Routine Costs Within IPF Rates: Implications of Over-Aggregation AcademyHealth 2009 Annual Research Meeting June 30, 2009 Presented by: Edward M. Drozd, Ph.D. RTI International www.rti.org RTI International is a trade name of Research Triangle Institute Funding and Contributors • This presentation based on a CMS-funded study: – – – – “Psychiatric Inpatient Routine Cost Analysis” CMS Contract 500-95-0058, T.O. 13 Jerry Cromwell, Ph.D., Project Director Frederick G. Thomas, Ph.D., CMS Project Officer • Co-authors – – – – – 2 Jerry Cromwell, Ph.D. (RTI) Barbara Gage, Ph.D. (RTI) Jan Maier, RN, MPH (RTI) Leslie Greenwald, Ph.D. (RTI) Howard Goldman, MD (U. Maryland) Background: Medicare’s Role in Payment for Inpatient Psychiatric Care • About 500 thousand hospitalizations in over 1,800 inpatient facilities annually. • Freestanding psychiatric hospitals (public and private) and hospital units – Focus in this presentation on freestanding hospitals and socalled distinct part units, formerly exempt from the Acute Inpatient PPS (IPPS) – About 25% of patients are treated in “scatterbeds” subject to the IPPS. • In contrast to Medicare inpatient care as a whole, a majority of these patients (about ⅔) are disabled, not aged. 3 Background: Medicare Payment for Inpatient Psychiatric Care • For over 20 years (from 1983 to 2004), payments based on facility-specific historic cost (updated annually). – Same payment regardless of expected patient cost. – Newer hospitals received relatively higher payments. • Balanced Budget Refinement Act (BBRA; 1999) mandated changes in reimbursement – Per diem, not per stay payments – “Such system shall include an adequate patient classification system that reflects the differences in patient resource use and costs among such hospitals” • Inpatient Psychiatric Facility PPS (IPF-PPS) implemented in 2004 4 Challenges in Meeting the Congressional Mandate • Congressional mandate presented two main challenges: – Challenge 1: Measuring patients’ true per diem cost. – Challenge 2: Developing a casemix classification system to estimate patient costs prospectively. • In this presentation we: – Describe the limitations of administrative data in estimating relative costliness of patients, especially when costs are driven heavily by differences in staff time with patients – Present results from the Cromwell (2005) study showing the impact of using incomplete information on resource use variation in setting payment rates 5 Limitations of Administrative Data • Existing administrative consists of cost report and claims data • Costs can be divided into routine (nursing) versus ancillary • Routine costs include nursing, aide, social work, etc. staffing costs as well as room costs • Ancillaries include drugs, labs, therapy, ECT services 6 Implications of Limitations of Administrative Data • Administrative data include patient-specific charges for ancillary services, but not for routine care • Of costs in IPFs: – Routine costs account for 80% – Labor costs account for 75% • Not distinguishing individual patients’ specific contributions to routine costs, especially labor, may underestimate cost of treating patients who use a great deal of staff time, and vice versa 7 Study Overview: Purpose & Data • Purpose: to collect background information for CMS for implementing the IPF-PPS – Collect patient-specific daily data on actual resources used – Identify characteristics associated with resource use that could be used in a payment system – 4-year effort, data collected from 40 inpatient psychiatric facilities • Data: Primary data on resource use combined with administrative data – Patient and staff times in activities for up to 7 days per patient. – Medicare patient diagnostic, behavioral, and demographic characteristics – 696 Medicare patients with matched claims and primary data 8 Study Overview: Methods (1) • Staff and patient times in activities combined to create facility-specific patient day staff resource intensity index • Per diem cost measure sum of three components: – Labor-related routine portion adjusted for patient day-level variation in resource use – Non-labor–related portion based on facility average – Ancillary portion based on patient stay average from claim (charges multiplied by cost-to-charge ratio) 9 Study Overview: Methods (2) • Patient casemix classification groups created based on patient average cost – Groups defined by hierarchical splits of patient characteristics, generated by CART with expert guidance over splits – Patients first be separated by clinician-advised diagnostic categories (Schizophrenia, Dementia, Mood Disorders, SubstanceRelated Disorders, and Others) for maximal clinical relevance – Other characteristics further separate patients hierarchically • Per diem cost regressions estimated using patient groups, dayof-stay groups, and facility characteristics • End result: 16-group model using some characteristics (ADLs, dangerousness to self or others) not found on claims 10 0 11 RN-Equivalent Minutes per Patient 1280+ 1121–1200 1201–1280 1041–1120 881–960 961–1040 801–880 641–720 721–800 561–640 481–560 321–400 401–480 241–320 81–160 161–240 ≤ 80 Percent of Patients Distribution of Staff Intensity Per Medicare Patient Day 25 20 15 10 5 Sources of Variation in Resource Intensity: Staff Intensity in Most Frequent Activities Top 10% Middle 80% Bottom 10% Top 10% ÷ Bottom 10% 835.0 361.6 150.1 5.6 Personal Care 48.5 21.0 7.5 6.5 Meals 28.6 19.5 11.3 2.5 Medications 47.3 30.3 12.2 3.9 Assessment/Treatment Planning 145.4 53.9 11.8 12.1 Observation/Seclusion 207.6 5.0 0.8 259.5 Checks 36.1 32.2 16.7 2.2 Other 74.5 53.3 26.5 2.8 All Routine 12 % of Patients 13 2.25 30 2.00 25 1.75 20 1.50 15 1.25 10 1.00 5 0.75 0 0.50 Relative Weight 35 Schizo 1 Schizo 2 Schizo 3 Schizo 4 Schzio 5 Dementia 1 Dementia 2 Dementia 3 Mood 1 Mood 2 Mood 3 Mood 4 Mood 5 Residual Substance 1 Substance 2 Compression in Estimates of Relative Costs for RTI Case Mix Groups % Covered Days Adjusted Not Adjusted Compression in Estimated Relative Cost for Above- and Below-Average Cost Patients Case Mix Groups With Above-Average Costs (Relative Weight > 1) Case Mix Groups With Below-Average Costs (Relative Weight < 1) Difference in Weight Between Adjusted and Unadjusted Per Diem Cost 0.16 0.05 Average Relative Weight 1.26 0.91 Proportion of Patients 29.3 70.7 Average Compression Percentage 12.4% 5.9% 14 Summary of Compression Findings (1) • There is significant variation in routine resource use per patient day – Coefficient of variation equal to 54% – Interquartile range equal to 64% of the median – Ratio of top decile to bottom decile equal to 5.6 • Differences driven by certain activities – Personal care, observation/seclusion, assessment/ treatment planning – Suggests there may be case mix characteristics that can identify these patients • This variation is unmeasured within facility, only incorporated into costs to the degree that facilities specialize is treating certain types of patients 15 Summary of Compression Findings (2) • Compression in estimated per diem cost less pronounced than in routine cost differences – Ancillary costs already patient-specific (though not patientday–specific – Other components of cost (e.g., overhead, non-labor–related routine costs) not identified as patient specific • More compression in estimated cost for high-cost patients than low-cost patients • On average, 12% underestimate of cost for high-cost patients; 6% overestimate of cost for low-cost patients 16 Implications • Systematic under- and overpayments for patients can lead to distortions in service delivery – More selective admission patterns – Dropping or adding service lines/specialties • How to collect more patient-specific resource use information? – Intermittent time studies, as also done for SNFs and currently for post-acute care – Modified charge structures for incremental nursing costs (e.g., 1-1 observation or seclusion) or recognizing different levels of care (as is done for intensive care, critical care, etc. units in acute hospitals) 17 Appendix: Patient Classification System Used in this Presentation (1) Patient Group Relative Weight Medicare Day % Definition Schizo 1 HiADL + Age>65 + HiPsy 1.98 0.61 Schizo 2 Age>65 + (HiADL + LoMed or LoADL + HiDanger) 1.01 3.66 Schizo 3 LoADL + Age>65 + LoDanger 0.96 5.40 Schizo 4 LoADL + Age<65 + HiPsy 0.90 3.99 Schizo 5 Age<65 + (LoADL + LoPsy or HiADL) 0.85 31.20 Dementia 1 HiADL + HiMed 1.45 2.95 Dementia 2 All Other Dementia 1.23 8.50 Dementia 3 LoADL + LoPsy 0.97 4.54 18 Appendix: Patient Classification System Used in this Presentation (2) Patient Group Relative Weight Medicare Day % Definition Mood 1 Age>65 + HiPsy + HiMed 1.40 2.32 Mood 2 Age>65 + HiPsy + LoMed + HiDanger 1.27 3.32 Mood 3 LoMed + OnECT + (Age<65 or Age>65 + HiPsy + LoDanger) 1.37 3.18 Mood 4 All Other Mood Disorder 1.00 14.43 Mood 5 Age<65 + LoMed + NoECT + NoDetox 0.92 10.37 Residual All Residual Disorder 1.15 4.35 Substance 1 HiDanger 1.09 0.41 Substance 2 LoDanger 0.85 0.79 19 Appendix: Patient Classification System Used in this Presentation (3) Characteristic Description of Clinical Criterion Diagnosis Category One of 5 diagnosis categories (schizophrenia, dementia, mood disorders, residual disorders, substance/related) determined through expert opinion (Hi/Lo)ADL Patient indicated as having 2 or more ADL deficits Age(<65/>65) Patient’s age is 65 years or greater (Hi/Lo)Psy Patient’s primary (psychiatric) diagnosis in a class of severe diagnoses determined through expert opinion (Hi/Lo)Med Patient’s has any comorbid medical diagnoses in a class of severe diagnoses determined through expert opinion (Hi/Lo)Danger Patient indicated as being dangerous to self or others (On/No)ECT Patient receiving electroconvulsive therapy during stay (On/No)Detox Patient receiving detoxification during stay 20