In 2020, there were over 926 000 non-fatal overdoses treated in emergency departments (EDs) in the USA, with approximately 10 non-fatal overdoses treated in EDs for every fatal overdose.¹ In 2022, fatal overdoses exceeded 107 000.² Individuals who experience a non-fatal overdose are more likely to experience a subsequent fatal overdose.³ While 87% of individuals who die from a fatal overdose do not have a documented history of prior non-fatal overdoses, this may be partially attributable to inadequate documentation.⁴ Changes in the illegal drug supply, such as a rise in synthetic opioids, and increases in stimulant-involved and polysubstance overdoses, have contributed to recent increases in overdose mortality.^{5 6} To address the US drug overdose epidemic, the Centers for Disease Control and Prevention (CDC) implemented the Overdose Data to Action (OD2A) programme in 2019, followed by OD2A in States (OD2A-S) in 2023. These programmes funded health departments to obtain high quality, comprehensive and timely overdose data to inform prevention and response efforts.⁷ Non-fatal overdose surveillance data are captured by the Drug Overdose Surveillance and Epidemiology (DOSE) system. One of DOSE’s key data sources is hospital discharge/billing data (hereafter, ‘discharge data’) from ED visits and inpatient hospitalisations. Discharge data are useful to estimate overall burden of non-fatal overdose and are the focus of this analysis.⁸

During the first OD2A programme, health departments submitted DOSE discharge data to CDC quarterly, with a 3.5-month lag between end-of-quarter closeout and submission (OD2A-S discharge data are reported annually, with a 6.5-month lag between end-of-year closeout and submission). By comparison, availability of state-level discharge data from the Agency for Health Research and Quality-sponsored Healthcare Cost and Utilization Project (HCUP) can often lag by several years, making accurate and timely estimates of overdose burden difficult to assess.⁹ However, DOSE discharge data have not been compared with other established state-level surveillance systems, such as HCUP.

The study objective was to compare non-fatal overdose rates obtained through DOSE during the first OD2A programme to rates obtained using HCUP databases. This comparison aimed to estimate whether DOSE is an accurate surveillance system for non-fatal overdoses.

Quarterly rates of non-fatal all drug, all opioid-involved, heroin-involved and all stimulant-involved non-fatal overdoses were mostly aligned between DOSE ED and SEDD data across all 12 states (figure 1, online supplemental table 1). For all states and drug overdose categories, 82% of quarterly ED rate differences between SEDD and DOSE were within ±0.5 overdoses per 100 000 (online supplemental figure 1, online supplemental table 1), with the IQR of the rate differences from −0.05 to 0.12 per 100 000 across all drug categories. The largest differences were among all drug overdoses, for which the average of the quarterly rate differences between SEDD and DOSE was 0.35 per 100 000, compared with 0.16, 0.05 and 0.05 per 100 000 for all opioid-involved, heroin-involved, and stimulant-involved overdoses, respectively.

Quarterly rates of inpatient hospitalisations for all four overdose categories were also similar between DOSE hospitalisation and SID data across 16 states (figure 2, online supplemental table 2). Ninety-three per cent of rate differences between SID and DOSE were within ±0.5 overdoses per 100 000 of each other (online supplemental figure 2, online supplemental table 2). All drug overdoses had the largest differences, with the average of the quarterly rate differences of −0.13 per 100 000, followed by −0.04 to –0.10, and −0.04 per 100 000 for all opioid, heroin-involved, and stimulant-involved overdoses, respectively. DOSE inpatient rates tended to be slightly higher than SID rates, with an IQR of −0.09 to 0 for the rate differences across all drug categories.

AQPC estimates were very similar between HCUP and DOSE for each state and overdose category (figures 3 and 4). The average difference in AQPC between the SEDD and DOSE ED data for all states and indicators was 0.08, with an IQR of −0.1 to 0.3 (not shown). For inpatient hospitalisations, the average difference in AQPC between the SID and DOSE data was 0.07, with an IQR of 0–0.2 (not shown).

This study suggests that non-fatal overdose surveillance through DOSE may be a valid source to estimate state-level non-fatal overdose burden. The minimal differences between DOSE and HCUP data from 2018 to 2020 may indicate DOSE’s potential to offer a more timely source of non-fatal overdose rates compared with HCUP data since the data are received more quickly, with a 3.5-month or 6.5-month lag between end-of-quarter or end-of-year closeout and submission, respectively, compared with HCUP’s lag of at least 1–2 years.

Variability in how states obtain and process discharge data may explain some of the minor differences observed between HCUP and DOSE discharge data. The datasets that health departments receive may differ slightly from what HCUP receives, and different data processing approaches for DOSE versus HCUP may also lead to variability. The higher variation in all-drug versus other indicators is likely due to more drug categories being included in the DOSE all-drug case definition versus the DOSE drug-specific case definitions; this could introduce additional variability between data sources. Future research could tease apart these differences more closely by examining specific categories of ICD-10-CM codes that are present in the all-drug definition.

Furthermore, details on the types of hospitals included in the SEDD and SID are limited and depend on the information reported to HCUP by data organisations, which could be another factor contributing to the minor differences between DOSE and HCUP.^{12 13} Despite these potential variations, DOSE hospital inclusion criteria were based on the HCUP definition of ‘community hospitals’, suggesting consistency in hospital selection methods between sources.^{12 13}

Overall, this comparison suggests that OD2A DOSE data are a dependable source of non-fatal overdose burden. Additionally, this analysis allowed CDC to better understand challenges working with state discharge data and offered opportunities for quality control of DOSE data submissions. Certain limitations should be noted. First, under OD2A (2019–2023), DOSE did not collect line-level data, and only collected aggregate data on four non-fatal overdose categories, which did not permit calculation of sensitivity or positive predictive value of DOSE data at the record level for a full validation.¹⁸ Second, only states that submitted data through both DOSE and SEDD or SID datasets that CDC purchased could be included in this analysis; however, under OD2A-S, 34 states began submitting DOSE discharge data as of 2024. Lastly, comparisons beyond 2020 were not possible as state-specific HCUP data were not yet widely available.

Discharge data collected from state health departments are an important complement to syndromic surveillance data for monitoring trends in US non-fatal overdoses.⁸ While HCUP data can be delayed by several years, DOSE discharge data may provide a more timely yet accurate source of non-fatal overdose discharge information.

This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.