Characteristics and Red Flag Correlates of Psychiatric Outpatients in a Mandated-Use Prescription Drug Monitoring Program State: A PBRN Card Study

Abstract

Objectives:

All 50 states have implemented a Prescription Drug Monitoring Program (PDMP) in efforts to control prescription drug abuse. Many now mandate PDMP checks before clinicians prescribe controlled substances. The aim of this study was to characterize the associations between patient characteristics, red flags found on PDMP reports, and prescriber behavior at community mental health agencies.

Methods:

Prescribers at 9 practice sites, in five regional community mental health centers, were recruited by a practice-based research network (PBRN) to participate in a Card Study. Prescribers completed a PDMP attitudes survey, and cards were completed for patients who had PDMP reports checked. Data were analyzed using descriptive and inferential statistics.

Results:

Thirty nine providers completed cards for n=249 unique patient encounters. Over 1/3 of all patients reported an addiction disorder (38%) or a diagnosis of chronic pain (34%). Twenty percent of PDMP reports were found to have red flags, most commonly multiple prescribers or multiple pharmacies. Red flags were associated with race (p<.0.05), presence of chronic pain (p<0.01), presence of an addiction diagnosis (p<0.05), use of opioids (p<0.001), and non-adherence with treatment (p<0.006). Among prescribers, red flags were associated with lower prescribing rates (p<0.01), and decisions to decrease dosage (p<0.002).

Conclusions:

Red flags were commonly found on PDMP reports done in community mental health settings, and were associated with important patient characteristics and diagnostic factors. PBRN research methods can be leveraged to obtain real-time observational data about psychiatric prescribers’ use of PDMP reports in clinical decision-making in different settings.

Introduction

Drug overdose is the leading cause of accidental death in the United States (U.S.) with prescription drug overdose continuing to grow at epidemic proportions^{1, 2}. Prescription Drug Monitoring Programs (PDMPs) have been recommended as a way to impact this trend by making law enforcement and healthcare providers aware of patterns of prescription drug abuse. A PDMP is a state administrated database that can provide a report upon request to healthcare and law enforcement personnel with information about an individual’s controlled prescription drug history. Each report conveys identifying information, and what prescriptions have been filled by a single patient during a certain period of time. These databases can only be accessed by registered users. Historically, PDMPs were designed as a law enforcement tool in which to identify drug misuse/abuse or extreme prescribing patterns³. PDMPs have also been categorized as a way to improve patient care, serving to aid the provider in more readily identifying patients’ needs for mental health and addiction service referral, confirming whether patients are being prescribed unsafe medication dosages, or whether they are at risk for deadly drug combinations⁴.

Statewide monitoring of prescriptions is pervasive in the US, and in Canada, with nearly every state and province having adopted a PDMP^{5, 6}. Despite the high prevalence of programs, state laws forming PDMPs vary widely, and programs vary considerably in their administration. Manasco et al. have reported that there is considerable variance across the US in: 1) access by licensure, 2) enrollment requirements, 3) prescriber mandate, 4) time-lapse to reported data, 5) data sharing between states, and 6) algorithm use to identify misuse trends⁵.

The Ohio Automated Rx Reporting System (OARRS) was signed into law in 2006, and data are available on reports from that day forward⁷. OARRS is available to all licensed healthcare providers, with enrollment mandated for physicians, APNs, Dentists, and PAs, as well as Pharmacists. The database tracks all scheduled meds, as well as other potentially abused medications including Muscle-Relaxers. Payment form, Prescriber name, Pharmacy name, and quantity are likewise tracked. The average time to current data is 24 hours. Data sharing is available with 17 states at present. Algorithms are used by OARRS administration to screen for “high risk” and “doctor shopping” behavior, and questionable prescribing behavior⁵. In 2014, prescribers were required to register and enroll in OARRS, and in April of 2015, Ohio law mandated that all prescribers of controlled substances check OARRS at initiation of medication and every 3 months subsequently for patients continuing controlled prescriptions⁷.

