1. PDMPs as Prevention Tools
Presenters:
• Tina Farales, Department of Justice Administrator, Prescription Drug
Monitoring Program, California Department of Justice
• Peter Kreiner, PhD, Senior Scientist, Brandeis University
• Chris Baumgartner, Drug Systems Director, Washington State
Department of Health
• Neal D. Traven, PhD, Epidemiologist, Prescription Monitoring Program,
Washington State Department of Health
PDMP Track
Moderator: John L. Eadie, Coordinator, Public Health and Prescription Drug
Monitoring Program Project, National Emerging Threat Initiative, National HIDTA
Assistance Center, and Member, Rx and Heroin Summit National Advisory Board
2. Disclosures
Chris Baumgartner; Tina Farales; Peter Kreiner,
PhD; Neal D. Traven, PhD; and John L. Eadie have
disclosed no relevant, real, or apparent personal
or professional financial relationships with
proprietary entities that produce healthcare
goods and services.
3. Disclosures
• All planners/managers hereby state that they or their
spouse/life partner do not have any financial
relationships or relationships to products or devices
with any commercial interest related to the content of
this activity of any amount during the past 12 months.
• The following planners/managers have the following to
disclose:
– John J. Dreyzehner, MD, MPH, FACOEM – Ownership
interest: Starfish Health (spouse)
– Robert DuPont – Employment: Bensinger, DuPont &
Associates-Prescription Drug Research Center
4. Learning Objectives
1. Explain how state and county public health officials
use de-identified PDMP data to coordinate opioid
abuse prevention and mitigation efforts.
2. Identify challenges of using PDMP data for public
health purposes.
3. Describe the Washington State model for providing
PDMP data to local jurisdictions to inform their
resource allocation and policy decisions.
4. Provide accurate and appropriate counsel as part of
the treatment team.
6. Mike Small has disclosed no relevant, real or
apparent personal or professional financial
relationships with proprietary entities that
produce health care goods and services.
Peter Kreiner has disclosed no relevant, real or
apparent personal or professional financial
relationships with proprietary entities that
produce health care goods and services.
7. PDMPs as Prevention Tools
De-Duplicated/De-Identified Data
Learning Objectives:
Identify challenges of using PDMP data for public
health purposes
Explain how state and county public health officers use
de-identified PDMP data to coordinate opioid abuse
prevention and mitigation efforts.
8. California Health and Safety Code section § 11165. (a)
To assist health care practitioners in their efforts to ensure appropriate
prescribing, ordering, administering, furnishing, and dispensing of
controlled substances, law enforcement and regulatory agencies in
their efforts to control the diversion and resultant abuse of Schedule II,
Schedule III, and Schedule IV controlled substances, and for statistical
analysis, education, and research, the Department of Justice shall . . .
maintain the Controlled Substance Utilization Review and Evaluation
System (CURES)…
9. The prescription drug epidemic is predominantly a public health
Problem. Public Health program design, implementation and
success measurement is typically data and research driven.
PDMP data can and should assist the public health sector with the
means to devise data driven mitigation strategies and the ability to
measure the success of those efforts.
Support the Public Health Sector
10. The clinical community requires a much more informative data
Presentation than CURES 1.0’s simple provisioning of a basic
12-month PAR.
Today’s technology can provide a better “eye” on prescribers’
patients; and is capable of providing both proactive and reactive
reporting of patient prescription activity.
Technology is also capable of denoting treatment exclusivity
compacts, and providing prescribers an ability to communicate
securely across health care plans.
Enhance Informational Delivery
11. The Public Can and Should Know
The PDMPs store the most informative data regarding the current
public health crisis.
The public debate should not be deprived of the vast, telling data
housed by the PDMP.
Analytics
An analytics engine, however expensive, is essential for the delivery of optimal
PDMP information.
12. De-Duplication
PDMP patient data lacks positive identifiers.
