Noise Health Home 

[Download PDF]
Year : 2008  |  Volume : 10  |  Issue : 40  |  Page : 69--73

Fraction of work-related accidents attributable to occupational noise in the city of Botucatu, São Paulo, Brazil

Adriano Dias1, Ricardo Cordeiro2,  
1 Research Support Center Coordinator from Botucatu Medical School, Sao Paulo State University, Botucatu SP, Brazil
2 Department of Preventive and Social Medicine from Medical Sciences School, Campinas State University, Campinas/SP, Brazil

Correspondence Address:
Adriano Dias
Rua Nelo Cariola, 252, Botucatu, Sao Paulo, 18603570


Background: Noise is the most common agent of occupational exposure. It may induce both auditory and extraauditory dysfunction and increase the risk of work accidents. The purpose of this study was to estimate the fraction of accidents attributable to noise occupational exposure in a mid-size city located in southeastern Brazil. Materials and Methods: In this population case-control study, which included 108 cases and 324 controls, the incidence rate ratio of work accidents controlled for several covariables was obtained by classifying occupational noise exposure into three levels, as well as determining the prevalence in each level. Results: Based on these data, the attributable fraction was estimated as 0.6391 (95% CI = 0.2341-0.3676), i.e., 63% of the work accidents that took place in the study site were statistically associated with occupational noise exposure. Discussion: The causes of this association as well as its implications in the prevention of work accidents are discussed.

How to cite this article:
Dias A, Cordeiro R. Fraction of work-related accidents attributable to occupational noise in the city of Botucatu, São Paulo, Brazil.Noise Health 2008;10:69-73

How to cite this URL:
Dias A, Cordeiro R. Fraction of work-related accidents attributable to occupational noise in the city of Botucatu, São Paulo, Brazil. Noise Health [serial online] 2008 [cited 2022 Sep 30 ];10:69-73
Available from:

Full Text


Since the beginning of time, noise has been part of man's life. Noise exposure became more intense and prevalent as it started to be an issue in working environments on a universal scale. Today, noise exposure is considered to be the most common occupational hazard [1],[2],[3],[4] afflicting millions of workers. Hearing loss, the most common problem caused by work noise, is an irreversible and chronic condition of its insidious onset that undermines the hearing, and hence, the communication ability of those exposed to the noise. In addition to auditory damage, noise exposure has also been associated with extraauditory effects such as cardiovascular, [5] endocrine, [6],[7] and gastrointestinal diseases. [8]

Besides auditory and extraauditory physiological changes, noise has also been associated with a greater risk of work accidents among workers exposed to intense occupational noise in comparison with unexposed workers. [9],[10],[11],[12],[13],[14],[15] Therefore, programs for noise reduction and control not only seek to prevent exposure and auditory damage, but also to allow the decrease of work accident risks. [13],[16]

The purpose of this study was to estimate the fraction of work accidents attributable to occupational noise exposure in a mid-size Brazilian city.

 Materials and Methods

This population-based case-control analysis is part of a broader study of the relationship between nonworkplace stress and work accidents entitled, "Work-related accidents research program". [15],[17] It was conducted in Botucatu, a mid-size city located in the state of Sγo Paulo in southeastern Brazil, with about 130,000 inhabitants, a small agricultural sector, a modest industrial park (though some large industries are also found), and well-developed service and commercial sectors. Ethical approval was obtained from the Botucatu Medical School Review Board.

The study sample consisted of the economically active population (EAP) residing in the city. Cases were defined as city residents who reported a work-related accident within the preceding 90 days prior to the study. Subjects were selected through a systematic random sampling of households in the urban area of Botucatu where 94.6% of the EAP lives.

Interviewers visited the selected households. If no person was home, the visit was rescheduled to a later date at a different time. If no contact with a household resident was made for the second time, the visit was once again rescheduled to a later date at a different time. If a third failed home visit occurred, or household members refused to be interviewed, the household was excluded without replacement.

