Analysis of Urine Drug Test Results From Substance Use Disorder Treatment Practices and Overdose Mortality Rates, 2013-2020

Key Points

Question  Do urine drug test (UDT) results correlate with overdose mortality rates?

Findings  In this cross-sectional study of 500 000 unique urine specimens collected from substance use disorder treatment practices, UDT results for 5 drug categories were correlated with overdose mortality rates at national, state, and county levels. Correlation was strongest for synthetic opioids and methamphetamine, and multivariate regression analysis using state-level data revealed that synthetic opioid and methamphetamine UDT positivity rates were significantly associated with overdose deaths.

Meaning  This study’s findings suggest that UDT is a valuable data source that is capable of informing real-time surveillance efforts.

Importance  Drug overdose deaths in the US are currently the highest ever recorded; data collected from public health surveillance sources can help to identify emerging drug use patterns associated with overdose mortality rates, but the time lag in results often limits utility. Urine drug testing (UDT) is one potentially underused source that could augment surveillance efforts through timely data collection.

Objective  To evaluate the correlation between real-time UDT results from a proprietary national database and overdose mortality data from the National Vital Statistics System.

Design, Setting, and Participants  This retrospective cross-sectional study included 500 000 urine specimens submitted for UDT by substance use disorder (SUD) treatment health care practices and collected between January 1, 2013, and December 31, 2020. Real-time UDT data were obtained from the Millennium Health proprietary national database, and overdose mortality data were obtained from the National Vital Statistics System of the Centers for Disease Control and Prevention (CDC WONDER). Specimens were analyzed for specific drugs in 5 categories (cocaine, heroin, methamphetamine, synthetic opioids, and other opioids) using liquid chromatography–tandem mass spectrometry. Participants were adults aged 18 years and older who provided urine specimens at SUD treatment practices.

Exposures  Urine drug testing.

Main Outcomes and Measures  The primary outcome was the correlation between UDT positivity rates and overdose mortality rates at national, state, and county levels. Univariate and multivariate regression models were also used to evaluate the association between state- and county-level overdose mortality and standardized UDT positivity rates.

Results  Among 500 000 unique patient specimens collected from SUD treatment practices between 2013 and 2020, 288 534 specimens (57.7%) were from men, and the median age of the study population was 34 years (IQR, 17-51 years). On a national level, synthetic opioids and methamphetamine were highly correlated with overdose mortality (Spearman ρ = 0.96 for both). When synthetic opioids were coinvolved, methamphetamine (ρ = 0.98), heroin (ρ = 0.78), cocaine (ρ = 0.94), and other opioids (ρ = 0.83) were also highly correlated with overdose mortality. In the absence of synthetic opioids, all drug categories were highly correlated (ρ = 0.75 for other opioids, 0.81 for heroin, and 0.88 for methamphetamine), with the exception of cocaine (ρ = −0.37). Synthetic opioids (ρ = 0.77) and methamphetamine (ρ = 0.80) had the strongest state-level correlations over time, whereas other opioids had the lowest correlation for both total positivity (ρ = 0.31) and positivity in the absence of synthetic opioids (ρ = 0.23). In Ohio, county-level correlation was strongest for synthetic opioids (ρ = 0.71), followed by heroin (ρ = 0.69) and methamphetamine (ρ = 0.67). At the state level, the multivariate incidence rate ratio (IRR) for synthetic opioids was 1.16 (95% CI, 1.14-1.19; P < .001), and at the county level, the IRR was 1.13 (95% CI, 1.09-1.17; P < .001), suggesting that for every 1-SD increase in the UDT positivity rate, there were 16.2% and 12.8% increases, respectively, in monthly overdose deaths. Both methamphetamine (11.7% increase per 1-SD increase in UDT positivity rate; IRR, 1.12; 95% CI, 1.09-1.14; P < .001) and cocaine (5.1% increase per 1-SD increase in UDT positivity rate; IRR, 1.05; 95% CI, 1.03-1.07; P < .001) also had significant positive associations with mortality rates, but the effect sizes were smaller than that of synthetic opioids (IRR, 1.16).

Conclusions and Relevance  In this study, UDT results were highly correlated with mortality rates at national, state, and county levels. These findings suggest that real-time UDT surveillance can help to quickly identify changes in drug use patterns that might inform targeted harm reduction strategies designed to prevent overdose deaths.

