Compliance and data reliability in sun exposure studies with diaries and personal, electronic UV dosimeters

Study population

This study was performed in Denmark (latitude, 56° north; and longitude, 12° east) over a period of 3 summer-half-years, from 1999 to 2001. Where a summer-half-year corresponds to median 137 participation days [Inter quartile range (IQR) 129–148 days] from April to October. Subject selection has previously been described in detail (1–6) but to give a brief summary, 340 participants (153 males and 187 females) were chosen to cover a wide age-span, including 97 children (4–15 years), 30 adolescents (16–19 years), 89 indoor workers (21–64 years). In addition, adult groups with expected high UVR exposure levels were included, 53 sun worshippers, 24 golfers and 47 municipal gardeners and rangers. As some subjects participated in more than 1 summer-half-year, the study sample comprised data from 407 summer-half-years, in the following referred to as 407 subjects.

Personal electronic UV dosimeter

The UV dosimeter, SunSaver, mounted in a housing the size and form of a wristwatch, was developed and assembled in our department. It includes a sensor, a battery-driven data logger and a digital watch. The measured UV data can be transferred to a personal computer. The subjects were instructed to wear the dosimeter continuously on the wrist not to cover it with a sleeve or immerse it in water. The UV dosimeter has been described in detail previously (1–6, 10). The UV dosimeters data were tapped in the middle and at the end of the study, and additionally, if any dosimeter problems occurred.

Sun exposure diary

The participants were provided with a diary, and instructed to answer ‘Yes’ or ‘No’ to the following questions about sun exposure behaviour: (1) Did you wear the SunSaver today? (2) Are you off work/school or on holiday today? (3) Are you abroad today? (If yes write country code). (4) Did you sunbathe today? (Sitting or lying in the sun (or sun-bed) with upper body or shoulders exposed to get a tan). (5) Have you exposed your shoulders or upper body outdoors today? (6) Have you been at the beach or at the sea today? (7) Have you applied a sunscreen today? (If yes, write the factor number). (8) Did you get sunburned today? (If yes: Red? Red and sore? Red, sore and blistered?) On a minimal, moderate or extensive area? We considered a ‘Yes’ response to question 4 or 5, to indicate, ‘risk behaviour’, because a large area of the body would be sun exposed. The answers to question (1) Did you wear the SunSaver today? were used to assess completion errors in this study.

Diary completing errors validated by corresponding SunSaver measurements

To validate if the subjects had completed the diary correctly, we compared the first question in the diary: ‘Did you wear the SunSaver today?’‘Yes’ of ‘No’ with the objective measurements from the SunSaver being either 0 standard erythema dose (SED) or above 0 SED. Only days where the diary was completed and the SunSaver functioning have been considered, see Table 1. We consider days in square 1 as correctly completed, because the SunSaver was reported worn and the SunSaver reading was above 0 SED on the day. We consider days in square 3 as incorrectly completed, as the SunSaver reading was above 0 SED, although the SunSaver was reported as not worn on the day in question. We are not able to control if the days in squares 2 and 4 are correctly completed. However, if we assume that the error rates are the same as in square 3, the total estimated error rate could be found by multiplying the error rate in square 3 by three.

Statistical analysis

We used non-parametric statistics because most of the data were not normally distributed. The results are therefore given as median (IQR). Mann-Whitney U-test was used to compare unpaired continuous data between groups and Spearman's rank correlation was used to investigate interactions between two continuous measurements. Wilcoxon' paired rank-sum test on paired continuous data. In each case a P-value of less than 0.05 was considered significant. We used SPSS for Windows, version 11.5 (SPSS Inc., Chicago) for data analysis.

