Table 1 Markov Model Parameter Estimates and Assumptions
From: Cost-effectiveness of artificial intelligence screening for diabetic retinopathy in rural China
Parameter | Value | Sensitivity Analysis Range |
|---|---|---|
1. DR transition probabilities | ||
  No to Mild [27] | 0.07 | 0.01–0.10 |
  Mild to Moderate [27] | 0.19 | 0.166–0.214 |
  Moderate to VTDR [27] | 0.17 | 0.147–0.193 |
  VTDR to Stable [28] | 0.90 | 0.881–0.919 |
  Stable to Blindness [29] | 0.02 | 0.002–0.03 |
  VTDR to Blindness [29] | 0.09 | 0.07–0.11 |
2. Utility | ||
  No DR [30] | 0.94 | 0.83–1.05 |
  Mild DR [30] | 0.87 | 0.73–1.01 |
  Moderate DR [30] | 0.87 | 0.73–1.01 |
  VTDR [30] | 0.83 | 0.74–0.92 |
  Stable DR [31] | 0.85 | 0.72–0.78 |
  Blindness [30] | 0.81 | 0.73–0.89 |
3. Disutility of DR [32] | 0.066 | - |
4. Mortality multipliers [26] | ||
  Blindness | 2.34 | 2.22–2.46 |
  Diabetes | 1.90 | 1.04–2.7 |
5. Sensitivity, % | ||
  AI Screening screening [23] | 90.79 | 86.40-94.10 |
  Ophthalmologist screening [24] | 96.00 | 94.79-97.21 |
6. Specificity, % | ||
  AI screening [23] | 98.50 | 97.80-99.00 |
  Ophthalmologist screening [24] | 94.67 | 94.57-97.43 |
7. Compliance of screening [33], % | 86.00 | - |