MILITARY MEDICINE, 172, 4:426, 2007

Statistical Analysis of Hearing Loss among Navy Personnel

Guarantor: Robert P. Trost, PhD
Contributors: Robert P. Trost, PhD*; Geoffrey B. Shaw, BS†

Hearing loss is one of the most common disabilities among of Labor at 90 dB for 8 hours/day.6 The U.S. Navy abides by the
sailors. Service members are posted to a variety of stations. standards given in Office of the Chief of Naval Operations In-

Downloaded from https://academic.oup.com/milmed/article/172/4/426/4578069 by Universidade do Porto user on 23 February 2023

Many of these posts have high noise levels, and particular struction 5100.19D, Occupational Safety and Health Program
ratings have high exposure. If assignments of higher risk can Manual for Forces Afloat, which specifies the 8-hour time-
be identified, then focused prevention programs can be imple-
mented. The focus of this study was to determine how hearing
weighted average as 84 dBA. This instruction defines noise-
loss relates to service time spent aboard ships. To investigate hazardous areas as areas in which the routine noise is ⬎84 dBA
which duty stations and ratings are at high risk for hearing or peak noise levels are ⬎140 dB. In such areas, hearing pro-
loss, this study looked at medical hearing test records for tection or administrative controls are required. This study did
nearly 268,000 enlisted sailors over the period of 1979 to 2004. not evaluate the degree of compliance with these exposure
Using both logistic and logarithmic binomial analyses, the limits.
study found that enlisted sailors who spent one-half of a 30- The focus of this study was to determine whether the length of
year Navy career assigned to a surface warship had a 13-per- service aboard ships affects hearing loss. During a career in the
centage point higher probability of leaving the service with a U.S. Navy, service members characteristically are posted to a
reduction in the ability to hear, compared with someone who variety of stations, both afloat and ashore. Many of these posts
spent his or her whole career in a shore billet. If the same
sailor spent two-thirds of his or her 30-year Navy career as-
have high noise levels (such as certain ships) and particular
signed to a surface warship, then the probability of hearing ratings have high exposure (such as jobs near machines). If
loss at the end of his or her career would increase by another assignments of higher risk in both location and tasking can be
5 percentage points. identified, then preventative methods may reduce damage to the
hearing of Navy personnel. This would preserve quality of life for
Introduction personnel and save millions of dollars for the government.
To undertake this analysis, medical records from the Defense
ermanent hearing loss is one of the most common disabili- Occupational and Environmental Health Readiness System
P ties among sailors. This problem in active duty military
personnel has been documented in several studies. For exam-
(DOEHRS) regarding individual hearing tests given by the Navy
since 1979 were merged with information on each individual
ple, using data from 1995 to 1999, Bohnker et al.1 showed a sailor’s duty stations. With this merged data set, logistic and
significant threshold shift (STS) (if an individual has a STS, it logarithmic binomial analyses were able to determine the im-
implies a reduction in the ability to hear normally) as high as pact of time spent aboard different types of ships versus time at
25% in some Navy personnel ratings, and Rovig et al.2 found shore duty stations on the probability of a reduction in hearing
higher risk of hearing impairment on an aircraft carrier among function.
flight deck and engine room workers, compared with adminis-
trative staff members. Similar findings of hearing loss among Methods
military personnel were discussed by Gwin and Lacroix3 for
submariners, Wolgemuth et al.4 for Navy personnel, and Ridgley Description of Data
and Wilkins5 for both Army and Navy personnel. In this study, we describe how the dependent and explanatory
In 2004, the Veterans Affairs spent $108 million dollars in variables were defined and determined and what data sets were
disability payments to 15,800 former Navy personnel for hearing used to obtain this information. We obtained our measure of
loss. This represents an increase of $65 million in Veterans hearing loss from the DOEHRS. Only audiograms correspond-
Affairs spending on Navy hearing disability payments since ing to active duty Navy enlisted personnel during the period of
1999. 1979 to 2004 were used.
Given the enormity of the increase in spending by the Veter- We obtained our data on time spent at different duty stations
ans Administration on hearing loss disability payments to from the Navy’s Enlisted Master Record (EMR). Demographic
former Navy personnel, it is important to find the sources of this data such as the age, gender, and race also came from the EMR
hearing loss and then investigate ways to reverse this trend. data set. This data set has information on the careers of all
National standards of exposure are established by the Occupa- individuals who served in the Navy during the period of 1982 to
tional Safety and Health Administration of the U.S. Department 2004.
*Department of Economics, George Washington University, Washington, DC
20052. Definition of STS
†CNA Corp., Alexandria, VA 22311. The dependent variable is a measure of hearing loss. To con-
No official endorsement by the CNA Corp. or by George Washington University is struct this measure, the study used only annual and termina-
intended or should be inferred.
This manuscript was received for review in November 2005. The revised manu- tion audiograms taken during the period of 1979 to 2004, ob-
script was accepted for publication in June 2006. tained in accordance with the periodic evaluation required by
Reprint & Copyright © by Association of Military Surgeons of U.S., 2007. the Hearing Conservation Program. They represented audio-

