Article Text

Methodology
Under-estimation of self-reported occupational exposure by questionnaire in hospital workers
  1. C Donnay1,
  2. M-A Denis2,3,
  3. R Magis4,
  4. J Fevotte3,5,
  5. N Massin6,
  6. O Dumas7,8,
  7. I Pin9,
  8. D Choudat1,4,
  9. F Kauffmann7,8,
  10. N Le Moual7,8
  1. 1Université Paris Descartes, AP-HP, Paris, France
  2. 2Hospices Civils de Lyon, Groupement Hospitalier Est, Médecine du Personnel, Bron, France
  3. 3UMRESTTE UMR T9405 (INRETS; Université de Lyon; Université Lyon 1; Institut de Veille Sanitaire), Bron, France
  4. 4Service Central de Médecine du Travail, AP-HP, Paris, France
  5. 5Institut de Veille Sanitaire, Département Santé Travail (DST/InVS), Saint Maurice, France
  6. 6Département Epidémiologie en Entreprise, INRS, Vandœuvre-lès-Nancy, France
  7. 7Inserm, CESP Centre for Research in Epidemiology and Population Health, U1018, Respiratory and Environmental Epidemiology Team, Villejuif, France
  8. 8Université Paris Sud 11, UMRS 1018, Villejuif, France
  9. 9Inserm, U823, Centre de Recherche Albert Bonniot, La Tronche, France
  1. Correspondence to Nicole Le Moual, Inserm U 1018/CESP, Centre for Research in Epidemiology and Population Health, Respiratory and Environmental Epidemiology, 16, avenue Paul Vaillant Couturier, 94807 Villejuif Cedex, France; nicole.lemoual{at}inserm.fr

Abstract

Objectives The aim of the study was to determine whether self-reported occupational exposure to cleaning/disinfecting agents in hospital workers is accurate, in comparison to expert assessment, taken to be the gold standard.

Methods In the Epidemiological Study of the Genetics and Environment of Asthma (EGEA), participants were interviewed on occupation with a specific questionnaire for hospital workers regarding tasks and cleaning/disinfecting agents. Two estimates of exposure were available: self-report and expert assessment. The expert assessment involved a standardised procedure to estimate intensity, frequency and probability of exposure for each job. The present analysis focused on eight exposures: formaldehyde, glutaraldehyde, bleach/chlorine, alcohol, quaternary ammonium components, ammonia, sprays and latex gloves. Agreement and differences between self-reported and expert estimates were studied by kappa and phi coefficients and McNemar tests, respectively.

Results In the survey of 1571 adults, 176 ever hospital workers (327 occupations) with both self-reported and expert exposure assessments were studied. An underestimation of self-reported exposure was observed especially for formaldehyde (26.5% vs 32.7%, p=0.01), ammonia (7.4% vs 18.8%, p<0.0001), alcohol (64.9% vs 93.0%, p<0.0001) and quaternary ammonium components (16.6% vs 70.9%, p<0.0001), compared to expert assessment.

Conclusion Occupational exposure to disinfecting/cleaning agents is common and high in hospitals. A large underestimation of self-reported exposure and a lack of knowledge of product components was observed. Our results show the relevance of expert assessment in epidemiological studies to limit measurement bias. This work underlines the need for health education programmes on the occupational risks induced by these types of products.

  • Healthcare worker
  • epidemiology
  • exposure assessment
  • disinfectant

Statistics from Altmetric.com

What this paper adds

  • Self-reported exposure was significantly under-estimated in comparison to expert assessment in this study, possibly because of a lack of knowledge of cleaning/disinfecting product components in hospital workers.

  • This study underlines the relevance of expert exposure assessment in epidemiological studies to limit measurement bias.

  • Healthcare workers need better training to improve their knowledge of occupational exposures to protect them from the occupational risks presented by disinfecting and cleaning products.

