Abstract

Objectives. The objectives of this study were to test for differences in prescription monitoring program (PMP) use between two states, Connecticut (CT) and Rhode Island (RI), with a different PMP accessibility; to explore use of PMP reports in clinical practice; and to examine associations between PMP use and clinician's responses to suspected diversion or “doctor shopping” (i.e., multiple prescriptions from multiple providers).

Design, Setting, Subjects. From March to August 2011, anonymous surveys were emailed to providers licensed to prescribe Schedule II medications in CT (N = 16,924) and RI (N = 5,567).

Outcome Measures. PMP use, use of patient reports in clinical practice, responses to suspected doctor shopping, or diversion.

Results. Responses from 1,385 prescribers were received: 998 in CT and 375 in RI. PMP use was greater in CT, where an electronic PMP is available (43.9% vs 16.3%, χ2 = 85.2, P < 0.0001). PMP patient reports were used to screen for drug abuse (36.2% CT vs 10.0% RI, χ2 = 60.9, P < 0.0001) and detect doctor shopping (43.9% CT vs 18.5% RI, χ2 = 68.3, P < 0.0001). Adjusting for potential confounders, responses by PMP users to suspicious medication use behavior were more likely to entail clinical response (i.e., refer to another provider odds ratio, OR, 1.75 [95% confidence interval, CI, 1.10, 2.80]; screen for drug abuse OR 1.93 [1.39, 2.68]; revisit pain/treatment agreement OR 1.97 [1.45, 2.67]; conduct urine screen OR 1.82 [1.29, 2.57]; refer to substance abuse treatment OR 1.30 [0.96, 1.75]) rather than legal intervention (OR 0.45 [0.21, 0.94]) or inaction (OR 0.09 [0.01, 0.70]).

Conclusions. Prescribers' use of an electronic PMP may influence medical practice, especially opioid abuse detection, and is associated with clinical responses to suspected doctor shopping or diversion.

Introduction

Increases in fatal overdose and opioid-related emergency department visits and hospitalizations since the mid-1990s have been driven by a substantial growth in opioid analgesic prescriptions and nonmedical use of prescription opioids [1–9], among other variables [10]. In Rhode Island (RI) and Connecticut (CT), overdose has surpassed motor vehicle crashes to become the leading cause of unintentional injury death [11]. National survey data indicate that RI has the highest per capita illicit drug use and ranks third for nonmedical use of prescription opioids among persons aged 12 or older, behind Oklahoma and Oregon [12].

Prescription monitoring programs (PMPs) are an emerging tool with potential to influence risks to patients associated with abusable medications, especially prescription opioids. PMPs offer more detailed information than patients themselves or single institution medical records often provide. For example, they permit verification of a patient's self-reported prescription history of abusable medications. Additionally, PMPs aid in the determination of filling multiple prescriptions of the same drug from multiple providers (i.e., questionable medication behavior or “doctor shopping”), and cataloguing of medications that may suggest contraindications or increased risk of adverse events.

From their inception, PMPs were conceptualized as tools of surveillance, set within a criminal justice model, and were funded largely through the Bureau of Justice Assistance at the U.S. Department of Justice. The few available studies evaluating PMPs have been conducted at the state level and suggest reductions in the circulating supply of prescribed medications [13] and slower growth in prescription medication sales when PMPs are used. Expanded PMP use by prescribers was coincident with a decrease in the county-wide opioid-related death rate in one study [14]; yet, another study [15] found that PMPs did not consistently reduce prescribed controlled substances abuse rates.

Little data exist on the effect of PMP use on clinical practice patterns. PMPs are available but are underutilized by prescribers. Most state PMPs report that less than 25% of health care professionals use PMPs to obtain patient reports [16]; only two states (Nevada and Delaware) require checking the PMP before dispensing medication. One small study suggested that PMP use in the emergency department could result in physicians prescribing some patients fewer opioids while prescribing other patients more opioids [17].

The aims of this study were 1) to test for differences in PMP use between two adjacent states with different PMP prescriber accessibility; 2) to explore how clinicians use PMP patient reports in clinical practice; and 3) to examine the association between PMP use and clinician's responses in instances of suspected doctor shopping or diversion.

Methods

CT and RI PMPs

Controlled substance data from licensed CT pharmacies are electronically uploaded and securely stored in a central database maintained by the Department of Consumer Protection, Drug Control Division. Since July 2008, health care professionals licensed to prescribe or dispense controlled substances in CT, who have registered with the PMP online system, can actively query the PMP database about a potential patient's Schedule II-V prescriptions. The RI PMP cannot be directly accessed or queried by health care professionals; it is housed at the RI Department of Health and overseen by the Board of Pharmacy. All dispensed Schedule II and III medications are reported to the RI PMP by pharmacies. Inquiries of the RI PMP can be called in, emailed, faxed, or mailed. Law enforcement and investigative queries outnumber medical queries of the RI PMP, whereas medical queries outnumber law enforcement queries to CT's PMP.

Sample

All providers licensed to prescribe scheduled medications with an email address (CT: N = 16,924; RI: N = 5,567), including all prescribers registered with the CT PMP at the time of the survey (N = 1,400), were provided to the research team by the CT Department of Consumer Safety and the RI Department of Health for the purpose of this study. Contacts were sent electronic invitations to respond to the survey. Response rates to mailed and online surveys of clinicians are notoriously low (cf. [18–21]). Monetary incentives provide only marginal benefit in clinician response rates [22,23] and were not supported by the study budget. Instead, we employed, with assistance from the state agencies, a brief introduction email describing survey aims, the importance of provider cooperation, and that results would be used to inform PMPs and public health.

Survey

Items were developed iteratively, in collaboration with the PMP staff. Initial items were generated from several sources: a review of the literature, a prior survey, consultation with prescribers, and PMP staff suggestions. We designed the survey to take no longer than 15 minutes to complete. For providers who have queried the PMP, feedback was sought on the reporting process and how PMP data were subsequently used clinically (e.g., urine-screened patient, referred patient to treatment, asked patient to leave practice, and ignored). For RI respondents, we also gauged clinicians' willingness to use an electronic PMP. All respondents were asked whether and how they currently screen for psychiatric problems and substance abuse; counsel patients on risk factors for overdose, overdose symptom recognition, and response; and give guidance on storage and disposal of unused medications. The survey was anonymous, with deletion of the email and IP address upon sending of the invitation. Responding to the survey was considered evidence of written informed consent. ZIP code of the provider and a categorical response to number of years practicing medicine were the only required fields.

