RE-ADMITTING IN HOSPITALS: MODELS AND CHALLENGES. Murali Parthasarathy Dr. Paul Damien

Transcription

1 RE-ADMITTING IN HOSPITALS: MODELS AND CHALLENGES Murali Parthasarathy Dr. Paul Damien April 11,

2 Major pain points Hospitals scored on five major pain points 1. Death rates among heart and surgery patients 2. Readmission (an event for which hospitals are now subject to penalties by the CMS) 3. Overuse of CT Scans 4. Incidence of Hospital acquired infections 5. Effective communication to patients about medications and discharge plans 2

3 Four states were selected to look at healthcare quality and readmission rates We analyzed four states and hospitals across them to : Identify major cost utilization for IP and OP hospitalization services Identify initiatives to improve QOC in hospitals QOC analysis arrived at significant measures to improve process of care, ED, surgical care, outcome of care, 30 day hospital mortality rate, healthcare associated infections and measures to identify areas of minimum and maximum care Illinois healthcare quality is expected to lie between TX and MA Massachusetts became the first state in the country to start healthcare reform for state residents Texas health care quality was rated the worst in the nation in the federal government s annual (2011) nationwide health-care report card TX CA California is one among the states with innovative ideas that improve quality, increase efficiency, and lower the costs of care 3

4 Avoidable hospital readmissions was significant with room for improvement and saving healthcare dollars THE CHALLENGE One in five patients is readmitted to the hospital within 30 days of discharge $ 2.5 Trillion spent in US on healthcare in % ($ 765B)of total spending went on waste Avoidable readmissi on 3% $740 B Waste 97% Avoidable readmissions cost Medicare $25 billion per year SOLUTION In a growing number of best practice studies on avoiding preventable readmissions, one of the principal calls-toaction is the utilization of technology to risk screen patients Total hospital Readmissions could be Reduced by up to 12% by Improving procedures and utilizing health Information technology 4

5 Understanding the problem - Help hospital administrators identify patients who might benefit from specific interventions STEP 1 : DATA BASE STEP 2 : MODELING & VALIDATION STEP 3 : MOBILE APP ROBUST DATABASE WITH DIFFERENT VARIABLES IDENTIFY AND VALIDATE SIGNIFICANT VARIABLES, MEASURE QUALITY OF HOSPITAL CARE RELATIVE TO ED VISIT AND TREATMENT OUTCOME APP HELPS IDENTIFY PATIENTS FOR FOLLOW-UPS AND RISK FACTORS DEMOGRAPHIC DATA DISEASE CONDITION PRESCRIPTION DATA IN-PATIENT/OUT- PATIENT COST LAB RESULTS CREATE MODELS THAT IDENTIFY SIGNIFICANT VARIABLES TRIAGE PATIENTS BASED ON ED DIAGNOSIS AND SEVERITY IDENTIFY DISEASES FOR WHICH PATIENT IS AT RISK MEASURE QUALITY OF HOSPITAL CARE RELATIVE TO TREATMENT OUTCOME AND FOLLOW UP VISITS CLOUD BASED MOBILE APP PATIENT X VISITED ED X MONTHS AGO AGE GENDER CONDITION HOSPITAL DATA 5

6 Potential relationship between process of care and excess readmissions was analyzed Timely and Effective Care Outcome Measures As seen by analyzing the CMS data from 2008 to 2010, one out of five patients admitted for treatment for HF, AMI and Pneumonia are likely to be readmitted within 30 days of initial treatment The potential impact of quality of POC in hospitals across four states on readmission was explored AMI, HF, PN Preventive Care Emergency Department Care Surgical Care Hospital 30-days Mortality Rate Hospital Associated Infections Readmission Rate 6

7 On comparing process of care in Texas with national average, care was seen to be better for AMI but lower for HF and PN Process of Care AMI, HF and PN 96% % TX Average National Average AMI 86% 84% HF 94% 95% PN 95% 96% Among POC measures for AMI, HF and PN, it was seen that poorest quality of care was given to Heart Failure patients 7

