Trends in drug use among young adult females: a 22-year retrospective analysis

Abstract Louisiana State University Health Sciences Center at Shreveport (LSUHSC-S) serves a largely minority-based, urban population. This study aims to identify trends in urine drug screen (UDS) results among females aged 18–35 visiting State University during 1998–2011 and 2012–2019. Using two databases extracted from the electronic medical record system, we performed statistical analysis of demographics and UDS results. Young females aged 18–35 mostly tested positive for cannabinoids and opiates during both periods, 1998–2011 and 2012–2019. African-American females had a higher percentage of positive UDS for illicit drugs, such as cannabinoids and cocaine, while Caucasian females had a higher rate for prescription drugs such as opiates, benzodiazepines, and amphetamine. Between 1998–2011 and 2012–2019, trends in drug screen results changed in both populations, with Caucasian females showing a drastic increase in amphetamines and African-American females showing increase in opiates and cannabinoids during 2012–2019. GatewayNet analysis (sequential-rule mining for inducing causation) for 2012–2019 indicated that a positive screen for amphetamines is likely preceded by a positive screen for cannabinoids, and benzodiazepines may be preceded by opiates. Our results emphasize the importance of drug use monitoring among young females of childbearing age. GatewayNet analysis implies a sequential nature to drug positivity on urine drug screening in this population.


Introduction
Women experience unique issues when dealing with substances, which stem from their biology, socioeconomic and sociocultural issues [1]. Stress, negative effects of relationships, and mental/emotional challenges may serve as initiators of drug abuse [2]. Recent studies indicate that women are more vulnerable to addiction [3], develop an addiction to drugs more quickly, and by using lower doses, when compared to men [1]. Moreover, after drug use is discontinued, women experience more severe signs of withdrawal, and are more likely to relapse [1,4]. According to the 2016 National Survey on Drug Use and Health, 19.5 million females ages 18 years and older (or more than 15% of all females) have used illicit drugs in the past year [5]. In addition, 8.4 million females ages 18 years and older reported misusing prescription drugs, such as opioids and benzodiazepines, in the past year [5]. According to the Centers for Disease Control and Prevention (CDC), an increase in prescription opioid use among women led to a 5-fold increase in overdose deaths between 1999 and 2010 [6]. Patients with Medicaid were more likely to be prescribed opioids, with higher doses and longer prescription durations, when compared to non-Medicaid patients [6]. Due to a limited number of clinical studies of trends in substance use among females, federal agencies emphasize the importance of inclusion of the female population in such clinical research [1].
In this study we identified the trends of urine drug screens (UDSs) for childbearing-aged women that visited University Health Sciences Center during two periods: 1998-2011 and 2012-2019. Louisiana State University Health Sciences Center, the only Level 1 Trauma Center in North Louisiana, serves a largely minority-based, urban population. In our previous study of the general population that visited the University Health Sciences Center Emergency Department during 1998-2011, we found that African-American and Caucasian patients had higher positive cannabinoid and opiate UDS rates [7]. In addition, Caucasians had higher positive benzodiazepine rates, while African-Americans had higher positive cocaine rates [7]. In this study, we analyzed UDS results of [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35] year-old African-American and Caucasian women that visited University Health Sciences Center from August 14, 1998 to December 11, 2019. This analysis includes UDS results for cannabinoids, opioids, benzodiazepines, amphetamine, and cocaine. We investigated trends in drug abuse among young females, relationships between reasons for the visit and use of a certain drug, and a gateway tendency of the drugs in this population.

