1.2.2 Interpreting Data on Risk Factors

Analysing data on illness and mortality

• Analysing data on illness and mortality is an essential part of the study of risk factors
• It is important to remember that correlation between a risk factor and a disease does not always mean that a causal relationship exists
  • Scientists cannot assume that because there is correlation between variables that one has caused the other
  • Many other factors will influence the prevalence and likelihood of disease and these factors need to be taken into account when analysing and interpreting data

Describing data

• This means identifying trends and stating what the results show e.g. the data show that the oldest age group has the highest relative risk of heart disease
  • When describing data it is always good to use numbers from the data to back up your descriptions, e.g. the data show that the oldest age group of 80+ has the highest relative risk of heart disease of 2.4

Drawing conclusions from data

• This means working out what the data show about the relationships between variables, e.g. the data show that there is an association, or correlation, between age and the relative risk of heart disease 
  • Conclusions should always be limited to what the data show
    • Causal relationships cannot be concluded from one data set, e.g. it cannot be concluded from one study that older age causes an increase in the relative risk of heart disease
    • Conclusions cannot be extrapolated beyond the setting of a study e.g. a study carried out in 40-50 year old adults cannot be applied to people over 70, and a study carried out in mice cannot be directly applied to humans

Evaluating the validity of data

• Larger sample sizes are more likely to give valid results as the sample is more likely to be representative of the population in question
  • Results are considered to be valid if they measure what they set out to measure, i.e. they are not influenced by external variables or poor experimental design, and have been analysed correctly
• Statistical analysis should be used to check that any differences between results are statistically significant
• Some studies need to have a control with which to compare the results
  • E.g. when testing a drug to treat heart disease, a control group that is not given the drug should be included in the study to ensure that any effect shown is due to the drug and not any other factor
• Studies should be repeated, or there should be many studies that show the same result, before conclusions can be drawn
• The study should be designed to control any variable that is not being tested
• Researchers should not be biased, i.e. looking for a particular outcome
  • This could be a problem if someone is being paid to come up with a particular result

1. A description for this data could include the following
   • Non-smokers exposed to 0 cigarettes per day have the lowest relative risk of CVD, with a relative risk of 1.00
   • Non-smokers exposed to 20 or more cigarettes per day have the highest risk of CVD, with a relative risk of 1.31
   • As the number of cigarettes smoked per day increases, the relative risk of CVD increases from 1.00 at 0 cigarettes per day, to 1.23 at 1-19 cigarettes per day, and then to 1.31 at 20 or more cigarettes per day  
2. A conclusion for the data could be
   • There is a correlation, or an association, between increased exposure to cigarette smoke and an increased risk of CVD
   • Note that while it could be concluded here that increased exposure to smoking causes an increase in the relative risk of CVD, it is best not to draw causal connections without more evidence
     • This is only a single study; there may be concerns over its validity, and other studies could show conflicting evidence
3. A commentary on the validity of the data could include
   • The study included 523 people; this is a fairly small sample size and may not represent an entire population
   • This is only one study; more studies would need to be carried out to back up these results
     • Being able to replicate, or repeat, the results of a study shows that the results are reliable
   • There is no information on how other risk factors might be interacting with smoking to influence the risk of CVD
     • Risk factors such as age, diet, biological sex, or exercise levels may be playing a role, as these factors may be interacting with the smoking variable e.g.
       • Smokers are often older
       • More men may smoke than women
       • Smokers may be less likely to exercise
   • The data doesn't comment on the use of any statistical tests so we cannot state the significance of the differences between the different levels of smoke exposure

Recognising conflicting evidence

• Evidence from one study is not enough to conclude that a risk factor is a risk to health or associated with a particular disease
• Studies similar in design would need to be analysed together to make links
  • Such an analysis is referred to as a meta-analysis
• Similar conclusions would need to be drawn from all studies in order to accept the findings
• Conflicting evidence may be found that leads to a different conclusion
  • Conflicting evidence is that which shows a different pattern to the evidence gained elsewhere
  • When conflicting evidence arises, more research is needed to show which pattern is correct
  • Conflicting evidence is often a sign that other variables are involved

• There are a number of factors to consider when evaluating the design of a study
  • The sample of people included in the study
    • Sample size should be representative of the population involved; larger samples are more likely to be representative
    • Randomly selecting participants removes bias and increases the likelihood of a representative sample
      • Selecting people from among friends or from among people at a gym, for example, is likely to select people who have similar lifestyles
  • Variables; the more variables that have been controlled the more reliable and valid the data
    • Reliable data can be reproduced by repeating an experiment
    • Valid data has only tested one independent variable whilst all other variables are controlled
  • Data; there should be no bias involved in the collection or analysis of data
    • Bias in data can come from human sources, e.g. by selecting a non-random sample or manipulating data to emphasise a certain outcome
  • Controls; the use of an experimental control provides a point of comparison and ensures that the results are due to the variable of interest
  • Repetition; repeats of the data need to be taken within a study, and similar results should be collected
    • Similar results are reliable 
  • Reproducible; it should be possible to reproduce a set of findings by repeating an entire investigation

• Risk is defined as the chance or probability that a harmful event will occur
• The statistical chance of a harmful event occurring needs to be supported by scientific evidence gained from research 
• An individual's perception of risk may be different to the actual risk of something occurring
  • Risk can be overestimated because of factors such as
    • Misleading information in the media
    • Overexposure to information
    • Personal experience of the associated risk
    • Unfamiliarity with the event
    • The event causing severe harm
  • Risk can be underestimated because of factors such as
    • Lack of information 
    • Misunderstanding of factors that increase the risk
    • A lack of personal experience of the associated risk
    • Unfamiliarity with the event
    • The harm being non-immediate