At each visit, an adult resident was interviewed to obtain biological, socioeconomic, and occupational data on all members of the household older than nine years of age. Respondents were also asked whether anyone had experienced any kind of accident (related to work, household chores, leisure, etc.) within the preceding 90 days. If the answer was positive, a new visit was scheduled for the interviewing of the presumed victim who would then be asked to confirm that an accident had occurred, and that it really took place during the study period, i.e ., 90 days before the first visit. Following confirmation, the accident was registered and its occupational character was checked. All cases of occupational accident were invited to take part in the study.

As each case was identified, three controls were randomly selected from a population of active workers who reported no injuries over the same 90 day period. Controls were matched by gender, age group (age ▒ one year), and census city zone. Data were continuously collected from controls throughout the sampling process following the same procedures for the identification and interviewing of study cases.

In addition to the nonwork variables detailed by Cordeiro and Dias, [17] the questionnaire included the following:

Worker's age: continuous variable measured in yearsEmployment work status: dichotomous categorical variable providing information on whether the worker was under an employment contractMean regular daily working hours in the 90 days preceding the study: continuous variable, measured in hoursMean weekly overtime 90 days prior to the study: continuous variable, measured in hoursMean number of co-workers in the same work section: discrete variableType of work shift: categorical variable, classified as "fixed day shift," "fixed night shift," and "alternated shifts". "Fixed day shift" was set as a point of reference, and categorized according to two dummy variables.Intense noise-exposure at current occupation, categorical variable expressing the worker's perception regarding the presence of intense and repeated noise in the work environment, classified as "always", "sometimes" and "never".Occupation group (Brazilian Standard Classification of Occupations): [18] categorical variable that classifies occupations into nine main groups. Taking the category, "scientists" as a reference, this variable was categorized according to eight dummy variables: police officers, white-collar workers, service providers, farmers, blue-collar workers, and maintenance workers.

Initially, simple conditional logistic regression models [19] were adapted (1:3 matching ratio) using the occurrence of an accident as a dependent dichotomous variable (control = 0; case = 1), and the response to each one of the categories described above as independent variables.

Subsequently, a multiple logistic regression model was fit. [19] In this model, accident occurrence was the dichotomous categorical response variable (control = 0; case = 1) and the variables that yielded incidence rate ratio estimates with P ≤ 0.25 in the univariate model were the predictive variables. [20] Among these predictive variables, the variable used to assess the relationship between occupational noise exposure and accident occurrence was the worker's perception of occupational noise exposure, which was classified into the three aforementioned categories and controlled for years of schooling, mean number of overtime hours, mean daily working hours, number of co-workers in the same section, alternative shift, and major occupation group.

Finally, the attributable fraction of work accidents related to occupational noise exposure was estimated using the usual expression [20],[21],[22] adjusted by the authors of this study to express exposure at different levels:



i = indicator of the different levels of noise exposure used in this study.

P i = prevalence of workers exposed to noise of level i in the source population.

RR i = incidence rate ratio (relative risk-RR) of work accidents, between workers exposed to level i noise and those not exposed to occupational noise in the study population.

The prevalence of workers exposed to level i noise in the source population were estimated by the prevalence of these workers in the control group. AF confidence interval was estimated by a logarithmic transform as proposed by Walter [23] and adapted by the authors of this study:



i = indicator of the different levels of exposure to noise used herein

AF = attributable fraction

z 1-a/2 = 100(1-a/2) percentile of normal standard distribution

n i = number of cases exposed to level i noise

n 0 = number of nonexposed cases

n = total number of cases

m i = number of exposed controls

m 0 = number of nonexposed controls

m = total number of controls


Sampling and interviewing, which resulted in the enrollment of 108 cases and 324 controls, took place between 16 th May and 15 th October, 2002 and included the 195 urban census zones of Botucatu. [Table 1] shows the age distribution among the workers who suffered accidents and were selected for this study. It is noteworthy that 34% of these accidents involved young adults aged up to 30 years with 10% of them being aged up to 20 years.

