Privacy and preparation
Healthcare no-show and reschedule datasets: defining each outcome
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
Label each appointment from its event history, not its final status. A no-show is a booked visit the patient neither attended nor cancelled; a cancellation ends the visit before it starts, split into late or timely by the clinic's policy window; a reschedule is a cancellation linked to a replacement appointment. Record who acted, when, under which policy version, and keep clinical detail out of the dataset.
Redaction
Healthcare no show
Key takeaways
- The final status in a scheduling system merges different events and is often overwritten; rebuild the timeline from status changes and messages.
- Three facts decide the label: initiator, lead time to the scheduled start and a linked replacement appointment.
- Late-cancellation windows vary by clinic and over time, so store the policy version with every label.
- Attendance evidence overrides a later no-show status, and voided records leave the dataset.
- Keep only scheduling fields; diagnosis, visit reasons and free-text notes stay out, and dates need a privacy reviewer's decision.
How do you tell a no-show from a cancellation or a reschedule in scheduling data?
Decide the label from three facts, not from the last status in the system: who ended the appointment, when they did it relative to the scheduled start, and whether a linked replacement appointment was booked. A no-show is a booked appointment the patient did not attend and never cancelled; a cancellation ends the appointment before it starts; a reschedule is a cancellation plus a new appointment for the same need, linked to the old one.
Those three facts live in the appointment's event history: interface messages, audit log entries and the timestamps on status changes. A healthcare no-show and reschedule dataset built from that history, with the definitions and reconciliation rules below, gives a model labels that mean the same thing across clinics and years.
Everything here is administrative. The task is scheduling operations, not care decisions, and the records it needs are scheduling records, not clinical ones.
Why the final appointment status is not enough
The final status collapses different events into one value and sometimes overwrites the event that matters. The HL7 FHIR interoperability standard defines statuses such as booked, arrived, checked-in, fulfilled, cancelled, noshow, waitlist and entered-in-error, and its noshow code means some or all participants, usually the patient, did not appear.
There is no rescheduled status. FHIR represents the link through a replaces reference on the new appointment, and notes that the cancellation details sit on the replaced appointment's cancellationReason. Its cancellationDate keeps only the most recent cancellation, so an appointment cancelled, restored and cancelled again shows one date. Under the R5 rules a cancellation reason is only valid on cancelled or noshow appointments.
Older interfaces carry the same distinctions as separate events. The HL7 version 2 scheduling messages include S13 for a notification of rescheduling, S15 for a cancellation and S26 for a notification that the patient did not show up. When a practice management system converts no-shows to cancellations at day end, or books a reschedule without linking it, only the event stream still holds the truth.
The administrative outcome definitions
These nine labels cover the outcomes most scheduling teams act on differently. Each requires specific evidence; a record without it gets the unresolved label rather than a guess.
| Label | Definition | Evidence required | Common mislabel |
|---|---|---|---|
| Attended | Patient arrived and the visit took place | Arrival or check-in event, or a linked encounter | Marked no-show when check-in happened in another system |
| No-show | Booked, not cancelled before start, patient did not arrive | Booked status at start time and no arrival or encounter | Same-day cancellations counted as no-shows without a stated rule |
| Late cancellation | Patient cancelled inside the policy window | Cancel timestamp, initiator, policy window version | Treated as a timely cancellation when the window is unknown |
| Timely cancellation | Patient cancelled before the policy window | Cancel timestamp and initiator | Merged with organization cancellations |
| Organization cancellation | Clinic, provider or system ended the appointment | Initiator coded as staff, provider or system | Counted against the patient in attendance metrics |
| Patient reschedule | Patient-initiated cancellation with a linked replacement | Cancel event plus a new appointment linked to it | Counted as a cancellation and a separate new booking |
| Organization reschedule | Clinic moved the appointment and booked a replacement | Staff or system initiator plus linked replacement | Treated as a patient reschedule |
| Voided in error | Appointment should never have existed | Entered-in-error status or a void audit entry | Left in the data as a cancellation |
| Unresolved | Evidence conflicts or is missing | Logged reason for non-resolution | Silently defaulted to the final status |
Who sets the late-cancellation window, and why it must travel with the label
The clinic's own policy sets the window, and it changes, so store the window version with each late-cancellation label. Research shows how much the cutoff varies: a primary care machine learning study defined a late cancellation as one made within 24 hours of the appointment, while a study of young people with type 1 diabetes treated cancellations within 8 days as last-minute and separate from no-shows.
