Mask-related skin changes among healthcare workers in a community-based hospital

Data collection

After IRB approval from the Beaumont Health IRB (IRB#2021-011), the study team observed an “opt in” recruitment method in which only those eligible Beaumont Farmington Hills healthcare workers who actively clicked on the SurveyMonkey [8] link provided to them through the health system’s email system were allowed to enter any study survey responses. The SurveyMonkey [8] software was programmed with both secure socket layers (SSL) and internet protocol security (IPsec) encryption to systematically conceal the identities of all respondents to members of the study team. Each respondent was advised in initial survey instructions that they are asked to avoid providing any identifying information in all survey fields. The study date range was from April 21, 2021 to July 1, 2021.

A cross-sectional survey study was performed. The primary 10-item survey instrument that was administered is the Dermatology Life Quality Index (DLQI) [9], a well-validated Likert-scale survey tool utilized by the team analyst (WC) to generate a composite score of between 0 and 30 points [10]. The composite DLQI score can be utilized to categorize patients into one of five overall categories, ranging from “no effect at all on my life” (0–1 point) through “extremely large effect on my life” (21–30 points) [11]. Respondents were asked a series of 10 additional questions concerning the degree to which abnormal mask-related skin conditions affect their skin symptoms, possible embarrassment/self-consciousness, and perceived impact of mask-related skin changes (Figure 1).

Figure 1:

Survey questions. Participants were asked 10 questions regarding mask usage and their associated skin changes.

The following eligibility criteria were observed: adults 21 years of age and older; Beaumont Health employees; and regular mask usage during work hours. Children less than 18 years of age and cognitively impaired individuals were excluded.

Minimal sample size calculations

Before data collection was begun for this IRB-approved project, the team analyst utilized G*Power 3.1.94 software [12] to generate a priori minimal sample size calculations for the hypothesized main effect influence of the predominant type of protective mask utilized (i.e., P100/N95/Kn95 vs. surgical/cloth mask) on variations in continuous DLQI composite scores. These calculations indicated that a minimal total sample size of at least 70 discrete respondents (i.e., at least 35 P100/N95/KN95 mask users vs. 35 surgical/cloth mask users) would afford the study team a 0.8005384 1 minus β level of statistical power to detect statistically significant variations between sample subgroups observing a two-tailed coefficient alpha of 0.05 to indicate statistical significance. This sample size was also based on a critical Z estimate for the hypothesized two-tailed influence of mask use subgroup on “moderate effect or higher” composite DLQI scores, as reported by Lan et al. [13] in 2020 (Critical Z=1.9599640).

Data analyses

After IRB approval, author WC conducted all analytic procedures. After cleaning the de-identified SurveyMonkey [8] data from string to numerical form, he created a “working” data set. For analytic procedures, the analyst examined the validity of the study teams’ hypothesized relationships utilizing SPSS Version 27 analytic software [14].

First, a series of descriptive statistics, cross-tabulation charts, and graphical examinations of data were utilized to evaluate sample subgroup and outcome distributional patterns. For most analyses, continuous study measures (e.g., continuous DLQI scores, etc.) were conservatively categorized into sample subgroups as prescribed by the original DLQI designers. Next, a series of Pearson r bivariate correlation coefficients between possible collinear predictive measures on the primary study outcome (i.e., mask-related DLQI score variations of 6 or greater) were utilized [15]. A series of simple inferential chi-squared tests of independence were also conducted [15].

Finally, the predictive significance of each selected measure on the selected dichotomous and categorical outcomes of interest were modeled out during a series of nonparametric (i.e., not based on any normal distribution assumptions) two-tailed forward “stepwise” (i.e., one model term entered at a time, retained if initial p values are less than 0.10) “main effects” multinomial regression procedures [16]. For all analytic procedures, the analyst observed a two-tailed coefficient alpha p value of less than 0.05 to indicate statistical significance.