Although there has been significant research on PDMPs on a state and national level, there are gaps that remain in our understanding of these important tools. In particular, it is unclear whether providers accept and utilize PDMPs in mandated states, such as Ohio. Further, mandated checking are likely to have consequences on provider behavior, clinical decision-making, and downstream clinical outcomes. Currently, there is a dearth of research available that specifically examines how psychiatric prescribers make use of PDMPs to inform their practices with patients on controlled substances as well as those with comorbid diagnoses (mental illness and addiction; mental illness and pain). Finally, there are few studies comparing patient characteristics, red flag rates, and provider PDMP-related behavior across medical specialties such as psychiatry, primary care, and pain management.

Purpose

The purposes of the current study are as follows:

• (1)Assess the attitudes and practices of prescribing psychiatric practitioners (psychiatrists and advanced practice nurses(APNs)) in a mandated PDMP environment, the Ohio Automated Rx Reporting System (OARRS).
• (2)To examine the prevalence and associations of red flags with patient characteristics in urban community mental health center clinic settings.

Methods

Study Design

We performed a provider attitude survey followed by a cross-sectional observational card study to obtain real-time data about encounters with patients involving the checking of OARRS data⁸. In a card study, a small survey is printed onto a pocket-sized card which is carried by the clinician and quickly completed following patient encounters that meet specified criteria. The survey and card questions were developed by the Behavioral Research and Innovation Network (BRAIN) Psychiatry Practice-Based Research Network steering committee in conjunction with membership feedback. . The BRAIN PBRN and the card study were supported by the PBRN Shared Resource at Case Western Reserve University. The study was approved by the University Hospitals Institutional Review Board in Cleveland, OH.

Study Participants

Study participants were psychiatrists (attending and resident physicians), and advanced practice nurses (APNs), who are members of BRAIN, and clinicians at one of six regional mental health centers with a total of 13 sites distributed throughout Cuyahoga County, Ohio. Clinicians were asked to consent to the study prior to completing the provider attitude survey, and accepting a packet of cards to complete. Clinicians were oriented to the study methods at site meetings by a research assistant, and were given opportunities to address logistical concerns and ask questions.

Data Collection

Data was collected by prescribers in outpatient mental health practices in the community affiliated with University Hospitals of Cleveland. Each participating psychiatrist and ANP was asked to complete a data card immediately following visits by 8 consecutive patients for which they checked the online OARRS database. Participating clinicians were instructed to not record any identifying information about patients or themselves on the data collection cards, and to complete cards only for patients age 18 and older. In addition, the clinicians tracked the number of patients seen during the total period of data collection in order to provide a denominator for calculating the frequency of the phenomenon being studied. Each participating clinician returned completed data collection cards to the research staff using a pre-paid mailed envelope. Data collection took place over a 12-week period following initial study deployment starting in January 2016. Prescribers who provided care multiple outpatient practices completed all of their cards at the site in which their first data card was completed.

Statistical Analysis

The statistical analysis was carried out in three phases. Descriptive statistics were used to examine the distribution of key outcome and predictor variables. Means and standard deviations were computed for continuous variables and percentages for categorical variables. These data were used to describe the study participants, address the descriptive research questions, and inform the statistical analysis. Next, patient characteristics including demographic and clinical variables were tested in terms of their association with the presence of “red flags.” Chi-square statistics were computed for categorical variables and independent t-tests and analyses of variance when one of the variables was a continuous level of measurement. Lastly, the significant associations found in phase two were re-examined after adjustment for multiple hypothesis testing and the nesting of patients within providers.

Results

Provider Demographics

Provider demographics were gathered in the provider attitude survey prior to study enrollment. As shown in Table 1, the sample of providers who completed cards (n=33) was majority female, white/Caucasian and Asian, and the majority of providers reported from the largest agency. The average years of independent prescribing was 13 years. The average number of patients seen prior to accumulating 8 OARRS checks was 48, with a median of 37 patients. Provider attitudes toward OARRS were largely positive, with majority satisfaction, despite some attitude that checking OARRS was burdensome. Longer practicing prescribers were less satisfied overall (p=.005).

