Using the NC Controlled Substances Reporting System to Identify Providers Manifesting Unusual Prescribing Practices

Transcription

1 Using the NC Controlled Substances Reporting System to Identify Providers Manifesting Unusual Prescribing Practices Prevention Research Center Penn State December 2, 2015 Chris Ringwalt, DrPH* Sharon Schiro, PhD** Meghan Shanahan, PhD* Scott Proescholdbell, MPH*** Harold Meder, MBA* Anna Austin, MPH,*** Nidhi Sachdeva, MPH *** *UNC Injury Prevention Research Center **UNC Department of Surgery ***NC Division of Public Health This study was supported by a grant #2012-R2-CX-0002 from the National Institute of Justice, Office of Justice Programs, U.S. Department of Justice. The opinions, findings, and conclusions or recommendations expressed are those of the authors and do not necessarily reflect those of the Department of Justice. 1

2 Study Goal To develop and validate a set of algorithms from metrics that utilize data from North Carolina s PDMP to develop a screening tool to identify prescribers who manifest unusual and uncustomary prescribing patterns 2

3 Problems with Use of PDMPs to Detect Inappropriate Prescribing Lack of clarity as to which PDMP indicators may serve as a good screening tool Concerns about the potential for multiple false positives Lack of resources to investigate providers identified by these screens Lack of information in PDMPs concerning provider specialty (e.g., oncologists, end-of-life treatment specialists) and level of responsibility Concern that providers treating chronic patients may: Dismiss those prematurely Treat them sub-optimally Decline to accept these patients into their practices 3

4 How do Regulatory Authorities Detect Inappropriate Prescribing Now? Complaints from patients and colleagues Audits of medical records Investigations by coroners or chief medical examiners However, currently, there is no standardized screening tool to apply to Prescription Drug Monitoring Programs for this purpose 4

5 What are Prescription Drug Monitoring Programs (PDMPs)? State-level electronic databases about controlled substances that are prescribed and dispensed, and the patients who receive them Purposes: to inform providers prescribing and pharmacists dispensing behaviors to reduce: the abuse and diversion of controlled substances inappropriate describing and dispensing Established by Congress in 2002 Operational in all states but Missouri 5

6 Key prescription-level variables (NC) Date prescription: Written Dispensed Pill quantity Days supply Refills authorized Schedule (1-IV) Drug class* National drug code** Drug name Strength and formulation Conversion factor to milligrams of morphine equivalents (MMEs) * e.g., opioid, benzodiazepine, stimulant **a unique 10-digit, 3-segment numeric identifier assigned to each medication listed under Section 510 of the US Federal Food, Drug, and Cosmetic Act. 6

7 Key patient-level variables Names Gender Date of birth Address, zip code and county Unique identifier Method of payment Species* *I am not making this up. 7

8 Key provider and dispenser variables Name DEA number* Address, including county and zipcode Date of registration with PDMP Query dates But not: Specialty *Pharmacies have DEA numbers, not individual pharmacists 8

9 Candidates for Metrics Providers who Write the Highest: Rates of prescriptions for daily doses of opioids >100 milligrams of morphine equivalents (MMEs) Average daily dose of MMEs Total MMEs for each prescription Rates of prescriptions for following drug classes, irrespective of dose: Benzodiazepines Opioids Stimulants Rates of co-prescribed benzodiazepines + opioids >100 MMEs Temporally overlapping prescriptions 9

10 Candidates for Metrics Providers with Patients who: Travel long distances from their homes to their: Providers Pharmacies Fill prescriptions received from multiple providers (doctor shopping) for: Opioids Stimulants Benzodiazepines Any controlled substance Fill prescriptions at multiple pharmacies (pharmacy hopping) 10

11 Example of metric distribution 2000 Average daily rate that NC providers write opioid prescriptions for >100 MMEs (2012 data) Number of Providers

12 Example: Distribution tail Average daily rate that NC providers write opioid prescriptions for >100 MMEs Number of Providers

