Work

Validation of self-reported sun exposure against electronic ultraviolet radiation dosimeters

Year: 2022

Type: article

Abstract: Ultraviolet radiation (UV) exposure is the main risk factor for skin cancer1 and skin cancer prevention research and health promotion programme evaluation relies on the accurate measurement of sun exposure using questionnaires. However, most previous studies have not had sufficient sample size to assess questionnaire measurement validity for different population subgroups and for different times of day. In this study, using data from the baseline measures in a randomized–controlled trial,2 we aimed to validate a self-reported questionnaire measure of sun exposure for different population subgroups and for different times of day against objective electronic UV dosimeters in a large Australian population-based sample (n = 896). A random sample of Australians aged 18–69 years living in all states and territories were invited to participate in the study via Medicare, the universal Australian healthcare system. Recruitment occurred during spring/summer from October 2017 to February 2019. Respondents were eligible if they had some European ancestry and no previous melanoma; 1025 (3%) participants gave consent and completed baseline measures for the trial, including a questionnaire and UV dosimetry. An online self-reported questionnaire was administered using REDCap and measured usual sun exposure from time spent outdoors on weekdays and weekends over the previous month. Participants were mailed a calibrated electronic UV dosimeter and asked to wear it on their wrist during daylight hours for 10 days, including on weekdays and weekends. Follow-up measures were also collected but this analysis used data from participants’ baseline assessment only to avoid potential confounding from the study intervention. From the dosimeter data we derived: (i) time spent outdoors exposed to UV, defined as any 8-s measurements with UV values of >0; and (ii) total standard erythemal doses (SEDs) as a measure of UV dose. The final analysis data set included 896 (87%) participants with weekday and weekend measurements that did not overlap with changeover to/from daylight savings. We used Spearman rank correlation and Bland–Altman plots to assess agreement3 between self-reported time spent outdoors and dosimetry-measured time outdoors and SEDs, for weekday, weekend and weekly exposure, overall and at different periods in the day (morning, peak, afternoon). We also performed subgroup analyses by time periods, gender, age, employment status, state of residence (grouped by latitude) and time spent outside during work. Statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA). See Supplementary material (available as Supplementary data at IJE online) for more details. Participants had a mean age of 46.9 years (SD 14.5) and 51% were female. Sun exposure was higher on weekends [median 0.8 SEDs, interquartile range (IQR) 0.3–1.7] than on weekdays (0.6, IQR = 0.3–1.2) and higher during the peak time (middle) of the day (0.3, IQR = 0.1–0.6) and afternoons (0.2, IQR = 0.1–0.5) than during mornings (0.1, IQR = 0.0–0.1). The weekly time per day spent outdoors estimated from the questionnaire (median 2.2 h/day, IQR = 1.3–3.3) was approximately double that estimated from the UV dosimeters (1.0 h/day, IQR = 0.6–1.6). Based on dosimeter-measured SEDs, there was a moderate correlation between weekday and weekend UV exposure (Table 1). Weekday–weekend correlation differed by age and employment type, but not by time of the day, gender, amount of time spent working outdoors or state of residence group. Spearman rank correlations between weekend and weekday ultraviolet radiation (UV) exposure measured as standard erythemal doses (SEDs) using electronic UV dosimeters rs, Spearman rank correlation; SA, South Australia; Vic, Victoria; Tas, Tasmania; Qld, Queensland; NT, Northern Territory; WA, Western Australia; NSW, New South Wales; ACT, Australian Capital Territory. Data were unavailable for the weekend morning period for two participants. The 95% CI for the Spearman correlation coefficients were calculated using the bias-corrected and accelerated (BCa) bootstrap method. Fisher’s r–z transformation and the ZPF statistic (Fisher r-to-Z transformed rs) was used to test whether the correlation coefficients between subgroups were different from each other. Spearman rank correlations between weekend and weekday ultraviolet radiation (UV) exposure measured as standard erythemal doses (SEDs) using electronic UV dosimeters rs, Spearman rank correlation; SA, South Australia; Vic, Victoria; Tas, Tasmania; Qld, Queensland; NT, Northern Territory; WA, Western Australia; NSW, New South Wales; ACT, Australian Capital Territory. Data were unavailable for the weekend morning period for two participants. The 95% CI for the Spearman correlation coefficients were calculated using the bias-corrected and accelerated (BCa) bootstrap method. Fisher’s r–z transformation and the ZPF statistic (Fisher r-to-Z transformed rs) was used to test whether the correlation coefficients between subgroups were different from each other. Table 2 shows a moderate correlation between questionnaire-measured self-reported time spent outdoors and UV dosimeter-measured time spent outdoors and SEDs, with a higher correlation for weekday than weekend exposure. The correlations were similar for dosimeter h/day and SEDs. The questionnaire–dosimeter correlations appeared weaker for students, but did not differ by time period, gender, age, state of residence or how much people worked outdoors. Spearman rank correlations between