Section 3. Examining Waste/Poor Quality

Our examination of waste/poor quality in health care is guided by the Chain of Effect for Quality as described in Section 2.4 and depicted in Exhibit 3. Specifically, we focus on three distinct levels to exemplify approaches for culling out waste/poor quality in health care: the Population Level, the Microsystem/Episode Level, and the Patient (care delivery) Level (Exhibit 7). These approaches for examining waste are linked to our analytic approach in Exhibit 8. For these three areas, we provide a more detailed description of the approach together with specific examples of how these approaches can be applied (and estimates when appropriate).

Exhibit 7. Identifying waste by levels

Approach	Level to Address Waste
National rates of potential overuse and underuse	Population Level
Analysis of waste represented by overuse, as reflected in regional variability	Population Level
Analysis of the cost of errors and adverse events	Microsystem/Episode Level (health delivery organizations)
Analysis of ATP, Six Sigma target areas	Microsystem/Episode Level (health care delivery organizations, microsystems)
Lean™ Observations	Patient Level (microsystems, individuals)

Exhibit 8. Efficiency/waste metric and analytic approach

Inefficiency/Waste Metric	Analytical Approach
	Dartmouth Atlas	Six Sigma, ATP	TPS	Cost of Unusual Occurrences
Cost per unit	No	Yes	Yes	No
Units per case	No	Yes	Yes	Yes
Number of cases	Yes	No	No	No

Return to Contents

3.1. Estimating Waste at the Population Level—the Dartmouth Atlas

In 1973, Wennberg and Gittelsohn published a seminal paper documenting massive geographic variation in the rates at which Americans are hospitalized. He later organized the Center for the Evaluative Clinical Sciences (CECS), based at Dartmouth University in Hanover, New Hampshire. In 1999, CECS first published the Dartmouth Atlas, which documents ongoing, massive variation in care delivery across the United States, as reflected in hospital admission rates. The CECS group has updated the Dartmouth Atlas on a 2- to 3-year schedule ever since.

Several years ago, Wennberg and the CECS group shifted the focus of their variation research to Medicare patients at the end of life (Wennberg et al., 2002). They have shown that (1) Medicare patients exhibit a very high degree of loyalty to a particular hospital and the physicians associated with that hospital, with about 90 percent of all services to a particular patient, on average, delivered through a single facility; and (2) care within a particular facility is consistent over time, despite major geographic variation among facilities. For example, health resource consumption in the last 2 months before death, which accounts for more than 20 percent of all Medicare expenditures, accurately predicts care resource consumption during the last 6 months of life, and accurately predicts health resource consumption during the last 2 years of life, well beyond the period associated with a final terminal illness. The CECS approach thus provides a unique method to account for and remove patient factors when measuring differences in clinical or cost outcomes. It directly adjusts for age, gender, ethnicity, and chronic disease burden, but then centers on patients who all are at an equivalent health outcome—all patients in this class have died. It then demonstrates that resource consumption at a particular hospital for this class of patients accurately predicts resource consumption at the same facility in other time periods before terminal illness (Wennberg, 2005; Wennberg et al., 2002).

The CECS analysis identified three classes of care: effective care, supply-sensitive care, and preference-sensitive care. "Effective care" refers to disease entities with good evidence for effective treatment, such as total hip arthoplasty following hip fracture. Effective care tends to be generally underutilized, but shows very little geographic variation. "Supply-sensitive care" centers on 13 chronic conditions, accounting for more than 50 percent of all Medicare expenditures (Exhibit 9). "Preference-sensitive care" centers on an additional 10 elective surgical procedures (Exhibit 10), where patients have legitimate choices among surgical, medical, or "watchful waiting" options. Together, supply-sensitive and preference-sensitive care account for almost 80 percent of all hospital-associated Medicare expenditures and explain the high rates of cost variation that the CECS group has documented. High resource utilization directly and significantly correlates with high rates of patient visits to specialists, with higher testing rates, higher hospital use rates, and more days spent in intensive care units. In turn, utilization of each of those resources was very strongly correlated with their availability in the geographic community. Fisher et al. (2003) demonstrated that high utilization for supply-sensitive conditions is associated with worse medical outcomes (mortality rates about 2 percent higher than expected).

