Electronic health records are a hot topic in the world of medicine, as hospitals implement new computerized systems to meet federal incentives. Proponents of replacing paper records with electronic health records (EHR) in hospitals and other health care settings argue that the update will improve the efficiency of health care, cutting costs and making life easier for patients and doctors. But a less popularized – and probably more immediate – effect of the EHR wave will be felt by clinical researchers, who will suddenly have a flood of medical data where there once was a drought.
This new EHR-enabled world of clinical research was featured in a recent lecture at the Department of Medicine Grand Rounds by Ari Robicsek, visiting from the Medical Center’s partner institution in Evanston, NorthShore University Health System. Robicsek is an infectious disease specialist and a self-described “accidental informaticist,” a physician and researcher who found himself drawn to EHRs as a tool to address important clinical questions. As an early adopter of paperless medical records, NorthShore has had 8 years to build a “data warehouse” that can be used for research projects. While the Medical Center works toward the next phase of its own EHR launch, called Phoenix, Robicsek’s examples were an exciting peek at how the new resource can be used to prevent hospital-acquired infections and make the most significant change to the definition of fever in 140 years.
“These are, I hope, a series of interesting illustrations of the increasingly amazing things that researchers and hospital systems are capable of doing because of the growing informatics resources available to us,” Robicsek said.
A top priority and concern for any hospital is reducing the spread of bacterial such as MRSA, which can infect sick patients with suppressed immune systems during their inpatient stay. In the last decade, hospitals have launched intensive screening programs to find patients who are carrying these bacterial strains as soon as they are admitted to the hospital, so that extra precautions can be taken. However, it’s not cheap to test every single patient, and false positives in the tests create unnecessary expense. Being able to target tests to patients more likely to be colonized by MRSA could save millions of dollars – a shift that Congress has ordered, without offering any help on just how to find those “magical” high-risk patients, Robicsek said.
Sounds like a job for the electronic health record! Because NorthShore has been adding the results of its MRSA screening tests to patients’ electronic records, Robicsek and colleagues were able to quickly comb through the data of more than 23,000 patients to find characteristics that predicted a high chance of carrying the bacteria. Instead of pulling each paper record by hand as in the old days, computer models could be built to find predictors of risk. When tested in a second batch of data (built from more than 26,000 patients), the models published earlier this year could identify the 30 percent of “high-risk” patients who account for the majority of positive MRSA tests. If implemented (as NorthShore plans to do later this year), such models could direct testing to those patients most likely to be an infection risk, rather than testing willy-nilly and racking up giant expenses.
Besides alerting physicians to clinical threats, electronic health records can also help them do more with data they’ve been collecting the old-fashioned way for centuries. Fever might be the most basic biometric, simple enough for Moms to test informally at home with the back of their hand. But the meaning of fever has changed little since Carl Reinhold August Wunderlich established the normal body temperature of humans (roughly 37° Celsius or 98.6° Fahrenheit) in 1871, Robicsek said.
“[Wunderlich] is thought over the course of his career to have taken the temperature of some 25,000 individuals, and it was his monograph on clinical thermometry that caused temperature vigilance to be introduced into routine clinical care,” Robicsek said. “Remarkably, there has been very little subsequent work validating his data…almost nobody has looked at this in the setting of physiological perturbation,” – in other words, asking what is a “normal” fever after a surgery, and when is it a cause for worry.
So Robicsek’s team took 440,000 temperature measurements from the Northshore EHR system, in patients who had received several different types of surgical procedures, from hip replacements to cardiovascular surgery. They found that the patterns of fever were radically different depending on the type of procedure, the patient’s age, the patient’s temperature before surgery, and many other factors. Rather than throw in the towel in the face of this complexity, Robicsek and his colleagues built unique models for each type of procedure, calculating the distribution of temperatures and the relevant factors for each case. To easily navigate this information, they built a resource called The Wunderlich Project, that can be accessed by any physician online curious about whether their post-op patient’s temperature is “normal.”
That web tool is the perfect example of how EHR data can transform complex questions – and complex answers – into useful information for doctors. Even more exciting, the Wunderlich Project remains hooked into the flow of EHR data, as each night, the models update with new information acquired from temperature measurements taken the previous day.
“Every morning when you use this tool, the numbers are a little better than they were the night before,” Robicsek said.
Electronic health records have their limitations, Robicsek warned. Data is only as good as the person who puts it into the system, and human errors and biases can persist in an electronic format. But even with those imperfections, the ability of researchers to almost instantly access thousands of data points instead of painfully, methodically flipping through file cabinets reflects a major leap in the way clinical research is conducted.
Robicsek A, Beaumont JL, Wright MO, Thomson RB Jr, Kaul KL, & Peterson LR (2011). Electronic prediction rules for methicillin-resistant Staphylococcus aureus colonization. Infection control and hospital epidemiology : the official journal of the Society of Hospital Epidemiologists of America, 32 (1), 9-19 PMID: 21121818