- In 2016, just under 50,000 people died of an opioid-related overdose.
Illegal substances such as heroin and fentanyl have taken an increasingly heavy toll on public health since the middle of the decade.
But the National Institutes of Health points out that close to 20,000 deaths in 2016 were directly attributed to overdoses on prescription opioid pain relievers.
Some of the opioid prescriptions involved in these deaths were written to help patients recover after accidents or surgeries.
Others started out as part of a well-intentioned pain management program for individuals with chronic pain.
Many were written for another person entirely, but improper safeguards left bottles of pills open to theft or misuse.
No matter how they started, these prescriptions all ended in a fatality, leaving healthcare providers struggling with the question of how to treat a person’s pain in a safe, effective, responsible manner.
Opioids don’t always have to be the first answer for patients in pain, an HHS task force recently stressed in a new draft national framework for pain management.
But when prescription pain killers are called for, providers are in need of data-driven tools and predictive analytics that can guide their decision making, alert them to potential risks, and foster meaningful discussions about care with patients searching for relief.
Healthcare providers expect their electronic health record (EHR) vendors to take a leading role in providing them with those decision-making resources, a KLAS survey from October 2018 revealed.
Seventy-six percent of provider participants said that opioid safety insights, such as clinical decision support (CDS), risk assessment tools, and prescription drug monitoring program (PDMP) data should be available directly within the EHR workflow.
Sixty percent of respondents said their EHR vendors are in the best position to deliver this data to providers at the point of care – yet only a third stated that they are currently satisfied with the capabilities they have to identify risks and take steps to prevent addiction.
At Cerner Corporation’s annual user meeting the same month, CEO Brent Shafer unveiled his company’s approach to delivering prescribing data into the workflow to improve opioid safety.
The Opioid Toolkit offers a blend of better access to existing data and personalized clinical decision support for individuals, explained Eva Karp, Senior Vice President and Chief Clinical and Patient Safety Officer at Cerner.
“You’d have to be living under a rock not to know that opioids are a problem for this country,” Karp told HealthITAnalytics.com. “Providers are much more alert to it now, as is the public. We know that we have to do something to address the roots of the problem.”
“There have been patterns of prescribing that contributed to a lot of the national issues, and we need to get better about how we resolve for pain. From our perspective, there is a huge opportunity to leverage the data we have to recognize patterns, identify what works, and equip providers to better intervene for patients.”
Navigating complex pain management relationships
Understanding suboptimal prescribing patterns requires a partnership between a vendor and its clients, Karp said. Provider organizations have treasure troves of patient data at their disposal, while health IT companies have the technical resources and expertise required to help providers mine that data for actionable insights.
“Since the beginning of the decade, we’ve been working very closely with providers and developing within our EHR capability to support clinicians as they identify risk factors, prescribe more safely, and manage populations outside of the four walls of the hospital or clinic,” she said.
“We’ve discovered that the underlying challenge is how to meet a person where they are, if they have chronic pain, without raising the risk of developing or exacerbating a substance abuse problem. The toolkit we’ve created is intended to support safer prescribing practices so that physicians and nurses receive alerts that can help them prevent a problem or manage an issue proactively.”
The first step is ensuring that expectations for pain control are clear and accessible to both patients and their providers, Karp said.
“Most chronic pain patients and providers have a pain management agreement that speaks to the issues of treatment,” she said. “The agreement is a document, signed by both parties, that outlines the plan for treatment, the risk factors that come with pain management, and the importance of not approaching other providers for pain medications.”
Like many other important but ancillary medical documents, a pain management agreement isn’t always transferred between disparate organizations when – or if – a patient’s records are shared.
A missing agreement can leave a dangerous gap in the new provider’s knowledge, potentially leading to accidental overprescribing or the opportunity for a patient to request more pain medication than is medically safe.
Cerner has integrated access to existing pain management agreements into its population health management platform, Karp said, allowing all its customers to have a clear idea of where the patient stands on his or her pain management treatment pathway.
“If a person presents to an acute care facility today and he already has a pain management agreement with a specialist at another organization, the acute care providers can access that document so that they can remain in step with what the patient’s other providers are delivering,” she said.
“We have also incorporated an integration with PDMP data and with NarxCare from Appriss Health so that the provider doesn’t have to spend time seeking out whether data from those sources exists,” she continued.