Military Medicine, Vol. 172, April 2007 426

Analysis of Hearing Loss 427

grams recorded on Department of Defense form 2216, reported, is ⬍10 dB, we need to look at the right ear. All three frequencies
and maintained in the DOEHRS. Only audiograms correspond- changed by 20 dB in the right ear, and the average for the right
ing to active duty Navy enlisted personnel were studied. ear is 20 dB. Therefore, this person had a STS.
The study sample included 267,658 enlisted sailors who had
occupational exposure above an 8-hour time-weighted average Ship Type
of 84 dBA and enlisted sailors who had reached the termination We computed the number of months each individual spent
of their service. We excluded from the analysis obvious data assigned to different types of ships and in different job ratings
errors such as age outside a reasonable range, cases in which before the last hearing test. We also computed the number of
the only hearing test available was the one the sailors took when months each individual spent assigned to shore duty stations
they joined the Navy, cases in which there was a gap in service,

Downloaded from https://academic.oup.com/milmed/article/172/4/426/4578069 by Universidade do Porto user on 23 February 2023

and to the Air Wing. We then estimated the impact of these
and cases in which there was not at least one hearing test “assignment time” variables on the occurrence of a positive STS.
available after 1998 (the first year electronic data were col- We classified three different ship types, namely, surface warships,
lected). There are no readily available data to determine what submarines, and support ships. Surface warships included sur-
percentage of the exposed Navy population above an 8-hour face combatants, aircraft carriers, and amphibious ships.
time-weighted average of 84 dBA was actually tested. The re-
sults of this study apply only to the population of enlisted sailors Occupational Class
who undergo periodic evaluations as required by the Hearing On the basis of the published military literature and our
Conservation Program. discussions with Navy and enlisted personnel, we used three
The indicator of hearing loss used in this study was whether indicators for job conditions, namely, jobs that are near ship
the person had a STS. The definition of STS for the Navy has engines, jobs that involve heavy machinery such as power tools,
changed over the past 25 years. For consistency, we calculated and jobs on the flight deck or near aircraft. We based these
the STS variable by using the earliest and latest data on hearing categories on ratings. Approximately one-third of the enlisted
function available for each individual. personnel in our sample had jobs that did not fit into any of
Our indicator for STS was based on the change in hearing for those categories, such as cryptographers or yeomen. For those
three frequencies (2 kHz, 3 kHz, and 4 kHz) between the first ratings, we defined an “other” group.
and last years of available hearing tests for each individual. We selected individuals’ ratings at the time of their last hear-
When a December test was followed by additional tests in Jan- ing tests. Generally speaking, ratings for individuals remain the
uary, we counted the January tests as being in the same year as same or change very little over their Navy careers once they
the December test. If there was only 1 year of hearing tests for an finish basic training and Navy schooling. Therefore, the as-
individual, then we used the STS indicator computed in the sumption that individuals’ basic ratings at the time of their last
original data set for that individual. Otherwise, we averaged all hearing tests were the same as they were for their previous Navy
of the tests in each of the 2 years and looked at the difference careers is reasonable.
between the average scores in the 2 years. If the average differ-
ence of the three frequencies for either ear was ⱖ10 dB, then we TABLE II
considered the result a STS. This is the current definition of STS
SUMMARY STATISTICS OF DATA
used by the military. We also tried an alternative definition of
STS, that is, if individuals had a 15-dB change in any of the Variable Mean SD Minimum Maximum
three frequencies, then they were also flagged as having a STS,
STS (%) 11.01 31.30 0 100
even when the average of the three was ⬍10 dB. In terms of the
Age (years) 27.51 7.17 17 64
impact of time spent aboard ships, this alternative definition did Female (%) 12.04 32.54 0 100
not change our results. It should be noted that the amount of Race (%)
hearing loss was not quantified in this study; rather, it was a Caucasian 61.40 48.68 0 100
yes/no determination. African American 18.34 38.70 0 100
As an example, we look at someone who joined the Navy in Hispanic 10.68 30.89 0 100
1986 and was given a last hearing test in 2001, with only one Asian 6.23 24.18 0 100
test in each of those 2 years. Table I gives the results of those Other 3.38 18.06 0 100
tests for the relevant frequencies. Duty station (years)
Between 1986 and 2001, the left ear changed by 15 dB at 2 Air wing 4.48 3.97 0.08 23
Amphibious 2.78 1.77 0.08 14
kHz, did not change at 3 kHz, and changed by 10 dB at 4 kHz.
Carrier 3.04 2.09 0.08 21
The average of those is (15 ⫹ 0 ⫹ 10)/3 ⫽ 8.3 dB. Because this Surface combatant 3.43 2.52 0.08 17
Submarine 4.74 3.32 0.08 17
TABLE I
Support ship 3.10 2.04 0.08 19
SAMPLE DATA FOR DETERMINING A STS Shore 3.37 3.95 0.08 30
Length of service 6.79 6.24 0.08 30
Response (dB) (years)
Left Ear Right Ear The mean and SD of ship, submarine, air wing and shore years does
2 kHz 3 kHz 4 kHz 2 kHz 3 kHz 4 kHz not include zero observations. The minimum and maximum values of
the percentage variables are 0 if the individual does not have the
1986 10 25 20 5 5 5
characteristic and 100 if the individual has the characteristic. Number
2001 25 25 30 25 25 25
of observations, 267,658.