8 Analysis of readmission rates in Texas 39% of hospitals (99) are subject to penalty 24% of patients admitted for heart failure, 20% for heart attack and 17% for Pneumonia readmitted within 30 days Diseases AMI HF PN Total Hospitals 257 Sum of Discharges Excess Readmission Ratio <= hospitals Excess Readmission Ratio > 1 99 hospitals Avg % Readmission 20% 24% 17% Avg Discharges / Readmission Excess Readmission Ratio Rate of readmission is higher for HF compared to AMI and PN Process of care is lowest among HF patients, subsequently increasing rate of readmissions Datasource : Hospital Compare

9 Among high penalty hospitals, Dimmit Regional hospital provides the least care for HF Comparison of care among top 10 penalty hospitals : HF- Discharge instruction TX average : 94% Dimmit Regional hospital provides 39% of effective care for process of care Heart Failure patients given discharge instructions- 9

10 Among high penalty hospitals, Dimmit Regional hospital provides the least care for HF Number of discharges Vs Percentage of Readmission among top 10 penalty hospitals Heart Failure TX average : 24% 10

11 TEXAS Hospitals Presence Demographic and Economic impact DESCRIPTION TEXAS CEDAR PARK REGIONAL MEDICAL CENTER CLEVELAND REGIONAL MEDICAL CENTER DALLAS REGIONAL MEDICAL CENTER DIMMIT REGIONAL HOSPITAL GOOD SHEPHERD MEDICAL CENTER ETMC HENDERSON TYPE OF HOSPITAL For Profit Hospital For Profit Hospital For Profit Hospital Public Hospital Non Profit Hospital Non Profit Hospital COUNTY 254 WILLIAMSON LIBERTY DALLAS DIMMIT GREGG RUSK POPULATION, ,448, ,359 76,349 2,453,907 10, ,741 54,013 % RURAL 12% 12% 63% 1% 39% 13% 27% PERSONS >=65 YEARS % 9.8% 12.1% 9.2% 14.6% 13.7% 14.6% BELOW POVERTY LEVEL, %, % 6.8% 17.0% 18.8% 27.0% 16.9% 15.3% PER CAPITA MONEY INCOME IN PAST 12 MONTHS (2012 DOLLARS), $25,809 $30,540 $20,114 $26,576 $15,995 $23,968 $21,696 # HOSPITAL

12 TX is seen to have lower readmission rates despite lower POC; POC is a low impacting measure on readmissions Quality Measure AMI, HF & PN TX IL MA CA National Avg Average AMI 86% 97% 94% 91% 84% Process of Care Readmission Average Heart Failure 94% 95% 96% 95% 95% Average Pneumonia 95% 95% 97% 96% 96% Average AMI, HF and PN 92% 96% 96% 94% 92% % HF readmitted 24% 26% 25% 24% 24% % AMI readmitted 20% 22% 23% 20% 20% % Pneumonia readmitted 17% 18% 19% 18% 18% Average Readmission 20% 22% 22% 21% 21% # Hospitals # Affected (High Discharges/Readmission) Subject to penalty 99 (39%) 87 (48%) 17 (25%) 325 (35%) 1519(49%) State TX IL MA CA Hospital Name Dimmit Regional hospital Franciscan St James Health Metrowest Medical Center Community Regional Medical Center Other steps that could positively impact readmissions Identify patients at high risk for readmissions and connect them to additional discharge support 12

