Nonfatal Overdose Biosurveillance: A Cross-Sectional Pilot Study

Abstract

Objective:

Nonfatal overdoses provide critical insights into the substance use crisis, offering opportunities for timely interventions and prevention. This study pilots a nonfatal overdose biosurveillance strategy to analyze the demographic, clinical, and toxicological profiles of overdose patients, aiming to identify patterns and risk factors associated with these incidents.

Methods:

We assessed residual urine specimens collected from emergency department patients experiencing a nonfatal overdose at 2 hospitals in Wisconsin from August 2022 through February 2024. We collected data on patient demographic characteristics, results of clinical toxicology screening, manner of overdose, risk factors for overdose, and discharge status. Statistical analyses identified associations and odds ratios (ORs) among patient characteristics, detected drugs, and discharge status.

Results:

Of the 79 patients in the study, many had risk factors for overdose, including substance use disorder (48%), history of a mental health condition (43%), and polysubstance use (72%). Synthetic opioids had a strong positive association with a history of overdose (OR = 3.86). The presence of stimulants and antidepressants showed moderate sex-based associations, while race was linked to differing discharge status. Polysubstance use had a positive association with some drug combinations, such as narcotic analgesics and cocaine (OR = 4.00).

Conclusions:

This study highlights the prevalence of polysubstance use and identifies key demographic and clinical factors associated with nonfatal overdoses. These findings underscore the need for comprehensive, real-time biosurveillance to inform targeted public health interventions and improve patient outcomes. Enhanced understanding of these patterns can lead to more effective strategies for overdose prevention and management, addressing a critical gap in current public health approaches.

Fatal and nonfatal overdoses are a substantial public health concern. ¹ An estimated 60% of overdose fatalities occur among people who have at least 1 opportunity to be linked with life-saving interventions before the fatal overdose. ² This high percentage reflects issues related to substance use, mental health care, and interventions in US communities.^3,4 While fatal overdoses often capture headlines^5-10 and drive policy responses,^11-14 nonfatal overdoses provide critical insights into the ongoing opioid crisis and other substance-related issues. These events offer a unique opportunity for intervention and prevention, because they precede potential fatalities and highlight populations at immediate risk of overdose.

Literature on nonfatal overdose biosurveillance reveals a growing recognition of its importance in understanding and addressing the substance use crisis.^15,16 Traditional overdose research has predominantly focused on fatal incidents,^5-14,17 often neglecting the insights provided by nonfatal cases. Recent studies, however, highlight that nonfatal overdoses are key indicators of substance use patterns ¹⁸ and risk factors,^19-24 offering opportunities for early intervention.^25-28 Literature underscores the prevalence of polysubstance use in nonfatal overdoses and identifies various demographic and socioeconomic factors, such as age, sex, race, and mental health status, as important contributors to overdose risk.^29-33 Additionally, increasing evidence suggests that real-time biosurveillance can enhance public health responses by identifying emerging trends and outbreaks of new or adulterated substances.^34-37 Despite these achievements, gaps remain in the integration of comprehensive biosurveillance data into public health strategies, underscoring the need for more robust and detailed research in this area.

This study focused on nonfatal overdose biosurveillance, leveraging an innovative approach that integrates comprehensive toxicological screening with detailed demographic and clinical data collection.^38,39 By examining a diverse patient cohort, this research aimed to elucidate the patterns and risk factors associated with nonfatal overdoses. The inclusion of detailed profiles of substance use and the analysis of co-occurring risk factors such as mental health conditions and housing instability provide a holistic view of the factors contributing to overdose incidents.

Methods

Study Design and Population

This cross-sectional pilot study re-analyzed residual urine specimens (ie, left over after completion of clinical testing) from patients experiencing overdoses and seeking care at 2 emergency departments (hereinafter, submitting partners) in Madison and Green Bay, Wisconsin, from August 2022 through February 2024. The submitting partners collected urine specimens as part of routine clinical care for patients presenting with suspected overdose. After all clinical testing was complete, hospital personnel deidentified specimens and sent them to the Wisconsin State Laboratory of Hygiene (WSLH) for toxicological screening. Emergency department physicians and clinical toxicology laboratory personnel identified nonfatal overdose patient samples for inclusion in this study. Prior to deidentification, hospital personnel performed medical record reviews to gather data for the study’s specimen submission form. This form was hosted on REDCap^40,41 and captured data on patient demographic characteristics, results of clinical toxicology screening, manner of overdose, risk factors for overdose, and discharge status.