Military Medicine, Vol. 172, April 2007

428 Analysis of Hearing Loss

Summary Statistics binomial function. The logistic function has the following math-
We obtained our data on time spent at different duty stations ematical form: P(STS ⫽ 1) ⫽ e␥⬘ Z⫹␤⬘ X/1 ⫹ e␥⬘ Z⫹␤⬘ X.
from the EMR. This data set contains historical data on where For the logarithmic binomial model, the functional form is:
individuals spent their careers in the Navy and their job ratings, P(STS ⫽ 1) ⫽ log(␥⬘ Z ⫹ ␤⬘ X).
for the period of 1982 to 2004. We used these data to determine the In these equations, ␥ and ␤ are vectors of unknown parameters,
total time individuals had been assigned to various types of ships, X is a vector of individual characteristics such as the natural
assigned to the Air Wing, or assigned to shore duty stations. As logarithm of age (to account for the nonlinear impact of age on
noted above, we defined three broad categories of ships, namely, hearing loss), gender, race, and an intercept term, and Z is a
surface warships, submarines, and support ships. The EMR data vector of variables that define a sailor’s career, such as the
length of time spent at various duty stations, including the time

Downloaded from https://academic.oup.com/milmed/article/172/4/426/4578069 by Universidade do Porto user on 23 February 2023

set was also used to assign each individual to one of four job
classifications based on the Navy rating, that is, engines, spent on different types of ships and on shore duty. The param-
machines, aircraft, or other. Demographic data, such as age, eters in ␥ are the focus of this study. In comparisons of the
gender, and race, also come from the EMR data set. magnitudes of the coefficients in specific “time spent” variables
Table II gives the summary statistics for the 267,658 enlisted in the Z vector, the larger the magnitude of the coefficient, the
sailors who were analyzed in this study. The main explanatory greater is the impact on hearing loss from time spent at that
variables of interest were the numbers of months each individ- particular duty station.
ual had been assigned to different classes of ships, the Air Wing,
or shore duty stations before his or her most recent hearing test. Results
Table II gives the mean and range of these variables for individ-
uals who spent ⱖ1 month at any of the posts. Table III gives the coefficient estimates of the logistic model for
enlisted sailors, and Table IV gives the estimates for the loga-
Model rithmic binomial model. Table V gives the relative risk (RR), as
computed from the logarithmic binomial model. The baseline or
The probability of a STS depends on various covariates Z and comparator group, with a RR of 1.0, is a Caucasian male of any
X and is modeled as both a logistic function and a logarithmic age, with no time spent onboard a ship or at a shore duty
TABLE III TABLE IV
LOGISTIC MODEL ESTIMATES FOR STS EQUATION LOGARITHMIC BINOMIAL MODEL ESTIMATES FOR STS EQUATION

