Determinants of Inpatient Smoking S ki Cessation C ti Quality Q lit D Douglas l Levy, L PhD dlevy3@partners.org Massachusetts General Hospital and Harvard Medical School Supported by the Robert Wood Johnson Foundation S b t Substance Ab Abuse Policy P li Research R h Program P C ll b Collaborators t Ray Kang, MA – Northwestern University, Feinberg School of Medicine Christine Vogeli, PhD – Massachusetts General Hospital, Harvard Medical School Nancy Rigotti Rigotti, MD – Massachusetts General Hospital, Harvard Medical School National Hospital Inpatient Quality Measures In 2002, Joint Commission adopted National Hospital Inpatient Quality (NHQ) measures Hospitals participating in the Hospital Quality Alliance / Hospital Compare report data to CMS Focus on 3 conditions – AMI, AMI heart h t failure, f il pneumonia i Provision of smoking cessation advice/ counseling (SCA) is i only l metric t i for f allll three th conditions diti D t Determinants i t off quality lit Patient characteristics – LOS? Race? Comorbidities? Hospital characteristics – Teaching hospital (better AMI process) – Percent minority patients (worse process measures for high minority hospitals) Region g – Performance on quality metrics vary by HRR R Research h questions: ti What entity has the greatest influence on the provision of SCA? – Patient? Hospital? Region? Wh t specific What ifi characteristics h t i ti off these th nested entities determine SCA rates? P Patient Patientti t-level l l data d t Acquired A i d from f CMS 2005--2008 2005 NHQ measures – yes/no SCA Patient demographic and clinical characteristics – Age, sex, race, insurance, LOS, discharge status, comorbidities Hospital and Regional level data Hospital – AHA survey data (2007) Size (# beds), ownership (public, NFP, FP), t teaching hi status t t Assume constant across study period – Aggregated patient characteristics from NHQ patientpatient-level file (by condition) race payer race, payer, smoking rate rate, non non--SCA quality Region – defined as Dartmouth HRR Population characteristics 2005--2008 2005 AMI HF PN Annual avg. # smoking patients in sample 141k 146k 200k Unique hospitals 3,688 3 688 4,588 4 588 4,828 4 828 Annual avg. % community hospitals 61 82 88 P ti t demographics Patient d hi Female White Black Hispanic <=64 65--79 65 80 >=80 AMI 31% 81 11 5 74 22 4 HF 39% 60 32 6 59 32 9 PN 50% 79 15 5 64 28 8 Additional patient characteristics Payer Medicare Medicaid Private Unk. Ins. LOS Dischg. to an Institution AMI HF PN 28% 9 48 12 4.9 48% 16 25 10 4.9 42% 15 31 11 6.0 6 10 14 C Common comorbidities biditi Hypertension Chronic Lung g Disease Diabetes Electrolyte/ Fluid Disorders Obesity Anemia AMI 57% HF 59% PN 44% 28 22 54 30 65 19 12 10 8 21 12 19 34 7 17 SCA rates t – patient ti t level l l 100% 90% 80% AMI HF PN 70% 60% 50% 2004 2005 2006 2007 2008 2009 SCA rates t – hospital h it l level l l 100% 90% 80% AMI HF PN 70% 60% 50% 2004 2005 2006 2007 2008 2009 What factors affect SCA rates? Use mixed effects models – Random intercepts for patient, hospital, HRR – Unconditional model (no covariates) Intraclass correlation coefficient as an estimate of how much of variability in SCA rates is at each level of hierarchy – Conditional model (adds covariates) Wh Where is i variability? i bilit ? AMI 80% 70% 60% ICC 50% Patient Hospital HRR 40% 30% 20% 10% 0% 2004 2005 2006 2007 2008 2009 Wh Where is i variability? i bilit ? HF 80% 70% 60% ICC 50% Patient Hospital HRR 40% 30% 20% 10% 0% 2004 2005 2006 2007 2008 2009 Wh Where is i variability? i bilit ? PN 80% 70% 60% ICC 50% Patient Hospital HRR 40% 30% 20% 10% 0% 2004 2005 2006 2007 2008 2009 Effect of patient characteristics on SCA AMI HF PN 2005 2008 2005 2008 2005 2008 -8% -1% -3% -1% -6% -2% Black -2 ns -1 ns -3 -1 Hispanic -2 ns -2 -1 -4 -1 Asian -3 -1 ns ns -6 -3 -.1/day ns .1/day .04/day -.1/day -.03/day -14 -4 -12 -5 -14 -7 Age >=80 LOS Disch’d to I tit ti Institution p <0.05 unless marked “ns” Effect of patient comorbidities on SCA AMI HF PN 2005 2008 2005 2008 2005 2008 2% ns 3% 1% 3% 1% Anemia -1 -.3 -1 -.3 -1 -.4 Chrn. Lung 1 .4 2 .5 4 1 Diabetes -1 ns -1 ns -1 -.3 Lymphoma -4 ns -2 ns -5 -1 Mets -8 -1 -6 -1 -6 -2 Paraplegia p g -5 -1 -3 -1 -4 -1 Tumor -6 ns -2 ns -5 -1 Alcoholism p <0.05 unless marked “ns” Effect of hospital characteristics on SCA AMI HF PN 2005 2008 2005 2008 2005 2008 10pt. %Smokers in dx 2.4 -.4 1.7 1.4 ns 2.7 10pt. non non--SCA quality 5 3 6 5 6 5 10pt % Medicaid 10pt. -.1 1 ns -.8 8 -.8 8 ns -1.2 12 # Beds (∆ (∆/10 beds) .1 .05 .2 .07 .2 .1 Private FP 2 3 2 ns 3 2 Public -5 -2 -3 -3 -4 -5 Maj. Teach -5 ns -6 ns -6 ns p <0.05 unless marked “ns” Summary S (1) SCA rates climbed substantially during study period i d SCA for HF and PN lags AMI – CCU effect? Major j shift from patientpatient p -level determinants to hospitalhospital-level determinants over study p period – Hospital initiatives resulting in all patients being p treated the same is a likelyy explanation Summary S (2) Hospitals w/ more smokers in a given diagnosis tend to do better at SCA SCA quality is predicted by quality on other measures Private FP hospitals generally doing better than private NFP hospitals – public hospitals lag both Major teaching hospitals have made up their deficit relative to nonnon-teaching Li it ti Limitations Do not know what SC advice/ counseling li took t k place l – SCA in hospital is not enough to meaningfully i f ll improve i cessation i rates – Revised SC measure being piloted All patients, includes postpost-discharge f/u Included many, y, but not all hospitals p Models not ideal for binary outcomes C Conclusion l i Most hospitals have been very effective at responding to the quality measurement initiatives – Some better than others Still some lag for vulnerable hospitals and patients Effect of patient characteristics on SCA AMI HF PN 2005 2008 2005 2008 2005 2008 -8% -1% -3% -1% -6% -2% Black -2 ns -1 ns -3 -1 Hispanic -2 ns -2 -1 -4 -1 Asian -3 -1 ns ns -6 -3 -.1/day ns .1/day .04/day -.1/day -.03/day -14 -4 -12 -5 -14 -7 Age >=80 LOS Disch’d to I tit ti Institution p <0.05 unless marked “ns” Effect of hospital characteristics on SCA AMI 2005 2008 HF PN 2005 2008 2005 2008 10pt. %Smokers in dx 2.5 -.4 .9 ns 3 10pt. %Black ns ns ns ns ns 10pt %Hispanic 10pt. ns -.8 8 ns -1 ns 10pt. % Medicaid -1.1 ns -3 ns 2 # Beds (∆ (∆/10 beds) .2 .2 .1 .2 .1 Private FP ns ns 3 ns 4 Public -.7 -8 -7 -7 -9 ns -6 ns -7 pMaj. <0.05Teach unless marked “ns”