Procedure

Survey items were programmed into Survey Monkey Professional and hosted by a secure server on the site (http://www.surveymonkey.com) for 6 months. Survey invitations were emailed by the study team, with a brief letter explaining the survey's purpose and the research collaboration with the state agencies. Three motivational reminders were sent to promote higher response rates. At the end of the survey period, responses were downloaded and locked for analysis.

Analysis

Descriptive, summary statistics were tabulated by state and by PMP use. To compare between states and PMP use experience, χ2 tests and t-tests were used. Nonresponse bias was assessed by comparing the demographics of early and late responders. Differences by state in attitudes toward, barriers to, and perceptions of effectiveness in reduction of diversion and doctor shopping of PMP use as well as current practices in overdose prevention counseling were also conducted using t-tests and χ2 tests with two degrees of freedom, unless otherwise noted. Regression analyses examined the association of PMP use on responses to suspected doctor shopping and diversion, controlling for age group, gender, years practicing, screening practices, frequency of prescribing opioids, and state. Analyses were conducted in SAS v. 9.2 (SAS, Cary, NC, USA). This study was approved by the Rhode Island Hospital Institutional Review Board. Funding for this study came from the Centers for Disease Control and Prevention (PI: TCG) and had no role in the study's design, conduct, or reporting.

Results

Responses from 1,385 prescribers were received, 998 in CT and 375 in RI. Due to duplicates, administrative not clinician email addresses, spam filters, preferences to opt-out of SurveyMonkey-sponsored surveys, incorrect and defunct addresses, prescribers who had relocated or retired, who treated nonhumans (e.g., veterinarians), or who did not prescribe opioids, it was not possible to generate an accurate count of those who received the survey invitation and were eligible to participate. On the basis of the state records of the number of prescribers with controlled licenses and conservative estimates of the problems noted above, we estimate an overall response rate of approximately 10% in each state.1 We estimate having obtained a 29.6% response rate (N = 414/1,400) among PMP users, based on the number of CT PMP registered users. Item-level denominators vary as answers were not required on most items. In terms of nonresponse bias, early responders were more likely to be aged 45–60, whereas late responders were more likely to be clinicians aged 25–35 (χ2[six degrees of freedom] = 16.22, P = 0.013). No other differences in demographic or PMP experience were detected between early and late survey responders.

Sample Characteristics

Table 1 reports sample characteristics. Respondents were primarily aged 45 or older, with CT prescribers slightly younger than RI prescriber respondents (χ2 = 11.8, P = 0.008). A total of 55.0% of the respondents were male. Most prescribers reported practicing medicine for over 10 years; CT prescribers reported less experience (χ2 = 13.5, P = 0.001). For the majority of respondents, prescribing of opioids was commonplace; however, greater proportions of CT compared with RI respondents prescribed opioids on a monthly basis and proportionately more RI respondents never prescribed opioids (χ2 = 141.3, P < 0.001). A comparable and high proportion of RI and CT prescribers reported routinely screening for psychiatric symptoms and for illicit drug abuse.

Table 1

Characteristics of prescribers responding to survey

Characteristic Rhode Island Connecticut Total Responses, Statistic (P Value) 
Age N = 1,146, χ2 = 11.84, P = 0.0079 
  Under 35 years 38/315 12.1% 141/831 17.0%  
  35–44 years 73/315 23.2% 227/831 27.3%  
  45–60 years 163/315 51.7% 339/831 40.8%  
  Over 60 years 41/315 13.0% 124/831 14.9%  
Gender N = 1,129, χ2 = 0.040, P = 0.84 
  Male 172/310 55.5% 449/819 54.8% 
  Female 138/310 44.5% 370/819 45.2% 
Time as a licensed, independent practitioner N = 1,146, χ2 = 13.53, P = 0.0012 
  1–5 years 63/315 20.0% 215/831 25.9% 
  6–10 years 34/315 10.8% 138/831 16.6% 
  >10 years 218/315 69.2% 478/831 57.5% 
Frequency of prescribing opioids N = 1,373, χ2 = 141.3, P < 0.0001 
  Several times per day 90/375 24.0% 266/998 26.7% 
  About once per day 37/375 9.9% 116/998 11.6% 
  Multiple times a week 62/375 16.5% 161/998 16.1% 
  Weekly 34/375 9.1% 85/998 8.5% 
  Monthly 39/375 10.4% 200/998 20.0% 
  A few times per year 66/375 17.6% 170/998 17.0% 
  Never 47/375 12.5% 0/998 0.0% 
Current screening practices      
  Screens for psychiatric symptoms 254/373 68.1% 672/995 67.5% N = 1,368, χ2 = 0.039, P = 0.84 
  Screens for illicit drug abuse 257/375 68.5% 725/998 72.6% N = 1,373, χ2 = 2.26, P = 0.13 
Characteristic Rhode Island Connecticut Total Responses, Statistic (P Value) 
Age N = 1,146, χ2 = 11.84, P = 0.0079 
  Under 35 years 38/315 12.1% 141/831 17.0%  
  35–44 years 73/315 23.2% 227/831 27.3%  
  45–60 years 163/315 51.7% 339/831 40.8%  
  Over 60 years 41/315 13.0% 124/831 14.9%  
Gender N = 1,129, χ2 = 0.040, P = 0.84 
  Male 172/310 55.5% 449/819 54.8% 
  Female 138/310 44.5% 370/819 45.2% 
Time as a licensed, independent practitioner N = 1,146, χ2 = 13.53, P = 0.0012 
  1–5 years 63/315 20.0% 215/831 25.9% 
  6–10 years 34/315 10.8% 138/831 16.6% 
  >10 years 218/315 69.2% 478/831 57.5% 
Frequency of prescribing opioids N = 1,373, χ2 = 141.3, P < 0.0001 
  Several times per day 90/375 24.0% 266/998 26.7% 
  About once per day 37/375 9.9% 116/998 11.6% 
  Multiple times a week 62/375 16.5% 161/998 16.1% 
  Weekly 34/375 9.1% 85/998 8.5% 
  Monthly 39/375 10.4% 200/998 20.0% 
  A few times per year 66/375 17.6% 170/998 17.0% 
  Never 47/375 12.5% 0/998 0.0% 
Current screening practices      
  Screens for psychiatric symptoms 254/373 68.1% 672/995 67.5% N = 1,368, χ2 = 0.039, P = 0.84 
  Screens for illicit drug abuse 257/375 68.5% 725/998 72.6% N = 1,373, χ2 = 2.26, P = 0.13 