Data sources and analysis
We extracted the UDS results for females aged 18-35 admitted to the Emergency Department (ED) or to the labor unit of University Health Sciences Center during 1998-2011 and 2012-2019 from two medical record system. The dataset covering 1998-2011 was manually extracted from medical records prior to implementation of an electronic health record (EHR) system and was more limited in scope. The implementation of the EHR, EPIC, in 2012 allowed for an expanded dataset. 21,665 female records from 1998 to 2011, and 22,350 records from 2012 to 2019, were included. These records contained UDS results, socio-demographic information (age, gender, race, payment/insurance type), and visit timestamps. The 2012-2019 data from the EPIC electronic medical record system also included the reason for the visit, drugs administered during the visit, and drug administration timestamps. We performed statistical analysis using R software (https://r-project.org) to ascertain the relationship between self-identified race, payment/ insurance type, visit type (in 2012-2019 records), and UDS results for African-Americans and Caucasians. We performed a pairwise Chi-square test to calculate statistical significance between visit type and UDS results.
We analyzed amphetamine, benzodiazepine, cannabinoid, cocaine, and opiate UDSs; other drugs either had too few records or extremely low positive rates. The UDSs screened for the presence of amphetamines, benzodiazepines, cannabinoids, cocaine, and opiates using the immunoassay OPI Flex reagent cartridges (Siemens Healthcare Diagnostics, Inc., Newark, DE) in the Siemens Dimension Vista® System (Siemens Healthineers Medical Solutions USA, Inc., Newark, DE). The opiate immunoassays are calibrated to morphine and cross-reacts with other opioids, such as codeine, hydrocodone, hydromorphone, oxycodone, and oxymorphone. The benzodiazepine tests are calibrated to lormetazepam. Visit types were categorized into one of thirteen reasons: pain, pregnancy, psychiatric/neurologic, sickle cell, drug misuse, medication refill, cardiac/pulmonary, gastroenterology, gynecologic, inflammation/allergy, oncology, integument, and other. The reason for visit was provided by the patient and subsequent problems/diagnoses were discovered during the visit which led to the UDS orders. For example, cardiac disease and chest pain could be due to drugs of abuse such as cocaine. Thus, both datasets contained UDS results only for female patients for whom a drug screen was ordered by their physician and not for all patients who visited the ED during the study period. We recorded administration times for prescribed medications in the 2012-2019 data. Administration data was not available for 1998-2011. To calculate a timeframe between drug administration and UDS test, we used the time in minutes between two dates in the EHR: the date and time the medication was scheduled to be given (administered) and the date and time that the UDS tests occurred. The maximum time between administration of the drug and UDS tests was 4,319 min, or a period of about 3 days. We attributed positive opiate screens to opioid analgesics given between 15 min and 72 h before the UDS test. We attributed positive benzodiazepine screens to doses given between 15 min and 240 h before urine collection.

GatewayNet
We used GatewayNet analysis to discover causative relationships between two drugs, such that a positive UDS for one drug indicates a past positive UDS for another drug in the same patient tested during a certain time period. GatewayNet is a form of sequential rule mining to establish causation relationships between a set of histories [8]. It uses a probabilistic definition of causation: given events x and y, y is caused by x if the probability of x being preceded by y is greater 0.5 [8]. The ratio of the probability that "x precedes y" and the probability that "x does not precede y" is referred to as certainty (C). Certainty has a value between 0 (no certainty) and infinity (absolute certainty), with 1.0 indicating equal probability (50/50 chances). Probabilistic causation is only satisfied when C > 1, and greater values indicate stronger causality. We also use confidence to remove unlikely gateway rules. Confidence is defined as the proportion of histories containing "y strictly preceded by x" to all x events. We only report gateway results with certainty greater than 1.0 and confidence at least 0.25. The GatewayNet software may be obtained from our prior work [8]. Originally, we developed GatewayNet to mine initiation rules, which are of the form x → y and are mined when a sufficiently large number of histories (at least 30 or more) involving y are strictly preceded by x. Initiation rules with C > 1 are called gateway rules and satisfy probabilistic causation. GatewayNet is also capable of windowed analysis, which only considers initiations within a given timeframe. We performed GatewayNet analysis on the 2012-2019 UDS data to identify relationships such that a positive screen for one drug preceded another in the same patient tested within a one-year window. Inclusion required that a patient have at least one positive results in two or more visits during a year. We excluded positive UDS results caused by administration of the drug in question.

