Get a free quote now.
Vancomycin Dosing in Elderly Populations: Improving Accuracy Through Precision Dosing Intelligence
Did you attend IDWeek 2020? If so, here’s a sneak peek of what Jon Faldasz, our Head of Clinical Applications, shared at our IDWeek 2020 Learning Lounge, now available on-demand.
In addition to providing detail about InsightRX Nova’s top-tier dosing decision support capabilities and state-of-the-art, real-time analytics, he touched on a third and incredibly important pillar of InsightRX Nova—continuous learning.
Using a specific example pertaining to vancomycin dosing in elderly populations, here’s an overview of how our Data Science team uses and implements continuous learning to keep InsightRX Nova fresh and current.
First, through our data-focused approach, we noticed an unexpected finding.
The great thing about our data-oriented approach is that we’re constantly looking at our partner hospitals’ deidentified data in search of meaningful signals. Our partner institutions’ clinicians are able to look at their data and notify us about interesting findings too.
In this particular case, we noticed an unexpected signal that also occurred at similar sites using similar software filters, specifically in sites using a model published by Goti et al. (2018). The Goti model is a population PK model that rounds serum creatinine (sCr) up to 1 mg/dL in patients older than 65 when calculating creatinine clearance.
Then, our Data Science team dug deeper into the issue in order to develop a solution.
After noticing this commonality, our Data Science team got to work. We ran an analysis on more than 2,853 adult vancomycin patients at 9 different institutions whose dosing regimens were impacted by our platform’s Goti rounding filter. These patients were >65 years old and had at least one sCr value less than 1 mg/dL.
To determine whether the age-adjusted sCr rounding in the Goti model was the cause of skewed data in this population, our Data Science team created a modified version of the Goti model. In this version, sCr values less than 1 mg/dL were not rounded up. Instead, the raw sCr values were used to calculate CrCL and dose vancomycin accordingly.
We rigorously tested our solution.
Using the published Goti model, we discovered a trend of CrCL underestimation, leading to underdosing in the elderly subpopulation of adult vancomycin patients.
In our modified Goti analysis which used raw, non-adjusted sCr values, CrCL was more accurate. Results from the modified Goti model reflected noticeable improvements in two important values:
- Mean percent error (MPE), which measures bias
- Relative root mean square error (rRMSE), which measures precision
Our modifications correlated with reduced prediction bias and error.
We incorporated our conclusions into our software for precision dosing, Insight RX Nova.
These results led us to recommend using raw sCr values rather than age-adjusted sCr in elderly patients receiving vancomycin. But continuous learning doesn’t end with our Data Science team. We decided to make these findings known by:
- Presenting them at our IDWeek 2020 Learning Lounge
- Publishing a letter to the editor in Therapeutic Drug Monitoring
- Encoding the modified Goti model into the InsightRX Nova platform so that users can benefit from the advancement at no additional cost
In this way, our approach to precision dosing is designed to complement clinical judgement, not to replace it. Through our approach to continuous learning, our precision dosing software gets better and more accurate over time and remains current with the latest trends in population PK.
Learn more about our precision dosing software, InsightRX Nova.