In the survey, 397 healthcare workers gave an affirmative answer to the question “Have you had hand eczema during the past 12 months.” Of these, 255 (64%) agreed to participate in the trial and no further recruitment was possible. Of the 142 healthcare workers contacted by email and telephone to ask about reasons for not participating, 103 (72%) responded: 66 did not want to participate, 15 were pregnant, 13 had changed job or moved, 4 could not spare the time, 3 were taking immunosuppressive drugs, 1 was on long term sick leave, and 1 had died. Overall, 123 participants were randomised to the intervention group and 132 to the control group (figure). Randomisation was stratified according to hospital, profession, and clinical scoring of disease severity.
Table 1 shows the characteristics of the participants in the two groups. One participant in the intervention group and two participants in the control group required treatment as usual for severe eczema. They were prescribed topical corticosteroids.
Table 1 Characteristics of participants in skin care education and counselling group and treatment as usual group (control) at enrolment. Values are numbers (percentages) unless stated otherwise
Dropouts and missing values
Follow-up data were available for 247 of the 255 (97%) participants. In the intervention group, 122 of 123 participants received the intervention as planned; one did not attend. Two participants were excluded at follow-up because they were pregnant and one did not attend. In the control group one participant was excluded because systemic corticosteroids had been prescribed and three participants did not attend follow-up.
Table 2 shows the absolute and percentage distribution of missing values for the outcome measures. Only for the outcome use of moisturisers at work was the percentage of missing values significantly related to the intervention (Fisher’s exact test P=0.034). Otherwise the differences in percentage of missing values between the intervention groups were not significant. In five instances more than 5% of the values were missing.
Table 2 Missing data for planned outcome measures
Disease severity, quality of life, knowledge of hand eczema, and skin protection
Table 3 shows the distribution of scores on the hand eczema severity index (the primary outcome), scores on the dermatology life quality index, and scores on knowledge of hand eczema and skin protection. The intervention group had significantly lower mean scores on the hand eczema severity index than the control group: difference of means: unadjusted −3.56 (95% confidence interval −4.92 to −2.14); adjusted −3.47 (−4.80 to −2.14), in both cases P<0.001 for difference. After the missing values had been imputed to obtain the worst case scenario the means of the groups still differed significantly (5.57 v 8.01, parametric test P=0.003). Since the assumptions of the parametric test were not quite fulfilled, a non-parametric test was also carried out after the imputation (Mann Whitney test P<0.001).
Table 3 Distributions of primary and two secondary outcome measures in each group and difference between groups. Values are means (95% confidence intervals) unless stated otherwise
The intervention group also had significantly lower mean scores on the dermatology life quality index than the control group: difference of means: unadjusted −0.78, non-parametric test P=0.003; adjusted −0.92 (−1.48 to −0.37). Knowledge of hand eczema and skin protection at follow-up was not statistically significantly different between the groups.
Self evaluated disease severity and skin protective behaviour
Table 4 shows the results of the ordinal regression analyses. Self evaluated disease severity was lower in the intervention group than in the control group (P=0.001). This may be assessed from table 5, which shows the estimated probability of each answer in the intervention groups for those outcomes showing a significant difference in answer profiles between the two groups. Thus the intervention group had a higher probability of self reporting “mild” disease than the control group (62.0% v 41.5%). By contrast the probabilities of self reporting “moderate” and “severe” were both higher in the control group. The groups also differed significantly for scores on hand washing at work (P=0.0047) and use of protective gloves during wet work (P=0.048), cooking (P<0.0005), and cleaning (P=0.0065). The estimated probabilities showed that in all cases the beneficial effect was associated with the intervention group (table 5).
Table 4 Ordinal regression analyses of secondary outcomes without and with adjustment by protocol specified stratification variables and baseline value
Table 5 Estimated probabilities of various possible answers in each intervention group for those questionnaires where there was a significant difference between the groups’ response patterns
The two groups did not differ significantly for use of disinfectants (P=0.12) or moisturisers at work (P=0.20).
Number of eruptions
Differences in the number of eruptions could not be calculated. Forty one participants reported having “constant” hand eczema instead of providing the number of eruptions.