Potentially Preventable Readmissions (PPRs) in the Texas Medicaid Population, Fiscal Year 2009 Hospital Seminars January 2011
Agenda 1. Overview 2. 3M All Patient Refined Diagnostic Related Groups (APR-DRGs) and PPRs 3. Study data sources and preparation 4. Study methods casemix adjustment 5. Statewide results 6. Hospital reports For Further Information www.hhsc.state.tx.us for the public PPR report PPR.report@tmhp.com for questions
OVERVIEW Presenters John Chapman, PhD Senior Consultant, Payment Method Development, ACS Government Healthcare Solutions Lisa Lyons, RN Product Marketing Manager, 3M HIS Kevin Quinn, MA, EMT-P Vice President, Payment Method Development, ACS Government Healthcare Solutions The study was performed by Affiliated Computer Services, Inc. (ACS), a Xerox Company, which is the parent company of Texas Medicaid and Healthcare Partnership (TMHP). 3M provided valuable technical assistance but is not responsible for how the PPR methodology was applied or for the results.
OVERVIEW Learning Objectives Understand what a PPR is and how it is identified Understand what data was used for the reports, both in general and for one s own hospital Understand how to compute and assess PPR rates Have familiarity with general PPR patterns Be able to assess and explain your hospital s PPR rate, and explore reduction strategies.
OVERVIEW Please Bear in Mind Statements and opinions are those of the presenters and not necessarily those of the Health and Human Services Commission (HHSC). This is the first year this analysis has been performed. Suggestions to improve data, methodology, and presentation are welcome. TMHP and ACS have no financial interest in any DRG algorithm or method of measuring readmissions. Results in this analysis were produced using data obtained through the use of proprietary computer software created, owned, and licensed by the 3M Company. All copyrights in and to the 3M TM Software are owned by 3M. All rights reserved.
OVERVIEW Concern over Readmissions Over 150,000 people a year in Florida 11 percent of the all-payer population studied were readmitted for potentially preventable reasons. -- Goldfield et al. 20 percent of Medicare inpatients were readmitted within 30 days, almost all unplanned Half of medical patients did not see a physician in the interval before readmission -- Jencks et al. Reducing readmissions at Park Nicolet: We ve kept it up out of a sense of moral obligation to these patients, but we re getting killed, said David K. Wessner, chief executive of Park Nicollet. We will totally run out of gas. -- NY Times The Congress should direct the Secretary to reduce payments to hospitals with relatively high readmission rates for select conditions -- MedPAC
OVERVIEW Texas Readmissions Statute (Gov. Code, Sec. 531.913) Enacted in 2009 Requires HHSC to measure and report on potentially preventable readmissions (PPRs) of Medicaid patients. Defined as the return hospitalization due to deficiencies in care or treatment during the initial hospital stay or in post-hospital discharge follow-up. Does not include readmission from unrelated events. But does include readmission for: Same condition or procedure. Infection or other complication resulting from care previously provided. Condition or procedure indicating that the previous admission s surgical intervention was unsuccessful in achieving the anticipated outcome. Another condition or procedure of a similar nature HHSC must create a program to identify PPRs and exchange PPR performance information with each hospital. Each hospital must distribute this information to its health care providers.
OVERVIEW Questions of Tone and Approach Focus on individual stays or overall rates Traditional approach is on medical errors in individual stays Alternative approach is focusing on hospital-wide rates Emphasis on potentially preventable events Punishing bad performance vs. enabling excellence Name/blame/shame vs transparency/collaboration Good people working in bad systems Medical errors vs. continuous quality improvement How to help hospitals improve themselves? Give hospitals information they can use
OVERVIEW 76-Year-Old Man with Heart Failure Admitted The patient Active 76-year-old male, retired investment broker History of seven chronic conditions Takes nine meds daily; coping with dietary restrictions Lives with wife of 50 years; she shows cognitive changes Three children with families living in other states The care Under the care of six specialists Primary care provider retired Admitted and treated for exacerbation of heart failure Based on a case study presented by Randall Krakauer, M.D., in a 12/2/09 presentation Aligning Reimbursement to Reduce Avoidable Hospital Readmissions, sponsored by the Healthcare Intelligence Network.
