AI enables 100% quality control coverage where manual sampling was previously the only economically viable option.
AI value · Quality control coverage
Verified
100%
The 10% baseline is a general industry practice estimate, not Maccabi's measured prior practice.
Before
Perform primary diagnosis on prostate biopsies → Manually review a small sample (e.g., 10%) of cases for quality control
After
Perform primary diagnosis on prostate biopsies → Analyze 100% of digitized slides as a second-read control and raise alerts for discrepancies → Review AI alerts (minimal time cost) and order additional cuts/stains if necessary
Human boundary
AI raises alerts; human pathologist reviews alerts and makes the final diagnostic and procedural decisions.
Why it matters
Second-read quality control is too labor-intensive to perform on 100% of cases.
How the work changed
Before
How the work ran before the change.
Step 1 of 2
Pathologist
Perform primary diagnosis on prostate biopsies
Step 2 of 2
Pathologist (Quality Control)
Manually review a small sample (e.g., 10%) of cases for quality control
ControlSample-based human QA
What changed
Second-read quality control is too labor-intensive to perform on 100% of cases.
Decision rightAI handles the default; humans own exceptions
After
How the same work runs now.
Step 1 of 3
Pathologist
Perform primary diagnosis on prostate biopsies
Step 2 of 3
AI System
Analyze 100% of digitized slides as a second-read control and raise alerts for discrepancies
Control100% AI QA coverage
Step 3 of 3
Pathologist
Review AI alerts (minimal time cost) and order additional cuts/stains if necessary
Process model built from the published workflow evidence for Maccabi Healthcare Services. Every step, actor, and control appears in full below.Every step, actor, and control
Exception path
Alerts indicating missed cancer or significant grading differences lead to additional cuts, stains, or third-opinion requests.
Decision authority
AI raises alerts; human pathologist reviews alerts and makes the final diagnostic and procedural decisions.
Before
#
Actor
Action
Control
01
Pathologist
Perform primary diagnosis on prostate biopsies
Not published.
02
Pathologist (Quality Control)
Manually review a small sample (e.g., 10%) of cases for quality control
Sample-based human QA
After
#
Actor
Action
Control
01
Pathologist
Perform primary diagnosis on prostate biopsies
Not published.
02
AI System
Analyze 100% of digitized slides as a second-read control and raise alerts for discrepancies
100% AI QA coverage
03
Pathologist
Review AI alerts (minimal time cost) and order additional cuts/stains if necessary
Human override and final diagnosis
Work that left the path
Manual screening of cases that have no discrepancies
Human role before
Perform sample-based manual second reviews for quality control.
Human role after
Review targeted AI alerts on 100% of cases to finalize quality control.
AI role
Act as a 100% coverage second-read control system for all prostate core-needle biopsies.
Outcomes
Quality control coverage
Verified
10%→100%
March 2018 to November 2019 · 941 cases / 11,429 H&E slides
The 10% baseline is a general industry practice estimate, not Maccabi's measured prior practice.
Time cost of quality control
Verified
Not specified→~1%
March 2018 to November 2019 · Pathologist time
Author estimate based on alert volume.
What leaders can reuse
Anti-pattern
Assuming 100% AI screening requires 100% human re-review.
Questions
01What critical QA processes are currently sampled at 10% because 100% is too expensive?
Portability conditions
Digital workflow capability
High volume diagnostic environment
Reputation risk
low
Evidence and authority
What the public record supports.
Current · updated
1 peer reviewed; publication outcomes are verified.
Bundle 1.0.0 · reviewed 2026-09-06 · stable ID f4f600d2dd9b256e
