Editorial: Methodological triangulation in public health research – advancement or mirage?

Public health researchers and practitioners alike are well aware of the complexity and multifaceted reality of their subject matter. It has therefore become de rigueur to call for an interdisciplinary research approach or, putting the idea into practice, to use more than one research method when studying a particular problem. The attached, albeit frequently tacit, expectations are high: researchers believe that the use of different methods will yield a more valid, or at least more complete, picture of their research subject. Where one method fails, the other will stand in. Social scientists have used this approach for decades; they call it methodological triangulation (for other types of triangulation see Denzin 1970). Regrettably, one rarely finds a systematic and critical reflection on the prerequisites and problems of a multiple-method approach in public health papers; see for example the conspicuously brief treatment of this topic in the BMJ series on qualitative research methods ( Pope & Mays 1995). It seems to be taken for granted that the use of multiple methods will inevitably enhance both the usefulness and the validity of results. Why then do authors so frequently fail to report on what basis they select their methods and synthesize the final results? Or even reject the findings of one method posthoc? Does a combination of methods ensure a better understanding of public health problems, or is it a methodological mirage? There is wide agreement how triangulation should not be used: it cannot cover up weaknesses or flaws in research design. A study in which one research method has been applied shoddily cannot be salvaged by simply adding a second method. Regarding its merits, methodological triangulation is seen as an opportunity to enrich research findings and deepen insight; in addition, findings from one method can presumably be validated by another method. While all this seems intuitively appealing, two practical problems arise that warrant further examination: how to deal with contradictory findings, and how to decide what methods can be combined. Our own experience suggests that an intuitive approach towards methodological triangulation can work: When studying the performance of an immunization programme in Zimbabwe, we used focus group interviews to help interpret the results of a coverage survey and to identify the underlying reasons for observed service problems ( Razum 1993). The results of the focus groups pointed in the same direction as the coverage survey data. Thus, the findings derived from one method seemed to validate those from the other. But what if the findings had been contradictory? Which method should then have been given preference? A frequent ‘solution’ is to favour one method over the other, effectively rejecting discrepant findings. What an untidy situation: After having meticulously followed a carefully considered study protocol, researchers use their common sense to decide which method they should trust more. What does the use of common sense lead to? Most likely to favouring the expected over the unexpected, thus wasting an opportunity to make startling discoveries or discard well-loved but outdated ideas. Favouring one method over another is not necessarily wrong. A study may aim to validate a new or simpler method of data collection against a ‘gold standard’, e.g. a questionnaire on smoking behaviour vs. urine cotinine measurements. Researchers implicitly accept the superiority of an established method over a new one and will reject the new one in case of contradictory findings. This validation of one method against another, however, is different from methodological triangulation. In methodological triangulation, the implicit assumption is that the methods used are of equal epistemological value and internal validity. Triangulation is thus most useful in situations where contradicting findings are welcome as an enrichment and not perceived as a flaw. The primary aim is deepened insight, not formal validation. The question then is: how different can methods be so that their combination will still be enriching? Feyerabend (1984) argues that different and possibly incommensurate epistemological approaches may nevertheless be internally consistent and useful to explain the reality in which they were contrived. Few researchers would go as far as proposing to triangulate epidemiological methods with astrology, i.e. two methods that follow different paradigms. Triangulating methods that follow the same – scientific – paradigm, for example epidemiological and anthropological methods, should be rewarding though ( Trostle & Sommerfeld 1996). How about slightly more adventurous combinations of methods, e.g. from the fields of science and art? Rorty (1991) suggests that more relevant detail about human affairs can be learned from the European novel than from approaches towards knowing reality that strive for purity of method. He, like Feyerabend, holds that artists strive to reflect reality as hard as scientists or philosophers do, though by different means. Oliver Twist ( Charles Dickens 1837–39) indeed provides rich information on underlying social causes of high childhood mortality in 19th century England – far more detailed than what could be learned in epidemiological studies, and certainly sufficient to support public health interventions. The familiar argument against using pieces of art as scientific evidence is that an artist's view is ‘subjective’; researchers, by implication, would be ‘objective’. Rosaldo (1993) believes that this subjectivity is actually an advantage. He argues that an individual, real-life phenomenon such as an illness can only be fully understood from a subjective position. When a researcher restricts study to ‘measurable’ disease he or she may be able to achieve general results. Many particularistic patterns, however, will appear unanalysable and thus may not even be included in an ongoing research agenda. If we follow Rorty and Rosaldo, a triangulation of epidemiological data and subjective observation would actually help avoid a superficial understanding of the reality of disease. Even if we refuse to go that far – methodological triangulation does help to learn more about reality. It can generate new ideas, raise interesting questions, provide new insights and point out inconsistencies – provided one accepts rather than rejects discrepant findings. Researchers striving for noncontradictory results may, for reasons of economic as well as scientific soundness, prefer to select just one research method at the design stage rather than discard incongruent results derived from a second method at the analysis stage. To fully appreciate the potential of methodological triangulation, it would be prudent to further advance both our theoretical understanding and empirical experience regarding methods which can be usefully triangulated. Clearly, the more similar the methods are, the less likely it is that the findings will disagree; but gains in insight will diminish simultaneously. Accepting bold combinations of methods, e.g. supplementing epidemiological data with subjective experience in a narrative form, poses a challenge to researchers as well as reviewers and readers of journals. Yet it may provide useful and relevant (though sometimes conflicting) in-depth information and make research more interesting.