PMP Experience

Fewer respondents (χ2 = 85.2, P < 0.001) reported having ever used the PMP in RI (16.3%) compared with CT (43.9%, Table 2). Willingness among RI prescribers to use an electronic PMP like CT's system was high (74.3%) and, among non-users in both states, there was interest (77.4% in RI, 72.9% in CT) in using the PMP. Reasons for not currently using the PMP relate to knowledge of the system (not knowing it existed, how to enroll), particularly in CT. Staff support concerns were mentioned by a minority of respondents in both states; limited time was a more frequently cited reason for not using the PMP among CT prescribers. In RI compared with CT, practical constraints such as having limited or no Internet or telephone access at work were more frequent, as was reporting employers forbade or discouraged PMP access. A sizable minority of RI prescribers who did not use the PMP, and more so than their counterparts in CT, reported they did not want to have access to the type of information in a PMP patient report. Compared with RI prescribers who did not use the PMP, CT prescribers who did not use the PMP more frequently reported that they did not know how they would use the PMP patient report information.

Table 2

Prescription monitoring program use by prescribers

PMP Experience Rhode Island Connecticut Total Responses, Statistic (P Value) 
Has ever used the PMP 58/356 16.3% 414/943 43.9% N = 1,299, χ2 = 85.17, P < 0.0001 
Times used PMP in past year N = 472, χ2 = 38.90, P < 0.0001 
  Never 13/61 21.3% 23/411 5.6%  
  A few (1–10 times) 38/61 62.3% 178/411 43.3%  
  Monthly 8/61 13.1% 67/411 16.3%  
  Weekly or more often 2/61 3.3% 143/411 34.8%  
Willingness to use an electronic PMP 255/343 74.3%   NA 
Aware of state PMP (among non-users of PMP) 52/290 17.9% 191/511 37.4% N = 801, χ2 = 33.11, P < 0.0001 
Interested in using the PMP (among non-users of PMP) 222/287 77.4% 365/501 72.9% N = 788, χ2 = 1.94, P = 0.16 
Reasons for not using the PMP      
  Did not know the system existed 216/258 83.7% 286/422 67.8% N = 680, χ2 = 21.07, P < 0.0001 
  Do not know how to enroll 31/258 12.0% 150/422 35.5% N = 680, χ2 = 45.38, P < 0.0001 
  Do not have internet access at work or Internet access is limited 90/258 34.9% 19/422 4.5% N = 680, χ2 = 109.80, P < 0.0001 
  Do not have time to access the system 1/258 0.4% 75/422 17.8% N = 680, χ2 = 48.75, P < 0.0001 
  Do not have the staff to access the system 41/258 15.9% 56/422 13.3% N = 680, χ2 = 0.90, P = 0.34 
  I do not want to have access to this information 51/258 19.8% 15/422 3.6% N = 680, χ2 = 48.02, P < 0.0001 
  My employer/practice does not permit/discourages PMP use 50/258 19.4%   NA 
  Do not have telephone at work or telephone is limited 24/258 9.3%   NA 
  My employer/practice does not require it 3/258 1.2% 58/422 13.7% N = 680, χ2 = 31.04, P < 0.0001 
  Do not know how I would use the information 15/258 5.8% 55/422 13.0% N = 680, χ2 = 9.04, P = 0.0026 
Knows how misuse shown on patient report N = 427, χ2 = 0.054, P = 0.82 
  Yes: flag in report/in text 20/53 37.7% 135/374 46.8% 
  No: not indicated, do not know/not sure 33/53 62.3% 239/374 63.9% 
PMP Experience Rhode Island Connecticut Total Responses, Statistic (P Value) 
Has ever used the PMP 58/356 16.3% 414/943 43.9% N = 1,299, χ2 = 85.17, P < 0.0001 
Times used PMP in past year N = 472, χ2 = 38.90, P < 0.0001 
  Never 13/61 21.3% 23/411 5.6%  
  A few (1–10 times) 38/61 62.3% 178/411 43.3%  
  Monthly 8/61 13.1% 67/411 16.3%  
  Weekly or more often 2/61 3.3% 143/411 34.8%  
Willingness to use an electronic PMP 255/343 74.3%   NA 
Aware of state PMP (among non-users of PMP) 52/290 17.9% 191/511 37.4% N = 801, χ2 = 33.11, P < 0.0001 
Interested in using the PMP (among non-users of PMP) 222/287 77.4% 365/501 72.9% N = 788, χ2 = 1.94, P = 0.16 
Reasons for not using the PMP      
  Did not know the system existed 216/258 83.7% 286/422 67.8% N = 680, χ2 = 21.07, P < 0.0001 
  Do not know how to enroll 31/258 12.0% 150/422 35.5% N = 680, χ2 = 45.38, P < 0.0001 
  Do not have internet access at work or Internet access is limited 90/258 34.9% 19/422 4.5% N = 680, χ2 = 109.80, P < 0.0001 
  Do not have time to access the system 1/258 0.4% 75/422 17.8% N = 680, χ2 = 48.75, P < 0.0001 
  Do not have the staff to access the system 41/258 15.9% 56/422 13.3% N = 680, χ2 = 0.90, P = 0.34 
  I do not want to have access to this information 51/258 19.8% 15/422 3.6% N = 680, χ2 = 48.02, P < 0.0001 
  My employer/practice does not permit/discourages PMP use 50/258 19.4%   NA 
  Do not have telephone at work or telephone is limited 24/258 9.3%   NA 
  My employer/practice does not require it 3/258 1.2% 58/422 13.7% N = 680, χ2 = 31.04, P < 0.0001 
  Do not know how I would use the information 15/258 5.8% 55/422 13.0% N = 680, χ2 = 9.04, P = 0.0026 
Knows how misuse shown on patient report N = 427, χ2 = 0.054, P = 0.82 
  Yes: flag in report/in text 20/53 37.7% 135/374 46.8% 
  No: not indicated, do not know/not sure 33/53 62.3% 239/374 63.9% 

PMP = prescription monitoring program; NA = not applicable.

