describes the distribution of costs for six chronic conditions in each of two categories: typical services and PACs. Approximately 30 cents of every dollar spent across these six conditions were consumed by PACs. The average percentage of costs for PACs relative to total costs for each ECR varied from 14.8 percent for hypertension to over 55 percent for CHF. By definition, hospitalizations and related professional services accounted for the majority of the PAC costs. Additionally, for most chronic conditions, about 20 percent of PACs were related to the index condition, 60 percent were due to comorbidities, and another 20 percent were due to patient safety failures.
illustrates the results of the regional variation analysis. For most of the conditions studied, there is a twofold difference in average PAC rates from states with lowest rates compared with those with the highest. These statewide PAC rates represent averages in each state, and within each state there is also significant variation. shows the potential savings that could be achieved across all the six conditions for the 688,550 patients studied. Similar to the work performed by the Dartmouth Atlas, reducing PAC rates in states with rates above the second decile in each of the conditions to second decile levels would yield U.S.$116.7 million, or 3.8 percent of the U.S.$3.1 billion total costs. This amount extrapolates to U.S.$6.5 billion in total national savings when considering a privately insured base of 200 million Americans. Because we did not study Medicare- or Medicaid-insured Americans, we cannot estimate the potential savings in that population.
Variation in Average Potentially Avoidable Complications Rates across States Ranked in Deciles
Savings Opportunity by Chronic Conditions When Reducing PAC Rates to Second Decile Values for States with PAC Rates above the Second Decile
summarizes our estimated results for a practice paid under the Prometheus Payment model compared with fee-for-service. The input variables are an estimate of the number of commercially insured patients with a chronic condition and their distribution in a practice. Two other input variables are the current observable PAC rates and target PAC rates. For each we used the actual observed in this study as reported in and the targets from . The ECR costs, FFS costs, actual PAC occurrence, and PAC costs are derived from our analysis of the claims database. In this model, the practice's risk associated to the management of patients with asthma are significant, representing the relatively small additional margin that could be derived from an ECR payment for those patients, relative to the potential gain from reducing PAC rates from current levels. Overall, the practice's net gain would be close to U.S.$20,000. However, there is no estimate here of the expenses that would likely be incurred by the practice in order to improve the management of patients. Further, because extrapolations are inherently uncertain, a more in-depth evaluation of the Prometheus Payment model is crucial to assess its effects.
Results of Modeling the Net Effect of the Prometheus Payment Model on a Practice with 1,000 Chronic Care Patients
We used a systematic review of the literature to identify the percent of events stemming from care failures that might be avoided for the chronic conditions studied to determine whether PAC rates could be compressed to yield savings for payers and margin opportunities for providers under the Prometheus Payment model. We found that several studies have shown reductions in avoidable complications for patients with the chronic illnesses studied, albeit in very varied clinical settings.
For example, a study in Germany (Michalsen, König, and Thimme 1998
) and a more recent one from Rich et al. (1995)
showed that the number of hospital admissions and readmissions for heart failure could be reduced by more than 50 percent with appropriate patient management and interventions. In addition, AHRQ recently reported that for patients with chronic illnesses, CHF, and bacterial pneumonia were the two most common causes of potentially preventable hospitalizations and accounted for half of the avoidable hospitalization costs (Jiang et al. 2006
). In their studies, Yuen (2004)
, Ahern (2007)
, and Kim (2007)
demonstrated that 32–36 percent of all diabetes-related hospitalizations were related to short-term complications and uncontrolled diabetes and were considered avoidable using criteria from AHRQ. In addition, Kim estimated these avoidable hospitalizations led to an economic burden of U.S.$2.4–U.S.$2.8 billion. Furthermore, complications from diabetes could also be significantly reduced. Narayan et al. (2000)
demonstrated that good glycemic and blood pressure control decreased the development of vascular disease by 25–35 percent and good lipid control decreased coronary events by 25–55 percent in diabetic patients. Similarly, a large study in the United Kingdom (U.K. Prospective Diabetes Study Group 1998
) showed that tight blood pressure control in patients with type 2 diabetes led to a 44 percent decrease in incidence of stroke and a 50 percent decrease in major cardiovascular events. Narayan et al. also reported that regular eye exams could lead to a 60–70 percent decrease in serious vision loss, and a study by Aiello (2001)
reported that intensive blood sugar control could lead to risk reductions ranging from 52 to 75 percent in development of severe retinopathy or the need for laser surgery. These studies also showed that pneumococcal vaccination could lead to a 32 percent decrease in pneumonia-related hospitalizations and a 64 percent decrease in development of respiratory conditions in diabetics. Finally, tight blood pressure control was also important in reducing the development of nephropathy (Ritz and Orth 1999
), and Lavery, Wunderlich, and Tredwell (2005)
demonstrated that good management of diabetic foot ulcers led to a 38 percent decrease in hospital admissions and a 47 percent decrease in foot amputations.
