Study to Validate Coded Medical Terms Used to Identify Opioid-Related Overdose in Databases Used for PMR Study 1B

Member Companies of the Opioid PMR Consortium

Status: Completed

Conditions: Opiate Addiction; Opioid-Related Disorders; Drug Abuse; Narcotic Abuse

Treatments: Other: Algorithm to determine overdose from opioid abuse

Study type: Observational

Funder types: Other; Industry

Identifiers

NCT02667197

3033-6 (Other Identifier)

Observational Study 3033-6

Details and patient eligibility

About

The purpose of this study is to determine reliability of codes and data from electronic medical records to predict and measure overdose and death in patients prescribed opioid analgesics. The study will compare this electronic data to data manually obtained from medical charts.

Full description

As part of a series of post-marketing requirement (PMR) studies for extended-release (ER) and long-acting (LA) opioid analgesics, the Food and Drug Administration (FDA) is requiring New Drug Application (NDA) holders of ER/LA opioids to conduct studies to estimate the incidence of misuse, abuse, addiction, overdose, and death among patients with chronic pain using long-term opioid therapy, and to validate the measures used to estimate the incidence of these adverse events.

The purpose of this study is to validate the measurement of opioid overdose events using diagnostic codes and data extracted from notes written in the electronic medical record (EMR), accompanied by diagnostic algorithms, to be used in a study of the incidence and predictors of opioid overdose and death (PMR Study 1B) among patients prescribed opioid analgesics. Diagnostic codes, accompanied by diagnostic algorithms, will be compared against manually abstracted medical chart reviews.

Code-based algorithms will be useful for identifying opioid overdoses in claims-based systems that include only coded data and will also find applicability in systems with EMRs. Code-based algorithms will be improved with text search of EMR clinical notations using Natural Language Processing (NLP) to identify overdose events not identified by diagnostic codes and to differentiate between intentional and unintentional overdoses. Yield from the resulting EMR-based algorithm will again be compared against manually abstracted medical chart reviews.

This EMR-based algorithm will be useful for identifying opioid overdoses in systems with EMRs, and for further differentiating between the causes of different types of overdoses. For example, overdose events can be due to misuse (e.g., therapeutic use not as indicated by a clinician), medication errors by patients, medical errors made by prescribers, abuse by patients, abuse by non-patients feigning to be patients in order to receive medications; and suicides. Overdose events therefore differ in intentionality, that is whether the person was attempting suicide or not. Unintentional overdoses can occur as a result of various causes, including misuse (therapeutic use but not consistent with clinician orders), abuse, adverse reactions to medications, anesthesia, and medication errors-both patient and provider-based. In addition, the distinction between unintentional and intentional overdoses can sometimes be unclear. This validation study will attempt to differentiate overdose by intentionality using both code-based algorithms and NLP-enhanced algorithms.

Currently, administrative databases use ICD-9 codes for nonfatal diagnoses and ICD-10 codes for fatal events. In October of 2015, ICD-10 codes are scheduled to replace ICD-9 codes for nonfatal diagnoses in administrative databases. This study will validate existing ICD-9 codes so that the study can meet the FDA-required timeline for a final report by November 2015.

This study will not evaluate misuse since this will be captured by instruments in a prospective study of patients with chronic pain (PMR Study 1A) using a combination of adapted validated instruments, and new instruments that will be evaluated in PMR Study 2. This study will not include a formal validation for opioid-related deaths, since processes for coding deaths vary from state to state, but will include some verification of opioid-related deaths relative to medical records for events with available state and national death data (there is a 12-month to 2-year lag in state death records).

Enrollment: 2,701 patients

Sex: All

Volunteers: No Healthy Volunteers

Inclusion and exclusion criteria

Inclusion:

Members of the KPNW integrated healthcare system located in the states of Oregon and southwestern Washington, between August 2008 and December 2014

• Approximately 1,200 events identified based on ICD diagnostic codes for opioid poisoning, overdose or opioid-related cause of death
• A random sample of approximately 1,250 individuals at increased risk of opioid overdose identified based on ICD diagnoses for opioid-related adverse effects, pain, mental health, or substance abuse

Exclusion:

None

Trial design

2,701 participants in 1 patient group

Opioid overdose and poisoning

Treatment:

Other: Algorithm to determine overdose from opioid abuse