Cost Savings Associated with VA Hospital-based p Palliative Care Joan D. Penrod, PhD James J. Peters VA Mount Sinai School of Medicine This work was funded by VA Health Services Research and dD Development l tS Service. i Co investigators Co-investigators Partha P th Deb, D b PhD Cornelia Dellenbaugh, MPH James F. Burgess, Jr., PhD y Zhu, PhD Carolyn Cindy L. Christiansen, PhD Carol Luhrs, MD Therese B. Cortez, NP Elayne Livote Livote, MPH MPH, MS R. Sean Morrison, MD Background Care off hospitalized C h it li d patients ti t with ith advanced d d disease is characterized by: • High prevalence of pain and symptom distress • High g use of burdensome, non-beneficial care, including ICU • Inadequate communication among patients, families, & medical team Background Palliative Care: Definition • Interdisciplinary • Assess and treat p pain and other symptoms y p • Identify and discuss goals of care Background P lli ti C Palliative Care: D Definition fi iti • Provide practical and psychosocial support • Coordinate care across specialists and settings • Provided simultaneously with all other appropriate medical treatment Background • 53% of all U.S. hospitals with 50 or more beds have palliative care programs. • Consultative team is most common model. • VA issued directives and funding to establish and expand PC consultation teams in acute care and long-term care. Background Is palliative care effective? • Systematic review in 2008 of results from 22 RCTs found: – Greater family satisfaction with care – Improved I d quality lit off life lif • Recent R t RCT off a nurse-led l d PC iintervention t ti ffound: d – Higher scores for patient quality of life and mood – No differences in symptom intensity scores scores, days in the hospital hospital, ICU or emergency department visits Background What is the effect of PC on care costs? • RCT of PC consultation for 448 patients from an integrated health plan at 3 hospitals in 3 cities found: – Lower costs overall – Fewer ICU admissions on hospital readmission • Observational study of 21,000 patients from 8 hospitals with wellestablished PC programs found: – Savings of $279 in direct costs per day compared to usual care patients Background We hypothesized: – Lower hospital p costs for patients p with PC consult compared to similar patients g usual care receiving Background PC teams t elicit li it goals l off care that th t reflect fl t patient ti t and family preferences – Less use of cardiopulmonary resuscitation (CPR) – Fewer transfers to ICU for intubation, intravenous p esso s pressors – Fewer invasive procedures p – More comfort care measures Methods Ob Observational, i l retrospective i design d i Study Population • All adult patients admitted with advanced disease to 5 VA hospitals • October 1, 2004 to September 30, 2006 Methods Ad Advanced d disease di – – – – – – – – Metastatic solid tumor Central nervous system (CNS) malignancies Metastatic melanoma Locally advanced head and neck cancer Locally advanced pancreatic cancer HIV/AIDS with secondary diagnoses C Congestive i h heart ffailure il (CHF) Chronic obstructive pulmonary disease (COPD) Methods Data Sources • Patient characteristics and clinical information from VHA Medical SAS Inpatient Dataset • Inpatient care costs from VA Decision Support System National Data Extracts Methods Variables • Key independent variable – Binary indicator of whether the patient received a palliative care consultation during the hospital stay Methods Oth Independent Other I d d t Variables: V i bl • • • • • • • • • Condition/Diagnosis Age (under 75 years, 75-84, 85 and older) Married Bl k white, Black, hit other th Comorbidity index Death Enrollment priority group Ln(LOS) Site Methods Selection problem • Non-random Non random assignment of patients to treatment • Unobserved and unmeasured patient and/or physician factors influence both treatment and costs • Instrumental variable estimation to address selection l ti Methods Assumptions for IV • Physician preference for PC varies • Patients referred to PC by some physicians would not be referred by other p y physicians • Patient preference for