Back pain Outcomes using Longitudinal Data (BOLD): Lessons for LIRE - PowerPoint PPT Presentation

Back pain Outcomes using Longitudinal Data (BOLD): Lessons for LIRE (and Other Pragmatic Trials) Jeffrey (Jerry) Jarvik, M.D., M.P.H. Professor of Radiology and Neurological Surgery Adjunct Professor Health Services and Pharmacy Director, Comparative Effectiveness, Cost and Outcomes Research Center (CECORC) University of Washington

Acknowledgements-BOLD • AHRQ: R01 HS019222-01 •NIH 1UH2AT007766-01 Disclosures • Physiosonix (ultrasound company) –Founder/stockholder • Healthhelp (utilization review) –Consultant • Evidence-based Neuroradiology (Springer) –Co-Editor

Key People UW Non-UW • Jerry Jarvik, MD,MPH- PI • Rick Deyo, MD, MPH-OHSU • Katie James, PA-C, MPH-Proj • Dan Cherkin, PhD-GHRI Dir • Heidi Berthoud MPH- GHRI • Bryan Comstock, MS- Biostats • Safwan Halabi, MD-HFHS • Nick Anderson, PhD- • Dave Nerenz, PhD- HFHS Biomedical Informatics • Dave Kallmes, MD- Mayo • Brian Bresnahan, PhD- Health • Jyoti Pathak, PhD- Mayo Economist • Patrick Luetmer, MD- Mayo • Patrick Heagerty, PhD- Biostat • Andy Avins, MD MPH-KPNC • Judy Turner, PhD- Psychologist/Pain expert

Inappropriate Imaging • 30-40% of imaging studies in the U.S. may be inappropriate Picano E. Sustainability of medial imaging. BMJ. 2004;328:578-580

Background and Rationale • Lumbar spine imaging frequently reveals incidental findings • These findings may have an adverse effect on: –Subsequent healthcare utilization –Patient health related quality of life

Prevalence of Disc Degeneration in Normals Modality Author/ Age Prev Year Range MR Boden/ 20-60 44% 1990 60-80 93% MR Stadnik/ 17-60 52% 1998 61-71 80% MR Weishaupt/ 20-50 72-100% 1998 MR Jarvik/ 35-70 91% 2001

Disc Degeneration

Back pain Outcomes using Longitudinal Data (BOLD) • CER for seniors with back pain • AHRQ funded- part of $1.1 billion American Recovery and Reinvestment Act (ARRA)

BOLD CHOICE (Clinical and Health Outcomes Initiative in CE) • Overall goal: establish registry to evaluate effectiveness, safety, and cost-effectiveness of interventions for pts > 65 with back pain • Setting: HMO Research Network • Sites – Kaiser Northern CA: Andy Avins – Henry Ford Health System Detroit: Dave Nerenz – Harvard Pilgrim/Vanguard Boston: Srdj Nedeljkovic

BOLD CHOICE: 3 Aims 1. Establish BOLD registry 2. Conduct observational cohort study of early imaging 3. Conduct RCT of epidural steroid injections plus local anesthetic (LA) vs. LA alone

BOLD Aim 1: Registry Measures • 1) Roland-Morris Questionnaire • 2) 0-10 pain NRS-avg pain past 7d • 3) pain interference with activity (BPI) • 4) patient expectation re recovery • 5) PHQ-4 Depression/Anxiety • 6) EQ-5D • 7) Brief fall screen

BOLD Aim 2: Early Imaging Cohort • Observational cohort • Compare early to no early imaging in elderly with new visit for LBP • Outcomes: Disability (RMDQ), pain, subsequent resource utilization • Propensity score matching to control for variables that affect receiving imaging

BOLD Aim 2: Early Imaging Cohort • Observational cohort • Compare early to no early imaging in elderly with new visit for LBP • Outcomes: Disability (RMDQ), pain, subsequent resource utilization • Propensity score matching to control for variables that affect receiving imaging

Lumbar Imaging with Reporting of Epidemiology (LIRE) Proposed Study Flow Primary Care Clinics With LBP Patients Randomize Clinics Macro with No macro with prevalence info prevalence info Outcomes Outcomes Assessment- Assessment- Resource Resource Utilization Utilization

LIRE Primary Aim • To determine whether inserting age- specific prevalence of imaging findings among asymptomatic subjects into lumbar spine imaging reports decreases back-related interventions (imaging, injections, surgeries, etc.) over the subsequent year

