Artificial Intelligence (AI) models are redefining the patient experience by improving a healthcare provider’s ability to deliver patient care and treatment. Whether it’s administrative work or clinical diagnosis, AI frees up time for healthcare providers to focus on patients. AI models can be trained to streamline administrative tasks and improve the accuracy and timeliness of diagnosis. By using AI models, healthcare organizations can sort through data and empower data-driven decision making. But how can healthcare organizations unlock these capabilities? We spoke with Seth Kindley, Pure Storage’s Principal Data Architect, on how healthcare organizations can improve medical services and patient experience by implementing the right technology solutions.
Future Healthcare Today (FHT): What are some of the ways healthcare organizations can use AI models to improve the delivery of patient care and treatment?
Seth Kindley (SK): AI is being used to track real time interactions within healthcare facilities, specifically focusing on the patient experience. AI is used to track patient interactions from the time they check in, to when they are put into an exam room for imaging, and to the distribution of prescribed medications when they leave. All these interactions are being tracked in real time with the idea of taking human error, supply chain shortages, and risk out of the equation. Ultimately, the goal is to drive efficiency. The more efficiently you can take care of a patient, the less time they have to sit and wait for treatment.
We have the ability to improve the way things are done, as long as we have access to and can train high quality data models. An AI model will never take the job of a human. However, AI models can be used in concert with healthcare providers to improve the delivery of patient care and treatment, while minimizing the effects of human error.
FHT: What insights can healthcare providers gain by using AI models to assist with clinical tasks like digital imaging?
SK: We can take data from picture archiving and communication system (PAC) studies and use it to train AI models to drive efficiency through pattern, or marker, recognition. The ideal scenario in the future would be that AI models could look at this data, glean an inference from these studies and come up with a preliminary diagnosis. This would allow healthcare providers to get to a course of action faster.
This can be especially beneficial in rural healthcare centers that don’t have the typical infrastructure, or where individual specialists may not be easily accessible. Their digital imaging can be done in a clinical study, then AI models can scan it, and then the course of care can be suggested based on their markers without having the radiologist interact with it. They can even determine statistical probabilities within a half, or a tenth of, a percentage based on the markers of this model. This streamlines the time to action of patient care and treatment.
The goal of automation in healthcare is to drive efficiency and decrease the number of people involved in the decision-making process. By training these models with extensive data, we can get a more definitive diagnoses or accurate recognition of a marker while minimizing human interaction. This will free up healthcare providers to focus more on the delivery of face-to-face patient care. Moreover, this shift will allow more patients to move through the system faster and will provide them with a better quality of care.
FHT: What role does data play in the expansion of AI capabilities?
SK: Data is the cornerstone for expanding the capabilities of AI models. Data can be used to train AI models to perform several tasks, such as recognizing the difference between DNA and disease markers, as well as study certain phenotypes at a higher level. The model will cross examine images pixel by pixel until it has enough positive identifications to definitively know the differences between any two images.
Data plays a critical role in expanding the capabilities of AI, but for so long data has been held in silos, data lakes, and data hubs without anyone being able to put it to work. Now we are going through and breaking down those silos and digitizing decades-worth of files, magnetic media, paper, or even film x-rays. But we’ve got important knowledge gaps emerging from this process based on where data was collected and who is represented in that data.
In order to fill these knowledge gaps, we have to go through this large inventory process where we assess the data we already have and see what it’s worth. This includes digitizing data to make it available for these AI pipelines to train against and redacting data that we definitively know won’t add any value to the data sets. In order to improve the capabilities of AI and patient experiences, federal and civilian healthcare agencies need to work together to come up with a data foundation where everyone can use that data to train their AI models.
FHT: How can healthcare organizations ensure the security of patient data as it’s used to train AI models?
SK: Ensuring the security of patient data is of the utmost importance. Healthcare organizations need to have the highest levels of security because they are often targets of cyberattacks and ransomware attacks.
To ensure patient data security, healthcare organizations need to have a robust physical and cyber security infrastructure in place. Physical security includes the protection of everyone who comes through the facility, medical supplies, and physical copies of records and the devices used to store and access them. Cybersecurity includes securing data and the in-transit encryption of new data. Advances in technology can enable a healthcare organization with the ability to unequivocally prove that the data stored is secure and data acquired is encrypted while in transit, and at rest. Cybersecurity systems should have Intrusion detection systems (IDS) and intrusion prevention systems (IPS) as well as a networking team to secure data and ensure it is backed up securely and reliably.
There is no single thing that a healthcare organization can do to ensure the safety of patients and their data. Rather a number of things working in concert with one another to ensure that the data is safe and available to make and train accurate AI models.
FHT: Any final thoughts to share with our readers?
SK: The process of training AI models can be difficult, and a bit overwhelming at times because the data is dynamic. It’s important that healthcare organizations build partnerships with organizations that specialize in developing reference architectures, have subject matter experts with clinical experience, and data scientists to help them on this journey. Building and training AI models that can assist in the advancement of patient care is a long-term process, but the process becomes shorter and less complex with a trusted advisor to help navigate through the ever-changing world of technology.
Learn how data can be used to train AI models to assist in redefining the patient experience here.
