Provider reasons for discontinuing long-term opioid therapy following aberrant urine drug tests differ based on the type of substance identified

Abstract

Objective:

Urine drug testing (UDT) is increasingly performed as a means of identifying aberrant behavior that may be grounds for discontinuation of long-term opioid therapy (LTOT). Little is known, however, about the ways in which positive UDT results may differentially inform decisions to discontinue LTOT based on the type of substance for which the UDT screened positive. The aim of this study was to examine the likelihood of clinician-initiated discontinuation of LTOT attributed to positive UDT results across three discrete categories of substances: (1) cannabis, (2) alcohol or illicit substances (excluding cannabis), and (3) controlled prescription medications that were not prescribed.

Design:

This retrospective study utilized the US Department of Veterans Affairs (VA) Health Care System. Corporate Data Warehouse to assemble a sample of 600 patients with substance use disorders and matched controls who were discontinued from LTOT in 2012. Comprehensive manual medical record review identified UDT results in the year prior to discontinuation and reason(s) for discontinuation.

Patients, Participants:

Patients with one or more UDTs positive for a single substance (N = 185) comprised the study sample.

Main Outcome Measure(s):

Likelihood of clinician-initiated discontinuation attributed to a positive UDT across the three categories.

Results:

Patients with one or more UDTs positive for cannabis were more likely to be discontinued from opioid therapy as a result of the positive UDT compared to those with one or more UDTs positive for nonprescribed prescription medication (adjusted odds ratio [OR] = 18.05, 95% CI = 7.29-44.66). Similarly, patients with UDTs positive for alcohol or illicit substances were more likely to be discontinued for the positive UDTs relative to patients who tested positive for nonprescribed prescription medications (adjusted OR = 13.10, 95% CI = 4.81-35.68). No difference in UDT-related discontinuation decisions was evident between patients with UDTs positive for alcohol/illicit substances versus cannabis (adjusted OR = 1.47, 95% CI = 0.57-3.77).

Conclusions:

High odds of UDT-related discontinuation were found in patients who tested positive for cannabis, alcohol, or illicit substances, relative to nonprescribed prescription medications.

There is growing recognition that long-term opioid therapy (LTOT) for the treatment of chronic pain comes with substantial risks, while limited evidence supports its long-term effectiveness.¹ Identified risks of LTOT include increased rates of fractures, sexual dysfunction, constipation, myocardial infarction, sleep-disordered breathing, opioid use disorder, overdose, and mortality.^1-4 Indeed, the increased rate of opioid prescribing in recent decades has been accompanied by a sharp and persistent rise in opioid-related adverse events.⁵

In response, clinical practice guidelines were issued by the Centers for Disease Control and Prevention and the US Department of Veterans Affairs/Department of Defense that aimed to encourage safe opioid prescribing.^6,7 One key recommendation included in these guidelines is routine urine drug testing (UDT) to identify substance misuse that may be grounds for tapering or complete discontinuation of LTOT.⁸

Routine UDT aims to identify aberrant behaviors such as concurrent use of prescribed opioids and nonopioid substances (including alcohol) or absence of prescribed opioids, which may suggest other forms of opioid misuse or diversion. The concurrent use of prescribed opioids and nonopioid substances has been linked to increased risk of emergency department visits, inpatient admissions for overdose, and overdose death.^9-12 Thus, UDT results that are positive for illicit substances or prescription medications that were not prescribed may alter providers’ opioid risk-benefit calculation for their patients, and lead to clinical decisions to decrease patients’ prescribed opioid dose or discontinue LTOT altogether. While research has begun to track the demographic and clinical correlates of LTOT discontinuation,^13,14 reasons for discontinuation,¹⁵ and patient outcomes following discontinuation,^16-18 little is known empirically about the ways in which positive UDT results may differentially inform decisions to discontinue LTOT based on the type of substance for which the UDT screened positive. Yet, providers have both medically and culturally informed beliefs about substances that may lead them to respond differentially to positive UDTs.

