Psychotropic Medications among Michigan Children Insured by Medicaid - - PowerPoint PPT Presentation

Rob Penfold, PhD Associate Investigator Group Health Research Institute penfold.r@ghc.org

Spatio-temporal Clusters of New Psychotropic Medications among Michigan Children Insured by Medicaid

Introduction

• Background
• Socio-political and practice environment
• Local culture
• Evidence-based practice & Practice-based evidence
• Spatio-temporal Surveillance
• Tool for knowledge translation and educational outreach
• Prescribing of psychotropic medications in Michigan
• Growth in prescribing
• Relationship to commercial detailing
• Implications for educational outreach visits

Acknowledgments

Kelly Kelleher, MD MPH Center for Innovation in Pediatric Practice Nationwide Children’s Hospital, Columbus, OH Kathleen Pajer, MD Center for Biobehavioral Health, NCH Brandon Strange, MD Division of Child Psychiatry, NCH Wei Wang, MSc Center for Innovation in Pediatric Practice, NCH

Engaging Providers in Practice Change

• Academic detailing, counter detailing
• Most effective approach for changing behavior

(in the “wild”) but expensive

• P&T committees
• Formulary restrictions, prior authorization, quantity

limits, step therapy, concurrent medication caps, drug utilization review

• Can be targeted
• Not all physicians need face-to-face interventions
• e.g. Trepka (2005) geographically defined intervention
• Must monitor behavior/performance

Surveillance ↔ Outreach

“As hospitals, health maintenance organizations, and

• ther health care organizations continue to develop

computerized systems to track resource utilization on a physician-specific basis, it will become increasingly practical to identify physicians with particular utilization problems and present them with focused educational interventions . . . Surveillance and feedback of this sort is another approach that may work best in certain settings with established lines of authority (e.g., teaching hospitals and staff model health maintenance organizations)” (Soumerai and Avorn, 1990, p552).

Spatio-temporal Surveillance

• Spatial Scan Statistic
• Identify communities, locations, or practice networks that

change behavior first/early

• Focus resources both spatially and temporally
• Use cluster characteristics to guide interventions

1 R03 MH077573-01A1

• Spatio-temporal Diffusion of Psychotropic Medications

to Rural Children

• Central hypothesis: the location and timing of prescribing

for psychotropic medications depends on collegial interaction during the referral process

• Hypothesis 2: the ratio of mental health specialists to

primary care physicians in a county is associated with the rate of prescribing.

First Step

• Hypothesis 1:
• Do prescriptions tend to cluster in a cross section of

new psychotropic medications?

• Hypothesis 2:
• Related to patient characteristics?
• Related to access to psychiatrists?

Why Psychotropic Medications?

• Rapid increase in utilization
• Low barriers to prescribing
• Expensive
• Cost/benefit debate
• Effectiveness, safety, side effects, non-adherence,

clinical appropriateness

• Preferred drug list and prior authorization strategies
• Preliminary results regarding olanzapine

Why Children?

• Weak evidence base regarding safety and efficacy of

these meds in young children

• No (legal) explicit data from Pharmaceutical reps

regarding children

• Collegial interaction likely to be a stronger influence
• Transmission of practice-based evidence
• Utilization versus prescribing
• Who is doing the “adopting”

Why Michigan?

• First preferred drug list in 2002 (MPPL)
• National attention
• Implemented by several states
• Prior authorization in 2002
• Representative vis-à-vis utilization and need for MH

services

Methods

Methods - Data

• Medicaid Analytic Extract (MAX) from Centers for

Medicare & Medicaid Services

• Personal summary file (age, sex, race, ZIP, SSN)
• Prescriptions file (NDC, date, Rx-ing MD)
• Michigan between Jan. 2000 and Dec. 2003
• Selected all individuals aged less than 21 years

Data cont’d

• Provider enrolment file from Michigan
• Physician ID number, specialty, ZIP
• Rural-urban commuting area codes by ZIP
• Area resource file by county
• Verispan Inc. - commercial detailing effort
• Personal selling audit
• Hospital selling audit
• Physician meetings and events audit

