|
Purpose |
Identify overdue recommendations for follow-up from radiology reports |
|
Tag(s) |
Actionable finding, critical result, incidental finding, follow-up |
|
Panel |
Non-Interpretive, COVID-19 |
|
Define-AI ID |
20160001 |
|
Originator |
COVID-19 Sub-Panel |
| Lead | Kellie Greenblatt |
|
Panel Chair |
Ben Wandtke, Woojin Kim |
|
Panel Reviewers |
Michael Francavilla, Rich Heller |
|
License |
Creative Commons 4.0 |
| Status | Public Comment |
Many radiologic findings prompt a recommendation for follow-up. As a result of the COVID-19 pandemic, there is the potential that important follow-up care was not obtained (delayed, canceled, no-show, or simply overlooked).
This program needs to identify radiology reports that recommended follow-up, then be able to determine if the follow-up was ever obtained. The output should contain each patient’s information, the radiology finding that requires follow-up, the modality or procedure that was recommended, and the time frame by which the follow-up was to occur. This can then be used to arrange any follow-up that has not been completed.
A 64-year-old male had a CT performed in April 2019 showing a 3.8 cm diameter infra-renal aorta. The radiology report recommended a follow-up CT in April 2020. That exam was scheduled but then canceled due to imaging center closure, related to COVID-19. The exam was never rescheduled. The follow-up examination is necessary due to the risk of the aortic aneurysm continuing to expand and eventually rupture, a life-threatening condition.
A 55-year-old female with no history of malignancy had a non-contrast abdominal CT performed in late February 2020 for pain, incidentally showing a homogenous-appearing 3 cm left adrenal mass with Hounsfield unit measurements of 30. The radiology report recommended a follow-up MRI with in- and opposed-phase imaging. That exam was never scheduled. The follow-up examination is necessary to determine whether the adrenal mass is benign or potentially malignant, necessitating referral to a surgeon.
The program analyzes radiology reports from the year and identifies instances of follow-up that were to be performed starting March 1, 2020 (or a starting date set by the end-user).
The program then uses the RIS and EMR to see if the recommended follow-up was obtained, including not only outpatient imaging but also ED and inpatient imaging exams and procedures. This is dependent on the program’s ability to correctly identify the follow-up exam to be completed and to also determine whether or not that particular (or another qualifying) exam was completed within an appropriate time frame in the RIS/EMR. Three possibilities exist:
Scheduled and obtained: no need for flagging
Scheduled but not obtained (exam scheduled in the future, exam canceled, exam rescheduled, patient no-show, exam no longer needed, or patient deceased): flagged for follow-up/monitoring
Never scheduled: flagged for follow-up
If follow-up was recommended but not obtained, it adds that patient to a database that contains information on the patient, the ordering clinician, the imaging finding, the specific recommendation such as modality, if an imaging examination was recommended, or action type (e.g, biopsy) and anatomy), and when it was due for follow-up.
The program needs to be able to work with existing systems/workflows related to the management of follow-up care of a given site. For example, the program could alert a navigator, who could subsequently reach out to either the ordering clinician or the patient or both. If the program has access to the scheduling system and the patient has an order for the follow-up exam, it can request the patient schedule the exam. Regardless of the system and workflow used at a given site, the program needs to continuously monitor to ensure the appropriate and correct follow-up exam is completed, along with reminders and escalation options.