Name:
Mike Small, Michael Small, Michael J. Small, Mikey Small, Mike Smalls
DOB:
06/19/1953, 06/19/1935, 06/19/1963
Address:
2101 Columbus Avenue, Sacramento, CA 95814
2101 Columbus Street, Sacramento, CA 95814
1201 Columbus Boulevard, San Diego, CA 95828
13. De-Duplication
Name and DOB and Zip(5) OR Name and Street Address and City
Mike Small Michael J. Small
04/19/1963 04/19/1963
2101 Columbus Ave 2100 Columbia Way
Sacramento, CA 95814 Sacramento, CA 95814
Mikey Small
04/19/1963
1201 Columbus Boulevard
San Diego, CA 92111
Michael Small Mike Smalls
04/19/1936 04/19/1963
2101 Columbus Avenue 2101 Columbus Ave.
Sacramento, CA 95814 Sacramento, CA 95814
One Mike Small
Entity
14. De-Duplication
Every day approximately 145K new Rx records are added to the CURES 2.0
data base. With this new data, the analytics engine must re-resolve
patient, prescriber and dispenser entities across the 1TB database every
night in order to produce daily CURES 2.0 Patient safety messaging alerts.
The de-duplicated data also contributes to the quarterly and annual
systematic production of a statewide and 58 county de-identified data sets
for use by public health officers and researchers.
15. De-Identified Data
Anonymized Patient ID
Anonymized Prescriber ID
Anonymized Pharmacy ID
Patient Birth Year
Patient Gender
Patient Zip Code
Patient County
Patient State
Prescriber Zip Code
Prescriber County
Prescriber State
Pharmacy Zip Code
Pharmacy County
Pharmacy State
Product Name
NDC
Drug Form
Strength
Quantity
Days Supply
Date Filled
Refill Number
Payment Code
Prescriber Specialty
Prescriber Board Certification
Indicator
• Personally identifying information redacted.
• Anonymized patient IDs maintained to be consistent from report to
report.
• Generated quarterly and annually for each county and the entire
state.
16. De-Identified Data Normalization
With PDMPs in 49 states and all territories, it is important to normalize
PDMP de-identified data sets for national level research and analysis.
17. Examples: CURES California County Data Shared with
State and County Departments of Public Health
• Opioid prescribing rates (minus buprenorphine
formulations thought to be associated with MAT)
• Average opioid dosage/Percent of residents with high
(> 100 MME) average daily dosage
• Concurrent opioid and benzodiazepine prescriptions
• Change in opioid prescribing rates, 2010 – 2013
• Change in average opioid dosage, 2010 – 2013
• Change in number of waivered physicians, 2010 -
2013
18. California Opioid Prescribing Rates per 1,000 Residents,
by County, 2013
Opioid prescriptions per 1,000 population
386.2 - 568.7
568.7 - 678.3
678.3 - 961.4
961.4 - 1163.8
1163.8 - 1767.1
19. California: Average Opioid Dosage per 1,000 Residents
in 2013, by County
Dosage in MMEs
306.2 - 599
599 - 745.9
745.9 - 1201.4
1201.4 - 1721.9
1721.9 - 2732.7
Resident
20. California: Percent of Opioid Patients Receiving > 100 MME
During a 30-Day Period in 2013, by County
Percent with > 100 MME
3.7 - 8.1
8.1 - 9.8
9.8 - 15.2
15.2 - 23.2
23.2 - 41.1
Residents per
1,000
Number of Residents per 1,000
For at Least 30 Days During 2013, by County
21. California: Patients with Concurrent Opioid
and Benzodiazepine Prescriptions, Per 1,000 Residents, by County, 2013
Concurrent prescription rate per 1,000
3.5 - 8
8 - 11.7
11.7 - 16.6
16.6 - 25
25 - 41
Residents per 1,000 with Both
by County, 2013
22. California: Change in Opioid Prescribing Rates,
2010 to 2013, by County
Change in opioid prescribing rates
< -3 Std. Dev.
-3 - -2 Std. Dev.
-2 - -1 Std. Dev.
