Communications of the ACM
Data Mining Technology Applied to Hospitals' Infection Control Programs
Credit: Data Into Results
Infection control within hospitals is a huge and growing problem. A recent study in the Archives of Internal Medicine found that conditions caused by healthcare associated infections (HAIs) killed 48,000 people and increased healthcare costs by $8.1 billion in 2006 alone.
In response, Medicare and private insurers are restricting payment for care of patients with HAIs. While many hospitals have made infection control a top priority, the historic challenge facing nurses, physicians, pharmacists and others charged with managing this problem has been combing through hundreds of records every day to identify trends or track individual patients.
All hospitals, even small ones, contain huge amounts of data in paper and electronic records, often contained in separate departments or software applications. Data mining technology, originally developed for use in business and scientific research, is now being used to automatically monitor clinical information among these disparate data sets including pharmacy, lab, admission/discharge and medical transcripts.
By applying algorithms, association rules, Bayesian classification and regression analysis to large data sets to identify patterns and meanings, hospital staff can spot trends that would otherwise be hidden. Data mining has been used for many years by business and government to analyze trends in massive blocks of data such as airline passenger records, census data and supermarket scanner information.
The overall problem of hospital infection control is complicated by the recent rise of multiple drug-resistant organisms (MDROs). Many hospitals have initiated antimicrobial stewardship programs which match the right antibiotic, at the correct dose, for the appropriate duration to cure the patient's infection, while minimizing the emergence of drug resistance.
According to Chun Wong, chief executive officer of Asolva, a developer of data mining software for hospitals, "Data mining of hospital information can help identify MDROs promptly. By linking pharmacy and lab results, infection control practitioners can quickly identify clusters of organisms found in the patient population and then administer the appropriate antimicrobial drugs."
Wong says his company's flagship product, Medici, was created in 2006 for clinicians and pharmacists to review a variety of patient data and make appropriate clinical decisions. The company's latest product, Orsini, adds a new capability: natural language processing. This enables clinicians to ask the questions they want in simple text and to obtain reports which identify trends across the entire patient population.
"This new system works much like a Google search. A hospital clinician can enter a phrase such as 'find all patients with abnormal creatinine who are on Vancomycin' and the Orsini engine will search across all of the hospital's data silos to create a report within seconds," Wong says.
"With natural language processing, clinicians don't need to enter software code or ask the IT department to develop a special report. Orsini dissects their query and decides how to get the best answer from relevant databases," Wong adds.
Asolva's software systems interface with all major electronic medical record and clinical laboratory systems. They can be used to provide a wide variety of reports to help meet Joint Commission and Medicare reporting requirements.
No entries found