OVERVIEW and Readmitted The handoff Three new medications ordered Oral and handwritten discharge instructions Told to schedule follow-up MD appointment within seven days The outcome Can t read discharge instructions Has questions about meds but doesn t know who to call Weak, dizzy, unable to eat First available MD appointment more than two weeks away Two weeks later, rehospitalized for acute heart failure Due to lack of adherence to prescribed therapies
OVERVIEW Which Solution Makes the Most Sense? Readmission with Implantation of Left Ventricular Heart Assist Good Instructions, MD Visit Scheduled, Home Visit by RN Price tag: $88,000 Price tag: < $500
OVERVIEW Steps in Our Analysis 1. Create analytical dataset based on claims extract and encounter files Includes extensive data validation 2. Group by APR-DRG Base DRG plus level of severity (e.g., DRG 123-4) 314 base DRGs x 4 = 1,256 total DRGs 3. Calculate PPRs using 3M PPR software Analyze admits in 11 months with PPRs in 12 months 4. Calculate risk-adjusted PPR rates by hospital Adjust for base DRG, severity, age, psych comorbidity
OVERVIEW PPR Results for Fiscal Year 2009 Total Readmissions Medicaid Care Category Initial Admits Readmit Chains Same Hospital Other Hospital All PPR Rate Pediatric Respiratory 27,239 649 552 170 722 2.4% Other medical 41,311 1,223 1,073 389 1,462 3.1% Other surgical 10,935 470 442 89 531 4.5% MH/SA 14,307 1,181 871 593 1,464 9.2% Subtotal 93,792 3,523 2,938 1,241 4,179 3.9% Adult Circulatory 13,809 1,025 835 409 1,244 8.2% Other medical 49,808 3,618 2,914 1,517 4,431 8.0% Other surgical 17,650 1,005 914 243 1,157 6.1% MH/SA 14,126 1,445 1,112 978 2,090 12.0% Subtotal 95,393 7,093 5,775 3,147 8,922 8.2% Obstetrics 155,038 1,180 1,001 216 1,217 0.8% Total 344,223 11,796 9,714 4,604 14,318 3.6% MH/SA: Mental health and substance abuse
STUDY DATA Data Sources Fee-for-Service (FFS) and Primary Care Case Management (PCCM) Based on standard claims extract Well-established and familiar to hospitals Augmented to include up to ten diagnoses and up to six procedures Managed care encounter data Claims adjudicated by the managed care plan and submitted to HHSC Required greater review and validation efforts: Combining multiple records for a stay Removing duplicate records Removing records with critical data issues
STUDY DATA Key Data Quality Questions for PPR Is there one, and only one, record in the dataset for each hospital stay in the real world? Are providers and recipients accurately and consistently identified? Are the diagnosis, procedure, and discharge status code fields accurate, complete, and consistent?
STUDY DATA Main Data Validation Edits Consolidation of multiple records for a single stay (claim chaining) Removed duplicates Removed invalid/unreliable discharge status Removed undocumented aliens, because not all stays are covered and so PPR assignment would be inappropriate.
