> ## Content Index
> Fetch the complete content index at: https://www.melaninblissmedia.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# A No-Show Is a Data Field Before It Is a Scheduling Rule
- URL: https://www.melaninblissmedia.com/a-no-show-is-a-data-field-before-it-is-a-scheduling-rule/
- Published: 2026-11-17T12:01:37.000Z
- Updated: 2026-11-17T12:01:37.000Z
- Description: A missed appointment gets coded, the code trains a prediction, and in some systems the prediction decides who is booked into a crowded slot. What the research says about who waits longer, and what actually brings patients in.
- Author: Amber McClendon
- Tags: Health, Hospital Policy, The Back Office of Care, #TheBackOffice

A no-show is a field in a database. Someone at the front desk, or a script in the scheduling software, marks a visit as missed. After enough visits are marked, the field stops being only a record of what happened. In some scheduling systems it becomes the input to a prediction about what will happen next, and the prediction can become a rule about who is booked into which slot.

This article follows that path. It goes to a place the usual access coverage does not. MB’s earlier Back Office of Care pieces ask who owns the next step after a visit. This one asks what happens before the visit, inside the schedule itself, and whether a practice’s way of coping with missed appointments changes how long some patients wait once they arrive.

The thesis is narrow. A missed visit is recorded as a no-show, the record trains a prediction, and the prediction can be turned into an overbooking rule, so patients who are most likely to miss visits for reasons a practice could address can also be the patients most likely to wait longest once they get there. The evidence is specific to one clinic’s data and to certain scheduling methods, and the sections below say where it stops. This article does not investigate any clinic.

| **AN ILLUSTRATIVE SCENARIO (NOT A REAL PERSON)** This is a composite built to show the mechanics. It is not someone MB interviewed, and it does not describe an actual clinic or software product. A clinic session runs from 9:00 to 11:00 with four 30-minute slots. Five patients are booked, so one slot holds two people. The scheduling software has estimated each patient’s chance of showing up, and it has placed the two patients with the lowest estimates in the doubled slot and the slot right after it. One of the two in the doubled slot has missed two visits this year because a bus route changed and a new shift was added at work. Both patients arrive. One of them waits thirty minutes, and the patient booked in the next slot may wait too. No one in this story had to make a biased decision. Each step, from the missed-visit record to the estimate to the slot, followed a rule. The sections below ask who writes those rules and what the research says about their effects. |
| --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

## A no-show is a label, and the label varies

Studies do not agree on a single definition. A systematic review of missed diabetes outpatient appointments, covering 34 studies, found that the definition of non-attendance varied \[14\]. A 2026 systematic review of primary care predictors included 27 observational studies and more than 13 million appointments. It reported non-attendance ranging from about 5% to 31% and noted that differences in definitions and analyses meant the results had to be synthesized narratively \[8\]. A Pennsylvania study of 15 family medicine clinics counted no-shows, 6.9% of 1,118,236 appointments, separately from late cancellations, 6.8% \[4\].

The code also records nothing about the reason. At Penn Medicine, a text-message survey of 186 patients who had missed appointments found that the most common answer was “I did not know I had an appointment,” given by 22% \[13\]. That is a small survey reported by the health system that ran the program. The code is entered the same way whatever the cause.

## What the models learn from

A prediction model can only use fields that exist. The primary care review lists the predictors that recurred across its 27 studies. Patient-level factors included younger age, socioeconomic disadvantage, minority status in North American studies, mental health burden, and lower literacy or greater social needs. Appointment-level factors included prior non-attendance and a longer time between booking and the visit. System-level factors included transportation barriers and scheduling friction \[8\].

Single-system models show the same pattern with numbers attached. In the Pennsylvania study, schedule lead time was the most important predictor of missed appointments \[4\]. In a rural Wisconsin health system with 1,260,083 appointments from 263,464 patients, the overall no-show rate was 6.0%. It was 7.7% for appointments booked more than 60 days ahead, 11.8% for patients aged 21 to 30, and 2.9% for patients over 60 \[6\]. In 11 federally qualified health center facilities in New York City, the strongest predictor was whether a patient was assigned to an empaneled provider, followed by lead time \[9\]. In a safety-net endoscopy unit, patients without an activated patient portal had no-show rates about 5 percentage points higher across the range of predicted risk \[7\].