Introduction

There is growing evidence that healthcare workers are at increased risk of asthma.1 2 Various cleaning/disinfecting products for healthcare hygiene are used every day by healthcare workers (nurses, auxiliary nurses, physicians, cleaners, dental assistants). To protect patients from nosocomial infections, the frequency of disinfecting or cleaning tasks has increased, as well as the use of cleaning and disinfecting sprays.3 New, potentially asthmogenic, substances have recently been introduced into hospitals to replace some suspected carcinogenic products (ie, formaldehyde).4

Healthcare workers may be exposed to high levels of a number of pollutants (cleaning and sterilising products, aerosolised medications) and to sensitisers in particular (latex, surface disinfectants, biocides).5 Disinfecting and cleaning products, such as bleach/chlorine and ammonia, are skin, nose and lung irritants or are sensitisers (low molecular weight agents), such as biocides.5 The most common products used in hospitals are disinfectants (chlorhexidine, glutaraldehyde, quaternary ammonium components) used to sterilise medical instruments and to clean floors and surfaces. Cleaning/disinfecting products are complex mixtures of many chemical components and it is difficult to identify which specific substances might impact on the respiratory health of healthcare workers and increase their risk of asthma.3 The types of products, and the intensity and the frequency of exposure vary according to the cleaning or disinfecting task (eg, quaternary ammonium components to clean floors, glutaraldehyde to sterilise endoscopic materials).

Assessment of the exposure of healthcare workers to cleaning/disinfecting agents remains a challenge in occupational epidemiology.6 Occupational exposures, based on precise description of working tasks and the cleaning/disinfecting products used, as documented in specific questionnaires, should be more precise than estimates based only on job titles. However, as self-reported exposure may be affected by recall and misclassification bias,6 7 assessment of exposure to cleaning and disinfecting products by experts might be more reliable.

The aim of the present study was to estimate whether self-reported occupational exposure to cleaning and disinfecting agents in hospital workers is accurate compared to expert assessment of exposure, which is considered the gold standard.

Methods

Population

The French Epidemiological Study of the Genetics and Environment of Asthma (EGEA) survey is a case–control study including the family members of cases with asthma. The first survey (EGEA1) was conducted between 1991 and 1995 with 2047 participants; the protocol and the characteristics of the participants have been described previously.8 Approximately 12 years later (2003–2007) in a second survey (EGEA2), 92.2% of the initial population who were still alive, completed a short self-questionnaire and 77.1% answered the main questionnaire.9 Detailed information on occupational history and from specific questionnaires for hospital healthcare and cleaning workers was recorded for 1571 adults. Responses to specific occupational questionnaires on tasks and exposures for hospital and cleaning workers were available for 294 participants with 468 jobs (figure 1). An expert assessment (see detail below) of exposure to pollutants was carried out for 198 participants, working in health or biological units or as cleaners in hospitals, potentially exposed according to their 430 job history descriptions.

Figure 1

Selection of the studied population. EGEA, Epidemiological Study of the Genetics and Environment of Asthma.

The present analysis includes 176 adults with 327 jobs, with available estimates of exposure to pollutants from (i) the specific questionnaires completed during interviews with the workers and (ii) the expert assessment of exposure.

Written consent was obtained from all participants for both surveys. Ethics approval was obtained for both surveys from the relevant committees (Cochin Port-Royal Hospital, Paris, for the first survey (EGEA1) and Necker Enfants-Malades Hospital, Paris, for the second survey (EGEA2)).

Asthma status

In this analysis, a participant was classified as having asthma if he/she was recruited as an asthma patient from a chest clinic, or if he/she answered positively to one of two standardised questions: (1) Have you ever had attacks of breathlessness at rest with wheezing? or (2) Have you ever had asthma attacks?