The label changes the workflow, which is why it matters. A no-show may trigger outreach, a missed-appointment charge under the practice's own policy, or a different overbooking rule; a late cancellation may open a slot for the waitlist. CMS's claims processing manual allows physicians and suppliers to charge Medicare beneficiaries for missed appointments only if they charge non-Medicare patients under the same policy, and Medicare does not pay those charges. A dataset that mixes the two labels teaches a model the wrong next step.
Record the policy as data: window length, whether it counts calendar hours or business days, the effective dates and which appointment types it covers. Then compute each label against the policy in force on the appointment date.
The reconciliation guide: from event history to one label
Run the steps in order for every appointment, and keep the output of each step so a reviewer can see why a label was chosen.
At a glance
The reconciliation guide: from event history to one label
Before the labels are used, test the procedure itself. Draw a random sample of reconciled appointments from each department and year, have two people who know the scheduling workflow label them from the raw history without seeing the automated label, and compare. Disagreements usually point to a missing rule, such as an unlinked reschedule pattern or an initiator code that means different things in two systems. Fix the rule, rerun, and record the share of appointments that end as unresolved; a rising share after a system change is the earliest sign that a new interface broke the evidence trail.
- Build the timeline: collect every status change, interface message and audit entry for the appointment, with timestamps normalized to the clinic's local time zone.
- Find replacements: link a new appointment when the system records the link (a replaces or previous-appointment reference, or a rescheduling message). Without a recorded link, use a stated rule such as same patient, same department and service line, created within a short interval after the cancellation, and flag those links as inferred.
- Identify the initiator: patient, staff, provider or system, from the cancel reason code, the user role on the audit entry or the message source. Free-text notes are a last resort and are reviewed, not parsed blindly.
- Compute lead time: minutes between the cancel timestamp and the scheduled start.
- Apply the policy version in force on the appointment date to separate late from timely cancellations.
- Check attendance evidence: an arrival, check-in or linked encounter overrides a noshow or cancelled status that was set later.
- Apply precedence: voided in error wins over everything; attendance evidence wins over status; a recorded link wins over an inferred one.
- Label or log: assign one label, or assign unresolved with the reason (for example, missing initiator or two conflicting arrival records).
Worked example (illustrative): four appointments, four different labels
These four appointments are invented to show the rules; no real patient or clinic is described. The clinic's policy at the time defines a late cancellation as one made less than 24 hours before the start.
Three of the four final statuses read cancelled, yet they describe a patient who moved a visit, a clinic that lost a provider for the day and a patient who dropped out at short notice. A model trained on the raw status would treat them as one behavior. The second row is the exception worth remembering: a no-show status set at the end of the day lost to a check-in recorded in the morning, because the status was a bulk end-of-day update while the check-in was a direct observation.
| Event history | Final status in system | Reconciled label | Rule that decided it |
|---|---|---|---|
| Booked; patient cancels by text 30 hours before; new booking 6 minutes later, linked | Cancelled | Patient reschedule | Recorded link plus patient initiator |
| Booked; status set to no-show at 5 pm; check-in event exists from 10:02 am | No-show | Attended | Attendance evidence overrides a later status |
| Booked; staff cancel 2 hours before because the provider is out; no replacement within 14 days | Cancelled | Organization cancellation | Staff initiator, no replacement |
| Booked; patient cancels 3 hours before; no replacement | Cancelled | Late cancellation (policy v2) | Lead time under the window in force |
Which administrative evidence is necessary, and which clinical detail stays out?
Keep only the fields the scheduling task needs, and leave diagnosis, clinical notes and visit reasons out. HHS's minimum necessary guidance asks covered entities to make reasonable efforts to limit uses, disclosures and requests of protected health information to what the purpose needs, and the scheduling outcome task needs very little.
Dates are the hard part. HHS's de-identification guidance says that under the Safe Harbor method, elements of dates more specific than the year that relate to an individual must be removed, while the Expert Determination method relies instead on a qualified expert applying statistical or scientific methods. Lead times, day-of-week and hour-of-day offsets often carry the signal a model needs, and whether they can be kept is a question for a qualified privacy reviewer, not a default.
- Keep: appointment and replacement identifiers (tokenized), scheduled start offset, status events with relative timestamps, initiator role, coded cancel reason, department or service line, appointment type, booking channel, reminder events, policy window version.