Study Medications

Medication FDA date Children aripiprazole (Abilify) 11/15/2002 13,076 atomoxetine (Straterra) 11/26/2002 14,087 escitalopram (Lexapro) 8/15/2002 15,570 ziprasidone (Geodon) 2/5/2001 18,302 methylphenidate OROS (Concerta) 8/1/2000 51,303

Control Medications

aripiprazole (Abilify) ziprasidone (Geodon) escitalopram (Lexapro) Atomoxetine (Strattera) methylphenidate oros (Concerta) clozapine (clozaril) olanzapine (zyprexa) quetiapine (Seroquel) risperidone (Risperdal Citalopram (Celexa) Fluoxatine (Prozac) Paroxetine (Paxil) Sertraline (Zoloft) amphetamine (Dexedrine) dextroamphetamine (Adderall) dexmethylphenidate (Focalin) methylphenidate (Ritalin)

• Matched by class

Spatio-temporal Cluster Detection

• Spatial Scan Statistic
• Two sets for each study medication
• Cases only (space-time permutation model)
• Cases versus controls (Bernoulli model)
• Focused on first 90 days after medication available
• Retrospective space-time analysis
• Time aggregation – month
• Space - 5 digit ZIP
• Clusters: circular, less than 50% period or geography

Hierarchical Models

• Hierarchical cross-classified generalized linear model
• Select unique children (first prescription, initiation)
• Model child getting study versus control medication
• Observations cross-classified by ZIP and month
• Level one: children
• Level two: ZIP and month (space & time)
• L1 covariates: age, sex, race
• L2 covariates: ZIP type, number of detailing visits
• NB: omitted Straterra

Time period for HCCGLMs ends December 31, 2003

Psychiatrist Access

• Ratio of Psychiatrist to PCP MDs within clusters versus

rest of Michigan

• Psychiatrists per capita within clusters versus rest of

Michigan

Results

Child count during first year

1000 2000 3000 4000 5000 6000 7000 1 2 3 4 5 6 7 8 9 10 11 12

Months on Michigan Medicaid Formulary Number of Children Abilify Concerta Geodon Lexapro

New Rx in the first year

500 1000 1500 2000 2500 3000 3500 1 2 3 4 5 6 7 8 9 10 11 12

Months on Michigan Medicaid Formulary Number of Children Geodon Lexapro Abilify Concerta

Commercial Detailing Effort

Abilify: National Monthly Expenditure on Meetings, and Detailing During the First 12 Months on the Market

2 4 6 8 10 Oct-02 Nov-02 Dec-02 Jan-03 Feb-03 Mar-03 Apr-03 May-03 Jun-03 Jul-03 Aug-03 Sep-03 Oct-03

Month of Availability Monthly Expenditure (millions)

Note: the observation for October 2002 is the sum of expenditures between June 2002 and October 2002

FDA approval 11/15/2002

Concerta: National Monthly Expenditure on Meetings, and Detailing During the First 12 Months on the Market

2 4 6 8 10 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00 Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01

Month of Availability Monthly Expenditure (millions) Note: pre-market expenditures on Concerta were zero.

Geodon: National Monthly Expenditure on Meetings, and Detailing During the First 12 Months on the Market

2 4 6 8 10 Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Jul-01 Aug-01 Sep-01 Oct-01 Nov-01 Dec-01 Jan-02

Month of Availability Monthly Expenditures Millions

Note: the observation for Jan 2001 is the sum of expenditures between June 1999 and Jan 2001

FDA approval 2/5/2001

Lexapro: National Monthly Expenditure on Meetings, and Detailing During the First 12 Months on the Market

5 10 15 20 25 30 Jul-02 Aug-02 Sep-02 Oct-02 Nov-02 Dec-02 Jan-03 Feb-03 Mar-03 Apr-03 May-03 Jun-03 Jul-03

Month of Availability Monthly Expenditures (Millions)

Note: the observation for July 2002 is the sum of expenditures between April 2001 and July 2002