Table 1:

Demographics of Providers (n=33)

	n (%)	n (%)
Gender	by Provider	by number of Cards
Male	12 (36)	94 (38)
Female	21 (64)	155 (62)
Race/Ethnicity
Asian	9 (27)	68 (27)
Caucasian/White	19 (58)	151 (61)
Other	5 (15)	30 (12)
Type of Prescriber
Attending/Resident physician	21 (64)	169 (68)
APN	12 (36)	80 (32)
	Mean (SD)
No. years as independent prescriber	13 (13)

Patient Demographics

As shown in Table 2, patients (n=249) were majority white (63%) and female (66%) by a large margin. This was unexpected by known demographics of CMHCs in the greater Cleveland area, which are majority African American. The average age of patients for whom OARRS was checked was also greater than expected by mean age of CMHC patients. The majority of patients (70%) were considered adherent or extremely adherent by their respective providers based on subjective provider assessment, and had been seen for more than 1 year (55%).

Table 2:

Patient Characteristics (n=249)

	n (%)
Gender
Male	77 (31)
Female	169 (68)
Trans/Other	2 (1)
Age (years)	Mean (SD)
	46 (18)
Race/Ethnicity	n (%)
African American/Black	58 (24)
Caucasian/White	155 (63)
Hispanic/Latino	22 (9)
Other	10 (4)
Treatment duration: with provider ≥ 1 year	131 (55)
1st appt	25 (11)
<1 year	83 (34)
Prescriber Estimated Overall treatment adherence	Adherent or Extremely Adherent
	157 (70)

Patient Characteristics, and Diagnoses

Table 3 lists the frequencies of non-addiction and addiction diagnoses in the patient sample. Most notably, anxiety and mood disorders were extremely common, with combinations of anxiety and depression comprising 36% of the sample. Mental health diagnosis comorbidity (greater than or equal to 2 diagnoses) was present in 62% of the total sample. Addiction Diagnoses were also prevalent, the three most common being Alcohol (15%), Opioid (14%), and Cannabis (9%) Use Disorders. Thirty four percent had 2 or more Addiction Diagnoses. A large proportion of the sample were reported to have Chronic Pain (34%). The vast majority of patients for whom OARRS was checked were still prescribed controlled substances at the end of their appointment (93%), with only 7% leaving their appointment with no prescribed controlled medications. Psychiatrists/APNs prescribed Benzodiazepines and Stimulants, but a small minority of prescribers prescribed opioids. It is known that some of the study psychiatrists are also buprenorphine prescribers. It is notable that the majority of the cards returned were from 1 mental health center, but this center had 4 regionally distributed sites, and thus represented a diverse population.

Table 3:

Frequencies of Diagnoses, PDMP Checks, Prescriptions Written, and Red Flags

Non-addiction psychiatric dx	n (%)
Anxiety	125 (50)
Depression	109 (44)
Bipolar Disorder	65 (26)
Schizophrenia	47 (19)
ADHD	46 (19)
Personality Disorder	38 (15)
Other	39 (16)
Addiction-related dx	n (%)
Any	95 (38)
Alcohol	38(15)
Opioid	35 (14)
Cannabis	22 (9)
Benzodiazepine	19 (8)
Stimulant/Cocaine	20 (8)
Other	7 (3)
Chronic pain disorder	n (%)
	85 (34)
Community Mental Health Center	n(%)
CMHC A	139 (56)
CMHC B	46 (19)
CMHC C	16 (6)
CMHC D	8 (3)
CMHC E	40 (16)
Mandated Checks	124 (50)
Controlled Prescriptions Prescribed	n(%)	Prescribed by Psych %
Opiate	68 (27)	11
Stimulant	53 (21)	92
Benzodiazepine	153 (61)	84
Other	30 (12)	60
Red Flag Frequencies	n (%)	Percentage of Total Red Flags
Total Red Flag Found	50 (20)
2+ Prescribers	28 (11)	56
3+ Pharmacies	17 (7)	34
Early Refills	4 (2)	8
Previously Unknown Rx	12 (5)	24
New Prescribers	8 (3)	16
OARRS conflicts	13 (5)	26
Other	5 (2)	10