Our dataset comprised only a subset of patients experiencing a nonfatal overdose at the 2 hospitals; data collection varied due to factors such as clinician awareness of the study, clinical practices, and availability of urine specimens. Not all patients presenting with a nonfatal overdose have urine collected, and sometimes urine samples are used up during routine clinical testing.

This project was determined to meet federal criteria for exempt human subjects research by the University of Wisconsin–Madison’s institutional review board (no. 2022-0393).

Study Measures

Patient demographic characteristics

During medical record review, hospital personnel recorded patient age, sex, race, and ethnicity on the specimen submission form, categorizing sex as male (including those assigned female at birth) or female (including those assigned male at birth) and race as American Indian or Alaska Native, Asian, Black or African American (hereinafter, Black), Native Hawaiian or Pacific Islander, White, or unknown. Multiple races could be selected for a patient. No submitted urine specimens were collected from a patient of known Asian or Native Hawaiian or Pacific Islander descent. We categorized race as Black, White, or Other (includes American Indian or Alaska Native and unknown) and ethnicity as Hispanic, not Hispanic, or unknown.

Results of clinical toxicology screening

WSLH researchers recorded analytes detected in positive and negative ionization modes for each urine specimen. Study team members divided detected substances into the following classes for data analysis: amphetamines, antidepressants, antipsychotics, benzodiazepines, cannabinoids, cocaine, dissociative anesthetics, fentanyl, narcotic analgesics, stimulants, synthetic opioids, antidepressants, medication used to treat opioid use disorder, and naloxone. Some substances were included in multiple classes. Narcotic analgesics included all opioids, synthetic opioids, fentanyl, and tramadol. Synthetic opioids excluded fentanyl but included analogs and nitazenes. Stimulants included amphetamines, cocaine, modafinil, and phentermine. The presence of alcohol was also included if submitting partners indicated a blood or breath alcohol concentration on the specimen submission form.

Manner of overdose

During medical record review, hospital personnel recorded data on the manner of overdose, categorized as intentional/suicide, unintentional/accident, assault, or unknown. No patients in this study selected assault as a manner of overdose.

Risk factors for overdose

During medical record review, hospital personnel indicated on the submission form if the patient experienced certain factors associated with increased risk of overdose. We included the following risk factors in our analysis: unhoused, chronic pain, HIV or hepatitis B/C infection, pregnancy, incarceration within the last 30 days, history of overdose requiring medical care, history of substance misuse, and mental health disorder. Hospital personnel could select multiple risk factors for each patient. No patients had a known or reported HIV or hepatitis B/C infection, were pregnant, or were incarcerated within the last 30 days. The lack of samples from pregnant or recently incarcerated patients may have stemmed from their awareness that urine specimen results could result in probation/parole violations or involvement with the Department of Children and Families. An additional risk factor of polysubstance use was determined by the presence of ≥2 distinct drug classes or the presence of 1 distinct drug class in the urine specimen along with the presence of alcohol.

Discharge status

Submitting partners noted the discharge status of the patient on the sample submission form. Options included official discharge, death, left without treatment, left against medical advice, admission to hospital, admission to intensive care unit (ICU), admission to a psychiatric facility, transfer to another facility, admission to detoxification or substance abuse treatment program, or discharge to law enforcement. Submitting partners selected all options that applied. No patients included in this study left without treatment, left against medical advice, or were discharged to law enforcement. However, patients who leave without treatment or against medical advice would be unlikely to submit a urine specimen and, therefore, would not be eligible for inclusion in this study. For data presentation, we categorized the discharge status of 69 patients (some categories had too few patients for statistical analysis) as discharge, admit to hospital, or admit to hospital or ICU.

Sample Preparation

Urine specimens were shipped on ice and stored at 2 °C to 8 °C upon arrival at the WSLH. Aliquots (250 µL) were pipetted into 2-mL vials, mixed with 250 µL of 1:1 water/methanol with 0.1% formic acid and 50 µL of (±)-warfarin-d₅ (100 ng/mL) internal standard. If less than 250 µL was available, aliquot volume was reduced and the water/methanol and internal standard ratios were maintained.