RI PMP users tended to report accessing the PMP 10 or fewer times in the past year, whereas CT prescribers tended to report monthly or more frequent access. Approximately one-third of prescribers in both states who had ever used the PMP correctly identified how misuse of prescription drugs was shown on the patient report.

Characteristics of PMP Users

PMP users differed from non-users in several ways and by state. In CT, PMP users were less likely to be over 60 years of age (11.5% PMP users vs 17.8% non-users, χ2[three degrees of freedom] = 9.19, P = 0.027) and PMP users were more likely to report that they screened for psychiatric problems (79.4% PMP users vs 59.2% non-users, χ2 = 43.7, P < 0.001) and illicit drug abuse (83.1% PMP users vs 65.6% non-users, χ2 = 36.3, P < 0.001) among their patients. For both states, more frequent prescribing of opioids was associated with the use of the PMP (RI: χ2 = −9.59, P = 0.002; CT: χ2 = 136.36, P < 0.001).

Influences on Medical Practice

Screening for Drug Abuse

When asked to specify how illicit drug abuse was screened, there were few state-based differences (Table 3). For both states, the most frequently endorsed methods of screening for illicit drug abuse were asking the patient directly, professional judgment, and urine drug screens. In CT, the PMP was mentioned by 36.2% as a tool used for screening for drug abuse and had higher use endorsement than any of the standardized screening assessments (e.g., CAGE, CRAFFT, DAST, and ASSIST).

Table 3

The prescription monitoring program and medical practice patterns

Medical Practice Rhode Island Connecticut Total Responses, Statistic (P Value) 
Screens for illicit drug abuse by:      
  Ask directly 228/249 91.6% 670/715 93.7% N = 964, χ2 = 1.33, P = 0.25 
  Professional judgment 171/249 68.7% 514/715 71.9% N = 964, χ2 = 0.93, P = 0.34 
  Urine screens 152/249 61.0% 412/715 57.6% N = 964, χ2 = 0.89, P = 0.35 
  PMP 25/249 10.0% 259/715 36.2% N = 964, χ2 = 60.93, P < 0.0001 
  CAGE 70/249 28.1% 210/715 29.4% N = 964, χ2 = 0.14, P = 0.71 
  CRAFFT 70/249 28.1% 19/715 2.7% N = 964, χ2 = 0.21, P = 0.65 
  DAST 10/249 4.0% 43/715 6.0% N = 964, χ2 = 1.42, P = 0.23 
  ASSIST 3/249 1.2% 20/715 2.8% N = 964, χ2 = 2.01, P = 0.16 
Medical Practice Rhode Island Connecticut Total Responses, Statistic (P Value) 
Screens for illicit drug abuse by:      
  Ask directly 228/249 91.6% 670/715 93.7% N = 964, χ2 = 1.33, P = 0.25 
  Professional judgment 171/249 68.7% 514/715 71.9% N = 964, χ2 = 0.93, P = 0.34 
  Urine screens 152/249 61.0% 412/715 57.6% N = 964, χ2 = 0.89, P = 0.35 
  PMP 25/249 10.0% 259/715 36.2% N = 964, χ2 = 60.93, P < 0.0001 
  CAGE 70/249 28.1% 210/715 29.4% N = 964, χ2 = 0.14, P = 0.71 
  CRAFFT 70/249 28.1% 19/715 2.7% N = 964, χ2 = 0.21, P = 0.65 
  DAST 10/249 4.0% 43/715 6.0% N = 964, χ2 = 1.42, P = 0.23 
  ASSIST 3/249 1.2% 20/715 2.8% N = 964, χ2 = 2.01, P = 0.16 
 Rhode Island
 
Connecticut
 
 Non-users PMP users Total Non-users PMP users Total 
Counsels patients prescribed opioids on: 
  Risks of co-ingestion 185/292; 63.4% 43/56; 76.8% N = 348, chi2 = 3.75, P = 0.05 273/497; 54.9% 329/409; 80.4% N = 906, χ2 = 65.50, P < 0.0001 
  Risk of addiction 172/292; 58.9% 38/56; 67.9% N = 348, chi2 = 1.57, P = 0.06 220/497; 44.3% 316/409; 77.3% N = 906, χ2 = 101.10, P < 0.0001 
  Unauthorized dose escalations 116/292; 39.73% 37/56; 66.1% N = 348, chi2 = 13.20, P = 0.0003 147/497; 29.6% 229/409; 56.0% N = 906, χ2 = 64.47, P < 0.0001 
  Symptoms of overdose 115/292; 39.38% 33/56; 58.9% N = 348, chi2 = 7.34, P = 0.007 171/497; 34.4% 207/409; 50.6% N = 906, χ2 = 24.23, P < 0.0001 
  Sharing medications 116/292; 39.73% 29/56; 51.8% N = 348, chi2 = 2.81, P = 0.10 146/497; 29.4% 220/409; 53.8% N = 906, χ2 = 55.53, P < 0.0001 
  Storage of medications 57/292; 19.52% 21/56; 37.5% N = 348, chi2 = 8.73, P = 0.003 76/497; 15.3% 131/409; 32.0% N = 906, χ2 = 35.66, P < 0.0001 
  Disposal of medications 38/292; 13.01% 18/56; 32.1% N = 348, chi2 = 12.73, P = 0.0004 65/497; 13.1% 100/409; 24.4% N = 906, χ2 = 19.48, P < 0.0001 
  None/does not counsel patient 61/292; 20.89% 7/56; 12.5% N = 348, chi2 = 2.10, P = 0.05 145/497; 29.2% 37/409; 9.0% N = 906, χ2 = 56.63, P < 0.0001 

