1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Self-reported sun exposure is commonly used in research, but how well this represents actual sun exposure is poorly understood. From February to July 2011, a volunteer sample (n = 47) of older adults (≥45 years) in Canberra, Australia, answered brief questions on time outdoors (weekdays and weekends) and natural skin color. They subsequently maintained a sun diary and wore an ultraviolet radiation (UVR) digital dosimeter for 7 days. Melanin density was estimated using reflectance spectrophotometry; lifetime sun damage was assessed using silicone casts of the back of the hand; and serum 25-hydroxyvitamin D (25(OH)D) concentration was assayed. Questionnaire-reported time outdoors correlated significantly with diary-recorded time outdoors (Spearman correlation rs = 0.66; 95% CI 0.46, 0.80; < 0.001) and UVR dosimeter dose (r= 0.46; 95% CI 0.18, 0.68; = 0.003), but not 25(OH)D concentration (rs = 0.24; 95% CI −0.05, 0.50; = 0.10). Questionnaire-reported untanned skin color correlated significantly with measured melanin density at the inner upper arm (rs = 0.49; 95% CI 0.24, 0.68; < 0.001). In a multiple linear regression model, statistically significant predictors of 25(OH)D concentration were self-reported frequency of physical activity, skin color and recent osteoporosis treatment (R2 = 0.54). In this study, brief questionnaire items provided valid rankings of sun exposure and skin color, and enabled the development of a predictive model for 25(OH)D concentration.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Sun exposure has both risks and benefits for human health [1] that are particularly apparent where there is a mismatch between the level of ambient ultraviolet radiation (UVR) and the skin type that has evolved to be best suited to that environment. Thus, Australians of European origin being fair skinned, but living in high ambient UVR conditions, have the highest skin cancer incidence in the world—skin cancer accounts for ca 80% of newly diagnosed cancer cases [2] and costs the health care system more than any other form of cancer [3]. On the other hand, there are now increasing reports of vitamin D deficiency in dark-skinned migrants moving to higher latitude (lower UVR) locations [4, 5]. Investigation of the role (and possibly pattern) of individual sun exposure as a cause of these conditions has been made challenging by the difficulty of measuring sun exposure at different life stages and across a range of study types. For small-scale studies assessing current sun exposure, recently developed digital UV dosimeters may be valuable, but their use is not feasible in large population cohorts or for data collection over long time periods. In these situations, there is little choice but to use self-report questionnaire data, and this is indeed the method most commonly used in epidemiological research.

Numerous factors related to both the environment (e.g. solar zenith angle and weather) and the person (e.g. time outdoors and use of sun protection) affect the level of UVR reaching the skin. These factors threaten the validity of self-reported sun exposure as a measure of received UVR dose. Several studies do support the validity of self-reported sun exposure [6-13], but these have most commonly assessed the validity of diary (rather than questionnaire) instruments. Herein, we extend this past work by comparing responses to a brief questionnaire-based measure of “time outdoors” to a more detailed (less subjective) diary measure of time outdoors, as well as to objectively measured UVR exposure obtained from personal digital dosimeters. We further compare questionnaire responses of time outdoors to an objective measure of actinic damage (silicone casts of the skin of the back of the hand), which provides a measure of cumulative UVR exposure, and to vitamin D status (as assessed by serum 25-hydroxyvitamin D concentration; 25(OH)D), thought to reflect recent sun exposure. This allows us to assess the association between the questionnaire measures and biologically effective UVR dose received by an individual. We also assess the validity of a questionnaire measure of natural skin pigmentation, as this is an important genetic moderator of the biologically effective UVR dose received by an individual. Finally, we investigate the value of simple questionnaire measures as predictors of vitamin D status.

The sun exposure measures investigated herein are based on those used in a number of studies [14-18], and are identical to those used in the 45 and Up Study, an ongoing population-based cohort study, involving more than 260 000 residents aged 45 years or over from Australia's most populous state, New South Wales (NSW; [19]). The 45 and Up Study aims to investigate the determinants of healthy aging, including environmental determinants such as sun exposure. The brief and easily administered questionnaire items utilized in this study are potentially of great value in assessing personal sun exposure and its association with health outcomes on a large scale. However, it is first essential that the association between these measures and actual sun exposure, and thus their validity, be investigated.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Study sample and setting

A volunteer sample of persons aged 45 years and over was recruited from the Australian Capital Territory (ACT), with data collection spanning 20 February to 18 July 2011 (end of summer to mid-winter). The ACT has a relatively dry continental climate, with records indicating that the year 2011 had below average rainfall at 580.1 mm (historical average = 614.4 mm), but close to average temperatures [20]. Considering the main study period (March–June), March was the warmest month and June the coldest, with data collected by the Australian Bureau of Meteorology at the Canberra Airport weather station indicating a mean maximum daily temperature of 23.3°C and mean daily global solar exposure of 19.2 MJ m−2 for March, and a mean maximum daily temperature of 13.6°C and mean daily global solar exposure of 10.2 MJ m−2 for June [21]. Study recruitment was by distribution of information brochures around university campuses, shopping centers, libraries and community centers, as well as by word of mouth, and snowball recruiting. Participants were required to be at least 45 years old, be English speaking and able to attend two interviews at the Australian National University.

Data collection

Participants were mailed a questionnaire of selected demographic and health-related information taken from the baseline questionnaire used in the 45 and Up Study (available at, including date of birth, height, weight, country of birth, work status, qualifications, smoking history, physical activity in the last week, dietary behaviors (e.g. foods never eaten, and weekly intake of meat, seafood and cheese) current and past health, medication use (including recent treatment for osteoporosis), sun exposure, sun sensitivity and past history of skin cancer and broken bones/fractures. The questions centrally relevant to sun exposure (which were identical to those on the 45 and Up Study questionnaire) were as follows: “About how many hours a DAY would you usually spend outdoors on a weekday and on the weekend?” (time outdoors), with open entry responses for a weekday and weekend day; and “What best describes the colour of the skin on the inside of your upper arm, that is your skin colour without any tanning?” (untanned skin colour), scored on a 6-scale item from “very fair” to “black.” Following self-completion of the questionnaire, participants attended two interviews with the researcher, before and after a 7-day period during which a sun diary was completed and UVR dosimeter worn.

