1. Quality of care: from theory to practice Kim A Eagle MD Albion Walter Hewlett Professor of Internal Medicine Chief, Clinical Cardiology Co-Director, Heart Care Program University of Michigan Ann Arbor, MI Mauro Moscucci MD Assistant Professor of Medicine Director, Interventional Cardiology Program University of Michigan Ann Arbor, MI

2. Factors that effect in-hospital mortality have been well defined. Risk of mortality can now be assessed on the basis of comorbid conditions. Standards for appropriate modeling, risk adjustment, and evaluation for percutaneous coronary intervention (PCI) have not been well developed In-hospital mortality PCI outcomes

3. Risk adjustment models Risk adjustment models have proven very effective in accounting for mortality rate  the northern New England risk adjustment model for in-hospital mortality  the Cleveland Clinic model Predictors of in-hospital mortality

4. New England model A prospective cohort study of in-hospital mortality after PCI in northern New England was conducted from 1994 to 1996. Data from 52 interventional cardiologists on 15 331 consecutive hospital admissions for PCI were collected (98.5% of all patients who underwent a PCI during the study period). The data were used to develop and internally validate a multivariate prediction equation for in- hospital mortality that required only routinely collected data known before the PCI. O’Connor, et al. J Am Coll Cardiol 1999;34:681-691 PCI and in-hospital mortality

5. New England model  older age  congestive heart failure  peripheral or cerebrovascular disease  increased creatinine levels  lowered ejection fraction  cardiogenic shock  acute myocardial infarction  urgent priority  emergent priority  preprocedure insertion of an intra-aortic balloon pump  PCI of a type C lesion O’Connor, et al. J Am Coll Cardiol 1999;34:681-691 Univariate assessment Variables associated with an increased risk of in-hospital mortality

6. New England model Variables included  age  indication for intra-aortic balloon pump (IABP)  procedural priority for IABP  and preprocedure use of an IABP  congestive heart failure  peripheral or cerebrovascular disease  elevated creatinine level  EF  intervention on a type C lesion O’Connor, et al. J Am Coll Cardiol 1999;34:681-691 Multivariate prediction equation Variables not included  sex  history of MI  use of preprocedure intravenous nitroglycerin  LVEDP  number of diseased coronary arteries  intervention on a proximal left anterior descending coronary artery

7. Cleveland Clinic model Data from 12 985 consecutively treated patients were taken from quality-controlled databases at 5 high-volume centers. Multivariable logistic regression models were used to examine individual and interaction relations between baseline characteristics of patients and death and also the composite of death, Q-wave infarction, or bypass surgery. These models were used for risk adjustment, and the relations between both yearly caseload and years of interventional experience and the 2 adverse outcome measures were explored for all 38 physicians who performed at least 30 procedures per year. Ellis SG, et al Circulation 1997;95:2479-2484 Model predictive of death after PCI

8. Cleveland Clinic model Risk-adjusted measures of both death and the composite adverse outcome were inversely related to the number of procedures each operator performed annually, but were not related to years of experience. High-volume operators had a lower incidence of major complications than did lower-volume operators, but the difference was not consistent for all operators. Both adverse outcomes were more closely related to the logarithm of caseload (for death, r=.37, p=0.01; for death, Q-wave infarction, or bypass surgery, r=.58, p<0.001) than to linear caseload. Results Ellis SG, et al Circulation 1997;95:2479-2484

9. Mathematical models Mathematical formulas are used to calculate the expected mortality rate of an institution. The formulas make adjustments for patient population and compare expected and observed mortality rates. The expected mortality rate of an institution serving a high-risk population is not necessarily higher than that for an institution serving a low- risk population. Expected mortality rates

10. University of Michigan Every operator receives cardiac reports that include observed and expected mortality rates and the baseline comorbidities of his or her patients. A multicenter registry is used to provide the same type of feedback to operators from 6 other hospitals in Michigan. Operator feedback

