Four US fisheries in Alaska, Oregon, and Massachusetts were sampled. These regions were selected to maximize demographic variation and fishery-sector representation. Salmon fishermen were recruited from Southeast Alaska. This commercial fishery population harvests fish using purse seines, trolling gear, and drift gill nets.²² Crews range in size from three to six members, trips range from 1 to 3 days at sea (DAS), and workdays range from 14 to 20 hours.²³ In Oregon, we recruited participants from the Dungeness crab fishery. In this fishery, crews harvest their catch with crab pots. Crew size varies from two to four crew members with DAS ranging from 1 to 7, and workdays typically ranging from 12 to 20 hours in length.^24–26 Scallop fishermen from Massachusetts who harvest using dredges were also recruited. Crew sizes for this fishery are typically 6 to 10 crew members with DAS between 10 and 14, although vessels are now limited to roughly 20 DAS for the entire year.^27,28 The average workday length is around 12 h.^29,30

In Massachusetts, most inshore lobster fishermen typically leave and return to port after a 6–10-h day and fish with lobster traps. Crews are typically one to two members, while some fishing vessel captains fish without a crew. This fishery was included because fishing trips of multiple days do not typically occur. These lobstermen, therefore, served as a comparator arm for the other three fisheries who typically work overnight on their vessels.

Human Subjects Research approval was received from the Mary Imogene Bassett Hospital Institutional Review Board under protocol #1481898. All recruited fishermen signed a consent form to participate in the research and discussed the study process and their rights as research subjects with the study team.

The formation of the research team emphasized a transdisciplinary approach and included an epidemiologist and statistician who have worked together for over 15 years with lobster fishermen in the Northeast developing and evaluating worker safety interventions for this industry. A second investigator who specializes in occupational safety and ergonomics was also included, along with a former fisherman and international marine safety instructor. Finally, a pulmonologist with training in sleep disorders partnered with a mechanical engineer to provide expertise on tracking sleep duration at sea.

The sampling frame for the salmon, Dungeness crab, and lobster fisheries consisted of a list of permit holders within each of these three fisheries. The original sampling plan was to distribute these surveys completely at random to individual permit holders using a combination of email, post, and telephone (provided a telephone number was available). In cases where the permit holder did not provide his/her telephone number, the Alaska Marine Safety and Education Association (AMSEA) trainees list was cross-checked to gather this information, or internet searches were conducted to locate phone numbers. As planning progressed, it became clear the wide geographic dispersal of the fishermen had to be taken into account. Therefore, the randomization scheme was revised to focus on randomly selecting fishermen in specific locations, typically ports that contained relatively high concentrations of fishermen (cluster sampling). Fishermen were randomly selected in these ports to participate in the survey and health assessments. This allowed healthcare workers to travel and administer the in-person health assessment.

Because an all-inclusive permit list (0 both crew and captains) did not exist for the scallop fishermen, they were recruited by contacting scallop settlement houses. These settlement houses provide administrative assistance to both captains and crew members who specifically harvest scallops. Services include processing pay checks, handling mail, and providing industry updates. Some settlement houses agreed to distribute surveys by putting a paper copy in each scalloper’s mailbox. Officials at the settlement houses also informed the scallopers of the study by posting information signs near fishermen’s mailboxes, as well as verbally describing the study. Fishermen could either complete the paper survey and return it or call a toll-free number to complete the survey with a study team member.

As explained above, the research design called for completing health assessments with all subjects who had completed the survey. However, an adequate sample size for the health assessments could not be recruited using only these subjects because not every subject completing the survey was able to or had agreed to completing a health assessment. Therefore, it was necessary to solicit additional health assessment subjects using radio, social media, partner promotions, word of mouth, and dockside engagement with fishermen. Fishermen expressing an interest via these recruitment methods were sent information on the study by mail or email and followed up with via phone calls, texts, or email. One example of media outreach that was particularly helpful in increasing interest in the study was an article in National Fisherman Magazine.³¹

The survey instrument captured self-reported data on demographics, vessel type and size, work role, sleep schedules, self-reported general health, alcohol and tobacco use, caffeine intake, information on diet, and sleep (including factors that affect sleep), as well as physical, emotional, and financial health. Data on prior fishing-related injuries were also captured. Anyone who completed the survey was then invited to complete a thorough occupational health assessment. Survey data gathered by telephone were entered directly into the REDCap database system. In cases where survey data were captured during dockside recruitment, the data were entered on paper for subsequent data entry in REDCap.