At Interview 1, participants were asked, in retrospect (i.e. after completion of the questionnaire), which of the following time periods they were thinking about when they answered the time outdoors question on the questionnaire: the previous week; the previous month; the current season; a sort of average over the whole year; other (please explain). Height and weight were then measured using a stadiometer and digital scales, and reflectance spectrophotometer (CM-2500d; Minolta, Japan) readings were taken of the left inner upper arm. Specifically, measurements were made half way between the midpoint of the axilla and the medial epicondyle of the humerus at a site without scars or moles, with reflectance readings at 400 and 420 nm used to calculate melanin density as previously described [22]. Silicone skin casts of the dorsum of each hand were made as previously described [23], but using Affinis light body silicone-based impression material. Casts were digitally photographed under a light microscope and double-graded by independent scorers on a scale of 1–6 using the Beagley–Gibson method to score actinic damage [24].

For the following seven consecutive days, participants completed a sun diary, recording time spent outdoors each hour between the hours of 6 A.M. and 8 P.M. (0, 0–15, 15–30, 30–45, or 45–60 min), indoors and outdoors physical activity, clothing worn and sunscreen use. A digital UVR dosimeter was worn concurrently. The UVR dosimeters (see Fig. 1) were developed at the University of Canterbury, New Zealand (and manufactured by Scienterra, New Zealand) to digitally measure personal UV exposures in behavioral studies [25]. They are lightweight, weather-proof, compact (36 mm diameter, 12 mm breadth) and battery operated (3-volt coin cell). They are based on a visible-blind AlGaN photodiode with a spectral response that closely matches the CIE erythemal action spectrum and a UV diffuser to mimic the cosine response of human skin. Dosimeters were supplied by the Australian Radiation Protection and Nuclear Safety Agency (ARPANSA), and the calibration, configuration and conversion of logged data into erythemally effective units (UV Index, standard erythemal dose [SED]) were performed at the ARPANSA laboratory (Melbourne, Australia). The badges were configured to make time-stamped measurements at 10 sec intervals from 6 A.M. to 8 P.M. Velcro bands were attached to the dosimeters to allow them to be worn around a participant's wrist on the outside of clothing. The wrist has been previously validated as a reliable site for personal dosimetry [26].


Figure 1. UVR dosimeter system: dosimeter as worn in study (a) and docking cradle for dosimeter (b).

Download figure to PowerPoint

At Interview 2, the diary and dosimeter were collected, and participants were asked whether they considered the badge-wearing week to have been usual for them in terms of time spent outdoors. A blood sample (5 mL) was taken, centrifuged within 4 h of collection and the serum aliquoted and placed in a −68°C freezer until completion of participant interviews. The serum 25(OH)D concentration was assayed using liquid chromatography–tandem mass spectrometry (LC-MS/MS), at RDDT Laboratories (Melbourne, Australia). This assay detects, separately, the levels of 25(OH)D2 and 25(OH)D3. The coefficients of variance ranged from 0.61% to 1.51% for the three 25(OH)D3 QC determinants, and their means fell within historical limits. No samples had detectable levels of 25(OH)D2. Results are presented as total 25(OH)D concentration, but are in effect the measured levels of 25(OH)D3.

Statistical analysis

Descriptive statistics for continuous data are presented as median (lower quartile [Q1], upper quartile [Q3]) due to the skewed distribution of a number of variables and the small sample size. Nonparametric Spearman rank correlations (rs) were used to analyze the rank association between variables, and the degree of association rated as very weak (rs = 0–0.19), weak (rs = 0.2–0.39), moderate (rs = 0.40–0.59), strong (0.6–0.79) or very strong (rs = 0.8–1.0; [27]). The Wilcoxon signed-rank test and the Wilcoxon rank-sum test were used to make comparisons between paired and unpaired data respectively.

From diary records, daily time outdoors was estimated by summing the median values in the range indicated as spent outdoors for each hour of the day (e.g. 0–15 min = 7.5 min). To minimize missing values, we used the time spent in outdoor physical activities, when recorded, to impute time spent outdoors when the latter was missing. For example, if time outdoors was not recorded from 10 to 11 A.M., but outdoor physical activity of 20 min was recorded for that same hour, time outdoors was imputed as 15–30 min. If the diary was not completed (after imputation) or the dosimeter not worn for any period of time between the peak UVR hours of 10 A.M. and 4 P.M., the day was treated as a missing value for analyses involving diary time outdoors and UVR exposure respectively. When fewer than 7 days of diary or UVR data were available for participants, data were averaged over the available number of days to derive weekday and weekend mean time outdoors (h/day) and UVR exposure (SED/day). Analyses were computed separately for weekdays and weekends, as well as using a derived daily average for the week ((weekday average × 5 + weekend average × 2)/7).

An intraclass correlation coefficient of the type 2,1 (ICC 2,1) as described by Shrout & Fleiss [28] was used to assess quantitative agreement between the questionnaire and diary time outdoors measures. Data were square-root transformed to improve normality prior to the computation, and agreement was rated as poor (ICC < 0.40), fair-to-good (ICC = 0.40–0.75) or excellent (ICC > 0.75; [29]). Agreement was further assessed using the Bland–Altman (BA) method [30], by plotting the difference between the questionnaire and diary measurements of time outdoors for each participant against the mean of these measurements, which allows both random and systematic errors to be examined visually [31].