Exhibit 9. Supply-sensitive chronic conditions (as identified by the Dartmouth CECS group)

Supply Sensitive Chronic Conditions	Percentage of Medicare Patients Exhibiting Each Condition at the End of Life
Congestive heart failure	32.7%
Cancer: solid tumors	27.6%
Chronic pulmonary disease	22.5%
Dementia	14.8%
Nutritional deficiencies	10.5%
Coronary artery disease	8.6%
Chronic renal failure	5.9%
Cancer: lymphomas and leukemias	5.5%
Peripheral vascular disease	5.2%
Functional impairment	2.6%
Diabetes with end organ damage	2.5%
Severe chronic liver disease	2.0%
AIDS	0.1%

Exhibit 10. Preference-sensitive conditions associated with elective surgical procedures (as identified by the Dartmouth CECS group)

Benign prostatic hypertrophy Abnormal uterine bleeding/uterine fibroids (hysterectomy) Breast cancer Early breast cancer: chemotherapy and hormone therapy Early breast cancer: breast-sparing surgery Breast cancer: breast reconstruction Breast cancer: DCIS Breast cancer: metastatic phase Coronary artery disease Treatment choices Chronic care Colorectal cancer screening Low back pain Acute management Chronic low back pain: treatment choices Herniated disc Spinal stenosis Hip osteoarthritis Knee osteoarthritis Prostate cancer/PSA testing Bariatric weight-loss surgery

In the latest release of the Dartmouth Atlas (Wennberg et al., 2006), the CECS group identified two integrated delivery systems—the Mayo Clinic hospitals and Intermountain Healthcare—as having the most efficient care delivery in the nation and excellent associated clinical outcomes. While their analysis focused exclusively on supply-sensitive conditions, they estimated that total Medicare costs to the nation (Part A and Part B) would fall by 32 percent if all other hospitals were to adopt similar care patterns. Preference-sensitive conditions offer even more savings. Internal CECS data suggest that, when patients are offered complete, fair, and unbiased choices around preference-sensitive conditions, procedure utilization rates fall by 20 percent to 60 percent.

We classified Dartmouth Atlas estimates of supply-induced demand as Population Level waste: health services that consumed resources but did not improve patients' clinical outcomes. To the extent that care practices for Medicare patients represent general patterns of care delivery and resource consumption, as the CECS analysis suggests, Medicare-associated waste levels would apply to total health care costs for the country. Supply-induced demand represents Chassin's "overuse" category. In Deming's scheme, it is inefficiency waste. It results from increased frequency of unit and bundled services, as described above.

The same Dartmouth Atlas report extended the CECS supply-sensitive variation measurement methodology to the level of individual hospitals (Wennberg et al., 2002). The Dartmouth Atlas Web site (http://www.dartmouthatlas.org) allows hospitals to estimate their rates of supply-induced overuse waste. The focus of our report is hospital care alone, so we excluded potential waste in Medicare Part B physician visits. On the basis of hospital costs alone, if the entire Medicare program achieved the blended Mayo Clinic/Intermountain Healthcare utilization rates, then Medicare Part A costs would fall by 14 percent.

Exhibit 11 presents the CECS methodology and provides blended base utilization rates for the combined Mayo Foundation and Intermountain Healthcare systems, compared with national averages. Exhibit 12 applies the Mayo-Intermountain baseline to the Medicare total hospital reimbursements for Providence Health System facilities to demonstrate how the method could be used to estimate Population Level waste for any U.S. hospital. These estimates are conservative, in that they exclude Medicare Part B payments, where the effect is larger; and they do not (yet) include estimates of overuse associated with preference-induced demand.