Several of those inputs describe the practice as much as the patient: how far ahead it books, whether it assigns a clinician, whether it can reach a person through a portal. A model that scores a patient on them is partly scoring the patient’s contact with the system. That is MB’s reading of the lists, not a finding from any single study.

Accuracy is also less settled than a single statistic suggests. A radiology group trained models on 4,546,104 appointments and reported cross-validation scores of 0.77 to 0.93 on the area under the curve. When the best model was tested prospectively for six weeks at one outpatient location, with 123 no-shows, the score was 0.73 \[5\]. In the Wisconsin model, a cut-off of 0.4 gave a sensitivity of 0.71 and a positive predictive value of 0.18 \[6\]. By definition, that means about 18 of every 100 appointments flagged at that cut-off were actual no-shows. A model can rank patients usefully and still be wrong about most of the individuals it flags.

## From prediction to rule: predictive overbooking

Overbooking means giving the same appointment time to more than one patient on the expectation that some will not come. A 2022 paper on the ethics of these systems explains the method used in what it calls state-of-the-art scheduling. A machine learning component predicts each patient’s probability of showing up. A separate optimization component then assigns patients to slots to minimize a weighted sum of patient waiting time, provider idle time and provider overtime. Efficiency is greatest when the patients with the lowest predicted show probabilities are placed in an overbooked slot or in the slot right after it, and those are the slots associated with longer waits \[2\].

The research on who ends up in those slots was led by Michele Samorani and colleagues. The paper is not indexed in PubMed under its title, so MB relied on its public abstract on the SSRN preprint server. The abstract describes data from a large specialty clinic in which Black patients had higher no-show probabilities. It reports that state-of-the-art scheduling caused Black patients to wait about 30% longer than non-Black patients, and that an objective function that minimizes the waiting time of the group expected to wait longest removed the disparity at a similar schedule cost \[1\].

The follow-up ethics paper by Shanklin, Samorani and colleagues explains the mechanism. The prediction can be accurate given the training data, and the schedule built from it can still concentrate long waits in one group. The authors attribute the higher no-show rate in the data to “a large complex of social, political, and economic reasons” and note that low income is associated with missing appointments and, in the United States, with being Black \[2\]. The table summarizes the approaches they describe, with results they attribute to the original experiments.

| **Scheduling approach**                                                                                       | **Gap in average wait, Black and non-Black patients** | **Schedule quality compared with standard method** |
| ------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------- | -------------------------------------------------- |
| Standard predictive overbooking, built to maximize efficiency                                                 | About 30% longer for Black patients \[1\]             | Reference point                                    |
| No prediction at all (every patient gets the same show probability)                                           | No disparity \[2\]                                    | 14% lower \[2\]                                    |
| Prediction without socioeconomic fields such as zip code, marital status and employment                       | About 10% \[2\]                                       | About 4% lower \[2\]                               |
| Prediction unchanged, scheduler minimizes the wait of whoever is expected to wait longest, without using race | About 8% \[2\]                                        | About 4% lower \[2\]                               |
| Prediction unchanged, scheduler minimizes the wait of the racial group expected to wait longest               | Not statistically significant \[2\]                   | About 1% lower \[2\]                               |

Two details matter. Removing the socioeconomic fields did not remove the gap, because, as the authors note, features summarizing a patient’s no-show history still correlate with race and ethnicity \[2\]. And the approach that came closest to removing the disparity left the prediction alone and changed what the scheduler was told to minimize \[2\]. The ethics paper notes that using race explicitly raises legal questions that differ by application and jurisdiction \[2\].

## A fair prediction is not the same as a fair schedule

Not every analysis finds a problem with the predictions themselves. The Pennsylvania team stratified performance by sex and by race and ethnicity and reported that no bias against patient characteristics was detected in its final model \[4\]. The radiology group, for its part, advised caution because bias could lower the quality of service for patients based on race, zip code and gender \[5\].

