There are two main types of research. Physicians mainly use quantitative research methods. Researchers who are influenced by the humanistic approach, e.g. anthropologists, use qualitative research methods that deal with factors such as values, concerns and behaviour. This way of research, however, is not unfamiliar to physicians 15 and may also cover case studies. Both ways of conducting research may contribute significantly Quantitative research methods in maritime medicine – maritime epidemiology

The application of evidence-based practice to occupational health requires the application of epidemiological methods 16. Observational epidemiological studies may be descriptive or analytical.

 Descriptive epidemiology is used to describe in quantitative terms the main features of a collection of data such the disease pattern in a defined population. For describing situations, the statistical summary measures include incidence, prevalence, cumulative incidence, mortality rate (which may be age-standardized), or years of potential life or working ability lost. Incidence and prevalence are commonly used terms that refer to measurements of disease frequency. The incidence of a disease is the rate at which new cases occur in a population during a specified period. The prevalence of a disease is the proportion of a population that is affected by the disease at a specific time.

 Descriptive epidemiology is distinguished from analytical epidemiology in that it aims to quantitatively summarize a data set, rather than to support inferential statements about the population that the data are thought to represent. In some cases it may still be relevant to study incident descriptive data even if the source population is unknown. A multinational surveillance study of accident reporting in a number of countries has provided valuable results in spite of this constraint 17.

 Analytical epidemiology relates the disease pattern to factors such as life styles, treatments, vulnerability, or exposures. For example, the morbidity in an exposed population is related to the morbidity in a non-exposed population selected in a way so that the non-exposed population would be expected to have comparable morbidity if the exposure being studied neither prevents nor causes the disease being investigated. In the clinical controlled trial, one has to select controls out of those with the disorder for which a treatment is to be studied by randomly allocating subjects to groups with different treatment. This design is typically chosen in the study of the effectiveness of a medication but a similar design may be feasible with an intervention directed towards preventing the effect of a particular exposure, such as ship’s motion leading to seasickness. Seafarers at risk are allocated at random to the treatment and the pattern of seasickness is compared in those with and without the treatment. In contrast, if new clinical or experimental variables are to be measured in a risk group (e.g. nosographical or pathogenetic factors such as obesity or metabolic syndrome among seafarers, or subjects exposed to a certain potentially harmful factor), the controls cannot belong to that risk group but should be taken from the general population. In this case one would aim to make the two groups that are to be compared as much alike as possible except for the studied factor by matching them with regard to factors such as sex, age and social status.

Even when a data analysis draws its main conclusions using inductive statistical analysis, descriptive statistics are generally presented along with more formal analyses. For example a paper reporting on a study involving human subjects will typically contain a table giving the overall sample size, sample sizes in important subgroups (e.g. for each treatment or exposure group), and demographic or clinical characteristics such as the average age, the proportion of males and females, and the proportion of subjects with related co-morbidities.

In research involving comparisons between groups, the major emphasis is often placed on the significance level of the groups being compared, i.e. whether the two groups differ to a greater degree than would be expected by chance. This significance level is often represented as a p-value, or sometimes as the standard score of a test statistic. In contrast, an effect size is a descriptive statistic that conveys the estimated magnitude and direction of the difference between groups, without regard to whether the difference is statistically significant. Reporting significance levels without effect sizes is often criticized, since for large sample sizes even small effects of little practical importance can be highly statistically significant and for small samples large effects may not reach significance. Effect sizes such as the absolute risk reduction, the number needed to treat (or harm), the odds ratio or relative risk are better for demonstrating the severity of the problem or the ability of an intervention to reduce the risk.

The statistical calculations will not be reviewed here but the reader is referred to textbooks on medical statistics 8 ,9 or web resources such as Wikipedia 11.


Design of epidemiological studies

Based on the collected data that are to be compared, three main designs of observational studies are particularly relevant in occupational and maritime medicine. Each design has its strengths and limitations (Table 1).

 Cross-sectional studies form a class of research methods that involve observation of a whole population, or a representative subset, at a defined time. They may be used to describe some feature of the population, such as the prevalence of an illness. Cross-sectional studies may also identify correlations by, e.g. supporting inferences of cause and effect. Registration of data on exposure and health effects at the time of investigation permits comparison of the disease prevalence in those with and without the exposure of concern. Cross-sectional studies cannot determine the direction of a relationship, that is what is the cause and what is the effect.