Estimation of exposure

The EGEA2 survey included an occupational questionnaire, with a detailed description of the tasks for each job or apprentice period, with specific questionnaires for cleaners and hospital workers adapted from those of the European Community Respiratory Health Survey, which described tasks as well as cleaning and disinfecting agents used. Feasibility and validity studies were conducted before the final versions of the questionnaires were selected.10 Participants decided which jobs had similar tasks and an individual who had jobs with similar tasks and exposures completed only one specific questionnaire although information about other similar jobs was collected (see online supplement). If the individual had worked as a cleaner and as a healthcare worker he completed both specific questionnaires. Potential exposure to cleaning/disinfecting products was evaluated using the information in the questionnaires. Each job in the study (n=3746, figure 1) was classified, based on task description, by two experts (JF, NM) blinded to asthma status, as potentially exposed or non-exposed to cleaning or disinfecting tasks. Cleaners working outside hospitals have heterogeneous exposure according to their industry, and it was difficult to find experts with knowledge of these exposures. After discussion with experts, it was decided to conduct an expert assessment of exposure only for healthcare workers and hospital cleaners (198 out of 294 participants with specific questionnaires; figure 1). The expert assessment, blinded to asthma status, used a standardised procedure described below, and the opinions of three experts (CD, RM, MAD—an occupational hygienist and occupational physicians in the hospital) were recorded for these hospital workers. The three experts had worked in various hospital units for more than 10 years, and had a good knowledge of current and previous occupational exposures. A structured assessment form was used to document three types of cleaning/disinfecting tasks (disinfecting, cleaning and both tasks) and 21 specific cleaning/disinfecting agents or groupings of these agents. Disinfecting and cleaning agents were identified by reported chemical agents or brand names that allowed the experts to identify their chemical components. Expert assessments were performed by job. Decision rules were defined according to the experts' knowledge and the available data on product components,11 and the intensity (non-exposed, low, moderate, high, environmental), frequency (non-exposed, <1 day/month, <1 day/week, 1–3 days/week, ≥4 days/week) and probability (non-exposed, <50%, ≥50%, 100%) of exposure were estimated. Information on the dates of marketing and the components of each product reported by participants was obtained from a list of cleaning/disinfecting products used in hospitals (using all updates from 2000 to 2007).11 General decision rules for coding the level of exposures were written for all large job categories such as nurses, auxiliary nurses, cleaners and doctors, taking into account calendar year and hospital unit (emergency, surgery). Participants working in surgery, emergency and paediatric units were considered to have a higher probability of exposure than workers in outpatient or inpatient units. The intensity of exposure to alcohol varied according to hospital unit and was considered ‘high’ for nurses working in an operating room or an intensive care unit, ‘moderate’ for nurses working in an inpatient unit or doing blood tests and ‘low’ for nurses working outside the hospital. Cleaners were considered less exposed to patient-care disinfectants than nurses and more exposed to surface cleaning/disinfecting products.

Exposures were estimated for each job, taking into account the description of tasks and self-reported exposures. Exposures were evaluated by job category (nurses, cleaners) to provide better reproducibility, as underlined in previous papers.12 Before each meeting, each expert evaluated, independently, exposure for a given list of jobs (that differed for each expert) from the same job category (eg, decisions regarding exposure for cleaners were made on the same day). Each expert's assessment of exposure, for each of the 18 specific exposures and the three tasks (disinfection, cleaning and combined cleaning/disinfecting), was discussed and a final decision was taken by consensus by the three experts. All expert assessments were reviewed at the end to check the standardisation of the decisions taken.

Two estimates of exposure were available: (i) self-reported: a ‘never’ exposure (see questionnaire in the online supplement) was considered as ‘non-exposed’ and a self-reported exposure (‘4–7 days/week’, ‘1–3 days/week’ or ‘<1 day/week’) as ‘exposed’; (ii) expert assessment: only jobs classified by the experts with a high probability of exposure (probability ≥50%) to a given product were considered as ‘exposed’ and other jobs (non-exposed or probability <50%) as ‘non-exposed’, as previously discussed.13

Statistical analyses

Analyses are presented for the eight specific exposures (formaldehyde, glutaraldehyde, bleach/chlorine, alcohol, quaternary ammonium, sprays, ammonia and latex gloves) for which at least 10 participants reported an exposure. The present analysis was performed at the job level. Self-reported exposures were compared to expert assessments, that were taken as the gold standard, for the 327 jobs with both questionnaire and expert assessment available (figure 1).

Specificity and sensitivity were computed by specific exposures, by asthma status and according to the level of exposure. Differences between self-reported and expert assessments were tested by the McNemar test,14 and agreement was studied by calculating Cohen's kappa (chance-corrected) and phi (chance-independent) agreement coefficients. According to a recent paper,6 kappa estimates, which are classically used to quantify agreement, tend to underestimate the level of agreement when there are marked differences in the prevalence of exposure in a given cell. Phi estimates seems more adapted in this case. The strength of agreement for both kappa and phi coefficients was interpreted, as recommended, as follows: poor: <0; slight: 0–0.2; fair: 0.2–0.4; moderate: 0.4–0.6; substantial: 0.6–0.8; and almost perfect: 0.8–1.15

Analyses were stratified according to three categories of occupation (nurses, auxiliary nurses and cleaners, physicians), socioeconomic status (‘low’ such as cleaners and auxiliary nurses, laboratory technicians; ‘high’ such as nurses, physicians, researchers), level of exposure according to task, age, gender and asthma status. As job exposure estimates were not independent (on average each participant had two jobs), two further analyses were performed by taking into account only either the last or else the first job per participant. As participants could respond to only one specific questionnaire (as jobs with similar tasks could be grouped according to information supplied by participants), to avoid bias, analyses were also performed after exclusion of jobs with similar tasks. Analyses were also performed for all jobs performed in the last 5 years to limit the effect of memory bias.

