Endocrine Abstracts (2016) 45 P26 | DOI: 10.1530/endoabs.45.P26

Ensuring complete data for the national paediatric diabetes Audit - An Approach Using Python

Victoria Price, Akhil Shah & Michael McGuigan


Leighton Hospital, Crewe, UK.


Introduction: Diabetes teams need to keep electronic data records for supporting participation in the National Paediatric Diabetes Audit (NPDA), and other governance activities. NPDA benchmarks diabetes services against each other, and services are keen to make sure their data is accurate so that it gives a good reflection of their service. Twinkle is a specialised database that is widely used for this purpose.

Twinkle allows units to produce specific reports on their patients as a CSV spreadsheet, the format NPDA requires for uploading to their website. However, we found limitations in this process. We found it difficult to identify missing data from the Twinkle database itself using the built in Report Builder. In addition, sometimes data entered onto Twinkle was not appearing on the CSV file.

We therefore explored an alternative approach to this problem.

Methods: We developed a computer programme in Python, a basic programming language, to specifically analyse the data in the Twinkle CSV file prior to uploading to the NPDA website. The algorithm could identify young people with missing care processes, whether this was due to the process not being performed, not being entered onto Twinkle, or not being exported correctly to the CSV file. Data could be analysed only for those completing a full year of care.

By generating this list in April, we were able to identify children with missing care processes and add this information if it was available, improving our data completeness. We then generated the list in December, which enabled us to target children with missing data in the last quarter of the April-April audit year, to try and improve the number of care processes completed.

This led to an improvement in key care processes as follows: albuminuria 53.7% to 88.1%; retinopathy 61.1% to 90.5%; foot examination 66.7% to 92.9%.

A similar programme was able to analyse the median HbA1c for a specific time frame. This is not possible directly within Twinkle.

Discussion: Alternative approaches may be able to achieve the same result using report builder, or through advanced features of Excel, but this was a solution which we found useful.

Article tools

My recent searches

No recent searches.