Mask-related skin changes

A total of 430 Beaumont Health employees completed the questionnaire and DLQI, out of which 62, or 12.6%, were excluded due to incomplete responses. A total of 370 out of 430 (86.0%) noticed skin changes since the work-hours face mask requirement was instituted at Beaumont, while 378 out of 430 (87.9%) felt that their skin was better when not wearing a mask. Among the participants, 167 (38.8%) reported worsening of a pre-existing skin condition in the area covered by the mask, 163 (37.9%) reported starting a new medication (over-the-counter or prescription), while 98 (22.8%) needed a prescription to manage a new or worsening mask-associated skin disorder. Further, 201 (46.7%) researched potential treatment options for mask-related skin changes. The majority, 267 (62.1%), of the respondents wore masks for greater than 31 h per week, with N95 (32.1%), cloth (30.0%), and surgical masks (23.5%) comprising the most reported mask types. The use of two or more total mask types was reported in 249 (57.9%). Many of the respondents, 272 (63.3%), wore the same mask for 1 day or less, whereas 83 (19.3%) wore the same mask for 2–3 days, and 63 (14.7%) wore the same mask for greater than 4 or more days. Responses to the questionnaire regarding mask-related skin changes are summarized in Table 1.

Dermatology life quality index (DLQI)

The majority of respondents, 283 (65.8%), reported having at least a little symptomatic skin (i.e., itchy, painful, sore, stinging) during the prior week. Furthermore, 311 (72.3%) reported that they were at least a little embarrassed or self-conscious of their skin. Problems caused by the treatment of their skin (i.e., taking up time) were reported in 178 (41.4%) of participants. Sample respondents’ mean DLQI composite scores were 4.5 (SD 4.3), ranging from 0 to 24 on a possible scale from 0 to 30. A DLQI score of 4.5 is interpreted as a small effect on a patient’s life. Due to the extremely nonparametric spread of composite scores with a lower mean DLQI score, our author group observed the DLQI designer’s five categories. Responses to the DLQI are summarized in Table 2.

Chi-squared analysis

Chi-squared (X²) analysis of composite DLQI tertile score (Table 3) by the number of different mask types utilized (Pearson X²=19.0, df=8, p=0.015), and some degree of symptomatic skin (Pearson X²=156.4, df 4, p<0.001), were found to be statistically significant. The number of mask types utilized and experiencing some degree of symptomatic skin were associated with a higher impact on the respondent’s quality of life.

Multinomial regression predictive modeling

A multivariate controlled two-tailed multinomial regression predictive modeling procedure was completed to examine for the significance of each respondent mask use characteristic on this categorical DLQI composite score outcome observing the DLQI designer parameters when controlling for respondent characteristics. The number of days the same mask was worn (Wald=22.3, df=12, p=0.035), and some degree of symptomatic skin (Wald=217.1, df=8, p<0.001), reached statistical significance on composite DLQI scores (Table 3). Thus, wearing a mask for a greater number of days and reporting some degree of symptomatic skin was associated with a higher impact on the respondent’s quality of life.

Limitations

This study is not without limitations. The data was only collected from one healthcare organization. The cross-sectional design of the study is also a limitation. The lack of demographic data introduces potential confounding factors, including age, gender, and healthcare roles. We recognize that the question “How many hours a day do you wear your mask during the work week?” with options “Less than 20, 20–30, 31–40, and over 40” is poorly worded because there are only 24 h in a day. Our results likely underestimate the effect that mask wearing has on skin conditions among healthcare workers because the survey was distributed to both frontline healthcare workers and those who work in a corporate office setting.

The lack of significant associations between some of the selected study measures on selected study outcomes could be partially attributed to (A) the lack of a sufficient-sized sample to detect meaningful sample subgroup differences; (B) an inadequately heterogeneous sample (e.g., more variation needed among selected measures or outcomes); (C) missing data patterns; and/or (D) perhaps the lack of any possible association to detect in the first place. The study team acknowledges that their results may have been skewed by “preferred response” or self-selection respondent biases. We also recognize that dermatologic conditions and symptoms were self-diagnosed and self-perceived.