13 Initial Validation Strategy Combed NC Vital Statistics records for deaths (N=465) in 2012 related to opioid overdose used t-codes representing drug-related poisonings Recorded DEA #s of providers who had prescribed opioids to these patients within 30 days of their death. Any given decedent could have received prescriptions from multiple providers (N=651; mean=1.4) Matched these to metrics relating to: List 1: Top 1% of prescribers of controlled substances in each tail List 2: Top 1% of prescribers in each tail + top 1% of prescribers for all controlled substances Thus List 2 is a subset of List 1 Note that because the number of providers in each full distribution varies, the number in the top 1% will also 13

14 Co-prescribed benzodiazepines + opioids >100MMEs 60 n= Providers who did not prescribe opioids to a decedent n= % 77% Providers who prescribed opioids to a decedent 0 Highest 1% of this metric Highest 1% of this metric + 1% of prescribers 14

15 180 Temporally overlapping prescriptions n= Providers who did not prescribe opioids to a decedent Providers who prescribed opioids to a decedent % Highest 1% of this metric n=18 61% Highest 1% of this metric + 1% of prescribers 15

16 180 Prescriptions for opioids >100MMEs 160 n= Providers who did not prescribe opioids to a decedent 100 n= Providers who prescribed opioids to a decedent 20 34% 43% 0 Highest 1% of this metric Highest 1% of this metric + 1% of prescribers 16

17 350 Prescriptions for any opioids 300 n= Providers who did not prescribe opioids to a decedent 200 n= Providers who prescribed opioids to a decedent 50 36% 42% 0 Highest 1% of this metric Highest 1% of this metric + 1% of prescribers 17

18 300 Prescriptions for any benzodiazepine n= Providers who did not prescribe opioids to a decedent n= Providers who prescribed opioids to a decedent 50 30% 32% 0 Highest 1% of this metric Highest 1% of this metric + 1% of prescribers 18

19 Non-Performing Metrics*: Providers with Patients who Travel long distances to their Providers Pharmacies Are: doctor shoppers pharmacy shoppers * With this validation effort, at least 19

20 Caveats Prescribing opioid analgesics within a month of a patient s death does not constitute causality There are other sources of opioids (e.g., heroin) Attributing deaths to opioid overdoses is not a perfect science Findings from these metrics only represent an initial screen Greater concurrent validity related to providers in top 1% of all prescribers of a controlled substance (2 nd bar) may be a function of greater exposure i.e., they write the most prescriptions Our PDMD: Lacks specialty information Lacked (until last year) payer information 20

21 Potential Uses for Study Findings State medical boards and other investigatory bodies Potentially problematic providers can be quickly identified Patients who have received problematic levels of prescriptions can be identified and their charts reviewed to determine if the prescriptions were appropriate Metric placement (rate & rank) can assist investigations by demonstrating to providers exactly where they lie on these distributions North Carolina Medical Board has just adapted and published several of our metrics, namely: 1. Top 1% of providers who prescribe 100 MMEs/patient/day 2. #1 above + Any benzodiazepine + Top 1% of all prescribers of controlled substances by volume Same technology can be brought to bear on potentially problematic pharmacies (dispensers) 21

22 Directions for Further Research Further validation required: Cases of prescription drug malfeasance known to medical boards and law enforcement, in a cross-sectional (but preferably longitudinal) context Other metrics may be of interest Multiple prescriptions for long-acting opioid analgesics for patients with chronic non-cancer pain Very high daily doses (>120 MMEs) of opioid analgesics Providers who appear in the tails of multiple metrics Providers and dispensers who share patients with multiple prescriptions for controlled substances Investigations of provider prescribing and dispenser filling behaviors in a multi-state context Effects of use of screening mechanisms should be carefully evaluated to determine potential for chilling effects on prescribing behaviors. Primum non nocere. 22

23 Barriers to Research Lack of interested funders Lack of access to state-level PDMP data PDMP data may be anonymized, constraining ability to link providers, dispensers, and patients Datasets are very large and require: Substantial cleaning and variable construction Secure servers with a very large capacity Sophisticated (and expensive) programming Analysis time Lack of access to multi-state PDMP databases Lack of key data elements, e.g.: Method of payment (particularly cash) Provider specialty and context Ability to link across multiple years Inability to link to administrative medical datasets (e.g., Medicaid) 23

24 Contact information Chris Ringwalt, DrPH Injury Prevention Research Center University of North Carolina Chapel Hill, N.C. 24