These statements do not contradict the Samorani analysis. Checking whether a model predicts equally well across groups answers one question. The Samorani analysis asks what the schedule does with the predictions. A model can predict accurately for every group while the groups still end up in different slots, because the scheduler places the lowest estimates where the waits are longest. That distinction is MB’s reading of how the studies were designed. The Pennsylvania abstract reports a test of prediction performance and does not report waiting times.

The efficiency evidence is itself mixed. A simulation study of 947,364 outpatient appointments at one hospital in Oman found that overbooking guided by a prediction model outperformed systematic overbooking \[15\]. A rapid systematic review of predictive-model interventions identified seven randomized trials and one non-randomized study and concluded that the effect of predictive model-based overbooking was uncertain, with limited information on cost-effectiveness, acceptability and equity \[3\].

| **WHAT THE EVIDENCE CAN AND CANNOT TELL US** The 30% figure comes from one specialty clinic’s data, with patients grouped as Black and non-Black in that data. It does not show that every clinic that overbooks produces the same gap, or that any named clinic does. The disparity results are reported from the authors’ scheduling experiments, as the follow-up paper describes them. They are not a count of waiting times in a deployed system. The retrieved literature does not say how many practices use prediction-based overbooking. The one systematic review of predictive-model interventions found the overbooking evidence uncertain. The papers link higher no-show rates to income, access and other circumstances. They do not identify a single cause, and nothing here supports treating a missed visit as a judgment about a person’s character. The Pennsylvania fairness check measured prediction performance by group. It did not measure how long anyone waited. The Penn Medicine results come from a report by people involved in the program. |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

## What brings patients in

The rapid review of predictive-model interventions found that the model is more useful when it directs help than when it directs slots. There was high-certainty evidence from one randomized trial that predictive model-based text reminders reduced no-shows, with a median risk ratio of 0.91\. There was moderate-certainty evidence that predictive model-based phone reminders reduced no-shows, with a median risk ratio of 0.61 across three trials, and that patient navigators did too, with a risk ratio of 0.55 (95% confidence interval 0.46 to 0.67) in one trial \[3\].

Penn Medicine’s published case study gives a concrete example. In a rapid randomized trial of 59,994 patients at high risk of missing an appointment, adding an automated voice call to the existing text reminders lowered the no-show rate from 11.3% to 9.6% and raised completed appointments from 75.9% to 77.8%. The authors report that the effect on completion was largest among Black patients, helping to reduce an existing gap, and that results held in follow-up data from more than 244,000 high-risk patients over six months \[13\]. The trial ran for four weeks in 2024 at one health system.

Reminders do not work everywhere. A systematic review of 61 behavioral-economics studies found reminders were the most common intervention, with 56 studies, and that evidence supporting reminders by text, phone or mail was significant across settings, while other approaches were rarely tested \[10\]. A review of interventions in safety-net settings reported a pooled risk ratio of 1.08 (95% confidence interval 1.03 to 1.13) for attendance with any intervention, found no single strategy clearly superior, and described the strategies as only modestly effective. Combined strategies came out at 1.16 (1.03 to 1.32) \[11\]. In the New York health center study, a reminder had minimal effect: the no-show rate was 41.6% before and 42.1% after \[9\].

Transportation help has some support. A systematic review of 12 studies of non-emergency medical transportation, among them three randomized trials, pooled seven studies and found missed appointments at a ratio of 0.63 (95% confidence interval 0.48 to 0.83) with the interventions. Every included study had some risk of bias, and evidence on cost, utilization and health outcomes was too sparse for conclusions \[12\]. The safety-net review, which pooled a different set of studies, put transportation at 1.05 (0.98 to 1.13), a result that does not clearly differ from no effect \[11\]. The two reviews do not give one clean answer.

Other levers show up as associations. Continuity with the same clinician, shorter lead times and, in more recent studies, telemedicine were linked to lower non-attendance \[4,8,9\]. Evidence on same-day or open-access scheduling is thin in what MB retrieved: one 2021 study of 4,815 visits found that visit status was associated with whether patients were booked with a preferred physician \[16\]. It is not a trial and cannot show that open access lowers missed visits.

A no-show prediction can be used to decide who waits longer. It can also be used to decide who receives a phone call, a ride or a navigator. The intervention studies above are about the second use.