Cannabis has become increasingly accepted for both medical and recreational use. Medical cannabis is legal in 28 states and the District of Columbia, while eight states legalized recreational cannabis between 2012 and 2016 alone.¹⁹ There is also preliminary support, albeit limited, that cannabis may be beneficial in treating some forms of chronic pain.^20,21 Thus, there is reason to believe that UDTs positive for cannabis may be viewed as less concerning than noncannabis substances, and clinicians may be subsequently less likely to discontinue prescribing opioids based on the cannabis-related aberrant UDT. In contrast, provider concern may be piqued by UDT results positive for substances that increase the likelihood of experiencing an adverse event or that indicate a possible substance use disorder (SUD). For instance, guidelines recommend tapering or discontinuation of LTOT among patients with untreated alcohol or other SUDs,⁷ based on research linking SUDs with a substantially heightened risk of overdose.^22,23 Thus, clinicians made aware of nonprescribed controlled substance use, illicit substance use, and/or alcohol use via a positive UDT may have concerns about patient safety and be more likely to discontinue opioids.

The aim of this study was to compare differences in the odds of clinician-initiated discontinuation of LTOT that were attributed to positive UDT results across three discrete categories: UDTs positive for (1) cannabis, (2) an illicit substance excluding cannabis (eg, cocaine, amphetamine) or alcohol, or (3) a prescription medication that was not prescribed by a veterans affairs (VA) clinician (eg, benzodiazepines, a different opioid). We hypothesized that patients testing positive for cannabis would be less likely to be discontinued from LTOT as a result of their positive test than patients testing positive for either illicit substances/alcohol or nonprescribed controlled substances.

METHODS

This study was approved by the VA Portland Health Care System Institutional Review Board.

Data source and sample selection

We used the VA Corporate Data Warehouse, which contains all information contained in VA patients’ electronic health record (EHR), to identify a national cohort of VA patients who were prescribed LTOT by a VA clinician. We first identified patients prescribed opioid therapy for all of 2011, allowing refill gaps of 30 days or fewer between opioid prescription fills. Prior studies examining discontinuation of LTOT among patients have allowed for short refill gaps to account for logistical barriers to refill that do not constitute discontinuation.^13,14 These patients comprised the initial cohort of patients on LTOT. We next identified patients from this cohort who ceased being prescribed opioids from the VA for 12 or more consecutive months starting in 2012. The “index date” was defined as the date of last opioid fill. The 12-month discontinuation interval allowed for intermittent discontinuations that stemmed from personal changes such as geographic relocation, rather than clinical decisions. Similar to prior studies, patients receiving opioids within the context of surgery, cancer pain, or opioid agonist therapy (ie, buprenorphine or methadone maintenance), were excluded, as were patients who had no VA contact or died in the year following discontinuation.^14,15 Sampling procedures identified a cohort of 7,247 patients who discontinued LTOT in 2012: 1,868 patients (26 percent) with a diagnosed SUD and 5,739 patients (74 percent) without SUD. SUD diagnosis was defined as having an ICD code for a SUD diagnosis linked to a medical visit encounter. A medical visit encounter included encounters with any treating clinician and could be for the purpose of medical, mental health, or other healthcare needs. To obtain a sample with high likelihood of discontinuation due to aberrant behaviors such as positive UDTs, and because an additional aim of the parent study was to compare reasons for opioid discontinuation between patients with and without SUD,¹⁵ we randomly sampled 300 patients with SUD and used propensity scores to match 300 patients with no diagnosed SUD. The sample size was selected as one sufficiently large to be able to detect differences in reasons for opioid discontinuation between patients with and without SUDs. Additional details about the propensity score models and matching procedures have been published elsewhere.¹⁵

Electronic health record review

A research associate with extensive experience reviewing the EHR for VA patients in opioid- and pain-specific studies conducted a comprehensive EHR review of the 600 patients in this sample. Chart review ascertained the reasons for discontinuation of opioid therapy. Discontinuation stemming from positive UDTs required explicit language associating the discontinuation decision with the UDT result, for example, “pt [patient] is + cocaine on UDS, violated pain contract. Cannot provide him any narcotics.” For this study, we utilized data pertaining to UDTs positive for a single substance only (see Figure 1 for sample selection). Specifically, patients were included if they tested positive one or more times for a single substance in the year prior to opioid discontinuation (n = 185), but were excluded if they had positive UDTs for more than one substance in the year prior to discontinuation (n = 99). This approach allowed us to isolate the association of the type of substance for which a UDT was positive on providers’ decisions to discontinue LTOT, as tests positive for multiple substances precluded us from determining the relative influence of specific substances on the opioid discontinuation decision. Because this study focused on opioid discontinuation decisions related to positive UDT results, we also excluded n = 55 patients whose only aberrant UDT results were due to the absence of a prescribed opioid.

Figure 1.

Sample selection for patients testing positive for a single substance and discontinued from LTOT.