Diary completion and subject compliance rates

The subjects participated in 54 943 days. The diary completion rate was 95% as we collected 52 333 completed diary days. The subjects had worn the UV dosimeter on 47 480 (86.4%) of the completed diary days, which we define as the subject compliance rate. Only 10 (2.5%) of the subjects (six children, one adolescent, one sun worshipper, one golfer and one gardener) had a subject compliance rate of 0. Table 2 shows the individual median rates within the subgroups, and among males and females. The children, adolescents and golfers had significantly lower subject compliance rates than the rest of the groups (P<0.05). There were no significant difference in subject compliance rate between males and females.

UV dosimeter record rates

UV dosimeter readings were collected on 46 944 days, 85.4% of the participation days. Eleven (3%) of the subjects (eight children, two adolescents and one gardener) did not have any dosimeter records. During the study, 478 SunSavers were utilized whereof 16 were lost by 11 children, three adolescents, one gardener and one sun worshipper. As most of the subjects who lost their SunSaver were also careless with the diary completion we only missed 493 days (0.9%) because of SunSaver loss. There was a tendency to wear the UV dosimeters less during weekends than on workdays and more in the beginning than at the end of each season. In October the first year the dosimeter compliance fell especially among the young children. This was mainly because of difficulties in keeping the SunSaver, uncovered by long sleeves. In addition, the subjects received very little UVR in October. We therefore decided not to collect UV data during October month the following years. The individual median UV dosimeter record rates within the different subgroups, and among males and females are shown in Table 3. The children had significantly lower UV dosimeter record rates than the indoor workers, golfers and gardeners (P<0.046). Males had significantly lower than females (P=0.02).

Sun exposure record rates

The sun exposure record rate was defined as days with corresponding UV dosimeter and diary data, where the subjects had also answered yes to wearing the UV dosimeter. We collected sun exposure records on 75% of the participation days (41 182 days) from 390 subjects (96%). The 17 subjects with zero sun exposure records comprised 10 children, three adolescents, one sun worshipper, two golfers and one gardener. Table 3 shows the distribution of the individual median sun exposure record rates. Children, adolescents and golfers had lower sun exposure record rates than the rest of the groups (P<0.05). While no significant difference was found between males and females (P=0.15).

Errors in data transfer from diary to computer

The transferring of data from the paper diary forms to the computer database caused errors, which were revealed when we controlled the data entry afterwards. In 1999 the diary data were keyed-in manually by a secretary, which led to 3.5% days with data transfer errors. The data transfer errors fell to 1.1% in 2000 and 0.8% in 2001, when the diary data forms were scanned directly into the database. Although scanning the data did not eliminate data transfer errors, the manpower used was considerably reduced. All revealed data transfer errors were corrected manually in the database.

Correctable diary completing mistakes

When scrutinizing the diaries, we found completing mistakes in 3.1% of the participating days, which we were able to correct. It could be a diary month where workdays and days off were interchanged consequently or school holidays that children had marked as school days. Children and adolescents made completing mistakes on significantly more days (4.4%) than adults (2.5%) (P<0.001), and females on significantly more days (3.5%) than males (2.5%) (P=0.012).

Completing errors in diaries validated by corresponding UV dosimeter readings

Table 4 shows the days with corresponding diary and dosimeter data. The data were grouped in four squares combining ‘yes’ or ‘no’ to wearing the SunSaver with the actual SunSaver readings being either 0 SED or above 0 SED as described in Table 1. As 1.1% incorrectly completed days were found in square 3, we presumed a total estimated error rate in completing diary of 3.3%. The error rate was higher for children (6.3%) and adolescents (3.6%) than for the adult groups (2.4%). A day in square 3 was changed from ‘no’ to ‘yes’ to wearing SunSaver and thereby transferred to square 1. In 167 out of the 481 days had the day before or after a dosimeter reading of 0 SED and a ‘yes’ to wearing SunSaver in the diary, in these cases were the 2 days just exchanged. The number of days with positive UV readings, were calculated by summing the days in squares 1 and 3. Positive UV readings were obtained on 67.4% of the days with both SunSaver and diary data. The children had the lowest (57.1%) and the gardeners the highest (81.6%) percentages of the days with positive UV readings. As only days where the SunSaver had been worn (squares 1 and 2) were analysed in our UV exposure studies (1–6), we had probably reduced the estimated error rate from 3.3% to 1.1% by excluding the days in squares 3 and 4.

