We build state of the art machine learning-based calculators, with the following characteristics: Specific to your patient population. Data scientists create your predictive model based on data from your patient population. This population-specific data will ensure that the predictions are as precise as possible to your patients. We build different models for a range of patient outcomes, including mortality, complications, readmissions, and cost. Personalized risk calculation. After our models demonstrate adequate precision, we turn them into prediction calculators. [Read More]
Medical image recognition through Deep Learning algorithms
Data Science Methods
Deep Learning is a machine learning algorithm often used in image recognition. This method reduces a medical image (MRI, CT, plain radiographs) to a numeric matrix, and an algorithm then analyzes this matrix in search of certain features associated with specific radiologic signs and diagnosis. It presents several advantages over the traditional way of image classification by clinical experts. Below we outline the comparison between the two, how the algorithm works, and what you would need if you would like to create a Deep Learning algorithm for your image recognition task. [Read More]
Machine learning approaches to enhance claims and registry data analyses
Our slides for the The Gerontological Society of America’s 71st Annual Scientific Meeting, taking place in Austin, Texas.
Machine learning for bundled payments applied to total knee replacement procedures
Our slides for a presentation on bundled payments applied to total knee replacements.
Data Science Methods
Machine learning is a sub-discipline of Artificial Intelligence (AI) that encompasses a broad range of algorithms to perform intelligent predictions of outcomes based on a dataset. The prediction algorithm takes in large sets of data, finds patterns, trains itself using this data, and outputs a result. Most importantly, unlike traditional statistical analyses that are focused on risk factors for groups of patients, machine learning allows for the prediction of outcomes for individual patients. [Read More]
Leveraging the economic value of participation in a trauma registry.
Mary is a trauma surgeon, and she leads a group participating in a clinical registry. That was undoubtedly the right decision. Sure, there is an annual fee, and collecting data does require some extra work, but Mary’s group does get benefits. First, registry leaders send out periodic benchmarking reports comparing their local clinical outcomes against other centers around the country, all risk-adjusted so that they wouldn’t get penalized for having more severe cases. [Read More]
Enhancing the effectiveness of chronic pain prevention through in-silico trials and Artificial Intelligence.
Mary just came back from a military deployment where she had a severe upper extremity injury in her dominant arm. Despite the best available care, both in the field and at hospitals in the US, Mary develops chronic pain. The consequences to her life are consequential: alcohol and drug disorders, poisoning with opioids, constant suicide ideation, and, more recently, self-inflicted injuries, including suicide attempts (Meerwijk et al. 2019). Mary is a fictitious character but represents a scenario lived by thousands of military personnel. [Read More]
Self-report assessment and its value to different stakeholders.
Mary goes for a clinical appointment at a health system. While waiting for her appointment, a nurse asks her to fill out a questionnaire about her quality of life. The questions are kind of silly, and Mary keeps wondering why she was asked to spend her time doing that. After all, her nurse didn’t make any comments about her multiple responses. Her physician also didn’t even seem to know that she had answered that questionnaire. [Read More]
Streamlining data collection by machine learning-based prediction of physical tests and self-reported scores.
Ava has a significant challenge ahead of her. She is the Principal Investigator for a cohort of patients with cardiac conditions, and tests – such as the Timed Up and Go (TUG) – and self-reported scores are vital to the success of her program. However, primary care physicians involved with the project consider a time-consuming data collection protocol to be a deal-breaker. In other words, if the pace of their clinics slows down, this will make their lives a nightmare, and they will be less likely to participate. [Read More]
Implementation and de-implementation science applied to machine learning models.
Louise is ecstatic. She just completed a machine learning model to predict hospital readmission that outperforms the previous best model by a large margin. To get that done, Louise brought in variables from multiple external databases. She also conducted an extensive feature engineering, meaning a lot of variable transformation. Now the only missing step is to get the model integrated into the clinical workflow at her healthcare system. But that should be easy, right? [Read More]