Patients discontinued as a result of a positive UDT were coded as positive for alcohol, cannabis, nonprescribed opioids, nonprescribed sedatives/hypnotics (including benzodiazepines), cocaine or amphetamines. We also assessed for positive UDTs resulting from use of hallucinogens, but no patients tested positive solely for hallucinogens. UDTs positive for nonprescribed prescription medications were cross referenced with the patient’s current VA prescriptions for opioids, sedatives/hypnotics and amphetamines at the date of the UDT. Results of UDTs were categorized as aberrant for a nonprescribed prescription medication when the date of the positive UDT result did not overlap windows of VA prescribing for the prescription medications. Because UDTs tested for amphetamines generally, we were unable to differentiate between illicit amphetamines (eg, methamphetamine) and prescribed amphetamines (eg, Adderall). When patients were not prescribed amphetamines by the VA, UDT results positive for amphetamines were assumed to be due to the illicit amphetamine use.

Some patients in the study sample who tested positive on one or more UDTs were discontinued from opioids for reasons other than the UDT result. Additional reasons for LTOT discontinuation included patients discontinuing of their own volition, other aberrant behaviors such as substance use (without UDT confirmation), opioid diversion, patient safety concerns (eg, due to prior history of overdose), lack of efficacy of opioids to improve pain or functioning, and transfer of care to a non-VA facility. In the current study, these patients were categorized as discontinuing opioids for reasons other than the aberrant UDT result. A detailed description of the chart review development, training, and ongoing quality assurance have been described elsewhere.¹⁵

Administrative data

Data abstracted from the VA Corporate Data Warehouse included demographic information (age, gender, race/ethnicity), rurality of one’s residence based on rural-urban commuting area codes,²⁴ military service-related disability status, average daily dose in milligrams of morphine equivalent, and medical comorbidities as assessed by the Elixhauser Comorbidity Measure.²⁵ Diagnoses of SUD and mental health disorders were obtained for the 12 months prior to discontinuation of LTOT. Finally, administrative data identified the number and results of UDTs for each patient in the year prior to discontinuation of LTOT.

Statistical analyses

Patients with UDTs positive for a single substance were grouped into one of three mutually exclusive categories: UDT positive for (1) cannabis only, (2) alcohol or illicit substances (amphetamines and cocaine) only, and (3) prescription medications that were not prescribed only (benzodiazepines, barbiturates, and opioids). Substances were sorted into categories based on prior literature suggesting that providers’ beliefs about the effects of such substances may differ based on the type of substance.^17,26 For instance, UDT positive for alcohol, a noncannabis illicit substance, or a nonprescribed prescription medication may trigger concern regarding the potential for an adverse event,^10,22,23 whereas cannabis may be viewed as a viable pain management strategy with limited potential harms.²⁰ The outcome of interest was whether the positive UDT was listed as the reason for discontinuation of LTOT in the medical record versus chart-documented discontinuation for a reason other than the positive UDT. Descriptive statistics characterized patient demographic and clinical variables, as well as reasons for LTOT discontinuation. Unadjusted logistic regression models examined the likelihood of clinician-initiated discontinuation attributed to a positive UDT for cannabis versus alcohol or illicit substances versus prescription substances that were not prescribed by VA clinicians. Logistic regression models were then adjusted for variables that have been associated with opioid discontinuation in prior research^13,14: age, gender, race, average daily opioid dose, the presence of a mental health diagnosis, and the presence of a SUD diagnosis. We also conducted two sensitivity analyses. In the first, we restricted the sample to patients who had a single aberrant test in the year prior to discontinuation, excluding those with multiple aberrant tests in this timespan, as decision to discontinue LTOT may differ based on the number of aberrant UDTs. In the second, we limited inclusion in the illicit substance/alcohol group to those aberrant for illicit substances only, excluding those with aberrant tests due to alcohol. Within VA primary care, alcohol is not routinely tested for, but must be added on to a standard UDT. Thus, patients were not consistently tested for alcohol, and the presence of a test likely indicates that providers had reason to believe that patients were at risk of alcohol use. Analyses were conducted using Stata version 14. An α = 0.05 and two-tailed significance tests were used for all inferential analyses.

RESULTS

From the sample of 600 patients included in the chart review, 284 had at least one positive UDT result in the year prior to opioid discontinuation. Of these patients, we excluded 99 who tested positive for multiple substances, leaving a total sample of 185 patients who tested positive for a single substance only. Clinical and sociodemographic characteristics of the study sample are provided in Table 1. Patients were 54 years old on average and predominantly male (95 percent). The majority, 68 percent, were White, non-Hispanic, 18 percent were Black, 3 percent Hispanic, and 11 percent did not have race/ethnicity data available. In the year prior to the date of LTOT discontinuation, 58 percent of patients had been diagnosed with a mental health disorder, and 57 percent had been diagnosed with a SUD.

