The Nurses’ Health Study II (NHSII)[16,17] is an ongoing prospective study which began in 1989 when 116,430 registered female nurses aged 25–44 years, from 15 U.S. states, completed a mailed questionnaire on their medical history, lifestyle characteristics and nursing job types. Every two years, follow-up questionnaires were sent to update information on potential risk factors, identify newly diagnosed diseases and nursing jobs. Since 2009, nurses have been asked two questions on the frequency of instrument disinfection and surface cleaning tasks.[6]

In 2014, we initiated a nested case-control study on asthma. In order to evaluate exposure to specific disinfectants among nurses, an occupational questionnaire was sent to nurses with and without asthma. In the current study to design JEMs and JTEMs, to avoid differential misclassification of exposure, we selected at random a sample of 12,280 non-asthmatic nurses (Figure E1) out of those who never reported asthma (from 1989 to 2011; n=94,758) and who were still in a nursing job at the time of the 2011 follow-up questionnaire. Out of 12,280 non-asthmatic nurses, 10,189 were selected among all types of nursing jobs and an additional sample of 2,091 nurses (enriched sample) was selected among operating room, emergency room and intensive care unit nurses in order to enrich the sample of less frequent nursing jobs with expected high exposure levels.

The NHSII study and the current investigation were approved by the Institutional Review Board at the Brigham and Women’s Hospital (Boston, Massachusetts, USA).

The 2014 occupational questionnaire was adapted to the U.S. context from a questionnaire used in the European Community Respiratory Health Survey (ECRHS)[18] and the Epidemiological Study on the Genetics and Environment of Asthma, Bronchial Hyper-responsiveness and Atopy (EGEA).[1]

Data on current nursing job was collected in the 2014 occupational questionnaire by the question “Which best describes your current employment status?”, with 8 categories provided: nursing in the emergency room (ER), operating room (OR), intensive care unit (ICU), other inpatient nurse, outpatient or community, other hospital nursing, nursing outside hospital and nursing education or administration (see question 1 on figure E2, online supplement). Two questions (4 and 5, Figure E2) regarding the frequency (days/week) of the main disinfecting tasks performed at work were also included: ≪ Thinking about your current job and the use of disinfectants (such as ethylene oxide, hydrogen peroxide, ortho-phthalaldehyde, formaldehyde, glutaraldehyde and bleach): (a) On how many days per week, on average, do you clean medical instruments with disinfectants? (b) On how many days per week, on average, do you clean surfaces (like floors, tables) at work with disinfectants? (never, <1 day/week, 1–3 days/week, 4–7 days/week) ≫. These tasks were chosen based on results from the Texas Health Care Workers study,[6] which suggested that they were the most relevant in terms of asthma risk.

Finally, questions were asked about the frequency of use (“On how many days per week do you use the following disinfectants at work?”; see questions 14a to 14p, online supplement, Figure E2) of 14 specific disinfectants (e.g., glutaraldehyde, bleach, quats). Participants could fill in the brand name of the products they use if they did not know the active compound. We searched the safety data sheets of all provided brand names to determine the products’ main active compounds. Self-reported exposure to each specific disinfectant was evaluated using crude report and this additional information, and was considered in most analyses as binary variables according to weekly use (1–3 or 4–7 days/week versus never or <1 day/week).

Current job and use of disinfectant were used to design JEMs and JTEMs by 3 methods based on weekly use of products (yes/no), frequency (days/week) and/or intensity (hours/day) of exposure (Table E1 and figure E2, online supplement). To create JTEMs, the two questions on disinfecting tasks (to clean medical instruments/to clean surfaces) were combined to create a 3-category variable, to define tasks performed weekly: no weekly disinfection tasks; weekly use of disinfectants to clean surfaces only; and weekly use of disinfectants to clean at least medical instruments (regardless the use of disinfectants to clean surfaces). The category “clean instruments only” was not studied separately because of the low number of participants in this category (2.9%); it was thus grouped with the category “clean instruments and surfaces” into the larger category “clean at least instruments”.

