Data Management

Database Management: Post-relational environment provides intelligent patient care apps

Partners Healthcare Systems is using EMAR to make better decisions about patient care


To avert the possibility of costly medical malpractice lawsuits, healthcare professionals these days are taking extra precautions to ensure patient safety. For many hospitals, it means implementing barcoding on medications, patients, and even nurses as a safety check to ensure patients are being given the right drugs. Boston-based Partners HealthCare System, Inc. is adding another layer of protection by augmenting its medication-administration application with sophisticated medical intelligence to help doctors and nurses make better decisions about patient care. The EMAR (Electronic Medical Administration Record), scheduled to roll out in January 2004, is just one example of the many IT innovations at the core of the multibillion-dollar healthcare network’s patient care systems.

The Partners Healthcare System network includes community hospitals, medical centers, specialty healthcare facilities, community health centers, and more than 4,000 primary care and specialty physicians throughout Massachusetts. Though Partners uses Oracle and SQL for its data warehousing, data mining, and OLAP applications, it selected CACHÉ from Cambridge, MA-based InterSystems Corp. as the cornerstone for its clinical applications. “We make a division between our mission-critical patient care systems and other applications. We chose CACHÉ for the high availability, high volume demands of our transactional patient care systems because that’s where it excels,” said Steve Flammini, CTO for Partners HealthCare System.

CACHÉ is both a highly scalable, post-relational database and a Web services application development environment, explained Flammini. What makes it particularly well-suited to Partners’ mission-critical clinical applications is that the algorithms and schemas used to store the data are extremely efficient. “The sparse matrix storage method makes the performance of the database breathtaking compared to its relational counterparts,” stated Flammini.

Because Partners deploys its applications over the Web, the efficiency of the CACHÉ technology enables the organization to deliver high-performance transaction processing on a massive scale. “The product’s distributed protocol significantly reduces network traffic,” explained Flammini, “which means faster transaction processing. It also makes it possible to scale up to serve many thousands of users without sacrificing performance.”

Paul Grabscheid, VP of Strategic Planning for InterSystems, pointed to another aspect of CACHÉ technology that makes it a particularly appealing tool for developing patient care applications. CACHÉ has an underlying way of looking at data that Grabscheid claimed is richer and more flexible than traditional relational database systems. “The clinical side of healthcare is characterized by pretty complex data,” stated Grabscheid, “Lots of data doesn’t fit well into a two-dimensional structure.” According to Grabscheid, CACHÉ can look at information as two-dimensional, like tables, but also in three or four or 10 dimensions if necessary.

Unifying access to all patient information
CACHÉ is the database foundation for hundreds of applications running on Partners’ technology infrastructure. One in particular—the Clinical Data Repository Viewer (CDRV)—leverages CACHÉ’s ability to seamlessly integrate information retrieval from multiple data repositories. With the CDRV, primary care physicians can instantaneously look up patient lab results over the Web, whether those results come from the radiology imaging platform or the clinical data repository that houses all the other clinical details about a patient in the Partners’ network. Because CACHÉ is both an agile Web services platform and an XML server, Partners could build an application that creates interoperability between the two discrete platforms, giving end users the impression that all the reports are coming from a single source. “Through an XML interchange,” explained Flammini, “the CDRV application can pull up lab results or operative notes or discharge summaries from the clinical data repository or message the radiology imaging platform to display X-rays to the doctor. From the doctor’s perspective, it’s all one application.”

To understand the sense of scale, the clinical data repository currently houses over 380 million lab results and 11 million radiology reports. Physicians and other healthcare professionals query the warehouse about 68,000 times a day through a secure intranet. “The CDRV, like most of our clinical applications, requires 24/7/365 criticality and high response time because doctors are using them while they’re actually caring for patients,” said Flammini. “CACHÉ shows itself very well in this space.”