Red Flag Frequencies and Correlations

Red Flags were found on OARRS reports in 20% (n=50) of unique patient encounters. Table 4 illustrates the associations between red flags and patient encounter characteristics. The most common red flags were 2 or more Prescribers (56%), and 3 or more Pharmacies (34%). Red flags were discussed by providers with their patients in 71% of visits. Red flags were associated with a number of specific patient characteristics. Patient characteristics including chronic pain, Non-African American ethnicity, addiction diagnosis (especially opioid), or who were judged to be non-adherent with treatment by their providers were significantly more likely to have red flags upon OARRS check. Gender, and Age were not correlated with Red Flags. Individual providers and facilities had statistically significant differential rates of red flags. Red Flags were associated with both pre-OARRS and post-OARRS decision-making on the part of providers. For instance, discretionary checks were more likely in patients who were later found to have red flags, but were not statistically associated with age, race, provider type, or patient sex. It is also notable that once adjustments were made for number of secondary analyses and nesting of data by provider, several associations were no longer statistically significant including discretionary and addiction diagnoses.

Table 4:

Percentage of Red Flags by Patient Characteristics

By Race	Total N	n with flag (%)	Chi Sq p-value
White/Caucasian/Hispanic/Other	186	45 (24)
African American	58	5 (9)	0.013+
By Patient Diagnosis
Chronic Pain Disorder: Yes	85	24 (28)
No	155	25 (16)	0.021
Addiction Diagnosis: Yes	95	25 (26)
No	154	25 (16)	0.054++
By Provider Judged Adherence
Adherent: Yes	157	22 (14)
No	68	23 (34)	0.001
By Controlled Prescription
Opioid: Yes	68	31 (46)
No	181	20 (11)	0.001
Benzodiazepine: Yes	153	32 (21)
No	96	18 (19)	0.992
Stimulant: Yes	53	8 (15)
No	196	41 (21)	0.257
Other: Yes	30	12 (41)
No	202	36 (18)	0.004
By Discretionary Check	122	32 (26)
Mandatory Check	124	19 (15)	0.026++

⁺Race associations were significant after adjustments as a dichotomous variable, but not as a 3 category variable (White, AA, Other)

⁺⁺Mandatory Checks and Addiction Diagnoses were no longer associated with Red Flags at p<0.05 after adjustment for testing and nesting.

Discussion

This observational card study is the first of its kind to document the real-time clinical utilization of a PDMP by psychiatric clinicians in a state where legal mandate requires clinicians to check the PDMP for any patient receiving controlled-substances⁹. The study design had several main strengths. First, we collected baseline data about routine practice with respect to psychiatric prescribers and their patients. This demonstrated not only the characteristics of patients for which OARRS checks are required, but also those patients for whom providers checked OARRS discretionarily. We must assume that this discretionary practice is based on clinical judgment in conjunction with anticipated utility of the OARRS database for clinical decision-making. Via the collection of data for 249 unique patients, we were able to determine a preliminary baseline red flag prevalence rate for the population seen regionally in Cleveland CMHCs. These data were associated with patient and provider characteristics, and these associations may be able to advise future studies in predictive analysis of red flag prevalence. Finally, we obtained information about clinical decision-making based on information from OARRS reports, the data from which will be reported in a separate publication.

Providers checked OARRS discretionarily 50% of the time, a practice which was associated with Red Flags. Required checking by legal mandate may increase may indirectly increase the frequency of discretionary checks due to perceived utility or induction of habitual checking. Surveys in non-mandated environments have reported much lower rates of discretionary checking^10–13. Red Flag rates varied depending on context, provider, and patient characteristics. It is notable that red flag rates depend on definition of red flags. We utilized the same definitions as Sowa et al¹⁴. Our Red Flag definitions also correlated with those defined by OARRS recommendations⁷. Sowa and colleagues (2015) describe presence of red flags as a sign of prescription misuse, but this is controversial in our opinion. For example, patients may have 2 or more providers if they require opioids for pain, and benzodiazepines for anxiety, which may be clinically appropriate. Future studies must clarify this, and realize that rates of red flags may vary, and correlate variably with substance misuse. It is further unclear whether red flags predict morbidity, mortality, substance diversion, or other poor outcomes.