STUDY DATA Identifying Patients and Hospitals Patients consistently identified by recipient number No names, Social Security numbers, or birth dates, even in confidential data to hospitals Hospitals identified by Texas Provider Identifier (TPI) FFS/PCCM claims show the Medicaid TPI Encounter claims show National Provider Identifier (NPI) NPI cross-walked to TPI using NPI, bill type, address, taxonomy, etc. TPIs reviewed for duplicates, anomalies
STUDY DATA Completeness of Dx and Px Coding Children s and psychiatric hospitals not paid by DRG may code less completely We compared the number of diagnoses and procedures reported for each stay, controlling for 1.25 1.00 0.75 0.50 0.25 - Chart A.2.4.3.1 Measure of Diagnosis and Procedure Coding Completeness 1.08 Children's Hospitals 0.96 DRG Hospitals See text for explanation of comparison 0.77 Psych Specialty Hospitals the mix of DRGs Children s hospitals showed no obvious evidence of under coding, while psychiatric hospitals did 1.13 DRG Hospitals
STUDY DATA APR-DRG and PPR Assignment APR-DRG and PPR status assigned to each stay 0.6 percent of stays had APR-DRG or PPR grouping errors and were omitted from the final dataset Major methodological exclusions from dataset Newborns, multiple trauma, metastatic cancer, left against medical advice, etc. Initial admissions in August 2009 Every remaining stay was either an initial admission or a PPR Initial admissions may or may not be the initial claim in a PPR chain
STUDY DATA Data Review and Preparation Results Adjustment FFS/PCCM Encounter Total Records received 484,995 245,418 730,413 Removed to assure each record represents a unique, IP stay: 50 21,822 21,872 Removed due to data issues: 1,633 19,688 21,321 Removed for study design reasons: 264,899 78,098 342,997 Final Analytic Dataset 218,413 125,810 344,223 Note: Further detail is in Appendix Table A.2.1 of the report.
STUDY METHODS Introduction to Casemix for PPRs Four characteristics strongly influence the likelihood that a stay will have a PPR Base APR-DRG Severity of illness Age Serious mental health or substance abuse co-morbidity
STUDY METHODS Variation by DRG Base DRG PPR Rate Cesarean delivery 1.4% Bronchiolitis and RSV pneumonia 2.6% Appendectomy 4.1% Diabetes 7.4% Heart Failure 10.2% Schizophrenia 14.7%
STUDY METHODS Variation by Severity Severity Level Base DRG 1 2 3 4 Cesarean delivery 1.1% 2.0 % 3.0 % 3.1 % Bronchiolitis and RSV pneumonia 2.1% 2.8 % 5.3 % 12.7 % Heart Failure 8.1 % 9.9 % 11.4 % 8.8 % Schizophrenia 15.3% 13.8% 17.3% N/A
STUDY METHODS Variation by Age Pediatric PPR Rates in Relation to Adult Rates 120% 100% 80% 60% 40% 20% 0% % of Adult Rate 750-2 Schizophrenia 753-1 Bipolar Dis 753-2 Bipolar Dis 751-2 Maj Depression 383-2 Cellulitis 139-2 Oth Pneumonia 463-2 Kidney/UTI 420-2 Diabetes 751-1 Maj Depression 139-3 Oth Pneumonia APR-DRGs shown are the ten most common adult DRGs that also had at least 100 pediatric initial admissions Adult Pediatric
STUDY METHODS Variation by MH/SA Co-morbidity Age Category MH/SA Co-morbidity Adj. Factor Pediatric No 0.993 Yes 1.337 Adult No 0.978 Yes 1.127 Note: Excludes obstetrics and MH/SA stays, for which the MH/SA co-morbidity is not a significant factor.
STUDY METHODS PPR Rates by Medicaid Care Category MCC Pediatric Adult Respiratory / Circulatory 2.4% 8.2 % Other Medical 3.1 % 8.0 % Other Surgical 4.5 % 6.1 % MH / SA 9.2 % 12.0 % Obstetrics 0.8 % MCCS are intended to be typical of internal hospital organization and Medicaid policy areas MCCs reflect three of the four sources of variation in the likelihood of a PPR.