The first method to generate the JEM and JTEM was based on the percentage of participants reporting exposure to a given disinfectant in a given nursing job (JEM) or for a given nursing job and task category (JTEM).[19,20] The job axis of the JEM included the 8 types of nursing jobs (OR, ER, ICU …). The “job-task” axis of the JTEM included the 24 possible combinations of 8 types of nursing jobs by 3 categories of cleaning tasks (surfaces only, at least instruments, none). For the exposure axis of the JEM and JTEM, the 14 disinfectants and the general use of sprays were considered. However, only 7 disinfectants for which at least 10% of the nurses in at least one nursing job reported weekly exposure (alcohol, hypochlorite bleach, peroxide bleach, glutaraldehyde, quaternary ammonium compounds, enzymatic cleaners, formaldehyde), and sprays, were retained. The other 7 disinfectants (acetic acid, ammonia, chloramine T, ethylene oxide, ortho-phthalaldehyde, peracetic acid, phenolics) were grouped together as “other” (with requirement of exposure to at least one of these agents) since less than 10% of the women reported exposure, regardless of the nursing job.

The second method was based on a score (ranged 0 to 6) combining frequency (0: <1; 1: 1–3; 2: 4–7 days/week) and intensity (1: <1; 2: 1–4; 3: >4 hours/day) of exposure, evaluated by the following questions “On how many days per week do you use the following disinfectants at work?” for frequency and “On days with disinfectant use, how many hours, on average, do you use disinfectants” for intensity. The score was calculated for each nurse with the assumption that intensity was the same for all disinfectants used. The third method was based on a weighted score according to the frequency of exposure to specific chemicals using the percentage of self-reported exposure weighted by 2 (1–3 days/week) and 5 (4–7 days/week). The second and third methods for both the JEM and JTEM are detailed online (Table E1).

Specific cut-offs were defined to classify exposure in “low”, “medium” and “high” levels for each disinfectant and each method developed (Table E2). Cuts-offs were chosen according to the distribution of the exposure prevalence (e.g., % nurses reporting weekly exposure; Figure 1) over the 24 categories defined by job types and cleaning tasks (i.e., 8 job types by 3 tasks) for each disinfectant. For a given method, the same cut-offs were used for the JEM and the JTEM. The first quartile (Q1) and median were used to define cut-offs for low and high exposure, respectively. For some chemicals, Q1 and the median were very low and minimum cut-offs were defined as follows: for the first method, we choose the maximum value between the median and 10% to define the cut-off for high level exposure; and the maximal value between Q1 and 5% to define the cut-off for low level exposure. Indeed, we considered that classifying a job as “highly” exposed to a specific disinfectant was not realistic if less than 10% of nurses in this job reported being exposed. Similarly, classifying a job as “moderately” exposed to a specific disinfectant was not realistic if less than 5% of nurses in this job reported being exposed.

Exposures evaluated by the JEM were compared to those evaluated by the JTEM, and both were compared to self-reported exposure alone. Specificity and sensitivity were computed for each exposure considering the JTEM as the reference, according to our a priori hypothesis. Both Cohen’s Kappa (chance-corrected) and Phi (chance-independent) coefficients were calculated to evaluate agreement between JEM and JTEM, as previously suggested.[10] To interpret strength of the agreement, standard cut-offs (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) were used.[21] Differences between JEM and JTEM assessment were tested by the McNemar test.

In addition, sensitivity analyses were performed by stratifying analyses on age (49–54; 55–59; ≥60 years).

All analyses were performed with SAS 9.3 (Cary, NC).

Out of 12,280 nurses invited, 11,134 (91%) completed the 2014 occupational questionnaire. After excluding 2,061 women (2,057 not in a nursing job in 2014; 4 declined study), the study population included 9,073 non-asthmatic women (figure E1).

The nurses were on average 59 years of age (standard deviation, 4; range, 49 to 68), with slightly younger nurses in the enriched sample (Table 1). In the random sample, 3% of the nurses worked in emergency room, 6% in operating room and 5% in intensive care unit. Most nurses reported working in outpatient or community (23%) and in nursing education or administration (16%). Fifty-four percent of nurses reported that they cleaned surfaces with disinfectants weekly, while 21% cleaned instruments with disinfectants weekly. Fifty-five percent of nurses reported using at least one of the 14 specific disinfectants weekly, 11% used them at least 1 hour per day and 20% performed only administrative tasks (Table 1). Across nursing jobs, the percentage of nurses using disinfectants 4–7 days/week ranged from 5% to 48%, while weekly use ranged from 19% to 88% (Table E3). Nurses in the emergency room, operating room and intensive care unit more often used disinfectants weekly to clean instruments or surfaces (>80%), as compared to other nursing jobs.