Cross-sectional studies are simple and comparatively rapid to perform. They differ from case-control studies (see below) in that they aim to provide data on the entire population under study, whereas case-control studies typically include only individuals with a specific characteristic. Unlike case-control studies, cross-sectional studies can be used to describe absolute risks and not only relative risks. Cohort studies (see below) differ from both in making more than one observation on each member of the study population over a period of time.

 The cohort study (longitudinal studies, follow-up studies) permits the identification of changes such as the incidence of a certain disease over time and can show the relative and absolute influence of an exposure on the disease.

In prospective cohort studies, a population in which some are exposed and others are not, or where exposure levels differ, is defined and then followed-up with regard to disease outcome. The two groups should be comparable with regard to everything but the studied exposure and should usually be followed-up during a long observation time. The disease incidences in the various exposure groups are then compared.

These studies are easy to understand and design but may be difficult and costly in practice. For rare outcomes, cohort studies are not suitable because a sufficient number of cases for analysis would demand a very large sample. The relative risk and risk difference is expressed by the quotient of and difference between the cumulative disease incidence in the exposed vs. the non-exposed subjects, respectively. Examples of maritime cohorts can be found in Strand et al. 18 and Hansen et al. 19.

A retrospective cohort study, also called a historic cohort study collects data from past records looking back at events that have already taken place, for example with regard to health or exposure. The starting point of this study is the same as for all cohort studies. The first objective is still to establish two groups - exposed versus non-exposed; and these groups are followed up in the ensuing time period.

Registry studies are observational retrospective cohort studies in which data are collected from records on events, which happen to subjects with a specific disease or condition without predefined treatment. Registries may be public statistics such as mortality or disease statistics, notification records, etc.

 The case-control study has the advantage of being relatively cheap and rapid because it does not require a follow up period. It is also useful for the study of conditions that are infrequent because it is based on a design in which all cases (all with the disease one wants to study) are compared with a representative sample of the population from which the subjects with the disorder derive but who do not have the disorder studied.

The odds-ratio is the quotient of the proportion of the diseased to the healthy subjects in the exposed and non-exposed groups, respectively.

This design should not be used if the exposure to be investigated is rare or if there is a major drop out of control subjects.

The odds-ratio may be close to the relative risk for rare disorders, but it does not represent the relative risk and is not a true measure of association.

 Any of the above study methods uses the whole or a sample of a population at risk. If the sample is not representative, the results cannot be extrapolated to the rest of the population. Care must also be taken when extrapolating study results from one population to another e.g. from studies of European seafarers to Asian Strengths and weaknesses of various epidemiological designs

Each epidemiological design has strengths and weaknesses. The strengths relate to the relative feasibility in various settings and to the timeframe and available resources. The weaknesses are due to the relative influence of factors such as confounding and bias that may disturb or even destroy results if not taken into account (Table 1).


Study type



Cross sectional studies

Relatively quick;


Can study multiple outcomes;


The best way to determine prevalence.

Do not themselves differentiate between cause and effect or the sequence of events;

Healthy worker effect.

Cohort studies

Describe incidence or natural history;


Analyze predictors (risk factors) thereby enabling calculation of relative risk;


Measure events in temporal sequence thereby distinguishing causes from effects;


Complete data on cases, stages, exposures;


Can study multiple outcomes (effects of exposure);


Can calculate and compare rates and risks;


Choice of factors available for study;


Quality control of data;


Can accommodate "nested" case-control study.

Must study large numbers;


May take many years, even decades;


Circumstances may change during study;


Expensive in money;


Retrospective cohorts where available are cheaper and quicker;


Incomplete control of extraneous factors;


Rarely possible to study disease mechanism;


Confounding variables are the major problem in analyzing cohort studies;


Subject selection and loss to follow up is a major potential cause of bias.

Case-control studies

Good for rare diseases;


Requires relatively little time to conduct;


Possibility of exploring multiple exposures;


Retrospectively compares two groups;


Aim to identify predictors of an outcome;


Permit assessment of the influence of predictors on outcome via calculation of an odds ratio;


Useful for hypothesis generation.

