18. Which rate is right? Impact of improved decision quality on surgery rates: BPH Knowledge of relevant treatment options and outcomes Concordance between patient values and care received
19.
20. The Dartmouth Atlas Project: 306 hospital referral regions Ongoing Study of Traditional Medicare Population USA
21.
22.
23.
24.
25. Knee Replacement: An Example of Preference-Sensitive Care Ratio of knee replacement rates to the U.S. average (2005 ) 1 .30 to 1 .75 (46) 1 .10 to < 1 .30 (78) 0 .90 to < 1 .10 (106) 0 .75 to < 0 .90 (53) 0 .41 to < 0 .75 (23) Not Populated
26. Total Knee replacement for Arthritis per 1,000 Medicare enrollees among 306 Hospital Referral Regions Red dot = U.S. average: 4.03 5.64 40% increase 1.0 3.0 5.0 7.0 9.0 11.0 1992-93 2000-01
27. Relationship Between Knee Replacement Rates Among Hospital Referral Regions in 1992-93 and 2000-01 0.0 2.0 4.0 6.0 8.0 10.0 12.0 0.0 2.0 4.0 6.0 8.0 10.0 12.0 Knee Replacement (1992-93) Knee Replacement (2000-01) R 2 = 0.75
36. Bottom Line Implication: Clinical Appropriateness should be based on sound evaluation of treatment options (outcomes research) To Avoid Wrong Patient Surgery, Medical Necessity should be based on Informed Patient Choice among Clinically Appropriate Options
37.
38. Association between hospital beds per 1,000 and discharges per 1,000 among Medicare Enrollees: 306 Hospital Regions Hip Fracture R 2 = 0.06 All Medical Conditions R 2 = 0.54 0 50 100 150 200 250 300 350 400 1.0 2.0 3.0 4.0 5.0 6.0 Acute Care Beds Discharge Rate
39. Association between cardiologists and visits per person to cardiologists among Medicare enrollees: 306 Regions R 2 = 0.49 Number of Visits to Cardiologists 0.0 0.5 1.0 1.5 2.0 2.5 0.0 2.5 5.0 7.5 10.0 12.5 15.0 Number of Cardiologists per 100,000
40.
41.
42. Contrasting Practice Patterns in Managing Chronic Illness During Last Two Years of Life (Deaths 2001-2005) Regions in Highest and Lowest HCI Index Quintiles Source: Dartmouth Atlas Database Resource input Lowest Quintile Highest Quintile Ratio H/L Medicare $ per capita $38,300 $60,800 1.59 Physician Labor/1,000 All Physicians 16.6 29.5 1.78 Medical Specialists 5.6 13.1 2.35 Primary Care Doctors 7.4 11.5 1.55
43. Contrasting Practice Patterns in Managing Chronic Illness in Regions (HRRs) Ranked in Highest and Lowest Utilization Quintile (patients in their last 2 years of life) Low HRRs High HRRs Ratio H/L End of Life Care Hospital Days (L6M) 8.5 15.6 1.83 Hospital MD Visit (L6M) 12.9 36.3 2.82 % Seeing 10 or more MDs 20.8 43.7 2.16 % Deaths in ICUs 14.3 23.2 1.63
44. End of life care for Chronic Illness at selected academic medical centers (deaths 2001-05) Hospital Name NYU Medical Center UCLA Medical Center Brigham and Women's Johns Hopkins Tufts-New England Beth Israel Deaconess Boston Medical Center Massachusetts General Cleveland Clinic Mayo Clinic (St. Mary's) University of Wisconsin Total Medicare spending 105,068 93,842 87,721 85,729 85,387 83,345 79,672 78,666 55,333 53,432 49,477 % of deaths with ICU admission 35.1 37.9 26.2 23.2 28.5 23.5 28.6 22.5 23.1 21.8 16.1
Under the normative assumption that the “right rate” for a given procedure should be based on the choices made by informed patients (free of undue influence by the practice style preferences of their physicians or other unwarranted influences), the systematic implementation of decision aids among patient populations would offer the opportunity to obtain valid benchmarks for the “true” demand for a given treatment option. Such an opportunity presented itself to our research group in the early 1990’s when a decision aid we had designed to help patients decide between watchful waiting and surgery for their enlarge prostate was introduced in the urologic clinics in 2 pre-paid group practices, Kaiser-Permanente in Denver and Group Health Cooperative in Seattle. After the implementation of shared decision making, the population based rates for prostatectomy fell 40% , providing a benchmark for demand under shared decision making. (Rates in the control group, Group Health Cooperative’s Tacoma site, did not change.) giving us a benchmark for demand under shared decision making. When we compared this benchmark to the rates among the 306 region (blue dots in the above figures), it was of interest that the shared benchmark was at the extreme low end of the national distribution, suggesting that the rates of surgery in most US regions exceeded the amount that informed patients want.