Presenting this data within the workflow can alert providers to other active prescriptions, even those that were written by providers across state lines.
“Purposeful drug seeking across state lines isn’t an everyday occurrence, but it’s certainly possible that someone would go to different prescribers in different regions to access narcotics,” Karp acknowledged.
“And we know that we still have some work to do before all of the PDMPs can communicate with each other.”
“So whatever we can do to present the provider with data in the workflow is important. It significantly reduces the hunt-and-peck that frustrates a lot of providers, and it ensures that this data is seen when and where it needs to be.”
Developing meaningful clinical decision support
Balancing access to important data with the need to reduce cognitive clutter can be a challenge for health IT developers. Overcrowded interfaces and annoying, low-value alerts can increase EHR dissatisfaction and contribute to another worrisome epidemic: burnout and exhaustion among clinical staff.
Cerner is among the many health IT vendors that have made adjustments to their user interfaces in order to meet growing demands for cleaner, more intuitive workflows.
“We’re constantly thinking about and refining what kind of information the provider needs within reach and how to present that data when it’s relevant,” Karp said. “When you’re prescribing, it’s helpful to have decision support that can create a safer environment without being a noise factor.”
“And if you’re someone administering a medication, you need to be presented with an alert if you’re giving a dose that’s too close to the previous one so that you can use your clinical judgement or ask the right person about whether it’s advisable to continue or not.”
Setting the right thresholds for when alerts should fire and when notifications can be overridden is a key component of making CDS tools helpful instead of harmful, she continued.
“We’ve worked with organizations on how they’re setting their decision support tools to monitor low and high dosing – we give them reports on how often those alerts are overridden so that they know if they’re creating valuable notifications or if they need to adjust the parameters.”
“Ultimately, organizations do tend to find a sweet spot where their users don’t perceive alerts as noise, and that’s because those types of messages meet the criteria of being the right data at the right time to the right person.”
The process can involve trial and error, and organizations should look closely at override statistics and patient outcomes while incorporating user feedback about functionality and flow.
“CDS to support safe prescribing is generally very well-received,” Karp observed. “As providers incorporate these tools into their practice, they get access to a lot of valuable insights on their own performance that they didn’t have before.”
“They can look at how their prescribing rates compare to their peer groups, and how their patients’ outcomes compare to similar providers. They love having that kind of data to help support their choices and provide a basis for making changes if they need to reevaluate some of their practices.”
Applying artificial intelligence to opioid safety
Applying artificial intelligence and machine learning to the task of opioid safety isn’t so much a conscious decision as an inevitability, Karp indicated.
“You can’t get a meaningful view of your performance without very rich and comprehensive analytics,” she said. “And increasingly, you can’t really generate the expected level of insights without turning to machine learning and artificial intelligence.”
“AI and machine learning have been a big focus for Cerner, and we can very clearly see the potential of these strategies to deliver the insights we need to address the opioid epidemic – there are patterns here that are far too complex for any human to identify, but we can now do so much with algorithms that we couldn’t do before.”
Artificial intelligence will be an integral part of Cerner's ongoing effort to improve prescribing safety and inform clinicians, Karp said.
“We’re planning to release a substance use disorder risk algorithm in 2019 that will identify individuals at elevated risk of SUD much more proactively than other methods so that providers can work with the patient on alternatives or pair them with the right intervention for their needs,” she said.
“Artificial intelligence and machine learning are a key component of how we’re looking to advance clinical practice, and we are laser focused on applying those strategies to the opioid substance abuse issue.”
Creating highly usable technologies with the ability to augment provider decision-making will be the key to tackling inappropriate prescribing and overutilization of opioid therapies, Karp predicted.
“We work closely with the EHR Association, ONC, and CMS on policies related to prescribing in addition to collaborating on the challenges of provider workflows and the clinical experience in general,” she said. “The ONC has definitely made it a very strong focus to reduce burdens on providers, and we believe technology is critical for supporting all of those objectives.”
“Part of the puzzle is figuring out how to elevate safer practices through standardization and measurement while enabling providers to make good decisions that are informed by timely and accurate information. As we make adjustments to our strategies and learn from them, we are confident that these tools will advance care across the entire continuum.”
Editor's note (1/4/2019): This article has been revised to better reflect the role of artificial intelligence in Cerner products.