Geographic Factors Associated with Emergency Department Super Utilizers

Transcription

1 Geographic Factors Associated with Emergency Department Super Utilizers Michael Horst, PhD, MPHS, MS 1 ; Alison Lauter 1 ; Jeffrey Martin, MD 2 ; Angela Gambler, MBA 1 ; Andrew Coco, MD, MPH 1,2 1 Lancaster General Research Institute, Lancaster, PA; 2 Lancaster General Family Medicine Residency, Lancaster, PA

2 Jeff Brenner, MD Executive Director, Camden Coalition of Healthcare Providers Faculty, RWJ Medical School

3 Medical Report The Hot Spotters Can we lower medical costs by giving the neediest patients better care? by Atul Gawande January 24, 2011

4 Factors Associated with ED Frequent Users: From Previous Studies Demographics Age Gender Race/Ethnicity Socio Economic Education Insurance/payment type Poverty Single parent Homeless Visit Type Higher visit acuity Non injuries/trauma Acute exacerbations of chronic conditions Health Status Mental health issues or classified as poor mental health Substance abuse Classified as poor health Psycho social issues Unmet health needs Health Service Utilization Previous/frequent ED use Previous hospitalization Frequent office/primary care clinic visits

5 Defining High Utilizers For this study >= 90 th percentile in ED or IP visits From the literature. Top percentiles (visits and or charges) in ED (4 12 annual visits) Inpatient Primary care Outpatient tests/procedures EDR (ED/ED+PC) Top utilizers representing 25% of visits or charges Calculation of observed and expected utilization: those exceeding expected

6 Purpose & Methods Purpose Identify high utilizers (PC, ED, inpatient) Patients linked to primary care offices Determine predictors of high ED and IP utilization Assess comorbidities, mental health diagnoses and time/date of ED visit Methods Pilot: Visits occurring in CY 2008 and 2009 Phase 2: Visits occurring in CY 2008, 2009 and 2010 Linking PC to ED and inpatient (identifying unique individuals) Basic demographics Calculation of top utilizers Determine predictors of high utilization: binary logistic (ED cohort) and binary random intercept (Primary Care cohort) model with practice as the random intercept

7 Study Area and Sites Rationale for defining target area: Horst M, Coco A. Observing the spread of common illnesses through a community: Using Geographic Information Systems (GIS) for surveillance. J Am Board Fam Med 2010;23:32 41

8 Study Overview

9 What did we learn from the pilot? Factors Associated with High ED Utilizers (90 th Percentile) that are affiliated with a primary care practice Age Number of primary care visits Number of inpatient visits Payment types Travel time to LG ED Travel time to non LG EDs!(

10 Target Area Primary Care (geocoded to street or rooftop), Inpatient and ED Visits from 1/1/ /31/2009 Number of Unique Individuals Total Mean SD Min Median Max Primary Care 151,898 Visits 1,370, ED 27,798(18.3%) Visits 52, Inpatient 21,179(13.9%) Visits 28, Mean, SD, Min, Median and Max are calculated per unique individual over the 2 year study time frame. For ED and Inpatient, it is calculated only for those having visits in those areas. Top 10 th Percentile Utilizers Criteria for Top 10 th Percentile Utilizer Number of Unique Individuals Number of Visits % of Total Visits Primary Care > 19 Visits 15, , % ED > 3 Visits 2,804 18, % Inpatient > 1 Visit 4,337 11, %

11

12

13 Phase 2A

14 Target Area Primary Care Cohort (geocoded to street or rooftop), Inpatient and ED Visits from 1/1/ /31/2010 Number of Unique Individuals Total Mean SD Min Median Max Primary Care 167,005 Visits 2,005, ED 51,728(31.0%) Visits 131, Inpatient 32,596(19.5%) Visits 50, Mean, SD, Min, Median and Max are calculated per unique individual over the 3 year study time frame. For ED and Inpatient, it is calculated only for those having visits in those areas. Top 10 th Percentile Utilizers Criteria for Top 10 th Number of Unique Number of % of Total Percentile Utilizer Individuals ED Visits ED Visits ED > 4 Visits 6,430 56, %

