Potential Impact of Implementing Research Findings: Modeling Approach - - PowerPoint PPT Presentation
Potential Impact of Implementing Research Findings: Modeling Approach Implementing PCOR Results to Inform Choices and ANIRBAN BASU Improve Healthcare Delivery basua@uw.edu Monday, June 25 th @basucally Academy Health, Seattle 2018
ANIRBAN BASU basua@uw.edu @basucally
Implementing PCOR Results to Inform Choices and Improve Healthcare Delivery Monday, June 25th Academy Health, Seattle 2018
Joint work by Blythe Adamson, Ph.D. & Anirban Basu, in collaboration with PCORI research staff, Joanna Siegel and William Lawrence Funded by PCORI Work carried out under Salutis Consulting LLC
> Common modeling framework across multiple disease conditions and interventions > Impact estimates for different stakeholders > Not a conversation about value… but enumerating costs and consequences
1. ANTIBIOTIC STUDY (Keren et al. JAMA Ped 2015): Antibiotics administered intravenously via a peripherally inserted central venous catheter (PICC) or orally for post-discharge treatment of children with acute osteomyelitis or complicated pneumonia using 2. STROKE STUDY (Xian et al. BMJ 2015): Warfarin versus to anticoagulation among anticoagulation naïve elderly acute ischemic stroke patients, who are discharged alive with documented persistent
3. DIABETES STUDY (Young et al. JAMA Int Med 2017) Use of self- monitor blood glucose (SMBG) for non-insulin-treated patients with type 2 diabetes 4. CHEST PAIN STUDY (Hess et al. BMJ 2016): Shared decision-making using a decision aid with usual care in the choice of admission for
evaluation in patients with possible acute coronary syndrome
Year 1 Year 2 Year 3 Year 4 Year 5 Cohort 5 Cohort 4 Cohort 3 Cohort 2 Incident Cohort 1
Outcome X Treatment 1 % currently used No X Current Practice Cohort Outcome X Treatment 2 % currently used No X Outcome X Treatment 1 Implementation Model 100% No X Outcome X Treatment 2 0% No X
Time horizon of 5 years (no discounting) Outcomes limited to length of study follow‐up Immediate, complete uptake
findings Patient‐ centered
US population for whom comparative‐effectiveness results are generalizable
*Focus of impacts on overall health sector, private payer, public payer and patients
A COMMON MODEL
Year 1 Year 2 Year 3 Year 4 Year 5 Cohort 5 Cohort 4 Cohort 3 Cohort 2 Incident Cohort 1
Outcome X Treatment 1 % currently used No X Current Practice Cohort Outcome X Treatment 2 % currently used No X Outcome X Treatment 1 Implementation Model 100% No X Outcome X Treatment 2 0% No X
Time horizon of 5 years (no discounting) Outcomes limited to length of study follow‐up Immediate, complete uptake
findings Patient‐ centered
US population for whom comparative‐effectiveness results are generalizable
*Focus of impacts on overall health sector, private payer, public payer and patients
A COMMON MODEL
Acute Osteomyelitis Complicated Pneumonia Number of children in the United States
73,673,073 73,673,073
Annual incidence of hospitalization, per 100,000 children
13 20
Percentage of hospitalized population the findings would be applicable to based on study inclusion/exclusion criteria
43% 59%
Annual number of index hospitalizations
4,090 8,857
Total number of children impacted over 5 years
20,448 44,283
Persons in the United States ≥65 years (n = 40,267,984) Who experience non‐valvular acute ischemic stroke each year (n = 299,270) Stroke patients with atrial fibrillation (n = 77,536)
Cohort* in Analysis
(n=19,759)
Type 2 Diabetes patients in the United States (n = 26,338,269) Age ≥30 years and Non‐insulin treated
Using insulin or no medication (43%) Excluded from study (531/1032 = 51.4%)
Impact Model Cohort of Non‐Insulin Treated Type 2 Diabetes Patients Using SMBG (n = 7,091,977) Eligible for study
Not using SMBG (112/450 = 24.9%)
ED visits each year for the principle reason of chest pain and related symptoms (not referable to body systems)
6,642,000 visits x (1 – 71.8% excluded)
Patients not excluded from study participation
= x
5 years
Time horizon years
9,369,818
Patients with ED visit for chest pain that could potentially benefit from decision aid in 5 years
x = x
Hess 2016
11
stress test Admit to observation unit
74%
52.1%
no stress test
26%
Current Practice Outpatient follow‐up stress test visit within 30 days
17% 98.8%
no stress test Do not admit
83%
ED visit for chest pain
47.9%
at low risk for acute No follow‐up visit within 30 days coronary syndrome
1.2%
stress test Admit to observation unit
74%
37.3%
no stress test Implementation
26%
Model: Using Decision Aid Outpatient follow‐up stress test visit within 30 days
17% 98.8%
no stress test Do not admit
83%
62.7%
No follow‐up within 30 days
1.2%
*Between study arms, there was no difference in stress test type, CCTA performed within 30 days, coronary revascularization, admittance to hospital from ED observation unit, repeated ED visit, readmission to the hospital, outpatient clinic difference, or cardiac events. These outcomes were not included in the final model because they are not changed with use of the decision aid.
