Loading · جارٍ التحميل
Evaluation lab
Our method by language, help needed and schooling, with denominators and confidence intervals. Groups with fewer than 30 people in decline are marked, not read.
Overall figures can hide differences: one group can be served worse. Coverage is how often a stable person's check lands inside their usual range. It should stay near 90% in every group.
| Group | n | In decline | Sensitivity (95% CI) | Specificity (95% CI) | Wrong referrals per 1,000 | Usual-range coverage |
|---|---|---|---|---|---|---|
| Everyone | 5,000 | 519 | 59.5% (55–64) | 93.3% (93–94) | 60 | 93.2% |
| Arabic, Emirati | 2,217 | 220 | 59.6% (53–66) | 94.3% (93–95) | 51 | 93.6% |
| Arabic, resident | 1,007 | 114 | 57.9% (49–67) | 92.6% (91–94) | 66 | 93.0% |
| English | 1,273 | 129 | 60.5% (52–68) | 93.2% (92–95) | 61 | 93.8% |
| Urdu / Hindi (assisted) | 503 | 56 | 60.7% (48–72) | 90.4% (87–93) | 85 | 90.4% |
| Education 0–9 years | 977 | 103 | 67.0% (57–75) | 90.5% (88–92) | 85 | 90.5% |
| Education 13+ years | 1,971 | 212 | 52.4% (46–59) | 93.8% (93–95) | 56 | 94.0% |
| Assisted, education 0–9 | 97 | 9 | Too few in decline to read (under 30) | 91.6% | ||
Computed from the 3-year cohort (seed 20261113). Local Gulf norms replace the literature priors in pilot phase 1.