Demographic characteristics
From 1998 to 2011, the majority of UDS records were for African-American and Caucasian females (Table 1). This was also true from 2012 to 2019, during which the number of UDS screens for African-Americans increased, and Caucasians decreased. During 1998-2011, average annual UDS rate was 1,627 ± 395 for females between aged 18-35 who visited LSUHSC. This number grows to 2,698 ± 211 annual UDS for 2012-2019, with a sharp increase in number of UDS for African-American females. The average number of UDS per person during these two periods also increased from 1.6 in 1998-2011 to 2 in 2012-2019 for African-Americans and from 1.5 to 1.7 for Caucasians females, respectively. The most common Admission Type was "Emergency" (45%), followed by "Urgent" (28%) and "Elective" (23%). Between 2012 and 2019, the most common visit type was "Psychiatric/ Neurologic" (25.6%), followed by "Pain" (23%) and "Pregnancy" (17%). Other visit types, representing 1-4% of the visits each, include "Cardiac/Pulmonary", "Sickle Cell", "Gynecologic", "Gastroenterology", and "Drug Misuse". African-Americans made up the majority of records for each category except Drug Misuse. Every Sickle Cell visit involved an African-American patient. Between 1998 and 2011, the most common payment/insurance type was "Self-Pay" (71%) followed by "Medicaid" (16%). Between 2012 and 2019, 60% of African-Americans and 53% of Caucasians used "Medicaid" to pay for their visits. Increase in the rate of Medicaid used by both African-American and Caucasian females may be explained by the fact that Medicaid program was expanded in Louisiana in 2016. Caucasians used Blue Cross, Commercial insurances and self-paid options at a higher rate, while African-American females used outpatient pharmacy in addition to Medicaid more often than Caucasian females.

General
During 1998-2011 and 2012-2019, cannabinoids, followed by opiates, were the most prevalent positive drugs in the UDS for 18-35 year-old females. Among Caucasians, there was an increase in the positive screen rate for cannabinoids (24.8% to 27%), opiates (17% to 20%), and amphetamine (6.8% to 16%), and a decrease in benzodiazepines (21% to 13.7%) and cocaine (8.4-3.6%) across periods (Figure 1a and b). Among African-Americans, there was an increase in the positive screen rate for cannabinoids (24% to 33%), opiates (8.9% to 16.9%), and amphetamine (0.9% to 3%) and a decrease in cocaine (8.4% to 5.6%) across periods (Figure 1a and b). The positive benzodiazepine rate among African-Americans remained roughly the same between both periods (3.4%). During 1998-2011 and 2012-2019, Caucasians had higher percentage of positive UDS for opiates, benzodiazepines, and amphetamines than African-Americans (Figure 1a and b). The rates for cannabinoids and cocaine were roughly equal between both groups in 1998-2011, but in 2012-2019, they were higher in African-Americans.   between visit types and drug screen results indicate several relationships. We observed that opiate UDS was positively associated with medication refill, pain, and sickle cell, and negatively associated with cardiac/ pulmonary, gynecologic, pregnancy, and psychiatric/ neurologic related reasons. Benzodiazepines UDSs were positively associated with drug misuse and psychiatric/neurologic reasons, but negatively associated with cardiac/pulmonary and pregnancy-related reasons. Amphetamine UDS was positively associated with drug misuse and psychiatric/neurologic reasons, but negatively associated with cardiac/pulmonary, pain, pregnancy-related, and sickle cell. Cannabinoid UDS had a positive relationship with gastroenterology and psychiatric/neurologic reasons, but negatively associated with cardiac/pulmonary patients. Cocaine UDS was positively associated with drug misuse and psychiatric/neurologic reasons and negatively associated with pregnancy and sickle cell. Analysis of correlation between visit types and UDS results during 2012-2019 demonstrated that "Drug misuse" visits positively associated with UDS for amphetamine, benzodiazepines, and cocaine (Table 2). Cardiac/pulmonary visits were positively correlated with amphetamines and opiates UDS and negatively correlated with use of benzodiazepines and cannabinoids. For gastroenterology records, the only relationship is for positive cannabinoid UDS. Gynecologic and medical refill-related visits were positively associated with opiates UDS. For pain patients, the positive relationship exists for both amphetamines and opiates. Pregnancy-related visits were negatively associated with amphetamines, benzodiazepines, cocaine, and opiates UDS. For psychiatric/neurologic patients, a positive relationship exists for amphetamines, benzodiazepines, cannabinoids, cocaine, but negative relationship for opiates. Sickle cell patients visits were positively associated with opiates but negatively -with amphetamines and cocaine UDS ( Table 2).

