What are they making?:
It is important to be clear about the focus of the research. This should be presented in the abstract as either a research question, or for more experimental research a hypothesis. This is confusingly usually presented as how the planned intervention will not make a difference – referred to as the ‘null hypothesis’ – and the planned research activity aims to demonstrate that the difference can be achieved and thus, disprove the hypothesis posed.
Who is going to be involved as subjects/population?
It is important to be given information about the people who will be involved, whether they have volunteered or been chosen because they have a specific characteristic of interest. This is one of the most usual places for bias to enter the research study as factors such as age, social status, educational level and ethnicity can all have an influence on whether people accept or decline participation in research. In order to reduce the risk of this, the randomised controlled trial identifies a total potential sample, for example, all women having their first baby in a London district. A specific system is used to select at random, the women who will be in the group testing the intervention (eg birth in a midwife led unit) and the women in the other group who will receive the current care (eg birth in a consultant unit); this group is called the ‘control group’. When collecting information about all the women in the study, it can then be verified that the random allocation has worked if the main characteristics of age, marital status, ethnicity, education status, are statistically similar in both groups. If there are any other key characteristics that are important in relation to the research topic, then again, the two groups should be similar if randomisation has been effective.
Where the study is not based on randomisation of the sample, any differences between groups being compared might have an effect on the overall results. An example of this is comparisons between women who have given birth before and primigravid women. The experience of giving birth is likely to make differences in many social and health aspects which means that if comparisons are made between groups with unequal representation from one or other group, in some circumstances the results could then be unreliable.
This refers to the use of questionnaires, interviews, data collection sheets and other research tools used to obtain information from the subjects. This covers a vast area of the research process but whatever tool or approach is used, it needs to be something that allows the respondent to freely give information about themselves, or information obtained about clinical care that would be recorded by an observer. This information and its link to the research subject is treated as confidential so that only those involved in the research have access to it .
This relates to the equipment and how it will be used. So, for example, having decided to use a semi structured questionnaire which will be completed by the respondents, this needs to be developed to seek out the information required without leading the respondent to reply in a way that might favour a specific direction. Once the research tools have been developed or existing ones validated for use in the study, information should also be provided about how they will be given to respondents, when this will be and how they will be returned. There is also a need to establish how the information obtained by these methods will be analysed.
The cooking process:
There are two parts to the cooking process: collecting in all the findings from the data collection and then undertaking a range of analytical tests to explore the relationship of data to the topics of interests, taking into account the interaction and effect of what are called ‘variable factors’ on the findings.
There are numerous pitfalls for quantitative studies where there is a need to count and compare numbers. Statistical methods range from the simple description of frequencies (how many women in the study were married?) to complex analytical tests that aim to pare away anything that might interfere with a route aiming to determine a cause and effect finding. The RCT is seen as the gold standard for research conduct, but even then, being able to prove that the intervention A causes the outcome B is fraught with difficulties as the variations within and between individuals are many and complex. So, even though you might be daunted by the language and description that accompanies statistical tests, you should still look at what is being presented to see what you can understand – the basic information is still there, it just needs seeking out.
Samples, response rates and percentages
One aspect of all research studies that deserves a mention is the sample number and how this relates to the overall results.
At the beginning of a study a total sample number is given followed by some explanation of who is eligible to take part and who is excluded. This results in another number of those who are potentially available for recruitment and finally, the number of those who agreed to take part. The article should include information about how respondents were approached and given information about the study and that they then gave written consent to take part.
This final figure then, all those who were both eligible and agreed to take part, is your total sample number – often referred to in the tables as n = I am going to refer to a sample number of 1500 primigravid women. At each point when the respondents are involved in providing data, it is important that the total number (n=1500) is set against the actual number of people who responded. This may then also be given as a percentage.
It is important to note whether the response rate is very different from the number of people who originally took part. Where the participants are in a randomised study, this is quite important if the allocation to treatment/intervention or control differs substantially. Where the response rate is less than half the original sample recruited, it is questionable whether the research findings will be valid. Response rates can vary within the study where different numbers of people answer questions as well as where data is missing. What is important is how the researchers identify the actual or potential effect of this on their analysis and conclusions.
Even where there are lots of tables and graphs, there should still be basic information given in the text so that by the time you have read through the results section, you should have a general feel for what the research has revealed.