Reliance on recall and/or historical data on exposure;


Comparability of cases and controls causes bias to be a major problem;


Can only look at one outcome;

Temporality can be difficult to establish;




Table 1. Advantages and disadvantages with various epidemiological study designs

Data collection, the importance of choosing the right measures and tests, and to perform them with accuracy


Data collection describes the process of preparing and collecting data as part of a research project. The pre-collection activity is a crucial step in the process. It is often discovered too late that the value of the data collected has to be discounted due to poor sampling. After the pre-collection activity is fully completed, data collection in the field can be carried out in a structured, systematic and scientific way. Data collection may be formalized through a data collection plan containing agreements on goals, target data, definitions and methods. Following data collection some form of sorting, analysis and/or presentation will take place.

 A formal data collection process will ensure that data gathered are both defined and accurate and that subsequent decisions based on arguments embodied in the findings are valid. The process provides both a baseline to measure from and, in certain cases, a means of identifying what to improve.

 The data collected may be of a biomedical character, or may represent exposures, either from direct or indirect estimates. Both can be measured or semi-quantified. The collected data should be valid and relevant to the studied issue.

 In most cases the conduction of a pilot study (a smaller version of a proposed research study) may serve to refine the methodology, and indicate if the approach is feasible and will provide the data needed. The pilot study should resemble the proposed study as much as possible, using similar subjects, the same setting, and the same techniques of data collection and analysis.

 Biomedical data include biological data from laboratories, e.g., biochemical or physiological data, data from physical examinations, records of symptoms, outcomes of physical examinations, records of diseases or deaths. Data may be derived from, e.g. questionnaires, interviews, census, sample surveys, or administrative registers. Each of these data sources has their respective advantages and disadvantages.

 For assessment of health effects, it is essential that the methods used, are able to reliably identify the subjects with the studied condition (sensitivity) and to distinguish them from those without the condition (specificity). Based on these data the predictive (positive and negative) ability of the test can be calculated. It is, however, important to realize that no test is perfect and that many commonly applied tests, e.g. many tests that are applied in the physical examinations, are really not very good for their purpose. It is also important to realize that the predictive value of a test cannot be uncritically transferred from one setting to another (see spectrum bias below).

 Questionnaires are a frequent source of data to which the same requirements as those for any other means of assessing disease, symptoms, level of functioning, well-being, or exposures apply. One special limitation for questionnaire data that should be mentioned here is that one only gets answers to the questions asked. Consequently, it is particularly important that the data that are to be collected by use of questionnaire should be scrutinized prior to the launch of the questionnaire.

 Exposure assessment has also many challenges. Exposure assessment is the process of estimating or measuring the magnitude, frequency and duration of exposure to a chemical agent or a physical or psychosocial factor, along with the number and characteristics of the population exposed. Ideally, it should describe the sources, pathways, routes, and the uncertainties that may occur in the assessment.

 Exposures may be assessed by, e.g. monitoring the pollutant concentrations reaching the respondents (the so-called direct approach). The concentrations are directly monitored on or within the person through personal sampling at point of contact, biological monitoring, or biomarkers. The point of contact approach indicates the concentration, or dose if measured over a period of time, reaching the host, while biological monitoring and the use of biomarkers infer the dosage of the pollutant through determination of the body burden. The respondents often record their daily activities and locations during the measurement of pollutants to identify the potential sources, microenvironments, or human activities contributing to exposure. An advantage of the direct approach is that the exposures through multiple media (air, soil, water, food, etc.) are accounted for through one study technique. The disadvantages include the invasive nature of the data collection and the associated costs. Evidently, this approach cannot be applied in retrospective studies.

The indirect approach measures the exposure in various locations or during specific human activities to predict the exposure distributions within a population. It focuses on the exposure within microenvironments or activities rather than the exposure directly reaching the respondents. The measured data are correlated to large-scale activity pattern data, to determine the predicted exposure. This exposure modeling determines the estimated exposure distributions within a population rather than the direct exposure an individual has experienced. The advantage is that the process is minimally invasive to the population and is associated with lower costs than the direct approach. It may be applied in retrospective studies if exposure patterns have been constant or can be reconstructed. A disadvantage is that the results have been determined independently of any actual exposure, so the exposure distribution is open to errors from any inaccuracies in the assumptions made during the study, the time-activity data, or the measured exposure data.

In general, direct methods tend to be more accurate but more costly in terms of resources and demands placed on the subject being measured. For both approaches it is crucial that measurements are accurate.

On occasions it is impossible to collect either direct or indirect exposure data and then less reliable alternatives such as job title and period working in area where exposure had occurred may need to be used.

