burnout among radiologists who used AI frequently, with a dose-response by frequency of use

6,726 radiologists, 1,143 hospitals (Liu 2024)

−6.0pts

Drop in adenoma detection when endoscopists went back to working without AI, after months of using it

1,443 colonoscopies, 4 centers (Budzyń 2025)

35%

Sensitivity of a commercial chest X-ray AI in a real-world, low-prevalence screening population

3,047 radiographs, 2 primary care centers (Kim 2023)

Automation bias is worse than I expected. In a prospective experiment, 27 radiologists read mammograms with a purported AI system that was deliberately wrong on 12 of 40 cases. Among the most experienced readers, the share of correctly assigned BI-RADS categories fell from 82.3% to 45.5% when the AI suggested the wrong category. Among the least experienced, it fell from 79.7% to 19.8%. Experience helped. It did not protect.

AI suggested the correct category AI suggested the wrong one

Experience buys you some protection. Not much. All six bars are zero-based on one shared scale, so every length is directly comparable. The three groups perform almost identically when the machine is right. The gap opens only when it is wrong. This was a prospective experiment with a purported AI system, not a deployed product, which makes it a clean measure of the effect and not a claim about any commercial tool. Dratsch et al., Radiology 2023.

Deskilling may be real. Four Polish endoscopy centers compared adenoma detection during unassisted colonoscopy in the three months before AI was introduced and the three months after. The rate fell from 28.4% to 22.4%, an absolute drop of 6 percentage points. This was a retrospective before-and-after comparison, not a randomized one, so seasonality, case mix, and staffing changes are all live alternative explanations. But the effect size is large enough that dismissing it would be motivated reasoning.

Lab performance does not transfer automatically. A commercial chest X-ray AI validated against CT findings in 3,047 consecutive radiographs from two primary healthcare centers, where the prevalence of significant disease was 2.2%, achieved a sensitivity of 35.3% and an AUROC of 0.648. The authors' conclusion is the sentence I would put on a poster in every department: regulatory approval and experimental performance may not translate to real practice, and the mismatch tends to be worst exactly where the need is greatest.

AI did not make radiologists less burned out. A survey of 6,726 radiologists across 1,143 Chinese hospitals found that frequent AI users had higher odds of burnout than non-users, with an adjusted odds ratio of 1.20 and a dose-response relationship with frequency of use, driven mostly by emotional exhaustion. It was worst among radiologists with high workloads. This is cross-sectional, so causality could run either way. But it points at something that matches my experience: if you speed up one task and leave the volume expectation untouched, you have not reduced anyone's suffering. You have just changed what they do all day and asked for more of it.

Equity does not happen by itself

The version of this future I want is one where a woman in a rural county gets her MRI read this week instead of in March. But the technology does not deliver that on its own, and there is a well-documented case showing exactly how it fails.

A commercial algorithm used across US health systems to identify patients needing extra care was found to be substantially biased against Black patients: at any given risk score, Black patients were considerably sicker than white patients. The mechanism was not malice or a bad training set in the usual sense. The algorithm predicted health care costs as a proxy for illness, and because less money has historically been spent on Black patients, the proxy encoded the disparity. Correcting it would have raised the share of Black patients flagged for additional help from 17.7% to 46.5%.

That is a design decision, not an accident of the math. Somebody chose a convenient outcome variable. The same choice is available to every group building an imaging model right now, and it will be made well or badly depending on who is in the room.

My own specialty has actually built something

Here is the part of this story I did not expect to be writing, and the part I am proudest of. While the broader AI conversation has been arguing about whether guardrails are even possible, radiology quietly went and built some.

In June 2024 the American College of Radiology launched ARCH-AI, the ACR Recognized Center for Healthcare-AI, described as the first national quality assurance program for AI in medical imaging. To earn the designation, a practice attests to a specific set of things: that it has an interdisciplinary AI governance group, that it keeps a documented inventory of every algorithm it runs, that it has a deliberate process for reviewing and selecting those algorithms, that it does acceptance testing before deployment, that it monitors performance afterward, and that it manages any models it built itself.

None of that is glamorous. All of it is exactly what was missing in the failure modes above.

ARCH-AI is deliberately a stepping stone. The ACR leadership behind it have written openly that it exists as a precursor to a formal accreditation program, on the same model the College has used since radiation oncology in 1966 and mammography in 1987, with council approval anticipated around spring 2027. Their stated reason for building it is the same observation this whole post keeps circling: real-world AI performance often differs from what premarket testing showed. That sentence is in the ACR's own road map paper. It is not a criticism from outside the field.

