Section 4. Recommendations for Next Steps
Cost of Poor Quality or Waste in Integrated Delivery System Settings
Our three-level model offers a strong framework to estimate total waste within health care delivery. It also helps identify several significant gaps, and thus suggests "next steps" in pursuing this useful topic. Specifically, we describe opportunities for furthering our work at each of the three levels in the overarching framework—population level, episode level, and patient level.
At the Population Level (Level 1), we relied on the Dartmouth Atlas's recent calculations of unnecessary specialty visits, testing, hospitalization, and ICU admits, to estimate overuse. The Dartmouth Atlas tools provide hospital-level estimates of clinical waste associated with unnecessary hospitalization. Wennberg and colleagues make compelling arguments that the patterns of overuse they identified within the Medicare program reflect general patterns of overuse in care delivery that extend to commercial insurance as well. However, that extension has not been empirically tested. In addition, the current Dartmouth Atlas analysis does not directly assess overuse associated with preference-induced demand. Preliminary, unpublished studies suggest that utilization for some preference-associated treatments may fall by as much as 50 percent when patients are given full information about their choices. However, the impact of better patient decision support on the total costs of care delivery has not been rigorously assessed. We note that Wennberg and colleagues at the Center for Evaluative Clinical Sciences (or CECS Group) are currently pursuing these analyses, which may be an opportunity to collaborate.
At the Episode Level ("number of units per case"—Level 2), we catalogued a large number of improvement projects that appeared to significantly reduce health care costs. Categorizing and generalizing this class of projects, to generate a broad framework that would lead to general estimates of this class of waste, remains the most challenging part of the project. It also presents a very promising avenue for future work.
For example, it may be possible to further classify Level 2 into quality waste (process failures, or defects) versus inefficiency waste, and processes that are inherently clinical versus those that supply subcomponents to clinical care. Care-associated clinical defects link this work to our parallel AHRQ Targeted Injury Detection System (TIDS) project. It appears that our TIDS work will provide a reliable lower bound estimate of broadly defined injury rates in inpatient settings. Additional research to estimate the marginal costs associated with different levels and types of injury could lead to reliable estimates of care-associated waste for this entire subcategory. Such estimates could form a critical piece of the business case for patient safety, and a major subcomponent of developing pay-for-performance efforts. Similar generalizations around inefficiency waste and quality waste in nonclinical hospital operations may also lead to detection and management tools.
Application of Toyota Production System (TPS) observation approach at the Patient Level (Level 3) is perhaps the most useful part of the current work. We developed a set of structured observation tools that were used to demonstrate that inefficiencies at the point of care are very common, offering large opportunities to streamline care processes. When a care delivery group uses our toolset to detect such waste, they can produce not just estimates of the size of competing opportunities for improvement, but also generate detailed knowledge about the nature of the failures that produce the waste, leading to testable changes that result in savings when addressed. Such savings represent not just financial resources but also staff resources. For example, reducing wasted time and effort on a nursing unit effectively expands the capacity of that unit. In the face of severe nursing shortages, waste elimination at this level has the same impact on patient care as increasing the number of nurses, but without additional costs. While we tested our TPS observation tools on a reasonable number of work processes within Intermountain Healthcare and validated their use on a small scale at the University of North Carolina Hospital, a definitive analysis will require a broader scope replicated at several institutions.
If we are able to broaden and deepen our present work, we will be able to much more accurately estimate total waste within health care and do so in a more convincing manner. More importantly, our Level 2 projects suggest that such analysis can lead to effective improvement action. The three levels of our waste model are functionally independent, in the sense that patient injuries (Level 2) are as likely to happen with unnecessary, "supply-induced demand" care (Level 1), as with value-adding care; and inefficiencies in front-line staff functions (Level 3) apply whether treating an avoidable complication (Level 2), delivering unnecessary care (Level 1), or delivering value-added care. A complete waste model would draw from all three levels. For example, a combined final estimate might look something like this:
At Level 1, supply-induced demand and preference-induced demand are complementary categories. Thus, total value added resource consumption can be estimated as:
ignoring preference-induced demand, for which estimates are not presently available. For Level 2, clinical quality waste can be estimated as:
(1 - marginal costs associated with treatment of avoidable injuries / total care costs)
Level 3, front-line waste in care delivery performance can be estimated as:
(1 - front-line time wasted / total work time) = (1 - 0.35) = 0.65
The total "value added" benefit of a health care delivery system is then the product of the three categories. For the estimates we have generated to date, while conservatively ignoring preference-induced demand and all Level 2 opportunities, this computes to 0.68 x 0.65 = 0.44. In other words, by our current very conservative estimates, only 44 percent of all resources consumed in health care delivery add value. Thus, 56 percent—more than half—represents potentially recoverable waste. We need to fill in the rest of the equation. Providing estimates in current gap areas has the potential to spur broad-spread waste reduction action; help to more clearly identify specific targets for savings; and enable those engaged in waste elimination to track which of their efforts are successful.