self-reported time outdoors (questionnaire) and UV dosimetry-measured time outdoors and standard erythemal doses (SEDs) rs, Spearman rank correlation; SA, South Australia; Vic, Victoria; Tas, Tasmania; Qld, Queensland; NT, Northern Territory; WA, Western Australia; NSW, New South Wales; ACT, Australian Capital Territory. The 95% CI for the Spearman correlation coefficients were calculated using the bias-corrected and accelerated (BCa) bootstrap method. Fisher’s r–z transformation and the ZPF statistic (Fisher r-to-Z transformed rs) was used to test whether the correlation coefficients between subgroups were different from each other. Spearman rank correlations between self-reported time outdoors (questionnaire) and UV dosimetry-measured time outdoors and standard erythemal doses (SEDs) rs, Spearman rank correlation; SA, South Australia; Vic, Victoria; Tas, Tasmania; Qld, Queensland; NT, Northern Territory; WA, Western Australia; NSW, New South Wales; ACT, Australian Capital Territory. The 95% CI for the Spearman correlation coefficients were calculated using the bias-corrected and accelerated (BCa) bootstrap method. Fisher’s r–z transformation and the ZPF statistic (Fisher r-to-Z transformed rs) was used to test whether the correlation coefficients between subgroups were different from each other. Bland–Altman plots (Supplementary Figure S1, available as Supplementary data at IJE online) show that the questionnaire overestimated the time spent outdoors compared with the UV dosimeter (weekly average mean difference = 1.40 h/day, 95% limits of agreement –1.57, 4.36). The plots indicate that with greater time spent outdoors on both weekdays and weekends, there was increased variability in differences between the questionnaire and dosimetry measures, with a tendency to greater overestimation in the questionnaire. In conclusion, we found that this sun exposure questionnaire adequately ranks personal sun exposure at different times of day for different population subgroups, but may over-estimate the duration of UV exposure. Thus, the questionnaire may be less appropriate to use if the intended purpose is to accurately measure the duration of UV exposure. Validity was lower for weekend UV exposure, indicating that its variable nature makes this exposure more difficult to measure. The levels of agreement were similar to previous studies.4–7 Measurement of UV exposure over both weekdays and weekends is important for accurate assessment in different population subgroups, particularly for students and younger people. Strengths of this study include the large population-based national sample and the use of objective time-stamped electronic dosimeters as a validation measure. The dosimeter was worn on the wrist, which is a reliable body site for measuring personal UV, but may only register ∼50% of the ambient exposure compared with when worn on the top of the head.8,9 This limitation may partly explain the greater time spent outdoors measured by questionnaire compared with the dosimeter, although any UV value of >0 was counted and the dosimeters should always measure some UV when outside in spring/summer due to the relatively strong diffuse component. Further, as non-parametric correlations compare ranks (not means), this has minimal impact on our correlation results. The consent rate was low, which may influence generalizability, although it was similar to other population-based research studies recruiting through the Medicare database.10 Ethics approval for the study was obtained from the Human Research Ethics Committee at the University of Sydney (project 2017/163) and informed consent was provided by all participants. De-identified data from this analysis are available on request by contacting the corresponding author. Ethics approval to share the data will be required. Supplementary data are available at IJE online. A.E.C. designed the study in collaboration with the Managing Your Risk Study Investigators (including M.A., M.G.K., S.N.L., P.A.K., M.H.L., G.J.M. and A.K.S.). A.K.S. and A.E.C. directed the study’s implementation. A.K.S. and G.R. co-ordinated the collection of dosimetry and questionnaire data. A.E.C. and D.E. designed the analytical strategy. R.Z. and D.E. conducted the analysis and prepared the tables and figures. R.Z. and A.E.C. drafted the paper. M.A. supplied the UV dosimeter devices and provided expertise on dosimetry calibration and measurement. S.N.L. provided advice on the analysis. A.R.S. assisted with data collection. All authors provided critical review of the draft paper. The study was funded by a National Health and Medical Research Council (NHMRC) project grant (1129822) and endorsed by the Melanoma and Skin Cancer Trials Group (ANZMTG 03.17). A.K.S. received a Research Training Program (RTP) Stipend Scholarship and a Merit Top Up Scholarship from the University of Sydney, and a Melanoma Institute Australia Postgraduate Research Scholarship. A.E.C. was supported by a NHMRC Career Development Fellowship (1147843) and NHMRC Investigator Grant (2008454). We thank other investigators and research staff who contributed to the Managing Your Risk Study including: Phyllis Butow, Kate Dunlop, Brooke Beswick, Hugh Dawkins, Suzanne Dobbinson, Georgina Fenton, Louise Keogh, Judy Kirk, Cynthia Low, Ainsley Newson, Jacqueline Savard, Lyndal Trevena, Sarah Wordsworth, Rachael Morton, Caro Badcock, Lauren Humphreys, Ainsley Furneaux-Bate, Rosa Evaquarta, Sinem Gultekin and Antigone Matsos. We thank Services Australia Assisted Mailout service for assistance with recruitment. None declared.

Source: International Journal of Epidemiology

DOI: 10.1093/ije/dyac179