SE of ␹2 for Null SE of ␹2 for Null

Coefficient Coefficient Hypothesis That Coefficient Coefficient Hypothesis That
Variable Estimates Estimates Coefficient Is 0 Variable Estimates Estimates Coefficient Is 0
Constant ⫺5.0622 0.1685 903.02 Constant ⫺4.9667 0.1454 1167.24
Logarithm of age 0.8147 0.0548 221.31 Logarithm of age 0.7652 0.0471 263.51
Female ⫺0.4216 0.0243 300.45 Female ⫺0.3848 0.0222 299.60
African American ⫺0.3270 0.0182 324.09 African American ⫺0.2886 0.0159 328.09
Hispanic ⫺0.1212 0.0219 30.55 Hispanic ⫺0.1068 0.0193 30.65
Asian ⫺0.2002 0.0264 57.57 Asian ⫺0.1694 0.0225 56.72
Other race ⫺0.0522 0.0389 1.80 Other race ⫺0.0498 0.0347 2.06
Machine air wing 0.0672 0.0210 10.24 Machine air wing 0.0546 0.0167 10.68
years years
Aircraft air wing years 0.0365 0.0028 175.74 Aircraft air wing years 0.0297 0.0023 167.42
Other air wing years 0.0391 0.0055 51.46 Other air wing years 0.0324 0.0045 52.67
Engine warship years 0.0512 0.0057 80.94 Engine warship years 0.0405 0.0046 76.07
Machine warship 0.0632 0.0035 328.30 Machine warship 0.0488 0.0028 303.65
years years
Aircraft warship years 0.0514 0.0051 100.22 Aircraft warship years 0.0402 0.0042 92.87
Other warship years 0.0795 0.0037 452.29 Other warship years 0.0603 0.0029 427.52
Engine submarine 0.1395 0.1285 1.18 Engine submarine 0.1061 0.0835 1.62
years years
Machine submarine 0.0304 0.0044 47.23 Machine submarine 0.0245 0.0037 44.02
years years
Other submarine 0.0549 0.0058 91.21 Other submarine 0.0436 0.0045 92.96
years years
Engine support ship 0.0246 0.0141 3.04 Engine support ship 0.0197 0.0116 2.86
years years
Machine support ship 0.0517 0.0058 79.84 Machine support ship 0.0409 0.0046 78.15
years years
Other support ship 0.0513 0.0086 35.39 Other support ship 0.0425 0.0068 38.52
years years
Non-ship years 0.0427 0.0024 325.83 Non-ship years 0.0348 0.0020 310.58
Number of observations, 267,658; STS, 11.01%. Number of observations, 267,658; STS, 11.01%.

Military Medicine, Vol. 172, April 2007

Analysis of Hearing Loss 429

TABLE V spent in shore duty stations was 0.0348, the coefficient for time
LOGARITHMIC BINOMIAL MODEL RR ESTIMATES spent in the machine room on surface warships was 0.0488,
and the coefficient for time spent on surface warships in the
95% Confidence “other” category was 0.0603. This means that time spent on
Variable RR Estimates Estimates of RR surface warships was far more damaging to hearing than time
Female 0.681 0.652–0.711 spent at shore duty stations. Although variables such as age,
African American 0.749 0.726–0.773 race, and gender were not the primary focus of this study, the
Hispanic 0.899 0.865–0.933
Asian 0.844 0.808–0.882
study did control for these effects in the analysis. Like other
Other race 0.951 0.808–1.018 studies, we found that hearing loss increased with age and

Downloaded from https://academic.oup.com/milmed/article/172/4/426/4578069 by Universidade do Porto user on 23 February 2023