GHC Test Template

Intervention Text

Stepped Wedge Design

Stepped Wedge Design • A one-way cluster, randomized crossover design • Temporally spaces the intervention • Assures that each participating clinic eventually receives the intervention • Within site comparison controls for between site differences (eg- CPT coding)

LIRE Sites • Kaiser Permanente • Group Health Northern California Research Institute/GHC – Dan Cherkin, PhD – Andy Avins, MD MPH • Mayo Clinic Health System • Henry Ford Health System – Dave Kallmes, MD – Safwan Halabi, MD

Site Characteristics Site # # PCPs # Patients # Back # L- Primary Pain spine Care Visits Imaging Clinics (2011) exams Kaiser PNCA 17 1096 2,430,000 149,300 44790 Henry Ford 26 230 187,000 23,900 7170 Group Health 24 303 347,000 37,700 11310 Mayo Clinic 61 269 1,500,000 106,700 32010 Total 128 1,898 4,464,000 317,600 95,280

LIRE Aims/Working Groups and Leaders 1. Refinement of benchmark text Jerry Jarvik, MD MPH 2. Implementation of cluster randomization Bryan Comstock, MS 3. Spine intervention intensity measure Brian Bresnahan, PhD 4. Electronic data capture Nick Anderson, PhD 5. IRB, Protocols, Subcontracts Katie James, PA, MPH

LIRE Aims/Working Groups and Leaders 1. Refinement of benchmark text Jerry Jarvik, MD MPH 2. Implementation of cluster randomization Bryan Comstock, MS 3. Spine intervention intensity measure Brian Bresnahan, PhD 4. Electronic data capture Nick Anderson, PhD 5. IRB, Protocols, Subcontracts Katie James, PA, MPH

LIRE Aim 3 • Develop/validate a composite measure of spine intervention intensity-a single metric of overall intensity of resource utilization for spine care

Aim 3 Progress • Working with site programmers to pull CPT data • Already established data pulls for 2 sites • Constructed density plots of CPT –For QC checks –Compare site use of codes

BOLD: CPT Code Frequencies By Site Site 1 Site 2 Site 3

BOLD: Density Plot of Radiology CPT Code Frequencies By Site for QC Site 1 Site 2 Site 3 �

BOLD: Density Plot of Surgery CPT Code Frequencies By Site Site 1 Site 2 Site 3

Aim 3 (cont.) • Converted CPT codes to RVUs as our primary metric of back-related utilization –Used total RVU (tech + pro) –Did not use geographic adjuster –Use 2012 values using CMS look-up files

Converting CPTs to RVUs • Validate CPT conversion by directly pulling RVUs from one site

Example RVU Values HCPCS DESCRIPTION TOTAL 2012 RVUs 72100 X-ray exam of 1.07 lower spine 99214 Office/outpatient 2.26 visit level 4 72131 CT lumbar spine 6.27 w/o dye 72148 MRI lumbar spine 11.31 w/o dye 63047 Laminectomy 32.89

CPT Proportion of RVUs 12.00% Office Site 1 level IV 10.00% 8.00% Subsequent Initial hospital care hospital 6.00% Office care level III Screening 4.00% mammo MR Office ED visit dig Lspine level V 2.00% 0.00% 99214 99233 99223 99213 99285 72148 99215 G0202

Challenges • CPT counts seem to differ by site –Step wedge design helps to address this since before-after comparison is within site –Using only back-related RVUs improves accuracy/reliability � using algorithm developed by Martin et al at Dartmouth • Different pharmacy data systems (e.g. not all sites have Rx filled data) –Within-system comparisons will be valid

Challenges • System differences will always be present in large pragmatic trials • When do pragmatic trials become meta- analysis of parallel trials?

Key People to Thank UW Non-UW • Katie James, PA-C, MPH-Proj • Rick Deyo, MD, MPH-OHSU Dir • Dan Cherkin, PhD-GHRI • Bryan Comstock, MS- Biostats • Heidi Berthoud. MPH- GHRI • Nick Anderson, PhD- • Safwan Halabi, MD-HFHS Biomedical Informatics • Dave Nerenz, PhD- HFHS • Brian Bresnahan, PhD- Health • Dave Kallmes, MD- Mayo Economist • Jyoti Pathak, PhD- Mayo • Patrick Heagerty, PhD- Biostat • Patrick Luetmer, MD- Mayo • Judy Turner, PhD- • Andy Avins, MD MPH-KPNC Psychologist/Pain expert