Register-based studies

Public and private registers may provide information that can be used for setting priorities, identifying risky exposures, designing and implementing interventions and assessing their effect over time. These registers include national registers of seafarers, hospital admission registers, registers of notified occupational accidents and work-related disorders, registries of other notified conditions, e.g. infectious diseases or cancer, insurance based registers, and union member files.

 Analyses of maritime register data as described previously (retrospective cohort studies) are characterized by a number of challenges similar to those required for making valid risk assessments and comparisons in any epidemiological analysis. Comparison of register based data, e.g. from a maritime to a shore-based population, is complicated and potentially misleading due to the various ways in which the data are provided. Register-based studies require that both the denominator (e.g. the number of persons at risk/exposed persons/seafarers) and the numerator (e.g. the number of persons who suffered a disease, or an accident) should be known or at least estimated with some certainty from the available data for the time period of the study.

 There are two main problems with regard to the denominator:

  • Epidemiological studies of seafarers are methodologically difficult to undertake because the denominator may be unstable, e.g. with workers who are employed on casual terms and lack of a stable career. In addition, comparisons between maritime occupations and other industries should take into account the difference between the onshore working day and the pattern of working and living at sea with periods of leave. Risks may change over time and also vary in between and over time for the individual seafarers, who do not all work the same schedule.
  • Another challenge is the difficulties in characterizing the denominator because public and other registers may vary from country to country, over time and from setting to setting. For this reason, external, e.g. international, comparisons tend to be difficult to undertake and may even be impossible, in particular if the details of the composition of the registry are not described. At times even the scope of a national register may change considerable over time.

The best way to overcome these problems is to attempt conversion of denominator figures into working hours.

 The challenges with regard to the numerator are no less.

  • Registers may be (and mostly are) incomplete, and usually information needs to be found from multiple sources.
  • Large differences between registers complicate or preclude comparisons. For example offshore accidents are notified to the authorities after one day off work in some countries and after three days in others.
  • The identification of a case may be difficult or inaccurate. E.g., the diagnostic distribution of upper limb disorders in a register of notified work-related diseases is likely to be inaccurate since there is little consensus of the diagnostic criteria to be applied and because the majority of work-related conditions are currently not covered by diagnostic criteria.

 A publication about maritime work-related mortality illustrates all these difficulties 19, and how the investigators sought to overcome them by meticulous studies of multiple databases, and files from various sources. In this case the measure of effect (death) vas rather robust although occasionally the assessment of the mode of death did cause difficulties. If, however, the outcome is a disease many additional constraints may arise, e.g. with respect to the reliability of the applied diagnostic criteria.

 When analyzing, e.g. hospital based registers, one should be aware of the fact that there is a threshold for referring to hospital and that this threshold depends on the diagnosis, the severity of the disorder, and the diagnostic practices in different hospitals and countries 20 ,21. Register-based studies tend to ignore uncommon diseases or events because only an excess of common ones come out significant. This may to some extent be overcome by increasing the observation time, but by doing so the data that are already historical becomes even older. Thereby the challenge relating to the relevance of historical data is further increased, and no preventive initiative can be based on such studies. Historical exposures responsible for a certain health event may have changed or may have gone. For example, the high number in 1970-85 of alcohol-related deaths among Danish seafarers in a previous mortality study 22 is unlikely to represent the current risk as the pattern of alcohol consumption has drastically changed since then. Consequently, there are no preventive perspectives in this observation. Studies of hospitalizations may suffer from similar constraints. In addition, data from public registers, e.g. on hospitalization, cannot always be related to occupational exposures even with increased incidence of hospitalizations due to certain diseases. In studies of Danish seafarers and fishermen 20 ,21, for example, this increase was not related to the duration of employment.

Interventional research in maritime medicine

Interventional research is experimental of nature. In spite of this, many factors that may influence the outcome may be (and mostly are) out of the influence of the researcher. In theory, all potential risk factors are kept at a stable level except the one being studied. These ideal circumstances can usually not be reached in the practical situation in which the composition of the study group or the work conditions may change over time. Downsizing of the maritime population may have occurred due to reduced manning or economical constraints of the trade. The fishing options may have been altered by weather conditions. The route of the vessel or the nature of the transported goods could have changed. New regulations may have induced new standards for occupational health and safety that may have counteracted previous risks.