The most frequently reported types of accidents were cuts, bruises, fractures, and acute articulatory injuries. The most commonly injured body parts were the hands, the remaining parts of the upper limbs, the head except the eyes, and the lower limbs except the feet. Improper operation of machinery or equipment, fall from a height, automobile accidents, and falling objects were the immediate primary causes. Most accidents were either mild or moderate, and 85 of them (78.7%) resulted in sick leave of up to 15 days.

[Table 2] shows the incidence rate ratio estimates obtained by applying the univariate logistic model mentioned above.

While fitting the multivariate logistic model, the variables, "work place is always noisy" and "workplace is sometimes noisy" were found to be risk factors for work-related accidents, with adjusted incidence rate ratio (RR) estimates being 3.660 ( P = 0.0003, 95% CI 1.817-7.370) and 4.995 ( P P > 0.05 and were discarded from the final result.

No statistically significant interaction term (α = 0.05) was observed among the selected variables. The analysis of the adjusted model residuals revealed no violation of the logistic model assumptions. Statistical analysis was performed using the SAS 8.12 software. [24]

Of the 108 work accidents studied, 18 workers reported being "always exposed to intense occupational noise" and 53 reported being "sometimes exposed to intense noise". Of the 324 controls, 32 and 68 reported "always" and "sometimes exposed to intense noise," respectively. By applying these results to equations (1) and (2), the fraction of work accidents attributable to constant and intermittent occupational noise and its respective confidence interval were estimated as 0.6391 and 95% CI = 0.5856-0.6858 respectively.


The attributable fraction is the proportion of all cases of a disease in the study population attributable to a specific exposure. It is an epidemiological concept which relates the relative risk of a disease to the populational prevalence rate of exposures believed to cause the disease. It is usually interpreted as the percentage of cases that could be prevented if a disease-causing exposure were eliminated. [20] This parameter can be very useful in public health, especially when there is a need to choose between different prevention strategies. [20] The concept of attributable fraction was originally laid down by Levin [25] who named it "attributable risk; " its properties were studied by Walter. [26] Unfortunately, several authors started to use the expression, "attributable risk: to refer to the difference in incidence rates between exposed and nonexposed populations. For this reason, we opted for the term, "attributable fraction" used for the first time by Walter. [23] The same concept has also been called, "population attributable risk percentage" by Cole and MacMahon, [27] "etiologic fraction" by Miettinen, [28] and "excess fraction" by Greenland and Robins. [29]

Occupational noise exposure is known to occur at different levels, depending on the characteristics of the occupation. Given that measuring noise exposure in each study case was impossible, the authors opted for working with the workers' own perception of such exposure, which was categorized into three levels. This approach resulted in two estimates of work accident risk as functions of the noise exposure level perceived, with nonexposure being considered as the baseline level. Thus, the incidence rate ratio estimates were 3.660 and 4.995 for workers who reported being "sometimes" and "always" exposed to noise, respectively.

In the original concept of AF, exposure is considered to be homogeneous. [20],[21],[22] In this study, however, it was estimated as the sum of two exposure levels found in the study population by generalizing the traditional AF estimator [20],[21],[22] for a scenario of different exposure strata. Noise exposure prevalence was determined by the prevalence of continuous or intermittent noise exposure reported in the control group, given that this is the role of controls in case-control studies. Indeed, several studies have shown that incidence rate ratio estimators, used herein to calculate AF, are likely to be found in population-based case-control studies. [30]

In this study, the fraction of work accidents attributable to work noise exposure in Botucatu in 2002, was estimated to be 63.9%. That is, more than half of the work accidents that took place in Botucatu could be avoided by eliminating noise exposure at work provided this is one of the causal factors of work accidents. Verifying this condition however, is difficult and complex. [31],[32] .

