A Health System Wide Evaluation of Clinical Decision Support Tools to Improve PDMP Utilization and Patient Outcomes

Prior research by this team has shown that when emergency department (ED) prescribers are less likely to prescribe an opioid analgesic to patients at high-risk of overdose or misuse when they consult the Colorado Prescription Drug Monitoring Program (PDMP). Despite extensive work to reduce workflow and accessibility barriers to use of the PDMP, the majority of providers still do not use the PDMP when making prescription decisions. Clinical decision support (CDS) that is embedded in the Electronic Health Record (EHR) has promise as a tool to increase PDMP use, advance presentation of important PDMP data to providers and improve opioid safety without interrupting provider workflow.

This pragmatic, randomized study is designed to determine if:

• CDS tools that deliver focused, timely clinical information about high risk prescriptions to prescribers can improve use of the PDMP
• The use of focused CDS tools can decrease high-risk opioid prescribing
• The use of focused CDS tools is associated with better patient outcomes: decreasing high-risk, long-term, and aberrant opioid use

Specifically, we will test the hypotheses that:

• Focused CDS tools will increase the percentage of high-risk prescriptions with a PDMP review, compared to the control group.
• Focused CDS tools will decrease the percentage of high-risk controlled medication prescriptions, compared to the control group.

The alerts will fire for discharge prescriptions in all system emergency departments, inpatient facilities and ambulatory care facilities. Evaluation of patient risk as described below will be conducted automatically and for two of the three visible alerts, the provider will only see the alert if the patient meets high risk criteria. Providers in emergency departments and inpatient facilities will be randomized at the individual level, while providers selected to participate in the ambulatory setting will be randomized so that providers who prescribe at the same site or set of sets ( such as a series of primary care clinics staffed by an identifiable set of providers) will be in the arm to reduce contamination between the groups. Approximately 4000 providers will be included, based on historical staffing of UCHealth facilities.

Health encounters will be identified on the basis of an encounter with an enrolled provider. Automated risk assessment steps will be carried out on patients seen at included UCHealth facilities in the course of the year-long activation of the alerts. Based on historical patient counts, this is approximately 950,000 unique patients. Visit data will be retrospectively collected. This population is expected to be approximately 200,000 individuals.

The CDS only changes the provider interface with the PDMP and does not include any treatment recommendations or limit providers in any way from providing opioids or benzodiazepines if that is the best course of action in the provider's clinical judgement. Providers are not required to check the PDMP. The CDS is designed to enable access to the PDMP specifically focused on high risk prescriptions.

As the study implements an automated CDS, which is routinely introduced into clinical practice with no consent process, and does not make any treatment recommendations or requirements, it is being conducted at the provider level under a waiver of informed consent. Patient outcome determination will be conducted as secondary research, with patients to be included identified on the basis of their interaction with an enrolled health care provider. Only data collected as part of routine care will be collected to measure secondary outcomes. All patient data will be de-identified by the research team's Honest Broker and a de-identified data set provided for analysis.

Two populations are included in the study description. Providers and the patients seen by those providers.

Providers are randomized to one of the four groups described previously.

Provider Inclusion Criteria:

• Licensed health care providers within the UCHealth system with a history (previous year) of prescribing opioids and/or benzodiazepines.

Provider Exclusion Criteria:

• Majority of prescribing is done at a facility outside the state of Colorado
• Provider specialty of oncology or pediatrics as recorded in the EHR
• Provider in an ambulatory clinic where a total of <50 opioid or benzodiazepine prescriptions were written in the year before study initiation

Patient Inclusion Criteria:

• Patient seen by a randomized provider within the UCHealth system who may receive an opioid prescription

Patient Exclusion Criteria:

• Less than 12 years of age or greater than 89 years of age at the time of the first study associated visit
• Active oncology or sickle cell diagnosis
• Active end of life care (defined as discharge to hospice or palliative care facility or specific palliative care orders in the medical record)