Wide-Scope Screening Assay

Details on the wide-scope (ie, untargeted data acquisition) screening assay used in this study are described elsewhere. ⁴² Briefly, chromatographic separation was performed on an Acquity HSS C₁₈ column (Waters Corporation) (2.1 mm × 150 mm, 1.8 μm particles) maintained at 50 °C. Mobile phases consisted of the following: A₁ = 5 mM ammonium formate at pH 3.0; B₁ = acetonitrile and 0.1% formic acid; A₂ = water and 0.001% formic acid; and B₂ = acetonitrile and 0.001% formic acid. Separation occurred via gradient during a 15-minute period in positive ionization acquisition mode and a 7.5-minute period in negative ionization acquisition mode. A Xevo G2-XS QToF mass spectrometer (Waters Corporation) with an electrospray ionization source was used in positive and negative ionization modes. Data-independent acquisition mode was used with 3 MS functions (MS^E).

Data Analysis

To evaluate the representativeness of our dataset, we queried the Centers for Disease Control and Prevention’s Drug Overdose Surveillance and Epidemiology ⁴³ system to identify the sex and race of all people who experienced a nonfatal overdose and sought care from the 2 emergency departments included in this work during the study period. We performed a χ² goodness-of-fit test to assess whether the distribution of sex and race in the study sample significantly differed from the expected distribution based on the broader population of individuals presenting with overdose to the participating emergency departments.

We summarized data as the frequency of observations and measures of correlation. We identified significant associations via cross-tabulation of 2 binary variables (ie, 2 × 2 tables) using previously described methods. ⁴⁴ The null hypothesis was that no correlation existed between 2 variables (ie, they are independent); it was assessed by the Pearson χ² test for groups with n ≥ 4 (eg, is amphetamine use [presence, absence] associated with being unhoused [yes, no]?). We excluded 2 × 2 tables with any cells ≤4. Large χ² values indicated a significant association between 2 variables, and the hypothesis of independence was rejected. We converted χ² values into P values, which we classified as weak association (.10 > P > .05), moderate association (.05 ≥ P > .01), strong association (.01 ≥ P > .001), and very strong association (P ≤ .001). We used previously described methods to calculate odds ratios (ORs) determined from the 2 × 2 tables to quantify directionality of the relationship. ⁴⁴ Calculated ORs in this cross-sectional study indicate the number of people who experienced an event divided by the number of people who did not experience it. ⁴⁴

Results

Seventy-nine urine specimens were submitted. By manner of overdose, 44% (n = 35) were unintentional, 37% were intentional (n = 29), and 20% were unknown (n = 15). Patient age ranged from 1 to 70 years (mean = 34 y; median = 32 y). Most (61%) were aged 10 to 39 years; the patients were evenly distributed across decades within that range (Figure). Both sexes were represented: 56% were male (n = 44) and 44% were female (n = 35). Most (91%) patients identified as White (n = 57; 72%) or Black (n = 15; 19%).

Figure.

Distribution of patients, by age, included in a study of nonfatal overdose, Wisconsin, 2022-2024. Data source: Residual urine specimens were collected from emergency department patients experiencing a nonfatal overdose at 2 hospitals in Wisconsin from August 2022 through February 2024.

In comparison, the population experiencing a nonfatal overdose (n = 842) presenting to submitting partners during the study period had the following demographic characteristics: by sex, 54% (n = 451) were male and 46% (n = 391) were female; by race, most (88%) were White (n = 611; 73%) or Black (n = 131; 16%). A χ² goodness-of-fit test confirmed no significant differences between the 2 samples for sex ( ${\chi }_1^2$  = 0.1; P = .70) or race ( ${\chi }_5^2$  = 5.4; P = .36).

For detected drug classes and associations with patient sex, race, and ethnicity, we found moderate associations between patient sex and use of any stimulant (P = .03; majority male use) and antidepressants (P = .03; majority female use) (Table 1). We observed no significant associations for race or ethnicity.

Table 1.