Exhibit 11. Medicare hospital reimbursements at Mayo Foundation and Intermountain Healthcare Hospitals (1999–2003 data)

Hospital/System	Location	Loyalty	Deaths	Per Decedent			Ratio to U.S. Average
				Total Reimbursement	Hospital Days	Reimbursement Per Hospital Day	$	=	Days	x	$/Day
Rochester Methodist Hospital	Rochester, MN	79.4	889	37,233	27.0	1,377	1.52	=	1.13	x	1.35
St. Luke's Hospital	Jacksonville, FL	88.0	2204	24,906	20.9	1,193	1.02	=	0.87	x	1.17
St. Mary's Hospital-Rochester	Rochester, MN	86.7	4314	29,733	22.9	1,299	1.21	=	0.96	x	1.27
Mayo Clinic Hospital	Phoenix, AZ	84.6	1484	24,240	18.6	1,304	0.99	=	0.78	x	1.27
Austin Medical Center	Austin, MN	89.6	953	21,845	19.9	1,097	0.89	=	0.83	x	1.07
Luther Hospital	Eau Claire, WI	93.1	1690	19,001	17.2	1,106	0.78	=	0.72	x	1.08
Franciscan Skemp-La Crosse	La Crosse, WI	95.1	1326	19,345	17.6	1,098	0.79	=	0.74	x	1.07
Fairmont Community Hospital	Fairmont, MN	91.7	584	18,249	17.8	1,027	0.75	=	0.74	x	1.00
Immanuel-St. Joseph's	Mankato, MN	92.8	1485	18,408	20.1	917	0.75	=	0.84	x	0.90
Naeve Hospital	Albert Lea, MN	92.5	920	18,129	16.3	1,110	0.74	=	0.68	x	1.09
Myrtle Werth Hospital	Menomonie, WI	91.3	510	17,429	15.4	1,133	0.71	=	0.64	x	1.11
Lake City Medical Center	Lake City, MN	89.8	161	20,383	17.0	1,198	0.83	=	0.71	x	1.17
Waseca Medical Center	Waseca, MN	82.1	207	18,733	14.0	1,342	0.76	=	0.59	x	1.31
Bloomer Medical Center	Bloomer, WI	85.3	172	18,218	16.4	1,113	0.74	=	0.69	x	1.09
Floyd County Memorial Hospital	Charles City, IA	87.9	337	17,633	16.5	1,068	0.72	=	0.69	x	1.04
Barron Memorial Medical Center	Barron, WI	88.3	315	16,456	14.6	1,127	0.67	=	0.61	x	1.10
Springfield Medical Center	Springfield, MN	92.7	234	16,344	14.5	1,126	0.67	=	0.61	x	1.10
St. James Health Services	St. James, MN	90.9	140	15,961	15.9	1,006	0.65	=	0.67	x	0.98
Franciscan Skemp-Sparta	Sparta, WI	84.2	192	13,721	12.4	1,107	0.56	=	0.52	x	1.08
Mayo Foundation Totals		88.9	18117	23,430	19.7	1,188	0.96	=	0.82	x	1.16
LDS Hospital	Salt Lake City, UT	90.1	1863	22,326	18.0	1,238	0.91	=	0.75	x	1.21
Dixie Regional Medical Center	St. George, UT	94.0	1665	20,135	17.2	1,173	0.82	=	0.72	x	1.15
Utah Valley Regional Med Center	Provo, UT	93.4	1647	20,392	19.2	1,062	0.83	=	0.80	x	1.04
Alta View Hospital	Sandy, UT	86.9	734	18,642	15.8	1,183	0.76	=	0.66	x	1.16
McKay-Dee Hospital Center	Ogden, UT	94.8	1726	18,796	15.5	1,214	0.77	=	0.65	x	1.19
Cottonwood Hospital	Murray, UT	90.2	1263	18,037	16.8	1,074	0.74	=	0.70	x	1.05
Logan Regional Hospital	Logan, UT	93.1	741	18,008	15.5	1,164	0.74	=	0.65	x	1.14
American Fork Hospital	American Fork, UT	91.4	547	16,429	12.0	1,365	0.67	=	0.50	x	1.33
Cassia Regional Medical Center	Burley, ID	92.7	482	17,301	19.4	893	0.71	=	0.81	x	0.87
Valley View Medical Center	Cedar City, UT	91.8	287	17,483	11.9	1,475	0.71	=	0.50	x	1.44
Garfield Memorial Hospital	Panguitch, UT	92.0	103	17,207	13.2	1,308	0.70	=	0.55	x	1.28
Heber Valley Medical Center	Heber City, UT	88.2	88	16,974	11.9	1,424	0.69	=	0.50	x	1.39
Delta Community Medical Center	Delta, UT	89.1	95	16,910	13.6	1,246	0.69	=	0.57	x	1.22
Sanpete Valley Hospital	Mt Pleasant, UT	89.3	117	15,685	12.5	1,255	0.64	=	0.52	x	1.23
Sevier Valley Hospital	Richfield, UT	90.6	263	14,255	12.2	1,166	0.58	=	0.51	x	1.14
Intermountain Healthcare Totals		92.0	11621	19,254	16.6	1,163	0.79	=	0.69	x	1.14
Combined Mayo/Intermountain		90.1	29738	21,798	18.5	1,178	0.89	=	0.77	x	1.15
U.S. Average				24,491	23.9	1,023	1.00	=	1.00	x	1.00