FDA approval 8/15/2002

Primary Clusters – Space-Time Permutation

Cadillac

ABILIFY

Grand Rapids Big Rapids Kalamazoo

GEODON CONCERTA

Ann Arbor

LEXAPRO

Lansing Flint

STRATERRA

Detroit

miles

25 50

Primary Clusters Identified

Drug r miles Base Pop Obs. Cases Exp’d Cases RR LLR P-value Begin end Abilify 20.9 1756 9 .44 27.98 19.9 0.001 1/14/03 2/13/03 Concerta 125.8 9290 686 240.2 4.3 370.4 0.001 10/01/00 10/30/00 Geodon 20.5 2382 26 2.2 17.3 44.5 0.001 4/7/01 5/6/01 Lexapro 24.2 1354 4 0.19 27.1 8.84 0.066 10/14/02 11/13/02 Straterra 110.2 1896 56 2.5 33.7 131.9 0.001 1/25/03 2/24/03

Cases only: space-time permutation model

Primary Clusters - Bernoulli

• Strattera
• Abilify
• Geodon
• Lexapro
• Concerta

50 miles 100 atomoxetine county ziprasidone escitalopram aripiprazole cluster centers methylphenidate extended release

HCGLM Models - Abilify

Fixed Effect Coefficient Odds Ratio p-value Interval INTERCEPT,theta0

• 8.720082

0.0002 0.000 (0.000,0.001) Number MDS, G01

• 0.000071

0.9999 0.330 (1.000,1.000) RUCA_2, G02 0.403517 1.4971 0.049 (1.002,2.237) RUCA_3, G03

• 0.001543

0.9985 0.995 (0.647,1.542) RUCA_4, G04 0.168276 1.1833 0.381 (0.812,1.724) # Psych details (000’s), B01 2.499292 12.1739 0.000 (3.282,45.163) AGE years, P1 0.101331 1.1066 0.000 (1.077,1.137) SEX Male, P2

• 0.237994

0.7882 0.050 (0.621,1.001) Black, P3

• 0.003198

0.9968 0.982 (0.759,1.308)

• Am. Indian, P4
• 0.017121

0.9830 0.978 (0.291,3.326) Asian, P5

• 1.060158

0.3464 0.306 (0.046,2.630) Hispanic, P6

• 0.792316

0.4528 0.090 (0.181,1.131) Unknown, P7

• 0.095521

0.9089 0.858 (0.319,2.591) ZIP variance component. 0.17143 >.500 Month variance component 0.18295 0.000

HCGLM - Concerta

Fixed Effect Coefficient Odds Ratio p-value Interval INTERCEPT,theta0

• 0.317087

0.728267 0.457 (0.316,1.679) Number MDS, G01 0.000021 1.000021 0.473 (1.000,1.000) RUCA_2, G02

• 0.003038

0.996966 0.958 (0.891,1.115) RUCA_3, G03

• 0.113565

0.892646 0.061 (0.792,1.006) RUCA_4, G04

• 0.181648

0.833895 0.006 (0.733,0.948) # Psych details (000’s), B01

• 0.150557

0.860228 0.331 (0.635,1.165) AGE years, P1 0.009108 1.009149 0.003 (1.003,1.015) SEX Male, P2 0.035417 1.036052 0.109 (0.992,1.082) Black, P3 0.128284 1.136876 0.000 (1.073,1.204) American Indian, P4

• 0.012426

0.987651 0.916 (0.785,1.242) Asian, P5 0.131874 1.140964 0.486 (0.787,1.654) Hispanic, P6 0.109314 1.115513 0.081 (0.986,1.262) Unknown, P7 0.064485 1.066609 0.413 (0.914,1.245) ZIP variance component 0.12851 0.000 Month variance component 0.96679 0.000