Results

The studied population (healthcare workers and cleaners in hospitals, n=176) did not differ significantly from the population with only specific questionnaire information (n=118) or from EGEA2 participants with a job history but no specific questionnaire (1112 out of 1406; figure 1) for age, asthma status and smoking habits. Participants were aged 46 years on average, most were women (74%), 52% had never smoked and 39% had asthma (this high prevalence is explained by the EGEA protocol) (table 1). In their last occupation, 25% of the subjects were working as nurses, 21% as physicians, 11% as auxiliary nurses and 6% as cleaners.

Table 1

Description of the study population from the EGEA survey

Self-reported estimates and expert occupational exposure assessments were compared for all 327 described jobs (table 2). Few participants were exposed to glutaraldehyde according to either estimate (only nurses or auxiliary nurses specialising in cold sterilisation of medical materials were exposed), so results concerning this exposure are not reported in our analyses.

Table 2

Comparison of self-reported to expert exposure assessments* for all jobs (n=327)

The results show significant under-estimation of self-reported exposure compared to expert exposure assessment for all hazards except sprays and latex gloves. For latex gloves and spray exposures, self-reported exposures were over-estimated. High sensitivity was observed for bleach/chlorine, spray and latex gloves (87%, 95% and 99%, respectively) with substantial to high agreement coefficients (phi: 0.72–0.84). Phi and kappa values were very similar for all products except alcohol, ammonia and quaternary ammonium components where phi values were slightly higher than kappa values. Low to fair agreement in estimates was observed for alcohol and quaternary ammonium components (phi: 0.15 and 0.32, respectively). For the self-reported data, high rates of missing values (‘don't know’) were observed for formaldehyde (8.0%), glutaraldehyde (9.5%) and quaternary ammonium components (7.3%).

Analyses were stratified according to three categories of occupation: auxiliary nurses and cleaners (n=55), nurses (n=89) and physicians (n=96). A lower rate of self-reported missing values was observed in physicians. For auxiliary nurses and cleaners, poor to fair kappa and phi coefficients were observed for quaternary ammonium components, alcohol and ammonia (phi: −0.43, 0.27 and 0.38, respectively) and for nurses poor to fair kappa and phi coefficients for quaternary ammonium components, alcohol and formaldehyde (phi: 0.05, 0.25 and 0.37, respectively), with underestimations of self-reports, whereas for physicians, only alcohol exposure was underestimated (results not shown). The underestimation of exposure to quaternary ammonium components was higher in cleaners/auxiliary nurses than in nurses. Auxiliary nurses and cleaners were more exposed to ammonia and bleach/chlorine than nurses (expert assessment: 65.4% and 83.3% vs 2.6% and 51.1%, respectively).

Separate analyses in younger (<45 years) and older participants (≥45 years) are presented in online supplementary table E1. Younger participants were less often exposed to formaldehyde (14.9 vs 41.4%; expert assessment) and more often exposed to sprays (47.8 vs 29.4%) than older participants. A higher underestimation of exposure was observed in the younger compared to the older participants for alcohol, quaternary ammonium components and ammonia, whereas the expert assessments of exposure prevalences were similar irrespective of age.

Separate analyses in men (n=96 jobs) and women (n=231 jobs) generally showed higher agreement levels, lower prevalences of exposure (except for formaldehyde) and higher sensitivity in men than in women (results not shown). More than 50% of jobs in men and fewer than 20% in women were physicians. Strong underestimations of self-reported exposures were observed in women for formaldehyde, quaternary ammonium components, ammonia and alcohol and in men for alcohol and quaternary ammonium components.