Limitation: There was some (8%) loss to follow‐up in the study; of 70 people lost, 68 were confirmed alive at 45 days (Hess 2016). Slightly more people were lost‐to‐ follow up under the “Using Decision Aid” arm as fewer were admitted to hospital.
Model Inputs
12
Parameter Value (Range) Source
Number of ED visits for the principal reason of chest pain and related symptoms (not related to body systems), total per year 6,642,000 (5,175,920 ‐ 8,108,080)
Calculated from Rui et al. 2014 Table 10
Excluded from participation in the study* 71.8%
Hess et al. 2016
2323/3236
Probability of admission for observation with usual care 52.1% (49.8‐54.4%)
Hess et al. 2016
225/447
Probability of outpatient visit within 30 days if not admitted with usual care 98.8% (98.7‐98.9%)
Calculated from Hess et al. 2016
(203+275)/(52+100+55+101+138+38)
Probability of stress test if admitted to hospital 74.1% (72.2‐76.0%)
Calculated from Hess et al. 2016
(169+120)/(225+165)
Probability of stress test at outpatient follow‐up visit 17.0% (15.6‐18.3%)
Calculated from Hess et al. 2016
35/(55+100+52)
Risk difference for admission using decision aid 14.8% (8.4‐21.2%)
Hess et al. 2016
*Patients were excluded from randomization to usual care or the decision aid for several reasons. The most common reasons were (1) known coronary artery disease, (2) acute ischemia on initial electrocardiogram (ECG), and (3) they were unable to use decision aid (for example a learning barrier or dementia). ED = emergency department
Population Size Calculation
13
ED visits each year for the principle reason of chest pain and related symptoms (not referable to body systems)
6,642,000 visits x (1 – 71.8% excluded)
Patients not excluded from study participation
Rui et al (2014)
5 years
Time horizon years
9,369,818
Patients with ED visit for chest pain that could potentially benefit from decision aid in 5 years
Hess et al (2016)
ED = emergency department
Cost Inputs
14
Cost Value (Range) Source
Hospital admission for observation and testing Cost of hospital admission for
$3,313 (3,136‐3,490)
Mallidi et al. 2013
Cost of inpatient stress testing $500 (SD = 26.75, 448‐1,867)
Assumed near the upper range of outpatient stress testing cost (below), with probabilistic draws based on standard deviation. One‐way sensitivity analysis uses upper range of $1,867 from Mallidi et al. 2013.
Outpatient follow‐up visit with cardiologist or primary care provider Cost of outpatient follow‐up visit $198 (186‐210) MEPS 2014 Average cost of outpatient stress test
General stress testing Echo stress testing Nuclear stress testing
$244 (234‐255)
$77.52 $274.51 $494.55
CMS Physician Fee Schedule Search, weighted by frequency of test type
The cost of hospital admission and cost of outpatient visit exclude the cost of the stress test. Some patients who are admitted for stress testing do not receive a stress test and some patients who follow‐up within 30 days do not receive a stress test at the follow‐ up visit.
