1. The information and opinions presented in this poster are those of the authors and in no way represent the policy of NIOSH or CDC. Mention of product names is for completeness and does not indicate endorsement by NIOSH or the authors. Abstract Comfort is a major barrier for consistent and effective hearing protector wear. Workers will reject the uncomfortable hearing protector. In 2011 we reported on a one-year longitudinal study (2004-5) looking at worker acceptance of a semi-custom hearing protector. Workers were tested four times over the course of the study for both comfort (using the Park & Casali survey) and personal attenuation rating (PAR) using the Michaels and Associates FitCheck system. Recently we were interested in the relationship between comfort measurements and PARs. Our hypothesis was there was an inverse relationship between PAR and comfort (increasing comfort => decreasing PAR). We found there was an inverse relationship for two of the variables (“loose- tight” and “hard-soft”) and a positive relationship (increasing comfort => increasing PAR) of one of the others (“rough-smooth”). Other scales (e.g. “comfortable-uncomfortable”) did not show significant results. The results of the third model were: CAPT (ret.) Rickie R. Davis 1, USPHS, Peter B. Shaw 2 1 Hearing Loss Prevention Team, Engineering and Physical Hazards Branch, 2 Statistics Team, Division of Applied Research and Technology, National Institute for Occupational Safety and Health, CDC, Cincinnati, OH 45226 Introduction. Hearing protectors which do not attenuate noise are worthless. However, hearing protectors which are not comfortable will not be worn. The ideal hearing protector would reduce a worker’s exposure to noise while being comfortable over the course of a shift. In 2004-05 NIOSH headed a hearing protector study with a major automobile manufacturer and union allowing for the longitudinal study of Personal Attenuation Rating (PAR) and comfort (Murphy et al., 2009; Davis et al., 2011). Recently we became interested in re-visiting these data to determine if there was a relationship between comfort scores as measured on a paper scale (Figure 1, adapted from Park & Casali, 1991) and field measures of personal attenuation rating. Methods All methods were approved by the NIOSH Human Subjects IRB. The methods have been previously published (Davis et al. 2011; Murphy et al. 2009). Briefly, 224 worker volunteers were randomly assigned to three groups. Group A (n=70) was fitted with a semi- custom earplug which could be adjusted to produce an optimal PAR. Group B (n=82) was counselled in the correct placement and attenuation of one of the 6 plant-provided earplug hearing protectors. Group C (n=72) was a control group who wore one of the 6 plant- provided earplugs but received no counseling until the end of the study. Half-way through the study an additional 65 workers were added to Group A. During the 4 NIOSH visits (start, 1 month, 4 months and one year) a number of tests were conducted: otoscopy, PT audiometry, hearing health history, FitCheck, comfort, and attitudes & beliefs. We will only be concerned with the FitCheck and comfort data. All data were collected over this one year period. Figure 1 is one of the two forms of the comfort survey administered on all four visits. Scoring consisted of assigning a numerical value for each scale adjusted from most comfortable (1) to least comfortable (5). The PAR testing was conducted by 4 Michaels and Associates FitCheck systems run in parallel in a 4 person sound-treated room. Testing consisted of “open ears” followed by “closed ears” tests. Data were analyzed using Stata 13 (College Station, TX). A linear mixed effects model, based on the repeated measures design of the study, was utilized to analyze the data. Attenuation (PAR) was the outcome variable while the fixed explanatory variables were the comfort score, hearing protector device (HPD) and date of measurement (date). Initially, the comfort score was treated as a sum of all component scales. In the second part of the analysis the individual comfort scales were used as explanatory variables. Figure 2 A-G. The data for total comfort scores. The model with the total comfort score, type of HPD and date of measurement as independent variables is where: y ijkl = observed PAR for subject i, HPD