Summary of detected drugs or drug classes, discharge status, and risk factors across sex and race among emergency department patients experiencing a nonfatal overdose, Wisconsin, 2022-2024 ^a

Category	Sex				Black (n = 15)			White (n = 57)
	Male (n = 44)	Female (n = 35)	${\chi }_1^2$ (P value)	OR b	No.	${\chi }_1^2$ (P value)	OR b	No.	${\chi }_1^2$ (P value)	OR b
Detected drugs and drug classes
Amphetamine (n = 15)	11	4	—	—	2	—	—	12	—	—
Antidepressant (n = 26)	10	16	4.7 (.03)c	0.35	4	—	—	20	0.44 (.51)	0.69
Antipsychotic (n = 7)	3	4	—	—	2	—	—	5	—	—
Any stimulant (n = 31)	22	9	4.8 (.03) c	2.58	7	0.4 (.51)	1.46	23	0.1 (.74)	1.46
Benzodiazepine (n = 21)	13	8	0.4 (.50)	1.42	2	—	—	17	—	—
Cannabinoid (n = 35)	22	13	1.3 (.25)	1.69	7	<0.1 (.84)	1.13	22	2.7 (.10) c	1.13
Cocaine (n = 19)	13	6	1.6 (.20)	2.03	3	—	—	16	—	—
Dissociative anesthetic (n = 18)	10	8	<0.1 (.99)	0.99	4	—	—	14	—	—
Fentanyl (n = 15)	10	5	0.9 (.34)	1.76	4	—	—	10	0.3 (.60)	1.75
MOUD (n = 7)	4	3	—	—	1	—	—	6	—	—
Naloxone (n = 8)	5	3	—	—	2	—	—	6	—	—
Narcotic analgesic (n = 30)	18	12	0.4 (.55)	1.33	5	0.2 (.68)	0.78	24	1.5 (.22)	0.78
Synthetic opioid (n = 18)	14	4	—	—	4	—	—	13	<0.1 (.99)	1.30
Discharge status
Discharge (n = 18)	10	8	0 (.99)	0.99	6	3.1 (.08) c	2.89	8	8.9 (.003) c	0.20
Admit to hospital (n = 45)	25	20	0 (.98)	0.99	3	—	—	41	—	—
Admit to ICU (n = 6)	3	3	—	—	2	—	—	4	—	—
Admit to hospital or ICU (n = 51)	28	23	<0.1 (.85)	0.91	5	7.9 (.005) c	0.20	45	18.5 (<.001) c	10.00
Risk factors
Unhoused (n = 5)	3	2	—	—	1	—	—	3	—	—
Chronic pain (n = 6)	3	3	—	—	0	—	—	5	0.4 (.52)	2.02
History of overdose (n = 13)	5	8	1.9 (.17)	0.43	4	—	—	8	0.9 (.35)	0.56
History of SUD (n = 38)	22	16	0.1 (.70)	1.19	7	<0.1 (.90)	0.93	30	1.7 (.20)	1.94
History of mental health condition (n = 34)	13	21	7.4 (.007) c	0.28	5	0.7 (.40)	0.60	26	0.6 (.46)	1.47
Polypharmacy (n = 57)	32	25	<0.1 (.90)	1.07	12	0.6 (.45)	1.69	40	0.4 (.53)	0.69

Abbreviations: —, does not apply; ICU, intensive care unit; MOUD, medication for opioid use disorder; OR, odds ratio; SUD, substance use disorder.

^aData source: Residual urine specimens were collected from emergency department patients experiencing a nonfatal overdose at 2 hospitals in Wisconsin from August 2022 through February 2024.

^bORs compare the odds of the event in 1 group relative to the reference group (eg, amphetamine detected vs no amphetamine detected, discharge vs no discharge, history of mental health condition vs no history of mental health condition).

^cP values from ≤.001 to .10 were considered significant. Pearson χ² values were converted into P values, which were classified as weak association (.10 > P > .05), moderate association (.05 ≥ P > .01), strong association (.01 ≥ P > .001), and very strong association (P ≤ .001).

In the examination of discharge status by race and ethnicity, we found a strong negative association between identifying as Black and admission to the hospital or ICU (P = .005; OR = 0.20) and a weak positive association between identifying as Black and discharge (P = .08; OR = 2.89). Identifying as White was positively and very strongly associated with being admitted to the hospital or ICU (P < .001; OR = 10.00) and negatively and strongly associated with discharge (P = .003; OR = 0.20) (Table 1). In the examination of drug classes detected and any association with discharge status, antidepressants were positively and moderately associated with admission to the hospital (P = .04; OR = 2.82) (Table 2). Cannabinoids were negatively and weakly associated with admission to the hospital or ICU (P = .09; OR = 0.45) and admission to the hospital (P = .07; OR = 0.44).