$0 $2,000 $4,000 $6,000 $8,000 $10,000 $12,000 $14,000 $16,000 $18,000
Total Cost (Millions $)
Private Public Out‐of‐Pocket
Usual Care Using Decision Aid
Outpatient Inpatient
Millions $ Savings $4,100 $608 $133
$0 $2,000 $4,000 $6,000 $8,000 $10,000 $12,000 $14,000 $16,000 $18,000
Total Cost (Millions $)
Private Public Out‐of‐Pocket
Usual Care Using Decision Aid
Outpatient Inpatient
Millions $ Savings $4,100 $608 $133
also increases the number of patients we expect will not follow‐up within 30 days.
chest pain) would have loss to follow‐up using the decision aid and 2,922
Outpatient Follow-Up in 30-days
18
Not admitted and no follow‐up within 30‐days Expected number of stress tests “missed” via loss to follow‐up
Current Practice 55,638 9,458 Implementation Model 72,829 12,380 Incremental 17,191 2,922
Xian et al. BMJ (2015)
Young et al. JAMA Internal Medicine (2017)
1000 2000 3000 4000 5000 ED Visits Rehospitalization
Number of Cihildren
2 4 6
Millions of Patients
Hospital Admission Outpatient Visit 10 20 30 40 Deaths Strokes
Thousands of Patients
2,000 4,000 6,000 8,000 10,000
Millions of SMBG Tests
Current Practice Implementation Model
A) Antibiotics B) Chest Pain C) AF Strokes D) Diabetes
3,080 ED visits avoided 2,032 hospitalizatins avoided 1,494 deaths from all cuases within 24 months avoided 1,391,349 hospital admissions avoided 1,374,096 additional
466 strokes avoided 10 Billion finger sticks avoided
Models publicly available: http://www.salutisllc.com/ portfolio.html# Locked Excel files or R‐shiny Manipulation of input parameter allowed
> Potential impact for implementation of research findings from many clinical studies can be estimated following common features in a simulation model > The population and patient-level impact and the distribution of impacts across stakeholders highlight the value of research and the potential for savings and benefits within and outside of the health care system > Potential expenditures avoided from perfect implementation of findings from four PCORI studies ranged from $ 7.1 million to $12 billion over a 5 year period.
Thank you
Results for Usual Care
29
Current Practice
Mean Credible Range ED Visits for Low‐Risk Chest Pain 9,366,301 (8,220,654 ‐ 10,473,772) Admitted to Observation Unit 4,874,757 (4,209,735 ‐ 5,512,097) Outpatient follow‐up within 30 days 4,435,972 (3,844,990 ‐ 5,004,545) Total Stress Tests 4,363,906 (3,825,275 ‐ 4,920,548) Inpatient Stress Tests 3,610,451 (3,146,849 ‐ 4,086,345) Outpatient Stress Tests 753,455 (645,451 ‐ 868,489) Total Costs ($) $19,028,267,777 ($16,490,288,227 ‐ $21,517,444,255) Inpatient Admission and Testing Costs $17,962,997,302 ($15,553,338,124 ‐ $20,365,886,425) Outpatient Visit and Testing Costs $1,065,270,475 ($917,279,789 ‐ $1,213,498,391)
Results for Intervention
30
Implementation Model: Using Decision Aid
Mean Credible Range ED Visits for Low‐Risk Chest Pain 9,366,301 (8,220,654 ‐ 10,473,772) Admitted to Observation Unit 3,470,686 (2,739,503 ‐ 4,247,010) Outpatient follow‐up within 30 days 5,822,674 (4,938,273 ‐ 6,783,162) Total Stress Tests 3,559,512 (3,012,090 ‐ 4,144,170) Inpatient Stress Tests 2,570,486 (2,013,695 ‐ 3,125,334) Outpatient Stress Tests 989,026 (826,194 ‐ 1,168,787) Total Costs ($) $14,187,760,172 ($11,369,925,418 ‐ $17,079,572,808) Inpatient Admission and Testing Costs $12,789,436,531 ($9,981,416,667 ‐ $15,686,209,863) Outpatient Visit and Testing Costs $1,398,323,642 ($1,168,842,847 ‐ $1,646,638,202)
Costs by Payer Type
31
Cost (Millions $)
Payer Type Total Public Private Out‐of‐Pocket Cost Usual Care Costs $15,889 $2,474 $666 $19,028 Inpatient Costs $14,999 $2,335 $629 $17,963 Outpatient Costs $890 $138 $37 $1,065 Using Decision Aid $11,847 $1,844 $497 $14,188 Inpatient Costs $10,679 $1,663 $448 $12,789 Outpatient Costs $1,168 $182 $49 $1,398 Total Incremental Costs ‐$4,100 ‐$608 ‐$133 ‐$4,841
‐$4,320 ‐$673 ‐$181 ‐$5,174
$220 $65 $48 $333
Impact on Cost Per Patient
Payer Type Total Public Private Out‐of‐Pocket Cost Incremental Cost Per Patient ‐$438 ‐$65 ‐$14 ‐$517
Sensitivity Analysis
32