Table 1.

Sample characteristics of patients with one or more UDT’s positive for a single substance

Variable	Cannabis (n = 85)	Illicit substance/alcohol (n = 39)	Non-RX controlled substance (n = 61)
Sociodemographic characteristics, percent (N)
Age (M ± SD)	52.8 ± 10.6	55.4 ± 7.6	55.5 ± 10.3
Male gender	95.2 percent (81)	97.4 percent (38)	93.4 percent (57)
Race/ethnicity
White, Non-Hispanic	64.7 percent (55)	79.4 percent (31)	83.61 percent (51)
Black, Non-Hispanic	16.5 percent (14)	7.7 percent (3)	4.92 percent (3)
Hispanic	11.8 percent (10)	10.2 percent (4)	6.56 percent (4)
Other/Unknown	7.1 percent (6)	12.82 percent (5)	4.92 percent (3)
Clinical characteristics, percent (N)
Service connected disability	62.4 percent (53)	51.3 percent (20)	52.4 percent (32)
Prediscontinuation mental health diagnosis	62.4 percent (53)	51.3 percent (20)	70.5 percent (43)
Anxiety diagnosis	18.8 percent (16)	28.2 percent (11)	42.6 percent (26)
Bipolar diagnosis	3.5 percent (3)	5.1 percent (2)	11.5 percent (7)
Depression diagnosis	30.6 percent (26)	25.4 percent (6)	24.6 percent (15)
Psychosis diagnosis	3.5 percent (3)	12.8 percent (5)	9.8 percent (6)
PTSD diagnosis	37.7 percent (32)	15.4 percent (6)	39.3 percent (24)
Prediscontinuation SUD diagnosis	49.4 percent (42)	61.5 percent (24)	63.9 percent (39)
Alcohol	18.8 percent (16)	41.0 percent (16)	34.4 percent (21)
Amphetamine	1.2 percent (1)	5.1 percent (2)	1.6 percent (1)
Cannabis	10.6 percent (9)	5.1% (2)	1.6 percent (1)
Cocaine	1.2 percent (1)	25.6 percent (10)	3.3 percent (2)
Opioids	8.2 percent (7)	10.3% (4)	26.2 percent (16)
Number of positive tests, percent (N)
1	75.3 percent (64)	74.4 percent (29)	77.1 percent (47)
2	20 percent (17)	23.1 percent (9)	13.1 percent (8)
3	2.4 percent (2)	0	3.3 percent (2)
4 +	2.4 percent (2)	2.6 percent (1)	6.6 percent (4)
Discontinuation linked to Positive UDT	76.5 percent (65)	74.3 percent (29)	21.3 percent (13)

To examine the proportion of patients testing positive for a substance who were eventually discontinued as a result of the positive test, patients were grouped into three mutually exclusive categories based on UDT positive test results: cannabis only (n = 85 patients), alcohol/illicit substances only (n = 39 patients), and nonprescribed controlled prescription substances only (n = 61 patients). The proportions of patients discontinued from LTOT due to the positive UDT result(s) were as follows: cannabis = 76 percent (n = 65), alcohol or illicit substances = 74 percent (n = 29), and nonprescribed prescription medications = 21 percent (n = 13). The remaining patients (n = 78, 42 percent of the entire sample) discontinued LTOT for reasons other than the positive UDT result. These included suspected use of substances without confirmation by UDT (n = 29), suspected misuse of the VA-prescribed opioid(s) (n = 19), discontinuation of patients’ own volition (n = 13), patients transferring care outside of the VA (n = 7), UDT negative for a prescribed opioid (n = 5), safety concerns due to patient opioid overdose (n = 1), ineffectiveness of opioid therapy at reducing pain (n = 1), and no reason given/other (n = 15).