Exhibit 12. Medicare total hospital reimbursements for Providence Health System facilities, using the Mayo Clinic/Intermountain merged baseline to estimate potential Population Level waste (1999–2003 data)

Hospital/System	Location	Loyalty	Deaths	Per Decedent			Ratio to Mayo/Intermountain					Proportion waste
				Total Reimbursement	Hospital Days	Reimbursement Per Hospital Day	$	=	Days	x	$/Day
Providence Holy Cross Med Center	Mission Hills, CA	85.9	996	35,180	29.4	1,198	1.61	=	1.59	x	1.02	0.38
Providence St. Joseph Med Center	Burbank, CA	91.6	2587	34,542	28.9	1,196	1.58	=	1.56	x	1.02	0.37
Little Company of Mary Hospital	Torrance, CA	87.3	1770	31,238	25.9	1,204	1.43	=	1.40	x	1.02	0.30
San Pedro Peninsula Hospital	San Pedro, CA	89.9	878	33,472	29.5	1,135	1.54	=	1.60	x	0.96	0.35
Providence Alaska Medical Center	Anchorage, AK	93.1	1229	27,390	23.3	1,177	1.26	=	1.26	x	1.00	0.20
Providence St. Vincent Med Center	Portland, OR	90.9	1645	21,229	17.2	1,235	0.97	=	0.93	x	1.05	-0.03
Providence Everett Medical Center	Everett, WA	92.2	2097	21,615	14.9	1,452	0.99	=	0.81	x	1.23	-0.01
Providence Portland Medical Center	Portland, OR	90.2	1622	20,403	16.1	1,266	0.94	=	0.87	x	1.07	-0.07
Providence St. Peter Hospital	Olympia, WA	92.6	1865	19,225	13.3	1,447	0.88	=	0.72	x	1.23	-0.13
Providence Centralia Hospital	Centralia, WA	93.3	1091	19,080	15.5	1,235	0.88	=	0.84	x	1.05	-0.14
Providence Medford Medical Center	Medford, OR	92.3	1214	17,788	15.0	1,185	0.82	=	0.81	x	1.01	-0.23
Providence Seaside Hospital	Seaside, OR	87.9	262	22,579	16.4	1,378	1.04	=	0.89	x	1.17	0.03
Providence Newberg Hospital	Newberg, OR	90.1	173	19,798	13.8	1,438	0.91	=	0.75	x	1.22	-0.10
Providence Milwaukie Hospital	Milwaukie, OR	85.3	274	18,116	11.9	1,516	0.83	=	0.64	x	1.29	-0.20
Providence Health System		90.8	17703	25,343	20.3	1,247	1.16	=	1.10	x	1.06	0.14

Return to Contents

3.2. Estimating Waste at the Microsystem/Episode Level—Quality Improvement Analysis

The ATP projects represent work by representatives of many U.S. health systems to target areas of waste and poor quality. The ATP projects span a number of years; our analysis used 58 projects for which summary reports were available. Each project was categorized by its major area of focus, based on the project summary report. The Six Sigma projects were conducted over 3 years at Providence Health System. Some projects were selected as "training" projects for the initial Six Sigma rollout, but there is reason to believe that the projects are representative of the kinds of waste reduction efforts found. Each project was categorized by its major area of focus based on materials presented at hospital briefing sessions attended by the author (BB) and shared with the other authors.