## Where the decisions sit

The table follows a missed visit through the system and lists, for each step, a question a patient or caregiver can fairly ask. The middle column describes what the retrieved studies report. It does not describe how any particular practice works.

| **Step** | **What the studies describe**                                                                                     | **A question to ask**                                                                                  |
| -------- | ----------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------ |
| Record   | Visits are coded as attended, no-show or late cancellation, and definitions differ \[4,5,14\].                    | How does the practice define a no-show, and how can a wrong code be corrected?                         |
| Predict  | A model scores upcoming appointments using fields such as lead time, past visits, age and portal use \[4,5,6,7\]. | Does the practice use a score, and has anyone checked it for differences across patient groups?        |
| Place    | A scheduler can book two patients into one slot and decides which patients \[2\].                                 | Does the practice ever book two people for the same time, and on what basis?                           |
| Wait     | In the studied method, patients in doubled slots and the slot after them wait longer when both arrive \[2\].      | Does anyone track waiting time, and for which patients?                                                |
| Respond  | A flagged patient can get a call, a ride or a navigator \[3,13\].                                                 | What does the practice do for patients it expects to have trouble coming, other than a different slot? |

| **QUESTIONS TO ASK A PRACTICE ABOUT HOW IT SCHEDULES** How will you remind me, and what happens if I do not get texts? Can I choose a phone call? If I miss a visit, how is it recorded? Can I explain why, and does the reason go in my record? Do you ever book two patients for the same time? If so, what is the usual wait, and what happens when both of us arrive? If I arrive late, how late is too late, and what happens then? How far ahead is your next available visit with my own clinician? Is there a cancellation list? Is telehealth an option for follow-ups, and can you help with transportation? Does the practice use a score or software to predict who will miss visits? What does it change about my appointment? If I need to cancel, whom do I call, and is there any consequence I should know about before I do? |
| ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |

## What MB will ask next

MB will ask practices and health systems to explain, in plain language, whether they predict no-shows, what the prediction changes about a patient’s appointment, and whether anyone measures waiting time by patient group. We will also ask for the written rule that tells front-desk staff what to do when two scheduled patients arrive for one slot. If you work in scheduling or patient access and want to describe how your practice does this, MB would like to hear from you.

General information, not individual medical advice. For emergencies, seek immediate help. The conversation continues.