 
Routinely tries to detect “dr. shopping” by: 231/347 66.6% 601/940 63.9% N = 1,287, χ2 = 0.77, P = 0.38  
  Clinical judgment 221/325 68.0% 565/874 64.6% N = 1,199, χ2 = 1.18, P = 0.28  
  Investigate if contacted by law enforcement, health department 199/325 61.2% 423/874 48.4% N = 1,199, χ2 = 15.63, P < 0.0001  
  Ask directly 194/325 59.7% 494/874 56.5% N = 1,199, χ2 = 0.97, P = 0.32  
  PMP 60/325 18.5% 384/874 43.9% N = 1,199, χ2 = 65.93, P < 0.0001  
  Electronic medical record 109/325 33.5% 279/874 31.9% N = 1,199, χ2 = 0.28, P = 0.59  
  Urine screens 102/325 31.4% 225/874 25.7% N = 1,199, χ2 = 3.33, P = 0.068  
  Screening tools 41/325 12.6% 88/874 10.1% N = 1,199, χ2 = 1.60, P = 0.21  
  Electronic pharmacy/prescription database 55/325 16.9% 65/874 7.4% N = 1,199, χ2 = 23.67, P < 0.0001  
  Insurance rejection 36/325 11.1% 86/874 9.8% N = 1,199, χ2 = 0.40, P = 0.53  
  Do not do anything 35/325 10.8% 107/874 12.2% N = 1,199, χ2 = 0.49, P = 0.48  
  Aberrant opioid use behavior assessment 13/235 5.5% 21/874 2.4% N = 1,199, χ2 = 2.19, P = 0.14  
 Rhode Island
 
Connecticut
 
 Non-users PMP users Total Non-users PMP users Total 
Counsels patients prescribed opioids on: 
  Risks of co-ingestion 185/292; 63.4% 43/56; 76.8% N = 348, chi2 = 3.75, P = 0.05 273/497; 54.9% 329/409; 80.4% N = 906, χ2 = 65.50, P < 0.0001 
  Risk of addiction 172/292; 58.9% 38/56; 67.9% N = 348, chi2 = 1.57, P = 0.06 220/497; 44.3% 316/409; 77.3% N = 906, χ2 = 101.10, P < 0.0001 
  Unauthorized dose escalations 116/292; 39.73% 37/56; 66.1% N = 348, chi2 = 13.20, P = 0.0003 147/497; 29.6% 229/409; 56.0% N = 906, χ2 = 64.47, P < 0.0001 
  Symptoms of overdose 115/292; 39.38% 33/56; 58.9% N = 348, chi2 = 7.34, P = 0.007 171/497; 34.4% 207/409; 50.6% N = 906, χ2 = 24.23, P < 0.0001 
  Sharing medications 116/292; 39.73% 29/56; 51.8% N = 348, chi2 = 2.81, P = 0.10 146/497; 29.4% 220/409; 53.8% N = 906, χ2 = 55.53, P < 0.0001 
  Storage of medications 57/292; 19.52% 21/56; 37.5% N = 348, chi2 = 8.73, P = 0.003 76/497; 15.3% 131/409; 32.0% N = 906, χ2 = 35.66, P < 0.0001 
  Disposal of medications 38/292; 13.01% 18/56; 32.1% N = 348, chi2 = 12.73, P = 0.0004 65/497; 13.1% 100/409; 24.4% N = 906, χ2 = 19.48, P < 0.0001 
  None/does not counsel patient 61/292; 20.89% 7/56; 12.5% N = 348, chi2 = 2.10, P = 0.05 145/497; 29.2% 37/409; 9.0% N = 906, χ2 = 56.63, P < 0.0001 

 
Routinely tries to detect “dr. shopping” by: 231/347 66.6% 601/940 63.9% N = 1,287, χ2 = 0.77, P = 0.38  
  Clinical judgment 221/325 68.0% 565/874 64.6% N = 1,199, χ2 = 1.18, P = 0.28  
  Investigate if contacted by law enforcement, health department 199/325 61.2% 423/874 48.4% N = 1,199, χ2 = 15.63, P < 0.0001  
  Ask directly 194/325 59.7% 494/874 56.5% N = 1,199, χ2 = 0.97, P = 0.32  
  PMP 60/325 18.5% 384/874 43.9% N = 1,199, χ2 = 65.93, P < 0.0001  
  Electronic medical record 109/325 33.5% 279/874 31.9% N = 1,199, χ2 = 0.28, P = 0.59  
  Urine screens 102/325 31.4% 225/874 25.7% N = 1,199, χ2 = 3.33, P = 0.068  
  Screening tools 41/325 12.6% 88/874 10.1% N = 1,199, χ2 = 1.60, P = 0.21  
  Electronic pharmacy/prescription database 55/325 16.9% 65/874 7.4% N = 1,199, χ2 = 23.67, P < 0.0001  
  Insurance rejection 36/325 11.1% 86/874 9.8% N = 1,199, χ2 = 0.40, P = 0.53  
  Do not do anything 35/325 10.8% 107/874 12.2% N = 1,199, χ2 = 0.49, P = 0.48  
  Aberrant opioid use behavior assessment 13/235 5.5% 21/874 2.4% N = 1,199, χ2 = 2.19, P = 0.14  

PMP = prescription monitoring program.

Counseling on Risks of Prescription Opioid Medications

Prescribers in both states reported counseling patients prescribed opioid medications at comparable rates and on similar content areas (Table 3). Counseling included risks of co-ingestion with alcohol or other central nervous system depressant (66%), risk of addiction (59%), unauthorized dose escalations (42%), symptoms of overdose (42%), and sharing medications (41%). Counseling on storage and disposal of medications was as common as not counseling the patient at all (approximately 20%). PMP users in CT were more likely to counsel on prescription opioid medication risks of any type (Table 3) compared with non-users; PMP users in RI were more likely than non-users to counsel on select risks, such as unauthorized dose escalations, overdose symptoms, storage, and disposal.

Detecting Doctor Shopping

Similar proportions of CT and RI respondents reported routinely trying to detect doctor shopping (Table 3). In CT, active efforts were more frequently used (e.g., checking the PMP) than passive efforts (e.g., investigating if contacted by the department of health or police), although clinical judgment and asking the patient directly were the more common methods of detecting doctor shopping in both states.

