Data-driven identification of potentially successful intervention implementations: a proof of concept using five years of opioid prescribing data from over 7000 practices in England
- Posted
- 29 Jun 2023
- Status
- Preprint
- Authors
-
- Lisa Hopcroft
- Helen Curtis
- Richard Croker
- Felix Pretis
- Peter Inglesby
- Dave Evans
- Seb Bacon
- Ben Goldacre
- Alex Walker
- Brian MacKenna
- Citation
- Hopcroft, LEM et al. 'Data-driven identification of potentially successful intervention implementations: a proof of concept using five years of opioid prescribing data from over 7000 practices in England.' Preprint. medRxiv 2023.06.28.23291704; doi: https://doi.org/10.1101/2023.06.28.23291704
- Categories
Background
We have previously demonstrated that opioid prescribing increased by 127% between 1998 and 2016. New policies aimed at tackling this increasing trend have been recommended by public health bodies and there is some evidence that progress is being made. We sought to extend our previous work and develop an unbiased, data-driven approach to identify general practices and clinical commissioning groups (CCGs) whose prescribing data suggest that interventions to reduce the prescribing of opioids may have been successfully implemented.
Methods
We analysed five years of prescribing data for three opioid prescribing measures: one capturing total opioid prescribing and two capturing regular prescribing of high dose opioids. Using a data-driven approach, we applied a modified version of our change detection Python library to identify changes in these measures over time, consistent with the successful implementation of an intervention. This analysis was carried out for general practices and CCGs, and organisations were ranked according to the change in prescribing rate.
Results
We present data for the three CCGs and practices demonstrating the biggest reduction in opioid prescribing across the three opioid prescribing measures. We observed a 40% drop in the regular prescribing of high dose opioids (measured as a percentage of regular opioids) in the highest ranked CCG (North Tyneside); a 99% drop in this same measure was found in several practices. Decile plots demonstrate that CCGs exhibiting large reductions in opioid prescribing do so via slow and gradual reductions over a long period of time (typically over two years); in contrast, practices exhibiting large reductions do so rapidly over a much shorter period of time.
Conclusions
By applying one of our existing analysis tools to a national dataset, we were able to rank NHS organisations by reduction in opioid prescribing rates. Highly ranked organisations are candidates for further qualitative research into intervention design and implementation.