No-Show Management

Overbooking Without a Model Is Just Guessing

A primary care clinic schedules 20 patients per provider per day. The historical no-show rate is 20%. On an average day, 4 patients do not appear. That is 4 empty slots — roughly 80 minutes of provider time wasted, 4 patients who could have been seen but were not. Annualized across 250 workdays, a single provider loses approximately 1,000 patient encounters per year to no-shows. At a blended reimbursement of $120 per visit, that is $120,000 in lost revenue and, more importantly, 1,000 patients who needed care and did not receive it.

The operational instinct is to overbook — schedule 24 or 25 patients to compensate. But overbooking without a calibrated model is a coin flip between two failure modes: too few patients show up (wasted capacity persists) or too many show up (overtime, cascading delays, patient dissatisfaction). No-show management is not a scheduling preference. It is a probability calibration problem.

No-Show Rates Are Not One Number

No-show rates vary by setting, population, and visit type. Treating them as a single clinic-wide statistic destroys the information needed to manage them.

Primary care: 5-20%, with most practices in the 15-18% range. New patient visits no-show at higher rates than established patients. Monday morning and Friday afternoon slots run higher than midweek.

Behavioral health: 20-40%, and frequently higher for initial intake appointments. Gallucci et al. documented initial-appointment no-show rates exceeding 50% in community mental health settings. The abandonment page (02-abandonment-and-access) covers why: long wait-to-appointment times drive no-shows monotonically upward. A patient who waits 4 weeks for a therapy intake is not twice as likely to no-show as one who waits 2 weeks — they are often three or four times as likely.

Specialty care: 10-25%, varying by specialty. Dermatology and ophthalmology tend toward the low end; psychiatry and pain management toward the high end. Referral-based specialties carry the additional burden of referral drop-off — the patient never scheduled at all, which is a different phenomenon than scheduling and not appearing.

Medicaid populations: Across all settings, Medicaid-insured patients no-show at rates 1.5-2x those of commercially insured patients. This is not a compliance problem. It is a transportation, childcare, employment flexibility, and communication access problem. Interventions that treat it as non-compliance (discharge from panel, cancellation fees) punish the population with the fewest alternatives.

The Overbooking Decision

The question is: given a no-show rate of p, how many patients should you schedule per slot?

This is a newsvendor problem — the same structure that governs how many newspapers to stock when demand is uncertain (Morse and Kimball, 1951). You balance two costs:

Cost of an empty slot (Cu): lost revenue, unserved patient, idle provider time.
Cost of an overbooked slot (Co): overtime pay, extended wait times, rushed visits, patient complaints, clinician burnout.

The optimal overbooking level occurs where the marginal probability of a no-show equals the cost ratio Co / (Co + Cu). When the costs are symmetric — an empty slot is roughly as bad as an overbooked one — the simple rule holds: schedule 1 / (1 - p) patients per slot.

At a 20% no-show rate, that yields 1 / 0.8 = 1.25 patients per slot. In practice: for every 4 scheduled slots, double-book one. Twenty slots become 25 scheduled patients, expecting 5 no-shows and 20 arrivals.

The Bailey-Welch rule (Bailey, 1952; Welch, 1964) refines this: front-load the overbooking. Schedule 2 patients in the first slot, then 1 per slot thereafter. The logic is that provider idle time at shift start is pure waste — there is no way to recover it. By front-loading, you ensure the provider begins working immediately even if one of the first two patients no-shows. Late in the session, an empty slot matters less because the provider may be running behind anyway.

LaGanga and Lawrence (2007) extended this analysis to outpatient clinic settings, demonstrating that overbooking policies calibrated to patient-level no-show probabilities outperform uniform rules by 15-30% on measures combining revenue, overtime cost, and patient wait time.

Why Uniform Overbooking Fails

A blanket 25% overbooking across all slots, all days, and all patient types is a blunt instrument applied to a problem with fine-grained structure.

No-show probability varies by:

Patient history. A patient who has no-showed twice in six months has a predicted no-show probability 3-5x higher than a patient with perfect attendance. This is the single strongest predictor.
Day and time. Monday morning and Friday afternoon slots carry higher no-show rates in most practices.
Appointment type. Follow-ups no-show more than new patient visits in some specialties, less in others. Procedure appointments (where the patient has prepared — fasted, arranged transport) no-show less than routine visits.
Wait time to appointment. The longer the gap between scheduling and the visit, the higher the no-show rate. Same-day and next-day appointments no-show at a fraction of the rate of appointments booked 3+ weeks out.
Weather and season. Not trivial: winter storms in rural areas can spike no-show rates by 10+ percentage points.

Predictive no-show models use these patient-level and appointment-level features to generate individualized no-show probabilities. Daggy et al. (2010) demonstrated that logistic regression models using patient demographics, appointment history, and scheduling lead time significantly outperform blanket overbooking rules. Zacharias and Pinedo (2014) formalized the appointment scheduling problem with heterogeneous no-show probabilities, showing that sequencing patients by predicted no-show probability — placing high-risk no-shows earlier in the schedule — reduces both expected idle time and expected overtime.

