Diagnosing System Failures Without the Authority to Fix.
I’ve been burnt out since I started working as a physician assistant. For a while, I assumed I was the problem: not efficient enough, bad boundaries, too slow with paperwork. Maybe I needed thicker skin.
Here’s the thing: I can only control what happens in the room with each patient. That’s it. Not who shows up. Not how long I have with them. Not whether I’m cleaning up someone else’s mess. Just what happens when the door closes and it’s the two of us figuring something out.
It’s not a skill problem. It’s structural. Healthcare operates at three layers: patient care, workflows, and organization. I’m trained for the first. I patch the second to survive. But the third is where everything actually breaks, and I have zero authority to fix it.
Who Absorbs the Friction?
I work in primary care. On paper, I see patients and manage their health. In practice, I’m the cleanup crew for everything the system failed to do right the first time.
Patient hasn’t been seen in over a year with multiple chronic conditions, but their doctor is booked three months out? That’s a 20-minute slot for what needs 45 minutes, plus chart prep. Patient on twelve medications from six specialists who don’t talk to each other? I untangle it. Prior authorization for a med they’ve been on for years because insurance changed their formulary? I write the appeal.
The system generates friction. Someone absorbs it. In primary care, that’s me.
I can see exactly what’s broken and why. But I can’t fix it because fixing it requires organizational authority, and I’m a physician assistant working in an industry famous for bureaucracy, politics, and profit. My control ends at the exam room door.
The Workarounds
Here’s my actual workday:
20 minutes for what needs 40. Complex patients get the same slot as someone with a cold. So I either run late and make everyone wait, or I do substandard work. I’ve gotten really good at triage: what must happen today, what can wait, what I can handle later via message. It’s a skill. It’s also a workaround for broken scheduling.
Zero continuity. I see patients I’ve never met for ‘follow up’ all the time, sometimes because patients want to be seen ASAP, but more often because a call center scheduler doesn’t understand that continuity matters. Then my own patients get scheduled with other doctors when I’m not available. It’s inefficient and alienating for everyone.
Bureaucratic nonsense. Bending over backward to head off insurance denials. Building shortcuts to satisfy billing and quality rules. Verifying if specialists are in-network, then watching patients get triaged out 12 months. Writing sick notes for adults who shouldn’t need permission to stay home. I see patients who need 5 minutes of medical care and 15 minutes of HR paperwork.
Feedback that goes nowhere. I have three supervisors who want to hear about problems. But only if it’s about fixing someone’s individual workflow, and I’d better have screenshots, and be ready to solve it myself. Anything systemic? “Maybe you’re just not suited to primary care.” We never acknowledge over-empanelment, illogical scheduling guidelines, or our inability to provide actual longitudinal care.
The Clinical Informatics Detour
I thought clinical informatics was the answer. I’m interested in information flow, workflow breakdowns, how data could improve care instead of just satisfying billing. I even got my Epic Builder certification to work directly on our EHR.
But clinical informatics is a bandaid. Most roles stay clinically tethered. You get influence without authority, extra responsibility without organizational power, extra hours with no extra pay. And if they reduce your clinical hours for admin work, they cut your salary because the admin stuff doesn’t bill. We’re salaried, no overtime, no comp time, so you just work more hours for less money.
It optimizes within existing constraints and charges you for the privilege. I want to question the constraints.
Data science is about testing whether the model itself is broken. Does this scheduling system actually work? What does fragmentation cost? What’s the downstream impact of poor continuity? It doesn’t just make workflows smoother, it evaluates whether we’re designing things correctly.
The Real Constraint
I can recommend. I can’t redesign. I can document. I can’t restructure. I can’t set policy, adjust staffing, or change incentives.
The problems I care about are organizational. My authority is individual. That mismatch is my burnout.
I need different authority. The kind that works on system problems without absorbing patient distress all day. Where being tired means I thought hard, not that I absorbed eight hours of suffering I couldn’t adequately address.
But it has to be supplementary. I don’t want my identity tied to a role where my impact is one patient at a time, at the expense of my mental health.
I want to work on systems. Design interventions that change how care is delivered, not just how clinicians cope with broken delivery. Do analysis that informs decisions, not the kind that gets filed under “nice to know.”
Why Data Science
Scale. In clinical medicine, I help one patient at a time. Data science changes that. One good analysis can shape decisions affecting thousands of patients, multiple clinics, entire systems. Instead of repeatedly compensating for design flaws in the exam room, I can help change the conditions that produce them.
Actual authority. If I can show a workflow creates measurable friction, quantify the cost, and propose a testable fix, organizations act on that. Not because I’m louder, but because I’m providing information decision-makers need and can’t get elsewhere.
Sustainability. Clinical work is emotionally intensive by design. Data work is cognitively intensive. I want to be tired from wrestling with hard problems, not from absorbing preventable distress. Deep engagement with complex systems, measured in projects and outcomes, not appointment slots.
The Plan
Self-study for a few months: statistics, Python, SQL. Then six months on an online analytics course in my off hours. After that, I’ll apply to a top online data science master’s program that won’t require quitting my job, but rigorous and credentialed enough to open doors. I’ll build a portfolio: workflow analytics, data visualization, maybe some predictive modeling. Things that prove I can do the work, not just talk about it.
And I’ll dial back clinical hours. Per diem work will keep me clinically sharp, connected to patient care, and financially stable while I figure out if this actually works. But not so much as to be back absorbing system friction full-time.
The Pattern
This isn’t just me. Skilled clinicians constantly diagnose system failures they can’t fix because the intervention is outside their authority. They burn out. They leave. They get cynical.
Healthcare asks clinicians to absorb system failure and calls it resilience.
Administrators respond with: “We value your feedback. Here’s a wellness seminar.” Sometimes the wellness seminars are really nice, but they’re not: “Here’s authority to redesign workflows.”
Coping isn’t solving. If your only tool is individual resilience, you’ll interpret every system failure as personal failure.
What’s Next
Am I running toward something or away from something? Both. Away from absorbing friction I can’t fix. Toward work that addresses root causes instead of symptoms. And all of this is before AI starts ‘helping’ by generating even more documentation requirements, so away from a role where AI automation is helpful now, but will probably turn the screws harder very soon. Toward the rooms where shaping that automation is actively being discussed.
Either way, I’m not leaving because I stopped caring about patients. I’m stepping out of the exam room to work on the problems that keep breaking it.
TL;DR
I’m a PA leaving primary care for data science. Not because I don’t care about patients, but because I’m tired of diagnosing system failures I have no authority to fix. Clinical informatics seemed like the answer but it’s just more responsibility without organizational power. Data science gives me the tools to actually address root causes instead of repeatedly compensating for the same design flaws.