Drug Overdose Data Sources

Description: Drug overdose fatality data comes from Minnesota death certificates. After a death occurs, a death investigation is completed, along with an autopsy (89% of drug overdose deaths had an autopsy completed in 2021), and a medical examiner or coroner determines the cause and manner of death. Both are added to the corresponding sections of the death certificate. Once the medical examiner certifies a death certificate, the Office of Vital Records (OVR) at the Minnesota Department of Health (MDH) performs data quality checks to ensure the death certificate was completed properly and accurately. After death certificates are sent to MDH's OVR, they are then sent to the National Center for Health Statistics (NCHS), where the literal text on the death certificate in the cause of death fields is converted from text to the International Classification of Diseases and Related Health Problems, 10th revision (ICD-10) codes. The NCHS uses a specially made computer program to code the literal text on the death certificate into ICD-10 codes according to the information listed on the death certificate. This program allows for one underlying cause of death and up to 20 contributing causes of death. Once this program is run and the data are assigned ICD-10 codes, the data is returned to the OVR in the MDH and made available for analysis.

ICD-10 codes used in identifying overdose deaths

To identify drug overdose deaths and the specific drugs involved from the entire death certificate file, a two-step analysis is completed. First, only death certificates with an ICD-10 underlying cause of death ICD-10 code in the ranges of X40-X44, X60-X64, X85, or Y10-Y14 are selected. From there specific ICD-10 drug codes in the contributing cause of death fields are searched to identify all drug overdose deaths that involved each drug category of interest (ICD-10 drug codes: T36-T50). The drug categories are not mutually exclusive, meaning a death with heroin (T40.1) and cocaine (T40.5) will be counted in both drug categories. Data is then summarized by demographic and geographic variables of interest.

Strengths and limitations

Strengths: All deaths in the state are reported to OVR, as well as Minnesota resident deaths that occurred in other states. There is a consistency of reporting across states, which aids analysis. Additionally, data can be easily grouped by some drug categories based on ICD-10 codes to allow for examination of trends over time.

Limitations: The limitations of the death certificate data include timeliness in completing the death certificates. The death investigation, autopsy process, and toxicology testing take time to complete. Due to Minnesota’s current medical examiner and coroner system, there is also variability in how drug overdose deaths are reported. This includes variation in the details and specificity listed on a death certificate in terms of the specific drugs involved in a death. Annual preliminary data becomes available in the second quarter of the following year and is not finalized for a full calendar year. Finally, death certificates cannot be used to quickly identify new drug trends because of the time it takes to complete toxicology and for death certificates to be finalized and made available for analysis.

Perform your own analysis:

• CDC WONDER The CDC WONDER online mortality databases use a rich online query system for the analysis of public health data. Analyses of death certificates can be conducted for the United States as well as for any state, including Minnesota.
• CDC WISQARS CDC’s WISQARS (web-based injury statistics query and reporting system) is an interactive, online database that provides fatal and nonfatal injury, violent death, and cost of injury data from a variety of trusted sources. Researchers, the media, public health professionals, and the public may use WISQARS data to learn more about the public health and economic burden associated with unintentional and violence-related injury in the United States.

Description: Developed by the Centers for Disease Control and Prevention (CDC), the State Unintentional Drug Overdose Reporting System (SUDORS) collects detailed information on overdose deaths occurring in Minnesota. The Minnesota Department of Health (MDH) currently partners with all Minnesota medical examiners and coroners to collect data on unintentional and undetermined overdose deaths. The data is used to better understand circumstances surrounding the incident as well as the type and origin of drugs involved. Starting in 2019, more than 97% of all overdose deaths occurring in Minnesota have been reported in SUDORS. More information can be found at State Unintentional Drug Overdose Reporting System (SUDORS).

The data includes:

• All deaths due to an overdose that occurred in Minnesota.
• Coverage: 2019 - ongoing.
• Fatal drug overdoses of unintentional (i.e., accidental) or undetermined intent.

Strengths and limitations

Strengths: Comprehensive toxicology  information is available, including all substances present at the time of death. Furthermore, detailed circumstances about the events leading up to and at the time of death are abstracted from medical examiner and coroner files. For example, data is available on people experiencing homeless and/or housing instability, as well as detailed mental health, past medical history, and overdose-specific circumstances.

Limitations: Only cases that meet the SUDORS case definition are included; therefore, it is possible that not all unintentional or undetermined overdose or substance-related deaths are present in the analysis. SUDORS attempts to collect 600-plus decedent, injury, and circumstance variables, but only data present in the medical examiner or coroner records at the time of abstraction is included. Due to this, many circumstance variables can be assumed undercounted or are reported as missing/unknown.

Description: Data for nonfatal hospital-treated overdose come from hospital discharge data. The Minnesota Department of Health (MDH) receives approximately 95% of hospital discharge data every three months from our partners at the Minnesota Hospital Association. The data covers all 87 counties and can include reports from all 123 acute care hospitals in the state as well as additional hospitals located in North Dakota.

ICD-10-CM codes used to identify nonfatal drug overdose

MDH uses the Centers for Disease Control and Prevention (CDC) Drug Overdose Surveillance and Epidemiology case definition to identify hospital visits related to a nonfatal drug overdose. To do this, International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes are used. ICD-10-CM codes are what medical professionals use for clinical diagnosis and insurance billing purposes. There are many ICD-10-CM codes for drug overdose. For a full list of the ICD-10-CM codes currently used to identify nonfatal overdose data, please review Appendix I in the Nonfatal Drug Overdose Dashboard Guidance.