Perceptions of Diversion, Doctor Shopping, and Prescription Opioid Abuse at the Practice and State Level

PMP users tended to perceive that the PMP was helpful in reducing abuse of prescription opioids in their practice. RI prescribers thought that the PMP was not helpful in reducing diversion (73.5%) or abuse (74.9%) of prescription opioids in the state, regardless of their PMP use. In contrast, CT PMP users, more so than non-users, viewed that the PMP helped to reduce diversion (58.0% vs 28.5%, χ2 = 68.3, P < 0.001) and prescription opioid abuse (47.7% vs 23.4%, χ2 = 49.8, P < 0.001) in the state. RI prescribers did not differ based on their PMP experience in their perception of how common diversion or doctor shopping is among their patients (28.0%) or how serious diversion or doctor shopping is in their practice (41.6%) or in general (76.2%). In CT, PMP use was associated with a greater likelihood of perceiving diversion or doctor shopping as common among one's patients (38.2% vs 0.1%, χ2 = 69.1, P < 0.001) and perceiving diversion and doctor shopping as a serious issue both in their practice (56.4% vs 23.4%, χ2 = 89.1, P < 0.001) and in general (82.2% vs 66.3%, χ2 = 25.9, P < 0.001). It is of note that, independent of location or PMP use, prescribers perceived the problem of diversion and doctor shopping to be relatively serious in their own practice, a far more serious issue in general, and less common among their own patients.

Multivariable Logistic Regressions: PMP Use and Responses to Suspected Diversion or Doctor Shopping

Regression results (Table 4) indicated that typical responses to suspected cases of diversion or doctor shopping differed by PMP use, controlling for other potential confounding variables. PMP users were more likely than non-users to refer the patient to another provider, screen the patient for drug abuse, revisit—but not initiate—a treatment agreement/pain contract with the patient, and conduct a urine drug screen. A trend suggested that PMP users were also more likely than non-users to refer the patient to substance abuse treatment. PMP users were less likely than non-users to contact the patient's other physicians, to call law enforcement to intervene, and to do nothing or ignore the suspicion of diversion or doctor shopping (Table 4).

Table 4

Differences in response when suspect diversion or doctor shopping by prescription monitoring program use

 Typical PMP User Actions vs Typical Non-user Actions (Ref) aOR*[95% CI] 
Contact the patients other physician(s) (if known) 0.31 [0.23, 0.41] 
Discuss the concerns with the patient 1.16 [0.81, 1.68] 
Refer the patient to another provider 1.75 [1.10, 2.80] 
Screen the patient for drug abuse 1.93 [1.39, 2.68] 
Initiate a treatment agreement/pain contract with the patient 0.92 [0.66, 1.27] 
Revisit treatment agreement/pain contract with the patient 1.97 [1.45, 2.67] 
Conduct a urine drug screen of the patient 1.82 [1.29, 2.57] 
Counsel the patient on potential overdose risk 1.21 [0.83, 1.51] 
Refer the patient to substance abuse treatment 1.30 [0.96, 1.75] 
Nothing; ignore it 0.09 [0.01, 0.70] 
Ask the patient to leave the practice 0.99 [0.67, 1.47] 
Notify law enforcement 0.45 [0.21, 0.94] 
 Typical PMP User Actions vs Typical Non-user Actions (Ref) aOR*[95% CI] 
Contact the patients other physician(s) (if known) 0.31 [0.23, 0.41] 
Discuss the concerns with the patient 1.16 [0.81, 1.68] 
Refer the patient to another provider 1.75 [1.10, 2.80] 
Screen the patient for drug abuse 1.93 [1.39, 2.68] 
Initiate a treatment agreement/pain contract with the patient 0.92 [0.66, 1.27] 
Revisit treatment agreement/pain contract with the patient 1.97 [1.45, 2.67] 
Conduct a urine drug screen of the patient 1.82 [1.29, 2.57] 
Counsel the patient on potential overdose risk 1.21 [0.83, 1.51] 
Refer the patient to substance abuse treatment 1.30 [0.96, 1.75] 
Nothing; ignore it 0.09 [0.01, 0.70] 
Ask the patient to leave the practice 0.99 [0.67, 1.47] 
Notify law enforcement 0.45 [0.21, 0.94] 
*

Adjusted for age, gender, years practicing, drug abuse screening practices, frequency of prescribing opioids, and state.

aOR [95% CI] = adjusted odds ratio [95% confidence interval].

Discussion

This survey found that for health care professionals with electronic PMP (CT) access, more than 50% use it at least monthly, whereas only 16% of those who have to call up, fax, or provide written request to access a PMP (RI) use it at least monthly. Health care professionals accessing electronic PMP data tend to use it to screen for abuse and doctor shopping among their patients and as a clinical tool for discussing a patient's health status. The form of the PMP was critical to its uptake: a paper-based PMP in RI was accessed to a far less extent than the electronic PMP in CT. The use of an electronic, interactive PMP was associated with greater awareness of prescription opioid abuse and diversion and perceived utility of the PMP in reducing these problems. Findings from this small, two-state sample indicate that PMP users take a more active approach to detecting abuse and doctor shopping in their practices than non-users and the tools employed and responses to detected concerns prioritize medical over legal action. Faced with a suspicious pattern of behavior in their patient, in our study, PMP users were more likely to screen for drug abuse, refer to treatment or to another provider, and revisit a pain treatment agreement and they were less likely to employ inappropriate actions (such as initiating a treatment agreement after concern about diversion or doctor shopping was raised), inaction, or to engage law enforcement.

Results indicate a need for better communication to providers about the PMP, user-friendly education materials and continuing medical education content. Despite efforts to publicize the CT PMP (Figure 1), many non-users reported not knowing about the system. Educational content could address orientation, registration, and demonstration of how to access a state PMP; information contained in the patient report and suggested possible actions upon its review; and research on PMP use such as practice-based influences reported here. Because use of the electronic PMP was associated with increased screening for drug abuse and referral to specialists and drug treatment among our respondents, ensuring access to and availability of drug treatment options is imperative. Effective drug treatment is critical to reducing the negative consequences of nonmedical prescription drug use and ultimately may have a more lasting effect on a community than incarceration and policing efforts alone [24].

Figure 1

Connecticut prescription monitoring program informational materials (source: http://www.ct.gov/dcp).

Figure 1

Connecticut prescription monitoring program informational materials (source: http://www.ct.gov/dcp).