STUDY METHODS But PPR Rates Also Vary Within MCCs Other Medical DRGs Pediatric Adult 383-1 Cellulitis & Oth Bact Skin Inf 0.7 2.8 249-2 Non-Bact Gastroenteritis, N&V 2.2 6.2 720-3 Septicemia & Disseminated Inf 6.0 9.6
STUDY METHODS Norms For this report, norms were established: For each combination of : Base DRG Severity level Age category Using average Texas Medicaid statewide rates Norms do not necessarily reflect best practices
STUDY METHODS Comparing TX Medicaid with FL All-Payer Chart 2.1.1 Comparison of Results: Texas Medicaid vs. Florida All-Payer TX Medicaid FL All-Payer 14% 12% 10% 8% 6% 4% 2% 0% Ped Resp Ped Oth Medical Ped Oth Surgical Ped MH/SA Adult Circ Adult Oth Medical Adult Oth Surgical Adult MH/SA Obstetrics Florida results have been made comparable to Texas results through adjustment for frequency by APR-DRG and adult/pediatric age split, Total
STUDY METHODS Expected Values The expected value, or PPR likelihood, for each stay is: The norm for that stay Multiplied by The applicable MH co-morbidity factor. The expected PPR rate for a group of stays such as all the stays of a hospital is the sum of the expected values of the stays in the group.
STUDY METHODS Expected PPR Rates: Illustration DRG Age MH Comb. MH Adj. Norm Indiv. Prob. of PPR A 14 No.9 10% 9% B 32 Yes 1.5 20% 30% C 54 No.95 22% 21% Sum 60% Group Expected Rate (average of individual probabilities) 20% Note: Numbers are hypothetical, for illustration
STUDY METHODS Expected PPR Rates: Example DRG Age MH Comb. MH Adj. Norm Indiv. Prob. of PPR 249-2 14 No.99 6.0% 5.9 % 249-2 32 Yes 1.127 10.8% 12.7 % 194-3 54 No 0.98 11.4 % 11.2 % Group Expected Rate (average of individual probabilities) 9.9%
STUDY METHODS Actual-to-Expected (A/E) Ratio There is need to assess PPR rates for various mixes of DRGs, severity, etc. Each hospital has its own mix Within a hospital, each MCC has a distinct mix To measure how each observed PPR rate compared to its norm we computed actual-to-expected ratios. An expected rate is computed for each hospital. The ratio of the hospital s actual rate to this expected rate is a standard measure of performance. This method is called indirect adjustment.
STUDY METHODS Actual-to-Expected Ratio: Example DRG Age Group MH Comb. Number of Stays Indiv. Prob. of PPR Actual # of PPRs Exp. # of PPRs 249-2 Pediatric No 100 5.9 % 5 5.9 249-2 32 Yes 100 12.7 % 12 12.7 194-3 54 No 100 11.2 % 10 11.2 Total 27 29.8 Actual-to-Expected Ratio 29.8 / 27 = 0.91
STUDY METHODS Interpreting A/E Ratios At one level it is what it is this study is using all applicable stays, so is not subject to sampling variation. BUT. It s not only tempting, but useful, to generalize. This can be done, but with caution and only when volumes are large enough.
STUDY METHODS PPR Rates with Few Stays Imagine a hospital with a true PPR rate of 5 percent for all its stays. It has 40 admissions a year. The expected number of PPRs each year is two. But it s not unlikely that it will have one or three in any given year. If it has one, its A/E ratio is 0.5 If it has three, its A/E ratio is 1.5
STUDY METHODS Protection Against Over-interpretation: Step 1 Minimum volume threshold At least 40 stays, and At least five actual PPRs, and At least five expected PPRs If the volume threshold is not met, we provide the actual and expected values, but not the A/E ratio. While it s important not to over-interpret a single high or low A/E ratio, it s still important to become aware of what PPRs are occurring.
