Helping clinical leaders identify prescribing patterns across the health system
Making health system changes to address the opioid crisis is challenging if you don’t have actionable data about prescribing patterns. With guidance from their chief medical officer, the business intelligence team at Jefferson Health in Philadelphia took advantage of the recent enterprise EHR implementation to create dashboards highlighting potential prescribing issues. Internal EHR development teams have already made changes to address the issues they found.
In a recent interview, Cara Martino, enterprise business intelligence manager at 14-hospital Jefferson Health, said that Jefferson was one of the first health systems to implement its Epic EHR and Qlik Sense at the same time. Many organizations that go live on Epic use Crystal Reporting, she said. “It was the combination of the two that allowed us to do this project,” she added. Before going live on Epic and using this new BI tool, Jefferson clinicians used a bunch of disparate systems for ambulatory, inpatient and emergency department physicians. “We weren’t able to aggregate the data,” she recalled.
Jefferson departments had tried to look at opioid prescribing previously, but “they were trying to manually abstract opioid data from the system, compile it in Excel, and try to see trending across the continuum of care, but they really weren’t able to do that very well,” Martino said. “Once we went live with Epic, we were able to standardize the work flow of entering the opioid order and prescription, and Qlik Sense allowed us to look at it from that enterprise view and to slice and dice the data a bunch of different ways.”
In response to the opioid epidemic in the Philadelphia area, Jefferson has created an opioid task force. In January 2018, Edmund Pribitkin, M.D., Jefferson’s chief medical officer, came to the BI team because he saw they didn’t really have good data to share. “We were bringing physicians and pharmacists together in a room to think through what we can change to address the opioid crisis,” Martino said, “but they didn't’ have baseline data to understand where we are going wrong.”
Pribitkin asked the BI team to develop a high-level dashboard with five key metrics to share at the clinical chair department level:
• Prescriptions by provider and service;
• Number of orders with over 100 pills;
• Prescriptions written for more than 7 days, 5 days and 3 days;
• Patients with two or more opioid prescriptions within 30 days; and
• Morphine-equivalent daily doses over 50.
The metric around patients with two or more prescriptions within 30 days had been difficult to see when data was being entered in a lot of different ordering systems, but is much easier now because everyone is ordering through Epic.
Within a month, the BI team had a prototype created. In the dashboard, you can pick a patient with the most scripts, and below it populates the different providers who are prescribing for them. “Sometimes you pick someone on the higher end and only one person is prescribing to them. Usually it is a chronic pain clinician, and it is probably appropriate,” Martino said. “But then you pick someone who has six different providers providing them opiates. They may have no idea about each other. Now we get to have that conversation about what we can do in the EHR to notify them at the time of ordering that the patient already has something prescribed. Also, our quality department can notify these six providers via e-mail. They might ask whether we should refer this patient to a chronic pain physician. If they really need this much pain medication, are we treating them the right way?”
Showing clinical leaders a trending graph such as prescription orders of 7 days or more can have an impact on ordering trends and EHR recommendations. Martino, who is a nurse herself, described an example of how her team worked with clinician informaticists who build EHR workflow tools to make some key changes.
They saw in the data that there were examples of doctors were prescribing six opioids to the same patient within two minutes. On further examination, they found that clinicians were entering orders incorrectly and quickly canceling them. But the e-prescription interface with the pharmacy system wasn’t necessarily canceling the incorrect orders, and the patients could have six prescriptions waiting for them at the pharmacy.
“Our first step was to get a message out to providers to let them know if you are prescribing and sending to an outside pharmacy, you need to pick up the phone and talk to the pharmacy about a change in an order and make sure they know which is the correct one,” Martino said. They also worked on the pharmacy interface to generate those canceled order messages, so that if something is ordered and then canceled within 30 seconds, it alerts the pharmacy not to prepare it.
The BI team saw another EHR change that they thought might make a big difference in prescribing patterns. In the EHR quick buttons, duration of the prescription defaulted to 10 days. The quick button on order days had choices of 7, 10, 15 or 30. “We knew those were too high,” she said. “We wanted 3, 5 or 7.”
They worked with an emergency department doctor who was also an informaticist to pilot that change in the ED. “In the first month, we saw a drop from 32 scripts over 7 days to just two,” Martino said. The following month they saw a reduction of more than 65 percent in prescriptions of more than 7 days. “The best part is no one complained or said they were looking for the 10 button and it wasn’t there,” she added. Now that change is being rolled out enterprise-wide.
Although there are some other requests from clinicians for data about opioid prescribing, Martino said the BI team’s current goal is to help more clinicians gain insights into the data that already exists. “It is really hard to go from an institution that was not data-rich and had disparate systems and manually collected data to one that has one EHR and a robust BI tool that allows us to slice data in a million different ways,” Martino said. “We are trying to get the clinicians comfortable with the tools. There is so much we can do, but we have to get the data to the right people and educate them about how we are pulling it out of the EHR. It is easy for this not be a priority for clinicians because they are so invested in taking care of their patients. It is our job as informaticists to go to their meetings and talk about changes we can make. It is also important for us to have clinicians on our team and to be able to offer that informaticist viewpoint.”