A major goal of the Six Sigma projects has been reduced delays (improved throughput). Throughput is the key to more efficient use of facilities and labor, both high cost factors in health care. Each additional day in the hospital requires additional nursing hours, blood draws, meals, medical consults, and so on. Typically these units per case are in proportion to the additional time spent.

Increased throughput can also be said to lower a facility's cost per unit by spreading fixed cost components (e.g., the cost of new equipment, the cost of hiring and training staff) over a larger volume. The amount of unit cost savings will depend on whether the cost of a lab test or a fully delivered medication is lower when the lab or the nursing floor is running at higher capacity.

Throughput also has second-order effects. Each additional day involves new staff that must be oriented to the patient, with attendant handoff problems and costs. Several projects targeted redundant information gathering that was the result of delays. Longer stays also mean additional chances for infections or other adverse events.

Third-order effects involve the administrative layers necessary to address the many forms of delay. There are often more complex staffing and scheduling plans, added facility space for holding areas, the need to sometimes house patients in inappropriate floors, and additional patient transport.

Caveat. The method of cost per unit reduction by substituting less expensive resources (e.g., labor) for more expensive ones was not exemplified by either ATP or Six Sigma to any great extent. As previously stated, these projects did not represent a cross-section of all situations because they were selected for training purposes without an underlying framework to guide distribution of what was exemplified.

Quantifying Waste. The ATP and Six Sigma projects suggest huge opportunities for savings, but we cannot be sure of the total available savings via synthetic estimate. These projects are early attempts used for training purposes and only reflect targeted opportunities for investigation that may not be likely to other clinical areas. Nevertheless, they do provide examples that demonstrate how such tools (i.e., Six Sigma) can be used to excavate waste/ inefficiency.

Examples include the following:

• A nutrition services project to address waste in snacks provided to patients: At one hospital, 29,700 between-meal snacks are ordered and produced annually for patients. Only 56 percent ever reached patients. Of the nourishments delivered, 70 percent were consumed. Overall, only 39 percent of all nourishments produced were consumed by patients. Annualized cost associated with current waste is $32,000.
• A project to reduce length of stay (LOS) for hip fracture patients who are discharged to skilled nursing facilities: This project reduced LOS from 120 hours to 94 hours (21% reduction) by addressing delays in the assignment of an acute care manager and the timely removal of urinary catheters.
• A project to reduce LOS for stroke patients: Patients were discharged on average 20 hours sooner if stroke preprinted orders were used AND an ACM reviewed their chart within 24 hours. The reduction from 78.9 to 58.6 hours constituted a 25 percent decrease in LOS.

Do these LOS reductions represent real cost savings to the hospital? Depending on the payment method, a hospital might see a loss in revenue. However, Providence and others are using a backfill approach (Cowan et al., 2006) to estimate the gain in net revenues attributable to a hospital's ability to admit more patients. With this approach, we assume that hospitals/units are admitting to capacity (typically, maximum capacity is an 80 percent to 85 percent census, to allow patient flow through the unit). Savings are then calculated as follows:

Marginal Cost Savings = backfill profit – loss in net revenues for early discharge (where backfill profit = revenues – variable costs/day)

This further assumes that the earlier days bring in more revenue than later days in a LOS (i.e., not per diem—based reimbursement) and that the freed bed can be readily filled by a new case.

Application of this method is illustrated by Cowan et al. (2006). In their study of staff substitution leading to improved care management, they found a marginal cost savings of $1,591 (average value of backfill profit = $1,707 per case less marginal loss of eliminated additional days of $116). The authors note that Ettner et al. (2006) analyzed the cost offset and net costs savings using the same dataset but a different methodology and found similar results. Thus, the stability of this approach seems reliable.

