Table 2.
Predictors of ‘The NHS of today is a good employer when doctors become ill themselves’ by cohort year, gender, specialty, and retirement status.
| Univariable analysis |
Multivariable analysis* |
||||||||
|---|---|---|---|---|---|---|---|---|---|
| Group | Disagreement (%) | Neither (%) | Agreement (%) | N | df | χ | p | OR (95% CI) | p |
| All | 42.6 | 29.0 | 28.4 | 2594 | |||||
| Cohort | 2 | 2.1 | 0.345 | ||||||
| 1977† | 41.2 | 30.4 | 28.5 | 1040 | – | – | |||
| 1974† | 43.6 | 28.0 | 28.4 | 1554 | – | – | |||
| Gender | 2 | 16.8 | <0.001 | ||||||
| Women‡ | 48.7 | 26.6 | 24.6 | 759 | 1 | ||||
| Men | 40.1 | 29.9 | 30.0 | 1835 | 1.4 (1.2, 1.8) | <0.001 | |||
| Specialty | 6 | 48.5 | <0.001 | ||||||
| GP‡ | 49.1 | 28.0 | 22.9 | 1226 | 1 | ||||
| Hospital medical specialties | 36.3 | 30.4 | 33.3 | 306 | 1.9 (1.4, 2.6) | <0.001 | |||
| Surgery | 36.2 | 29.4 | 34.4 | 282 | 1.8 (1.3, 2.5) | <0.001 | |||
| Other hospital | 37.3 | 29.7 | 32.9 | 780 | 1.9 (1.5, 2.4) | <0.001 | |||
| Retirement status | 6 | 24.5 | <0.001 | ||||||
| Retired, not now working in medicine‡ | 47.3 | 25.6 | 27.0 | 1132 | 1 | ||||
| Retired, and ‘returned’ for some medical work | 41.3 | 31.1 | 27.6 | 758 | 1.0 (0.8, 1.3) | 0.973 | |||
| Working full-time in medicine | 36.1 | 31.1 | 32.7 | 440 | 1.2 (0.9, 1.6) | 0.139 | |||
| Working part-time in medicine | 37.1 | 33.3 | 29.5 | 264 | 1.4 (1.0, 2.0) | 0.031 | |||
*The multivariable analysis is based on the comparison of the ‘% agreement’ results with the other two response categories combined.
†Cohort was not significant univariably and so was excluded from the model.
‡Reference group for multivariable model.
Note: A total of 2691 of 3550 doctors (Appendix 1) answered the question. ‘Univariable’ denotes single factor χ2 test for each predictor. The univariable analysis excluded doctors falling outside of the four specialty groups in the table above: this reduced the sample size to 2594. ‘Multivariable’ denotes binomial logistic regression result for each predictor with all other predictors in the model. We excluded cases where one or more predictors were missing, where the dependent variable was missing, or where the respondent was undecided: this reduced the sample size for the logistic regression from 2594 to 1843.
The odds ratio (OR) indicates whether a (randomly chosen) member of the group in question was more, or less, likely than a member of the reference group to agree with the statement. For example, a man would be 1.4 times as likely as a woman to agree with the statement.