Exploring high opioid prescriptions among nephrologists in the United States using machine learning algorithms
Background and aims: The opioid pandemic has contributed to deaths globally, and prescription opioids have played a crucial role in these deaths. Addressing overdose requires understanding the reasons behind prescription, especially in cases of chronic diseases. Several factors play a role in the in...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | Emerging Trends in Drugs, Addictions, and Health |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2667118224000242 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Background and aims: The opioid pandemic has contributed to deaths globally, and prescription opioids have played a crucial role in these deaths. Addressing overdose requires understanding the reasons behind prescription, especially in cases of chronic diseases. Several factors play a role in the increased prescription of opioids, relating to patients’ lifestyle, characteristics, and disease. As these factors are complex in nature, understanding them requires machine learning approach. This study explored overprescribing opioids among nephrologists in the US using unsupervised machine learning algorithms. Design: Two types of unsupervised clustering were applied to the Medicare Provider Utilisation and Payment Data Part-D Prescriber Summary. Setting: The dataset had 50,134 records with 85 features relating to opioids prescription per US state. Univariate and bivariate analysis were applied first to gain understanding of the data followed by K-mean clustering and Gaussian Mixture Models. Findings: Unsupervised clustering showed that prescription issued to males were three times higher than those issued to females. Moreover, male nephrologists were higher prescribers than female nephrologists, and a third of male nephrologists were high prescribers of opioids. The highest rates of prescriptions were seen in California. Conclusions: Unsupervised machine learning algorithms enabled understanding of high opioid prescription across gender and US state by analysing multiple features. Both K-mean clustering and Gaussian Mixture Models achieved the same outcomes. Future work will benefit from applying deep learning in order to understand in-depth patterns in prescription and contributing factors related to over-prescribing. |
|---|---|
| ISSN: | 2667-1182 |