From questionnaire and interview data for height and weight, Body Mass Index (BMI) was calculated using the standard formula. BMI measured at interview was used in all analyses. Participants' overall physical activity levels were estimated from questionnaire measures related to the number of weekly sessions and total duration of each of: walking continuously (for at least 10 min), moderate physical activity and vigorous physical activity. A weekly average number of sessions (total sessions) was obtained by summing the total number of sessions of each activity [32]. A weighted weekly average of time spent doing physical activity (total duration) was calculated by summing time spent doing each activity, with vigorous physical activity receiving twice the weighting [32].

As 25(OH)D levels show seasonal variation, we adjusted for date of blood draw using methods previously described [33-36]. As a first step, we fitted sine and cosine curves to the data, providing a periodic regression model of the form: 25(OH)D = β1sin(2πt/365) + β2cos(2πt/365) + β0, where t is the day of the year the blood sample was collected. To obtain a deseasonalized 25(OH)D level for each participant, we subtracted the predicted (i.e. model derived) 25(OH)D concentration from the corresponding measured value, and then added the overall predicted mean. Using these deseasonalized values as the dependent variable, a backwards automated stepwise regression was used to produce an initial model of 25(OH)D determinants using data from questionnaire responses. A significance level of 0.1 was set for removal from the model. Potential predictor variables were selected on the basis of identified correlations with the seasonally adjusted 25(OH)D concentration in the present study, as well as plausible associations based on existing evidence. Following the automated regression to derive an initial model, we used a purposive selection method to refine and develop a final model [37]. Variables omitted from the initial model, but that had a plausible association with 25(OH)D concentration, were inserted back into the model and nested models compared using a likelihood ratio test.

Distributions of the residuals in the final model were examined, and assumptions of normality and homoscedasticity verified using the Shapiro–Wilk and Breusch–Pagan tests respectively (violations indicated if < 0.05). To assess the model variance attributable to each predictor, the full final model was compared with reduced models in which the variable of interest was omitted. The sum-of-squares (SSR) attributable to that predictor was calculated as the regression SSR for the full model (SSRfull) minus the SSR for the reduced model (SSRreduced), and expressed as a proportion of the SSR for the full model (i.e. (SSRfull − SSRreduced)/SSRfull).

Inter-rater agreement for cast-scoring was assessed using a linearly weighted kappa statistic, with a 95% confidence interval calculated using a bootstrapped (1000 repetitions), bias-corrected method [38]. Level of agreement was interpreted using the guidelines suggested by Altman ([39]). Where discrepancies occurred in the grading, scorers reviewed and discussed the features of these casts and assigned a single grade. Of 94 silicone casts, two (from the same individual) were of too poor a quality to score. Only grades from left-hand casts were used for analyses [23]. Logistic regression was used to compute odds ratios for high skin cast damage scores (6, compared to <6).

Stata Statistical Software (version 9 for Windows) was used for data analysis (StataCorp. (2005) Stata Statistical Software: Release 9. College Station, TX: StataCorp LP.). For all measures of effect, P-values and 95% confidence intervals (95% CI) are presented in parentheses, with the lower and upper bounds of the 95% CI separated by a comma. Statistical tests were two-sided and considered statistically significant if < 0.05.


Ethics approval was obtained from the Human Research Ethics Committee of the Australian National University prior to study commencement. All participants gave written informed consent.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Characteristics of study sample

Characteristics of the study sample are presented in Table 1. The 47 study participants ranged in age from 46 to 75 years (median 58) at the time of questionnaire completion. The volunteers were predominantly university educated and there was a slight female preponderance (1.2:1). The majority (64%) rated their health as very good or excellent.

Table 1. Characteristics of study sample (n = 47)
Characteristica n Median (Q1, Q3) or %
  1. a

    All variables derived from questionnaire except BMI (measured at interview), 25(OH)D concentration, diary time outdoors and UVR exposure (digital dosimeter monitoring).

Age (years)4758 (52, 64)
Highest qualification
Less than university degree1531.9%
University degree or higher3268.1%
Overall health
Very good2348.9%
BMI4726.7 (23.4, 30.2)
25(OH)D level (nmol L−1)4672.1 (56.0, 86.7)
Physical activity
Total sessions (sessions/week)4713 (8, 18)
Total duration (h/week)4610.5 (7.0, 16.5)
Questionnaire time outdoors (h/day)
Weekday471.0 (1.0, 2.0)
Weekend474.0 (2.0, 5.0)
Diary time outdoors (h/day)
Weekday461.3 (0.8, 1.8)
Weekend462.6 (1.0, 3.3)
UVR exposure (SED/day)
Weekday410.12 (0.05, 0.45)
Weekend400.42 (0.11, 1.06)
Untanned skin color
Very fair817.0%
Light olive1429.8%
Dark olive12.1%
Skin response to sun exposure
Never tans36.4%
Gets mildly or occasionally tanned817.0%
Gets moderately tanned2144.7%
Gets very tanned1531.9%
Ever had
Skin cancer (not melanoma)1021.3%
Treatment in the last month for
Osteoporosis or low bone density612.8%

Participant adherence

Most participants wore the UVR dosimeter and completed the sun diary as instructed (both dosimeter and diary data were available for 302 of 329 days, 92% compliance). However, dosimeter malfunction resulted in the loss of 56 days of data (17%), reducing the number of days with both diary and dosimeter data available to 250 (76% or 5.3 days/person on average).

Questionnaire, diary and dosimeter measures of sun exposure

Questionnaire and diary time outdoors

The distribution of questionnaire time outdoors was positively skewed, with a median weekly average of 2 h per day (1.6, 3.0 h per day). Regarding the reference period used when responding to the time outdoors question, most people indicated they were referring to “recent” sun exposure, that is, the previous week (34%), the previous month (9%) or the current season (23%). Approximately one-third of the participants (32%) said they were considering an average over the whole year.