Despite the enactment of essential multipronged public health strategies aimed at reducing drug overdose deaths,^1-3 more people in the US died of drug overdoses in 2020 than in any previously recorded year,⁴ increasing 30% since 2019. Deaths involving synthetic opioids, stimulants, and prescription opioids increased between 2019 and 2020, whereas heroin-involved deaths decreased.⁴ The coinvolvement of synthetic opioids across all drug categories also increased, generating concerns about polysubstance use.⁵ These statistics suggest that additional measures are needed to reduce the harmful impact of substance use.

One priority identified by the US Department of Health and Human Services is the need to strengthen epidemiological efforts through data-based public health surveillance.⁶ Data sources include poison control centers, emergency departments, electronic health records, and crime laboratories^7-10; data from some of these sources have been correlated with drug overdose mortality.^10-14 For example, an increase in synthetic opioids identified in crime laboratory data was strongly correlated with drug overdose deaths in Ohio counties.¹⁰

Another data source that can potentially identify drug use patterns in communities is urine drug test (UDT) results. There are primarily 2 types of UDT technologies: immunoassay and mass spectrometry.¹⁵ Compared with immunoassay, mass spectrometry (eg, liquid chromatography–tandem mass spectrometry [LC-MS/MS]) is a more accurate method of detecting specific drugs and metabolites at quantitative levels.¹⁵ National UDT guidelines have established recommendations for the use of LC-MS/MS in the clinical setting.^15-18

The objective of this cross-sectional study was to use the Millennium Health proprietary national database of real-time UDT results to evaluate the correlation between UDT results collected in a substance use disorder (SUD) treatment setting and overdose mortality data identified in the National Vital Statistics System of the Centers for Disease Control and Prevention (CDC WONDER).⁴ Although UDT data can help to track timely changes in substance use,^19-22 no studies, to our knowledge, have evaluated the correlation between a national database of UDT results and overdose mortality rates. Establishing the association of specific UDT results with overdose mortality rates would provide a foundation to explore whether real-time drug use surveillance data can be used to inform emerging patterns in risk of drug-specific overdose mortality that might allow for more agile responses to harm reduction and fewer overdose deaths. We hypothesized that emerging patterns in opioid and stimulant drug use captured in aggregated UDT results would correlate with overdose deaths.

We conducted a retrospective cross-sectional study of UDT results collected from patients between January 1, 2013, and December 31, 2020. Drug testing was ordered by clinicians who specialized in SUD treatment and was based on medical necessity. Ordered UDT results were obtained from all 50 US states. Specimens were analyzed using LC-MS/MS. The LC-MS/MS assays used were laboratory developed, with performance characteristics determined by Millennium Health in San Diego, California. Millennium Health is certified by the Clinical Laboratory Improvement Amendments and accredited for high-complexity testing by the College of American Pathologists. The study protocol was approved by the Aspire Independent Review Board (Santee, California) with a waiver of informed consent because of the use of deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.

To align with the National Vital Statistics System (CDC WONDER) multiple cause of death (MCD) mortality files,²³ we selected the following 5 drug categories for analysis (with tested analytes given in parentheses): cocaine (benzoylecgonine), heroin (6-monoacetylmorphine), methamphetamine (methamphetamine), synthetic opioids (fentanyl, N-desmethyl-tramadol, norfentanyl, O-desmethyl-tramadol, and tramadol), and other opioids (codeine, hydrocodone, hydromorphone, morphine, norhydrocodone, noroxycodone, oxycodone, and oxymorphone). If any parent drug or metabolite within a category was detected, the category was considered to have positivity for that specimen. A list of prescribed medications for each study participant was noted on the requisition form by the clinician ordering the UDT. We excluded patients with positive test results for cocaine, fentanyl, methamphetamine, tramadol, or other opioids that were reported by clinicians to be currently prescribed to patients.

A single specimen for each patient 18 years of age or older was selected based on the earliest specimen collection date to remove repeated measurements. A total of 500 000 unique patients from the Millennium Health proprietary UDT database were randomly selected with equal probability, and their data were used for subsequent analyses. Monthly, quarterly, and yearly total UDT positivity rates were calculated for each of the 5 drug categories. Drug positivity rates were further stratified by codetection of synthetic opioids and by the state or county of the patient’s residence.