The piece I find genuinely impressive is the second one. In November 2024 the ACR launched Assess-AI, a registry inside the National Radiology Data Registry that monitors how deployed imaging AI is actually performing, in real practices, over time. Participating sites send de-identified algorithm outputs, report text, and study metadata. The registry extracts surrogate labels from the radiology reports, computes concordance between what the algorithm said and what the radiologist ultimately said, and returns it as dashboards. Sites can compare themselves against national benchmarks and against peers matched on facility type, region, trauma level, and urban versus rural. They can drill into discordant cases and look at whether the disagreements cluster by demographic or technical factor. It currently covers intracranial hemorrhage, pulmonary embolism, pneumothorax, large-vessel occlusion, bone age, and cervical spine fracture.

Sit with what that is for a moment. It is post-market surveillance for algorithms, built by the specialty that uses them, that measures whether a model still works in your department, on your scanners, with your patients, after the vendor demo is over. Model drift is not hypothetical; departments change their protocols, equipment, and case mix constantly, and performance moves with them. Assess-AI is the mechanism for noticing.

In 2026 the ACR and SIIM also approved a formal practice parameter covering tool selection, predeployment evaluation, ongoing monitoring, and privacy, and the program has begun expanding internationally, with the University Hospital of Bern named its first site outside the United States.

Radiology did not wait to be regulated. It built the registry, wrote the parameter, and put a badge on the wall. I would like the rest of the AI industry to notice that this was possible.

I want to be measured about it. ARCH-AI is attestation, not audit: a practice affirms it is doing these things rather than being inspected. Assess-AI depends on voluntary participation and on surrogate labels pulled from report text by a language model, which is a reasonable approximation of truth and not truth itself. Neither program stops a department from buying a bad algorithm. What they do is make it much harder to buy one and never find out.

Before your department buys an imaging AI, ask these

ARCH-AI covers the institutional layer. These are the clinical questions underneath it.

• What population was it validated on, and what was the disease prevalence in that population compared to ours?
• What are the reported subgroup results by age, sex, race, body habitus, and scanner vendor? If there are none, that is the answer.
• Is it a triage tool, a second reader, or a concurrent reader? Each one fails differently and each one needs a different workflow.
• What happens to the reading list when it is wrong, and how would we ever find out that it was?
• Does the volume expectation change when the tool goes live? If throughput goes up and staffing does not, we have bought a burnout accelerator.
• Are we submitting to Assess-AI, and if not, what is our alternative plan for catching drift?
• How do we preserve the skills of residents and junior attendings who will train alongside it?

What this post does not tell you

Two posts in, I have argued that the work is moving rather than vanishing, and that the technology is real but conditional: good in the trial, fragile in the field, safe only with a loop and a registry behind it.

Both of those are arguments about what medicine should do. They assume medicine gets to decide the pace.

I no longer think that is true. Several hundred million people a week are already asking these systems health questions, and a growing number have connected their own medical records to them. The knowledge asymmetry that defined the exam room for a century is closing from the patient's side, and nobody asked us.

That is the last post.

---

How this piece was built

Every trial number here was checked against the source abstract rather than a summary of it, and where a study is observational, small, or before-and-after rather than randomized, I have said so in the same sentence as the finding rather than in a footnote. Two figures in this post carry warnings about how to read them; please read them.

AI disclosure. The argument and the point of view are mine. I worked with Claude (Anthropic) as a research and drafting partner: it searched PubMed and Consensus, verified each trial's numbers against the primary source, produced the two figures, and helped organize the draft. I reviewed every claim and citation before publishing.