STUDY METHODS Protection Against Over-interpretation: Step 2 Test of how likely it is that the observed A/E ratio differs from 1.00 simply by random chance Depends on how different the observed A/E ratio is from 1.00 and on the volume of stays Cochran-Mantel-Haenszel (CMH) statistic A/E ratio in hospital-specific reports is flagged * if the p-value < 0.10 ** if the p-value < 0.05
STATEWIDE RESULTS Overview of Results Total Readmissions Medicaid Care Category Initial Admits Readmit Chains Same Hospital Other Hospital All PPR Rate Pediatric Respiratory 27,239 649 552 170 722 2.4% Other medical 41,311 1,223 1,073 389 1,462 3.1% Other surgical 10,935 470 442 89 531 4.5% MH/SA 14,307 1,181 871 593 1,464 9.2% Subtotal 93,792 3,523 2,938 1,241 4,179 3.9% Adult Circulatory 13,809 1,025 835 409 1,244 8.2% Other medical 49,808 3,618 2,914 1,517 4,431 8.0% Other surgical 17,650 1,005 914 243 1,157 6.1% MH/SA 14,126 1,445 1,112 978 2,090 12.0% Subtotal 95,393 7,093 5,775 3,147 8,922 8.2% Obstetrics 155,038 1,180 1,001 216 1,217 0.8% Total 344,223 11,796 9,714 4,604 14,318 3.6%
STATEWIDE RESULTS Most PPRs Table 2.4.1 PPR Rates by APR-DRG: Top 20 APR-DRGs in Terms of Total Readmissions Base DRG Initial Admits Readmit Chains Readmit Stays Stays per Chain PPR Rate 753 Bipolar Disorders 11,283 1,176 1,530 1.3 10.42% 750 Schizophrenia 5,082 745 1,129 1.5 14.66% 751 Major Depression 4,998 475 615 1.3 9.50% 540 Cesarean Delivery 41,035 565 577 1.0 1.38% 560 Vaginal Delivery 91,865 543 560 1.0 0.59% 194 Heart Failure 2,861 291 369 1.3 10.17% 140 COPD 3,188 301 355 1.2 9.44% 139 Other Pneumonia 9,990 296 339 1.1 2.96% 420 Diabetes 2,535 187 266 1.4 7.38% 138 Bronchiolitis & RSV Pneumonia 9,270 236 252 1.1 2.55% 662 Sickle Cell Anemia Crisis 1,611 177 252 1.4 10.99% 720 Septicemia & Disseminated Infections 2,335 192 226 1.2 8.22% 053 Seizure 3,808 167 209 1.3 4.39% 249 Non-Bacterial Gastroenteritis 5,673 162 195 1.2 2.86% 279 Hepatic Coma & Oth Major Liver Disorders 737 139 190 1.4 18.86% 280 Alcoholic Liver Disease 765 147 188 1.3 19.22% 383 Cellulitis & Other Bacterial Skin Infections 6,492 168 178 1.1 2.59% 460 Renal Failure 1,431 137 167 1.2 9.57% 463 Kidney & Urinary Tract Infections 4,572 140 163 1.2 3.06% 282 Disorders of Pancreas except Malignancy 1,338 118 155 1.3 8.82% Note: The APR-DRG shown is the DRG for the initial admission.