Diary records for time outdoors were obtained for 46 of 47 participants (312 of 329 days; 95%). The distribution of diary time outdoors was positively skewed, with a median weekly average time outdoors of 1.7 h per day (1.1, 2.2 h per day). When asked whether the diary-completing (and badge-wearing) period had been usual for them in terms of time spent outdoors, 12 participants (26%) said it had not been. Of these, all but one indicated that they had spent less time outdoors than usual, with the reason most typically cited as the weather (too cold), or illness/injury. Time outdoors on the weekend was significantly greater than on weekdays (< 0.001) in both questionnaire and diary data, but weekday and weekend time outdoors were not significantly correlated in either the questionnaire (= 0.18) or diary (= 0.12) data.

Dosimeter measured UVR exposure

Personal UVR doses were available for 41 participants on weekdays and 40 participants on weekends. Weekday averages were based on fewer than 5 days for 8 of 41 participants, and weekend averages for 8 of 40 participants were based on only 1 day of dosimeter readings. The median average weekly UVR exposure was 0.30 SEDs/day (0.12, 0.59 SEDs/day). UVR exposure was significantly higher (= 0.005) on weekends (median = 0.42 SEDs/day; 0.11, 1.06) than weekdays (median = 0.12 SEDs/day; 0.05, 0.45) and there was a significant correlation between weekday and weekend exposure (rs = 0.38; 95% CI 0.08, 0.62; = 0.01).

Correlations between questionnaire, diary and dosimeter measures

Table 2 presents the correlation analyses performed between questionnaire, diary and dosimeter measured sun exposure according to day type. Moderately sized and statistically significant correlations were observed between questionnaire and diary reports of time outdoors on weekdays and weekends and between questionnaire time outdoors and dosimeter UVR exposure. Excluding participants who indicated the study period had been non-typical did not appreciably alter the correlations (questionnaire-diary rs = 0.62; 95% CI 0.36, 0.79; < 0.001 and questionnaire-dosimeter rs = 0.44; 95% CI 0.08, 0.69; = 0.02).

Table 2. Spearman rank correlations between questionnaire, diary and dosimeter measures of sun exposure
Measures compared n r s (95% CI)P-value
Questionnaire and diary
Weekday460.54(0.30, 0.72)<0.001
Weekend460.51(0.26, 0.70)<0.001
Week460.66(0.46, 0.80)<0.001
Questionnaire and dosimeter
Weekday410.50(0.22, 0.70)0.001
Weekend400.40(0.10, 0.63)0.01
Week400.46(0.18, 0.68)0.003
Diary and dosimeter
Weekday400.51(0.23, 0.71)<0.001
Weekend390.48(0.19, 0.69)0.002
Week390.41(0.10, 0.64)0.01
Agreement between self-report measures of time outdoors

Questionnaire reports of time outdoors were overestimates of that obtained from the diary for 54% of participants on weekdays (median difference 0.14 h per day; −0.45, 0.75), and 74% of participants on weekends (median difference 1.13 h per day on a weekend; −0.06, 2.25). Intraclass correlation analyses indicated fair-to-good agreement on weekdays (ICC = 0.49; 95% CI 0.24, 0.68), weekends (ICC = 0.43; 95% CI 0.04, 0.68) and for average weekly time outdoors (ICC = 0.56; 95% CI 0.22, 0.76). A Bland–Altman plot for weekly time outdoors is shown in Fig. 2. This indicated a greater tendency toward overestimation among participants with greater mean time outdoors. The degree of overestimation was significantly correlated with questionnaire time outdoors (rs = 0.60; 95% CI 0.38, 0.76; P < 0.001).


Figure 2. Bland–Altman plot of the discrepancy between questionnaire and diary records of average weekly time outdoors. Horizontal lines mark the mean discrepancy (solid line) and 95% limits of agreement (dashed lines).

Download figure to PowerPoint

Questionnaire measures and 25(OH)D concentration

Blood samples were obtained for 46 study participants (98%). Over half (54%) of the study sample had 25(OH)D levels (unadjusted) below 75 nmol L−1; 13% had levels less than 50 nmol L−1. Seasonally adjusted 25(OH)D levels were compared to questionnaire time outdoors using univariate correlation analyses. Overall, there was no significant correlation between 25(OH)D levels and questionnaire measures of weekday (rs = 0.22; 95% CI −0.07, 0.48; = 0.13), weekend (rs = 0.09; 95% CI −0.21, 0.37; = 0.56) or cumulative weekly time outdoors (rs = 0.24; 95% CI −0.05, 0.50; = 0.10).

Questionnaire measures and skin casts

As assessed by a weighted kappa statistic (κw = 0.63, 95% CI 0.50, 0.74), inter-rater agreement was good [39]. In this older population living in a high ambient UVR environment, the distribution of skin cast scores was highly skewed toward the highest grade of skin damage; around 52% of casts (n = 24) were scored as grade 6 and another 35% (n = 16) as grade 5. Odds ratios for a cast score of six (compared to < 6), adjusted for age (data not shown) were greater than 1 in association with greater “time outdoors,” a history of skin cancer and with being born in Australia, but were not statistically significant (P-values ranged between 0.05 and 0.25).

Questionnaire versus objective measures of skin tone

The median melanin density derived from spectrophotometer readings was 2.2% (1.7, 3.2%) and melanin density was moderately correlated with self-reported untanned skin color (rs = 0.49; 95% CI 0.24, 0.68; < 0.001). There was also a significant association between melanin density of the inner upper arm and self-reported skin response to sun exposure in summer (rs = 0.30; 95% CI 0.01, 0.54; = 0.04). However, Fig. 3 shows that there was considerable overlap in the melanin densities of those rating their skin “very fair” or “fair,” and for those reporting that their skin “never” or “gets mildly tanned” in response to repeated sun exposure.


Figure 3. Schematic box-plots of melanin density (of inner upper arm) by self-reported “untanned skin color” (a) and “skin response to sun exposure” (b). Plots show the median (line within the box), interquartile range (box), upper and lower adjacent values (bars at extreme of plots) and outside values (dots). Sample sizes for each category are shown in Table 1.