 Interventions may take the form of controlling a risk on board by technical means, by changing the administrative or managerial environment, by altering cultural issues or, e.g. by self-assessment of personal risk factors and taking action to address these. Involvement of those directly afflicted by the risk may be an effective way to shift control to or improve control by the individual, and thus empower seafarers to help themselves better. It does, however, require an employer-attitude that recognizes the nature of health risks, sees seafarers as an asset and not as a cost, accepts their involvement, and is prepared to invest in this view.

 Such a participatory approach leaves ownership with the involved seafarers or fishermen because they themselves have designed the intervention with assistance from researchers who can provide them with material and advice, validated methods for intervention, and – in the end – demonstrate whether the intervention has provided benefits to the maritime population studied. One example of this experimental design is a comparison between the influence of various navigation methods in simulators 23 on mental workload and performance.

Clinical research

Clinical research is a branch of medical science that determines the safety and effectiveness of medications, devices, diagnostic products and regimens intended for prevention, treatment, or diagnosis, or for relieving symptoms of a disease. Clinical research can well be applied in maritime medicine, e.g. in a trial of the efficacy of transdermal scopolamine for seasickness 24.

Confounding and bias

Confounding arises because health effects may have multiple determinants. E.g., it has been shown that the incidence of lung cancer is increased in the maritime environments and that this in particular is the case for the engineering crew 25. The challenge is whether exposure to carcinogens in the maritime setting, e.g. in the engine room, may explain this finding or whether a higher prevalence of smoking may influence the health of the maritime population and in particular that of the engineers. If this is the case, studies of the relation of lung cancer in engineers to exposures in the engine room should control for the excess smoking among this group. The concept requires that the confounder should be statistically associated with the exposure (that engineers smoke more than the comparison population), that it is an independent cause of the disease (smoking causes lung cancer), and that the confounder does not constitute a link between exposure and disease. One could also decide to study the role of smoking (or the exposure to passive smoking) to morbidity on board ships. In this case, smoking could be regarded as an occupational risk secondary to cultural factors or access to cheap cigarettes in the maritime setting. With this research question, exposures to carcinogenic substances in the engine room or from the cargo would be potential confounding factors to be considered.

 Bias may occur for several reasons such as because the information/data on which the study is based is wrong (information bias), or the results of the study are distorted due to the participants not being representative of the wider maritime population (selection bias).

 Information bias may come from an inaccurate assessment of the exposure or the effect, both of which are frequent sources of erroneous conclusions. E.g. in a study of upper limb disorders among fishermen, the physical examination may indicate only a limited number of disorders that the researcher decides to look for, or disorders that fulfill the defined diagnostic criteria. Consequently, some disorders, and possibly even the most prevalent ones may not be identified. Information bias relating to exposure may be caused by the larger propensity of those with a certain disease to remember exposures relevant to that disease than that of healthy colleagues. 

 Selection bias may occur if the persons studied do not represent the target group, e.g. if the motivation for participation is dependent of disease or exposure status. For instance, the response rate to a questionnaire on accidents may differ in between various groups of seafarers with those who have experienced an accident being more (or less) inclined to answer. Another example may be that noise-exposed seafarers would be more inclined to participate in a study on hearing impairment than non-exposed seafarers.

The “healthy worker effect” is a frequent source of selection bias in occupational and maritime medicine. Seafarers with certain diseases are more likely to leave the trade – in particular diseases that make it hard for them to manage their work or to pass the medical examination. Consequently, studies of an active maritime population may result in finding relatively less disease and an underestimate of risk.

 Spectrum bias may occur when the sensitivity and specificity of tests used differs between sub-groups of patients with different disease prevalence. Diagnostic test performances may be artificially overestimated in a case-control design where a healthy population is compared with a population with advanced disease ('sickest of the sick'). If properly analyzed, recognition of heterogeneity of subgroups can lead to insights about the test's performance in varying populations.

 Bias due to factors related to the researcher may occur from personal desires or views, from presuppositions that are wrong, or from selecting arguments in the discussion that favors certain viewpoints and tends to neglect others. Presuppositions may be based on, e.g. political viewpoints or selective sympathies and antipathies. Well-intended desires to benefit an exposed group of workers or to protect the employer from expenses are other examples of presuppositions that may lead to bias when these actions are not justified from the research.

