Missing data

-



  • What does it mean?

When a participant outcome is unavailable, due to a missing questionnaire or non-attendance to a trial related clinical appointment, their results are unknown. We call this missing data. In the toothpaste trial, this would mean we don’t know a participant's pain score at the end. We have a number of options available to try to account for the missing values. For example, we can use statistical tools that aim to predict what happened to patients that did not provide results, taking into account these predictions are uncertain, ie we do not know for sure what happened to those participants.

You can find out more about what missing data is and its impact on interpretation of treatment claims here.

  • How could patients be involved?

Patients can help us predict missing values so we can assess treatment effect with more certainty.

  • Example 

Some participants in the toothpaste trial will have their pain score missing. For those participants and to ensure we investigate different and realistic scenarios, we might need to predict what their value might have been, had they answered the question. We have different sources of information to make that prediction, like whether they attended urgent dental treatment and dentists’ opinions. However, it is still an informed guess. Should patients help make that informed guess? 

Comments

  1. Unknown11 March 2022 at 07:30

    Patients may be able to identify factors leading to missing data which will help the informed guess.

    ReplyDelete

Post a Comment

Popular posts from this blog

Analysis methods: how the trial data will be analysed

-
 What does it mean? When patients enter a trial, we collect their information. We usually collect their information at baseline (who is taking part in the trial?), and during follow-up (with questionnaires and/or clinical visits). This information is gathered in a database and analysed to answer the ultimate trial question - does the treatment work when compared with a control (usually another treatment)?  Trial analysis is complex and involves statistical models that estimate the treatment effect, ie how well the treatment works when compared with a control.  The statistical models we use to analyse trial data may be "adjusted for" patient's characteristics. This usually happens if we believe those characteristics are important to understand the treatment effect. Statistical models make assumptions about the information collected and it is important to know those assumptions in order to interpret the findings of the trial correctly. An example In the toothpaste trial, we
Read more

Harms

-
What does it mean? Most treatments come with a risk attached - they may lead to harms and side effects, even when they benefit patients. It is crucial to measure harms and side effects and show this information by treatment. This ensures patient's safety is always considered when running a trial and interpreting findings. An example As Emma starts using toothpaste A, she notices she has developed a strange side effect - she develops mouth ulcers and is concerned. She contacts her dentist who immediately reports this side effect to the trial team. The trial team keep note of this side effect, along with others that are reported from other patients. During the trial, and at the end, the trial team analyses and presents this information. How can patients be involved? Patients can help researchers when summarising and presenting data on safety, to ensure it is clear and can be discussed by the whole trial team. Patients can be involved in interpreting the safety analysis findings.
Read more