All Content

Mining for Data


The expansion of information technology (IT) in U.S. hospitals is an evolutionary process. Billing, collections, and admission and discharge records have long been computerized, but now electronic medical administration records, patient electronic health records, and computerized physician order entry (CPOE) systems are joining the ranks.

Hospitalists are more likely to encounter sophisticated IT systems in larger, urban, or teaching hospitals, according to a 2005 survey by the American Hospital Association.1

The Hospitalist’s first installment about hospital informatics (“Charts to Screens,” January 2007, p. 25) focused on the challenges of health IT and the barriers to effective adoption of computer-based documentation systems. This installment explores the potentially rich vein of data available to hospitalists from those information systems and the opportunities for research and QI applications.

The mechanics of conducting clinical research and QI projects will depend to a large extent on the progress each hospital medical group’s institution has made in the IT adoption process. Some say hospitalists have powerful contributions to make in influencing how the IT process evolves so their research opportunities will also improve.

QI Topics

Data in information systems differ from hospital to hospital, says Tejal K. Gandhi, MD, MPH, director of patient safety at Boston’s Brigham and Women’s Hospital (BWH) and assistant professor of medicine in the Department of Medicine at Harvard Medical School in Boston.

Dr. Gandhi’s research focuses on redesigning hospital and outpatient processes to improve patient safety. She notes that hospitalists can take advantage of data the hospital is collecting to satisfy its reporting requirements to spearhead more quality-improvement efforts.

“For example,” says Dr. Gandhi, “the hospital has to document how it’s doing on pneumonia measures, acute myocardial infarction measures (was the patient having a heart attack given aspirin and a beta-blocker?), and others. These are fruitful topics for quality-improvement projects.”

Hospitalist Andrew Karson, MD, MPH, associate director of the Decision Support and Quality Management Unit and associate program director for the Internal Medicine Residency Program at Massachusetts General Hospital, Boston, also focuses on patient safety issues in his research. Given hospitalists’ knowledge of decision-making systems in the hospital, they are in a unique situation to initiate such projects, he believes.

For example, Dr. Karson participated in a study initiated by colleague Christopher L. Roy, MD, associate director of the hospitalist program at BWH.2 Dr. Roy posited that pending test results could be an important patient safety issue and, at the very least, might affect continuity of care. The researchers identified 2,644 consecutive patients discharged from BWH and Massachusetts General between February and June 2004. During that time, a mixture of hospitalists and non-hospitalists were responsible for discharging patients on house staff and non-house-staff services. Using a Results Manager application integrated into each patient’s electronic medical record (EMR) at the hospitals, the team identified and tracked pending laboratory and/or radiologic test results that had been returned after the patients were discharged.

The team used physician reviewers to determine whether the pending test results were potentially actionable. They found that 41% (1,095) of the discharged patients had a total of 2,033 test results return after their discharge. Of those tests, the physician reviewer determined that 9.4% (191) were potentially actionable. Examples of actionable results of which discharging physicians had been unaware included a levofloxacin-resistant Klebsiella infection in a patient being treated with levofloxacin, and a thyroid-stimulating hormone level that was dangerously low in a patient with rapid atrial fibrillation. A coauthor of the study, Eric G. Poon, MD, MPH, Division of General Medicine and Primary Care at BWH, is working on a results-management system that will automatically alert hospitalists and other physicians in the process of discharging patients when those patients are awaiting test results.

The next step will be to tie CPOE with bedside charting and decision support, so that verification of order sets will also be stored in the hospital system’s information warehouse and will be accessible to physician researchers.

Research Potential

CPOE systems afford opportunities to delve further into clinical research and QI projects.

The flow of care in hospitals is inextricably linked with writing orders—for medications, tests, consultations, or interventional care processes. “Interfacing with CPOEs, therefore, can help influence the way care is practiced more broadly for our patients,” says Dr. Karson. “By embedding rules and decision support elements within our CPOE systems, we can improve the quality and safety of the care that we provide.”