 Publication bias may also be personal if the researcher prefers to publish positive findings and so is less likely to publish negative ones, but may also relate to the greater willingness of journals to publish positive rather than negative results.

 Differentiated misclassification of information or participation occurs with a flawed measure for instance of disease among the exposed that differs from the measure applied for the non-exposed. Differentiated misclassification may cause bias in both directions, and the effect of exposure may be consequently over- or underestimated. In contrast, non-differentiated misclassification, in which the measures are equally flawed in the two groups, tends to mainly lead to bias towards the hypothesis that there is no or less than the actual association between exposure and effect.

 An effective approach that can address some of the aspects of bias is to design the study in such a way that neither the persons studied (e.g. patient and control) nor the researcher knows to which group the being examined belongs. Single blinding refers to ignorance of the examiner about status. Double blinding requires that study subjects also do not know their status. With blinding, the risk of bias due to pre-existing views or expectations from the researcher as well as the study subjects can be reduced. Blinding may be rather easy in a randomized controlled trial in which some subjects are given the potentially active medication and others the non-active medication (placebo) and neither the researcher nor the person studied knows to which group each person belongs (double blind study). In a diagnostic trial, one must frequently merely accept blinding of the examiner (single blind study) because the study subjects may be aware of their disease-status. A single blind approach may also apply in cohort studies, if the study subjects know their disease status at follow-up. In occupational and maritime intervention studies, blinding of the examiner as well as those studied is often impossible because both are likely to know who are subjected to the intervention and who serves as controls. A ‘placebo’ intervention, where a sham intervention is introduced, is usually not feasible.

Qualitative research methods in maritime medicine

“Qualitative methods” in research have been applied in medicine and social sciences as well as in arts and humanities, but in different manners. Qualitative research does not try to imitate the requirement of natural science for universal truths and methodological rigor. Rather than simply attempting to determine, e.g. association, qualitative research is based on an interaction between theory and empirical data. By approaching a research question from multiple angles it can provide significant arguments that can contribute to better understanding of such relations and help to determine the interventions that are likely to be helpful. Qualitative research also differs from other research in the interplay between theory and method. The four tools that are represented by methodology, concepts, theories, and theoretical perspective cannot be separated and their interaction is the focus in a qualitative perspective.

Qualitative research therefore aims to constitute a relational, holistic and contextual approach with interaction between people as the key focus. The implication is that even though a particular dimension deals with, for instance, culture, this is not considered in isolation but in conjunction with other dimensions. The concept of "embeddedness" is often applied and has proved useful for studies in the maritime setting: Culture is embedded in social, political, economic, and other factors. This contextual approach means that empiricism often concerns the daily life on board a ship, and the fieldwork takes place at the micro level. Fieldwork requires drastic (thematic or geographical) boundaries, but the empirical data generated from fieldwork must be placed in a larger context.

Qualitative research can neither be a neutral science free of values nor can it represent a definitive truth. Regardless of status as a researcher or a researched subject, the ‘reality’ is always viewed from a given position and may be potentially disturbed by a set of biases or prejudices. The point is to include our positioning, our bias or our assumptions as parts of the tool for the analysis. Only by positioning, one can access the universe of others. Research outcomes always represent provisional truths that may still be under debate. Especially in applied research, one may be aware of the ‘truths’ and perspectives of various players. The role here is to embrace more contexts than the individual player is able to hold and to subsequently reflect on the findings with both players and academic colleagues. While the statements or observations can sometimes be judged as either true or false, it may well be more interesting to address the question of how and why the various observed behaviours and recorded statements have arisen.

 The prime methods in qualitative research consist of interviews, focus groups, participant observations and case studies. Participant observation has an obvious purpose, but an interview may also provide new answers, if following the statements of the respondent is emphasized rather than just following the interview guide. The questions asked determine the answers you get, but one of the strengths of qualitative research is that you are also likely to get answers to questions you have not asked. The interviewee takes the lead with the support of the interviewer. This has advantages on several levels: The interview promotes exchange, positioning and balance between guiding and laissez-faire, and at the same time, the person interviewed is acknowledged as an active, creative and interpretive co-worker. Getting unexpected answers is more likely when the interaction is open to all potentially sources of data-production. It is important to save strange, inexplicable or contradictory observations or statements. They may eventually turn out to be important keys to a more complex level of analysis.