|
Data Element |
Data Type |
Description |
Notes |
|
Radiologist report |
Text |
Reports dating back to 3/1/2019* through the COVID-19 crisis to be processed for possible missed radiologic follow-ups. |
*This date can be determined/modified by the developer and/or end-user(s). While most follow-up recommendations are 1 year or less, some actionable findings have a much longer imaging interval (e.g., the aorta that ranges from 3 months to 5 years based on aortic diameter (Khosa et al.)). As a result, the developer will want to give end-users the ability to adjust this look-back period. |
|
Radiology report metadata |
Various |
Various metadata about the exam that contains the follow-up recommendation, such as patient information, ordering clinician, reporting radiologist(s), modality, CPT, date of examination, other timestamps, etc. |
|
|
Evidence-based follow-up recommendations/guidelines |
Object |
Information on follow-up recommendations that are evidence-based and have been published |
One sample list (REF 2) |
|
Scheduling records |
Various |
Data on scheduled appointments for managing an actionable finding with follow-up recommendation(s), such as date(s) of the exam(s), patient identifiers, ordering clinician, exam type(s), exam modality, and the location where the exam(s) will be performed |
|
|
Obtained exams |
Various |
Various data on imaging that was acquired for managing an actionable finding with follow-up recommendation(s) |
Imaging Finding
|
Definition |
Imaging finding that triggered follow-up |
|
Data Type |
Text |
|
Value Set |
N/A |
|
Units |
N/A |
Follow-Up Recommendation
|
Definition |
Follow-up recommendation phrase/sentence |
|
Data Type |
Text |
|
Value Set |
N/A |
|
Units |
N/A |
Follow-Up Recommendation Anatomy
|
Definition |
Anatomy (if present) |
|
Data Type |
Text |
|
Value Set |
Liver, right lower lobe, etc. |
|
Units |
N/A |
Follow-Up Recommendation Modality Type
|
Definition |
Modality (if present) |
|
Data Type |
Text |
|
Value Set |
US, MRI, X-RAY, CT, PET, etc. |
|
Units |
N/A |
Follow-Up Recommendation Time Interval
|
Definition |
Time interval (if present) |
|
Data Type |
Numeric |
|
Value Set |
2 weeks, 6 months, annual, etc. |
|
Units |
Time unit/ description |
Follow-Up Recommendation Action Type (other than imaging)
|
Definition |
Action Type (if present) |
|
Data Type |
Text |
|
Value Set |
Biopsy, laboratory test, clinical appointment, etc. |
|
Units |
N/A |
Evidence-Based Follow-Up
|
Definition |
Matches the recommended action with follow-up recommendations from published whitepapers/guidelines |
|
Data Type |
Boolean |
|
Value Set |
True, False |
|
Units |
N/A |
Follow-Up Status
|
Definition |
The status of a patient’s imaging follow-up |
|
Data Type |
Categorical |
|
Value Set |
|
|
Units |
N/A |
Contact Ordering Clinicians and/or Patients to Schedule Follow-Up
|
Definition |
For flagged cases with a necessary follow-up that has not yet been scheduled, send an automated message to ordering clinician and/or patient with scheduling information |
|
Data Type |
Text |
|
Value Set |
N/A |
|
Units |
N/A |
Date Follow-Up Obtained
|
Definition |
Identifies when the patient received follow-up if obtained |
|
Data Type |
Date |
|
Value Set |
Date |
|
Units |
N/A |
Patient Status
|
Definition |
Identifies where the patient received follow-up if obtained |
|
Data Type |
Categorical |
|
Value Set |
|
|
Units |
N/A |
A patient can receive his/her recommended follow-up exam at a different institution/imaging center. To prevent unnecessary exams and notifications, if given access to an image-sharing network, the AI model can run the same algorithm for the detection of follow-up exam(s) and either close the loop automatically or notify the ordering clinician and/or patient.
Related, based on prior history, the AI model should be able to determine the likelihood of a patient going outside the system to receive his/her follow-up. Depending on the intended use, this information can aid in closing the loop or possible intervention to keep the patient within the system.
The AI model should be able to determine the likelihood of patient adherence to the follow-up recommendation to allow an appropriate level of intervention to increase the likelihood of adherence - similar to a patient no-show algorithm.
The AI model should alert the radiologist whenever there is a follow-up recommendation that is missing from the radiology report when there should be one.
For notification, the AI model should be able to determine the best/most appropriate person of contact based on the finding. Once determined, it should send the notification with integration into other systems to allow monitoring of confirmation of receipt. In addition, the model can also incorporate an escalation procedure in case the first person of contact fails to acknowledge receipt despite multiple attempts. The patients are also contacted based on local policies. Based on the level of integration with other systems, the proper follow-up exam can even be suggested for ease of ordering.
Larson PA, Berland LL, Griffith B, Kahn CE Jr, Liebscher LA. Actionable Findings and the Role of IT Support: Report of the ACR Actionable Reporting Group. J Am Coll Radiol. 2014;11(6):552-8. doi: 10.1016/j.jacr.2013.12.016
https://publish.smartsheet.com/42d18e874a164318a0f702481f2fbb70
Khosa F, Krinsky G, Macari M, Yucel EK, Berland LL. Managing incidental findings on abdominal and pelvic CT and MRI, Part 2: white paper of the ACR Incidental Findings Committee II on vascular findings. J AM Coll Radiol. 2013;10(10):789-94. doi: 10.1016/j.jacr.2013.05.021