Intelligently controlling patient med dispensing
The latest application under development in the Partners patient care arsenal is a wireless Electronic Medication Administration Record (EMAR), scheduled to pilot in early 2004 at Boston’s 700-bed Brigham and Women’s Hospital. The application will replace the paper trail nurses now use to document every medication they give to a patient. But unlike the paper log, the electronic program will incorporate a host of advanced decision support and safety checking features. Drugs stored in the pharmacy dispensing machine will be barcoded so that nurses can scan the containers to make sure the dosage is correct. Patient ID wristbands will also be barcoded so that nurses can scan the ID to make sure they’re administering the drug to the right person. And once the drug is administered, the nurse scans his or her own ID badge to provide a complete audit trail of the medication process.

“The biggest check is pharmacy,” explained Linda Winfield, registered nurse and corporate team lead for clinical systems management at Partners. “Once the physician enters the medication order into the system, it needs to be checked by [the] pharmacy [system] before it’s available for the nurse to administer.” Some of the intelligence applied at that level will include such vital issues as possible adverse reactions between medications, contra-indications due to patient allergies, and potential overprescribing based on previously administered medications.

The EMAR will also include multiple electronic reminders to the nursing staff, such as visual cues that a drug administration is overdue or a reminder to record the patient’s pain scale to determine if a pain management drug is being effective. “Adding a layer of electronic checking doesn’t replace the nurse thinking and making determinations,” insisted Winfield. “It’s just that extra check that adds to everybody’s peace of mind that we’re tightly focused on patient safety.”

EMAR will be rolled out on wireless laptops with wireless barcode readers, giving nurses the freedom to travel with the computer to the patient’s bedside. Winfield predicted that bringing the scanning as close to the patient as possible will help avoid potentially life-threatening errors. As an example, even if a physician enters a countermand order after the drug has been picked up at the dispensary, the nurse will be alerted on the laptop screen, the second she scans the patient’s wristband before administering the medication.

“It’s one thing to develop applications with the intelligence to support medical policy,” shared Flammini. “But it’s another thing to really deploy them across thousands of physicians where the benefits of the decision support can really play out on a network-wide scale. CACHÉ has the flexibility and scalability that enables us to innovate at that level.”

Improving workflow between healthcare and benefits providers
With insurance companies and pharmaceutical companies so inextricably linked to the delivery of healthcare services, Flammini plans to create applications in CACHÉ that will enhance the workflow between Partners HealthCare and third-party organizations. One example he offers is designing a real-time query between Partners and insurance providers’ systems to ascertain benefits eligibility. “That has impact on your ability to collect revenue for the encounter as well as improve the overall workflow,” said Flammini.

Flammini also sees potential for developing applications that would link Partners’ physician order-entry system to pharmaceutical company databases to discover whether certain drugs are covered by certain payers and even certain plan types. “By actually building this feature into the prescribing flow, you’ve gone a long way toward improving the workflow around the prescribing process and ensuring that Partner’s gets reimbursed for the drugs being prescribed,” stated Flammini.

The pros and cons of choosing CACHE
When pressed to cite any downsides to working in the CACHÉ environment, Flammini did note that despite its widespread prominence in the healthcare industry, the database technology and programming environment aren’t nearly as well-known by developers as tools by companies like Oracle and SQL. Flammini was quick to point out, however, that there are languages like Java that ride on top of CACHÉ that can insulate it from the developer. So unfamiliarity with the CACHÉ language ultimately becomes a non-issue.

Flammini was quicker to cite the benefits of CACHÉ:
  1. Because it uses open standards, it can interoperate with a variety of platforms. So it’s a great asset when implementing Web services to integrate disparate systems.
  2. It’s a lightweight platform with a low cost of entry. You can achieve mission-critical, enterprise-level performance from commodity hardware, so you don’t need a huge capital investment to get up and running. But you can run on big iron if you need it.
  3. It requires a lot less management, care-and-feeding, and systems administration than comparable Oracle and SQL platforms. System support staff on the CACHÉ side is approximately a third less than on the other two platforms. So CACHÉ is very accessible and very cost-effective to maintain the environment.

“It’s really in that super mission-critical space, where physicians are using the computer directly, and there’s a level of vital clinical decision support,” said Flammini. “We found CACHÉ to be the best overall synthesis of high performance and high scalability that is essential in our high availability systems.”

Editor's Picks