An age- and sex-adjusted population-attributable relative risk study by Li et al. (2005)
demonstrated that 45 percent of strokes in hypertensive patients might be attributable to uncontrolled blood pressure (≥140/90). These authors suggest that a substantial portion of first-time strokes in hypertensive patients may be preventable with better management of hypertension.
The principal reasons for admission for COPD or asthma are either exacerbation of the underlying condition or pneumonia (Merrill, Stranges, and Steiner 2008
; Stranges, Merrill, and Steiner 2008
;). Aggressive pharmacological management of COPD and asthma has helped reduce hospitalizations or ED visits by 31–56 percent (Akazawa et al. 2008
), and in their study Flores and Abrev (2005)
reported that 15–54 percent of asthma-related hospitalizations in children could have been prevented through better education of the family, closer outpatient follow-up by PCPs, and avoiding known disease triggers.
Umscheid et al. (2008)
used 2002 estimates of hospital-acquired infections (HAI) and determined the range of HAI risk reductions from U.S. studies (Ranji et al. 2007
). They report that 18–82 percent of blood-stream infections, 46–55 percent of ventilator-associated pneumonia, 17–69 percent of urinary tract infections, and 26–54 percent of surgical site infections are preventable. Healy et al. (2002)
analyzed complications in hospitalized surgical patients and reported that between 39 and 61 percent of major complications (wound infections, pneumonia, urinary tract infections, arrhythmias, respiratory failure, gastrointestinal complications, and deep vein thrombosis) and about an equal percent of minor complications could have been avoided. The National Pressure Ulcer Advisory Panel reported in 2001 that pressure ulcer prevention programs had reported 50 percent or greater reductions in facility-acquired pressure ulcers (Cuddigan, Berlowitz, and Ayello 2001
). Similarly, appropriate prophylaxis could reduce the risk of venous thromboembolism by 45 percent in acutely ill medical patients (Leizorowicz et al. 2004
), and a recent study found a 50 percent reduction in thromboembolic events with extended pharmacologic prophylaxis (Hull et al. 2007
). Adequate evidence-based treatment protocols in preventing contrast nephropathy and adequate drug dosing have demonstrated a risk reduction between 52 and 90 percent in the incidence of acute renal failure in patients in the intensive care unit (Singri, Ahya, and Levin 2003
). Additionally, a study of the use of a hospital electronic medical system with prompts for nursing care protocols demonstrated that infection rates dropped 88 percent, bedsores were reduced, and compliance to guidelines for care of patients on ventilator increased by 77 percent (Landro 2009
Overall, our review of the literature suggests that current PAC rates might be reduced by about 50 percent for CHF and CAD, 40 percent for diabetes, 60 percent for COPD and asthma, and 75 percent for hypertension, as compared with the current rates. These potential reductions are significantly higher than what we used in our estimations of both (1) cost savings if providers in states with high average PAC rates were to decrease them to second decile levels, and (2) our analysis of the prototypical practice. However, there is no evidence that the results achieved in these settings could be replicated widely across the U.S. delivery system.