Many patients with diabetes have complications leading to reduced visual acuity, and this requires that the print size of self-administered questionnaires is larger than that used in other populations, to ensure that misreading biases are not introduced. Likewise, other sources of bias should be considered and eliminated, such as with tools designed to assess quality of life, where a relative or friend of the patient may fill in the form for them.
The prevalence of some forms of diabetes is high in certain ethnic groups; language difficulties can make recruitment and consent difficult in some populations. It will be important to consider cultural differences when translating patient information sheets and consent forms, and when dealing with individual patients. For instance, in some UK minority ethnic groups it is not appropriate to ask a young man to interpret for his mother regarding questions of an intimate nature.
Many of the biochemical measures in diabetes are not normally distributed. For instance, plasma insulin and triglyceride commonly follow a logarithmic distribution, while fasting plasma glucose at diagnosis has a lower limit by definition. To enable parametric statistical tests to be used, transformation of the data (usually by taking logarithms) may be needed. Geometric mean and the associated one standard deviation range should be reported; for variables where such a transformation is not possible, the median and 25th and 75th centiles provide measures of centrality and dispersion. Care should be taken to ensure that statistical tests are only used on data which conform to the underlying assumptions of those tests, and that suitable checks of distributions are made, before embarking on analysis. Particular care should be taken with data that consist of counts or ordered categorical scores.
In many cases, the reporting of family history of diabetes causes difficulties in interpretation. In order to assess the strength of these data it is important to know how many relatives a subject has in an age group by which diabetes might have been diagnosed when they die, and this could include the number of siblings. Consulting an expert geneticist is important when considering this kind of information.
Was this article helpful?