References

1. Hernström V, et al. Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel-group, non-inferiority, single-blinded, screening accuracy study. Lancet Digit Health. 2025;7(3):e175-e183. doi:10.1016/S2589-7500(24)00267-X
2. Hernström V, et al. Interval cancers and screening outcomes in the MASAI trial. Lancet. 2026;407(10430):505-514. doi:10.1016/S0140-6736(25)02464-X
3. Lång K, et al. Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (MASAI): a clinical safety analysis. Lancet Oncol. 2023;24(8):936-944. doi:10.1016/S1470-2045(23)00298-X
4. Eisemann N, et al. Nationwide real-world implementation of AI for cancer detection in population-based mammography screening. Nat Med. 2025;31(3):917-924. doi:10.1038/s41591-024-03408-6
5. Dratsch T, et al. Automation bias in mammography: the impact of artificial intelligence BI-RADS suggestions on reader performance. Radiology. 2023;307(4):e222176. doi:10.1148/radiol.222176
6. Budzyń K, et al. Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study. Lancet Gastroenterol Hepatol. 2025;10(10):896-903. doi:10.1016/S2468-1253(25)00133-5
7. Kim JH, et al. Clinical validation of a deep learning-based software for lung nodule detection in chest radiographs in a health screening population. Eur Radiol. 2023;33(11):7823-7833. doi:10.1007/s00330-023-09761-3
8. Liu Y, et al. Artificial intelligence use and burnout among radiologists in China. JAMA Netw Open. 2024;7(12):e2448714. doi:10.1001/jamanetworkopen.2024.48714
9. Qin ZZ, et al. Tuberculosis detection from chest x-rays for triaging in a high tuberculosis-burden setting: an evaluation of five artificial intelligence algorithms. Lancet Digit Health. 2021;3(9):e543-e554. doi:10.1016/S2589-7500(21)00116-3
10. Jiang Y, et al. Effectiveness of computer-aided detection for active pulmonary tuberculosis screening in resource-limited settings. J Med Internet Res. 2025;27:e69109. doi:10.2196/69109
11. Munjal P, et al. Assessing the reliability of an AI-based chest radiograph interpretation system on over one million radiographs. NPJ Digit Med. 2025;8(1):318. doi:10.1038/s41746-025-01693-0
12. Obermeyer Z, et al. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447-453. doi:10.1126/science.aax2342
13. Larson DB, et al. A road map for the ACR Recognized Center for Healthcare-AI. J Am Coll Radiol. 2025;22(5):586-592. doi:10.1016/j.jacr.2025.02.008
14. Coombs LP, et al. The ACR Assess-AI registry: national performance monitoring of clinical AI. J Am Coll Radiol. 2026;23(9):1557-1566. doi:10.1016/j.jacr.2026.04.024

Peer-reviewed sources were located through PubMed and Consensus. Nothing in this post describes any individual patient or protected health information, and nothing here represents the position of my employer.

Ten Years Ago, They Told Us to Stop Training Radiologists

The 30-second version

• In 2016 a Nobel laureate told the world to stop training radiologists. There are more radiologists now than there were then, and the wait for an MRI is still measured in months.
• Everyone tells the ATM story half-finished. Teller jobs did rise, and are now projected to fall 13.2%. Cashiers are down roughly 700,000 since 2019 and still falling.
• But radiology is not retail, and the reason is demand, not prestige. Retail demand is roughly fixed, so faster checkout means fewer checkers. Imaging demand is nowhere close to met. Free capacity in a system with a backlog gets absorbed, not shed.
• In one BLS projection round, office administration sheds 752,100 jobs while health care adds 2.2 million, about 37% of all new jobs in the economy. The work is moving, not vanishing. The fear is aimed at the wrong target.

Ten years ago a Nobel laureate told the world to stop training radiologists. I am a radiologist. I am still here, and so is everyone I trained with, and the waiting list for an MRI in this country is still months long. That gap between the prediction and the reality is worth understanding, because we are about to make the same category of mistake again.

In 2016, Geoffrey Hinton stood at a seminar in Toronto and said that anyone working as a radiologist was "like the coyote that's already over the edge of the cliff" and had not yet looked down. He said people should stop training radiologists immediately, because deep learning would beat them within five years. He is one of the most important scientists of the century and he was not being cynical. He genuinely believed it.

Medical students believed it too. Some of them changed their plans.

Here is where things actually landed. The number of practicing radiologists in the United States rose 12% between 2010 and 2022, from 34,328 to 38,306, which works out to a move from 11.1 to 11.5 radiologists per 100,000 people. One large midwestern academic medical center has reportedly grown its radiology staff by 55% since 2016, to roughly 400 radiologists, though that figure comes from press reporting rather than a peer-reviewed source. Meanwhile, imaging volume kept climbing and turnaround times kept stretching. The prediction did not just miss. It missed in the opposite direction.

The interesting part is who is quoting the story now. Jensen Huang has spent the last year telling audiences at Davos and elsewhere that radiology is the proof that AI creates jobs rather than destroying them. I want to be careful here, because a lot of people have heard "Jensen Huang says AI is coming for radiology" and that is backwards; he is arguing the opposite. But his version has its own problem. He claims hospitals are hiring more radiologists because of AI. That causal story is not supported. Imaging volumes have been growing steadily for decades for reasons that have nothing to do with machine learning, and interpretation speed has barely moved. We are hiring because we are drowning, not because algorithms freed us up.