Table 2 provides results of unadjusted and adjusted logistic regression analyses that examined the likelihood of provider-initiated discontinuation from opioid therapy based on the type of substance for which UDTs were positive. In both the unadjusted and adjusted models, patients who tested positive for cannabis only did not differ from patients who tested positive for alcohol or illicit substances only in UDT-related reasons for discontinuation of LTOT (adjusted odds ratio [OR] = 1.47, 95% CI = 0.57-3.77). In contrast, those with one or more UDTs positive for cannabis only were more likely to be discontinued from opioid therapy as a result of the positive UDT compared to those with one or more UDTs positive for nonprescribed prescription medication only (adjusted OR = 18.05, 95% CI = 7.29-44.66). Similarly, in both unadjusted and adjusted models, patients with UDTs positive for alcohol or illicit substances were more likely to be discontinued for the positive UDTs relative to patients who only tested positive for nonprescribed prescription medications (adjusted OR = 13.10, 95% CI = 4.81-35.68.)

Table 2.

Likelihood of discontinuation due to a positive UDT based on the type of substance for which patients test positive (N = 185)^*

Substance type	Association with decision to discontinue LTOT
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Cannabis versus nonprescribed controlled substances	12.00 (5.44-26.48)	18.05 (7.29-44.66)
Cannabis versus alcohol/illicit substances	1.12 (0.47-2.69)	1.47 (0.57-3.77)
Alcohol/illicit substances versus nonprescribed controlled substances	10.70 (4.16-27.54)	13.10 (4.81 35-68)

^*Note. Adjusted models controlled for age, gender, race, presence of a military-related disability, average daily dose of opioids in morphine equivalents, prediscontinuation mental health diagnosis, and prediscontinuation SUD diagnosis. Bolded text indicates a statistically significant finding at p < 0.0.

Sensitivity analyses

We conducted two sensitivity analyses. In the first, we restricted the sample to patients discontinued following a single aberrant UDT (n = 140 patients), excluding those with multiple aberrant tests (n = 45). Although the significance of the results was unchanged, the magnitude of the OR depicting relationships between type of substances for which patients tested positive and discontinuation due to the test results decreased. Specifically, the odds of discontinuing opioids due to a positive UDT result for cannabis or illicit substances/alcohol relative to nonprescribed controlled substances declined in the fully adjusted models, from 18.05 (95% CI = 7.29-44.66) to 12.23 (95% CI = 4.58-32.72) for cannabis, and from 13.10 (95% CI = 4.81-35.68) to 9.63 (95% CI = 3.11-29.80) for illicit substances/alcohol. In the second sensitivity analysis, we tested the effect of excluding patients with UDTs positive for alcohol only from the illicit substance/alcohol group. Excluding the seven patients who tested positive for alcohol from the initial group of 39, results remained statistically significant. The coefficient on the illicit substance group increased to 27.07 (95% CI = 8.01-91.45), relative to 13.10 (95% CI = 4.81-35.69) in the original model that included patients with UDTs positive for alcohol.

DISCUSSION

This study examined the differences in odds of LTOT discontinuation based on the type of substance for which patients tested positive. Contrary to our hypothesis, patients who tested positive for cannabis had a similar odds of being discontinued from opioids due to the UDT result as patients who tested positive for alcohol or illicit substances. That patients were seen in the VA, a federal entity that does not recognize state legalization of medical or recreational cannabis, may contribute to clinicians’ decisions to discontinue LTOT as a result of cannabis use. It is also possible that cultural norms around cannabis use were less liberal in 2012 than they are today. Indeed, national polls reveal that the percent of the United States. population supporting cannabis legalization is changing rapidly, increasing from 45 to 57 percent between 2011 and 2016 alone,²⁷ and some have proposed cannabis as a “harm reduction” approach to managing pain in lieu of using other substances, including prescription opioids.²⁸ Thus, our results may not reflect changing opinions and clinical practices concerning opioid therapy for patients who are concurrent cannabis users. The finding that UDTs positive for illicit substances or cannabis were both more likely to lead to discontinuation, relative to controlled prescription medications not prescribed by a VA clinician, aligns with results from a prior study, which found that pain clinicians rank illicit drug use by patients as a more problematic aberrant behavior than prescription medication misuse.²⁶

A surprising finding was the low odds of discontinuation resulting from UDTs that were positive for nonprescribed controlled medications such as benzodiazepines and other prescription opioids, substances that when used concurrently with prescribed opioid medication heighten the risk for opioid overdose and death.^10,11,23 Some patients in our sample were previously prescribed, and discontinued, substances by VA clinicians for which they later tested positive on UDT. Providers may be more willing to overlook UDTs positive for medications that patients had been previously prescribed and potentially saved at home. However, whether these patients had leftover medication from a prior prescription is unknown and no patient EHRs contained data providing this explanation. An alternative explanation is physician misinterpretation of UDT results. Prior studies indicate that family physicians sometimes misinterpret urine drug test results, and interpretation may be more complex when results include prescription medications.^29-31 Additionally, some UDTs may produce false positive results due to cross reactivity of other substances,^32,33 and while confirmatory testing can all but rule out false positives for most substances, confimiatory testing was not ubiquitously performed at all VA sites during the study period. Thus, it is possible that a provider may not have acted on a positive UDT because they interpreted the result as a false positive, though no medical records we reviewed included notes to that effect.