HCGLM - Geodon

Fixed Effect Coefficient Odds Ratio p-value Interval INTERCEPT,theta0

• 6.850898

0.001059 0.000 (0.000,0.003) Number MDS, G01

• 0.000064

0.999936 0.452 (1.000,1.000) RUCA_2, G02 0.327247 1.387144 0.076 (0.966,1.993) RUCA_3, G03

• 0.3333

0.716555 0.184 (0.438,1.171) RUCA_4, G04 0.282133 1.325955 0.258 (0.814,2.160) # Psych details (000’s), B01

• 0.016795

0.983345 0.446 (0.942,1.027) AGE years, P1 0.168962 1.184075 0.000 (1.151,1.218) SEX Male, P2 0.234435 1.264194 0.039 (1.011,1.580) Black, P3

• 0.237075

0.788932 0.115 (0.588,1.059) American Indian, P4

• 0.662615

0.515501 0.365 (0.123,2.157) Asian, P5

• 0.082287

0.921007 0.911 (0.219,3.881) Hispanic, P6

• 0.562153

0.56998 0.224 (0.230,1.410) Unknown, P7 0.191463 1.21102 0.629 (0.557,2.635) ZIP variance component 0.1868 >.500 Month variance component 0.51313 0.000

HCGLM - Lexapro

Fixed Effect Coefficient Odds Ratio p-value Interval INTERCEPT,theta0

• 7.667514

0.000468 0.000 (0.000,0.002) Number MDS, G01

• 0.000126

0.999874 0.082 (1.000,1.000) RUCA_2, G02 0.022983 1.02325 0.848 (0.809,1.294) RUCA_3, G03 0.130965 1.139928 0.304 (0.888,1.463) RUCA_4, G04 0.214549 1.239302 0.072 (0.980,1.566) # Psych details (000’s), B01 0.711641 2.037331 0.002 (1.323,3.137) AGE years, P1 0.028097 1.028495 0.001 (1.012,1.045) SEX Male, P2 0.002897 1.002901 0.965 (0.884,1.138) Black, P3

• 0.147092

0.863214 0.147 (0.708,1.053) American Indian, P4

• 0.314006

0.730515 0.392 (0.356,1.499) Asian, P5 0.890239 2.435712 0.028 (1.098,5.405) Hispanic, P6 0.022396 1.022649 0.897 (0.729,1.436) Unknown, P7

• 0.545872

0.579336 0.176 (0.263,1.278) ZIP variance component 0.17782 0.000 Month variance component 0.18034 0.000

PCP to Psychiatrist ratio?

Medication In cluster Rest of MI p-value aripiprazole 2.436 2.436 1.000 atomoxetine 1.186 2.639 0.153 ziprasidone 8.597 2.108 0.000 escitalopram 3.851 2.341 0.297 methylphenidate

• ros

2.03 3.120 0.132

Psychiatrists and Child Psychiatrists per 1000 PCPs

Psychiatrists per 100,000 people

Medication in cluster rest of MI p-value aripiprazole 7.45 5.37 0.635 atomoxetine 5.24 5.5 0.923 ziprasidone 26.0 4.43 0.000 escitalopram 6.38 5.41 0.806 Methylphenidate

• ros

3.96 8.130 0.031

Discussion

Patient Commonalities?

• Age the only consistent covariate across the medications
• Detailing effort significant for Abilify and Lexapro but not

Geodon or Concerta

• Random spatial effect also inconsistent

Discussion

• Process of diffusion
• Clustering points to a contagion process
• Repetition across meds in MI identifies a target
• Implications for academic outreach
• Identify early adopters and/or champions
• Identify and capitalize on systemic properties of

clusters (environment)

• Target resources
• Rural access to new medications
• Physician networks, Pharmacy benefit managers

Future Directions

• Use contagion models with patients as a vector of

transmission in referral networks between primary and specialty care

Primary Care MD 1 Psychiatrist Primary Care MD 2

Patient 2, 3, 4 Patient 2, 3, 4

Conclusion

• Nexus of innovation theory, surveillance and outreach
• Psychiatrist-PCP density does not appear to be a main

factor

• Rural utilization does not appear to lag (at least in MI)
• Need to begin to model socio-professional networks

directly (e.g. within provider networks)