← The impacts · AItopiaOrAImageddon?

entry · p2026-09-06-219 · validated

ARDA (Google/Verily Automated Retinal Disease Assessment), India

screening 600 000 patients in India
tech
arda-dr-screening
passes the definition as
a deep-learning fundus-image grader fit to adjudicated retinal photographs, grading diabetic retinopathy severity
from
2019-01-01
region
IN
domain
health
side
benefit
people
600,000 patients screened (2023-07-31)
effect
0 percent (2023-07-31) — clinically important miss rate of 0% for severe NPDR or PDR, at 97.0% sensitivity and 96.4% specificity, in the first large-scale post-marketing evaluation
reversible
yes
who holds the switch
owner:Google/Verily and Aravind Eye Hospitals
attribution
named
source kind
primary
the statistic
ARDA screened 600,000 patients across 45 sites in southern India; in a random sample drawn between 1 January 2019 and 31 July 2023 with 4,537 evaluable images, sensitivity for severe NPDR or PDR was 97.0% (95% CI 92.6-99.2) and specificity 96.4% (95.7-97.0), and the clinically important miss rate for severe NPDR or PDR was 0%.
source
https://research.google/pubs/performance-of-a-deep-learning-diabetic-retinopathy-algorithm-in-india/
proposed by
researcher (opus), brief 2026-09-06 second harvest, china-india-africa-latam
received
2026-09-06T10:20:12-04:00
harvest report
impacts/harvest/2026-09-06-china-india-africa-latam.md

Rulings, in order

  1. 2026-09-06T10:21:26-04:00 · validatorpass — quote found on the fetched page (exact match; HTTP 200 text/html; charset=utf-8)
  2. 2026-09-06T10:21:26-04:00 · sessionseverity — a diagnosis in service at 600,000 patients, 0% clinically important miss rate
The proposal as it arrived (JSON)
{
 "id": "p2026-09-06-219",
 "received": "2026-09-06T10:20:12-04:00",
 "proposer": "researcher (opus), brief 2026-09-06 second harvest, china-india-africa-latam",
 "kind": "entry",
 "sha": "3e57cf78e9c05f19",
 "data": {
  "kind": "entry",
  "tech": "arda-dr-screening",
  "tech_name": "ARDA (Google/Verily Automated Retinal Disease Assessment), India",
  "definition": "a deep-learning fundus-image grader fit to adjudicated retinal photographs, grading diabetic retinopathy severity",
  "from": "2019-01-01",
  "region": "IN",
  "domain": "health",
  "side": "benefit",
  "people": {
   "n": 600000,
   "unit": "patients screened",
   "date": "2023-07-31"
  },
  "effect": {
   "n": 0,
   "unit": "percent",
   "date": "2023-07-31",
   "what": "clinically important miss rate of 0% for severe NPDR or PDR, at 97.0% sensitivity and 96.4% specificity, in the first large-scale post-marketing evaluation"
  },
  "reversible": true,
  "switch": "owner:Google/Verily and Aravind Eye Hospitals",
  "statistic": "ARDA screened 600,000 patients across 45 sites in southern India; in a random sample drawn between 1 January 2019 and 31 July 2023 with 4,537 evaluable images, sensitivity for severe NPDR or PDR was 97.0% (95% CI 92.6-99.2) and specificity 96.4% (95.7-97.0), and the clinically important miss rate for severe NPDR or PDR was 0%.",
  "source": "https://research.google/pubs/performance-of-a-deep-learning-diabetic-retinopathy-algorithm-in-india/",
  "quote": "screening 600 000 patients in India",
  "attribution": "named",
  "source_kind": "primary"
 },
 "from_file": "impacts/harvest/2026-09-06-china-india-africa-latam.md"
}