Sensitivity analyses also yielded interesting results. In the first, the magnitude of the OR depicting relationships between type of substance for which patients tested positive and discontinuation due to the test results decreased in the subsample of patients discontinued following a single aberrant test, although the magnitude of effect remained statistically significant. This result suggests that providers are more likely to discontinue opioids due to serial aberrant UDTs versus a single aberrant UDT. In the second analysis, excluding patients with UDTs positive for alcohol only from the illicit substance/alcohol group, results remained significant, but the magnitude of the OR increased substantially. This suggests that illicit substances may be more likely to lead to a UDT-related discontinuation relative to alcohol, although the small number of patients positive for alcohol alone did not allow for statistical comparison.

Finally, it is notable that while the majority of patients who tested positive for a substance were ultimately discontinued from LTOT as a result of a positive UDT, a significant minority of patients were discontinued for reasons other than the positive UDT result. This may suggest that some clinicians are not routinely reviewing the results of UDTs or not factoring these results into their decision to discontinue LTOT. In a prior study that examined pain care practices following aberrant UDTs in patients prescribed LTOT, 31 percent of patients had no documentation in the medical record of planned changes to their pain care.³⁴ The authors hypothesized that some opioid-prescribing clinicians may be unaware of patients’ UDT results, pointing to the possible need for improved opioid safety monitoring practices. Alternatively, it may be that providers did factor the positive UDT into their decision to discontinue LTOT, but simply did not document this in the medical record. Others have found that pain care documented in the medical record does not fully capture the pain care patients receive.³⁵

Limitations

This study has several limitations. First, the sample comprises VA patients diagnosed with SUD and matched patients without SUD, and results are not intended to generalize to the populations of VA or non-VA patients who discontinue LTOT. Second, this study was not designed to determine the true odds of discontinuation due to positive UDT results as our sample was drawn from a population of patients who discontinued LTOT. Excluded from our sample were patients prescribed LTOT who tested positive for a substance on a UDT but were never discontinued from opioid therapy. Third, data were collected through EHR review and administrative data abstraction. There is a risk of bias resulting from a failure of clinicians to document discontinuation due to a UDT result in the EHR. Fourth, controlled medications prescribed outside of the VA are not routinely included in patients’ EHRs, and the national scope of this study did not allow us access to state prescription drug monitoring program data, which is limited to those certified within each state. We were thus unable to verify the means by which some patients received non-VA controlled substances for which they tested positive on UDTs (ie, via “legitimate” prescriptions or other means). Fifth, drug screening in the absence of confirmatory testing is subject to false positives and potential cross reactivity. Different types of UDT also have varying levels of specificity and ability to test for particular substances.^32,33 Given the national scope of this study, and the lack of consistent UDT type and substances tested for across VA facilities, we were not able to incorporate test-specific attributes that may partially account for the observed outcomes. Likewise, physician knowledge of cross reactivity that may yield a false-positive result is not accounted for in these analyses, though no records identified suspected cross reactivity as a reason UDT results did not lead to LTOT discontinuation. Sixth, patients were not consistently tested for alcohol, and the presence of a test likely indicates that providers had reason to believe that patients were at risk of alcohol use. Finally, patients with positive UDT results for more than one substance were excluded as we would be unable to determine which substance may have led to LTOT discontinuation. Future studies should examine discontinuation reasons in the context of polysubstance use.

CONCLUSION

The aim of this study was to compare the odds of clinician-initiated discontinuation of LTOT attributed to positive UDT results between three discrete categories of patients: those testing positive for: (1) cannabis only, (2) illicit substances (excluding cannabis) or alcohol only, and (3) non-VA prescribed prescription medications only (eg, benzodiazepines and opioids). High odds of UDT-related discontinuation were found in patients who tested positive for cannabis, alcohol, or illicit substances, relative to nonprescribed prescription medications. Future research should clarify and explain the reason for this substance-specific discontinuation gap, and develop approaches to improve opioid treatment monitoring via UDT that leads to appropriate clinical action to reduce patients’ risk of opioid-related adverse events.