Drug overdose deaths were identified in the CDC WONDER MCD mortality files²³ by using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), underlying cause of death codes, including codes for unintentional death (X40–X44), suicide (X60–X64), homicide (X85), and death with undetermined intent (Y10–Y14). Stratification by drug involvement was defined using the following ICD-10 MCD codes: cocaine (T40.5), heroin (T40.1), methamphetamine (psychostimulants with abuse potential; T43.6), synthetic opioids (T40.4), and other opioids (T40.2). Overdose deaths from January 1, 2013, to December 31, 2019, were examined and stratified by state, MCD drug category, and synthetic opioid coinvolvement. Raw death rates were reported per 100 000 people. Yearly population estimates were used for both monthly and yearly mortality rates.²⁴ Overdose deaths of fewer than 10 per period and/or region are masked in the CDC WONDER database.

Ohio resident county-level provisional overdose deaths (2013-2020) were collected from the Ohio Department of Public Health website on October 8, 2021.²⁵ Stratification by coinvolvement of the 5 drug categories was performed, and monthly overdose deaths were summed quarterly for analysis. The Ohio Department of Public Health does not mask county overdose mortality estimates based on a low-threshold number of deaths. All overdose mortality rates are available in eTable 1 in the Supplement.

Race and ethnicity were not considered in the current analysis due to a lack of corresponding attributes in the UDT positivity data. Other demographic characteristics, such as sex and age, were also not considered.

Correlation analyses were performed to understand the correlation between overdose mortality and drug-specific UDT measures at the national, state, and county levels. Univariate and multivariate regression models were used to assess the independent (univariate) and additive relative associations of each UDT rate with overall overdose mortality rates.

Spearman ρ was used to estimate the correlation between corresponding UDT results and overdose mortality rates. Correlations were performed at the national, state, and county levels. States were required to have data for both overdose and UDT rates for all 7 years (2013-2019) for all 5 drug categories, and the mean number of UDT patient specimens for each state had to equal or exceed 100 for the same strata. This filtering ensured that only states with sufficient population size, mortality rate, and UDT sample volume were included. County-level analysis was based on Ohio counties that had data on both overdose and UDT rates for all 8 years (2013-2020) for all 5 drug categories, and the mean number of UDT patient specimens for each county had to equal or exceed 20 for the same strata.

Poisson regression analysis was used to model total overdose deaths at the state and county levels without stratification by drug class. Regression models included standardized (mean [SD], 0 [1]) total monthly UDT positivity rates for all 5 drug categories as continuous explanatory variables and total overdose deaths as the dependent variable. State-level models included the state as a categorical variable to account for state-level differences in mortality. County-level models also used all 5 drug rates calculated per quarter (32 total quarters) as continuous explanatory variables. County was added as a categorical covariate with a function similar to that of the state models. Both multivariate models (with UDT rates for all 5 drug categories included) and univariate models (with each UDT drug rate modeled separately) were evaluated. Population size (log transformed) was used as an offset variable in all models. Standard error estimates were robust (sandwich estimates) to account for mild distributional violations. Incidence rate ratios (IRRs) with 95% CIs were reported. Data were analyzed using R software, version 4.0.3 (R Foundation for Statistical Computing). Statistical significance was set at P < .05; all tests were 2-tailed.

A total of 500 000 unique patient specimens were collected from SUD treatment practices; 288 534 specimens (57.7%) were from men, and 211 466 (42.3%) were from women, with a median age of 34 years (IQR, 17-51 years). A total of 261 998 specimens (52.4%) were from patients younger than 35 years (Table 1). Most referring health care practices were in the East North Central (116 716 practices [23.3%]), Pacific (96 948 practices [19.4%]), and South Atlantic (86 732 practices [17.3%]) US census regions. The UDT positivity rates ranged from 27 278 specimens (5.5%) for heroin to 105 966 specimens (21.2%) for other opioids. The number of specimens analyzed per year varied from 51 130 in 2017 to 89 503 in 2015.

Since 2013, overdose deaths associated with cocaine, heroin, methamphetamine, and other opioids increased when found in conjunction with synthetic opioids (Figure 1, dark blue line). In contrast, since 2016, overdose deaths associated with all of these drugs, with the exception of methamphetamine, decreased or remained stable when synthetic opioids were absent (Figure 1, light blue line). In general, UDT positivity rates aligned with overdose mortality data (Figure 1, light and dark orange lines). Overall, UDT positivity for total synthetic opioids increased from 2.1% in 2013 to 19.1% in 2020 (a 792.5% increase) (eTable 2 in the Supplement). This increase was, in part, associated with coinvolvement with other drugs and was greatest for co-occurrence with other opioids (from 1.1% in 2013 to 9.2% in 2020) and methamphetamine (from 0.1% in 2013 to 7.9% in 2020) over the study period. As observed with overdose deaths, UDT positivity in the absence of synthetic opioids decreased for all drug categories from 2013 to 2020 (7.7% to 4.7% for cocaine; 3.9% to 1.6% for heroin; 20.5% to 6.9% for other opioids) with the exception of methamphetamine (2.1% to 13.1%). Coinvolvement of synthetic opioids for all 4 drug categories increased between 2019 and 2020, with methamphetamine having the largest increase at 98.7% (from 4.0% to 7.9%) (eTable 2 in the Supplement).