Strengths and limitations

Strengths:

• The data includes:
  • Minnesota residents who were treated at Minnesota and North Dakota hospitals.
  • All emergency department visits and inpatient hospitalizations.
• The data represents most hospital visits for drug overdose in Minnesota, regardless of the patient’s insurance carrier or payer.
• The data is de-duplicated prior to analysis, which means that a hospital visit is counted only once.

Limitations:

• The data does not include:
  • Nonfatal drug overdoses occurring out in the community.
  • Nonfatal drug overdoses treated by emergency medical services (EMS) and not transported to a hospital.
  • Nonfatal drug overdoses treated at federally funded facilities (like Veteran’s Affairs or Indian Health Service), tribally operated facilities, stand-alone psychiatric facilities, and other out-of-state facilities.
• Hospital discharge data is primarily used for billing purposes, not for public health purposes.
• ICD-10-CM diagnosis codes sometimes lack detail. For example, using a code for "unspecified opioids" does not help us understand what type of opioid was involved in an overdose.
• All drug overdoses are suspected overdoses. Drugs that are suspected to be involved in hospital visits are often self-reported by the patient or determined by presenting symptoms at the hospital. Toxicology tests are not typically run to determine the specific drug types involved. It is important to interpret drug-specific trends with caution.
• The transition to ICD-10-CM in the final quarter of 2015 (October – December 2015) makes comparisons inaccurate, if not impossible, to previous years. Because of this, data is provided from 2016 onward.

Other data considerations

• Most often, the data shared by MDH will include nonfatal drug overdoses of unintentional (i.e., accidental) and undetermined intent. Nonfatal drug overdoses determined to be intentional (i.e., related to self-harm) are not usually included, unless otherwise noted. For access to data on nonfatal intentional drug overdoses, please visit the Hospital-Treated Suicidal and Self-Harm Injury Dashboard.
• Drug categories are non-exclusive. For example, a nonfatal overdose involving a synthetic opioid would be counted as an ‘All opioid’ overdose and an ‘All drug’ overdose.
• In the hospital discharge data, the collection of race and ethnicity information has not always been consistent. However, it has been improving over the past several years. In 2016, 23% of hospital-treated nonfatal drug overdoses had missing race and ethnicity data. By 2021, the proportion of visits with missing race and ethnicity data had decreased to 2%. This context is important to keep in mind when looking at historical trends by race or ethnicity of the patient.

Perform your own analysis:

• The Nonfatal Drug Overdose Dashboard is a tool that provides the public with data to better understand which people are experiencing nonfatal drug overdoses, where they are happening, and the impacts on communities. The one-page dashboard includes three sections (trends over time, demographics, geography), allowing the user to examine data by suspected drug type, hospital setting (emergency department, hospitalization), population groups (county of residence, five-year age groups, gender, race/ethnicity), and various epidemiologic measures (counts, rate, proportion). The dashboard also features dynamic narratives to aid the user in data interpretation and provides helpful tips for navigation.

Description: The Minnesota drug overdose and substance use surveillance activity (MNDOSA) tracks cases of substance misuse that result in emergency department visits or hospitalization in near-real time. Toxicology testing is rarely performed for overdoses that are treated in hospitals. The MNDOSA program fills this information gap by identifying substances used in severe and unusual cases and collects data on the circumstances and risk factors involved in these cases.

Participating sites report cases treated in emergency departments where the principal diagnosis is attributable to substance misuse, to the MNDOSA program. An overview of MNDOSA data, including the substances commonly detected and substance trends over time, can be found at Minnesota Drug Overdose and Substance Use Surveillance Activity.

Strengths and limitations

Strengths:

Strengths of MNDOSA data is the availability of enhanced toxicology data, which is rarely available for nonfatal overdoses. This data shows trends in substance use over time and helps detect emerging and novel substances. Circumstance data provides important context that informs prevention among high-risk populations.

Limitations:

MNDOSA currently operates at select hospitals in Northeast Minnesota and the Twin Cities Metro area; this data does not represent substance use trends seen at all hospitals across the state or in all regions of Minnesota. Cases are reported through MNDOSA if they are being treated in a hospital for a reason primarily due to substance use; this often includes drug overdoses but may include other acute problems related to substance use. Finally, data collected from electronic health records vary in completeness, so MNDOSA findings must be interpreted with caution.

Description: Syndromic surveillance emphasizes the use of near real-time data with statistical tools to detect and characterize unusual activity for further public health investigation.

Syndromic surveillance systems may be used for situational awareness to further characterize an outbreak beyond initial detection and notification, to monitor the spread of an outbreak, and/or to monitor the effectiveness of outbreak response and interventions. These data help public health officials detect, monitor, and respond quickly to local public health threats and events of public health importance. Please access the MDH Syndromic Surveillance website to learn more about how the data is currently used at MDH. Additional examples are on the Situational Awareness page of the MDH COVID-19 Situation Update website.

Strengths and limitations

Strengths: Syndromic surveillance differs from other data sources, such as case or electronic case reports, All Payer Claims Database (APCD), and Electronic Health Record (EHR) Consortium data, in the breadth of data collected resulting in the ability to monitor and identify emerging public health concerns in near-real time. Syndromic surveillance includes data on all visits treated at hospitals, regardless of condition. Analysis of this data based on signs and symptoms before a confirmed diagnosis or lab test significantly improves the timeliness of analysis and allows identification of concerns across a broad range of conditions. Other data sources mentioned above are focused on a specific diagnosis or lab test, resulting in an inability to examine other relevant conditions or identify new health issues.

Limitations: Because the syndromic surveillance data is based on presenting signs and symptoms when a patient first walks through the door, conditions are considered suspected and are not confirmed. Subsequently, this data is not exact measures and should be used to understand trends, rather than to count exact cases. In addition, recent data is not final and is subject to change. To account for the variability during this time, the syndromic data typically include "lag period" in published charts.