15 Cohort of Primary Care: ED SU Rate (%)

16 Predictors of ED super utilizer status in a cohort of primary care subjects (binary logistic random intercept model: primary care practice). Variable Odds Ratio (95% CI) P value Gender (ref = female) 1.0 ( ) Comorbidity (any Charlson/Elixhauser encounter) 3.2 ( ) <0.001 Mental Health Diagnosis 3.1 ( ) <0.001 Primary Care Visits 4 16 (ref = 1 3) 0.9 ( ) Primary Care Visits >16 (ref = 1 3) 1.4 ( ) <0.001 Inpatient Visits = 1 (ref = 0) 1.7 ( ) <0.001 Inpatient Visits >1 (ref = 0) 8.7 ( ) <0.001 Age 0 17 (ref = 65+) 4.4 ( ) <0.001 Age (ref = 65+) 3.4 ( ) <0.001 Age (ref = 65+) 1.8 ( ) <0.001 Travel Time 5 min to LG or other (ref = > 5 min both) 1.3 ( ) <0.001 Travel Time 5 min both (ref = > 5 min both) 3.9 ( ) <0.001 Payer MA (ref = commercial) 1.3 ( ) <0.001 Payer Medicare (ref = commercial) 1.3 ( ) <0.001 Payer Other (ref = commercial) 2.8 ( ) <0.001

17

18 Phase 2B

19 Target Area ED Cohort (geocoded to street or rooftop), Inpatient and ED Visits from 1/1/ /31/2010 Number of Unique Individuals Total Mean SD Min Median Max Primary Care 52,873 (42.5%) Visits 856, ED 124,279 Visits 276, Inpatient 43,178 (34.7%) Visits 73, Mean, SD, Min, Median and Max are calculated per unique individual over the 3 year study time frame. For Primary Care and Inpatient, it is calculated only for those having visits in those areas. Top 10 th Percentile Utilizers Criteria for Top 10 th Number of Unique Number of % of Total Percentile Utilizer Individuals ED Visits ED Visits ED > 4 Visits 11,823 98, %

20 ED Cohort: ED SU Rate (%)

21 Predictors of ED super utilizer status in a cohort of all subjects who visited the ED (binary logistic model). Variable Odds Ratio (95% CI) P value Gender (ref = female) 0.8 ( ) <0.001 Comorbidity (ED Charlson/Elixhauser) 3.9 ( ) <0.001 Mental Health Diagnosis 3.3 ( ) <0.001 Part of Primary Care Network 1.6 ( ) <0.001 Inpatient Visit (within health system) 1.7 ( ) <0.001 Age 0 17 (ref = 65+) 3.3 ( ) <0.001 Age (ref = 65+) 2.8 ( ) <0.001 Age (ref = 65+) 2.2 ( ) <0.001 Travel Time 5 min to LG or other (ref = > 5 min both) 1.1 ( ) Travel Time 5 min both (ref = > 5 min both) 2.2 ( ) <0.001 Payer MA (ref = commercial) 4.4 ( ) <0.001 Payer Medicare (ref = commercial) 4.1 ( ) <0.001 Payer Other (ref = commercial) 1.1 ( ) No Family Doctor 2.9 ( ) <0.001

22 Percent No Family Doctor by Census Tract

23

24 Time, Day and Means of Transport to ED Percent of Visits Non Super Utilizers Super Utilizers p Weekday (vs. Weekend) 70.3% 72.6% <0.001 Day (vs. Night) 47.6% 48.0% During Normal Primary Care Hours 50.6% 52.1% <0.001 Emergency Vehicle Transport 23.3% 21.1% <0.001

25 Difference in Super Utilizer Rate of Primary Care Hour Visits to ED by Census Tract (Super Utilizer Rate Non Super Utilizer Rate)

26 Conclusions Phase 2A (cohort of all primary care patients) Comorbidities, mental health diagnosis, primary care/inpatient visits, age, payer status and travel time A lot of variability in super utilizer rates across practices Concentration of super utilizers in area around ED and urban community to the west Phase 2B (cohort of all ED patients) Female, comorbidities, mental health diagnosis, par of primary care network, inpatient visits, age, payer status, travel time and no family doctor Census tracts with high distress and higher levels of non white Does not seem to be a large impact on time of visit to ED or transport means Concentration of super utilizers in area around ED and urban community to the west

27