National monthly UDT results for all 4 drug categories were consistent with these patterns. Overall, methamphetamine and synthetic opioids were highly correlated with overdose mortality rates (ρ = 0.96 for both), and all 4 drug categories were highly correlated with overdose mortality rates when synthetic opioids were coinvolved (Table 2). Spearman correlation coefficients in the presence of synthetic opioids over this period were 0.98 for methamphetamine, 0.94 for cocaine, 0.83 for other opioids, and 0.78 for heroin. In the absence of concurrent synthetic opioids, UDT correlations with overdose mortality rates persisted across all drug categories (ρ = 0.88 for methamphetamine, 0.81 for heroin, and 0.75 for other opioids) with the exception of cocaine (ρ = −0.37).

After finding a correlation between national overdose mortality and specific UDT positivity rates, we examined the data at a more granular level using a subset of 24 states with sufficient power to explore these correlations. The correlation of UDT positivity rates with overdose mortality rates varied significantly between drug classes and the year being compared (Figure 2; eTable 3 in the Supplement). For example, Spearman correlation coefficients ranged from 0.04 for other opioids in 2013 to 0.85 for methamphetamine in 2016. The median correlation of UDT positivity rates with total overdose mortality rates from 2013 to 2019 was greatest for methamphetamine (ρ = 0.80), followed by synthetic opioids (ρ = 0.77), cocaine (ρ = 0.73), heroin (ρ = 0.54), and other opioids (ρ = 0.31). Stratified rates also revealed high median correlation with methamphetamine in the absence of synthetic opioids (ρ = 0.79) and cocaine in the presence of synthetic opioids (ρ = 0.81). Other opioids had the lowest correlation in the absence of synthetic opioids (ρ = 0.23).

Within-state correlation of UDT positivity rates with overdose mortality rates was also evaluated across years rather than within years to assess whether rates were consistent over time within a given state (Table 2; eTable 4 in the Supplement). The correlations varied significantly by state and drug class. For example, the median yearly correlation for included states was highest for methamphetamine (ρ = 0.91), followed by synthetic opioids (ρ = 0.80), cocaine (ρ = 0.43), heroin (ρ = 0.32), and other opioids (ρ = 0.18).

We also evaluated the correlation of temporal patterns for 19 counties in Ohio to assess whether yearly UDT rates were correlated with overdose mortality rates at finer spatial resolution. Spearman correlation coefficients within counties and across years (2013-2020) for different drug categories are shown in Table 2 and eTable 5 in the Supplement. Synthetic opioids had the highest median correlation with overdose deaths across counties (ρ = 0.71), followed by heroin (ρ = 0.69), methamphetamine (ρ = 0.67), other opioids (ρ = 0.34), and cocaine (ρ = 0.24).

Poisson regression analysis was used to better understand the association between total state-level overdose deaths and UDT positivity rates collected monthly. The IRRs for both multivariate and univariate models (with each UDT rate modeled separately) are shown in Table 3. All UDT rates at the state level, with the exception of heroin, were found to be significant in the multivariate model, and the overall model fit was good (McFadden R² = 0.62). The multivariate IRR for synthetic opioids was 1.16 (95% CI, 1.14-1.19; P < .001), suggesting that for every 1-SD increase in the UDT positivity rate, there was a 16.2% increase in monthly overdose deaths. Both methamphetamine (11.7% increase per 1-SD increase in UDT positivity rate; IRR, 1.12; 95% CI, 1.09-1.14; P < .001) and cocaine (5.1% increase per 1-SD increase in UDT positivity rate; IRR, 1.05; 95% CI, 1.03-1.07; P < .001) also had significant positive associations with mortality rates, but the effect sizes were smaller than that of synthetic opioids (IRR, 1.16). In contrast, other opioids had a significant negative association with mortality rates (IRR, 0.86; 95% CI, 0.84-0.88; P < .001). Univariate coefficients suggested a similar pattern, with the only major difference found for heroin, which had a significant negative association with mortality rates in the univariate model (IRR, 0.94; 95% CI, 0.92-0.96; P < .001) but no association with mortality rates in the multivariate model (IRR, 1.01; 95% CI, 0.99-1.03; P = .37).