Opiates
From 1998 until 2016, Caucasian females had higher rates of positive opiate UDS results than African-American females, but for 2017 to 2019 African-Americans had higher rates of positive opiate results (Figure 2a). This may be due to a close to 2-fold increase in the opiate administration rate during the ED visits in the African-American female population starting from 2017 (Figure 2b). The highest opiate positivity was among sickle cell patients (72%), followed by oncology (50%), and medication refill patients (42%) ( Table 2). By removing administered positive results between 15 min and 75 h before the UDS we observed a drop in the overall opiate positivity from 18% to 9%. Instances of administration of opiates during a visit were related to pain, then pregnancy and sickle cell. Opiate-positive patients were also positive for cannabinoids in 32% cases, benzodiazepines (15%), and amphetamines (8%) ( Table 3). Interestingly, opiate-positive females had higher positivity for benzodiazepines than the overall rates (7.1%, Table 2, row "Overall" and column "Benzo").

Benzodiazepine
During every year of 1998-2019, Caucasian females had higher positive rates for benzodiazepines than African-American females (Figure 2c). The benzodiazepines prescription rate was also higher in Caucasians during every year with available prescription data (2012-2019), increasing in 2017 to about 12% before declining to about 5% in 2019 (Figure 2d). The highest benzodiazepine positivity was among drug misuse patients (13%), followed by psychiatric/neurologic patients (10%) ( Table 2). The most common reason for visiting in benzodiazepine-positive patients was psychiatric/neurologic (41%), followed by pain-related (27%), and pregnancy-related (7.4%). Patients who were administered benzodiazepines during their visit were mostly admitted for psychiatric/neurologic (73%) or pain-related (10%) reasons. We analyzed benzodiazepine positivity by omitting administered benzodiazepines between 15 min and 10 days before the positive benzodiazepine UDS screen as negative, resulting in a drop of the overall benzodiazepine positivity from 7.1% to 6.2%. Patients positive for benzodiazepines were also positive for cannabinoids in 43% of cases, opiates (38%), and amphetamines (23%) ( Table  3), all notably higher than the overall rates ( Table 2).
Recent studies indicated that an increasing number of people are using a combination of opiates and benzodiazepines, leading to an increase in the overdose death rate [9]. Thus, we investigated co-use of opiates and benzodiazepines during 2012-2019. From 2,505 opiate-positive females, 547 (21.8%) also testing positive for benzodiazepines, of which only 141 (25.8%) had records of administered benzodiazepines during their ED visit.
Cannabinoid-positive patients also tested positive for opiates in 19% of cases, amphetamines (13%), benzodiazepines (10%), and cocaine (8%) during the same visit ( Table 3). The rates for amphetamines, benzodiazepines, and cocaine are all higher than the overall rates (Table 2) among the cannabinoid UDS positive group.
Amphetamine-positive patients also tested positive for cannabinoids in 50% of cases, benzodiazepines (21%), and opiates (18%) during the same visit ( Table 3). The cannabinoid and benzodiazepine rates were both higher than the overall rates (Table 2)  Cocaine-positive patients also tested positive for cannabinoids in 50% of cases, amphetamines (18%), and opiates (18%) during the same visit (Table 3). Cannabinoid and amphetamine positivity were both higher than the overall rates (Table 2) among the cocaine UDS positive group.

GatewayNet analysis
To determine whether any drug serves as a stepping-stone to other drugs, we performed GatewayNet analysis using the 2012-2019 dataset and a one-year window. Overall, we analyzed 1,719 patients (6,017 records) who had a positive UDS result at two or more visits during one year, excluding results that corresponded to administered opiates or benzodiazepines. GatewayNet indicated that cannabinoids positive UDS results may precede amphetamine UDS positivity (certainty = 3.1, confidence = 0.3), suggesting that cannabinoids may lead to abuse of amphetamines by young females. Greater than 75% of patients who screened positive for amphetamines exhibited this behavior. Additionally, we observed that benzodiazepine UDS positivity has a marginally greater chance of being preceded by a positive opiate UDS (certainty = 1.1, confidence = 0.4). This result suggests that young females who tested positive for benzodiazepines also tested positive for opiates at elevated rates during a previous visit.