Download figure to PowerPoint

Predictive modeling of 25(OH)D level

Questionnaire variables considered as potential predictors of seasonally adjusted 25(OH)D concentration were age, sex, BMI, untanned skin color, osteoporosis treatment, vitamin D supplementation, physical activity (both total duration and sessions per week) and time outdoors. The results from univariate and multivariate modeling are shown in Table 3. The major contributor to the final model was total weekly sessions of physical activity, with treatment of osteoporosis or low bone density in the last month also important. Untanned skin color was the other significant predictor—self-reported darker skin tones were associated with higher 25(OH)D concentration. The forced re-insertion of specific variables, notably time outdoors, BMI and age, did not significantly improve the model, with likelihood ratio tests used to check for any improvements in model fit. The final model had an R2 of 54% (adjusted R2 of 50%).

Table 3. Univariate regression analyses of possible predictors with seasonally adjusted 25(OH)D and predictive model derived from backwards stepwise regression
VariableUnivariate analysisMultivariate model
β (95% CI), P-valueβ (95% CI), P-value% SSRa
  1. a

    Proportion of regression sum-of-squares (SSR) attributable to variable.

    Overall Wald P-value.

Physical activity (sessions/week)1.19 (0.54, 1.84), P = 0.0011.14 (0.61, 1.67), P < 0.00139%
Untanned skin color
Very fair (n = 7)ReferenceReference16%
Fair (n = 24)14.26 (−3.63, 32.15)15.57 (1.99, 29.16)
Light or dark olive (n = 15)24.79 (5.72, 43.85), P = 0.04a19.76 (5.15, 34.38), = 0.03a
Osteoporosis treatment in last month
No (n = 41)ReferenceReference30%
Yes (n = 5)35.02 (16.88, 53.16), P < 0.00128.48 (13.27, 43.70), P < 0.001
Time outdoors (weekly; h/day)3.43 (−2.08, 8.94), P = 0.22Not included in final model
Vitamin D supplementation
No (n = 36)Reference 
Yes (n = 10)17.43 (2.48, 32.39), P = 0.02Not included in final model
Age0.32 (−0.49, 1.13), P = 0.43Not included in final model
Male (n = 21)Reference 
Female (n = 25)2.50 (−10.62, 15.62), P = 0.70Not included in final model
Physical activity (h/week)1.18 (0.28, 2.08), P = 0.01Not included in final model
Body Mass Index−1.50 (−2.64, −0.36), P = 0.01Not included in final model


  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Our results indicate that the brief “time outdoors” questions are valid measures to rank study participants according to their current sun exposure. This is based on the finding of moderate, positive correlations between questionnaire time outdoors and both sun exposure diary and objectively measured UVR exposure. We found no significant association between current time outdoors (as assessed by questionnaire) and the silicone cast score in this older population; investigation of this was hampered by the lack of variation in the cast scores. Almost the entire study sample had evidence of high skin damage, which is likely to be the result of a lifetime of exposure to the Australian sun.

We observed a pattern of increasing overestimation of time outdoors as questionnaire time outdoors itself increased. This could result from the study being conducted during the cooler months of the year (so that the diary measure of current sun exposure is less than that for usual sun exposure recorded by the questionnaire measure), or alternatively, from a proportional bias in underreporting of time outdoors in the diary. Although this observation would need to be confirmed by additional studies, conducted over a greater range of seasons, it warrants caution in the interpretation of inferences related to the absolute values of time outdoors.

Few previous studies have directly addressed the validity of brief questionnaire measures of sun exposure with which we can compare our findings. Our findings are consistent with the significant correlations (generally in the weak-to-moderate range) reported elsewhere between questionnaire and diary measures [7, 8] and questionnaire and dosimeter measures of sun exposure [6]. It is difficult to specifically compare findings beyond this for reasons including a diversity in study samples (e.g. Tasmanian adolescents [6] or lifeguards and parent–child pairs [8]), variability in the nature and application of self-report instruments (e.g. time outdoors between 9 A.M. and 3 P.M. [7], or time in the sun between 10 A.M. and 4 P.M. during summer [8]) and differences in the choice of gold-standard instruments (e.g. polysulfone dosimeter [6] or digital dosimeters similar to those used in this study). Despite difficulties in making direct comparisons between studies, together they point to a reasonable validity of self-reported sun exposure. Notwithstanding this, there is a need for further research conducted in a greater range of study contexts, to more comprehensively assess the validity of questionnaire measures of habitual sun exposure. It would be of particular interest to assess the comparative validity of different measures, perhaps to define optimal instruments for different study types and research questions.

In addition to our examination of the validity of the time outdoors questions as measures of sun exposure, we also examined the validity of self-reported skin color. We found a more modest correlation (rs = 0.49), than that reported from a New Zealand study (rs = 0.75; [40]), but the sample size was larger in that study (n = 289), with a greater range and representation of skin types allowing for better assessment across skin type categories than was possible in our study. Our findings nonetheless indicate that the questionnaire measure reliably distinguishes between those with light olive and fairer constitutive skin tones. The poor distinction in the measured melanin densities between those with self-rated very fair or fair skin suggests it may be appropriate to collapse these categories in analyses where constitutive melanin density is of interest.

We examined whether data from selected items included on the 45 and Up Study questionnaire could be used to develop a predictive model of current 25(OH)D concentration for use in future studies. We were able to explain ca 54% of the variance in 25(OH)D concentration using only three variables, in contrast to the 28% explained in a model containing six variables developed by Giovannucci et al. [41] and used to infer associations between various cancers and vitamin D status. Physical activity in the last week, and specifically total sessions, was the major contributor to our model. Physical activity has been included in two other predictive models for serum 25(OH)D concentration [41, 42].