Table 2.

Summary of detected drugs and drug classes across discharge status among emergency department patients experiencing a nonfatal overdose, Wisconsin, 2022-2024 ^a

Category	Discharge (n = 18)			Admit to hospital (n = 45)			Admit to hospital or ICU (n = 51)
	No.	${\chi }_1^2$ (P value)	OR b	No.	${\chi }_1^2$ (P value)	OR b	No.	${\chi }_1^2$ (P value)	OR b
Detected drugs and drug classes
Amphetamine (n = 15)	5	1.2 (.28)	1.96	8	0.1 (.75)	0.83	8	1.0 (.31)	0.59
Antidepressant (n = 26)	3	—	—	19	4.1 (.04) c	2.82	20	2.6 (.11)	2.37
Antipsychotic (n = 7)	2	—	—	4	—	—	5	—	—
Any stimulant (n = 31)	9	1.1 (.29)	1.77	17	<0.1 (.76)	0.87	19	0.2 (.63)	0.79
Benzodiazepine (n = 21)	3	—	—	15	2.4 (.12)	2.33	17	—	—
Cannabinoid (n = 35)	11	2.7 (.10)	2.42	16	3.2 (.07) c	0.44	19	2.9 (.09) c	0.45
Cocaine (n = 19)	5	0.2 (.67)	1.29	11	<0.1 (.92)	1.05	13	0.2 (.69)	1.25
Dissociative anesthetic (n = 18)	4	—	—	10	<0.1 (.89)	0.93	13	0.6 (.44)	1.57
Fentanyl (n = 15)	5	1.2 (.28)	1.96	8	0.1 (.75)	0.83	9	0.2 (.68)	0.79
MOUD (n = 7)	2	—	—	5	—	—	5	—	—
Naloxone (n = 8)	2	—	—	5	—	—	6	—	—
Narcotic analgesic (n = 30)	7	<0.1 (.93)	1.05	20	1.9 (.17)	1.92	20	0.9 (.76)	1.16
Synthetic opioid (n = 18)	5	0.3 (.57)	1.42	11	0.2 (.69)	1.25	12	0.1 (.83)	1.13
Risk factors
Unhoused (n = 5)	2	—	—	2	—	—	2	—	—
Chronic pain (n = 6)	1	—	—	3	—	—	3	—	—
History of overdose (n = 13)	5	2.2 (.14)	2.55	5	2.2 (.14)	0.41	6	2.3 (.13)	0.40
History of SUD (n = 38)	6	2.0 (.15)	0.45	24	1.2 (.28)	1.63	28	2.7 (.10)	2.19
History of mental health condition (n = 34)	8	<0.1 (.89)	1.08	18	0.4 (.53)	0.75	20	0.9 (.35)	0.65
Polypharmacy (n = 57)	14	0.4 (.54)	1.47	33	<0.1 (.79)	1.15	38	0.4 (.53)	1.38

Abbreviations: —, does not apply; ICU, intensive care unit; MOUD, medication for opioid use disorder; OR, odds ratio; SUD, substance use disorder.

^aData source: Residual urine specimens were collected from emergency department patients experiencing a nonfatal overdose at 2 hospitals in Wisconsin from August 2022 through February 2024.

^bORs compare the odds of the event in 1 group relative to the reference group (eg, amphetamine detected vs no amphetamine detected, discharge vs no discharge, history of mental health condition vs no history of mental health condition).

^cP values from ≤.001 to .10 were considered significant. Pearson χ² values were converted into P values, which were classified as weak association (.10 > P > .05), moderate association (.05 ≥ P > .01), strong association (.01 ≥ P > .001), and very strong association (P ≤ .001).

Risk factor distributions across sex, race, and ethnicity demonstrated no significant associations by race or ethnicity (Table 1). Having a history of a mental health condition was strongly associated with sex (P = .007; majority female). Having a history of overdose was positively and moderately associated with synthetic opioids (P = .03; OR = 3.86) and positively and weakly associated with narcotic analgesics (P = .06; OR = 3.20) (Table 3). Having a history of a mental health condition was positively and weakly associated with antidepressants (P = .06; OR = 2.44). Polysubstance use, observed in 57 of 79 patients (72%), was positively and strongly associated with stimulant presence (P = .001; OR = 10.36) and positively and weakly associated with cannabinoids (P = .06; OR = 2.76).