Regression models were also used to better understand the association between Ohio county-level total overdose mortality and standardized UDT positivity rates collected on a quarterly basis (2013-2020). Multivariate model fit was good (McFadden R² = 0.27), although not as high as the state-level model (Table 3). Cocaine, synthetic opioids, and other opioids were found to be significantly associated with overdose deaths in the multivariate model. Synthetic opioids had the highest relative IRR of 1.13 (95% CI, 1.09-1.17; P < .001) in the multivariate model compared with 1.15 (95% CI, 1.11-1.20; P < .001) in the univariate model. As observed in the state-level model, other opioids had a significant negative association with overdose deaths (IRR, 0.89; 95% CI, 0.83-0.95; P < .001) and cocaine had a significant positive association with overdose deaths (IRR, 1.07; 95% CI, 1.03-1.12; P = .002) in the multivariate model. Notably, methamphetamine was not associated with overdose deaths in the multivariate model (IRR, 1.01; 95% CI, 0.96-1.07; P = .63), although it was significantly associated with overdose deaths in the univariate model and had the largest relative effect size, with an IRR of 1.16 (95% CI, 1.10-1.22; P < .001). As observed in the state-level regression model, heroin was not associated with overdose deaths in the multivariate model (IRR, 0.98; 95% CI, 0.92-1.05; P = .63) but had a significant negative association with overdose deaths in the univariate model (IRR, 0.89; 95% CI, 0.85-0.94; P < .001).

This cross-sectional study found strong correlation between UDT data from SUD treatment practices and overdose mortality across 5 drug categories over a similar period at national, state, and county levels. At the national level, UDT positivity for synthetic opioids and methamphetamine were both highly correlated with overdose mortality rates. Given their potency and synthetic nature, fentanyl and methamphetamine can be inexpensively produced by drug trafficking organizations, contributing to substantial profits and widespread availability^26,27 that have increased the involvement of these substances in overdose deaths, with fentanyl ranking first and methamphetamine ranking second for overdose deaths in 2020.⁴ In this study, correlation with overdose deaths was strongest for synthetic opioids and methamphetamine, which is likely explained by their availability and lethality as well as the finding that all drug categories were highly correlated with overdose deaths when synthetic opioids were coinvolved. In the absence of synthetic opioids, all drug categories, with the exception of cocaine at the national level, were highly correlated with overdose deaths.

Similar to the national findings, synthetic opioids and methamphetamine had the highest median state-level correlation with overdose deaths over time, likely because of the notable involvement of both drugs in current overdose mortality patterns. Other opioids had the lowest correlation for both total positivity and positivity in the absence of synthetic opioids. Deaths involving prescription opioids remained lower than peak rates in 2017,⁴ concurrent with a decrease in opioid prescribing practices and manufacturing.²⁷ Many counterfeit tablets purported to be prescription-grade opioids are now being identified as containing illicitly manufactured fentanyl.²⁷ These factors may have had implications for the correlation between UDT results and mortality rates for other opioids, especially in the absence of synthetic opioids.

The multivariate models used to assess the associations of each UDT rate with overall overdose deaths allowed a more holistic understanding of the association of UDT measures with mortality rates. For example, at a county level, we were able to observe that synthetic opioids and methamphetamine were positively associated with mortality rates, whereas heroin and other opioids were negatively associated. This observation makes sense given that mortality rates increased at a relatively consistent rate over the study period, similar to methamphetamine and synthetic opioid UDT rates. However, heroin and other opioid deaths that did not coinvolve synthetic opioids decreased or remained stable,^4,5 which may explain the coefficients observed.

Some fluctuations were observed in the strength of the correlation within a given year. Not all synthetic opioids (eg, fentanyl analogues) were included in the analysis, although some have been associated with increases in drug overdose deaths in certain areas.²⁸ Other factors that may have had implications for yearly correlation include UDT sample volume, external interventions (eg, naloxone distribution) in certain areas, and the emergence of the COVID-19 pandemic, which had wide-ranging consequences for public health and the opioid crisis. These factors may have also been associated with fluctuations in state- and county-level findings, particularly where variations existed in community-led interventions.