11. Mathematical models Modern mathematical science gives physicians outcome data that is mathematically robust in terms of risk assessment. Because the mathematical model provides an accurate estimate of patient risk, risk- adjusted data can be used to help patients understand the risks of certain procedures. Statistical confidence

12. Mathematical models The mathematical models may identify situations in which the expected risk of death may be so high as to render care futile. In such situations, a realistic estimate of the likelihood of death can be provided to the patient, so the patient will not have unwarranted expectations. Predicting death

13. Mathematical models A model has been developed to assess the risk of mortality in acute renal failure patients in ICU requiring dialysis. Significant factors  male gender  respiratory failure requiring intubation  hematologic dysfunction  bilirubin < 2.0 mg/dL  the absence of surgery  serum creatinine on the first dialysis treatment day  an increasing number of failed organ systems  an increased BUN from the time of admission Application to other conditions Paganini EP, et al. Clin Nephrol 1996;46(3):206-211

14. A controversy Several studies have shown that there appears to be a relation between operator volume and outcome. However, with new technology (particularly coronary stent) even low-volume operators can still have a good outcome. Operator volume and outcome

15. ACC recommendations Statistical data support the premise that a physician who performs coronary interventional procedures infrequently is unlikely to be as proficient as one who performs them often. The low-volume operator has fewer opportunities to maintain skills, and is less able to acquire the additional skills needed to become proficient in the use of new techniques and devices. The low-volume operator is likely to be less experienced at recognizing and managing procedural complications. Statistical data demonstrate that operators who perform <75 procedures annually have the highest complication rates; this trend is most pronounced in institutions with an annual procedural volume <600. Hirshfeld JW, et al. J Am Coll Cardiol 1998; 31(3):722-743 Coronary interventional procedures

16. Volume and outcome Evidence suggests that quality might be acceptable for operators who perform fewer than 75 procedures annually but who do them in a high-volume center. There is substantial evidence suggesting that the introduction of new technology such as coronary stent has led to a significant improvement in acute outcome. Quality of procedures

17. Understanding process Analyzing practice variations among operators and among institutions is very beneficial. Benchmarking and comparisons identify differences among operators and institutions. Once areas needing improvement are identified, processes can be studied to determine why discrepancies exist and changes can be implemented. Benchmarking

18. University of Michigan A team visits other sites to evaluate and compare processes. Problem areas and important differences have been identified using this strategy. Team cohesion is fostered that is reflected in in-hospital work. Team members  a physician  a cath lab technician  a cath lab nurse  a nurse manager Benchmarking among cath labs

19. University of Michigan The common goals of the team make implementing changes easier. Knowing that a particular process is working elsewhere makes workers less resistant to change. Implementing changes

20. University of Michigan Cost procedure analysis  higher use of coronary stents  higher use of GP IIb/IIIa receptor blockers (abciximab)  much higher use of expensive devices Total lower procedure cost  decreases in vascular complications  decreases in number of transfusions  decreases in number of emergency bypasses  decreases in length of stay  fewer emergency cath lab procedures  potential for less restenosis and fewer repeat procedures Justifying new technology

21. University of Michigan Collaborative effort with department of clinical affairs. Annual cost data are available on all procedures performed (direct and indirect costs). Cost data are linked to outcome database. Reliable cost data

22. University of Michigan Interventional cardiology program will have access to clinical, financial, and patient data. Areas where costs are the result of practice variation, not actual procedures, can be identified. Costs can be closely monitored and areas can be identified where costs can be reduced. New database in development

23. Assessing appropriateness If a procedure is not appropriate, the care provided is not good care. Some payers that are beginning to assess appropriateness in a clinical fashion. In Michigan, Blue Cross is assessing the appropriateness of procedures performed in the past 2 years by applying criteria based on national guidelines. Indicator of procedure performance

24. Assessing quality Having a data base is not the final answer  a good system to collect data is required  data must be assessed in a clinical fashion  problem areas identified must be addressed By examining, benchmarking and steadily improving, change is possible. High-cost / high-risk procedures