Discussion
In this study, we analyzed UDSs of young women who visited Louisiana State University Health Sciences Center during two periods, 1998-2011 and 2012-2019. Most UDSs of 18-35 year-old females were for African-Americans and Caucasians patients during 1998-2011. In general, our dataset matches the population distribution of Shreveport, which is 57.1% African-American, 31% White, and 1.7% Asian according to the 2019 Census [10]. More than 50% of patients used Medicaid insurance during 2012-2019, which is more than double that of the general Louisiana population (20.8%) [11]. Most of the visits for both African-Americans and Caucasians during 2012-2019 were due to psychiatric/neurologic reasons, pain or pregnancy. African-American females had higher number of visits in all categories except for drug misuse, in which Caucasians demonstrated the higher number of visits during 2012-2019. Like the general population [7], young females mostly tested positive for cannabinoids and opiates during both 1998-2011 and 2012-2019. Overall, African-Americans had a higher percentage of positive UDS results for illicit drugs (cannabinoids and cocaine) compare to Caucasian females. Caucasians used prescription drugs (opiates, benzodiazepines, and amphetamines) at a higher rate compared to African-American female patients. These trends suggest that there are different drug use trends among African-American and Caucasian populations probably based on availability of drugs for African-American vs Caucasian populations.
From 1998 to 2016, the positivity for opiates among Caucasians was higher than African-Americans, peaking at 25% in 2011-2013, and correlating with the nationwide peak among Caucasians in 2010 [12]. It was suggested that higher opioid use among non-Hispanic/White females could be explained by higher opiate prescription rates [13]. In our study, we observed similar administration rates for opiates among both African-Americans and Caucasians during 2012-2016. Among all opiate-positive females during 2012-2019, half did not have a record for opiate administration during the visit. Moreover, we found that starting from 2017, African-Americans tested positive for opiates at higher rates than Caucasians, which could be a result of the drastic increase in opiate administration rates to African-Americans also starting from 2017.
The positive benzodiazepine UDS rate among Caucasians was three-to five-fold higher than in African-Americans during 1998-2019, similar to a trend in general population [7]. A recent study demonstrated that patients who identify as Non-Hispanic White are prescribed benzodiazepines more often [14,15]. Benzodiazepines are usually prescribed to treat generalized anxiety disorders, insomnia, seizures, and panic disorder, due to the ability of these drugs to reduce anxiety and stress levels. Women are more likely to misuse benzodiazepines to cope with stress [16]. In our current study, we observed that females with the most benzodiazepine-positive UDSs visited University Health Sciences Center for psychiatric/neurologic and pain-related reasons. According to a recent epidemiologic analysis, benzodiazepines and other tranquilizers were the third-most misused drugs, following cannabinoids and opioids in 2017 [16]. The National Survey on Drug Use and Health study indicated that females aged 18-49 had higher odds of misusing benzodiazepines than their male counterparts, while no correlation exists between gender and benzodiazepine use among other age groups [17]. Benzodiazepine misuse was associated with both access to prescribed drugs and misuse of other substances [16]. An increase in misuse was also associated with higher levels of polysubstance use [18,19]. Our GatewayNet analysis revealed that young females are more likely to use benzodiazepines with opiates rather than alone, which is in agreement with reports indicating that people with opioid-use disorders tend to misuse benzodiazepines at higher rates than the general population [16,[20][21][22]. Benzodiazepine and opiate abuse may lead to the development of tolerance, dependence, and withdrawal syndrome [23]. Benzodiazepine misuse includes an increased risk of opioid overdose, impaired cognitive function, declined quality of life, greater pain severity, and mortality due to opioid overdose or suicidal ideation [9,12,16,24,25].
Similar to the general population [7], Caucasians females had much higher amphetamine positive UDSs compared to African-Americans each year during 1998-2019 period. This may be explained by easier access to the drug by Caucasian population. Amphetamines and their derivatives are controlled substances, prescribed to treat attention deficit/hyperactivity disorder, narcolepsy, and depression. In our study, we observed that amphetamine positivity correlated with psychiatric/neurologic visits. Amphetamines are often abused because they energize and help the user stay alert. However, dangerous side effects including elevated blood pressure and drug-induced psychosis are associated with use. Amphetamines could lead to psychological and physical dependence and development of tolerance leading to overdose. Women may experience unique side effects after amphetamine use, such as mood disturbances, panic attacks, paranoia, hallucinations, and hormonal disbalance [26].
Cocaine UDS positivity steadily decreased in both the African-American and Caucasian female populations from 1998 to 2019, mirroring the general population [7]. Interestingly, while the positive cocaine UDS rate in African-Americans decreased almost 5-fold during 1998-2019, the cannabinoid and opiate UDS positivity increased by 1.8-and 6-fold, respectively, suggesting that these two drugs partially replaced cocaine. Similar to the general population [7], cannabinoid UDS positivity was highest among all drugs in the female population, reaching 36% in African-Americans and 31% in Caucasians, with a higher percentage of African-American each year starting from 2009. Considering that decriminalization of marijuana in Louisiana occurs on June 15, 2021 [27] our data indicate that during 1998-2019 all cannabinoid positive UDS were due to illegal use of the drug.
We observed that cannabinoid UDS positivity may predict future amphetamine positivity in young females. These results differ significantly from our previous study of the general population, in which we observed that cannabinoid UDS positivity may be preceded by positive benzodiazepines, cocaine, and opiates UDSs [8].