-1 - 0 Std. Dev.
Mean
0 - 1 Std. Dev.
1 - 2 Std. Dev.
23. California: Change in Average Opioid Dosage Rate,
2010 to 2013, by County
Dosage change 2010 to 2013
< -3 Std. Dev.
-3 - -2 Std. Dev.
-2 - -1 Std. Dev.
-1 - 0 Std. Dev.
Mean
0 - 1 Std. Dev.
1 - 2 Std. Dev.
2 - 3 Std. Dev.
24. California: Change in Number of Waivered Physicians,
2010 to 2013, by County
Change in waivered physicians
-2 - -1 Std. Dev.
-1 - 0 Std. Dev.
Mean
0 - 1 Std. Dev.
1 - 2 Std. Dev.
2 - 3 Std. Dev.
25. Observations
• Several northern counties with relatively small
population were highest in rates of risk indicators
(e.g., Del Norte, Lassen, Plumas, Tehama, Trinity),
suggesting need for treatment and prevention
• Two of these (Plumas and Trinity) also had high
percent increases in average MMEs per resident,
2010 – 2013, and low percent increases in number of
physicians waivered to prescribe buprenorphine for
medically-assisted treatment over the same period
27. PDMP Track: Linking and
Mapping PDMP Data
Chris Baumgartner, WA State Dept. of Health
Neal Traven, WA State Dept. of Health
28. Disclosure Statement
• Chris Baumgartner and Neal Traven have
disclosed no relevant, real or apparent
personal or professional financial relationships
with proprietary entities that produce health
care goods and services.
29. Learning Objectives
1. Explain how state and county public health officials
use de-identified PDMP data to coordinate opioid
abuse prevention and mitigation efforts.
2. Identify challenges of using PDMP data for public
health purposes.
3. Describe the Washington State model for providing
PDMP data to local jurisdictions to inform their
resource allocation and policy decisions.
4. Provide accurate and appropriate counsel as part of
the treatment team.
30. Unintentional Prescription Opioid Overdose Deaths
Washington 1995-2014
Source: Washington State Department of Health, Death Certificates
31. Unintentional Opioid Overdose Deaths Washington 1995-2014
Source: Washington State Department of Health, Death Certificates
32. WA State Unintentional Poisonings
Workgroup (UPWG)
• Began quarterly meetings in June 2008
• Representatives from public & private organizations:
• State/local health agencies, tribal authorities, insurers, law enforcement,
substance abuse prevention/treatment, poison control, health professional
associations, academic institutions, etc…
• Developed short-term actions
• Increase provider and public education
• Identify methods to reduce diversion through emergency departments
• Increase surveillance
• Support evaluation of practice guidelines for providers treating chronic,
non-cancer pain
• Support prescription monitoring program
33. 2016 Washington State Interagency
Opioid Working Plan
33
Goal 1: Prevent opioid misuse and abuse
• Improve prescribing practices
Goal 2: Treat opioid dependence
• Expand access to treatment
Goal 3: Prevent deaths from overdose
• Distribute naloxone to people who use heroin
Goal 4: Use data to monitor and evaluate
• Optimize and expand data sources
34. Opioid Plan - Goal 4 Strategies
1. Improve PDMP functionality to document and
summarize patient and prescriber patterns to
inform clinical decision making
2. Utilize the PDMP for public health surveillance
and evaluation
3. Continue and enhance efforts to monitor opioid
use and opioid-related morbidity and mortality
4. Monitor progress towards goals and strategies
and evaluate the effectiveness of our
interventions
35. 35
County Profiles Project
• Provide information – counts, rates, maps, analyses – to
Local Health Jurisdictions (LHJs), for use in building their
programmatic solutions
• Time trends in prescription drug use
o Which drugs are commonly prescribed?
o How frequently are they used?
o In combination with other Controlled Substances?