medical intervention varies Methods Assumptions for IV • Because attending physicians assigned i d tto patients ti t quasi-randomly i d l ((2 – 4 week rotation), ) p physician y characteristics are orthogonal to patient characteristics such as unobserved health status Methods Statistical Analyses • Generalized Linear Models (GLM) (gamma distribution with log link) to estimate the unadjusted effect of PC consultation on costs • Probit model to estimate the unadjusted effect of PC consultation on the probabilityy of an ICU stay Methods Adjusted Adj t d effects ff t were estimated ti t d in i models d l th thatt extend the basic framework in two ways: • Observed characteristics introduced models • Both models estimated with nonlinear instr mental variables instrumental ariables (NLIV) – Using simulated likelihood methods for the GLMgamma models – Bivariate probit for ICU admissions R Results l Palliative Care ( N = 606) Usual Care (N = 2715) Percent Percent Less than 65 28 1 28.1 33 2 33.2 65-74 21.6 24.9 75-84 33.8 31.8 85 and older 16.5 10.1 White 63 0 63.0 65 5 65.5 Black 31.0 29.2 Other 6.0 5.3 Married 34.7 33.1 0.44 1–6 90 5 90.5 90 0 90.0 0 49 0.49 7, 8 9.5 10.0 92.2 40.5 Patient Characteristics (N= 3321) Age, years P Value <0 0001 <0.0001 Race 0 50 0.50 VA enrollment priority groups g study yp period Died during <0.0001 Palliative Care ( N = 606) Usual Care (N = 2715) 61.6% 15.8% <0.0001 Chronic Obstructive Pulmonary Disease 36.3 54.5 <0.0001 Congestive Heart Failure (CHF) 27.6 50.6 <0.0001 HIV/AIDS 31 3.1 13 1.3 0 0012 0.0012 Comorbidities, mean (SD) 2.5 (1.5) 2.5 (1.4) 0.69 Number of Hospitalizations Hospitalizations, mean (SD) 2 18 (1 2.18 (1.6) 6) 1 94 (1 1.94 (1.5) 5) 0 0004 0.0004 Hospitalizations (N= 6595) Palliative Care (N = 824) Percent Usual Care (N = 5771) Percent Proportion with an ICU stay 32.9 37.3 0.013 Hospital Length of Stay, days, mean (SD) 20.6 (23.3) 11.5 (15.9) <0.0001 ICU Length of Stay, days, mean (SD) 9.7 (15.3) 5.0 (9.4) <0.0001 Advanced Diseases: Advanced cancer Differences in hospital costs for p palliative versus usual care Effects of Palliative Care on Costs and ICU Use -6 60 -40 $ -20 $ -5 500 -400 -300 -20 00 -100 0 Pharmacy Cost 0 Total Direct Cost Unadjusted Adjusted I-beams indicate 95% confidence intervals Unadjusted Adjusted Effects of Palliative Care on Costs and ICU Use Laboratory Cost -60 0 -200 0 -150 -40 $ $ -100 -2 20 -50 0 0 Nursing Cost Unadjusted Adjusted I-beams indicate 95% confidence intervals Unadjusted Adjusted Effects of Palliative Care on Costs and ICU Use ICU Stay -.5 5 --20 -10 0 $ 0 10 perrcentage p points -.4 -.3 -.2 2 -.1 0 20 Radiology Cost Unadjusted Adjusted I-beams indicate 95% confidence intervals Unadjusted Adjusted Discussion • PC is i associated i t d with ith llower h hospital it l costs t compared to usual care for patients with advanced diseases • VA findings are consistent with the recent multimulti site RCT and the large observational study in non-government hospitals. • PC teams work with patients/families to establish treatment goals and arrange care consistent with goals. Discussion • Shift to lower intensity treatments • Fewer laboratory and radiological tests • Decreased use of the ICU Implications • Medically complex patients like the ones in the study y account for a g growing gp proportion p of admissions and bed days – Lower costs for PC patients could make a big difference in hospital financial performance Limitations • Data are not from an RCT: potential selection bias • Severity S it off illness ill • May not generalize to newly established PC tteams and/or d/ lless organized i d PC programs Conclusions • PC improves symptom control and family satisfaction with hospital care while also reducing costs compared to usual care care. • Expansion of palliative care programs is indicated Background • Sources for potential selection bias – Severity of illness – Patient/family P ti t/f il preferences f – Doctor attitude • Non-referrers think patient/family too unrealistic about prognosis • Non-referrers do not “need” a consultant • Referrers feel comfortable with end-of-life care issues