Table 3.

Summary of detected drugs and drug classes across risk factors among emergency department patients experiencing a nonfatal overdose, Wisconsin, 2022-2024 ^a

Detected drugs and drug classes	History of overdose (n = 13)			History of SUD (n = 38)			History of mental health condition (n = 34)			Polypharmacy (n = 57)
	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR
Amphetamine (n = 15)	1	—	—	11	—	—	4	—	—	13	—	—
Antidepressant (n = 26)	6	1.2 (.27)	1.97	11	0.5 (.47)	0.71	15	3.4 (.06) c	2.44	22	—	—
Antipsychotic (n = 7)	1	—	—	4	—	—	2	—	—	6	—	—
Any stimulant (n = 31)	5	0 (.95)	0.96	22	10.7 (.001) c	4.89	10	2.4 (.12)	0.48	29	11.6 (.001) c	10.36
Benzodiazepine (n = 21)	2	—	—	11	0.2 (.65)	1.26	10	0.2 (.62)	1.29	20	—	—
Cannabinoid (n = 35)	4	—	—	17	<0.1 (.94)	1.03	17	0.8 (.38)	1.50	29	3.6 (.06) c	2.76
Cocaine (n = 19)	4	—	—	16	—	—	7	0.4 (.53)	0.71	19	—	—
Dissociative anesthetic (n = 18)	3	—	—	10	0.5 (.47)	1.47	7	0.2 (.69)	0.80	17	—	—
Fentanyl (n = 15)	4	—	—	9	1.0 (.30)	1.81	2	—	—	15	—	—
MOUD (n = 7)	2	—	—	6	—	—	2	—	—	7	—	—
Naloxone (n = 8)	4	—	—	3	—	—	2	—	—	8	—	—
Narcotic analgesic (n = 30)	8	3.7 (.06) c	3.20	16	0.5 (.47)	1.40	10	1.9 (.17)	0.52	29	—	—
Synthetic opioid (n = 18)	6	4.8 (.03) c	3.86	10	0.5 (.47)	1.47	5	2.2 (.14)	0.42	18	—	—

Abbreviations: —, does not apply; MOUD, medication for opioid use disorder; OR, odds ratio; SUD, substance use disorder.

^aData source: Residual urine specimens were collected from emergency department patients experiencing a nonfatal overdose at 2 hospitals in Wisconsin from August 2022 through February 2024.

^bORs compare the odds of the event in 1 group (eg, amphetamine detected) relative to the reference group (eg, no amphetamine detected).

^cP values from ≤.001 to .10 were considered significant. Pearson χ² values were converted into P values, which were classified as weak association (.10 > P > .05), moderate association (.05 ≥ P > .01), strong association (.01 ≥ P > .001), and very strong association (P ≤ .001).

In the assessment of patterns of polysubstance co-occurrence, narcotic analgesics were positively and strongly associated with cocaine (P = .009; OR = 4.00), while antidepressants were negatively and moderately associated with cannabinoids (P = .03; OR = 0.33). Dissociative anesthetics were positively and moderately associated with stimulants (P = .03; OR = 3.22) and positively and weakly associated with cocaine (P = .09; OR = 2.60) (Table 4).

Table 4.

Summary of use of co-occurring substances among emergency department patients experiencing a nonfatal overdose, Wisconsin, 2022-2024 ^a