• Geographic patterns of drug use
o Apply online mapping tools
o “Overdose and At-Risk Behaviors”
o Identify “treatment deserts”
36. 36
Our Inspiration
Oregon PDMP County Reports
• Approximately 20 tables
o Age-group counts and rates
o Specific drugs or drug classes
• Little analysis
o No comparisons between counties
o No time trends
o No graphics or maps
o Brief, generic discussion
• One-time effort?
o County reports not published for
2013, 2014, 2015
Using this as our takeoff point…
37. 37
Question:
• What kinds of information will be most valuable to Local Health
Jurisdictions in developing programs regarding Controlled
Substances?
Answer:
• We aren’t really sure, so let’s ask them!
Action:
• Invited all LHJs to join Advisory Workgroup, to collaborate with
the PMP in designing a report framework that will contain the
most useful information.
LHJ Advisory Workgroup (I)
38. 38
Seven county-level LHJs volunteered to
participate in shaping the profile reports
Department of Health convened
teleconferences, which discussed:
• Cross-referencing LHJ wishlists to available PMP
data fields
• Useful counts, groupings, summaries selected
• Decision to adjust, where appropriate, by age
group and gender
LHJ Advisory Workgroup (II)
39. LHJ Advisory Group counties
Clallam Snohomish
Grant
Spokane
Klickitat
Thurston
Clark
40. Table 3. Top 10 Controlled Substances by Number of County
Residents Receiving Such Medications
Table 5. Unique Recipient Count and Usage Rate for Most
Common Opioid, by Age-Sex Group
Table 13. Unique Recipient Count and Usage Rate for All
Benzodiazepines and for Most Common Benzodiazepine,
by Source of Payment
Table 19. Unique Recipient Count and Usage Rate for Opioid and
Benzodiazepine Combination, by Age-Sex Group
Figure 3. Time Trends in the Proportion of Patients Exhibiting At-risk
Behavior Among Opioid Users, in County and Statewide
Proposed Profile Content: Examples
41. So … where are we now on
the County Profiles project?
We ran into a few
problems and issues in the
PMP dataset
42. 42
PMP Data Issues (I)
• Database size, Security
o Highly confidential information
Analysis on non-networked computer
Encryption with BitLocker
o 45.0 million prescription records as of 07/20/2015
Add almost 1 million records per month
o Processing power
Dedicated SQL server
Analytic workstation with lots of RAM
• Fully-identified Data
o Prescribers (>130K), Dispensers (~3,300) – DEA #, Address
o Recipients (>5.2M, or is it really 4.1M??) – Name, Address, DOB
o Create alternate identifiers for use by external researchers
Maintain crosswalks between full and alternate identifiers
43. 43
PMP Data Issues (II)
• Clustering and Linking to Individual Recipients
o Tradeoffs in Under- or Over- clustering
Under- = Overestimate Number of Recipients
Over- = Overestimate Number of High-Risk Recipients
o Improve accuracy of clustering
Machine learning
Better clustering algorithms
• Data cleaning and editing
o Non-human recipients (Species Code?)
o Malformed or unknown identifiers (DEA, NDC, Zip Code)
o Data entry and/or upload errors
Really? 11.9 billion doses of tramadol?
Correct street, city, Zip, county … but state code is blank
State code defaults to WA, so we see things like:
Atlanta, 30318, Fulton, WA
Louisville, 40206, Jefferson, WA
44. 44
PMP Data Issues (III)
• Reference Databases
o DEA Numbers
Available at no charge to State Agencies
Real-time snapshot, possibly retrospective views
o NDC Codes
Obtain from FDA’s database, very frequently updated
Linking Packaging and Product tables
Morphine Equivalent Dose reference
o NPI
Prescriber specialty
o Zip Code
Frequent redrawing, addition of new ones
Use 3-digit to identity state
What to do about non-existent Zip codes?
