Table 5. Performance measures for diagnosis date–based hindcasting approaches in Nowcasting by Bayesian Smoothing, applied to daily case counts of New York City residents with mpox diagnosis from July 13 through September 27, 2022 (metrics calculated on last 7 days of hindcast, excluding weekends).
| Stratification and scenario number | Window length (days)a, n | Mean absolute error | Relative root mean square error | Number of estimates when the 95% prediction interval included the final case count (95% prediction interval coverage) | Number of estimates evaluated (number of models run) | Average score |
|---|---|---|---|---|---|---|
| Unstratified | ||||||
| 1b | 14 | 9.04 | 0.23 | 48 (96.00) | 50 (10) | 0.27 |
| 2 | 21 | 8.73 | 0.25 | 42 (93.33) | 45 (9) | 0.28 |
| 3 | 28 | 7.18 | 0.25 | 37 (92.50) | 40 (8) | 0.27 |
| 4 | 35 | 5.09 | 0.27 | 35 (100.00) | 35 (7) | 0.41 |
| 5 | 42 | 3.93 | 0.29 | 29 (96.67) | 30 (6) | 0.44 |
| 6 | 49 | 2.88 | 0.30 | 24 (96.00) | 25 (5) | 0.47 |
| Black or African American | ||||||
| 7b | 14 | 2.90 | 0.30 | 49 (98.00) | 50 (10) | 0.39 |
| 8 | 21 | 2.16 | 0.32 | 44 (97.78) | 45 (9) | 0.41 |
| 9 | 28 | 1.77 | 0.33 | 39 (97.50) | 40 (8) | 0.41 |
| 10 | 35 | 1.43 | 0.37 | 35 (100.00) | 35 (7) | 0.48 |
| 11 | 42 | 1.10 | 0.41 | 30 (100.00) | 30 (6) | 0.49 |
| 12 | 49 | 1.24 | 0.51 | 25 (100.00) | 25 (5) | 0.49 |
| Hispanic or Latino | ||||||
| 13b | 14 | 3.42 | 0.34 | 46 (92.00) | 50 (10) | 0.32 |
| 14 | 21 | 3.09 | 0.35 | 42 (93.33) | 45 (9) | 0.33 |
| 15 | 28 | 2.70 | 0.39 | 37 (92.50) | 40 (8) | 0.33 |
| 16 | 35 | 1.69 | 0.37 | 35 (100.00) | 35 (7) | 0.49 |
| 17 | 42 | 1.60 | 0.48 | 30 (100.00) | 30 (6) | 0.50 |
| 18 | 49 | 1.28 | 0.50 | 25 (100.00) | 25 (5) | 0.52 |
| White | ||||||
| 19b | 14 | 2.10 | 0.32 | 48 (96.00) | 50 (10) | 0.45 |
| 20 | 21 | 1.69 | 0.39 | 44 (97.78) | 45 (9) | 0.52 |
| 21 | 28 | 1.38 | 0.41 | 39 (97.50) | 40 (8) | 0.54 |
| 22 | 35 | 1.20 | 0.46 | 35 (100.00) | 35 (7) | 0.64 |
| 23 | 42 | 1.10 | 0.49 | 30 (100.00) | 30 (6) | 0.70 |
| 24 | 49 | 0.76 | 0.49 | 25 (100.00) | 25 (5) | 0.75 |
| All stratified | ||||||
| 25b | 14 | 2.81 | 0.32 | 143 (95.33) | 150 (10) | 0.38 |
| 26 | 21 | 2.31 | 0.35 | 130 (96.30) | 135 (9) | 0.41 |
| 27 | 28 | 1.95 | 0.38 | 115 (95.83) | 120 (8) | 0.42 |
| 28 | 35 | 1.44 | 0.40 | 105 (100.00) | 105 (7) | 0.53 |
| 29 | 42 | 1.27 | 0.46 | 90 (100.00) | 90 (6) | 0.55 |
| 30 | 49 | 1.09 | 0.50 | 75 (100.00) | 75 (5) | 0.57 |
a
14-, 21-, 28-, 35-, 42-, and 49-day nowcasts started on July 26, August 2, August 9, August 16, August 23, and August 30, 2022, respectively, to provide 2, 3, 4, 5, 6, or 7 weeks of Wednesday-Tuesday data since study start date July 8, 2022. We mimicked nowcasts weekly, ending September 27, 2022, as the last Tuesday during the study period.
b
Indicates scenario applied in real time at the New York City Health Department.