13 Develop an early warning - Help hospital administrators identify patients who might benefit from specific interventions STEP 1 : DATA BASE STEP 2 : MODELING & VALIDATION STEP 3 : MOBILE APP ROBUST DATABASE WITH DIFFERENT VARIABLES IDENTIFY AND VALIDATE SIGNIFICANT VARIABLES, MEASURE QUALITY OF HOSPITAL CARE RELATIVE TO ED VISIT AND TREATMENT OUTCOME APP HELPS IDENTIFY PATIENTS FOR FOLLOW-UPS AND RISK FACTORS DEMOGRAPHIC DATA DISEASE CONDITION PRESCRIPTION DATA IN-PATIENT/OUT- PATIENT COST LAB RESULTS CREATE MODELS THAT IDENTIFY SIGNIFICANT VARIABLES TRIAGE PATIENTS BASED ON ED DIAGNOSIS AND SEVERITY IDENTIFY DISEASES FOR WHICH PATIENT IS AT RISK MEASURE QUALITY OF HOSPITAL CARE RELATIVE TO TREATMENT OUTCOME AND FOLLOW UP VISITS CLOUD BASED MOBILE APP PATIENT X VISITED ED X MONTHS AGO AGE GENDER CONDITION HOSPITAL DATA 13

14 Estimating the risk factors for hospital readmissions within 30-days of discharge A Bayes Logistic Regression model is constructed to estimate the most significant factors responsible for hospital readmissions within 30-days of discharge in four selected states: CA, TX, MA & IL Summary of Modeling analysis State Significant co-morbid conditions Other significant risk factors CA TX Co-Morbid Conditions Heart Failure, Cancer, COPD, Ischemic Heart Disease, Kidney disease, Stroke Kidney disease, Heart failure, Alzheimer's, COPD, Stroke Other significant factors Alzheimer s Kidney Disease Average # discharges in 2008 Heart Failure Arthritis Age group Heart Disease Osteoporosis Length of hospital stay Stroke Depression # of OP visits COPD Cancer Number of hospital discharges in the previous year and # of Out patient visits Number of hospital discharges in the previous year and # of Out patient visits MA Heart failure, Osteoporosis # of Out patient visits IL Kidney disease, Cancer, COPD, depression, stroke Number of hospital discharges in the previous year and # of Out patient visits A high percentage of Heart failure, Heart disease, Stroke, COPD and Kidney disease patients are at a higher risk for readmissions in the US 14

15 Testing Robustness of the model by Simulation Risk of Readmissions within 30 days of discharge DATA SET Readmission Risk Frequency Percent POPULATION (903)RECORDS) NO YES 44 5 SAMPLE1 (100 RECORDS) NO 89 YES SAMPLE2 (100 RECORDS) NO 93 YES 7 7 SAMPLE3 (100 RECORDS) NO 92 YES 8 8 SAMPLE4 (100 RECORDS) NO 92 YES 8 8 SAMPLE5 (100 RECORDS) NO 85 YES SAMPLE6 (100 RECORDS) NO 91 YES 9 9 SAMPLE7 (100 RECORDS) NO 92 YES 8 8 SAMPLE8(100 RECORDS) NO 88 YES SAMPLE9(100 RECORDS) NO 93 YES 7 7 SAMPLE10(100 RECORDS) NO 90 YES Percentage of risk for hospital readmissions is found to be 5% in the population of 903 patients from Texas in Robustness of the model is tested by selecting ten samples* of size 100 records each from the population and the percentage of risk for readmissions was estimated for each sample (Table) Based on the sample simulated, Risk of readmission rates is found to be between 7% and 15% * Samples are selected from a uniform distribution x following U (0,1) 15