In the survey, participants were asked to share their least total hours of sleep, average consecutive hours of sleep, and least consecutive hours of sleep per 24-h period. They were also asked to provide the number of consecutive days with the least hours of sleep.

There were two questions that specifically addressed the extent to which pain interfered with sleep:

The second of these two questions specifically addressed the relationship between pain and sleep while fishing. Due to the fact that lobster fishermen do not sleep on their boats (with rare exceptions), this question did not form the basis for a contrast among all four fisheries. For this reason, the more general question relating to sleep in the past 12 months was used for pain and sleep comparisons between fisheries. A more specific analysis that compares factors that impact sleep while fishing between the three remaining fisheries (salmon, crab, and scallop) will be reported in a separate manuscript.

Health assessments included a survey where fishermen self-reported health data and a physical examination where health endpoints were directly measured. A pre-health assessment questionnaire was also administered just prior to the examination and captured information on medical history and recent injuries. In the initial assessment, an occupational technician measured height, weight, and waist circumference (with body mass index also calculated), Snellen visual acuity, peak flow respiratory measurement, exhaled carbon monoxide (CO), blood pressure, random glucose, cholesterol, Framingham scores,³² urinalysis, the short blessed test (which measures early cognitive impairment),³³ post-traumatic stress disorder screener,³⁴ and the Epworth Sleepiness Scale.³⁵ Following this, an occupational physician or nurse practitioner conducted a full physical examination that covered Mallampati scores (to predict the likelihood of sleep apnea),³⁶ ENT (ears, nose, and throat), skin, neurological function, musculoskeletal issues, and hand-grip strength. Fishermen were also asked in the health assessment questionnaire “Do any of the following medical conditions or treatments apply to you?” with “arthritis”, “back pain”, or “chronic pain” being listed as potential responses. Responses were grouped into “yes” or “no” based on whether they had pain in the past or currently (categorized as “yes”) or whether they had never had pain (categorized as “no”).

All health assessment data, including both self-report and physical examination data, were recorded on paper copies and then entered into the REDCap database system.

The first set of the univariate analyses compared levels of variables thought to be relevant for their relationship to pain and sleep between the four fisheries. Categorical variables were compared between the fisheries using chi-square tests or Fisher’s Exact Test in cases where the data were too sparse for chi-square. Continuous variables were compared between the fisheries using a 1 by 4 Kruskal–Wallis non-parametric analysis of variance. In the second step, any variable whose levels differed significantly between the fisheries, were then tested for a significant relationship to the subjects’ self-reported relationship between pain and sleep. Categorical variables were tested for their relationship to binary sleep outcomes using chi-square or Fishers Exact Test. For sleep variables with more than two levels the relationship was tested using Spearman’s Rank-Order Correlation Coefficients (Spearman’s Rho). Continuous variables were compared between levels of binary sleep variables using the Wilcoxon Rank–Sum Test and were tested for their relationship to sleep variables with more than two levels using Spearman’s Rho.

The variables tested included gender, self-reported health, smoking, nicotine use, coffee, tea, energy drinks, diet, financial issues, work schedule, work relationships, boat noise, stress, family relationships, regulations, crew reliability, weather, total cholesterol, high-density lipoprotein, Mallampati score, blood pressure, body mass index, and post-traumatic stress disorder. Any variable whose levels differed significantly between the fisheries and that was also related to the subjects’ self-reported association between pain and sleep were entered into a five-level, ordinal logistic regression equation to identify independent predictors of this outcome.

