“In 1980, it took 10 years to recruit 33, 000 patients for the hypertension study ALLHAT,” reflects Dr. Jonathan Perlin, President of Clinical Services and Chief Medical Officer of HCA. In contrast, Dr. Perlin’s team has just completed a study with over 600,000 patients from 53 hospitals in the space of 12 months and he is among the first to reap what he refers to as “the data dividend of meaningful use.” As part of meaningful use, America’s hospitals were required to acquire patient data on electronic platforms that could be shared and analyzed. Dr. Perlin has spearheaded a program of “Pragmatic Research” whereby measuring outputs from routine patient care can be used to develop best practices.
HCA’s first foray in pragmatic research was the REDUCE MRSA study for ICU patients. Research study partners from CDC, AHRQ, Harvard and others worked with 43 different HCA hospitals to compare three different approaches to eliminate Methicillin-resistant Staphylococcus aureus (MRSA) in patients in intensive care units: traditional isolation, isolation and treatment of patients identified as MRSA carriers with a nasal swab and chlorhexidine bathing, and treating all patients with the swab and bathing, a process called universal decolinization . Through the miracle of EHR, HCA enrolled and observed 75,000 patients, the study was completed in 18 months. The result: positive MRSA cultures were reduced by 37% and all cause blood stream infection decreased by 44%. On the strength of these results the “treat everybody” or universal decolonization approach is now recommended as best practice by the CDC and SHEA.
Traditionally, there has been a “17 year translation gap” before a scientific break-through becomes a disseminated best practice. In the case of the REDUCE MRSA study, the best practice was implemented into the entire HCA system in six months. Reducing MRSA had already been a priority at HCA, and MRSA infections system-wide were 70% of expected before the study. After implementing the universal 5-day disinfection of ICU patients, the rate fell to 42% of expected.
Dr. Perlin calls this a transformation becoming a “Learning Health System.” The data acquired through routine care can be fed forward to improve future care. Does the reduction of MRSA in the ICU mean that MRSA is reduced in hospital populations across the board? It takes a bigger study, and that’s where the 600,000 patients come in. Expect to see an answer later this year.
One challenge to the REDUCE MRSA study was theoretical possibility of resistance to the topical antibiotic, mupirocin. To study the effectiveness of mupirocin vs. an antiseptic iodinated nose swab, the research team is now conducting a 2-arm, 140 center trial with the goal of enrolling 1,000,000 patients. Enrollment in the study should be completed by the end of 2017, with results available in early 2018.
Dr. Perlin is taking big data in many directions at once. One of his teams looked at using computers to analyze patient records and ER CT scans for early, treatable cancers and to prevent any potentially overlooked tumors. Patients had come in to the ER and had required CT scans for reasons unrelated to cancer screening. Matching radiologists with machine analysis, they analyzed 33,000 cases. The computer analyzed each patient’s medical record and CT scan. Man and computer agreed that 1,000 cases appeared suspicious and ultimately 55 early tumors were identified.
With automated systems monitoring vital signs and laboratory data, Dr. Perlin sees a role for computers in the early detection of sepsis. His analytical tools have been validated at two hospitals thus far. Their approach is being placed in the public domain to help all hospitals improve sepsis detection as well.
Dr. Perlin calls these results the “data dividend.” He notes that the landmark ALLHAT study took 10 years to determine the efficacy of various hypertensives to prevent mortality, which meant 10 years without a life-saving answer. He says to have conducted the REDUCE MRSA study in 1 hospital, it would have taken 64 years to complete. With big data, researchers answered the question across 43 hospitals in 18 months. With the capabilities offered through machine learning, physicians and other clinicians have an improved ability to avoid missing diagnoses, such as the lung cancer example, and make diagnoses faster, such as HCA’s improved monitoring for sepsis. In a world with big data and little time, he is delighted to be working in a time of accelerated learning and decision support. Machines can help us quickly recognize patterns that we would see if we were looking at all of the data all-of-the-time.
Dr. Jonathan Perlin was interviewed by Dr. Thomas Masterson, editor, Vox Salutem.