- Drop or generalize: names, contact details, record numbers, exact dates, provider names, free-text cancel notes, visit reason text, diagnosis and procedure codes, insurance member identifiers.
- Review case by case: small departments or rare appointment types, where a service line alone could point to a condition or a person.
Edge cases that break simple rules, and when to stop
Some appointments do not fit a single patient, a single slot or a single cancel event. Handle them explicitly rather than letting the default rules decide.
At a glance
Edge cases that break simple rules, and when to stop
- Group visits: FHIR's noshow covers some or all participants, so record attendance per participant instead of one label per appointment.
- Multi-resource bookings: a room or device cancellation can end the appointment for reasons the patient never saw; code the initiator as organization.
- Bulk reschedules: when a provider's whole day moves, label each as an organization reschedule and keep the batch identifier so a model does not learn that the patients chose to move.
- Telehealth: a patient who never joined a video visit is a no-show only if the platform's join log is part of the evidence; otherwise mark unresolved.
- Stop and request specialist review when cancel reasons are mostly free text that mentions symptoms or conditions, when the initiator cannot be determined for a large share of cancellations, or when the records sit under a business associate agreement or data use terms that do not clearly allow this use. A healthcare administrative expert and a qualified privacy reviewer should approve the field list before any extraction.
Can scheduling outcome records be licensed?
Scheduling outcome histories are useful to AI developers working on access, reminders and capacity planning, and they are only useful when the labels hold up. A provider organization licensing scheduling records through SourceX gets rights review, contracting and preparation as part of the process, with the de-identification itself handled by SourceX through approved processing routes.
Privacy permission, ownership of the records and the contractual right to license them are separate questions, and removing identifiers does not by itself create a right to license. The organization would approve the scope and sign the agreement ahead of any delivery, having first checked its eligibility in a brief assessment that involves no uploaded data.
Frequently asked questions
Is a same-day cancellation a no-show?
Not by definition. A no-show means the patient never cancelled and did not arrive; a same-day cancellation is a late cancellation under most policies. Some clinics and studies count same-day cancellations with no-shows for reporting, so store the raw label and let the reporting rule combine them.
Does FHIR have a rescheduled status?
No. FHIR records a reschedule as a cancelled appointment plus a new appointment whose replaces element points to it. HL7 version 2 interfaces send a separate rescheduling message, S13, which helps recover the link in older systems.
Who counts as the initiator when a reminder system cancels an appointment?
The patient, if the patient replied to the reminder to cancel; the organization, if the system cancelled on its own, for example for an unconfirmed booking. Keep the channel as a separate field so both facts survive.
Can a no-show label carry billing consequences?
It can. A practice may charge for missed appointments under its own policy, and CMS allows charging Medicare beneficiaries only when non-Medicare patients are charged under the same policy; Medicare does not pay those charges. That is one reason the label must be accurate before it drives any workflow.
How many years of scheduling history are worth keeping?
Enough to cover at least one change of policy or system, so a model sees how labels behave across versions. Each year must carry its policy version and system of record, or the history mixes definitions.
Sources
- HL7 FHIR defines appointment statuses including booked, arrived, checked-in, fulfilled, cancelled, noshow, waitlist and entered-in-error; noshow means some or all participants, usually the patient, did not appear. Source
- FHIR's Appointment resource links a replacement through replaces, keeps the cancellation details on the replaced appointment, records only the most recent cancellationDate and allows cancellationReason only on cancelled or noshow appointments. Source
- The HL7 version 2 event table lists S13 notification of appointment rescheduling, S15 notification of appointment cancellation and S26 notification that the patient did not show up. Source
- A primary care machine learning study defined a late cancellation as a cancellation within 24 hours before the appointment. Source
- A study of teens and emerging adults with type 1 diabetes defined last-minute cancellations as those within 8 days of the appointment, separate from no-shows. Source
- CMS's Medicare Claims Processing Manual allows charging Medicare beneficiaries for missed appointments when non-Medicare patients are charged under the same policy, and Medicare does not pay those charges. Source
- HHS guidance says covered entities must make reasonable efforts to limit uses, disclosures and requests of protected health information to the minimum necessary for the purpose. Source
- HHS de-identification guidance says Safe Harbor removes date elements more specific than the year, and names Expert Determination as the other method. Source
See if your company qualifies
A short company assessment. No data uploads are needed.