THE BRAVE NEW WORLD OF Enhanced Pattern Recognition
AI-AUGMENTED TELEPHONE TRIAGE
The work of teletriage is nothing if not uncertain, and humans intensely dislike uncertainty. However, it is reasonable and prudent to acknowledge the limitations of this work, and to question our assumptions about it. Safer practice and evidence-based tools can reduce uncertainty and improve outcomes.
“Triage (on site) is not an endpoint but a beginning.” American College of Emergency Physicians, 1999. Uniform Triage Scale in Emergency Medicine, Information Paper
"Physician expertise and professionalism alone could not prevent common error.” Killip. (2007),
“All learned occupations have a definition of professionalism, a code of conduct. It is where they spell out their ideals and duties. They all have at least three common elements: expectation of selflessness; of skill; and of trustworthiness. Aviators have a fourth expectation -- discipline”. (Gawande, 2010)
‘We don’t look for patterns of our recurrent mistakes or devise and refine potential solutions for them. But we could, and that is the ultimate point”(Gawande, 2010).
“Any mistake or failure in the diagnostic process can lead to a diagnosis that is wrong, missed or delayed. More diagnostic errors occur in ambulatory settings, but inpatient diagnostic errors tend to be more lethal”. Groszkruger, 2014.
AI Safety
CDSS that employs AI without a governing framework and guard rails - can produce unsafe outcomes. Errors of Assessment, Communication, Continuity, Informed Consent and Human Error.(Joint Commission) are Root Causes of unsafe outcomes in teletriage.
Malpractice cases in teletriage are typically related to these predictable and preventable errors. Human and system error can be addressed by improving systems —structure and process — a group of related parts that work together for common goals — patient safety, practitioner safety, system effectiveness.
Meets Manchester Triage Group
Triage Requirements
1. Practice Management:
a. Clinical Training - Five CE courses
b. Risk Management - Clinical training, Error Feeback
c. Clinical Standards: Practice & Call Center
d. Prospectively: Universal Guideline - EBM certified
2. Clinical Trends/Health care Technology: CDSS, EMR
3. Financial Performance: Tracking and Trending Capabilities
4. Patient Experience:
a. Patient Brochure on how to use Telehealth Nurse Service
b. Patient Educational Material for Symptom-Based Calls
Prospectively Meets Developer Standards for POC AI-CDSS
Designed to be EBM-tested (demonstrably safe, valid and reliable without AI)
Prospectively AI/PA-Ready
A Single, Robust Template streamlines development for remaining 50 Guidelines
Adaptable to All Ages
Rapid integration, updating, modifying & testing
Multi-Purpose, Multi-Setting
Evolution of A Universal Guideline to AI-Ready CDSS
1993 Master Template A skeletal format to develop a set of Guidelines with minimal disposition details, three acuity levels
1995 Contingency Protocol, NPF = No Protocol Found (when no Protocol seems to apply),k acting as a “Fall-Back” Protocol with Dispositions (5 level acuity, described and defined categories, with a time and place for further evaluation.
1998 Pre-Emptive Protocol. The Go-To Protocol for the first pass prior to selecting a specific guideline
2000 Master or Generic Guideline A well -developed Framework Protocol and Template
2002 Universal Guideline More robust with broad Key Questions to rule out common emergent to non-acute symptoms,
2002 Universal Guideline as Training adjunct to introduce broad principles and policies, for new staff. Also useful for Face-to-face triage in ambulatory settings: on-site — ED and Urgent Care, any ambulatory or pre-hospital setting: Office, Clinic, College Student Health.
2020-2025 Universal Guideline Proof of Concept- with Reminders, QA questions, Built-in Nursing Process, Error Feedback, EBM ready — As-Is and later AI-ready.
A Universal Guideline for
Telehealth Nursing or Remote Nurse Triage
More Opportunities to be right
Includes Patients’ Own Degree of Worry Score
More Opportunities to anticipate/correct Common Error
Ease of Use: Workflow
Ease of Use: Guidelines and Form
Built-In Prompts for Red Flags and Rules of Thumb
A Robust Template for a Clinical Decision Support System (CDSS) Some pre-hospital triage CDSS are confusing, complicated and impractical.
In development for four decades, the Universal Guideline is based on the nursing process and has a five-level triage system with clear dispositions (when, where and reason for further evaluation), Anticipated resources, like equipment, medication, lab work, x-rays and/or procedures may determine the ultimate venue where a patient will ultimately be evaluated.
Many pre-hospital triage protocols and CDSS designs are unwieldy for the purpose of research or even updating or revising. Difficult designs interfere with proving that given CDSS is reliable and valid or evidence-based medicine (EBM).. Inadequately tested CDSS might actually interfere with the decision making process (Wachter, 2017).
A Universal Guideline Online: Fast & Cost-Effective to Test & Update
Telephone Triage Protocols and CDSS designs must be proven valid and reliable. Nurses need user-friendly tools— elegant, transparent and seamless. Standards or guard rails are essential for the next generation of AI-assisted CDSS. Meaningful research on how to safely incorporate AI/PA into current CDSS is urgently needed.
All-Purpose CDSS: Telephone Triage, Telehealth or ED Triage
Generative AI, predictive analytics, advanced EMRs and biotechnology will be standard components of CDSS in the near future. These powerful technologies, require a reliable system within which to operate — clinical guardrails.
Typically, clinical specialties require systems: clinical training, clinical practice standards, safe CDSS, EMR and experienced clinicians trained in remote clinical decision making. In other words, clinical systems must meet the Duty of Due Care — a legal principle underlying established subspecialties. Additional duties are the Duty to Communicate, to Warn and to Document.
A Universal Guideline with embedded standardized components — structure and process — serves as a virtual office visit — abbreviated History and Physical, Relevant past and recent medical history or “back story” - provide contextual information about the patient’s risk level. Essential assessment questions elicit more information about the chief complaint.
Early on, protocols were likened having a physician at one’s side while making difficult clinical decisions— supporting, reminding and prompting with additional questions. AI and predictive analytics may supplant the expert physician advisor, supporting pattern recognition and suggesting an acuity level and triage disposition - as a co-pilot.
“Improved Systems improve safety”
Giesen et al., (2011), Annals of Medicine
“If you fail to plan; plan to fail” (IOM)
Does your organization have a formal system? Or is it just “bits and pieces”? The Institute of Medicine defines system error as “failure to use a plan; using the wrong plan; or failure of planned action to be completed as intended”. Institute of Medicine, 2000. [PubMed]
CERNO…..I Perceive - -
Discern – Distinguish – Sift – Separate - Bear in Mind – Reckon – Determine – Resolve – Contend - Decide