The most frequently reported disinfectants were alcohol (weekly use: 39%), hypochlorite bleach (22%) and sprays (20%) (Figure 2). These agents were followed by quats (14%), peroxide bleach (9%), glutaraldehyde (7%), formaldehyde (5%) and enzymatic cleaners (4%).

The distribution of self-reported occupational exposure varied widely according to disinfectants (Figure 1). Therefore, product-specific cut-offs were defined as described in Methods and table E1. For example, for alcohol, Q1 was 30.0 and median was 49.5; a nursing job in which less than 30% of the nurses reported weekly exposure was thus classified as “low exposure”, between 30% and 49.9% as “medium exposure” and more than 50% as “high exposure” (Table E2). For other disinfectants, exposure levels were similarly assigned based on product-specific Q1 and median. For glutaraldehyde, enzymatic cleaners, peroxide bleach and formaldehyde, values of Q1 and median were very low (less than 5% and 10%, respectively; Figure 1), and minimum cut-offs were used.

The JEM and JTEM design strategies are presented in Table 2 (examples) and Table E4 (full matrices). Self-reported exposures varied considerably according to nursing jobs and tasks. Among emergency room nurses, 38% reported weekly use of hypochlorite bleach and were classified as highly exposed by the JEM (Table 2). Operating room nurses reported less use of hypochlorite bleach (24%) than emergency room nurses and were classified with medium exposure by the JEM. Exposures within a job also varied according to disinfecting tasks. For example, operating room nurses were assigned a low exposure level to hypochlorite bleach when they performed no cleaning tasks (weekly use reported by 9% of the nurses); those who cleaned only surfaces (weekly use: 24%) were assigned a medium exposure level to hypochlorite bleach; and nurses who cleaned at least instruments (weekly use: 36%) were assigned a high exposure level.

More nurses were classified as exposed overall (i.e., medium or high exposure) by the JEM (alcohol 84%, hypochlorite bleach 84%, sprays 84%) and JTEM (62%, 62%, and 59%, respectively) than by self-report. In addition, more nurses were classified highly exposed by JTEM (41%, 34%, and 41% for alcohol, hypochlorite bleach and sprays, respectively) than by the JEM (21%, 26%, and 30%, respectively; Figure 2).

JEM and JTEM estimates of exposure were heterogeneous for most nursing jobs and disinfectants, except for nurses working in the emergency room and in education or administration (e.g. 89% and 81% classified similarly by the JEM and the JTEM for hypochlorite bleach, respectively; Table E5). For exposure to formaldehyde, JEM and JTEM estimates were similar.

Comparing high versus medium/low exposure, more nurses were classified highly exposed by the JTEM than by the JEM whereas the opposite was observed comparing high/medium versus low exposure (Table 3). For alcohol, for example, 21% of the nurses were classified as high exposure with the JEM, versus 41% with the JTEM; however, 84% were classified as high/medium exposure with the JEM and 62% with the JTEM. For most disinfectants, except enzymatic cleaners and formaldehyde, the JTEM classified more nurses in the low and high categories whereas the JEM classified more nurses in the intermediate category (Figure 2). Agreement between the JEM and JTEM was fair-to-moderate (Kappa coefficient: 0.3–0.5) for all disinfectants except for formaldehyde (0.8). Phi values were slightly higher than kappa values for all disinfectants.

Using methods 2 and 3 to design the JEM and JTEM, exposure assessments were mostly similar to those observed with method 1. We observed discordance between the three methods for peroxide bleach, glutaraldehyde and formaldehyde (2/0/1) for some nursing jobs (Table E6).

Self-report, JEM and JTEM exposure assessment were stratified according to three age categories (49–54/55–59/≥ 60 years). Older nurses were less often classified highly exposed than younger nurses by the JEM and the JTEM, consistently with self-report assessments (Table E7).