The effect of CPOE on ICU patient care was highlighted in a 2005 study conducted by intensivist Stephen P. Hoffmann, MD, medical director, ICU, and associate professor of medicine at Ohio State University Medical Center, Columbus, and his colleagues. The team compared orders for ICU care before and after modification of a CPOE system and found that use of higher-efficiency CPOE order paths led to significant reductions in orders for vasoactive infusions, sedative infusions, and ventilator management.3

Paul D. Hain, MD, interim chief of staff and director of the Pediatric Hospitalist Program at Monroe Carell Jr. Children’s Hospital at Vanderbilt, Nashville, Tenn., and his colleagues at Vanderbilt University’s School of Medicine have been able to use their institution’s advanced CPOE to increase adherence to evidence-based treatments.

With the help of IT support staff, Dr. Hain inserted a pop-up window into the CPOE to remind providers that bronchodilators (albuterol) and steroids are ineffective for the treatment of bronchiolitis. Working with a third-year medical student, Ryan Bailey, Dr. Hain compared orders for these treatments in the years preceding installation of his pop-up reminder with those afterward. There was a significant drop in the non-evidence-based treatments, he notes, based on the installation of the pop-up window. “The reminder actually worked!” he exclaims. “It got people to stop using inappropriate therapies.”

This type of quality improvement, says Dr. Hain, is good for the hospital, for the hospitalists, and for their non-hospitalist colleagues. “This type of reminder allows us to share evidence-based guidelines with other admitting physicians in real time, and it appears to be a much more effective way to communicate information, as evidenced by our success in decreasing non-evidence-based treatments for bronchiolitis,” he asserts. The pop-up window includes a link to the treatment guidelines, so it also offers users an educational opportunity.

Close the Loop

Dr. Hoffmann and others caution about the limitations of using CPOE data. Most CPOE systems, notes Dr. Hoffmann, do not have a way of capturing whether an order or intervention was actually carried out.

“With CPOE, you can get a very good handle on how many order sets for processes of care have been ordered, but it doesn’t complete the loop—it doesn’t tell you whether that process of care happened once it has been ordered,” Dr. Hoffman says. “If you use the CPOE data set alone and stop there, the process is going to be fraught with unreliable information.”

CPOE can be a good tool for organizing clinical improvement projects but may not be the perfect tool for verifying outcomes of the order set. This was underscored by a project Dr. Hoffmann and his team conducted in collaboration with the University HealthSystem Consortium (UHC) on ventilator-associated pneumonia (VAP). The team wrote policy and processes based on current evidence for preventing VAP—such as raising the heads of patients’ beds to 30 degrees when they are mechanically ventilated—and created a flowchart of those processes. The aim of the project was to tie these care processes to the order for a ventilator, so that each time one was ordered, the other care items were bundled with it to trigger changes at the bedside. Now, it won’t be possible for a provider to order a ventilator without at least reviewing and ordering the additional care processes.

For the UHC project, Dr. Hoffmann and his team had to manually review charts and documentation to verify that the VAP bundle had been ordered and then utilized. “We looked at what documentation needed to be done, and we have modified nursing and respiratory therapy documentation to ensure that all these bundled-care process steps are adequately documented,” he says. The next step will be to tie CPOE with bedside charting and decision support, so that verification of order sets will also be stored in the hospital system’s information warehouse and will be accessible to physician researchers. The project is ongoing, states Dr. Hoffmann, and another evaluation will be conducted manually after a six-month interval to validate the data collection points that will be most useful in the automation process.

There are other ways to verify CPOE implementation. At BWH, says Dr. Gandhi, the electronic medication administration record (eMAR) provides a powerful adjunct to the CPOE.

“The fact that we have eMAR data is really advantageous,” notes Dr. Gandhi, “because now we can actually tell what patients have received.” For example, she says, Narcan (naloxone hydrochloride) is usually ordered and kept at the bedside for a patient receiving opioids in case the patient develops respiratory depression. Before institution of the hospital’s eMAR, “we could never tell how much Narcan was actually being given without doing laborious chart reviews. Now, with our eMAR, we can easily track how many times it was given, and this supplies a much better indicator of potential problems with the use of narcotics.”