Machine air wing years 1.056 1.022–1.091 that men and Caucasian subjects had a higher prevalence of
Aircraft air wing years 1.030 1.026–1.035 hearing loss, compared with women and African American
Other air wing years 1.033 1.024–1.042 subjects.7
Engine warship years 1.041 1.032–1.051 There are two ways to look at the importance of coefficient
Machine warship years 1.050 1.044–1.056
estimates, that is, statistical significance and practical signifi-
Aircraft warship years 1.041 1.033–1.050
Other warship years 1.062 1.056–1.068
cance. We tested for statistical significance between two coeffi-
Engine submarine years 1.112 0.944–1.310 cient estimates with a logarithmic-likelihood ratio c12 (␹2). How-
Machine submarine years 1.025 1.017–1.032 ever, with sample sizes this large, most coefficients have
Other submarine years 1.045 1.035–1.054 statistical significance, as was indeed the case for this study. A
Engine support ship years 1.020 0.997–1.043 better way to look at the importance of coefficients is practical
Machine support ship years 1.042 1.032–1.051 significance. For practical significance, we look at the impact of
Other support ship years 1.043 1.030–1.058 these coefficients on the predicted probability of getting a STS in
Non-ship years 1.035 1.031–1.039 a Navy career. This is done in Figure 1 for the logistic estimates
and in Figure 2 for the logarithmic binomial estimates. Figure 1
station. The RR of 1.062 for “other” surface warships means that indicates that, if one-half of an individual’s time during a 30-
1 year of duty on a surface warship in the “other” category year Navy career is spent on a warship in the “other” category,
increases the RR from 1.0 to 1.062. If the same 1 year is spent then the probability of a STS at the end of his or her career is
at a shore duty station instead, then the RR increases only from 0.48; if the individual spends two-thirds of his or her time on a
1.0 to 1.035. In comparisons of the coefficients of different surface warship, then the probability increases to 0.53. As a
length of time variables, a larger positive coefficient means that comparison, if an individual spends his or her entire career at a
time spent in that particular location (ship, shore duty station, shore duty station, then the probability of a STS is only 0.35.
or Air Wing) has a greater impact on hearing loss than the The conclusion that can be drawn from this is that time spent on
corresponding smaller positive coefficient. For example, for en- surface warships is far more damaging to hearing than time
listed personnel, the logarithmic binomial coefficient for time spent at shore duty stations.

Fig. 1. Comparison of three different Navy careers and their impact on STS by using logistic analysis.

Military Medicine, Vol. 172, April 2007

430 Analysis of Hearing Loss

Downloaded from https://academic.oup.com/milmed/article/172/4/426/4578069 by Universidade do Porto user on 23 February 2023

Fig. 2. Comparison of three different Navy careers and their impact on STS by using logarithmic binomial analysis.

Conclusions wearing of hearing protection devices has the potential to reduce

the number of sailors who might have a potential hearing loss in
Permanent hearing loss is one of the most common disabilities their naval careers by 18 percentage points.
among sailors. The primary focus of this study was to determine
how service aboard ships affects hearing loss. The DOEHRS med-
ical records of individual hearing tests given by the Navy since References
1979 were merged with information on each individual sailor’s
1. Bohnker BK, Page JC, Rovig GW, Betts LS, Sack DM: Navy hearing conservation
duty stations.
program: threshold shifts in enlisted personnel, 1995–1999. Milit Med 2002; 167:
With this merged data set, logistic and logarithmic binomial 48 –52.
analyses were able to determine that time spent on a certain 2. Rovig GW, Bohnker BK, Page JC: Hearing health risk in a population of aircraft
type of ship had a much greater impact on hearing loss than did carrier flight deck personnel. Milit Med 2004; 169: 429 –32.
time spent at shore duty stations. Time spent on surface war- 3. Gwin LP, Lacroix PG: A prevalence study of hearing conservation in the United
States naval submarine force. Milit Med 1985: 150: 652– 6.
ships had the greatest impact on potential hearing loss. 4. Wolgemuth KS, Luttrell WE, Kamhi AS, Mark DJ: The effectiveness of the Navy’s
For enlisted personnel, the predicted probability of a potential hearing conservation program. Milit Med 1995; 160: 219 –22.
hearing loss if one-half of the time in a 30-year career was spent 5. Ridgley CD, Wilkins JR: A comparison of median hearing thresholds from U.S.
assigned to a surface warship was 0.48; if two-thirds of the time Navy and U.S. Army audiometric data bases. Milit Med 1991; 156: 343–5.
were spent on a surface warship, then the probability increased 6. Wallace RB (editor): Maxcy-Rosenau-Last Public Health and Preventive Medicine,
Ed 14, p 642. New York, McGraw-Hill, 1998.
to 0.53. If the same sailor spent all of the 30-year career at shore 7. Bohnker BK, Page JC, Rovig GW: Navy hearing conservation program: 1995–1999
duty stations, then the predicted probability would be only 0.35. retrospective analysis of threshold shifts for age, sex, and officer/enlisted status.
Reducing noise levels on surface warships or enforcing the Milit Med 2004; 169: 73– 6.

Military Medicine, Vol. 172, April 2007