46. 46
Prescriptions per 1,000 Population:
All Controlled Substances, 2014
2,050-2,800
1,800-2,050
1,650-1,800
1,450-1,650
700-1,450
Whatcom
Skagit
Clallam
San
Juan
Island
Jefferson
Grays
Harbor
Snohomish
Mason
King
Kitsap
Pierce
Thurston
Pacific Lewis
Wahkiakum Cowlitz
Clark
Skamania
Douglas
Chelan
Whitman
Okanogan
Walla
Walla Asotin
Spokane
Pend
OreilleFerry
Stevens
Kittitas
Yakima
Grant
Klickitat
Lincoln
Adams
Benton
Garfield
Columbia
Franklin
47. 47
1,330-1,700
1,060-1,330
925-1,060
850-925
400-850
Whatcom
Skagit
Clallam
San
Juan
Island
Jefferson
Grays
Harbor
Snohomish
Mason
King
Kitsap
Pierce
Thurston
Pacific Lewis
Wahkiakum Cowlitz
Clark
Skamania
Douglas
Chelan
Whitman
Okanogan
Walla
Walla Asotin
Spokane
Pend
OreilleFerry
Stevens
Kittitas
Yakima
Grant
Klickitat
Lincoln
Adams
Benton
Garfield
Columbia
Franklin
Prescriptions per 1,000 Population:
Pain Relievers (Opioids), 2014
48. 48
375-500
340-375
317-340
265-317
140-265
Whatcom
Skagit
Clallam
San
Juan
Island
Jefferson
Grays
Harbor
Snohomish
Mason
King
Kitsap
Pierce
Thurston
Pacific Lewis
Wahkiakum Cowlitz
Clark
Skamania
Douglas
Chelan
Whitman
Okanogan
Walla
Walla Asotin
Spokane
Pend
OreilleFerry
Stevens
Kittitas
Yakima
Grant
Klickitat
Lincoln
Adams
Benton
Garfield
Columbia
Franklin
Prescriptions per 1,000 Population:
Tranquilizers (Benzodiazepines), 2014
49. 49
225-300
198-225
165-198
150-165
80-150
Whatcom
Skagit
Clallam
San
Juan
Island
Jefferson
Grays
Harbor
Snohomish
Mason
King
Kitsap
Pierce
Thurston
Pacific Lewis
Wahkiakum Cowlitz
Clark
Skamania
Douglas
Chelan
Whitman
Okanogan
Walla
Walla Asotin
Spokane
Pend
OreilleFerry
Stevens
Kittitas
Yakima
Grant
Klickitat
Lincoln
Adams
Benton
Garfield
Columbia
Franklin
Prescriptions per 1,000 Population:
Stimulants, 2014
50. 50
165-315
145-165
132-145
119-132
65-119
Whatcom
Skagit
Clallam
San
Juan
Island
Jefferson
Grays
Harbor
Snohomish
Mason
King
Kitsap
Pierce
Thurston
Pacific Lewis
Wahkiakum Cowlitz
Clark
Skamania
Douglas
Chelan
Whitman
Okanogan
Walla
Walla Asotin
Spokane
Pend
OreilleFerry
Stevens
Kittitas
Yakima
Grant
Klickitat
Lincoln
Adams
Benton
Garfield
Columbia
Franklin
Prescriptions per 1,000 Population:
Sedatives, 2014
51. 51
Drug Name
N of
Tablets/Capsules
Prescriptions per
1,000 State Residents
Hydrocodone 2,690,470 386
Oxycodone 1,779,532 255
Zolpidem 737,864 106
Alprazolam 600,700 86
Lorazepam 587,326 84
Dextroamphetamine/Amphetamine 547,771 79
Clonazepam 494,936 71
Codeine 458,487 66
Methylphenidate 440,009 63
Morphine 312,270 45
Ten Most Frequently Prescribed Drugs, 2014:
Statewide, Tablets and Capsules only
Population estimate = 6,968,170
WA Office of Financial Management, Population Unit
52. 52
Clallam Clark Garfield Snohomish
Hydrocodone 472 Hydrocodone 386 Hydrocodone 903 Hydrocodone 392
Oxycodone 449 Oxycodone 258 Morphine 191 Oxycodone 329
Codeine 74 Codeine 64 Oxycodone 189 Codeine 67
Methadone 74 Morphine 56 Codeine 113 Morphine 50
Morphine 67 Tramadol 31 Tramadol 84 Buprenorphine 41
Five Most Frequently Prescribed Opioids, 2014:
Selected Counties, Prescriptions per 1,000 Population
Population estimates:
Clallam 72,500
Clark 442,800
Garfield 2,240
Snohomish 741,000
WA Office of Financial Management, Population Unit
53. Since we started the County Profiles project…
• Greatly increased attention has been paid to opioids –
nationally, statewide, and locally
o Frequent reports in newspapers, TV news
o Locally produced documentaries
o Frontline on PBS, reported from King and Kitsap Counties
