Designing an effective method for data collection is essential for without data no research can be fulfilled. Still, very few statistics textbooks pay attention to data collection methods and issues (Bauer, 2009). Therefore, it is advised that healthcare decision-makers must be well acquainted with the art and science of data collection methods available in statistics. As medical research is performed on a person or a group of people, so information regarding each member is essential for the research and needs to be recorded. This data may be obtained from direct measurement (e.g. weight on scales), or asking questions or by observing, from results of a diagnostic test (e.g. diagnosis of coronary heart disease) or other methods. Sometimes the unit of observation or research may not be humans, in that case, similar information has to be extracted through direct data collection methods i.e. by observation or direct measurement. Other researches may be dealt with financial performance for instance comparison of two hospitals in terms of efficiency and cost. Here the unit of observation is hospitals, and information, in this case, will be gathered from the hospital financial database. Therefore, an effective data collection method and certainty as to the nature of data required for the research is quintessential for any research process and is no exception to healthcare.
What is data? Assuming a study of medical students is being conducted, where demographic data have to be gathered like age, sex, city of birth, socio-economic background, etc. Each of these demographic information i.e. Age or Sex is considered variables while the information obtained in these areas from each student is called data.
Broadly, data can be divided into two categories viz. qualitative and quantitative data. Qualitative data refers to qualities and quantitative data to quantities. Qualitative data are non-numeric data. For example, name, sex, or city of birth is qualitative data. Further, some variables may appear to be quantitative but essentially, they are not. For instance, data variables related to the socio-economic background of the participant like age, income, occupation, education, etc. This is because these numeric variables are used to depict the socio-economic variable and are not used for quantitative analysis. They are a “tag or label and socio-economic group cannot be considered to be a quantitative data” (Daly & Bourke, 2000, p. 2).
Quantitative variables, also known as categorical or nominal variables are used to collect data of quantitative nature and used for statistical interpretation. When quantitative data are gathered in two categories like alive/dead or hypersensitive/normotensive, they are called binary which means a “dichotomous or an attribute variable” (Daly & Bourke, 2000, p. 2). Any variable, which has an intrinsic quantitative meaning, is quantitative. These data also have an intrinsic numerical meaning. These data are called metric or numeric variables. They arise from actual measurement (e.g. Age) or count (e.g. Number of siblings).
Quantitative data can be continuous and discrete. A discrete (quantitative) variable is one whose “values vary by finite steps” for instance variable mike ‘number of siblings’ or ‘number of children in a family who takes an integral value (Daly & Bourke, 2000). Continuous variables are those, which may take any value. Thus, the data may or may not be an absolute numeric form rather can take any number between two absolute numbers. Examples of continuous variables are weight, age, time, body temperature, etc. Practically, continuous variables are measured in discreet units, and data is usually collected nearest to the closest unit. For instance, weight may be measured to the nearest kilogram or height to the nearest centimeter.
Now the question that needs to be answered is how these different kinds of data are measured. In the case of qualitative data, basic data are used. This is done to avoid the cumbersome process of dealing with a long list of qualitative responses. Here the basic rule is to count the number of occurrences of each qualitative data in a category, which are then presented in form of frequencies. This allows the presentation of the data in a compact form. However, a problem is encountered when presenting the percentages in decimal places. Traditionally one decimal place is considered sufficient to round off the percentage to its nearest value (Daly & Bourke, 2000). A second problem that arises is when the rounded-off percentages do not equal 100%.
Qualitative data may be analyzed using statistical diagrams like bar charts or pie charts. Bar charts are usually used when one axis represents qualitative data against a qualitative variable like measuring the representing major reasons for endoscopy vis-à-vis frequency of occurrence. Usually, bar charts are constructed to show frequency, relative frequency, etc. Pie charts or diagrams are used to demonstrate qualitative data. Essentially the pie represents the total area of the frequency observed in each category. However, pie is also useful, but car charts are preferred in representing qualitative data.
Now considering quantitative data, we must understand how the data collected are shown or a data overview is presented. The scales that can be used for representing quantitative data are nominal, ordinal, ratio, interval scales, etc. The first known method is through presenting a frequency table. Now here the method is the same as in the case of qualitative data representation, but here categories must be created to represent the values in variable groups. These groups are formed by making class limits and class intervals. Data may also be represented using histograms or frequency polygons. A histogram is a bar chart presented for quantitative data. They provide a good picture regarding the frequency distribution and the shape of the distribution. Another method of presenting frequency distribution is through frequency polygons. This is drawn by joining the mid-points at the top of each bar using straight lines. As a drawing, these diagrams can become tedious at a time for a large number of data. In such a case, a stem-and-leaf diagram, dot-plots, or cumulative frequency polygon are used.
Given these methods of data collection and data representation in healthcare researchers, it is now important to understand why a researcher should take the trouble of conducting such a complicated process. The reasons are as presented by Cook, Netuveli, and Sheikh:
- “Data collection and processing are important steps that need to be considered in detail before embarking on a research project.
- The careful selection of an appropriate statistical package will pay a dividend in long run.” (2004, p. 78)
Further, there are ethical issues that need to be considered while representing the data. Here a researcher must segregate his/her personal views from the data and present unbiased data in the data presented.
Reference
Bauer, J. C. (2009). Statistical Analysis for Decision Makers in Healthcare (2nd Eds.). New York: CRC Press.
Cook, A., Netuveli, G., & Sheikh, A. (2004). Basic skills in statistics: a guide for healthcare professionals. London: Class Publishing Ltd.
Daly, L. E., & Bourke, G. J. (2000). Interpretation and uses of medical statistics. Malden, MA: Wiley-Blackwell.