The analysis of patient characteristics revealed some additional interesting findings. Surprisingly, the majority of patients for whom OARRS was checked were female, and white. This defies the expected demographics of the regional CMHC population and suggests that providers may preferentially prescribe controlled substances for this demographic. Indeed, large scale studies of prescription patterns suggests this pattern¹⁵. This raises further questions about clinical decision-making bias amongst providers. Provider bias may affect the relevance of PDMP data to clinical decision-making, and bears further study.

Anxiety is clearly the most commonly managed symptom amongst psychiatrists, which increases the likelihood of benzodiazepine prescription. Chronic pain and corresponding opioid prescription, as well as opioid use disorders, further complicate the decision-making around this practice. The nexus of anxiety, depression, and chronic pain is clearly documented¹⁶. Thus it is not surprising that these diagnoses and treatment approaches are present in a large proportion of the sample. It is notable that even in the presence of substance use disorders, controlled substance prescription is also prevalent.

We were able to document that Red Flags found in OARRS reports correlate with reduced prescribing, which suggests that prescription of controlled substances may be moderated by the information found in statewide prescribing databases. If providers are fully aware of these potential conflicts, they may prescribe less or cease to prescribe more dangerous combinations of medications. PDMPs have been successful in changing statewide patterns of controlled substance prescription⁷. Several states have reported significant change in key prescription drug metrics over time following PDMP implementation¹⁷. These include increased number of database checks annually, decreased opioid and benzodiazepine doses dispensed/year, and decreased “doctor shopping” as defined and detected by scanning algorithm⁷. Several studies have correlated PDMP implementation with decreased population rates of opioid abuse, opioid-related inpatient admissions, and overdose deaths^{1, 17–21}. Some study findings are mixed, however^{3, 22}. A few studies have examined whether PDMPs influence prescription-writing practices of providers. Baehren and associates (2010) found that when emergency department physicians reviewed PMP report data after seeing patients (but prior to discharging them), they found that prescribers actually altered their original opioid prescription-writing decision 41% of the time²³. Another study of emergency department providers found that emergency providers changed plans to prescribe opioids at discharge in 9.5% of cases, with 6.5% of patients receiving opioids not previously planned and 3.0% no longer receiving opioids, after consulting patients’ PMP report data²⁴. Sowa and associates (2012) examined how psychiatry residents that were helping patients manage mental health diagnoses utilized PDMPs to inform their practices prior to writing controlled prescriptions. Results indicated that these providers changed their prescription-writing of controlled substances approximately 2.2% of the time¹⁴. More study is needed in mandated PDMP check environments to observe changes in provider behavior.

The OARRS Card study had some limitations including sample size, and lack of control group. Indeed, some of the trends discovered were found to no longer be statistically significant when adjustments were made for cluster analysis and test number. A larger sample would clarify these trends. Another possible source of bias in the study was that the majority of prescribers worked at agency. However, the agency provided data from 5 geographically diverse sites, so although agency culture might be a source of bias, the sample of patients was likely to reflect a broad population. Lastly, the study was not conducted with a control group (for which PDMP was not checked). A control group would have allowed for a true comparison of the effect of PDMPs on prescribing behavior. While the strengths of the study stemmed from a naturalistic design, more rigorous methods would be necessary to generalize findings to psychiatric prescribers in other settings.

Our data adds to the body of knowledge which will inform best practices with respect to PDMP monitoring. State policy makers, clinicians, and researchers must seek to further understand the impact of these important tools on providers, patients, and the system-at-large. Mandated state regulatory environments may not be similar to non-mandated ones, and may increase discretionary checks, or induce more systematized behavior on the part of clinicians. States must balance the ease of access to this information with the impact on privacy, and provide education on the appropriate use and interpretation of PDMP information. More research must be done on the impact of PDMPs on prescriber rates, dosing, and overall impact on health conditions and mortality. Future studies might focus on different sets of providers, and diverse patient populations. PBRN methods may allow clinicians to access outpatient populations that differ significantly from inpatient populations or academic research centers. Finally, prescriber bias may be a significant issue affecting the desired impact of PDMPs and patient outcomes globally.