If the scientific data going into these repositories is flawed, so are your clinicians' decisions.
Garbage in, garbage out is a common expression that's especially relevant to health IT. The quality of the data that goes into an e-prescribing program or clinical decision support system determines the accuracy of the diagnoses and treatment decisions coming out of your doctors and nurses.
Two examples illustrate the need to keep the GIGO mantra front of mind. One of my colleagues recently told me that he'd been prescribed a statin for a heart condition. Statins are quite effective in lowering serum cholesterol but they have some drawbacks. I mentioned the fact that statins can cause myopathy, a catch-all term for a variety of muscle complications, to my co-worker but he quickly dismissed that concern because his physician told him myopathy rarely occurs.
My previous experience as a nutritionist working with cardiac patients, plus conversations with several friends who have used statins, suggest otherwise. But when you look at randomized clinical trials involving statins, you get the impression that muscle-related problems occur in 1% to 5% of patients on both statins and placebos, indicating that it rarely happens, if at all.
So chances are the e-prescribing tools your clinicians use are telling them not to be too concerned. But a deeper dive into the research literature uncovers the fact that in everyday clinical practice statin-related myopathy might affect as much as 20% of patients.
Why the discrepancy? Randomized double-blind clinical trials often have exclusion criteria that weed out older patients, women, and patients with co-existing disorders that might increase the risk of muscle complications. In everyday clinical practice, physicians can't cherry pick their patients to find those who are least likely to develop complications, so they get a more realistic sense of how common myopathy is.
Think for a moment about how these divergent stats can affect clinical outcomes. If doctors and patients are "certain" that myopathy is nothing to be concerned about, they're less likely to follow up on suspicious signs and symptoms. If early symptoms such as muscle pain are ignored, they can lead to a more severe complication such as rhabdomyolysis, which is potentially life threatening. If, on the other hand, caregivers realize that as many as one in five patients are at risk, they're obviously going to pay more attention.
Here's another example worth pondering: For decades, U.S. clinicians were taught that gluten intolerance--celiac disease, the inability to digest the protein in wheat and certain other grains--was rare, affecting as few as 1 in 10,000 Americans. But in recent years, the introduction of a specific diagnostic test for the disorder, coupled with more epidemiologic studies, indicates it affects as many as one in 250 Americans and Europeans.
If your providers are relying on old data, they're likely missing a lot of these diagnoses, which means unhappy patients will leave your practice looking for the help they need from a clinician who's up-to-date. And for those patients too timid to look for a second opinion, odds are they'll continue to suffer all the gastrointestinal symptoms, anemia, and weight loss that characterize gluten intolerance. That in turn translates into higher costs as patients undergo more diagnostic tests in the search for answers.
So the question that IT managers must ask is: How good is the information going into our databases? If your organization is compiling its own knowledgebase from which to make diagnostic and treatment decisions, are your expert reviewers looking at all the relevant research? If they rely solely on Medline, they're missing a lot of useful data. Although Medline covers most of the important healthcare journals, it doesn't cover them all. Other valuable resources include EMBASE, Micromedix,Medi-Span, and the FDA's Adverse Event Reporting System (AERS). Of course, evaluating these data sources to find the best programs to populate your IT system is an entire discussion in itself, which I'll cover in my next column.
Even if your organization is using all the best data sources, how often do you update your database with the latest clinical research? If your knowledgebase is being updated once a year, you might already be falling behind. And patient care will suffer.
If, on the other hand, you rely on a vendor to provide order sets for a computerized physician order system, for example, do you trust its data-collecting skills? What sources is the vendor tapping?
The best vendors comb hundreds of thousands of research reports, clinical guidelines, and symposium proceedings. They have a stable of top-notch experts to review this content, cherry picking the best of the best. Zynx Health, Provation, and UptoDate come to mind.
Similarly, you want the vendor's product to integrate smoothly into your facility's systems, using standards such as HL7 and InfoButton API, and the right terminology, including SNOMED CT, RxNorm, LOINC, CPT, and ICD-9.
So if you've noticed your patient satisfaction surveys are less than glowing or see a growing number of patients leaving your practice for no apparent reason, it's time to get your clinical databases up to speed. It's worth the time and expense.
When are emerging technologies ready for clinical use? In the new issue of InformationWeek Healthcare, find out how three promising innovations--personalized medicine, clinical analytics, and natural language processing--show the trade-offs. Download the issue now. (Free registration required.)
InformationWeek Must Reads Oct. 21, 2014InformationWeek's new Must Reads is a compendium of our best recent coverage of digital strategy. Learn why you should learn to embrace DevOps, how to avoid roadblocks for digital projects, what the five steps to API management are, and more.