The practical version: double-book the slots assigned to patients with high predicted no-show probability. Leave slots for reliable patients alone. This concentrates the overbooking where capacity loss is most likely and avoids creating chaos where it is not.

A Behavioral Health Example

A community mental health center sees 16 patients per therapist per day across 4 therapists. The overall no-show rate is 30%. The clinic applies uniform 30% overbooking: 21 scheduled per therapist, expecting ~15 arrivals.

The problem: no-show rates are not uniform. New intake patients no-show at 45%. Established weekly patients no-show at 12%. Patients rescheduled after a previous no-show have a 55% probability of no-showing again.

Under uniform overbooking, days heavy with established patients see 19-20 arrivals against 16 slots — overtime, compressed sessions, burned-out therapists. Days heavy with intakes see 11-12 arrivals — wasted capacity. The variance in daily arrivals is high and the schedule absorbs it entirely through staff suffering or empty rooms.

A differentiated approach: double-book intake slots and high-risk reschedule slots. Leave established-patient slots at 1:1. The expected daily arrivals stabilize closer to 16 with lower variance. Overtime drops. Idle time drops. The same capacity serves more patients with less chaos.

Product Implications

Surface no-show rates by segment, not clinic-wide averages. The aggregate number hides the actionable variation. A scheduling product should display no-show rates by visit type, provider, day-of-week, and patient risk tier.

Embed predictive no-show scoring. Even a simple model (prior no-shows, days-until-appointment, visit type) outperforms blanket rules. The scheduling system should flag high-risk appointments and suggest overbooking those specific slots.

Track overbooking consequences, not just overbooking levels. Measure overtime hours, patient wait-time variance, and same-day complaints alongside no-show rates. Overbooking is only successful if the net effect is positive. A system that overbooks aggressively but does not measure the downstream damage is flying blind.

Earliest degradation metric: the ratio of actual arrivals to scheduled slots, tracked daily. When this ratio drifts away from 1.0 in either direction — consistently above (overtime signal) or consistently below (capacity waste signal) — the overbooking calibration is off. This metric moves before financial results, patient complaints, or staff turnover make the problem visible.

Quick Diagnostic

Five questions an operator should answer about their no-show management:

Do you know your no-show rate by visit type and patient segment, or only the clinic-wide average? If only the average, you are managing a distribution with a single number and losing the signal.
Is your overbooking rule calibrated to a model, or is it a round number someone chose years ago? “We overbook by 20% because we always have” is not a policy. It is inertia.
Do you track what happens when too many patients show up? Overtime hours, wait-time spikes, and compressed visit lengths are the cost of overbooking. If you do not measure them, you cannot balance the equation.
Do you use patient no-show history when scheduling? A patient who has no-showed 3 of 4 recent appointments has a different optimal scheduling strategy than a patient with 100% attendance. Treating them identically wastes information.
Do you know the relationship between scheduling lead time and no-show rate in your own data? If same-day appointments no-show at 5% and appointments booked 3 weeks out no-show at 30%, your access model should favor shorter booking horizons — which connects directly to open-access scheduling design.

Warning Signs

Blanket overbooking percentage with no segment differentiation. The same overbooking rule applied to Monday morning intakes and Wednesday afternoon follow-ups is guaranteed to be wrong for both.
No measurement of no-show rate by population, visit type, or time slot. You cannot manage what you do not stratify.
No tracking of overbooking consequences. If the clinic tracks no-shows but not overtime, wait-time variance, or patient complaints on overbooked days, it is optimizing one side of the ledger.
No-show “policy” that consists entirely of patient-facing penalties. Cancellation fees and panel discharge address patient behavior while ignoring system design. The no-show rate is substantially a system property — set by wait times, access barriers, and reminder systems — not purely a patient choice.
Provider idle time attributed to “bad luck” rather than analyzed as a scheduling design problem. Idle time is the signal that overbooking is undercalibrated or misallocated.

Integration Hooks

Module 2: Abandonment and Access. No-shows are a specific form of pre-service abandonment — the patient entered the queue (scheduled an appointment) and abandoned before service (did not appear). The mechanisms described in the abandonment page apply directly: wait-to-appointment time drives no-show probability through the same patience-threshold dynamics that drive LWBS, referral drop-off, and prescription abandonment. Managing no-shows without understanding the abandonment mechanism treats a symptom while ignoring the cause. Reducing wait-to-appointment is often the most powerful no-show intervention — more effective than reminders, penalties, or overbooking.

Human Factors Module 4: Decision Under Uncertainty. Why patients no-show is partly a behavioral economics question. Present bias (valuing today’s competing demands over a future appointment), status quo bias (inertia toward not acting), and intention-action gaps all contribute. Reminder systems work not because patients forget but because reminders re-activate intention at the moment when a decision to attend or skip is being made. Effective no-show interventions account for both the system dynamics (wait times, access barriers) and the decision architecture (reminders, commitment devices, friction reduction).