Predict Patient No-Shows For Radiology Appointments



Prediction of no-shows for radiology appointments




Business Facing Subpanel

Define-AI ID



Ashley Prosper, Hari Trivedi


Ashley Prosper, Hari Trivedi

Panel Chair

Rich Heller

Non-Interpretative Chairs

Alexander J. Towbin, Adam Prater

Panel Reviewers

Business Facing Subpanel


Creative Commons 4.0 


Public Comment

Clinical Implementation

Value Proposition

Efficient and appropriate outpatient scheduling is critical to the operational and economic success of a radiology department. When patients fail to appear for their scheduled appointments, revenue opportunities are lost, and employee productivity declines. In some cases, time-sensitive medications or tools are specially ordered for a specific patient’s appointment, resulting in economic losses when a patient fails to appear. Many of these factors may be out of a patient’s control, such as comorbidities or lack of transportation. Identifying patients that are at risk of “no-showing” may allow interventions that increase the likelihood of appointment adherence, or enable the department to quickly fill time slots in the event of a missed appointment. 


A 65-year-old male patient is being evaluated for an oncologic clinical trial. Trial enrollment requires that he receive a staging Dotatate PET/CT before beginning therapy. Eager to start their patient on treatment, the oncologist requests a PET/CT ASAP and the patient is given an appointment for Monday morning at 7 AM. A 4 mCi dose of Dotatate is ordered for the patient the Friday before for Monday morning delivery. A PET technologist and nurse are assigned an early shift with a 6 AM start time to receive the dose, prep the department, and prepare for the patient’s arrival. Monday at 7:30 AM arrives and the patient has not shown up for his appointment. The patient is unreachable by telephone. The next patient to receive Dotatate is not scheduled until tomorrow, all others are receiving 18-FDG. Realizing that the Dotatate dose will soon expire, the technologist speaks to the schedulers upon their arrival in the department at 8 AM. A patient cannot be found to fill the slot before the dose expires. The department loses 1 hour of the technologist and nurse’s time, 1 dose of Ga-68 Dotatate, and 1 hour of PET/CT scanner time.

A 72-year-old woman has an adenocarcinoma in the left upper lobe of the lung. Deemed a poor surgical candidate, she is scheduled to undergo a CT-guided ablation of the nodule. Given how peripheral the nodule is, her interventional radiologist selects cryoablation. Her local hospital owns a microwave probe but rents a cryoprobe for special procedures. Her IR physician requests a cryoprobe for Tuesday at 11 AM and the CT scanner is held for this procedure. The IR physician blocks their schedule for 2 hours. The patient fails to report to pre-op as scheduled for 10 AM that day. The department loses rental fee for the cryoprobe, 1 hour CT scanner time, and 2 hours of procedure time for the IR.

Workflow Description

A patient order is received for an outpatient imaging exam and/or image-guided procedure. A time slot is then identified in the schedule. This triggers three pathways – a reminder system for the scheduled patient, modification to other exams scheduled the same day, and a backup patient list. Exams requiring special imaging contrast, radionuclides, or equipment are preferentially clustered together. 

  1. Scheduling one exam requiring a special dose or tool triggers subsequent exams of the same type to be scheduled on that same day. In the event that a patient does not show, the dose or equipment can be used by backup patients scheduled later in the day. 

  2. A reminder system is triggered based on multiple patient factors. An algorithm can be trained using data about previously missed appointments to identify risk factors such as patient age, location, access to transportation, co-morbidities, previously missed appointments, days between scheduling and appointments, modality type, insurance type, etc. This algorithm can then assign a risk score for missed appointments to each patient and a cutoff threshold can be used to trigger interventions such as patient phone calls and text messages, transportation arrangements, overbooking, or backup patient lists. 

  3. Patients with an appointment later than they would like are placed on a standby list. A missed appointment automatically triggers a phone call to patients scheduled for similar exams as the missed appointment until a replacement can be found.

Considerations for Dataset Development


  1. Imaging Only 

    1. PET/CT

    2. MRI

    3. Nuclear medicine (aside from PET)

    4. CT

    5. US

    6. Mammography

  2. Image-Guided Procedures

    1. Ablation – CT and fluoroscopically guided

    2. CT and US-guided biopsy

    3. Breast biopsy

    4. Vascular interventional


  1. Is time slot available on the same day as another exams/procedure requiring the same equipment already scheduled for that week?

  2. Days between scheduling and appointments

  3. Was the schedule made for the patient or did the patient make the schedule (latter has been shown to improve no-show rates)?

  4. Weather at the time of appointment (temperature, snow fall, etc.)


  1. Imaging/procedure required

  2. Special medication or equipment required?

  3. Appointment attendance history including history of calling to cancel or reschedule (as these would be used as exclusion criteria). Also, this should include information on punctuality. 

  4. Primary language

  5. Means of transportation

  6. Insurance type

  7. Deceased?

  8. ICD 10, Axis I mental health diagnoses (major depression, bipolar disorder, posttraumatic stress disorder [PTSD], etc), alcohol dependence, substance dependence, lack of housing, homelessness, or unspecified housing status

  9. Socioeconomic factors?

  10. Distance

  11. Admission? [can the system automatically see if they were admitted]


  1. Imaging/procedure required

  2. Special medication or equipment needed?

  3. Proximity/travel time to the imaging center

  4. Best means of contact

  5. No-show history

Technical Specifications



Access to Facility Schedule



Information on a facility’s patient and staffing schedule

Potential Feature

Upcoming scheduled appointments with exam information, Patients still awaiting an appointment with exam information, Staffing with expertise by day


Access to Patient Data



Information about a particular patient, around their history with imaging, interactions with staff, and attendance to appointments.

Potential Feature

Imaging/procedure required, Medication required, Equipment required, Equipment required, Appointment attendance, Primary language, Primary means of transportation


Access to Exam Data



Information about the imaging services at the facilities and the associated pharmaceuticals for each exam

Potential Feature

type of imaging, type of image-guided procedure, medication for a given exam


Access to Patients on Standby



Information on the patients awaiting an appointment and their exam.

Potential Feature

Imaging/ Procedure required, Medication required, proximity to facility, contact information, appointment attendance history


Primary Outputs


Recommended Schedule Modifications


Recommend a schedule of appointments which clusters similar procedures together such in the case of a no-show, resources can be diverted with minimal loss



 No-Show Probability



For a given patient and appointment, predict the probability of a no-show

Data Type


Value Set





Secondary Outputs




 Identify patient(s) who could replace the no-show



When a no-show patient is confirmed, identify a patient with a similar procedure on the standby list who could be swapped in.





 Scheduled reminders to patients on schedule and on standby



With the patient’s primary mode of communication, notify them of their upcoming appointment. Notify those on standby that they could possibly be called if a no-show occurs.






H.B. Harvey, C. Liu, J. Ai, et al.
Predicting no-shows in radiology using regression modeling of data available in the electronic medical record
J Am Coll Radiol, 14 (2017), pp. 1303-1309

B. Satiani, S. Miller, D. Patel
No-show rates in the vascular laboratory: analysis and possible solutions
J Vasc Interv Radiol, 20 (2009), pp. 87-91