These UDT results were collected among a population seeking health care in an SUD treatment setting. Patients in this setting typically have higher rates of UDT positivity for illicit drugs than the general population, placing them at increased risk for overdose; this factor may have played a role in the strength of correlation.

Although UDT serves as a medically necessary tool for clinicians to detect potentially problematic illicit drug use and relapse that may lead to overdose events and deaths among patients during the treatment and recovery periods, secondary use of this information can help public health officials and community leaders develop real-time responses to the substance use crisis. Because of the long lag time between data availability for specific drugs and overdose deaths, timely UDT results that are both temporally and geospatially linked may provide a window into local and regional changes to inform reallocation of harm reduction strategies, such as distribution of naloxone or fentanyl test strips, in communities in which emerging patterns of increased rates of specific drug categories or co-occurrence of synthetic opioids and stimulants might be used to estimate an increasing risk of overdose deaths.

We were able to observe associations between the extent of UDT positivity rates and the community rates of drug-specific overdose mortality. Using state-level data, every 1-SD increase in synthetic opioid UDT positivity was associated with a 16.2% increase in overdose deaths. This finding supported our hypothesis that UDT results may provide early insights into emerging drug use patterns associated with overdose mortality rates. Extension of our findings to the development of predictive models incorporating UDT results will be important to assess whether secondary use of these data can inform agile responses at a community, state, or national level to reduce overdose deaths.

This study has several limitations. Not all psychoactive substances (eg, methadone, benzodiazepines, and fentanyl analogues) that affect overdose mortality rates were included in this study. Additional social factors associated with health (eg, unemployment and poverty rates) were not included in the models, despite their known impact.¹⁰ Various interventional data, such as opioid prescribing rates and naloxone use, were also not included. Data from UDTs may be subject to selection bias that was not accounted for in these models and may impact UDT rates spatially and temporally. In addition, these UDT results were collected in SUD treatment practices, and findings may not reflect patterns in the general population. Patients included in the analysis may have had an incomplete or inaccurate medication list from the ordering clinician, which would result in possible inclusion of positive results associated with an unreported prescription medication. The COVID-19 pandemic may have affected UDT rates, and its consequences for these models is unknown. The Ohio 2020 mortality data are currently provisional and not estimated in the same way as CDC WONDER data. A recent study²⁹ found that MCD designations on death certificates are incomplete and often erroneous, which could impact correlation and prediction studies.

This cross-sectional study found that UDT results were correlated with overdose mortality rates at national, state, and county levels, with synthetic opioids and methamphetamine having the strongest correlations with mortality rates. These findings suggest UDT is a valuable data source that is capable of informing real-time surveillance efforts to alert communities about changes in the drug environment that might be associated with drug overdose deaths.

Accepted for Publication: April 7, 2022.

Published: June 3, 2022. doi:10.1001/jamanetworkopen.2022.15425

Open Access: This is an open access article distributed under the terms of the CC-BY-NC-ND License. © 2022 Whitley P et al. JAMA Network Open.

Corresponding Author: Penn Whitley, BA, Millennium Health, 16981 Via Tazon, San Diego, CA 92127 (penn.whitley@millenniumhealth.com).

Author Contributions: Mr Whitley had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: All authors.

Acquisition, analysis, or interpretation of data: Whitley, LaRue, Fernandez, Dawson, Jackson.

Drafting of the manuscript: Whitley, LaRue, Passik, Dawson.

Critical revision of the manuscript for important intellectual content: Whitley, LaRue, Fernandez, Jackson.

Statistical analysis: Whitley, Fernandez.

Administrative, technical, or material support: Whitley, LaRue, Jackson.

Supervision: Passik, Dawson.

Conflict of Interest Disclosures: Dr LaRue reported receiving personal fees from Millennium Health during the conduct of the study. Dr Jackson reported receiving grants from the National Institutes of Health (NIH) outside the submitted work. No other disclosures were reported.

Funding/Support: This research was supported by funding from Millennium Health, which included employment salaries for all authors (with the exception of Drs Fernandez and Jackson), and partially supported by grant UL1TR002733 from the National Center for Advancing Translational Sciences of the NIH (Drs Fernandez and Jackson).

Role of the Funder/Sponsor: Millennium Health provided technical assistance on the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, and approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Additional Contributions: We thank Angela Huskey, PharmD, CPE, of Millennium Health, for her support of the study and assistance with manuscript review. Dr Huskey received no additional compensation outside of usual salary for her contribution.