The use of data analytics is changing the way healthcare is being provided. The ability to more accurately predict patient populations, combined with the analysis for managing outcomes has enabled healthcare providers to optimize (both for the patient and provider) treatment plans.
Data scientists pull together many different medical and other relevant data sources, then build models to predict the potential outcomes, and offer decision support at the critical point of service. Many of the decisions facing healthcare professionals are life and death. The ability to confidently make these decisions by combining data driven insights with provider expertise saves lives, increases workforce efficiency, limits risk and liability, manages costs, and enables more effective care.
Mosaic is a great partner to consider for healthcare analytics projects.
Mosaic helped a Children’s Hospital better predict their ICU saturation
- Obviously, when the ICU is saturated, more beds and medical personnel are needed for the at-risk patients, leaving other patients to be sent to less competent facilities for care.
- With better visibility into the ICU and how the ICU is impacted by Emergency Room operations, the Hospital can staff more dynamically, giving everyone the care they need, and save lives.
Mosaic built a machine learning email classifier for Purchase Order identification
- Mosaic developed and implemented a machine learning tool that identifies purchase orders that need attention from a leading hospital system’s supply chain group.
- After comparing a variety of models, Mosaic selected on an logistic regression.
- Under different thresholds, the model performs quite well, correctly predicting at 91% and 85%.
- Mosaic’s tool not only provides a reduction in time spent evaluating PO confirmations, but also eliminates almost all of the missed exceptions due to human error in reviewing thousands of purchase orders on a weekly basis, resulting in a number of beneficial downstream financial and patient experience effects.
Mosaic used text analytics to extract public health insights from social media for the CDC.
- As rates of chronic disease increase in the United States and around the world, it is important for researchers and policymakers to understand how people manage long-term health conditions.
- Social media posts provide a rich source of first-person lifestyle information, but it is difficult to extract meaningful information about the health of a population from these posts.
- Over just a few short months, the team scraped two health-based online forums, HealingWell.org and Mothering.com, and integrated the posts into a web-based exploratory data analysis tool called the Mosaic Social Media Context Awareness Platform.
- Mosaic used semi-supervised machine learning algorithms to characterize posts grouped by particular word patterns that can then be applied to predict characteristics of additional social media posts
- The prototype platform can be accessed at this website: https://cdctool.mosaicatm.com
Mosaic has developed a disease progression model for a leading healthcare technology company.
- We designed and applied innovative machine learning techniques to financial and clinical data streams.
- Several techniques were experimented with including estimation, clustering and Markov modeling.
- Clinical and statistical results were validated, lending new insights to the problem of understanding disease progression and the different types of outcomes patients experience.
- These findings help caregivers inform and motivate patients toward different treatment strategies and personal efforts for better outcomes.
Division of Johnson & Johnson – Medical Device Demand Forecasting Assessment
- Mosaic, an innovative provider of healthcare data science consulting, designed and delivered an analytics assessment centered around demand forecasting for medical devices.
- Data streams included clinical, financial, diagnostic, and patient demographic data.
- Mosaic delivered a detailed report as to how they could take advantage of current data with various modeling techniques, how to make the most of demand forecasting software, what additional data would need to be acquired to improve forecasting accuracy, and what types of modeling techniques could help them get there.