When the analyses were restricted to jobs without similar tasks, phi and kappa coefficients were higher (except for ammonia) compared to the results from table 2, and significant underestimations of self-reported exposure were observed for formaldehyde, alcohol, quaternary ammonium and ammonia (table E2 in online supplement). Similar underestimations were also observed when analyses were restricted to the first job of each participant (n=176, data not shown). For two further analyses, when analyses were restricted to the last 176 jobs or the 136 jobs during the last 5 years, similar results were observed with significant strong underestimations of self-reported exposure to quaternary ammonium components, alcohol and ammonia, whereas no difference between expert and self-reported assessments were observed for formaldehyde (data not shown).

Analyses performed separately in participants with and without asthma, are described in table 3. Missing data rates were higher among participants without than with asthma for the self-reported exposures, especially for formaldehyde (11.3% vs 4.0%). The level of agreement between the two exposures estimates was similar or slightly stronger in participants with than without asthma, except for quaternary ammonium components. Large underestimations of self-reported exposures were observed in participants without and with asthma for quaternary ammonium components, ammonia and alcohol.

Table 3

Comparison of self-reported exposure to expert exposure assessments in participants without and with asthma

For jobs with exposure to cleaning/disinfecting tasks, stratified analyses according to the level of exposure showed higher phi and kappa coefficients for low than for high exposure jobs (table E3 in online supplement). There were more missing values for self-reported exposure in high exposure jobs compared to lower exposure jobs for exposure to formaldehyde (14.9% vs 5.9%) and quaternary ammonium components (12.6% vs 5.9%). Large underestimations of self-reported exposures were observed, especially for formaldehyde, quaternary ammonium components, alcohol and ammonia.

Stratified analyses according to socioeconomic status, showed lower phi and kappa coefficients and higher rates of missing values, in low versus high socioeconomic level. Larger underestimations of self-reported exposures were observed for quaternary ammonium and ammonia in low versus high socioeconomic level (data not shown).

Discussion

High occupational exposure to disinfecting and cleaning agents is very common in hospital workers. The present analysis of several hundred jobs demonstrates a significant underestimation of self-reported occupational exposure, especially for formaldehyde, ammonia, alcohol and quaternary ammonium components, compared to expert assessment. These findings were confirmed when the analyses were stratified on socioeconomic level, age, gender, asthma status, level of exposure or job category.

Underestimation of exposure to cleaning/disinfecting products

In the present study, healthcare workers, in particular nurses, auxiliary nurses and cleaners, had a high probability of being exposed to cleaning and disinfecting products. The common use of such products has been previously reported.3 4 In our study, self-reported exposure was underestimated compared to expert exposure assessment, especially for nurses, auxiliary nurses and cleaners. This underestimation of exposure was large for formaldehyde, ammonia and alcohol and hospital workers had little knowledge of the components in cleaning/disinfecting products used, especially for quaternary ammonium.

We tested the hypothesis that declaration biases are greater for jobs with similar tasks because the participants responded only once for two or more jobs with similar tasks. However, analyses performed after excluding jobs with similar tasks are consistent with an underestimation of exposure. Similar results were observed when analyses were performed for the first job only (one job per subject). Furthermore, analyses restricted to the last jobs and to jobs in the last 5 years, confirmed a large underestimation of self-reported exposure for ammonia, alcohol and quaternary ammonium components. No differences between self-reported exposure and expert exposure assessment were observed for formaldehyde, which is probably explained by a decrease in formaldehyde use in recent years. The present analyses suggest that the underestimation of self-reported exposure is not explained by declaration or memory bias. However, the hypothesis that declaration bias might explain this underestimation cannot be totally excluded.

The under-reporting is likely to be explained by a lack of knowledge of the components in cleaning/disinfecting products used by nurses, auxiliary nurses and cleaners. Agreement level seems to be affected by sociodemographic characteristics, as previously underlined in a study comparing self-report to job exposure matrix (JEM) exposure estimates.16 We have observed differences in agreement level according to age, gender and socioeconomic status. The higher underestimation observed in younger subjects might be explained by better knowledge of exposure in older workers or by a cohort-based difference in tasks or in the perception of exposures. The differences observed by gender may be explained by socioeconomic status (more men than women were physicians). Low socioeconomic status might partly explain the underestimation of exposure. Nurses seem to have a better knowledge of exposure to quaternary ammonium components than cleaners or auxiliary nurses. Physicians seem to have a better knowledge of the components of products than other hospital workers, although they underestimate exposure to alcohol and quaternary ammonium components. In hospitals, physicians participate in the development of cleaning/disinfecting guidelines, which may in part explain their better knowledge of product components.