Both the doom version and the triumph version get radiology wrong in the same way. They treat the job as a task, and the task as the job.

What I think, and where this series is going

This is the first of three posts, and I want to state my position rather than meander toward it. I think this is the most opportune moment in my professional lifetime to change how care is actually delivered, and that almost nobody is arguing about the right thing. The public fight is over whether a model can out-read a human. What will actually decide whether patients are better off is whether we can finally match expertise to need.

Across the three posts, here is what I believe:

• The work redistributes rather than disappears. Health care cannot fill the jobs it already has. The fear of mass unemployment and the reality of a staffing crisis are unfolding in the same room, and the thing standing between them is retraining nobody has funded. That is this post.
• Triage becomes the organizing principle, and the pair outperforms either half, but only if the human is genuinely in the loop and somebody is monitoring the machine forever. That is the second post, on what the evidence actually shows.
• Reach matters more than accuracy, and patients level up the most. The knowledge asymmetry that defined the exam room for a century is closing from the patient's side, whether or not medicine approves. That is the third.

None of it is inevitable. It is a choice about what we build, not a forecast we are waiting on.

The ATM story, and the half of it nobody tells

Whenever this argument comes up, someone brings up bank tellers. It is the most-cited analogy in the automation debate and it is worth getting right.

The economist James Bessen went and collected the data. As ATMs spread across the United States, the number of tellers needed to run an average urban branch fell from about 20 to 13 between 1988 and 2004. But a cheaper branch is a branch worth opening, so banks opened 43% more of them in urban areas. Total teller employment held steady and even rose. The remaining work shifted toward relationships and problem-solving, the parts a cash dispenser could not touch.

That is the version everyone tells. Here is the part that gets left out: it did not last. The Bureau of Labor Statistics now projects teller employment to keep falling, because online and mobile banking eventually automated not just the cash-handling task but most of the reasons to walk into a branch at all.

Cashiers are the sharper case, and it is worth being concrete, because people wave at self-checkout without ever citing a figure. In May 2019 there were roughly 3.60 million cashiers in the United States, one of the largest occupations here. By 2025 there were 3.11 million. The BLS projects 2.91 million by 2035, a further loss of 200,600 jobs and a 6.5% decline, still the largest projected drop of any occupation in the country. Tellers are on that same table now, projected to fall 13.2%.

That is close to 700,000 cashier jobs gone or going in fifteen years. Not zero. Not a myth. The largest.

So the two stories are not the same story, and the difference is the whole point. Tellers rose while machines handled part of the job and are only now sliding, decades later, once phones removed the reason to enter a branch at all. Cashiers are further down that same road: the scanner and the payment terminal took the core of the work, the customer absorbed the rest, and one attendant now watches six lanes where four people used to stand.

The real lesson is not "automation never displaces anyone." It is a lesson about how much of a job gets automated, and it has a shape.

The question is not whether automation displaces people. It is where on the curve you are standing, and which way you are moving. The curve itself is a conceptual diagram drawn to organize the argument; nobody has measured this shape directly, and the horizontal axis has no units. The cashier figures below it are real: about 3.60 million in May 2019 (BLS Occupational Employment and Wage Statistics), 3.11 million in 2025 and a projected 2.91 million in 2035 (BLS National Employment Matrix, 2025–2035 round). Those come from two different BLS programs with slightly different methods, so read the trend rather than the exact differences. Teller figures from Bessen's IMF analysis.

So which part of the curve is radiology on? Before I answer that, there is a second half to the cashier story that the employment figures do not capture, and I notice it every single week.

I used to dread Costco. The checkout line was the thing you planned the trip around, the reason you talked yourself out of going. Now I walk through it. The scanners, the app, the reconfigured front end: the experience is dramatically better, for me and honestly for the person working there, who is solving problems instead of dragging four hundred items across a piece of glass. The company benefits, the customer benefits, and the work that remains is more interesting than the work that went.

Both things are true at once. The experience got much better and several hundred thousand of those jobs went away. I am not going to pretend otherwise to make my argument tidier.