Detected drugs and drug classes	Alcohol (N = 15)			Antidepressants (N = 26)			Benzodiazepines (N = 21)			Dissociative anesthetics (N = 18)			Narcotic analgesics (N = 30)			Stimulants (N = 31)
	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR	No.	${\chi }_1^2$ (P value)	OR
Amphetamine (n = 15)	0	—	—	4	—	—	4	—	—	4	—	—	7	0.6 (.44)	1.56	15	—	—
Antidepressant (n = 26)	2	—	—	0	—	—	8	0.4 (.56)	1.37	7	0.4 (.54)	1.41	11	0.3 (.58)	1.31	11	0.2 (.70)	1.21
Any stimulant (n = 31)	4	—	—	11	0.2 (.70)	1.21	10	0.8 (.36)	1.60	11	4.7 (.03) c	3.22	15	2.4 (.12)	2.06	31	—	—
Benzodiazepine (n = 21)	5	0.4 (.51)	1.50	8	0.4 (.56)	1.37	0	—	—	7	1.8 (.18)	2.14	10	1.1 (.29)	1.73	10	0.8 (.36)	1.60
Cocaine (n = 19)	4	—	—	6	<0.1 (.89)	0.92	7	1.4 (.24)	1.92	7	2.8 (.09) c	2.60	12	6.7 (.009) c	4.00	19	—	—
Cannabinoid (n = 35)	7	<0.1 (.84)	1.13	7	4.7 (.03) c	0.33	10	0.1 (.72)	1.20	6	1.1 (.29)	0.55	12	0.4 (.55)	0.75	13	0.1 (.73)	0.85
Dissociative anesthetic (n = 18)	4	—	—	7	0.4 (.54)	1.41	7	1.8 (.18)	2.14	0	—	—	7	<0.1 (.93)	1.05	11	4.7 (.03) c	3.22
Fentanyl (n = 15)	2	—	—	5	0 (.97)	1.02	3	—	—	4	—	—	15	—	—	7	0.4 (.51)	1.46
Narcotic analgesic (n = 30)	5	0.2 (.68)	0.78	11	0.3 (.58)	1.31	10	1.1 (.29)	1.73	7	<0.1 (.93)	1.05	30	—	—	15	2.4 (.12)	2.06
Synthetic opioid (n = 18)	0	—	—	4	—	—	5	<0.1 (.90)	1.08	3	—	—	18	—	—	11	4.7 (.03) c	3.22

Abbreviations: —, does not apply; OR, odds ratio.

^aData source: Residual urine specimens were collected from emergency department patients experiencing a nonfatal overdose at 2 hospitals in Wisconsin from August 2022 through February 2024.

^bORs compare the odds of the event in 1 group (eg, amphetamine detected) relative to the reference group (eg, no amphetamine detected).

^cP values from ≤.001 to .10 were considered significant. Pearson χ² values were converted into P values, which were classified as weak association (.10 > P > .05), moderate association (.05 ≥ P > .01), strong association (.01 ≥ P > .001), and very strong association (P ≤ .001).

In the evaluation of overdose intentionality by sex, race, discharge status, and detected drug classes, intentional overdose was strongly associated with sex (P = .001; majority female) and positively and strongly associated with hospital or ICU admission (P = .003; OR = 4.28). Having a history of a mental health condition was positively and very strongly associated with intentional overdose (P < .001; OR = 7.47). Intentional overdose was positively and moderately associated with alcohol use (P = .04; OR = 3.30), negatively and moderately associated with using narcotic analgesics (P = .05; OR = 0.37), and positively and weakly associated with using antidepressants (P = .09; OR = 2.31) ( eTable in the Supplement).

Discussion

This study offers a comprehensive overview of the demographic characteristics, drug use patterns, and discharge status of nonfatal overdoses in a sampled population. Our analysis reveals insights into overdose characteristics, risk factors for overdose, and opportunities for targeted intervention. In addition, it provides a proof-of-concept for using residual specimens in comprehensive toxicological analyses.

Demographic Characteristics and Manner of Overdose

The mean age of patients in our study sample was 34 years, with 61% aged 10 to 39 years, underscoring the vulnerability of younger populations to overdose events. Of those who had a fatal overdose in 2023 in Wisconsin, 51.9% were aged ≤44 years. ⁴⁵ Balanced representation between males (56%) and females (44%) underscores that both sexes are equally at risk of experiencing an overdose, necessitating inclusive approaches to overdose prevention and treatment strategies. However, males were overrepresented (70.2%) in Wisconsin’s fatal overdoses in 2023, necessitating further inquiry.

The manner of overdose, categorized as unintentional (44%), intentional (37%), and unknown (20%), reflects the complexity of overdose scenarios. The substantial quantity of intentional overdoses emphasizes the need for robust mental health support and interventions aimed at people at risk of self-harm or suicide.