Strategies need to be developed, with workplace interventions, to protect workers from cleaning/disinfecting exposures thus improving their health.3 4 It would be useful to limit the use of harmful or allergenic products and to provide safer disinfectants and cleaning products.2–4 In addition, more training of healthcare workers would improve workers' knowledge of exposure.

Participants with and without asthma

Our results are consistent with previously reported differential misclassification bias in self-reported exposure in work-related asthma or rhinitis studies.6 7 17 We observed an underestimation of self-reported exposure irrespective of asthma status, that was slightly lower in subjects with asthma. No differential misclassification related to asthma was reported for other environmental exposures, such as environmental tobacco smoke or exposure to pets, which may modify the clinical characteristics of subjects with asthma.18 19

Participants with asthma appear to have better knowledge of their occupational exposure or seem to pay more attention to products used than participants without asthma. Missing value rates were lower in participants with than without asthma and higher in the more exposed groups. Our results are consistent with previous reports6 7 which suggest higher sensitivity and lower specificity in self-reported exposure (compared to JEM estimates) by participants with asthma compared to participants without asthma.

For exposure to bleach/chlorine, a described risk factor for asthma,20 21 no significant differences in exposure estimates were observed in participants with asthma, which was not the case in participants without asthma. In hospitals, bleach, unlike other products, is often used undiluted, producing a strong smell. Odours are recognised triggers in asthma6 which may partly explain the greater perception of exposure to bleach/chlorine in participants with than without asthma.

Exposure assessment

Occupational exposures change with time and are different now to 20 years ago. The tasks of healthcare workers vary according to evolving medical techniques, substitution of products and new guidelines for healthcare hygiene. In studies of associations between exposures and respiratory health, exposure in most cases is estimated by self-report and, more recently, by JEMs.13 22 23 To date, two JEMs have been developed in healthcare workers24 25 to estimate exposures to large groups of pollutants and tasks but provide no specific information regarding the components of disinfecting/cleaning agents, or the probability and frequency of exposure. One of the strengths of our study is that self-reported exposures were compared to expert assessments rather than to JEM assessments. Case-by-case expert assessment is often considered to be the most accurate method for retrospective exposure assessment.26 Expert assessment might reduce misclassification bias and is considered to be more effective than the use of JEM estimates, as it takes into account all individual occupational information. Some authors emphasise the limitations of JEM estimates, such as the lack of variability in exposure within jobs, even if in theory differential misclassification is less likely to be present.6 27 The expert method is lengthy and expensive, especially regarding the search for the chemical components in products, which limits use of this method. It also depends on the competency of the experts who need a good knowledge of workplace exposure for all time periods under study. Some authors have highlighted the limitations of expert assessment, and have indicated that it is not necessarily reliable for all hazards.26 Mannetje et al showed high agreement for exposure between eight groups of experts when there was detailed information in the questionnaire, and low agreement for specific exposures that were difficult to directly estimate from tasks (eg, chromium dust). In our study, the available information was detailed, and included job, activity, precise description of tasks, start and end years of each job, and products used, which should increase the reliability of our expert assessments. This allowed, for example, a nurse working in a medical outpatient clinic in dermatology to be classified as less exposed to disinfectant than a nurse working in endoscopy, limiting the measurement bias. Standardisation of expert assessment is also an important issue.26 Our expert assessment was blinded to health status and by job category (nurses, cleaners) to provide better reproducibility, with standardised decision rules as suggested previously.12

The low agreement observed between self-reported and expert estimates suggests that it is necessary to get answers to all questions in the two specific questionnaires regarding tasks, products and exposures for each job with a positive response to targeted questions in the main questionnaires (similar to questions 18 and 15 from the specific questionnaires; see online supplement). Although the tasks may be similar, exposures may change according to cleaning/disinfecting guidelines. For example, formaldehyde (a risk factor for diseases other than asthma) has been replaced by quaternary ammonium components to clean surfaces. The experts have taken into account the job period to estimate exposure. We observed for most specific exposures, a high specificity and a more variable sensitivity according to hazards, these results being consistent with previous studies.26 As recommended, we used both kappa and phi to assess concordance and we observed, as expected, higher phi than kappa values when the McNemar test was significant, consistent with the fact that Cohen's kappa agreement underestimates the level of agreement when there are marked differences in exposure prevalence.6 However, irrespective of the agreement coefficients used, the conclusions were the same.