A second observation on LOS efficiency improvements must be made, however. Shortening LOS can sometimes result in very sick patients being discharged from acute care hospitals to other sites, such as long-term care hospitals. This is important at the Population Level. The care efficiency efforts on the part of acute care hospitals is driving a new level of care for high acuity patients. The question is would the patient have been better off (in terms of quality of life and overall admission time and cost to CMS) if these patients were not discharged so quickly from an acute care facility? Conversely, are these patients better off in these specialized facilities? The key question here is have we taken a systems perspective to understand unintended consequences of fragmented facility/reimbursement decisions?

Finally, there is concern that too-short LOS may have an impact on readmission rates. For example, the Leapfrog Group is rating hospitals as to their LOS, taking readmission rates into account.

ATP projects and Six Sigma projects address waste in similar ways. Both have a set of QI principles and tools at their core. Terminology and project structure differ. Six Sigma projects appear to be successful in reducing waste, although the experience of Providence is that many Six Sigma projects are taking longer than anticipated. Providence and Intermountain Healthcare have both adopted TPS methods as a tool for capturing "low hanging fruit" by involving frontline workers in short-term, Kaizen efforts.

Possible next steps at the Episode Level. While not able to estimate total health care waste at the Episode Level, our final efforts did produce a hypothetical analytic structure for possible use in the future. Several health care delivery systems are beginning to organize their operations around Deming's "key process" concept. Using a Baldrige quality award model, these institutions identify and then prioritize clinical work processes, with an aim to organize a relatively small list of processes (typically, about 10 percent of all clinical work processes in the organization) that produce the vast majority of clinical and financial results (typically, more than 90 percent of all clinical outcomes and costs), as suggested in Crossing the Quality Chasm (IOM, 2001). One such effort, which has been under development at Intermountain Healthcare for almost 10 years, has led to a "key process" list (this list is incomplete but illustrates the concept):

1. Patient Safety:
   • Adverse drug events (medication selection, preparation, and delivery).
   • Hospital-acquired infections (especially postoperative deep wound infection).
   • Pressure injuries.
   • Mechanical device failures.
   • Complications of central and peripheral venous lines.
   • Venous thromboembolism.
   • Patient falls and injuries (strength, agility, and cognition).
   • Blood product transfusions.
   • Patient transitions.
2. Clinical Programs (condition-related clinical processes):
   • Cardiovascular:
     • Ischemic heart disease:
       • Chest pain/r/o myocardial infarction.
       • Diagnostic and interventional cardiac catheterization.
       • CABG surgery.
     • Congestive heart failure:
       • Medical management of congestive heart failure.
       • Valve surgery.
       • Heart transplant.
     • Rhythm disorders.
   • Neuromusculoskeletal.
   • Surgical specialties.
   • Women and Newborn.
   • Intensive (inpatient) medicine.
   • Intensive (inpatient) pediatrics.
   • Intensive (inpatient) behavioral.
   • Oncology.
   • Primary care (outpatient clinics).
   • Health maintenance and latent risk prevention (preventive medicine).
3. Clinical Support Services:
   • Pharmacy.
   • Imaging.
   • Pathology (lab, microbiology, blood bank, and surgical pathology).
   • Central supply.
   • Procedure rooms (anesthesiology).
   • Intensive care units.
   • Nursing units.
   • Therapy (e.g., physical, respiratory).
   • Other (e.g., dietary).
4. Other:
   • Office of Research.
   • etc.
5. Service Quality.
6. Administration processes.

Such a list anticipates the creation of quality control measurement and management systems for each key process, which would create the ability to directly estimate waste and then manage it out. Essentially, all of the ATP and Six Sigma projects examined can be regarded as falling within this general structure, at either the Episode Level or the Patient Care Level.

Our work to date indicates that such an approach could have very high potential for identifying and managing waste, perhaps on the same order of magnitude as that found by the Dartmouth CECS group at the Population Level.

Return to Contents
Proceed to Next Section