But here is the structural difference, and it is the hinge of everything that follows. Retail demand is roughly fixed. There is a certain amount of shopping to be done in a week, so making checkout twice as fast means needing about half as many checkers. Imaging demand is not fixed, and it is nowhere close to met. There are scans sitting unread right now. There are people waiting four months for an MRI. There are patients whose scan will never be ordered at all because the queue makes ordering it pointless, and whose disease will therefore be found later than it should have been.

When you free capacity in a system with a fixed amount of work, you shed people. When you free capacity in a system drowning in unserved need, the capacity gets absorbed. Radiology is the second kind, and that is an argument about demand, not about how special radiologists are.

Which means the question that matters is not where we sit on that curve. It is how big the backlog is. So let me show you.

The constraint is not accuracy. It is capacity.

Almost every public argument about AI and radiology is an argument about accuracy: can the model see the nodule, can it beat the human. That is the wrong axis. Accuracy is not what is failing patients right now.

What is failing patients is that there is not enough of us, and there is more and more imaging.

A study of nearly 136 million imaging examinations across seven US health systems and all of Ontario found that between 2000 and 2016, CT use in older adults rose from 204 to 428 exams per 1,000 person-years, and MRI from 62 to 139. Both roughly doubled. Growth slowed in later years but never reversed. Over a broadly overlapping period, the number of radiologists per capita in this country moved by less than half a radiologist per 100,000 people.

Two curves that were never going to meet. The imaging figures come from Smith-Bindman's 2019 JAMA cohort (2000–2016); the workforce figure from a 2026 JACR analysis (2010–2022). The windows do not match, so read this as two separate trends placed side by side rather than a single like-for-like comparison. It is a picture of direction, not a calculation.

And the people absorbing that gap are not doing well. In a survey of a large coalition of physician-owned US radiology practices, 46% of radiologists met criteria for burnout and only 27% reported professional fulfillment. Taking call was the strongest associated factor. That survey had a 20.6% response rate, which means burned-out people may have been more motivated to answer, so treat the exact number loosely. The direction is not in dispute by anyone who works in a reading room.

In England, a 2026 review reported a 30% shortfall in clinical radiologists, projected to reach 40% by 2028.

This is the actual problem. Not "can a machine see the lesion." It is that scans are sitting unread, MRI slots are months out, and the people reading are running on empty. If you frame AI as a contest for who is the better reader, you have not even engaged with the thing that is hurting patients.

And it is not only radiologists

Step back from imaging and the picture gets genuinely strange, because the thing everyone is afraid of is not the thing that is happening.

The staffing crisis in American health care is not approaching. It is here, it has been here for a decade, and it keeps getting worse. In the ASRT's 2025 staffing survey, the CT technologist vacancy rate reached an all-time high of 19.4%. MRI was 17.4%. Cardiovascular interventional technology, 17.4%. Every imaging discipline surveyed sat above its 2020 level. Two years earlier the radiographer vacancy rate had hit 18.1%, up from 6.2% in 2021. The survey drew 475 department managers, so hold the decimal points loosely, but nobody who runs a department needs a survey to know this.

The jobs are open right now. Bars are zero-based and share one scale. Every discipline in this survey sits above its 2020 rate. From the ASRT 2025 Radiologic Sciences Staffing and Workplace Survey, which collected responses from 475 US radiology department managers. That is a small sample, so treat the decimal points as indicative rather than precise.

Now read that next to the volume figures from earlier in this post. CT use in older adults doubled. The CT technologist vacancy rate is nearly one in five. Those two facts together describe a queue, and the queue is made of people.

Then look at what the government actually projects, from the same 2025–2035 release I have been quoting on cashiers. Private health care and social assistance is projected to add more than 2.2 million jobs, the most of any sector, accounting for roughly 37% of all new jobs in the entire economy. Healthcare support and healthcare practitioners are the two fastest-growing of all 22 major occupational groups, and together they are expected to supply almost a third of every new job created through 2035. Nurse practitioner is the single fastest-growing detailed occupation in the country, at 41%.

In that same release, the group projected to shrink fastest is office and administrative support, shedding 752,100 jobs, the largest decline of any major occupational group. The BLS attributes it in plain language to automation, including AI.

Same document. Same decade. Opposite directions. Bars are drawn to a common scale from zero, so their lengths are directly comparable. The health care figure is reported by BLS as "more than 2.2 million," so the bar is a floor rather than an exact value. Source: BLS Employment Projections, 2025–2035, released 27 August 2026.

One government document, one projection round: office administration sheds three quarters of a million jobs while health care absorbs a third of all the new ones. That is not a story about work disappearing. It is a story about work moving.