Interestingly, the questionnaire measure of time outdoors was not a significant predictor of 25(OH)D concentration. A case–control study of multiple sclerosis reported significant, although weak, correlations between self-reported sun exposure and 25(OH)D levels, however, the measures related to time in the sun (rather than outdoors), and time spent outside due to activities [43]. A number of factors are known to moderate the association between time outdoors and 25(OH)D concentration, such as ambient UVR (a function of season, time of day and cloudiness) and personal factors (e.g. clothing and genetic variation; [44]). We did not have data on the latter, although some of the variation in ambient UVR was removed by using the deseasonalized 25(OH)D concentration as the outcome. With prolonged sun exposure the synthesis of vitamin D3 plateaus [45]; adequate 25(OH)D levels can be obtained in under 10 min of daily exposure in the Australian summer [46] so that time outdoors beyond this may not be associated with higher 25(OH)D levels. In this setting, total duration of time in the sun may be less important than the frequency of sun exposure—with this possibly better reflected in the number of sessions of physical activity, than self-reported time outdoors.

Self-reported untanned skin color was a significant predictor of 25(OH)D levels in this study sample, with higher levels observed in those with darker (self-reported) skin tones, suggesting less sun-seeking and/or more sun-protective behaviors in individuals with fairer skin tones. Although not directly comparable due to the use of a skin type classification system (Fitzpatrick), a UK population-based study involving 1414 Caucasian female adults similarly found that those with darker skin types had higher 25(OH)D levels [47].

A limitation of the study is its small size, as reflected in the widths of the confidence intervals, which means that null findings should not be emphasized. A further limitation is the use of volunteers, who were informed of the specific purpose of the study and motivated to take part. It is possible, although we have no supporting evidence, that such participants may give more accurate answers than would otherwise have been obtained from a different sample. Consequently, time outdoors assessed by questionnaire may perform better as a measure of sun exposure in this sample and study context than in the general population. A Hawthorne effect, which is when subjects' behavior changes as a consequence of being under observation [48], can also not be ruled out, although the opportunity for this was reduced by the recording of data over multiple days. Diary and dosimeter data were collected over only a 1 week period for each participant, which may not represent their habitual sun exposure over longer periods. Furthermore, the correlation analyses indicate that the brief questionnaire measures are good for studies where exposure is ranked, however, research related to the absolute amount of sun exposure would need to take account of differences between questionnaire responses and actual time outdoors. Another important consideration is that the questionnaire asked about “time outdoors” rather than sun exposure, as the association between time outdoors and actual sun exposure might vary in different population groups and settings.

Finally, the study participants were drawn from the Canberra population 45 years and over, and so our findings might not be generalizable to other age-groups or in other locations, or a wider (or different) range of skin types. However, the findings do not appear to be plausibly attributable to characteristics of study participants, and while of central concern in descriptive studies, representativeness is not a requirement for external validity and generalizability in an analytical study such as this one where internal comparisons are made [49, 50].

Strengths of the study were the use of multiple methods, including self-completed but very detailed sun exposure diaries and objective measures including digital UVR dosimeters, 25(OH)D concentration and silicone casts of the hand, to assess the time outdoors questions as measures of sun exposure over different time courses—current, recent and cumulative, respectively. We additionally asked participants to consider what period of time they were thinking of when answering the “time outdoors” question—it was of interest to note that the majority of participants were considering recent or current sun exposure.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

This study supports the validity of brief questionnaire-based measures of “time outdoors” and “untanned skin color” as measures of habitual time outdoors and UVR exposure, and melanin density of the inner upper arm respectively. The measures are likely to be useful in epidemiological studies for improving our understanding of the relationship between sun exposure and health.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

