Artificial intelligence (AI) has begun to achieve its potential by focusing on the major challenges healthcare professionals face. For example, Flagler Hospital is using an intelligent application designed to reduce clinical variation that improves patient outcomes and is the key to lowering health costs and succeeding with financial risk.
Clinical variation is tied to the 30% of US healthcare that experts say is unnecessary and costs the nation roughly $750 billion per year. Reducing clinical variation requires the ability to analyze a giant amount of data. With the help of AI and the vast computational power now available, even medium-sized and smaller hospitals can safely reduce variations in care.
Flagler Hospital in St. Augustine, FL has successfully used an AI solution to improve and standardize care paths for pneumonia, sepsis, COPD, heart failure and is on track to apply the same technology to 18 more conditions.
After we got the green light to proceed with our pneumonia pilot, the medical informatics department wrote nine queries for five systems, including our electronic health record (EHR), our enterprise data warehouse, and our surgical, financial, and corporate performance management systems. We also engaged a workgroup of physicians from each department in the hospital. This committee determined which variables to assess, including “continuous” variables like costs and length of stay (LOS) and “categorical” variables like patient origin and comorbidities.
The AI solution used unsupervised machine learning to understand the structure of the data and find patterns that revealed the best pneumonia treatment approaches. The software created groups of patients with similar outcomes and showed the treatments they received at what times and in which sequences. The program then showed the direct variable costs, average lengths of stay, readmission rates, and mortality rates for each cohort, along with the statistical significance of the data. Comorbidities were also factored into the program’s calculations.
To determine how the pneumonia care path should be changed, our committee selected the “Goldilocks” cohort with the shortest length of stay, least readmissions, and lowest mortality at the lowest cost. In some cases, the treatment of this patient group differed significantly from the then-current care path.
The optimal events, sequence, and timing of care were presented to the physician team using an intuitive interface that allowed them to understand exactly why each step (and its timing) was recommended. We then reviewed evidenced-based guidelines for the care path for pneumonia in the Goldilocks group to ensure evidence was being followed and created new order sets in the EHR system.
Reports generated by the AI software indicate that many physicians were not following the new pneumonia care path, but we are working to correct this, and adherence is improving weekly. We attribute the positive response largely to the use of our own data with the AI program, instead of data from scientific studies. As a result, the physicians had confidence that the results were based on data for patients like theirs. With the new care path, Flagler Hospital saved $1,350 per patient, reduced LOS by an average of 2 days, and dropped the readmission rate from 2.9% to 0.4%.
Following the successful pilot, we used the AI solution to improve our sepsis COPD, and heart failure care paths. We are about to deploy a care path for total knee and total hip replacements. Scheduled for the future are myocardial infarction, CABG, hysterectomy, and diabetes. Flagler Hospital expects to not only save at least $20 million from this program in the next 3 years, for a return on investment of about 22:1, but also to further enhance patient safety.
Other community hospitals can do the same—even without a data scientist on board—and if successful, can improve patient safety and outcomes, increase efficiency, and boost bottom lines. As hospitals move toward risk, moreover, clinical variation must be managed, and this is a sure-fire way to accomplish that goal without breaking the bank.