The statistical association between noise and work accidents demonstrated here and also by many others [10],[11],[12],[13],[14],[15] supports the existence of a causal relationship between these two variables. Nonetheless, the fact that noisy work environments could (and generally do) pose accident risks other than noise should not be overlooked. The relationship between noise and accidents could be a bias. However, according to the above logistic analysis, [17] the risk calculation used in this study to estimate the attributable fraction was controlled for several covariables. This strategy was used to at least partially control possible confounding effects resulting from the lack of comparability between cases and controls in relation to occupational risks other than noise.

Another important question in causality evaluation is the so-called plausibility. It seems very plausible that noise is an accident causal factor when it imposes difficulties in communication (in the detection, discrimination, location, and identification of sound) on the worker, as well as in speech intelligibility, [13],[33] maintenance of attention and concentration, [34],[35] memory problems, [35] stress, [14],[36],[37] and excessive fatigue. [34],[36] These factors are known to be involved in the genesis of work accidents.

Breslow and Day [22] recommend that without causality evidence, AF should be cautiously interpreted as the proportion of cases explained by the exposure, where the term, "explained" is used in the strict sense of statistical association.

Work accidents constitute a major problem in public health in both developed and developing countries. Differently from what the words suggest, these are not accidental events, [38] but socially determined [39] and preventable phenomena. The attributable fraction estimated here amply justifies investments in auditory protection programs, particularly those focusing on controlling sound emission at the source in order to maintain auditory health as well as to reduce the number of work accidents. The achievement of such reduction would emphasize the causal role of noise in work accidents.