The authors thank and acknowledge the volunteers who participated in this study, the Australian Radiation Protection and Nuclear Safety Agency (ARPANSA) for supplying the dosimeters, and the RDDT laboratory for their assistance with the 25(OH)D assays. Emily Banks and Robyn Lucas are supported by the Australian National Health and Medical Research Council.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  • 1
    Lucas, R., T. McMichael, W. Smith and B. Armstrong (2006) Solar ultraviolet radiation. Global burden of disease from solar ultraviolet radiation. In Environmental Burden of Disease (Edited by A. Prüss-Üstün, H. Zeeb, C. Mathers and M. Repacholi), pp. 1258. World Health Organization, Geneva.
  • 2
    Australian Institute of Health and Welfare (AIHW) and Australian Association of Cancer Registries (2005) Cancer in Australia: An Overview, 2008. Cancer Series no. 46. AIHW, Canberra.
  • 3
    Australian Institute of Health and Welfare (AIHW) (2005) Health System Expenditures on Cancer and Other Neoplasms in Australia. Series Number 22. AIHW, Canberra.
  • 4
    Holvik, K., H. E. Meyer, E. Haug and L. Brunvand (2004) Prevalence and predictors of vitamin D deficiency in five immigrant groups living in Oslo, Norway: The Oslo Immigrant Health Study. Eur. J. Clin. Nutr. 59, 5763.
  • 5
    Hintzpeter, B., C. Scheidt-Nave, H. K. Müller, L. Schenk and G. B. M. Mensink (2008) Higher prevalence of vitamin D deficiency is associated with immigrant background among children and adolescents in Germany. J. Nutr. 138, 14821490.
  • 6
    Dwyer, T., L. Blizzard, P. Gies, R. Ashbolt and C. Roy (1996) Assessment of habitual sun exposure in adolescents via questionnaire-a comparison with objective measurement using polysulphone badges. Melanoma Res. 6, 231239.
  • 7
    Chodick, G., M. D. Freedman, R. K. Kwok, T. R. Fears, M. S. Linet, B. H. Alexander and R. A. Kleinerman (2007) Agreement between contemporaneously recorded and subsequently recalled time spent outdoors: Implications for environmental exposure studies. Ann. Epidemiol. 17, 106111.
  • 8
    Glanz, K., P. Gies, D. L. O'Riordan, T. Elliott, E. Nehl, F. McCarty and E. Davis (2010) Validity of self-reported solar UVR exposure compared with objectively measured UVR exposure. Cancer Epidemiol. Biomarkers Prev. 19, 30053012.
  • 9
    O' Riordan, D. L., W. R. Stanton, M. Eyeson Annan, P. Gies and C. Roy (2000) Correlations between reported and measured ultraviolet radiation exposure of mothers and young children. Photochem. Photobiol. 71, 6064.
  • 10
    Thieden, E., M. S. Ågren and H. C. Wulf (2001) Solar UVR exposures of indoor workers in a Working and a Holiday Period assessed by personal dosimeters and sun exposure diaries. Photodermatol. Photoimmunol. Photomed. 17, 249255.
  • 11
    Sullivan, S. S., J. L. Cobb, C. J. Rosen, M. F. Holick, T. C. Chen, M. G. Kimlin and A. V. Parisi (2003) Assessment of sun exposure in adolescent girls using activity diaries. Nutr. Res. 23, 631644.
  • 12
    Chodick, G., R. A. Kleinerman, M. S. Linet, T. Fears, R. K. Kwok, M. G. Kimlin, B. H. Alexander and D. M. Freedman (2008) Agreement between diary records of time spent outdoors and personal ultraviolet radiation dose measurements. Photochem. Photobiol. 84, 713718.
  • 13
    Herlihy, E., P. H. Gies, C. R. Roy and M. Jones (1994) Personal dosimetry of solar UV radiation for different outdoor activities. Photochem. Photobiol. 60, 288294.
  • 14
    English, D. R., B. K. Armstrong and A. Kricker (1998) Reproducibility of reported measurements of sun exposure in a case–control study. Cancer Epidemiol. Biomarkers Prev. 7, 857863.
  • 15
    Vajdic, C. M., A. Kricker, M. Giblin, J. McKenzie, J. Aitken, G. G. Giles and B. K. Armstrong (2002) Sun exposure predicts risk of ocular melanoma in Australia. Int. J. Cancer 101, 175182.
  • 16
    Karagas, M. R., M. S. Zens, H. H. Nelson, K. Mabuchi, A. E. Perry, T. A. Stukel, L. A. Mott, A. S. Andrew, K. M. Applebaum and M. Linet (2007) Measures of cumulative exposure from a standardized sun exposure history questionnaire: A comparison with histologic assessment of solar skin damage. Am. J. Epidemiol. 165, 719726.
  • 17
    English, D. R., B. K. Armstrong, A. Kricker, M. G. Winter, P. J. Heenan and P. L. Randell (1998) Case control study of sun exposure and squamous cell carcinoma of the skin. Int. J. Cancer 77, 347353.
  • 18
    Hughes, A. M., B. K. Armstrong, C. M. Vajdic, J. Turner, A. E. Grulich, L. Fritschi, S. Milliken, J. Kaldor, G. Benke and A. Kricker (2004) Sun exposure may protect against non-Hodgkin lymphoma: A case–control study. Int. J. Cancer 112, 865871.
  • 19
    45 and Up Study Collaborators (2008) Cohort profile: The 45 and up study. Int. J. Epidemiol. 37, 941947.
  • 20
    Australian Bureau of Meteorology (2012) Canberra in 2011: A Normal Year for the ACT. Available at: Accessed on 3 May 2012.
  • 21
    Australian Bureau of Meteorology (2012) Climate Data Online. Available at: Accessed on 5 May 2012.
  • 22
    Dwyer, T., L. Blizzard, R. Ashbolt, J. Plumb, M. Berwick and J. M. Stankovich (2002) Cutaneous melanin density of caucasians measured by spectrophotometry and risk of malignant melanoma, basal cell carcinoma, and squamous cell carcinoma of the skin. Am. J. Epidemiol. 155, 614621.
  • 23
    Lucas, R. M., A. L. Ponsonby, K. Dear, B. V. Taylor, T. Dwyer, A. J. McMichael, P. Valery, I. Van Der Mei, D. Williams and M. P. Pender (2009) Associations between silicone skin cast score, cumulative sun exposure, and other factors in the ausimmune study: A multicenter Australian study. Cancer Epidemiol. Biomarkers Prev. 18, 28872894.
  • 24