Exposure assessment is a crucial step to obtain reliable results when studying associations with the disease.[13] Several methods were developed to assess occupational exposure to cleaning and disinfecting products but none is optimal. The expert method, which is often considered to be the most accurate method for retrospective exposure assessment, takes into account individual occupational information (tasks, specific exposure).[1,22] However, this method is lengthy, expensive, depends on the competency of the expert, and is not practical for large epidemiologic studies; moreover, this method is not reliable for all hazards.[1]

Self-reported exposure is a simple method that allows variations in exposure within job titles[11] and is easily applied in large epidemiologic studies. However, reporting or recall bias might be present and lead to differential misclassification.[11] In the present study, results showed lower prevalence of most self-report exposures compared to JEM and JTEM exposure assessments. Exposure is often under-estimated by healthcare workers in other studies, possibly because some workers do not know the components of the cleaning and disinfecting products they use. In the study by Donnay et al, investigators observed an under-estimation of self-report compared to expert exposure assessment for all hazards.[1]

In respiratory epidemiology, few JEMs have been designed to evaluate occupational risk factors for asthma[23,24] or chronic obstructive pulmonary disease.[19,25] An asthma-specific JEM assessing exposure to asthmagens (known risk factors for occupational asthma), including exposure to non-specific disinfectants[23] has been widely used.[26,27] Two JEMs have been developed in healthcare workers to estimate exposure to a large group of agents[28] and tasks[6] in France and in the U.S., but do not provide specific information regarding the components of cleaning or disinfecting agents.

Quantitative exposure estimates, such as exposure of volatile organic compounds (VOCs), may provide a more accurate characterisation of exposure,[29] but there are also limitations. First, detection limits prevents determination of precise measurements.[30,31] Moreover, for some agents such as quats, atmospheric measurements are difficult due to the low volatility of the agent.[2] Finally, duration of exposure cannot be taken into account, and other sources of VOCs such as building materials can induce errors in measurement.[32] To date only one study with measurement data in U.S. healthcare workers is available.[29] In this study, personal VOC exposures varied among occupations, but different nursing jobs were not distinguished.[29]

We believe that we are the first to develop a nurse-specific JTEM, to evaluate occupational exposure to specific disinfectants while taking into account the variability of exposure within a given nursing job according to the disinfecting tasks performed.

In previous epidemiologic studies in healthcare workers, registered nurses were considered as a single job.[10,33] The importance to consider tasks in a job has been suggested in the literature.[12,14,15] Taking into account tasks, Droste el al, found significant associations between lung cancer and occupational exposure to carcinogens evaluated through a JTEM, but not with self-report.[15] Our results suggest that the JTEM is more accurate than the JEM to evaluate exposures for most nursing jobs, except for emergency room and education/administration nursing in which exposure seems more homogeneous within the job.

Using the JTEM rather than a JEM reduces the loss of information due to the grouping of individual data.[34] In the current study, compared to the JTEM, the JEM estimates lacked sensitivity for high exposure level, and lacked specificity for medium exposure level. Both lack of specificity and lack of sensitivity lead to important bias toward the null when evaluating associations with health outcomes.[35] For future work, the JTEM will be applied to study the association between occupational exposure to disinfectants and asthma in the NHSII cohort.

Cleaning and disinfecting products are complex mixtures of many chemicals components that can cause or exacerbate asthma because of their irritant or sensitizer properties[3] by mechanism still not well understood. In our study, we designed JEM and JTEM to evaluate cleaning products and disinfectant considered as irritants (e.g., bleach, ammonia) and sensitizers (e.g., quats, glutaraldehyde).[36,37] Exposure estimates for each nursing job and task are consistent with the typical use of specific disinfectants in hospitals. For instance, quats are commonly used as non-critical surface (e.g., floors, furniture) disinfectants but also for disinfection of medical equipment that contacts intact skin;[38] glutaraldehyde is used especially for high-level disinfecting of medical equipment such as endoscopes.[38]

In the current study, use of formaldehyde at work was mainly reported by operating room nurses, and all of them were assigned the highest exposure level by the JTEM regardless of the cleaning tasks. A possible explanation is that operating room nurses use formaldehyde for specific tasks like biopsy and not for disinfecting tasks.[38] In addition, given the low proportion of nurses reporting using formaldehyde, we had to use a minimum cut-off value of 10% for “high” exposure and 5% for “medium” exposure. As formaldehyde exposure is likely heterogeneous, this cut-off may poorly discriminate high exposures to formaldehyde. For example, it is unlikely that operating room nurses without disinfecting tasks have high-level exposure (see Table 2). Formaldehyde has been classified as a human carcinogen by the International agency for research on cancer (IARC) and a probable carcinogen by the U.S. Environmental Protection Agency (EPA). Accordingly, its use has probably decreased in hospitals as limited contact with formaldehyde has been recommended.[38]