• Developing the state’s Interagency Opioid Working Plan
o PMP database now seen as a vital data source for public health
efforts at surveillance, monitoring, and evaluation
o As part of the Working Plan, the County Profiles project will
provide information on trends in opioid prescribing and use
o Dissemination of PMP reports, including the Profiles project,
beyond Local Health Jurisdictions
54. And as we look ahead…
• We believe we are close to resolving the pitfalls and problems
we have encountered
• Documentation is being written so that the scripts and
programs that emerged from our deep dive into the PMP data
will be maintained and, when necessary, updated
• Going back to the raw datasets obtained from our vendor, we
will build “clean” data files that will be placed on our secure
SQL server
• The one-time code written thus far will be converted to scripts
and macros so as to “automate” production of reports and
analyses
• GIS views of the PMP data and other layers will continue to be
developed and studied
• And maybe we’ll finally be able to catch our breath!
55. Contacts
Chris Baumgartner, PMP Director
chris.baumgartner@doh.wa.gov
Neal Traven, PMP Epidemiologist
neal.traven@doh.wa.gov
55
56. PDMPs as Prevention Tools
Presenters:
• Tina Farales, Department of Justice Administrator, Prescription Drug
Monitoring Program, California Department of Justice
• Peter Kreiner, PhD, Senior Scientist, Brandeis University
• Chris Baumgartner, Drug Systems Director, Washington State
Department of Health
• Neal D. Traven, PhD, Epidemiologist, Prescription Monitoring Program,
Washington State Department of Health
PDMP Track
Moderator: John L. Eadie, Coordinator, Public Health and Prescription Drug
Monitoring Program Project, National Emerging Threat Initiative, National HIDTA
Assistance Center, and Member, Rx and Heroin Summit National Advisory Board
Hinweis der Redaktion
Correction: average opioid dosage in MMEs per resident in 2013.
Correction: Number of residents per 1,000 receiving > 100 MME daily for at least 30 days during 2013
Correction: Residents per 1,000 with both opioid and benzodiazepine prescriptions for at least 30 days
Change in MME per resident per year.
30
31
I wanted to begin this talk with a bit of history about how this work started. Back in 2008, the Washington State Department of Health began a quarterly workgroup in June 2008 focused on preventing prescription, misuse, abuse and overdose. The purpose of the group was to coordinate the prevention activities already underway, set up a forum for continuing communication, and to come up with short term actions that we could work on together.
I’ve included examples of who is represented on the workgroup. It is relevant to this discussion to point out that there were several emergency department physicians who attended these meetings.
During the first few meetings we developed a charter, which outlines the short term actions.
County 2014 populations, notes:
Clallam (72 K) – Olympic Peninsula, rural, mountainous, Forks
Clark (440 K) – across from Portland, urban, America’s Vancouver
Grant (93 K) – rural, agricultural, Grand Coulee
Klickitat (21 K) – rural, many residents shop in Oregon
Snohomish (740 K) – urban/suburban
Spokane (485 K) – WA’s 2nd largest city
Thurston (264 K) – Olympia
Point out that we’ve combined names as displayed in the database. We believe that it doesn’t matter whether it comes with acetaminophen, ibuprofen, or aspirin – we’re interested only in the hydrocodone.