At Vanderbilt, says Dr. Hain, a dosage checker application installed behind the CPOE has allowed his colleague Neal Patel, MD, MPH, to verify that medication errors in the Pediatric ICU dropped dramatically after its implementation. But, on other projects, researchers must know in advance that they intend to follow up on order entries so that they can convert order entries into binary procedures. The CPOE and EMR systems have the capability of inserting text boxes, drop-down menus, and click buttons for verifying medications, procedures, or even safety check-offs. If the CPOE is not set up in advance for this feature, however, it’s back to manual extraction to confirm the data—“a painful process, just as it is from paper charts,” Dr. Hain notes.

Privacy and Other Issues

Are there privacy issues of which hospitalists should be aware when using their hospital information system databases for their research?

“In general, if you’re doing quality improvement projects solely for the sake of improving the quality of patient care at your institution,” says Dr. Karson, “you do not need IRB [institutional review board] approval.” Whenever hospitalists plan to publish or present the data to external audiences, however, prior IRB approval must be obtained to show that patients’ identities will be protected and that use of the data will cause no harm.

There could be wrinkles in following these guidelines if the results of a QI project reveal surprisingly good results or important lessons about quality patient care that researchers think are worth sharing. Although it is possible to apply post-hoc for IRB approval, Dr. Gandhi and others suggest obtaining approval prior to the start of the project if researchers think there is any chance they may want to share results externally. Researchers must also adhere to the quality rules during QI projects, asserts Dr. Hain, to make sure patients’ identities are protected.

The IT/MD Interface

Whether hospitals use off-the-shelf or custom-built, institution-specific CPOEs, hospitalists are well positioned to play important roles in enhancing their designs, believes Dr. Karson. “If you’re going to support [clinicians’] decisions with computerized decision support, then CPOE systems are a great way to broadly affect the care of patients,” he says.

As those CPOE systems are designed, they require decisions along the way so they will achieve the quality, safety, and efficiency goals for the hospital and for the patients that the hospital cares for. Who better to interface with information systems designers than process-oriented hospitalists? As a hospitalist, Dr. Karson is taking a lead role in updating the pneumonia order set in his hospital’s provider order entry system.

It is sometimes possible for hospitalists to extract data manually to effect a proof of concept as justification for an IT system upgrade, says Dr. Hain. For example, in Vanderbilt’s outpatient clinic, one physician wanted to know whether all diabetic patients received foot exams at their regular visits. They inserted a paper form with check boxes into patients’ charts and then aggregated these forms to show it was possible to track quality measures for diabetics. This has led to a diabetics dashboard on the outpatient clinic computers that tracks foot exams by the day, week, or month.

Hospitalists report varying degrees of expertise with IT. Dr. Hoffmann’s introduction to IT came when he assumed the medical directorship of Ohio State University’s ICU. Since that time, he has been charged with collaborating with the medical center’s information systems (IS) personnel to improve the CPOE. “We have a group here that embraces the system—so much so that the IS people sometimes are inundated with our enthusiasm to make changes,” Dr. Hoffman says.

Dr. Hain, who has a background in engineering, relies on IT support when designing changes to the CPOE. “Our IT department here has done a really good job of reaching out to its users,” he says. Several physicians in the medical informatics department specialize in the CPOE, as is the case in many academic institutions. “It’s important that the gap be bridged between computer programmers and MDs,” he says. “The best way to do that is to have MDs with master’s degrees in informatics working with the programmers, making it all the more seamless.” TH

Gretchen Henkel is a frequent contributor to The Hospitalist.


  1. American Hospital Association. Forward momentum: hospital use of information technology. October 2005. Available at Last accessed April 10, 2007.
  2. Roy CL, Poon EG, Karson AS, et al. Patient safety concerns arising from test results that return after hospital discharge. Ann Intern Med. 2005 Jul 19; 143(2):121-128.
  3. Ali NA, Mekhjian HS, Kuehn PL, et al. Specificity of computerized physician order entry has a significant effect on the efficiency of workflow for critically ill patients. Crit Care Med. 2005 Jan;33(1):110-114.

Next Article:

   Comments ()