16 Sample report output can be generated for one patient or groups of patients PT_ID DIAGNOSIS PROCEDURES IN 2009 YEAR STATE GENDER RACE AGE_GR PROB READMISSI PROB NOT ON WITHIN WITHIN 30 30_DAYS DAYS RE-ADMISSION RISK FOR C6770 E9E5A Cardiac dysrhythmias Atrial cardioversion 2010 TX FEMALE BLACK 64 & Below 8% 92% LOW 02D7D53671C B0D6B Cardiac dysrhythmias 2010 TX FEMALE WHITE % 96% VERY LOW 055A23F1BBC 9B8E1 Cardiac dysrhythmias Int insert lead in vent 2010 TX MALE WHITE % 93% LOW AEDA 62C89 Acute pulmonary heart disease 2010 TX FEMALE BLACK % 96% VERY LOW 23F6F5EDB28 F44E9 Cardiac dysrhythmias Dx ultrasound-heart 2010 TX MALE WHITE % 92% LOW B BB Cardiac dysrhythmias Atrial cardioversion 2010 TX FEMALE WHITE 24BDB08EFFA 0B90E Cardiac dysrhythmias 2010 TX MALE WHITE 85 & Older 10% 90% LOW 85 & Older 15% 85% HIGH 2F9148D1EB7 6A2F0 Acute pulmonary heart disease 2010 TX MALE WHITE % 95% VERY LOW A 9EFD1 Cardiac dysrhythmias 2010 TX MALE WHITE % 96% VERY LOW A3F Cardiac dysrhythmias Dx ultrasound-heart 2010 TX MALE WHITE % 94% VERY LOW From the risk scored data, the percentage of high risk readmissions within 30 days is found to be 2% 16

17 Across the four states, the percentage of high risk for readmissions within 30 days of discharge is lowest for the state of Texas State Min Score Max Score # Patients % Very Low % Low % High Texas % 50.1% 1.9% Massachusetts % 83.0% 9.5% Illinois % 81.1% 7.0% California % 90.1% 6.1% From the risk scored data, the percentage of high risk for readmissions within 30 days of discharge is found to be 1.9% for Texas, 9.5% for Massachusetts, 7.0% for Illinois and 6.1% for California Estimated the risk factors for all hospital readmissions and scored patients from 2010 for the risk of readmissions 17

18 Estimating the risk factors for Heart Failure readmissions within 30- days of discharge A Bayes Logistic Regression model is constructed to estimate the most significant factors responsible for HF hospital readmissions within 30-days of discharge in four selected states: CA, IL, MA & TX Other significant factors Average # discharges in 2008 Age group Length of hospital stay # of OP visits State CA IL MA TX Summary of Modeling analysis Significant risk factors Number of hospital discharges in the previous year and # of Out patient visits Number of hospital discharges in the previous year and average length of Stay Average number of Out patient visits Number of hospital discharges in the previous year and average length of Stay A high percentage of COPD, Depression Stroke and Chronic Kidney disease patients are at a higher risk for HF beneficiaries readmissions in the US 18

19 Testing Robustness of the model by Simulation for HF readmission risk in Texas Risk of Readmissions within 30 days of discharge DATA SET Readmission Risk Frequency Percent POPULATION (2285 RECORDS) YES NO SAMPLE1 (100 RECORDS) YES NO 90 SAMPLE2 (100 RECORDS) YES 9 9 NO 91 SAMPLE3 (100 RECORDS) YES 8 8 NO 92 SAMPLE4 (100 RECORDS) YES 6 6 NO 94 SAMPLE5 (100 RECORDS) YES 9 9 NO 91 SAMPLE6 (100 RECORDS) YES 5 5 NO 95 SAMPLE7 (100 RECORDS) YES 9 9 NO 91 SAMPLE8(100 RECORDS) YES NO 89 SAMPLE9(100 RECORDS) YES 6 6 NO 94 SAMPLE10(100 RECORDS) YES 8 8 NO Percentage of risk for Heart Failure readmissions is found to be 8% in the population of 2285 patients from Texas in Robustness of the model is tested by selecting ten samples* of size 100 records each from the population and the percentage of risk for Heart Failure readmissions was estimated for each sample (Table) Based on the sample simulated, Risk of Heart Failure readmission rates is found to be between 5% and 11% * Samples are selected from a uniform distribution x following U (0,1) 19