## Sources (Peer-Reviewed Work Retrieved From Pubmed; Doi Links Included)

> \[1\] Samorani M, Harris S, Blount LG, Lu H, Santoro MA. Overbooked and overlooked: machine learning and racial bias in medical appointment scheduling. Manuf Serv Oper Manag. 2022;24(6):2825-2842\. Working-paper abstract retrieved from SSRN: https://papers.ssrn.com/abstract=3467047\. Not found in PubMed under this title; MB relied on the abstract only, not the full text, and no DOI was retrieved.  
>  
> \[2\] Shanklin R, Samorani M, Harris S, Santoro MA. Ethical redress of racial inequities in AI: lessons from decoupling machine learning from optimization in medical appointment scheduling. Philos Technol. 2022;35(4):96\. DOI: https://doi.org/10.1007/s13347-022-00590-8 (PMID 36284736; full text in PMC9584259). Conceptual and modeling paper with an ethics framework; reports results of the original experiments.  
>  
> \[3\] Oikonomidi T, Norman G, McGarrigle L, Stokes J, van der Veer SN, Dowding D. Predictive model-based interventions to reduce outpatient no-shows: a rapid systematic review. J Am Med Inform Assoc. 2023;30(3):559-569\. DOI: https://doi.org/10.1093/jamia/ocac242 (PMID 36508503)  
>  
> \[4\] Tuan WJ, Yan Y, Abou Al Ardat B, Felix T, Chen Q. Predicting missed appointments in primary care: a personalized machine learning approach. Ann Fam Med. 2025;23(4):294-301\. DOI: https://doi.org/10.1370/afm.240316 (PMID 40721333)  
>  
> \[5\] Rothenberg S, Bame B, Herskovitz E. Prospective evaluation of a machine-learning prediction model for missed radiology appointments. J Digit Imaging. 2022;35(6):1690-1693\. DOI: https://doi.org/10.1007/s10278-022-00670-3 (PMID 35768754)  
>  
> \[6\] Shour AR, Jones GL, Anguzu R, Doi SA, Onitilo AA. Development of an evidence-based model for predicting patient, provider, and appointment factors that influence no-shows in a rural healthcare system. BMC Health Serv Res. 2023;23(1):989\. DOI: https://doi.org/10.1186/s12913-023-09969-5 (PMID 37710258)  
>  
> \[7\] Blalock L, Corona E, Martinez D, Patel S, Sewell JL, Somsouk M. Estimated no-show rate from the electronic record strongly correlates with endoscopy no-show. BMC Gastroenterol. 2026;26(1). DOI: https://doi.org/10.1186/s12876-026-04874-7 (PMID 42067757). Single safety-net hospital, retrospective.  
>  
> \[8\] Jad AZ, Alotaibi SM, Asiri AM, Alobaid SK, Algahamdi RA, Alaqeel LA, et al. Global predictors of appointment non-adherence in primary care settings: a systematic review. Healthcare (Basel). 2026;14(5). DOI: https://doi.org/10.3390/healthcare14050623 (PMID 41827577). Narrative synthesis of 27 observational studies; quality modest to moderate.  
>  
> \[9\] Ruggeri K, Folke T, Benzerga A, Verra S, Büttner C, Steinbeck V, et al. Nudging New York: adaptive models and the limits of behavioral interventions to reduce no-shows and health inequalities. BMC Health Serv Res. 2020;20(1):363\. DOI: https://doi.org/10.1186/s12913-020-05097-6 (PMID 32336283). Retrospective observational study.  
>  
> \[10\] Werner K, Alsuhaibani SA, Alsukait RF, Alshehri R, Herbst CH, Alhajji M, et al. Behavioural economic interventions to reduce health care appointment non-attendance: a systematic review and meta-analysis. BMC Health Serv Res. 2023;23(1):1136\. DOI: https://doi.org/10.1186/s12913-023-10059-9 (PMID 37872612)  
>  
> \[11\] Crable EL, Biancarelli DL, Aurora M, Drainoni ML, Walkey AJ. Interventions to increase appointment attendance in safety net health centers: a systematic review and meta-analysis. J Eval Clin Pract. 2020;27(4):965-975\. DOI: https://doi.org/10.1111/jep.13496 (PMID 33064929)  
>  
> \[12\] Shekelle PG, Begashaw MM, Miake-Lye IM, Booth M, Myers B, Renda A. Effect of interventions for non-emergent medical transportation: a systematic review and meta-analysis. BMC Public Health. 2022;22(1):799\. DOI: https://doi.org/10.1186/s12889-022-13149-1 (PMID 35449011)  
>  
> \[13\] Kavanaugh L, Thomas S, Lynch S, Leri D, Xiong RA, Xu L, et al. Automated calls added to SMS reminders reduce missed appointments among high-risk patients. NEJM Catal Innov Care Deliv. 2026;7(6):CAT250322\. DOI: https://doi.org/10.1056/CAT.25.0322 (PMID 42418615). Case study with a randomized comparison, reported by the implementing organization.  
>  
> \[14\] Brewster S, Bartholomew J, Holt RIG, Price H. Non-attendance at diabetes outpatient appointments: a systematic review. Diabet Med. 2020;37(9):1427-1442\. DOI: https://doi.org/10.1111/dme.14241 (PMID 31968127)  
>  
> \[15\] Alawadhi A, Jenkins D, Palin V, van Staa T. Development and evaluation of prediction models to improve the hospital appointments overbooking strategy at a large tertiary care hospital in the Sultanate of Oman: a retrospective analysis. BMJ Open. 2025;15(4):e093562\. DOI: https://doi.org/10.1136/bmjopen-2024-093562 (PMID 40306993). Retrospective analysis with simulation.  
>  
> \[16\] Gant-Farley HY, Ross MK, Hudak RP. After COVID-19: improving the patient’s outpatient appointment experience. J Patient Exp. 2021;8:23743735211039320\. DOI: https://doi.org/10.1177/23743735211039320 (PMID 34869833). Single-site observational study of 4,815 visits.