I find the public conversation about this disorienting. I read that AI is about to leave people without jobs, and then I go to work in an industry that cannot fill the jobs it already has, where those unfilled positions are the direct reason somebody waits four months for a scan, and where the shortage has been deepening steadily since before large language models existed. The openings are posted. Where are the people?

So the question is not whether there is work. There is an enormous amount of work. The question is whether we are willing to move people toward it.

That requires being honest about which way things travel. Some roles genuinely are easier to automate, and cashiering is the clean case: what the customer needs is a fast, accurate, low-friction transaction, and a machine now delivers that well. Other roles are close to unautomatable on any near horizon, and they are disproportionately in health care. Turning a patient. Getting a difficult IV. Positioning someone who is in pain for a scan without hurting them more. Noticing that a person is frightened and doing something about it. Those are not knowledge tasks with a hands-on component. They are hands-on tasks with a knowledge component, and the order matters.

None of the moving happens by itself. Somebody has to fund the training pipelines, build bridges out of declining occupations into growing ones, and pay for the years in between. Enrollment in radiologic technology programs has been falling while the vacancies climb, which tells you the market signal is not reaching the people who could act on it. A labor market does not clear just because an economist can see that it ought to.

This is the part I would most like people to hear, because the fear is real and it is aimed at the wrong target. The jobs are not vanishing. They are relocating, into work that is harder to automate and, in most cases, more worth doing. What we owe people is the ladder to get there.

What this post does not tell you

Everything above is an argument about demand and labor. It is the argument I most want people to hear, because the fear is real and it is pointed at the wrong target. But it is also, deliberately, an argument that dodges the hardest question.

None of it tells you whether the technology actually works.

A staffing crisis is a reason to want a tool. It is not evidence that the tool is any good, and "we are desperate" is close to the worst frame available for a purchasing decision. There are real randomized trials now, some of them genuinely impressive. There is also a study in which very experienced radiologists went from scoring 82% of mammograms correctly to 46%. The difference was that the AI handed them the wrong answer.

That is the next post.

---

How this piece was built

I structured this on Randy Olson's And, But, Therefore framework, which keeps an argument from collapsing into a list of statistics. Every figure here was checked against the primary source, and where a survey is small or two datasets are not strictly comparable, I have said so in the text rather than hiding it in a footnote.

AI disclosure. The argument, the experience, and the point of view are mine. I worked with Claude (Anthropic) as a research and drafting partner: it verified every number against BLS tables, ASRT survey releases, and peer-reviewed sources, produced the four figures, and helped organize the draft. It corrected two things I had wrong going in, which I have left visible in the text: the 2016 prediction was Geoffrey Hinton's rather than Jensen Huang's, and my cashier figures were from a superseded BLS projection round. I reviewed every claim and citation before publishing.

References

1. US Bureau of Labor Statistics. Employment projections: 2025–2035 summary. USDL-26-1422, August 27, 2026. bls.gov
2. US Bureau of Labor Statistics. Occupations with the largest job declines, 2025 and projected 2035. bls.gov
3. US Bureau of Labor Statistics. Occupational Employment and Wage Statistics, largest occupations, May 2019. bls.gov
4. Bessen J. Toil and technology. Finance & Development (IMF), March 2015. imf.org
5. American Society of Radiologic Technologists. 2025 Radiologic Sciences Staffing and Workplace Survey. asrt.org
6. Smith-Bindman R, et al. Trends in use of medical imaging in US health care systems and in Ontario, Canada, 2000-2016. JAMA. 2019;322(9):843-856. doi:10.1001/jama.2019.11456
7. Malhotra A, et al. The evolving US radiologist pipeline: trends in residency positions, resident workforce, and practicing radiologists per unit population. J Am Coll Radiol. 2026;23(8):1587-1592. doi:10.1016/j.jacr.2026.04.005
8. Parikh JR, et al. Prevalence of burnout of radiologists in private practice. J Am Coll Radiol. 2023;20(7):712-718. doi:10.1016/j.jacr.2023.01.007
9. Spalding A. A retrospective mixed-methods service evaluation of radiographer-led adult nephrostomy exchange service. Radiography. 2026;32(4S1):103316. doi:10.1016/j.radi.2025.103316

Peer-reviewed sources were located through PubMed and Consensus. Nothing in this post describes any individual patient or protected health information, and nothing here represents the position of my employer.