Advancing the Evidence and Innovation Agenda

Transcription

1 Advancing the Evidence and Innovation Agenda Jon Baron, Coalition for Evidence-Based Policy Dave Patterson, South Carolina Budget and Control Board, Division of Research Statistics John Laub, University of Maryland Betsey Stevenson, White House Council of Economic Advisors

2 Integrated Data Systems, Program Delivery, and Quasi-experimental Program Evaluation W. David Patterson, Ph.D. AISP Annual Conference Washington DC November 5,

3 THE PURPOSES OF AN INTEGRATED DATA SYSTEM 3

4 Elder Services & Assessments Disabilities & Special Needs Vocational Rehabilitation Department of Commerce LEGEND Homeless MIS Law Enforcement Legal/Safety Services Disease Registries Juvenile Justice Social Services Education Public Safety Claims Systems Health Department Environmental Conditions Alcohol & Drug Services Probation, Pardon & Parole Corrections Child Care All Payer Health Care Databases Health Professions Behavioral Health Mental Health Health Professions Free Clinic Visits Hospitalizations Emergency Room Visits Outpatient Surgeries Home Health Care Social Services Medicare Medicaid Services State Employee Health Services Community Health Centers Health Department Education Disease Registries Other State Agencies

5 Roles of the Data Warehouse in the SC Telepsychiatry Initiative Program Operation Program Evaluation Provision of Medicaid Data Provision of DMH data into SC Health Information Exchange (SCHIEx) Integration of DMH Electronic Medical Record system with SCHIEx Linkage of program specific data into Integrated System Provision of additional linked elements from the data warehouse, most notably UB 92/04 Statistical and analytic support

6 PROGRAM SUPPORT: AN EXAMPLE 6

7 Primary Goals Timely Psychiatric Assessment and Rapid Initiation of Treatment Increased Quality of Care Reduced Lengths of Stay (LOS) Comprehensive Discharge Planning Savings to the Hospital and Community Last Update: 12/07/2012 3

8 Nov 2007 TDE 1 st Award Benchmarks May 2009 Presented at APA Convention Aug 2009 EMR Operational Jun 2010 Computerworld Finalist; TDE 2 nd Award Oct 2011 Silver Award, APA Convention Feb ,000 th Consult May 2012 Presented at APA Convention Jun 2012 TDE 3 rd Award Oct 2012 SCORH Outstanding Rural Program of the Year Apr ,000 th Consult Last Update: 04/18/2013 4

9 Clinical Office Locations Columbia, SC (3) Charleston, SC (2) Aiken, SC (2) Future Site(s) Last Update: 04/18/2013 8

10 Consultation Process Patient Presents in ED ED Physician Requests Consult Psychiatrist Reviews CIS/SCHIEx, EMR Patient Consulted Video Encounter Ends Psychiatrist Electronically Signs Consult Recommendations Sent to ED Hospital Dispositions the Patient Last Update: 04/18/2013 9

11 SPARTANBURG UPSTATE CAROLINA WALLACE THOMSON BAPTIST EASLEY CHESTER SPRINGS CLARENDON CAROLINAS MARION CAROLINAS MCLEOD DILLON LORIS LAURENS FAIRFIELD used with permission from Ralph Strickland, DMH TELEPSYCHIATRY CONSULTATION PROGRAM SEACOAST CONWAY WACCAMAW GEORGETOWN Legend Current Hospitals (18 ACTIVE) Telepsychiatry Clinical Offices HAMPTON RURAL / URBAN SHADING DESIGNATED BY US GOV T SOURCE: RNS42@SCDMH.ORG file present int dmh affairs telepsych power map revised.ppt LAST UPDATE 07/02/2013

12 EVALUATION SUPPORT: A RELATED EXAMPLE 12

13 Strategy Propensity scoring with optimal matching used to match patients treated at intervention EDs to those treated at non-intervention EDs in South Carolina Compared two groups on utilization and cost outcomes using standard econometric techniques Narasimhan, Druss et al NIMH and NIH R01

14 Matching criteria Patients matched on: Age Sex Race Diagnosis (psychiatric) Timing of ED visit (1-month window, weekday versus weekend) Narasimhan, Druss et al NIMH and NIH R01

15 Utilization Measures -Differences in odds of admission from ED -Differences in follow up at 30 and 90 days. Narasimhan, Druss et al NIMH and NIH R01

16 Baseline Characteristics Telepsychiatry Control N=7,261 N=7,261 Age Female 49.8% 49.8% White 73.1 % 73.1 % Black 23.8% 23.8% Weekend Admission 38.6% 38.6% Narasimhan, Druss et al NIMH and NIH R01

17 Service Use Telepsychiatry Control P N=7,261 N=7,261 Admission 22% 11% <0.001 LOS at index visit (in <0.001 days) day OP f/u 46% 16% <0.001 < day OP f/u 54% 20% <0.001 Index 30 day IP cost $8,290 $11,224 Index 30 day hospital cost (IP+ED)* $12,634 $14, Narasimhan, Druss et al NIMH and NIH R01

18 Utilization models Two-part model to predict admission and then LOS conditional upon being admitted. 30 and 90 day f/u were examined using logistic regression with hospital random effects. Narasimhan, Druss et al NIMH and NIH R01

19 Utilization results Utilization Measure Effect Size N=14,522 Odds of admission 0.41 (0.022) LOS in days (0.003) Combined effect in days (<0.001) Odds of OP f/u within 30 days 5.48 (<0.001) Odds of OP f/u within 90 days 5.68 (<0.001) P-values in parentheses are derived from robust standard errors clustered at the hospital level (61 clusters); all models include hospital random effects and are adjusted for weekend versus weekday visit, sex, age, and race. Narasimhan, Druss et al NIMH and NIH R01

20 Cost results Effects of being treated using telepsychiatry: Cost N=16, day inpatient costs -$2,336 (0.041) 30 day hospital costs -$831 (0.523) 30 day total costs -$649 (0.619) P-values in parentheses are derived from robust standard errors clustered at the hospital level (61 clusters); all models include hospital random effects and are adjusted for weekend versus weekday visit, sex, age, and race.

21 Limitations Unmeasured patient-level differences patients in telepsychiatry program could be sicker or more complex at baseline than controls Unmeasured hospital-level differences telepsychiatry hospitals are less likely to have psychiatry expertise or inpatient services

22 Next Steps: Quality, Economic Outcomes and Sustainability of Telepsychiatry R01MH Create a synthetic control group of hospitals from surrounding states and examine outcomes for patients treated there relative to SC hospitals before and after telepsychiatry (diff in diff/triple difference). Examine effects on disease specific quality measures (depression, bipolar d/o, schizophrenia) Budget impact analysis: costs from the managerial insurance, and societal perspectives

23 LESSONS? 23

24 What can we learn from this effort? Integrated data systems can close the loop between practitioners, applied analysts, and basic researchers Integrated data systems can help create and sustain public, private, and not for profit partnerships around issues Integrated data systems reduce costs by repurposing existing data, producing economies of scale