Table 4. Variation in odds ratios (95% cIs) and percentage bias for demographic risk factors for hospital readmission within one year according to hesId, comparing data linkage algorithms.
| Reference standard (n = 175,773) |
Deterministic (n = 181,395) | Det+Probabilistic (n = 176,990) |
||||||
|---|---|---|---|---|---|---|---|---|
| Odds ratio (95% CI) |
Odds ratio (95% CI) |
Biasa | Odds ratio (95% CI) |
Biasa | ||||
| 1-year readmission rate | n | 18.7% | n | 18.4% | n | 18.7% | ||
| Total | 175,773 | 181,395 | 176,990 | |||||
| Maleb | 85,887 | 1.08 (1.05, 1.10) |
88,243 | 1.07 (1.04, 1.10) |
6% | 90,303 | 1.07 (1.05, 1.10) |
2% |
| Femaleb | 89,710 | (reference) | 92,818 | (reference) | 86,484 | (reference) | ||
| Age group | ||||||||
| 0 to 3 | 119,742 | 3.39 (3.22, 3.58) |
123,712 | 2.88 (2.74, 3.02) |
13% | 120,186 | 3.40 (3.23, 3.58) |
0% |
| 4 to 7 | 14,856 | (reference) | 15,510 | (reference) | 15,137 | (reference) | ||
| 8 to 11 | 10,787 | 1.08 (1.00, 1.16) |
11,306 | 0.99 (0.92, 1.06) |
119% | 10,945 | 1.07 (0.99, 1.15) |
15% |
| 12 to 15 | 8,920 | 0.94 (0.87, 1.02) |
9,186 | 0.83 (0.77, 0.89) |
9,049 | 0.93 (0.86, 1.00) |
||
| 16 to 19 | 10,080 | 1.19 (1.11, 1.28) |
10,239 | 1.04 (0.97, 1.12) |
77% | 10,243 | 1.18 (1.10, 1.27) |
6% |
| 20 to 23 | 11,388 | 1.31 (1.23, 1.41) |
11,442 | 1.15 (1.07, 1.23) |
50% | 11,430 | 1.31 (1.23, 1.41) |
1% |
| Ethnic group | ||||||||
| Missing | 44,733 | 0.52 (0.50, 0.54) |
47,518 | 0.51 (0.50, 0.54) |
−1% | 45,044 | 0.52 (0.51, 0.54) |
1% |
| White | 103,831 | (reference) | 106,070 | (reference) | 104,603 | (reference) | ||
| Mixed | 3,978 | 0.96 (0.88, 1.05) |
4,005 | 0.98 (0.90, 1.07) |
3,989 | 0.97 (0.88, 1.05) |
||
| Asian | 11,481 | 0.98 (0.93, 1.04) |
11,731 | 0.98 (0.93, 1.04) |
11,543 | 0.99 (0.94, 1.04) |
||
| Black | 6,457 | 0.78 (0.73, 0.84) |
6,557 | 0.81 (0.75, 0.87) |
13% | 6,491 | 0.8 (0.75, 0.86) |
9% |
| Chinese/Other | 5,293 | 0.73 (0.67, 0.79) |
5,514 | 0.72 (0.67, 0.78) |
−3% | 5,320 | 0.73 (0.68, 0.79) |
3% |
| Socio-economic group | ||||||||
| Missing | 50,322 | 0.04 (0.03, 0.04) |
51,746 | 0.03 (0.02, 0.04) |
−1% | 50,546 | 0.04 (0.03, 0.04) |
0% |
| Not deprived | 89,966 | (reference) | 92,727 | (reference) | 90,631 | (reference) | ||
| Deprived | 34,563 | 1.17 (1.14, 1.21) |
35,749 | 1.16 (1.12, 1.19) |
9% | 34,884 | 1.17 (1.13, 1.20) |
2% |
| No fixed abode | 782 | 0.58 (0.46, 0.74) |
1,031 | 0.40 (0.28, 0.56) |
−70% | 787 | 0.60 (0.47, 0.77) |
7% |
| Foreign | 140 | 0.42 (0.20, 0.86) |
142 | 0.42 (0.20, 0.87) |
2% | 142 | 0.47 (0.23, 0.95) |
14% |
| Data year | ||||||||
| 1998 to 2003 | 49,456 | (reference) | 53,463 | (reference) | 49,895 | (reference) | ||
| 2004 to 2009 | 50,348 | 1.83 (1.77, 1.91) |
51,429 | 1.92 (1.85, 1.99) |
−7% | 50,751 | 1.82 (1.76, 1.89) |
1% |
| 2010 to 2015 | 75,969 | 2.08 (2.01, 2.16) |
76,503 | 2.2 (2.12, 2.28) |
−7% | 76,344 | 2.08 (2.00, 2.16) |
0% |
a
This refers to the percentage by which the log odds coefficient in each model is over-or under-estimated, compared to the reference standard model 100*[(logitreference–logitcomparison/logitreference)], shown where the subgroup has a significantly increased risk of readmission in one year
b
Models exclude records were sex is missing (n = 334 after deterministic match, 203 after probabilistic match and 176 for reference standard)