Population

In our population sample, the proportion of subjects younger than 20 years equalled that in the Danish population overall, while adults engaging in outdoor work, outdoor sport and sun worship were overrepresented. Thus, we selected an adult population sample that might have a higher UVR exposure than the average Danish population. Seventeen (4%) of the subjects (11 males) had no days with corresponding UV and dairy readings and were considered non-responders. Only two children and one golfer became non-responders solely because of technical UV dosimeter problems. The remaining 14 non-responders comprised five 4–5-year-old children whose parents appeared not to be cooperative; six children and adolescents who were too careless, and either lost their SunSaver or did not complete the diary. One sun worshipper, one golfer and one gardener did not complete any diary. As no exposure data were available from the non-responders the skin pigmentation expressed as pigment protection factor (PPF) were used as an estimate of their sun exposure. The PPF is equivalent to the number of SED to induce just perceptible erythema by a minimal erythema dose phototest on the buttocks (11). PPF were measured by a skin reflectance meter (UV-Optimize, Model Matic 555; Matic, Naerum, Denmark) (12) on all subjects at the start of the study, and again after the summer holidays for all but three of the non-responders. No significant differences in PPF measured on non-exposed buttock skin or sun-exposed skin locations were found between the responders and non-responders. This indicate that the non-responders did not have a different sun exposure behaviour than the responders (1–6, 13).

General efforts to increase compliance

As the participation period was at least a summer season, a great deal of effort was put into making the subjects wear the UV dosimeter and complete the diary. To obtain the highest possible compliance, the study was performed on the premise where the participants were doing the study for us and not vice versa. E-mail contacts were regularly established with the schoolteachers and as many of the participants as possible. Our private phone and cell phone numbers were printed on the diaries. In addition, the subjects were told not to hesitate to contact us, if they had any problems also during the evenings, weekends or holidays. Thorough study information was given at the start and written in the diaries, too. The subjects except the sun worshippers were all part of a group and team spirit was encouraged within the groups. All subjects were promised and given detailed information about their individual sun exposure records. The adolescent group had UVR included in their biology curriculum and a workshop were arranged for them at our laboratory after study completion. The golfers were offered a full skin examination by our professor.

Efforts to increase UV dosimeter record rate

We developed the SunSaver to work as a UV dosimeter as well as a wristwatch; because the wrist position would be the easiest to accept during a long-term study. In addition, we have previously found that UV doses received to the wrist correlated significantly with UV doses received to the head (14). UV Dosimeter maintenance was an important issue to prevent loss of UV data and we aimed at solving UV dosimeter malfunction within 24 h. This means that when UV dosimeter problems occurred, we asked the subject to come to our ward the same or the next day to have the dosimeter repaired. If this was not possible, we picked up the UV dosimeter at their home/workplace, repaired it and mailed or delivered it back the same day. If they were on holiday we mailed them another UV dosimeter and asked them to return the broken UV dosimeter, when they received the new one. In 1999, 36% of the subjects' required additional UV dosimeter maintenance outside the scheduled tapping days. In addition, before the second season the electric circuit in the UV dosimeter was water protected with a silicon layer, and a condenser was included in the dosimeter for better control of the data logger. Thereby, the maintenance percentage fell to 19% in 2000 and to 3% in 2001; where all subjects had also participated in a preceding year and were familiar with the UV dosimeter. Furthermore, we went with a laptop, a printer and maintenance equipment to all the teams. The dosimeters were tapped on the spot, the batteries exchanged, and each participant given a print of their individual UV dose curve. This was performed to emphasize the importance of wearing the UV dosimeter every day, and to show each subject that important findings came out of measuring their individual sun behaviour.