j, date k and trial l. µ = overall mean. α j = fixed effect due to HPD type j, j=1…7 β k = fixed effect due to date k (k=1…4) (αβ) ij = fixed effect due to interaction between HPD type j and date k γ = coefficient representing fixed effect of total comfort score on PAR x ijkl = total comfort score for subject i, HPD j, date k and trial l s i = random effect due to subject i ε ijkl = random error for measurement subject i, HPD j, date k and trial l Data for the Total Comfort Scores Using Individual Components of Comfort Score as Explanatory Variables Identified Three Statistically Significant Components where: y ijkl = observed PAR for subject i, HPD j, date k and trial l. µ = overall mean. α j = fixed effect due to HPD type j, j=1…7 β k = fixed effect due to date k (k=1…4) (αβ) ij = fixed effect due to interaction between HPD type j and date k γ h = coefficient representing fixed effect of comfort score component h on PAR x ijkl = comfort component h score for subject i, HPD j, date k and trial l s i = random effect due to subject i ε ijkl = random error for measurement on individual i, HPD j, date k and trial l Test of Fixed Effects EffectNum DFDen DFF Valuep-value Painful-Painless110860.290.588 Comfortable-Uncomfortable110860.110.746 Uncomfortable Pressure-Not Uncomfortable Pressure110860.030.871 Tolerable-Intolerable110860.030.862 Loose-Tight110863.620.057 Bothersome-Not Bothersome110860.510.476 Light-Heavy110860.640.424 Not Cumbersome-Cumbersome110860.080.777 Hard-Soft110867.640.006 Hot-Cold110860.580.445 Rough-Smooth110866.260.013 Feeling of Complete Isolation- No Feeling of Complete Isolation110860.80.372 Ear Blocked-Ear Open110860.040.849 Ear Full-Ear Empty110861.290.257 HPD6108613.02<.0001 Date310861.860.135 HPD*Date1810861.630.047 where: y ijkl = observed PAR for subject i, HPD j, date k and trial l. µ = overall mean. α j = fixed effect due to HPD type j, j=1…7 β k = fixed effect due to date k (k=1…4) (αβ) ij = fixed effect due to interaction between HPD type j and date k γ h = coefficient representing fixed effect of comfort score component h on PAR, h=1,2,3 (hard, loose, rough) x hijkl = comfort component h score for subject i, HPD j, date k and trial l s i = random effect due to subject i ε ijkl = random error for measurement on subject i, HPD j, date k and trial l A Reduced Model Incorporating the Three Statistically Significant Comfort Components (above) Was Then Used. Test of Fixed Effects Num DFDen DFF ValuePr>F Loose-Tight110974.440.0353 Hard-Soft110979.520.0021 Rough-Smooth110976.370.0117 HPD6109713.69<.0001 Date310972.690.0453 HPD*Date1810971.830.0182 Figure 1. Version A of modified Park & Casali Comfort Survey. References Davis, R. R., W. J. Murphy, D. C. Byrne and P. B. Shaw (2011). "Acceptance of a Semi-Custom Hearing Protector by Manufacturing Workers." Journal of Occupational and Environmental Hygiene 8(12): D125-D130. Murphy, W., R. Davis, D. Byrne and J. Franks (2009). Advanced Hearing Protector Study. Conducted at: General Motors Metal Fabricating Division, Flint Metal Center, Flint, MI. Cincinnati, OH, NIOSH. 312-11a: 1-42. Park, M. Y. and J. G. Casali (1991). "An Empirical-Study of Comfort Afforded by Various Hearing Protection Devices - Laboratory Versus Field Results." Applied Acoustics 34(3): 151- 179. (Continued) Fit Statistics -2 Res Log Likelihood9723.1 AIC (smaller is better)9727.1 AICC (smaller is better)9727.1 BIC (smaller is better)9733.8 Test of Fixed Effects EffectNum DF Den DFF ValuePr>F Loose-Tight110974.440.0353 Hard-Soft110979.520.0021 Rough-Smooth110976.370.0117 HPD6109713.69<.0001 Date310972.690.0453 HPD*Date1810971.830.0182 Third Model Least Squares Means of PAR HPDDateEstimateSt. Err. AirSoft21.44592.1352 ComFit21.82676.4818 DeciDamp22.49990.7387 MaxLite25.50331.5276 PuraFit26.10420.8726 Quiet14.2182.2861 Sonomax16.53560.8863 123.3321.4054 221.82021.4124 319.59691.4296 419.89851.4486 EffectEstimateStd. Err. Rough-Smooth-0.95240.3773 Loose-Tight0.63880.3031 Hard-Soft1.17070.3794 The estimated values for the coefficients representing the effect of the comfort components were: Conclusion: As the variables loose-tight and hard-soft become more uncomfortable PAR increases. As the variable rough-smooth becomes more comfortable, the value of PAR increases. For this model Total Comfort was not significant but HPD and HPD*Date were.