Strength and limitations
A key strength of this study is the combination of analyses of UDS with drug administration records and reasons for the hospital visit, allowing for the identification of trends in drug abuse among specific groups and the prediction of illicit drug use. Another strength of this study is the use of GatewayNet to investigate initiation relationships between drugs, which revealed that cannabinoid use by young females may initiate the use of amphetamines and opiate-positive females tended to use benzodiazepines at elevated rates. There are several limitations of this study. One of the limitations is that it includes information only for patients who underwent a drug screening ordered by the clinician. The changes in the UDS ordering patterns may affect the number of tests in different categories of patients and, thus, the rate of UDS positivity for a particular drug. The other limitation is that these UDS results were not confirmed by other methods of drug detection such as GC/MS or another chemical assay. According to several reports, UDSs may yield false-positive and false-negative results and should always be interpreted as presumptive [28][29][30][31]. Another limitation of our study is that in most cases it was impossible to determine whether positive UDS results for opiates, benzodiazepine or amphetamine were due to prescribed or illegal use of drug. Also, it is possible that the use of, or prescription for, a particular drug was not captured in our patients' records. GatewayNet relied on the use of retrospective data and infers causation through sequence, and cannot by itself account for false positive UDS results or for gaps in the patients' records. We partially accounted for this by omitting UDS results that occurred as a result of administration of a substance that subsequently caused the UDS to be positive, however, the aforementioned caveats still apply.

Conclusions
Young females who visited the Emergency Department at Louisiana State University Health Sciences Center at Shreveport during 1998-2019 mostly tested positive for cannabinoids and opiates. While a decrease in the positive opiate UDS rate and opiate prescription rates among Caucasian females during the past few years provides hope for recovery from the opioid epidemic, the increased opiate administration rate during the ED visit and increased opiate UDS positivity among young African-American females is concerning. Another concern is that females with positive benzodiazepine UDS were also positive for opiates, cannabinoids, and amphetamines at a higher than overall rate. Finally, a rapid increase in amphetamine positive UDSs among young females during the last 8 years, especially in the Caucasian population, provides reason for great concern considering the addictive nature of amphetamines. Our study emphasizes the importance of monitoring for drug use among young females of childbearing ages, especially cannabinoids and opiates that may lead to abuse of amphetamines and benzodiazepines, respectively, according to our GatewaNet analysis.