separating apples and oranges (can you find the face?)
In the world of veterinary medicine clinical outcomes are often evaluated using sometimes difficult to classify responses. For example, gauging the level of reduced pruritis (itching) as a positive effect can not really be exactly categorized. You can’t just ask the animal; how much less do you itch? Even if you could, the answer would still be hard to pin down objectively. This also occurs in human medicine.
On the other hand, it is possible to glean meaningful information from these types of qualitative phenomena. Loosely grouping observations into general patterns of response and pooling them into data bases -though not numerically exact- can certainly provide researchers and clinicians with important information.
The crux of the matter here is to correctly interpret this type of information using the right analytical tools. This is just what this interesting review looked at. Though focused on how methods of analysis have been utilized in veterinary dermatology the authors discovered an erroneous pattern consistent with similar reviews in human medicine (i.e.; anesthesia, rheumatology, and nursing specialties).
Different data sets often require different kinds of statistical methods because what is being measured isn’t always the same type of thing. You can not use formulas designed for numerical scales (where there is a continuous spectrum of connected numbers) for evaluating groups of separated discontinuous numbers. For example, the authors mention data groupings (gathered information) like “mild, “chronic”, and “severe” as common separated categorical scales where the wrong kind of statistical analysis was used leading to inappropriate conclusions which can have detrimental clinical ramifications.
They noted that of a total of 62 peered reviewed articles that used categorical scales 57.5% of them used the wrong type of analytical methods. They add “The frequency of inappropriate presentation and analysis methods of ordered categorical data in the veterinary dermatology literature is similar to that reported for several fields in the human medical literature. In order to reduce the likelihood of making unwarranted implications or conclusions regarding ordinal data, authors should follow established guidelines for methods of presentation and analysis of ordered categorical scales.” In other words, researchers need to use orange methods to study “oranges” and apple methods for “apples”- you can’t switch them around.
The authors have a nice description of the differences between the types of gathered information used for analysis. Understanding these criteria reveals where a lot of studies go wrong and misinterpret the significance of their information.
Data types can be classified into four increasing levels of quality:
1) Nominal data- Represent the lowest quality and simply describe the nature of the data. For example they could be "breed, color, or even diagnosis". They are not numerical values and can not be ranked one over the other.
2) Ordinal data- This is a little higher quality data set, can be incorporated into groups and “can be ranked in mutually exclusive hierarchical categories.” For example, findings can be scaled into mild, moderate and severe but they are still separate and discontinuous observations- “the interval or distance between the groups is unknown.”
The next two represent true numerical and continuous scales and can be evaluated differently:
3) Interval scales- are characterized by having an arbitrary zero value (i.e.; body temperature).
4) Ratio scales- have a meaningful zero reference point (i.e.; length).
The authors go on to describe the importance of using the right kind of methodological analysis for the specific type of data available. Even lower quality data can offer revealing information if the correct tools are implemented.
They also state that “Well-designed studies with proper statistical and analysis methods are essential given the prominence of evidence-based medicine in veterinary dermatology. Because of the extensive use of ordinal scales, researchers, reviewers, editors, and readers of the veterinary dermatology literature should be familiar with statistical properties governing ordinal data.” This goes for other areas and disciplines in medicine beyond veterinary dermatology.
One interesting and sad finding noted by the authors is that “upon reviewing the curricula of 29 North American veterinary colleges, we found that only six incorporate a course in biostatistics, potentially leaving many veterinarians poorly prepared to critically evaluate statistical methodology.” In addition to more statistical training they suggest researchers make better use of statisticians and that care should be taken when “emulating previously published material.”
The importance of having the skills and tools to properly analyze information is increasingly critical and indispensable in today’s medicine. These reviews offer important insight into the challenges of correctly interpreting data.
These skills are as important for general clinicians as they are for researchers and offer up real opportunities to separate what is real from what isn’t. It’s like Sagan’s “candle in the dark” analogy -used so many times before- and familiar to the scientist in all of us. Yet, even so, never really gets old.
No comments:
Post a Comment