The essence of practice variation studies is the comparison of rates of use of medical care among defined populations. Sometimes the “population at risk” is the resident population living in a region. For example, the incidence of Medicare hospitalizations for hip fracture is measured by counting the number of residents who were hospitalized in a given period of time (the numerator of the rate) and dividing by the number of Medicare enrollees living in the same region (the denominator). The rates for discretionary surgery in this lecture are calculated this way as are a few examples supply-sensitive care. Sometimes, the populations selected for comparison are those at the same stage in the course of illness or health care needs.. Most effective care quality measures are calculated this way. For example, the quality of care for diabetic patients measure used in this lecture is based on a numerator that is a count of all diabetic patients who received the needed eye examination at least once over a 2 year period and a denominator is a count of all diabetic patients living in the region. The measures of supply-sensitive care at the end of life are also based on the experience of specific subpopulation. In these cases, the numerator is the number of events experienced by patients during the last six months of their life; the denominator is the number of patients who died. In the lecture, practice variations were viewed two ways: (1) the traditional Atlas strategy which examines variation among Medicare residents living in 306 hospital referral regions across the United States. (2 A newer method which examines variation on a hospital-speciific basis among patients with chronic illness who receive most of their care from well known academic medical centers (selected because they appeared on US News and World Reports 2001 list for the “Best Hospitals” for geriatric care and for treating cancer, heart disease or respiratory disease.)
21 As the name implies, supply-sensitive services are related to the supply of the resource that provides the service. This figure shows the association between supply of staffed hospital beds per 1,000 residents and the hospitalization rate for medical (non-surgical) condition among Medicare enrollees. More than half of the variation in discharge rates is associated with bed capacity. By contrast, hospitalization for hip fracture--one of the few conditions for which the pattern of variation is determined by the incidence of illness--shows little correlation with resource supply. The denominator for the utilization rates is the Medicare population resident in the region; the denominator for beds per 1,000 is the entire population of the region. The behavioral basis of this association must rest in Roemer’s law--the long- held hypothesis that hospital beds, once built (and staffed), tend to be filled. In my experience, the impact of beds per capita on clinical decision making is subliminal in the sense that clinicians are unaware of differences in practice style associated with the context of bed capacity. I gained this impression from interviews with clinicians practicing in Boston and New Haven and who were not aware of the 60% differences in hospitalization rates for medical conditions, even though some had practiced in both communities.
23 This figure illustrates the relationship between the number of cardiologists per 100,000 and the number of visits per person to cardiologists among the 306 regions. About half of the variation is “explained” by supply. The behavioral basis of this association seems clear: The Medicare population comprise a large shared of the patient load for cardiologists. Appointments to see physicians characteristically are fully “booked”--very few hours in the work week go unfilled. Most office visits are for established patients and the interval between revisit is governed by the size of the physician’s panel of patients. On average, regions with twice as many cardiologists per 100,000 will have twice as many available office visit hours. In the absence of evidence-based guidelines on the appropriate interval between revisits, available capacity governs the frequency of revisit. The strength of the association between physician supply and physician visit rates among Medicare population depends on the specialty. The association between internists and visits to internists is similar to that of cardiologists (and, together, these 2 specialties account for N% of visits to primary care and medical specialists). However, for family practice physicians, the association is much weaker with only about X% of visits (R2 = .xx) I believe the likely explanation rests in the much small proportion of their total visits that family practice physicians dedicate to patients 65 years of age and older: XX% of family practice visits are for patients 65 years of age and older, compared to yy% for general internists. The denominator for physician supply is census count for the region; for Medicare visits it is the number of enrollees living in each region.