Big data analysis

From big data to design recommendations: Analysing usage data to see behaviour

What

Medical infusion pumps are designed with number entry interfaces that assume all numbers are equally likely. This approach means that there is more chance for slip errors to occur. Investigating and redesigning the interface could reduce these slip errors.

Why

Human error in the medical domain results in serious harm to patients and sometimes death. The devices, whilst technically fit for the job, do not appear to be rigorously designed and tested with real users in real scenarios. Applying good user centre design principals could result in design recommendations that could save lives.

Action

Testing these devices in hospitals is difficult, time consuming and costly. The ethical process is understandably arduous. I realised there was an opportunity for conducting research on exisitng data. Using the usage logs as a form of “desire path” I would be able to understand how these devices were being used and what numbers were needed.

I acquired log data from a number of infusion devices in one hospital across multiple departments. I process the huge data sets using R. From this I was able to extract clear patterns in the numbers used: largely 1, 2, 5 and 0, and in doing so shed light on real infusion pump usage.

Impact

From this data I was able to design number entry interfaces that better matched the task they were used for - entering medical dosage data. These design recommendations were part of a white paper created to inform and improve medical device design.

The work was shortlisted for a Human Factors and Engineering Society award.

Previous
Previous

Participant Incentives

Next
Next

UCD Leadership