Agency for Healthcare Research and Quality Advancing Excellence in Health Care • www.ahrq.gov Hospital Readmissions, Cost/Year and Multiple Chronic Conditions Bernard Friedman Joanna Jiang Anne Elixhauser _ June 25, 2006 1 Objectives Examine “complexity” of chronic illness (number of different chronic conditions) in relation to: number of admissions per person in a year total cost per person over a year Test and control for other contributing factors patient, hospital, area characteristics Compare persons with “early” readmissions (within 30 days) vs. persons with other readmissions 2 Background IOM, Chasm study (2001): “optimizing care for patients with several chronic conditions…an extraordinary challenge...” Literature on Disease Management gives mixed results and modest savings (CBO, 2004) Some optimistic demonstrations on coordination of care for chronic illness (Friese, 1998; Chen 2000) Recent pessimism about Medicare demonstrations for chronically ill (Gold, 2005) 3 Targeting Interventions Health plans are concerned about cost of opening the benefits to new types of services for chronic illness. Suggestion from our recent study (MCRR, 2006): target persons when initial hospital stay indicates 3, 5 or more different chronic conditions. But how many of observed stays are actually readmissions? What proportion of persons have X+ different chronic conditions when first hospitalized? How many readmissions occur afterward and at what cost? 4 Population and Data Sources HCUP State Inpatient Databases for: California Nebraska New York Pennsylvania Tennessee Utah Adults, hospitalized during Jan.-Nov. 2002 No maternity-related stays 5 Measurement Issues Chronic versus non-chronic conditions at ICD-9-CM level http://www.hcup-us.ahrq.gov/toolssoftware/chronic/chronic.jsp Grouped to identify “different” chronic conditions no more than one per Clinical Classification Software (CCS) or CCS cluster Hospital cost: CCR based on accounting reports to CMS Person identity: supplied by states and verified (age, gender) Episodes -- include same-day transfers to another hospital Cannot observe death outside hospital 6 Definitions of chronic condition Lasts at least 12 months or longer AND Places limitations on task performance, self-care, independent living, social interactions OR Results in need for ongoing medical intervention (meds, therapy, special equipment, special schooling, protocols affecting diet or activity Hwang et al. Concurrent diseases that impose additional clinical vulnerability in the patients, and that may in turn, influence resource use Silber et al. 7 Measurement Issues Chronic versus non-chronic conditions at ICD-9-CM level http://www.hcup-us.ahrq.gov/toolssoftware/chronic/chronic.jsp Grouped to identify “different” chronic conditions no more than one per Clinical Classification Software (CCS) or CCS cluster Hospital cost: CCR based on accounting reports to CMS Person identity: supplied by states and verified (age, gender) Episodes -- include same-day transfers to another hospital Cannot observe death outside hospital 8 Overview of Study Population No readmissions N (%) Any readmission 3,517,898 (79.8%) 889,310 (20.2%) N readmissions 0 1.51 Hospital cost/yr $7,377 $19,641 Days in hospital 5.7 16.9 N chronic conditions 2.9 3.6 20.6% 31.4% 59.4 64.1 Female 55.0% 53.5% APR-DRG 3-4 17.3% 27.6% Elective admission 26.7% 19.0% Major ther. procedure 36.2% 23.7% % with 5+ chronic conditions Age, years 9 Characteristics of Patients with Readmissions Early readmit Readmit > 30d both 376 (42.3%) 358 (40.2%) 155 (17.4%) N readmissions 1.29 1.13 2.92 Hospital cost/yr $18,681 $15,925 $30,629 Days in hospital 16.6 13.2 26.1 N chronic conditions 3.5 3.6 3.8 % with 5+ chronic cond. 30.7% 30.3% 35.4% Age 63.8% 64.5% 63.8% Female 52.7% 54.6% 52.7% Severity level 3 or 4 29.4% 24.1% 31.1% Elective admission 18.3% 21.3% 15.7% Major ther. procedure 20.0% 18.4% 20.4% N in 1000s (%) 10 Risk of Readmission by Principal Diagnosis at Index Admission Principal Diagnosis at Index Admission Persons (1000s) Risk of any readmit RR early vs. other readmit Mental disorders Malignancy Coronary atherosclerosis Pneumonia 357 304 244 227 40.7% 42.5% 38.9% 42.6% 0.21 0.27 0.24 0.20 CHF Stroke Chest pain 223 203 188 59.4% 33.3% 27.9% 0.16 0.25 0.17 Asthma Conduction/dysrhythmia AMI 181 168 139 48.8% 39.1% 36.2% 0.17 0.21 0.36 11 Multivariate Regression: Adjusted Cost per person in 2002 Cost of each episode divided by index of relative cost of discharge with the principal DX Then Adjusted Cost is summed over episodes Semi-log model log(Adj. Cost)= α*log(Wage) + β*log(months) + linear {X} 12 Influences on Cost/Person, Based on Multivariate Model Patient Level Months observed Severity of illness Medicare or Medicaid Self-pay Secondary cancer dx Secondary HTN dx Secondary DM dx Hospital Level % admitted from ED + Wage index (hospital area) + Large hospital + Investor-owned hosp Teaching hosp + Safety indicators (hosp) + Area Level + N hosp beds (area) N primary care MDs (area) Quality indicators (area) + ns + + ns ns ns 13 ns Percent Increase in Cost per Person, by Number of Chronic Conditions N of different chronic conditions % increase in cost for year 2 % of cases 18.5 3 17.1 7.6 4 14.4 9.6 5 10.4 11.0 6 6.4 13.2 7+ 5.8 17.9 6.4 Persons=774,000; hospitals=1029, all results P<.001 14 Simulation of Potential Savings If the cost and readmissions for adults with 5+ different chronic conditions were reduced to level of other adults Target: 23% of all defined adults first admitted during Jan-Nov. of 2002. Control for all other determinants in the regressions. Potential savings: $3,548 per person in target group (9% of total resource costs for entire population covered) Extrapolated national savings: $22.7 billion in hospital costs 15 Issues Number of chronic conditions Some may not be diagnosed at the index admission Some may be known but not recorded Some may emerge during the year Lost some persons due to death outside the hospital. Additional tests of model specification. 16 Discussion Hospital admission is a prime opportunity for systematic and thorough diagnosis Even “early” readmissions are associated with complexity of chronic conditions Meaningful potential savings within a year How? educational services, pharmacy guidance, “coordination” (Lorig; Chen) Limited number of persons in target population 17