Specific exposures

As previously reported, the reliability of exposure estimates depends on the products.26 For exposures to formaldehyde, alcohol and quaternary ammonium components, our results regarding agreement coefficients were similar to those of Delclos et al.6 Joffe et al28 emphasised that lower sensitivities were observed for exposures described in chemically specific terms, which is consistent with our observations for ammonia, quaternary ammonium components and glutaraldehyde.

Self-reported exposure to quaternary ammonium components was always underestimated in men and women, irrespective of job category, age, socioeconomic level or asthma status. Lack of knowledge of the type of components in daily use as cleaning products in hospitals, underlines the difficulties in obtaining appropriate self-reported exposure estimates for these components. Open questions to collect information on brand name were useful in such instances (see the questionnaire in the online supplement). Sometimes the experts identified quaternary ammonium exposure only from the brand name, which may partly explain the under-reporting of the hospital workers. Similarly, alcohol is a component in many hand-cleaning products and it is possible that participants had forgotten to report exposure to these products, which may partly explain the under-reporting.

An overestimate of exposure to latex gloves and sprays was observed. Our hypothesis was that this overestimate was explained by jobs with similar tasks at different periods. The expert group was able to correct the exposure estimates by taking into account the marketing dates of each product. Analyses on jobs without similar tasks show no overestimate for latex gloves or for sprays.

Conclusion

Occupational exposure to disinfecting or cleaning agents is very common in hospitals. Workers do not know or underestimate their exposure compared to expert assessment. Our results underline the relevance of expert assessment in epidemiological studies to limit measurement bias. Occupational safety and health education programmes on the occupational risks of disinfecting and cleaning products, need to be developed. Healthcare workers require training to improve their knowledge of the toxic effects of cleaning and disinfecting products and should be instructed how to handle these products, so that they are protected. The use of personal protective equipment (gloves, face mask, glasses) must be adapted to each specific task of healthcare workers.

Acknowledgments

The authors thank all those who participated in the study and in the various examinations involved: interviewers, those involved in lung function tests, skin prick tests, blood sampling and IgE determination, coders, those involved in quality control, data and sample management and all those who supervised the study in all centres. The authors are grateful to the three Clinical Investigation Centres (Necker (Paris), Grenoble and Marseille) who supported the study and in which subjects were examined. They thank G Vasseur for job coding, Y Iwatsubo for expert assessment consulting (InVS) and B Balkau (Inserm). They are indebted to all individuals who participated without whom the study would not have been possible.

References

Supplementary materials

Footnotes

  • EGEA Cooperative Group: Coordination: F Kauffmann, F Demenais (genetics), I Pin (clinical aspects); Respiratory epidemiology: Inserm U 700, Paris: M Korobaeff (EGEA1), F Neukirch (EGEA1); Inserm 707, Paris: I Annesi-Maesano; Inserm CESP/U 1018, Villejuif: F Kauffmann, N Le Moual, R Nadif, MP Oryszczyn; Inserm U 823, Grenoble: V Siroux; Genetics: Inserm U 393, Paris: J Feingold; Inserm U 946, Paris: E Bouzigon, F Demenais, MH Dizier; CNG, Evry: I Gut, M Lathrop; Clinical centres: Grenoble: I Pin, C Pison; Lyon: D Ecochard (EGEA1), F Gormand, Y Pacheco; Marseille: D Charpin (EGEA1), D Vervloet; Montpellier: J Bousquet; Paris Cochin: A Lockhart (EGEA1), R Matran (now in Lille); Paris Necker: E Paty, P Scheinmann; Paris-Trousseau: A Grimfeld, J Just; Data and quality management: Inserm ex-U155 (EGEA1): J Hochez; Inserm CESP/U 1018, Villejuif: N Le Moual; Inserm ex-U780: C Ravault; Inserm ex-U794: N Chateigner; Grenoble: J Ferran.

  • Funding This study was funded by Merck Sharp & Dohme, the Hospital Program of Clinical Research (PHRC)-Paris, the National Research Agency (Health Environment, Health-Work Program) and the French Agency of Health Safety, Environment and Work (AFSSET, EST-09-15).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Cochin Port-Royal Hospital, Paris, for the initial survey and by Necker-Enfants Malades Hospital, Paris for the follow-up.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.