Self-reported counseling on risk of overdose was relatively low in this survey. Prescribers in our sample were no more likely to report counseling patients on risk of overdose when made aware of suspicious behavior. Patterns of doctor shopping may precede negative health outcomes such as opioid overdose. Evidence of recent doctor shopping was noted in 63 of 295 (21.4%) unintentional pharmaceutical overdose deaths [1] in a West Virginia study. An Australian study of opioid overdose deaths detected patterns in national pharmacy claims data of increasing doctor shopping and prescriptions of abusable medications in the year prior to death [25]. In neither study location was PMP data accessible to health professionals.

The potential role of PMPs in detecting and intervening with patients at high risk for overdose may be underutilized [16]. For instance, the PMP report could flag co-prescription of multiple central nervous system depressant medications to indicate increased overdose risk. However, PMP program materials rarely address how the PMP patient report could be used for overdose risk recognition or counseling. Therefore, it is not surprising that PMP use had little effect on prescriber's responses regarding overdose counseling in our study, even when abuse or doctor shopping was suspected. Paulozzi et al. [26] examined but failed to detect associations between existence of PMPs and decreased overdose mortality. However, the study period was not one in which electronic PMPs were widely implemented and use by health care professionals was limited; paper-based PMPs are more apt to be used primarily by law enforcement and not health care professionals [16]. Our findings suggest a critical need for specific guidance on incorporating elements of the PMP patient report into overdose prevention and an evaluation of its effect on clinical practice and overdose mortality.

Future research could evaluate the PMP patient report as part of a larger collection of prescriber tools for safer prescribing of opioid analgesics (i.e., patient agreements, screening tools, etc.); as part of clinical algorithms for more advanced monitoring of high-risk patients; or its use in flagging indications for increased safety measures such as co-prescription of naloxone [16,27,28]. In addition, future research could test for evidence of physician behavior change on patient outcomes. In CT, analyses are underway to test whether changes in physician behavior as reported among PMP users result in reduced prescription opioid overdose mortality rates. Other outcomes of interest to explore include changes on prescribing rates or types of medications, and reduced evidence in the rate of doctor shopping over time.

There are important limitations to this study. The true denominators in each state for the survey are unknown. External validity of these findings may be limited as the response rate was low, despite concerted efforts to recruit clinicians. Survey response rates by health professionals average about 10% lower than studies of the general population [29], and response rates to Web-based and emailed surveys for health professionals can vary wildly [30]. Our overall and PMP user response rates are comparable to other Web-based survey studies employing large health professional sampling frames [31–33]. However, survey respondents were comparable in gender and age to known CT and RI prescriber characteristics (cf. [34]), survey nonresponse bias was limited, and the higher estimated response rate among CT PMP users suggests greater external validity of reported practices among this group. Findings may not generalize to many other states with different PMP systems or clinician demographics. In addition, frequent use of the PMP was associated with frequent prescribing of abusable medication; wider or mandatory use of PMPs by clinicians who less frequently prescribe abusable medication may be associated with different clinical practice affects. Social desirability may have influenced reports of patient counseling on prescription opioid risks (Table 3) and therefore the true prevalence of counseling behaviors may actually be lower than that reported. Similar rank and order of rates of counseling topics by state and within PMP users suggest nondifferential measurement error. Additionally, this was a cross-sectional study, so associations may not be causal and temporality of associations cannot be established. Strengths of this survey include the large, two-state sample, the contrast in time (i.e., prior to implementation of an electronic PMP), and the breadth and detail of the items posed.

Conclusions

PMP use was associated with greater awareness of potential abuse of prescription opioids and with taking clinical rather than legal or no action when faced with suspicious medication use behavior. Wider use of electronic PMPs by prescribers may have a more direct influence on opioid abuse and ultimately overdose risk than non-use, paper-based PMPs, and predominant law enforcement use.

Acknowledgments

We are grateful to George McKenna, Nathaniel Katz, William Becker, Nickolas Roeder-Hanna, and Mike Simoli for their comments and suggestions on draft questionnaire content and survey methods.

Note

1
In CT: 16,924 − (0.08 dentists) − (0.05 veterinarians) − (0.10 email bounces, spam filters, opt outs, etc.) − (0.10 retired or relocated) − (0.10 residents or do not prescribe opioids) = 10,307 for 9.7% response rate. In RI: 5,567 − (387 dentists) − (213 veterinarians) − (0.10 email bounces, spam filters, opt outs, etc.) − (0.10 retired or relocated) − (0.10 residents or do not prescribe opioids) = 3,449 for 10.9% response rate.

References

1
Hall
AJ
Logan
JE
Toblin
RL
et al
Patterns of abuse among unintentional pharmaceutical overdose fatalities
.
JAMA
 
2008
;
2613
20
.
2
Paulozzi
LJ
.
Opioid analgesic involvement in drug abuse deaths in American metropolitan areas
.
Am J Public Health
 
2006
;
96
(
10
):
1755
7
.
3
Paulozzi
LJ
Ballesteros
MF
Stevens
JA
.
Recent trends in mortality from unintentional injury in the United States
.
J Safety Res
 
2006
;
37
(
3
):
277
83
.
4
Paulozzi
LJ
Budnitz
DS
Xi
Y
.
Increasing deaths from opioid analgesics in the United States
.
Pharmacoepidemiol Drug Saf
 
2006
;
15
(
9
):
618
27
.
5
Paulozzi
LJ
Ryan
GW
.
Opioid analgesics and rates of fatal drug poisoning in the United States
.
Am J Prev Med
 
2006
;
31
(
6
):
506
11
.
6
Coben
JH
Davis
SM
Furbee
PM
et al
Hospitalizations for poisoning by prescription opioids, sedatives, and tranquilizers
.
Am J Prev Med
 
2010
;
38
(
5
):
517
24
.
7
Centers for Disease Control and Prevention (CDC)
.
Emergency department visits involving nonmedical use of selected prescription drugs—United States, 2004–2008
.
MMWR Morb Mortal Wkly Rep
 
2010
;
59
:
705
9
.
8
Braden
JB
Russo
J
Fan
M-Y
et al
Emergency department visits among recipients of chronic opioid therapy
.
Arch Intern Med
 