 Standardization of qualitative methods is neither possible nor desirable. This does not mean that everything is permitted. In contrast, guidelines for good science are constantly refined (Table 2).


Head principle

Enabling conditions

Counter principle or impending/ precluding conditions

Awareness of perspectives

Make one's experience, position and assumptions explicit;


Consider pre-existing interpretations of the subject.

Readiness to revise one's position in the course of a survey.

Internal logic or consistency




Empirical anchorage;





Internal consistency;




Respondents’ validation;




Are questions answered in the conclusion?

There may be differences between what people say, what they do, and what they say they do;


The category true/false may be inappropriate;


The analysis should not necessarily refute or resolve paradoxes and contradictions;


Respondents, being differently positioned, can have different validation criteria. They may have good reason not to recognize themselves in a report that is critical of them;


Questions always determine answers. Finding answers to non-addressed questions may be more fruitful.


Would others come to the same conclusion?


Can the results be transferred to other settings?

Awareness of perspectives: no two researchers are evenly positioned;


The question of transferability must be answered by the reader.

Ethical values

Personal integrity;


Social responsibility;


Scientific honesty;


Protecting anonymity.


Pragmatic values and social responsibility may counteract personal integrity;


Protecting anonymity may counteract scientific honesty.

Aesthetic values

Clarity of structure;


Richness of meanings;


Reinstating the author;


Rhetorical persuasion.

Respect for complexity and ambiguity.


Richness of meanings limits clarity and vice versa.

Pragmatic values



Assess the consequences;


Emancipating value.

Relevance to whom?


One can never foresee all the ways a survey may be used;


Should one conceal some results if they may have bad consequences?

Table 2. Quality in qualitative research: Enabling and impending conditions 26 


Qualitative researchers should be conscious and explicit about their perspectives and articulate the basis for their interpretation in terms of the empirical data and assumptions, e.g. the preconceptions from their own relevant personal experiences that may have influenced the theoretical approach. These are obviously difficult to define, and the researcher is not always conscious about them although their theoretical approach is well defined. Such reflection on personal influences is a part of the social context that the researcher needs to describe.

 The researcher would always aim for internal logic and consistency. Empirical anchorage can be strengthened through triangulation by viewing the issue from various angles and by using multiple data or methods. While, however, triangulation cannot verify convergence (consensus) between sources, differences may serve as important means of detecting contradictions such as when two parties have two different versions of an event. The same applies to paradoxes. A major strength of qualitative research is that it can take account of and deal with contradictions and paradoxes.

 A key issue is whether the players accept the outcome of an analysis. Their validation could indicate empirical anchorage, but may also lead to an unwanted introspection. This conflict of interests may be so large that the full approval by one sort of players indicates an evidence of imbalance! The researcher is aiming for internal consistency meaning that "each piece must fit into the puzzle". However, the pieces that do not match may well be the most interesting ones. It is preferable to leave several optional interpretations open rather than ignoring elements that disturb the clear picture. In spite of having answered the questions posed, new issues may emerge and others may prove less relevant than anticipated. It is important to consider whether others have reached similar results. However, this should not necessarily be expected if their approaches have been very different. Checking the results and discussions with others, however, are always fruitful.

Explanatory power and transferability reflects good science, whether it is qualitative or quantitative. However, the qualitative researcher should not determine the external validity (whether results can be generalized to other contexts), but must ensure that the reader is given sufficient information to judge whether and to what extent something may be transferred to another context. This requires a respect for complexity and ambiguity. New angles enabling the reader to view concepts or contexts in a new way may be relevant not only by suggesting options for action, but because it lets the players view their practices from new angles.

 Three examples illustrate how qualitative research in the maritime setting may contribute to cognition:

  • In a study of shipboard illness behaviour, Bloor has shown that accommodation to symptoms was overlain by the economic imperative to keep the ship functioning. While management feared that the right to the sick role would allow malingering, the workers feared that adaption to the sick role would exclude them from employment 27.
  • The imminent challenges that arise with regard to working at sea and psychosocial health problems have been recently reviewed after an IMHA workshop 28. Many of these can only be approached by qualitative research methods.
  • The job tasks and exposures to occupational hazards of fishermen have been mapped by collecting and analyzing qualitative data 29 ,30.


This brief review aims to give the reader some understanding of qualitative research and how it uses its own methods and paradigms and how it may contribute to improved health, safety and wellbeing of seafarers