20 The percentage of high risk for readmissions for HF within 30 days of discharge is found to be lowest for the state of Illinois Risk of Readmissions within 30 days of discharge STATE TOTAL POPULATION RISK INTERVAL HIGH RISK PERCENTAGE TEXAS 506 (5,11) 0.6% MASSACHUSETTS 205 (5,15) 0.7% CALIFORNIA 645 ILLINOIS 322 (5,10) 0.7% (5,11) 0.5% From the risk scored data, the percentage of high risk for readmissions within 30 days of discharge is found to be 0.59 % for Texas, 0.68% for Massachusetts, 0.72% for California and 0.49 % for Illinois 20

21 Summary from Modeling 19% of all hospital admissions of beneficiaries with chronic conditions in the US resulted in readmissions within 30 days of discharge, 13% within 15 days and 9% within 7days, in 2008 A high percentage of Heart failure, Heart disease, Stroke, COPD and Kidney disease patients are at a higher risk for readmissions in the US A Bayes Logistic Regression model is constructed to estimate the most significant factors responsible for hospital readmissions within 30-days of discharge in four selected states: CA, TX, MA & IL. Number of out patient visits and hospital discharges in 2008 were found to be significant factors for hospital readmissions within 30 days of discharge in 2009 Robustness of the model is tested by computer simulation by selecting ten samples of size 100 records each from the population and the percentage of risk for readmissions was estimated for each sample. Based on the sample simulated, Risk of readmission rates is found to be between 7% and 15% for Texas, 5% to 12% for Massachusetts, 5% to 15% for California and 5% to 14% for Illinois. The percentage of high risk for readmissions within 30 days of discharge is found to be lowest for the state of Texas and highest for Massachusetts, in

22 Develop app - Helps hospital administrators identify patients who might benefit from specific interventions STEP 1 : DATA BASE STEP 2 : MODELING & VALIDATION STEP 3 : MOBILE APP ROBUST DATABASE WITH DIFFERENT VARIABLES IDENTIFY AND VALIDATE SIGNIFICANT VARIABLES, MEASURE QUALITY OF HOSPITAL CARE RELATIVE TO ED VISIT AND TREATMENT OUTCOME APP HELPS IDENTIFY PATIENTS FOR FOLLOW-UPS AND RISK FACTORS DEMOGRAPHIC DATA DISEASE CONDITION PRESCRIPTION DATA IN-PATIENT/OUT- PATIENT COST LAB RESULTS CREATE MODELS THAT IDENTIFY SIGNIFICANT VARIABLES TRIAGE PATIENTS BASED ON ED DIAGNOSIS AND SEVERITY IDENTIFY DISEASES FOR WHICH PATIENT IS AT RISK MEASURE QUALITY OF HOSPITAL CARE RELATIVE TO TREATMENT OUTCOME AND FOLLOW UP VISITS CLOUD BASED MOBILE APP PATIENT X VISITED ED X MONTHS AGO AGE GENDER CONDITION HOSPITAL DATA 22

23 Analysis of 70 Healthcare companies across Home Healthcare, Insurance, HIE and E H R Analytic Companies underway Home Healthcare Companies 22 Understanding the services for readmission reduction program Identify potential opportunity List of Healthcare Companies in US Insurance Companies 39 4 HIE Vendors 5 Comparison of services with Saaraa 70 E H R Data/ Analytics Companies Understanding the services for readmission reduction program 23

24 Medical equipment/ medication Tele monitoring Provides software Skilled nursing Companies providing Homecare service to payors have significant opportunities for growth Patients Physicians Hospitals Payors Scoring

25 Capability demonstrated Development of Bayes models to identify high risk patients Ability to do this by disease conditions, significant co-morbid conditions - by state and hospital Enable Hospitals to identify high risk patients early and better provide care during and after they leave the hospital to reduce readmission rates Next steps Quantify benefits in terms of reduced readmission rates and map to services patient need 25