    Holman, C. D. J., B. K. Armstrong, P. R. Evans, G. J. Lumsden, K. J. Dallimore, C. J. Meehan, J. Beagley and I. M. Gibson (1984) Relationship of solar keratosis and history of skin cancer to objective measures of actinic skin damage. Br. J. Dermatol. 110, 129138.
  • 25
    Allen, M. and R. McKenzie (2005) Enhanced UV exposure on a ski-field compared with exposures at sea level. Photochem. Photobiol. Sci. 4, 429437.
  • 26
    Thieden, E., M. Agren and H. Wulf (2000) The wrist is a reliable body site for personal dosimetry of ultraviolet radiation. Photodermatol. Photoimmunol. Photomed. 16, 5761.
  • 27
    Swinscow, T. (1996) Statistics at Square One. BMJ Publishing Group, London.
  • 28
    Shrout, P. and J. Fleiss (1979) Intraclass correlations: Uses in assessing rater reliability. Pyschological Bulletin 86, 420428.
  • 29
    Shrout, P. and J. Fleiss (1983) The Design and Analysis of Clinical Experiments. John Wiley & Sons, New York.
  • 30
    Martin Bland, J. and D. G. Altman (1986) Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 327, 307310.
  • 31
    Schmidt, M. and K. Steindorf (2006) Statistical methods for the validation of Questionnaires—Discrepancy between theory and practice. Methods Inf. Med. 45, 409413.
  • 32
    Australian Institute of Health and Welfare (AIHW) (2003) The Active Australia Survey: A guide and Manual for Implementation, Analysis and Reporting. Catalogue no. CVD 22. AIHW, Canberra.
  • 33
    van der Mei, I. A. F., A. L. Ponsonby, T. Dwyer, L. Blizzard, B. V. Taylor, T. Kilpatrick, H. Butzkueven and A. J. McMichael (2007) Vitamin D levels in people with multiple sclerosis and community controls in Tasmania, Australia. J. Neurol. 254, 581590.
  • 34
    Munger, K. L., L. I. Levin, B. W. Hollis, N. S. Howard and A. Ascherio (2006) Serum 25-hydroxyvitamin D levels and risk of multiple sclerosis. JAMA 296, 28322838.
  • 35
    Lucas, R., A. L. Ponsonby, K. Dear, P. Valery, M. Pender, B. Taylor, T. Kilpatrick, T. Dwyer, A. Coulthard and C. Chapman (2011) Sun exposure and vitamin D are independent risk factors for CNS demyelination. Neurology 76, 540548.
  • 36
    Simpson Jr, S., B. V. Taylor, L. Blizzard, A. L. Ponsonby, F. Pittas, H. Tremlett, T. Dwyer, P. Gies and I. van Der Mei (2010) Higher 25-hydroxyvitamin D is associated with lower relapse risk in multiple sclerosis. Ann. Neurol. 68, 193203.
  • 37
    Hosmer, D. and S. Lemeshow (2000) Applied Logistic Regression. John Wiley and Sons, Hoboken, NJ.
  • 38
    Reichenheim, M. E. (2004) Confidence intervals for the kappa statistic. Stat. J. 4, 421428.
  • 39
    Altman, D. G. (1991) Practical Statistics for Medical Research. Chapman and Hall, London.
  • 40
    Reeder, A. I., V. A. Hammond and A. R. Gray (2010) Questionnaire items to assess skin color and erythemal sensitivity: Reliability, validity, and “the Dark Shift”. Cancer Epidemiol. Biomarkers Prev. 19, 11671173.
  • 41
    Giovannucci, E., Y. Liu, E. B. Rimm, B. W. Hollis, C. S. Fuchs, M. J. Stampfer and W. C. Willett (2006) Prospective study of predictors of vitamin D status and cancer incidence and mortality in men. J. Natl Cancer Inst. 98, 451459.
  • 42
    Millen, A. E., J. Wactawski-Wende, M. Pettinger, M. L. Melamed, F. A. Tylavsky, S. Liu, J. Robbins, A. Z. LaCroix, M. S. LeBoff and R. D. Jackson (2010) Predictors of serum 25-hydroxyvitamin D concentrations among postmenopausal women: The Women's Health Initiative Calcium plus Vitamin D Clinical Trial. Am. J. Clin. Nutr. 91, 13241325.
  • 43
    Van der Mei, I. A. F., L. Blizzard, A. L. Ponsonby and T. Dwyer (2006) Validity and reliability of adult recall of past sun exposure in a case–control study of multiple sclerosis. Cancer Epidemiol. Biomarkers Prev. 15, 15381544.
  • 44
    Wang, T. J., F. Zhang, J. B. Richards, B. Kestenbaum, J. B. van Meurs, D. Berry, D. P. Kiel, E. A. Streeten, C. Ohlsson, D. L. Koller, L. Peltonen, J. D. Cooper, P. F. O'Reilly, D. K. Houston, N. L. Glazer, L. Vandenput, M. Peacock, J. Shi, F. Rivadeneira, M. I. McCarthy, P. Anneli, I. H. de Boer, M. Mangino, B. Kato, D. J. Smyth, S. L. Booth, P. F. Jacques, G. L. Burke, M. Goodarzi, C.-L. Cheung, M. Wolf, K. Rice, D. Goltzman, N. Hidiroglou, M. Ladouceur, N. J. Wareham, L. J. Hocking, D. Hart, N. K. Arden, C. Cooper, S. Malik, W. D. Fraser, A.-L. Hartikainen, G. Zhai, H. M. Macdonald, N. G. Forouhi, R. J. F. Loos, D. M. Reid, A. Hakim, E. Dennison, Y. Liu, C. Power, H. E. Stevens, L. Jaana, R. S. Vasan, N. Soranzo, J. Bojunga, B. M. Psaty, M. Lorentzon, T. Foroud, T. B. Harris, A. Hofman, J.-O. Jansson, J. A. Cauley, A. G. Uitterlinden, Q. Gibson, M.-R. Järvelin, D. Karasik, D. S. Siscovick, M. J. Econs, S. B. Kritchevsky, J. C. Florez, J. A. Todd, J. Dupuis, E. Hyppönen and T. D. Spector (2010) Common genetic determinants of vitamin D insufficiency: A genome-wide association study. Lancet 376, 180188.
  • 45
    Olds, W. J., A. R. McKinley, M. R. Moore and M. G. Kimlin (2008) In vitro model of vitamin D3 (Cholecalciferol) synthesis by UV radiation: Dose–response relationships. J. Photochem. Photobiol., B 93, 8893.
  • 46
    Mason, R. (2010) Vitamin D: A hormone for all seasons. Climacteric 14, 197203.
  • 47
    Glass, D., M. Lens, R. Swaminathan, T. D. Spector and V. Bataille (2009) Pigmentation and vitamin D metabolism in Caucasians: Low vitamin D serum levels in fair skin types in the UK. PLoS ONE 4, e6477.
  • 48
    Parsons, H. M. (1974) What happened at Hawthorne?. Science 183, 922932.
  • 49
    Ponsonby, A. L., T. Dwyer and D. Couper (1996) Is this finding relevant? Generalisation and epidemiology. Aust. N. Z. J. Public Health 20, 5456.
  • 50
    Mealing, N. M., E. Banks, L. R. Jorm, S. D. Steel, M. S. Clements and K. D. Rogers (2010) Investigation of relative risk estimates from studies of the same population with contrasting response rates and designs. BMC Med. Res. Methodol. 10, 26.