2010
;
170
(
16
):
1425
32
.
9
Wisniewski
AM
Purdy
CH
Blondell
RD
.
The epidemiologic association between opioid prescribing, non-medical use, and emergency department visits
.
J Addict Dis
 
2008
;
27
(
1
):
1
11
.
10
Shah
NG
Lathrop
SL
Reichard
RR
Landen
MG
.
Unintentional drug overdose death trends in New Mexico, USA, 1990–2005: Combinations of heroin, cocaine, prescription opioids and alcohol
.
Addiction
 
2008
;
103
(
1
):
126
36
.
11
United States Centers for Disease Control and Prevention (CDC)
.
National Center for Health Statistics Homepage
 . Available at: http://www.cdc.gov/NCHS/.
2011
(accessed September 2011).
12
SAMHSA
.
State Estimates of Substance Use and Mental Disorders from the 2008–2009 National Surveys on Drug Use and Health
 .
2011
Rockville, MD
: HHS Publication No. (SMA) 11-46412011.
13
U.S. Government Accountability Office
.
Prescription drugs: State monitoring programs provide useful tool to reduce diversion
 .
Washington, DC
2002 May.
2002
Contract No.: GAO-02-634.
14
Albert
S
Brason
FW
2nd
Sanford
CK
et al
Project Lazarus: Community-based overdose prevention in rural North Carolina
.
Pain Med
 
2011
;
12
(
suppl 2
):
S77
85
.
15
Twillman
R
.
Impact of prescription monitoring programs on prescription patterns and indicators of opioid abuse
.
J Pain
 
2006
;
7
:
S6
.
16
Green
TC
Zaller
N
Rich
J
Bowman
S
Friedmann
P
.
Revisiting Paulozzi et al.'s “Prescription drug monitoring programs and death rates from drug overdose.”
Pain Med
 
2011
;
12
(
6
):
982
5
.
17
Baehren
DF
Marco
CA
Droz
DE
et al
A statewide prescription monitoring program affects emergency department prescribing behaviors
.
Ann Emerg Med
 
2010
;
56
(
1
):
19
23
. e1–3.
18
Anness
BA
Willey
CJ
Taubman
AH
.
Physician attitudes toward the Rhode Island duplicate prescription law and self-reported prescribing practices for schedule ii medications
.
Clin Res Regul Aff
 
1995
;
12
(
4
):
283
306
.
19
Shosteck
H
Fairweather
WR
.
Physician response rates to mail and personal interview surveys
.
Public Opin Q
 
1979
;
43
(
2
):
206
17
.
20
Shirts
BH
Perera
S
Hanlon
JT
et al
Provider management of and satisfaction with laboratory testing in the nursing home setting: Results of a national internet-based survey
.
J Am Med Dir Assoc
 
2009
;
10
(
3
):
161
6
. e3.
21
Braithwaite
D
Sutton
S
Smithson
WH
Emery
J
.
Internet-based risk assessment and decision support for the management of familial cancer in primary care: A survey of GPs' attitudes and intentions
.
Fam Pract
 
2002
;
19
(
6
):
587
90
.
22
Halpern
SD
Ubel
PA
Berlin
JA
Asch
DA
.
Randomized trial of 5 dollars versus 10 dollars monetary incentives, envelope size, and candy to increase physician response rates to mailed questionnaires
.
Med Care
 
2002
;
40
(
9
):
834
9
.
23
Halpern
SD
Asch
DA
.
Commentary: Improving response rates to mailed surveys: What do we learn from randomized controlled trials?
Int J Epidemiol
 
2003
;
32
(
4
):
637
8
.
24
Justice Policy Institute
.
Substance Abuse Treatment and Public Safety
 .
Washington, DC
:
Justice Policy Institute
2008
.
25
Martyres
RF
Clode
D
Burns
JM
.
Seeking drugs or seeking help? Escalating “doctor shopping” by young heroin users before fatal overdose
.
Med J Aust
 
2004
;
180
(
5
):
211
4
.
26
Paulozzi
LJ
Kilbourne
EM
Desai
HA
.
Prescription drug monitoring programs and death rates from drug overdose
.
Pain Med
 
2011
;
12
(
5
):
747
54
.
27
Beletsky
L
Ruthazer
R
Macalino
GE
et al
Physicians' knowledge of and willingness to prescribe naloxone to reverse accidental opiate overdose: Challenges and opportunities
.
J Urban Health
 
2007
;
84
(
1
):
126
36
.
28
Rich
JD
Green
TC
McKenzie
MS
.
Opioids and deaths
.
N Engl J Med
 
2011
;
364
(
7
):
686
.
29
Cummings
SM
Savitz
LA
Konrad
TR
.
Reported response rates to mailed physician questionnaires
.
Health Serv Res
 
2001
;
35
(
6
):
1347
55
.
30
Braithwaite
D
Emery
J
De Lusignan
S
Sutton
S
.
Using the Internet to conduct surveys of health professionals: A valid alternative?
Fam Pract
 
2003
;
20
(
5
):
545
51
.
31
Kim
HL
Gerber
GS
Patel
RV
Hollowell
CM
Bales
GT
.
Practice patterns in the treatment of female urinary incontinence: A postal and internet survey
.
Urology
 
2001
;
57
(
1
):
45
8
.
32
Rodriguez
HP
von Glahn
T
Rogers
WH
et al
Evaluating patients' experiences with individual physicians: A randomized trial of mail, internet, and interactive voice response telephone administration of surveys
.
Med Care
 
2006
;
44
(
2
):
167
74
.
33
McMahon
SR
Iwamoto
M
Massoudi
MS
et al
Comparison of e-mail, fax, and postal surveys of pediatricians
.
Pediatrics
 
2003
;
111
(
4 Pt 1
):
e299
303
.
34
Association of American Medical Colleges
.
State Physician Workforce Data Book
 .
Washington, DC
:
Author
;
2011
.
Disclosures: The authors have no conflicts of interest or financial relationships with any company whose products may be related to the topic of this article to disclose.
Funding: This work was supported by a grant from the Centers for Disease Control and Prevention (CDC) (R21CE001846-01 Green [PI]).
All authors have significantly contributed to the design of the study, interpretation of data, drafting/revising the article for important intellectual content, and will be involved in the final approval of any published version. The concept, design, and analysis of the study were accomplished by the first author.