In a study conducted in 5 U.S. hospitals, cleaning and disinfecting tasks (at least once per shift), were frequent among registered nurses (66%),[9] consistent with our results. Among U.S. hospital workers[6], self-reported exposures to cleaning and disinfecting products during instrument cleaning (42%) and building surface cleaning (78%) were higher than in the present study. In a French study of hospital workers,[1] 15% reported exposure to formaldehyde, 39% to bleach, 64% to alcohol, 14% to quats, 7% to ammonia and 39% to sprays, which is higher than in the current study for all disinfectants, except for quats. However, the study population included cleaners in hospitals, who may have higher exposure levels than nurses. Interestingly, the ranking (most to least frequently used) of the disinfectants was the same as in our study. In another French study, much higher exposure levels were observed (e.g., 98% of registered nurses reported occupational exposure to quats).[2] However, exposure was defined as reported use of products at least once a month whereas in the present study, weekly exposure was considered. Weekly use of disinfectant or cleaning products has been associated with asthma whereas few studies underline the impact of sporadic exposure (except for a high peak of exposure).

Strengths of our study are its large sample size (n=9,073) and high participation rate (91%). As shown by Delclos et al for cleaning and disinfecting products,[10] it is helpful to construct a JEM blinded to health outcomes to avoid differential misclassification of exposure. Accordingly, we designed the JEM and JTEM using a sample of nurses without asthma to assess occupational exposure independently of the disease. Moreover, we designed literature the JEM and JTEM through a detailed occupational questionnaire providing specific information on cleaning or disinfecting agents, and on instrument and surface cleaning tasks, previously shown to be relevant in terms of asthma risk.[6] We have further collected data on the brand names of the products used by the nurses, and re-evaluated individual exposures to specific chemicals using information from safety data sheets. We used a standardized method to determine product-specific cut-off for high, medium and low exposure levels. In addition, we designed the JTEM using three different methods, based on weekly exposure only or further considering exposure intensity and frequency. The method based on intensity required more assumptions (e.g., that intensity was the same for all disinfectants used), and its accuracy may be further limited by the relatively low proportion of participants reporting to be exposed more than one hour per day. However, for all three methods used to design the JTEM, close exposure assessments were observed for most specific disinfectants (which was less the case for JEMs), supporting the validity of the JTEM estimates. Finally, using the proposed JTEM to evaluate exposure to many disinfectants among nurses is not costly as it is only based on 3 simple questions (type of nursing job and two general cleaning tasks). This tool is thus of particular interest for applications to epidemiological studies of large populations.

The JEM and JTEM we developed also have limitations. First, for some disinfectants (acetic acid, ammonia, chloramine T, ethylene oxide, ortho-phthalaldehyde, peracetic acid, phenolics) the exposure assessment was not available due to the low exposure prevalence (<10%) in all nursing jobs.[39] However, nurses’ exposure to these chemicals is likely to be limited or passive. For example, peracetic acid is used in automated machines for instrument sterilization. Ammonia and phenolics, sometimes used to clean environmental surfaces (e.g., bedside tables, bedrails, floors),[3,38] may be used more often by other workers (cleaners, technicians).[9] Furthermore, the JEM and JTEM were developed in a population of U.S. registered nurses, and the reproducibility of these methods in other populations requires further study. Finally, we could not formally validate the JEM and JTEM estimates because of a lack of gold standard, which is a classical limitation of JEMs.[40]

In conclusion, cleaning and disinfecting tasks, which involved the use of various potentially asthmagenic products, are frequent among registered nurses. Occupational exposure among nurses varied widely according to both type of nursing job and tasks. We designed a nurse-specific Job-Task Exposure Matrix that allows investigators to take into account the variability of exposure within a given job. Creating reliable tools to evaluate occupational exposure to specific agents is crucial to quantify their adverse health effects and further develop optimal strategies to prevent occupational asthma. Going forward, we plan to apply the JEM and JTEM to study the association between occupational exposure to disinfectants and asthma in the whole NHSII cohort.