Efforts to increase Diary completion rate

The diary was a shortened version of a more comprehensive diary used in a pilot study (15). We chose a higher compliance to a few questions rather than a lower compliance to a more comprehensive diary (1–6). Questions about the clothing worn, amount of sunscreen applied and skin area covered were therefore deliberately left out. Crossing yes or no or writing a number or a country code could answer all questions. The diary covering the first month was handed over with a thorough instruction. Thereafter letters were mailed at the end of each month with a report on the study progress, a new diary for the coming month, and a stamped and addressed envelope to return the completed diary at the turn of the month. A second letter was mailed after 2 weeks to about 10% of the subjects who had not returned the diary, and a third after 1 month to 4% of the subjects who still had not returned their diary. Eventually, the remaining 1% of the subjects were contacted by phone.

Compliance rate within-subject control and total sun exposure record rate

The compliance rate within the subjects' control was 85.5%, as the subjects in total did not complete their diary or wore SunSaver in 13.6% of the days and missed 0.9% of the days because of lost SunSavers. We were able to collect complete sun exposure records in 74.8% of the participating days. However, the additional 10.7% days lost because of UV dosimeter malfunction, not prevented by our maintenance service, were out of subject control. If the UV dosimeter were further improved, it would thus be possible to collect readings on a maximum of 10% more days. As the subjects had already completed diaries in 95% of the days it would be difficult to increase the diary completion rate in a long-term study comprising many subjects including young children. However, completion of diaries using the Internet or cell phones could perhaps increase the diary completion rate a little and result in even more reliable data. Although about 10% of the days were lost because the subjects did not wear the UV dosimeter, it should be taken into consideration that the loss primarily came from a few persons. They were mainly children and teenagers who did not wear the UV dosimeter at all or during few days only, while the subjects in median had worn the UV dosimeter on 93.5% of the days. An increase in days wearing UV dosimeter would therefore be difficult in studies comprising children and teenagers.

Compliance comparison with other studies

Compliance is only briefly discussed in most UV dosimeter studies. Nearly all studies are conducted with UV badges only showing accumulated UV doses. The UV data are mostly collected on a daily basis in a few days of a season and supplemented with information about activities in a questionnaire. The total exposure in a season or a year is then extrapolated from these few days measurements (8, 16–19). Therefore, compliance in those studies are difficult to compare with our results. One study has been conducted in Australia over an 18-month period on 12 post deliverymen and five physical education teachers. It seems that approximately 20% of the possible UV badges have been collected, but no information was given on the number of corresponding questionnaires (20). In a British study 180 children and adolescents wore UV badges over a 3-month period and 58% of UV dosimeters and exposure records were selected for analysis. However, no information was given on the percentage of successful record per subject (21). In our study the individual child and adolescent had complete UV exposure records for median 63% of the days and the individual adult for 89% of the days.

Does high or low sun exposure influence on compliance and data reliability?

We investigated if there was a correlation between having a high or low sun exposure expressed as measured mean UV dose per day and subject compliance and we found a significant but weak correlation (r=0.14; P=0.005). This indicates that sun exposure dose-size is no strong indicator of subject compliance and that being a good or bad complier is independent of sun exposure dose-size. However, if a subject is a good complier the UV exposure dose seems reliable. We found no significant correlation between the measured mean UV dose per day and the total UV exposure record rate or number of completing errors and mistakes among the individuals.

The UV doses were collected electronically along with the time every 10 min. We were, therefore, able to discriminate at what time of the day the UV doses were received. By comparing the UV dose and diary information day by day, we were able to disclose the number of completing errors. We found 3% mistakes, which we could correct, and an estimated error rate of 3.3%, which we were able to reduce to about one third. Because of our serious efforts to keep a high compliance, we have the impression that the subjects made few mistakes. In other dosimeter studies with dosimeters providing accumulated doses, it is not possible or difficult to reveal these errors and it is seldom discussed. However, it would be important to try to estimate the failure rate and thereby the reliability of the data.