Out of 577 attempted contacts, the study obtained 185 completed surveys. To calculate a response rate, 75 cases were removed from the denominator for the following reasons: “no number listed, n = 7”, “number disconnected, n = 8”, “number not in service, n = 33”, and “wrong number, n = 27”. This yielded a survey response rate, for the initial cluster sampling, of 185/502 = 36.9%. As explained above, the study originally intended to provide physical examinations to every subject who completed the survey. However, only 85 such examinations were completed for those original 185 surveyed individuals. This was attributable to the difficulty subjects experienced attending the physical examinations. Given the rural nature of the work, subjects often would be required to fly to the examination locations or were not available given the fishing season. Therefore, to ensure a sufficient sample size for physical examinations, 77 subjects were recruited (51 dockside, 18 through local promotions, and 8 from unknown sources). In addition to attending the physical examination, all 77 subjects also completed a survey, which resulted in a total 262 completed surveys and 162 completed physical examinations. Given the challenges of tracking promotional and dockside recruiting, it was difficult to calculate an exact numerical response rate for the physical examinations; however, few among those who were invited to participate in an examination via these dockside recruitment efforts refused.

Significant differences were seen between the fisheries for the least total hours of sleep (p < .0001) and the least consecutive hours of sleep (p = .001) (See Table 1). The main contributor to these two significant differences was the considerably limited hours of sleep on both endpoints observed in the Dungeness crab fishery.

It is noteworthy that the majority of fishermen in all four fisheries indicated back pain, with a prevalence ranging from 55.6% to 71.2%. As shown in Table 2, there was a significant difference in the prevalence of arthritis in these fisheries with the highest level being among the lobstermen. However, after adjusting for age, fishery was no longer a significant predictor in any of the three pain endpoints.

Fishermen were also asked “To what extent has significant pain and discomfort interfered with your sleep in the last 12 months?” Responses were rated on a 5-point Likert scale as follows: 1 (not at all), 2 (slightly), 3 (moderately), 4 (quite a bit), and 5 (extremely). Although the average rating was similar across fisheries, the average response to this question among all fisheries was 2.3. Further, 102 (39.8%) of the respondents checked 3, 4, or 5 indicating the considerable impact that pain has on their ability to sleep.

Univariate analyses were conducted to identify predictors of the extent to which pain interfered with sleep, which was measured as described above. Among the candidate variables considered were the subject’s: 1) work schedules, 2) work relationships, 3) boat noise, 4) pain, 5) stress, 6) family/relationship concerns, 7) financial concerns, 8) regulations, 9) reliability of crew/captain, 10) weather, and 11) safety. All these 11 variables were measured on a scale of 1 to 10, with 1 being the least, and 10 being the most. Additional tests were done to measure the relationship of this variable to the number of years the subject had been fishing and to whether or not they had experienced a work-related injury in the last 12 months. Finally, another variable measuring the subject’s ability to control the important things in life (measured as a 5-point variable) was also included in the analyses.

The results of these univariate analyses identified five variables whose levels differed significantly between the fisheries and were related to the response about pain and sleep. These five variables were captured in responses to the following questions:

These five variables were included, along with the subjects’ fishery, in a five-level ordinal logistic regression model to predict the extent to which pain interfered with sleep. The significant predictors were the inability to control important things in life (OR = 1.44, 95% CI 1.16–1.79, p = .001), having a work-related fishing injury in the last 12 months (OR = 2.12, 1.22–3.69, p = .008) and years in the fishing industry (OR = 1.02, 1.01–1.04, p = .013). Also, marginally significant in this model was the degree to which weather interfered with sleep (OR = 1.10, 1.00–1.20, p = .056). When the model was adjusted for this covariate set, the subject’s fishery was not a significant predictor of this outcome.

As noted in the methods section, researchers used a combination of random recruitment, recruiting through promotional materials, and dockside recruiting. Given these various methods, it is impossible to say whether the results gathered in these surveys and health assessments are entirely representative of the general population of fishermen. Complex weighting schemes were considered in combining the results obtained from the various recruiting methods, but they were found to be infeasible. Additionally, survey responses were self-reported, which raises the possibility of subject bias and incomplete recollection of events, a study limitation associated with many survey results. Despite these limitations, it is worth noting that the data captured in this study appear to be the most comprehensive set of health and survey data on fisheries sleep patterns and health outcomes in the U.S. and most importantly demonstrate that sleep studies in this population are indeed possible. Additionally, data were collected to assess commercially available hardware for use in sleep monitoring at sea during this study (publication in progress).