STATEWIDE RESULTS Statewide Results: By Delivery Method Medicaid Care Category Pediatric Initial Admits Table 2.2.1 PPR Results by Health Care Delivery Method Fee-for-Service Primary Care Case Management Managed Care Organization Actual PPR Rate Expctd PPR Rate Actual / Expctd Ratio Initial Admits Actual PPR Rate Expctd PPR Rate Actual / Expctd Ratio Initial Admits Actual PPR Rate Expct d PPR Rate Actual / Expctd Ratio Respiratory 7,442 3.0% 2.9% 1.05 10,259 2.2% 2.2% 0.98 8,816 2.3% 2.3% 0.98 Other medical 11,937 4.0% 3.7% 1.07 13,942 2.6% 2.7% 0.96 13,970 2.8% 2.9% 0.97 Other surgical 4,532 5.6% 4.9% 1.12 2,992 4.2% 4.4% 0.97 2,880 3.2% 4.1% 0.78 MH/SA 4,766 9.2% 9.3% 0.99 3,390 7.0% 8.7% 0.80 4,687 10.8% 9.4% 1.14 Subtotal 28,677 4.8% 4.6% 1.05 30,583 3.1% 3.4% 0.92 30,353 3.9% 3.9% 1.02 Adult Circulatory 5,372 8.1% 8.3% 0.98 7,011 8.3% 8.1% 1.02 182 6.0% 7.1% 0.85 Other medical 20,900 8.4% 8.0% 1.04 22,360 7.9% 8.1% 0.98 2,117 4.4% 5.6% 0.78 Other surgical 8,669 6.2% 6.2% 1.00 6,873 6.4% 6.2% 1.03 951 2.9% 4.5% 0.65 MH/SA 3,602 10.8% 11.4% 0.94 2,836 9.1% 11.5% 0.79 5,598 14.3% 12.6% 1.13 Subtotal 38,543 8.1% 8.0% 1.01 39,080 7.8% 8.0% 0.97 8,848 10.5% 10.0% 1.06 Obstetrics 17,408 0.6% 0.8% 0.82 53,472 0.7% 0.8% 0.87 82,941 0.8% 0.8% 1.13 Total 84,628 5.4% 5.4% 1.02 123,135 3.5% 3.7% 0.95 122,142 2.3% 2.2% 1.06 Note: Actual/expected ratios were calculated using more decimal places in the actual and expected PPR rates than are shown here.
STATEWIDE RESULTS Statewide Results: Reasons for Readmission Reason Share Medical readmissions for the same condition as the initial admission 23% Medical readmissions for a different acute condition that could plausibly have had a clinical association with the initial admission Mental health or substance abuse readmissions that followed an initial admission for mental health or substance abuse 29% 24% Post-surgical complications 2% Only 11% of readmissions following surgery are in this category Other reasons 22%
STATEWIDE RESULTS Statewide Results: Variation among hospitals Table 2.6.1 Number of Hospitals by PPR Performance Ratio of Actual PPRs to Expected PPRs Interpretation Hospitals Stat Sig Diff Lower than 0.75 Much lower than expected 23 11 0.75 to 0.89 Lower than expected 45 8 0.90-1.10 About as expected 83 0 1.11 to 1.25 Higher than expected 45 9 Higher than 1.25 Much higher than expected 34 23 Total 230 51 Notes 1. Low-volume hospitals are excluded. Low-volume hospitals do not meet the criteria of having at least 40 initial admissions, at least five expected readmissions, and at least five actual readmissions. 2. Stat Sig Diff shows the number of hospitals where the difference from 1.00 is statistically significant at the 90% confidence level.
STATEWIDE RESULTS Statewide Results: Variation among hospitals Chart 2.6.1 PPR Actual-to-Expected Ratios by Hospital Rank 2.50 Actual-to-Expected Ratio 2.00 1.50 1.00 0.50 Each dot is a hospital. A hospital with an actual/expected ratio below 1.00 had fewer PPRs than expected; a hospital with an actual/expected ratio above 1.00 had more PPRs than expected. - 0 50 100 150 200 250 Hospitals Ranked by Actual/Expected PPR Ratio (From Low to High)
STATEWIDE RESULTS PPRs by Days Since Discharge Chart 2.7.1 Patterns in PPR Initiation by Days Since Discharge Number of PPR Chains Initiated (Broken Blue Line) 1200 1000 800 600 400 200 18,000 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 Cumulative PPR Chains Initiated (Solid Red Line) 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Days Since Discharge -
HOSPITAL REPORTS Hospital-Specific Data Reports Provided confidentially to each hospital Each hospital to share with its providers Information: Hospital-specific version of PPR report Excel file includes Excel version of tables in PPR report plus individual claim data