Resident requests can't be an afterthought

Most residency programs collect vacation requests, rotation preferences, and pairing asks every year. The process usually looks familiar: a form goes out, emails trickle in, and someone later tries to honor as many of those asks as the draft will allow. That sequence treats resident requests as cleanup—work done after coverage, capacities, and program rules already locked the shape of the year.

Building any feasible schedule by hand is hard but often doable. Building a schedule that is also fair, rule-compliant, and responsive to what residents asked for is a different problem. Preference collection only helps if those preferences are part of how the schedule is built, not a final pass after the hard decisions are already made.

Why requests get squeezed into a finished draft

Program coordinators and chiefs are under real pressure: services need coverage, clinics need continuity, and ACGME-related duty and educational requirements cannot slide. It is natural to stabilize those constraints first and leave vacation windows and soft preferences for later. Forms and inboxes make that habit worse. Requests arrive as unstructured notes—“block 7 off if possible,” “please pair me with X on nights”—while the working schedule lives in a spreadsheet that already feels fragile.

Once a draft exists, every granted request becomes a local edit. Move one resident for a wedding, and another loses a preferred elective. Approve a hard-rotation pairing, and a third service dips below capacity. The backlog of asks turns into a negotiation about who is flexible enough to absorb the change, not a coherent plan that weighed those asks alongside the rules.

The search space is larger than intuition suggests

Part of why late-stage request fitting fails is scale. Even a mid-size program faces an enormous combinatorial space. Roughly eighty residents, twenty-six two-week blocks, and ten rotations produces on the order of 8×10²⁷ possible schedules—comparable to the number of atoms in a human body. That figure is not a claim that every assignment is equally likely or equally valid. It is a reminder that “just tweak the draft” is searching a space too large for manual trial and error to explore well.

Humans are good at building one workable schedule. They are poorly equipped to compare millions of near-feasible alternatives to find ones that also honor vacation timing, friendship pairings on difficult months, and equity caps across the cohort. When preferences enter only after a single draft is frozen, most of that search space is already closed. The remaining edits are the ones that do not break coverage—not the ones that best balance what residents asked for.

Treat preferences as scheduling inputs

A better framing is to treat resident requests as first-class inputs next to program rules, service capacities, and educational constraints. Vacation in a specific block, avoiding a blackout month, or working a hard rotation with a friend are not polish items. They are preferences with different strengths that should compete—transparently—with coverage floors and fairness rules while the schedule is still flexible.

That does not mean every ask is granted. It means denials and partial grants come from a process that saw the tradeoffs early: this vacation conflicts with clinic continuity; that pairing is possible only if another resident takes an extra night stretch; these three requests cannot all be true under current caps. Residents and faculty understand “we optimized under constraints” more readily than “we ran out of room in an already published grid.”

Operationally, that shift looks like structured preference collection—typed request kinds, blocks or date ranges, and priority where it matters—fed into the same planning pass that enforces capacities and ACGME-related guardrails. Tools built for residency block scheduling, including Oso, exist to keep those inputs in one model so requests are weighed while alternatives still exist, not after the draft is treated as final.

Resident requests cannot be an afterthought if fairness and trust matter. Collect them with enough structure to use them, put them beside the rules that actually constrain the year, and build the schedule once with the full picture in view. Manual cleanup will always leave someone feeling like their ask arrived too late. Designing requests into the process is how programs avoid that outcome.