Table 1.
Percentage of missing data and estimated odds ratios of significant exposures identified by different variable selection methods of logistic regression
| No. of missingness (%) | Estimated OR (95% CI) | ||||
|---|---|---|---|---|---|
| Exposure | Cases | Controls | Full model† | Backward variable selection | Forward variable selection |
| Direct contact with birds | 3 (2·4) | 0 (0) | 3·4 (1·3–8·4) | 3·2 (1·4–7·1) | 2·7 (1·1–6·1) |
| Direct contact with snakes | 2 (1·6) | 0 (0) | 4·3 (0·9–18·7) | 3·7 (1–13·4) | 3·9 (1·0–14·6) |
| Direct contact with a gecko* | 2 (1·6) | 0 (0) | 4·6 (0·8–30·5) | 4·0 (0·9–20·5) | 6·2 (1·3–34·2) |
| Living with dog having diarrhoea* | 20 (15·7) | 1 (0·1) | 3·3 (0·5–16·1) | 10·4 (3·3–33·4) | |
| Living with cat having diarrhoea* | 16 (12·6) | 6 (0·9) | 2·8 (0·3–18·5) | 7·0 (2·0–25·2) | |
| Visit pond or lake | 2 (1·6) | 0 (0) | 0·3 (0·1–0·7) | 0·3 (0·1–0·6) | 0·2 (0·1–0·5) |
| Consumption of | |||||
| Uncooked chicken | 5 (3·9) | 3 (0·4) | 1·4 (0·8–2·5) | 1·6 (0·9–2·6) | |
| Any food containing chicken | 10 (7·9) | 8 (1·2) | 2·6 (1·1–6·2) | 2·7 (1·5–5) | |
| Chicken cooked at home | 13 (10·2) | 9 (1·3) | 0·5 (0·2–1) | 0·4 (0·2–0·8) | |
| Chicken cooked outside home | 15 (11·8) | 11 (1·6) | 1·1 (0·6–2·2) | 2·1 (1·3–3·3) | |
| Uncooked ground beef* | 4 (3·1) | 0 (0) | 109·9 (8·6–3142·3) | 97·9 (9·2–2350·2) | 112·5 (13·1–2611·3) |
| Steak | 7 (5·5) | 5 (0·7) | 1·2 (0·7–2·1) | 1·8 (1–2·9) | |
| Roast beef | 7 (5·5) | 2 (0·3) | 0·5 (0·2–1·1) | 0·6 (0·3–1·1) | |
| Ground beef in spaghetti sauce, tacos | 14 (11) | 4 (0·6) | 0·5 (0·3–0·8) | 0·5 (0·3–0·8) | 0·6 (0·3–1) |
| Burger cooked at home | 19 (15) | 10 (1·5) | 1·0 (0·3–2·9) | 1·7 (1–3·1) | |
| Pasteurized milk | 3 (2·4) | 2 (0·3) | 0·6 (0·4–1·1) | 0·6 (0·3–0·9) | |
| Eggs | 10 (7·9) | 3 (0·4) | 6·5 (2·5–17·4) | 4·5 (2·2–9·4) | |
| Eggs cooked at home | 11 (8·7) | 4 (0·6) | 0·2 (0·1–0·8) | 0·2 (0·1–0·5) | |
| Scrambled eggs at home | 12 (9·4) | 4 (0·6) | 0·5 (0·2–1·3) | 0·6 (0·3–1·1) | |
| Boiled eggs at home | 13 (10·2) | 4 (0·6) | 1·5 (0·5–4·1) | 2·0 (0·9–4·2) | |
| Poached eggs at home* | 13 (10·2) | 4 (0·6) | 2·7 (0·5–13·4) | 6·3 (2–19·5) | |
| Scrambled eggs outside home | 13 (10·2) | 6 (0·9) | 0·2 (0–1·4) | 0·5 (0·2–1·2) | |
| Poached eggs outside home* | 13 (10·2) | 6 (0·9) | 0·1 (0–3211·9) | 5·8 (1·3–24·5) | |
| Cookie dough with raw egg | 2 (1·6) | 0 (0) | 0·2 (0–1·4) | 0·2 (0–1·0) | 0·2 (0–1·4) |
| Alfalfa sprouts | 6 (4·7) | 3 (0·4) | 0·4 (0–2·9) | 0·3 (0–1·4) | |
| Uncooked carrots | 9 (7·1) | 2 (0·3) | 0·5 (0·2–0·8) | 0·5 (0·3–0·8) | 0·5 (0·3–0·8) |
| Cantaloupe | 6 (4·7) | 1 (0·1) | 0·3 (0·1–0·8) | 0·5 (0·2–0·9) | 0·5 (0·2–1·0) |
| Watermelon | 5 (3·9) | 1 (0·1) | 2·1 (1·1–4·1) | 2·4 (1·3–4·3) | 2·4 (1·3–4·5) |
| Other melon | 5 (3·9) | 1 (0·1) | 1·2 (0–24·6) | 9·9 (1·1–76·7) | |
| Grapes | 13 (10·2) | 2 (0·3) | 1·4 (0·8–2·4) | 1·5 (0·9–2·4) | |
| Other pork | 9 (7·1) | 1 (0·1) | 0·5 (0·2–1·1) | 0·5 (0·3–1·0) | 0·4 (0·2–0·8) |
| Lamb | 1 (0·8) | 0 (0) | 0·1 (0–0·9) | 0·2 (0–1·0) | 0·2 (0–1·0) |
| Fish | 9 (7·1) | 2 (0·3) | 0·6 (0·3–1·1) | 0·6 (0·4–1·1) | |
OR, Odds ratio; CI, confidence interval.
*
Top ranked six exposures for permutation variable importance by random forest.
†
Estimates for the full model include only variables identified as significant by either the backward or the forward selection methods.