1Alberti PW. Deficiκncia Auditiva Induzida pelo Ruνdo. In: Lopes Filho O, Campos CAH, editors. Tratado de Otorrinolaringologia. Sγo Paulo: Roca; 1994. p. 934-49.
2Seligman J. Sintomas e sinais na PAIR. In: Nudelmann AA, Costa EA, Seligman J, Ibaρez RN, editors. PAIR: Perda Auditiva Induzida pelo Ruνdo. Porto Alegre: Baggagem; 1997. p. 143-51.
3Baker DB, Landrigan PJ. Workers. In: Detels R, McEwen J, Beaglehole R, Tanaka H, editors. Oxford Textbook of Public Health. Oxford: Oxford University Press; 2002. p. 1639-56.
4Costa EA, Morata TC, Kitamura S. Patologia do ouvido relacionada com o trabalho. In: Mendes R, editor. Patologia do trabalho. Sγo Paulo: Atheneu; 2003. p. 1254-82.
5Rosenman KD. Cardiovascular disease and environmental exposure. Br J Ind Med 1979;36:85-97.
6Anticaglia JR, Cohen A. Extra-auditory effects of noise as a health hazard. Am Ind Hyg Assoc J 1970;31:277-81.
7Odescalchi CP. Effetti extra-auditivi del rumore: Rumore a rendimento industriale. Med Lav 1970;61:629-44.
8Okamoto VA, Santos UP. Outros efeitos do ruνdo no organismo. In: Santos UP, editor. Ruνdo: riscos e prevenηγo. Sγo Paulo: Hucitec; 1999. p. 89-91.
9Cohen A. Industrial noise and medical absence and accident record data on exposed workers. In: The International Congress on Noise as a Public Health Problem. Washington DC: 1976. p. 441-53.
10Kjellberg A. Subjective, behavioral and psychophysiological effects of noise. Scand J Work Environ Health 1990;16:29-38.
11Barreto SM, Swerdlow AJ, Smith PG, Higgins CD. A nested case-control study of fatal work related injuries among Brazilian steel workers. Occup Environ Med 1997;54:599-604.
12Moll van Charante AW, Mulder PG. Perceptual acuity and the risk of industrial accidents. Am J Epidemiol 1990;131:652-63.
13Berger EH, Royster LH, Royster JD, Driscoll DP, Layne M. The noise manual. 5th ed. Akron: American Industrial Hygiene Association; 2000.
14Melamed S, Luz J, Green MS. Noise exposure, noise annoyance and their relation to psychological distress, accident and sickness absence among blue-collar workers: The Cordis Study. Isr J Med Sci 1992;28:629-35.
15Cordeiro R, Clemente AP, Diniz CS, Dias A. Exposiηγo ao ruνdo ocupacional como fator de risco para acidentes do trabalho. Rev Saude Publica 2005;39:461-6.
16Cohen A. The influence of a company hearing conservation programme on extra-auditory problems in workers. J Safety Res 1976;8:146-62.
17Cordeiro R, Dias A. Stressful life events and occupational accidents. Scand J Work Environ Health 2005;31:336-42.
18Instituto Brasileiro de Geografia e Estatνstica. Classificaηγo de Ocupaηυes - Censo Demogrαfico de 2000. 2000 [cited 2002 Jul 1]. Available from:
19Hosmer Jr. D, Lemeshow S. Applied logistic regression. New York: John Wiley and Sons; 2000.
20Rothman KJ, Greenland S. Modern epidemiology. Philadelphia: Lippincott-Raven; 1998.
21Schlesselman JJ. Case-control studies: design, conduct, analysis. New York: Oxford University Press; 1982.
22Breslow NE, Day NE. Statistical methods in cancer research. Vol. 1. The analysis of case-control studies. Lyon: YARC; 1980.
23Walter SD. The estimation and interpretation of attributable risk in health research. Biometrics 1976;32:829-849.
24SAS Institute. The SAS System for Windows. 8.2 ed. In. Cary: SAS Institute; 2001.
25Levin ML. The occurrence of lung cancer in man. Acta Unio Int Contra Cancrum 1953;9:531-41.
26Walter SD. The distribution of Levin's measure of attributable risk. Biometrika 1975;62:371-4.
27Cole P, MacMahon B. Attributable risk percent in case-control studies. Br J Prev Soc Med 1971;25:242-4.
28Miettinen OS. Proportion of disease caused or prevented by a given exposure, trait or intervention. Am J Epidemiol 1974;99:325-32.
29Greenland S, Robins JM. Conceptual problems in the definition and interpretation of attributable fractions. Am J Epidemiol 1988;128:1185-97.
30Cordeiro R. O mito da doenηa rara. Rev Bras Epidemiol 2005;8:111-6.
31Weed DL. On the logic of causal inference. Am J Epidemiol 1986;123:965-79.
32Susser M. What is a cause and how do we know one? A grammar for pragmatic epidemiology. Am J Epidemiol 1991;133:635-48.
33Hιtu R, Quoc HT. Psychoacoustic performance in workers with NIHL. In: Axelsson A, Borchgrevink H, Hamernik RP, Hellstrom P, Henderson D, Salvi RJ, editors. Scientific basis of noise-induced hearing loss. New York: Thieme; 1996. p. 264-85.
34Ribeiro HP, Lacaz FA. Acidentes de trabalho. In: Diesat, editor. De que adoecem e morrem os trabalhadores. Sγo Paulo: DIESAT; 1984. p. 65-85.
35Berglund B, Lindvall T. Community noise. Stockholm: World Health Organization; 1995.
36Ferreira Jr. M. Perda auditiva induzida pelo ruνdo. In: Ferreira M Jr, editor. Sa˙de no trabalho. Sγo Paulo: Roca; 2000. p. 262-85.
37Gessinger R, Castoldi L, Fensterseifer LM. Efeitos psicossociais da perda auditiva induzida pelo ruνdo (PAIR). In: Nudelmann AA, Costa EA, Seligman J, Ibaρez NR, editors. PAIR: perda auditiva induzida pelo ruνdo. Porto Alegre: Baggagem; 1997. p. 251-4.
38Tsai SP, Bernacki EJ, Dowd CM. The relationship between work-related and non-work-related injuries. J Community Health 1991;16:205-12.
39Dwyer T. Life and death at work: Industrial accidents as a case of socially produced error. New York: Plenum Press; 1991.