January / February 2012
Taking CDS to the Next Level for Medication Management
Consider the following scenario. A patient in a hospital acquires pneumonia. When the attending physician enters an order for the antibiotic ceftriaxone, an alert pops up indicating that the patient is allergic to the medication. Based on the information provided, the physician chooses another appropriate medication.
In this case, clinical decision support (CDS) did its job. It eliminated potential harm to the patient by providing an alert at the point of order. And while this scenario might seem to be a success, what the physician did not know is that extenuating circumstances during the patient’s stay rendered the replacement antibiotic ineffective.
The provision of decision support tools is an important first step to improving patient care. But increasingly, clinicians are discovering that there are two notable limitations to its use for medication management: 1) lack of data rich enough to deliver informed alerts to clinicians and 2) the inability to monitor patient activity after an initial order is made.
As healthcare organizations continue to focus on enhanced quality metrics intended to improve effectiveness of care and patient safety, medication decision support in particular will need to take on an expanded role that not only aggregates patient data throughout care in real-time but also provides contextually relevant data regarding the patient’s most current status. This way, care guidance can be effectively applied based on this data. This is a critical next step to address medication management needs and to meet expectations for quality and safety going forward.
The Challenge of Alert Fatigue
It’s no secret in the industry that alert fatigue has become a problem for physicians and other care providers. As a result, many have learned to ignore or bypass electronic warnings at the point of care because the data often lacks relevance.
Study after study confirms this assertion. For instance, consider a 2009 study conducted by researchers at Dana-Farber Cancer Institute and Beth Israel Deaconess Medical Center. A review of the electronic prescriptions and associated medication safety alerts generated by 2,872 clinicians revealed that 90% of the drug interaction alerts were overridden, along with 77% of the drug allergy alerts.
The reason? The vast majority of the alerts were for a potential interaction with a drug a patient was already taking. In these cases, the CDS alerting system did not recognize that the patient was already tolerating the medication in question without any notable adverse effects related to the interaction.
Treating patients is complex. It is both art and science. Physicians who have been working with patients over time often understand the nuances of their care better than the alerts provided by current CDS applications. Alerts often are based on general population knowledge rather than contextual information that is relevant to a particular patient. Physicians frequently complain that CDS is simply not able to provide the types of alerts that can really make a difference in care because the systems aren’t intuitive enough to compile relevant information in a meaningful way.
The end result is that care providers begin to ignore all alerts. Of course, the risk here is that some of the alerts provided may be quite relevant and have dramatic implications on the patient’s well-being. Alerts based on clinical rules that pull together many data elements can often detect problems that a busy clinician may miss.
The Quest for Richer Data
One approach to advancing physician acceptance of care alerts is to make the CDS applications more intuitive. Though improving, current CDS is limited in its reach and analysis of patient data. Most CDS tools are not able to consider all the contextual data that exists—such as labs and prior history—to present the most relevant information to physicians.
Consider that an alert pops up indicating a prescribed medication shouldn’t be used in conjunction with another medication a patient is already taking. In reality, there are very few situations that absolutely demand the avoidance of using one medication alongside another. Most often the situation will require the clinician be aware of the interaction and consider methods for reducing risk, such as monitoring for any signs of undesired effects.
For one Michigan-based health system, pharmacists received a total of 347 alerts over a 24-hour period for any combination of potassium, ACE/ARB and K sparing diuretics, notifying clinicians for the risk of hyperkalemia. Currently, pharmacists are overriding these alerts because this risk, while real, is not immediate in most cases. The organization fully expects physicians will have the same reaction to alerts that will appear in the new CPOE system. The current quandary is to find a way to make the alerts more relevant and clinically significant, by drawing from better patient data and criteria.
Further, most CDS alerts appear at the time of order entry (i.e., new drug order), with no additional follow-up over time regarding changes in patient status. This lack of ongoing monitoring limits the efficiency and effectiveness of medication management because it does not take into account the dynamic nature of a patient’s condition.
To address these issues, many hospitals are deploying advanced surveillance technology to enhance the scope of CDS at both the time an order is made and after dispensing.
The Promise of Surveillance Technology
Surveillance technology enables the aggregation of richer data covering the overall scope of care and ultimately rendering an alert that is more relevant to the patient’s true condition. With surveillance data, the previously-mentioned health system was able to create preconfigured CDS rules that incorporate richer data and more contextually consistent data.
Using the surveillance technology, a specific clinical rule was written to identify patients experiencing a real change in serum potassium and alert clinicians for follow-up, rather than sending an alert out across the board for any combination of the three medications. The rule addressed patients whose lab results fell within specific parameters of the combination of potassium, ACE/ARB and K sparing diuretics. It also alerted pharmacists of patients whose serum potassium increased by a particular percentage or rose above a defined level. The rule could be further refined to consider the patients renal function, another risk factor for the development of hyperkalemia.
Combining surveillance technology with point-of-order alerting provides flexibility for care providers in determining how to best structure alerts and CDS rules. Some types of CDS interventions may be better captured and managed using the surveillance approach. This may reduce unnecessary or non-critical alerts during the order entry process while ensuring ongoing proper care management. Ultimately, this approach may be one strategy for improving provider confidence in CDS alerts.
Most CDS applications are also limited in that once an alert is presented at the time of order, there is no follow-up regarding the patient’s status or condition. Patients are dynamic in nature and need ongoing monitoring. Down the road three days, something may change, and the initial alert may become more relevant.
Surveillance systems lend themselves to this kind of ongoing monitoring. The availability of smart logic contained within the technology can constantly monitor the lab results as well as other patient factors that can lead to improved decision-making about the best medications and the right dosage. Because information is available in real-time, these systems enable rapid analysis of patient needs, driving the best clinical decisions in a timely manner.
The rules engine available in automated surveillance systems enables data elements to be analyzed and processed in predefined ways. This infrastructure provides the basis to write a rule that addresses a specific question that is constantly monitored by the system.
One example of a basic question might be, “Which of my patients on the anticoagulant (blood thinner) warfarin has an INR value outside the target therapeutic range of 2 to 3?” This rule would identify patients at possible risk for bleeding or blood clot formation (ie, DVT) and would trigger investigation and possible therapeutic adjustments. Another somewhat more complex rule example is “Identify patients with recent microbiology culture results that are on antibiotics that are not effective for the bacteria present.” This would ensure a timely review and possible change in the antibiotic to ensure eradication of the infection.
These clinical rules can be defined to make alerts more meaningful and provide better care direction. They can also address specific issues related to quality measures in a hospital. For example, one of the CMS core measures is that all appropriate patients admitted and diagnosed with a myocardial infarction (“heart attack”) receive an ACE inhibitor medication. To improve compliance with this requirement, a rule can be written to identify patients that have had a heart attack (based on diagnosis or lab values) who are not on an ACI inhibitors.
Clearly, there are high expectations that information technology can and will greatly assist care providers and systems in improving the safety and quality of care. Current CDS alerting systems often generate an unacceptable rate of alerts that have little to no clinical value. Going forward, improving alert quality, appropriateness, and relevance will be key. If clinicians are desensitized to alerts provided through CDS, the ability of a health organization to improve quality and drive the practice of evidence-based medicine will be limited.
Surveillance technology offers clear advantages in delivering timely and contextually relevant CDS to improve care in areas such as medication selection and management. These systems have the ability to provide ongoing monitoring of a patient’s care and ultimately the most effective outcomes.
Steve Riddle is vice president of clinical affairs with Pharmacy OneSource, part of Wolters Kluwer Health. He can be reached at firstname.lastname@example.org.