Racial and Ethnic Disparities

Data on race and ethnicity indicated a majority of White patients (73%), followed by Black patients (19%), with smaller representations of other races. This distribution mirrors broader societal demographic characteristics ⁴ but also highlights potential racial disparities in overdose incidents and reporting. The lack of significant associations between drug use and race or ethnicity in our analysis suggests that drug use patterns may be more uniformly distributed across racial and ethnic groups than previously assumed.^46,47 However, strong associations between race and discharge status in our study (eg, Black patients being less likely than non-Black patients to be admitted to the hospital or ICU) indicate potential disparities in health care responses that warrant further investigation.

Risk Factors for Overdose and Patterns of Substance Use

Risk factors identified in our study—substance use disorder (48%), history of a mental health condition (43%), history of overdose (16%), chronic pain (8%), and being unhoused (6%)—are consistent with established knowledge about populations at risk of overdose.^15,19-21 The high prevalence of polysubstance use (72%) underscores the complexity of these cases and the need for comprehensive drug screening in overdose management.

Notably, having a history of a mental health condition was strongly associated with female patients. Intentional overdoses were very strongly associated with females, having a history of a mental health condition, the presence of alcohol, and the presence of antidepressants. These findings suggest that mental health interventions and support services to reduce intentional overdoses should be tailored to the needs of women. Similarly, the positive association between synthetic opioids and a history of overdose aligns with the known dangers of these potent substances ¹⁷ and underscores the need for targeted overdose prevention strategies for people with previous overdose experiences.

Drug Class Associations With Discharge Status

Our study found significant associations between some drugs or drug classes and discharge status. For example, the presence of antidepressants was moderately associated with hospital admission and weakly associated with intentional overdose, suggesting that patients using these medications may have severe or complex overdoses requiring inpatient care for stabilization or to prevent self-harm. Conversely, the weak association of cannabinoids with hospital or ICU admission and hospital admission alone may indicate a lower immediate health risk compared with other substances. However, this weak association does not mitigate the overall risks associated with cannabis use.

Polysubstance use was a common theme, with narcotic analgesics strongly associated with cocaine presence and antidepressants moderately associated with cannabinoids. These associations highlight the interconnected nature of substance use and the necessity for integrated treatment approaches to address the use of multiple substances simultaneously. However, it is not possible to determine from urine specimen data whether patients intentionally consumed multiple substances or whether the unregulated drug supply was contaminated. ⁴⁸

Limitations

This study had several limitations. First, the small sample size and specific population (ie, people living in a midsize Midwestern city) may limit generalizability. Second, sampling bias could be present: some people experiencing an overdose may not seek care in an emergency department or may refuse treatment, and a portion of those in the sample may not seek care or may refuse treatment because of concern about probation or parole violations or Department of Children and Families involvement. Third, the cross-sectional study design limited causal inferences. While urine toxicology results are comprehensive, they lack granularity on dose, timing, or whether multiple substances were intentionally consumed or if the drug supply was contaminated. Fourth, the timing of urine collection depended on human factors (eg, clinician decision-making, patient condition, workflow constraints), introducing bias. Future research should include larger, more diverse populations and longitudinal studies to explore overdose risk factors and outcomes. Investigating the effect of socioeconomic factors, health care access, and community resources on overdose incidents may provide a deeper understanding of effective prevention and intervention strategies.

Conclusion

Our findings have several implications for clinical practice and public health. First, the prevalence of polysubstance use and other risk factors for overdose suggests that health care providers should adopt comprehensive, multifaceted screening and intervention strategies. Second, the significant associations between race and discharge status call for further exploration into potential systemic biases in health care delivery and the development of more equitable care practices. Moreover, the clear link between mental health conditions and overdose risk underscores the critical need for integrating mental health services into substance use disorder treatment. Tailored interventions that address the needs of younger populations and women, who show distinct patterns of risk factors and substance use, are essential.

Our study highlights the multifaceted nature of overdose incidents and the importance of comprehensive approaches in addressing this public health crisis. Biosurveillance strategies such as the approach described here provide detailed, real-time analysis of overdose patterns and capture critical data on polysubstance use and demographic-specific risk factors among people who experience a nonfatal overdose. This approach enables identification of emerging trends and disparities in overdose incidents, informing precise and effective public health interventions and policy decisions. By understanding the interplay of demographic factors, substance use patterns, and health care outcomes, we can develop targeted and effective strategies to reduce the incidence and impact of overdoses.