Quality improvement's new focus yields quantitative results.

Health care quality improvement methods are now undergoing fundamental change. The emphasis is shifting from inspection of physician practices to continuous improvement of clinical processes validated by quantitative results, This change is long overdue, Traditional quality assurance methods that operate retrospectively and alienate physicians are not useful in a marketplace where quality and cost control are a matter of survival. Physician practices are only as good as the institutional processes they rely upon, and any quality improvement method that alienates physicians is doomed. Quality improvement is impossible without the support of physicians, because true improvement is driven and quantified by clinical data. Physicians are needed to interpret that data.

For the past 20 years, quality assurance methods have done little to improve quality and control costs. They have failed because quality cannot be "inspected in" after the fact. Quality must be inherent in the treatment process, from the time the patient is admitted to the hospital to the time the patient is discharged. By taking an entirely different approach to quality--one that emphasizes the need to reduce process variation--today's continuous quality improvement (CQI) methods factor quality into the clinical process and champion physician support.

Hospital results remain fairly stable over time/indicating that outcome differences between hospitals are not random. If the outcomes are not random, the processes that produce them are not random either. The difference, then, between hospitals that consistently achieve clearly superior outcomes and those that don't is a process difference. Hospitals producing superior outcomes for particular diseases do so by consistently applying the correct clinical-care processes for those diseases. In these hospitals, the patient admitted at 3 a.m. on a Sunday gets the same diagnostic, therapeutic, and monitoring process as the patient admitted at 8 a.m. on a Monday.

Variation in the clinical process arises from three sources: the patient, the clinician, and the institution. The process begins with assessment of the patient's health status and ends with measurement of results of care. It is between these two points that opportunity for improvement exists. The physician interprets the patient's medical history and key clinical findings at admission into diagnostic probabilities and an admission severity and determines a degree of urgency. The diagnostic probabilities and urgency form the basis of the physician's management plan for diagnosis, treatment, and monitoring of the patient (figure 1, page 33).

Most variation in care effectiveness (morbidity and mortality) results from institutional inconsistency in executing the physician's management plan. CQI's goal is to standardize the execution of the correct plan. It is a goal rooted in the belief that consistency and constancy of the process produce better results, both in outcomes and in utilization. This is not "cookbook medicine." It is standardization of the institutional consistency of the execution of clinician's orders, not standardization of the orders themselves.

Insight into the Process

Conversion of clinical data into information about the clinical process is critical to CQI. To improve the process, we must understand it. CQI management philosophy says quality can be continuously improved by studying outcomes data, evaluating the clinical processes that produce the outcomes, and making the required changes.

The greatest potential for savings lies in focusing process improvements on high-volume, high-risk diagnoses and procedures, which manifest significant variation in outcomes and process.(1) Increasing the effectiveness of the clinical process for one disease also will likely improve the processes for other diseases in the same body system. For example, a change in the diagnostic process for pneumonia is likely to benefit the diagnostic processes for other respiratory diseases.

Our research indicates that improving patient outcomes only half a standard deviation in 10 top-volume DRGs can yield an average annual savings of about $4.1 million for a typical 350-bed hospital.(1) Achieving benchmark results can

yield even greater savings. Clinical benchmarking is an important component of CQI.

Clinical Benchmarking

The goal of clinical benchmarking is twofold--create a sense of what is achievable, and find out how to achieve it. It is not enough to know that better outcomes are being achieved. It also is essential to understand how those outcomes are being achieved. After severity-adjusted clinical information is used to identify hospitals with the best outcomes for particular diseases, the clinical processes at the benchmark institutions must be studied and the findings documented and adapted to other processes.

Clinical benchmarking applies CQI principles to specific diagnoses and procedures, because studying how specific processes of care are applied to specific diagnoses is the best way to gain insight into what produces superior results. The outcome to be studied in defining the benchmark may be different for each disease or procedure. Mortality may be the outcome measure used for diseases or procedures with high mortality rates. A combination of mortality and resource use may be the outcome measure for diseases with moderate mortality rates, high charges, and/or long lengths of stay. Appropriateness may be the measure where appropriateness is the key issue, as with procedures such as appendectomies or caesarean-sections (figure 2, below).

Whatever the outcome being measured, its validity is directly related to the quality of the data involved. Outcome measures need to be adjusted for patient severity of illness at admission, and outcomes must be measured on the basis of objective clinical findings, such as laboratory values.

Key Clinical Clues

Understanding the clinical process begins with understanding the patient's condition. The key clinical findings (KCFs) presented by the patient at admission, not the discharge diagnosis, drive process results.

Research has shown that certain clusters of key clinical findings are associated with higher mortality risk in certain diseases.(2-6) For example, the presence of even one of five abnormal clinical findings in DRG 89 (simple pneumonia, pleurisy) patients at admission is associated with a significantly higher average mortality rate--18 percent mortality for patients with one of the five findings, as opposed to 4 percent for patients with none of the five (figure 3, below).(1)

These dramatic differences are not reflected in the secondary discharge diagnoses. The abnormal five findings (systolic BP, BUN, lethargy, creatinine, low or high pulse) are associated with a diagnosis of pneumonia in approximately 38 percent of admissions (figure 4, page 35).

In clinical benchmarking, KCFs are ranked by their association with mortality in a particular disease. This ranking reveals low-risk and highrisk patient subpopulations. Hospitals with benchmark outcomes have defined clinical processes for these subpopulations.

The Forbes Experience

Forbes Health System, a 784-bed hospital system in Pittsburgh, Pa., is an example of how knowledge of clinical processes and KCF subpopulations for a particular diagnosis can influence CQI. In 1988, comparative data showing Forbes with a 28 percent higher combined rate of major morbidity and mortality in pneumonia (DRG 89) than expected for its patient mix(1) was the impetus for process improvements that lowered the hospital's overall pneumonia mortality rate and reduced charges.

The changes began with a P-value calculation for statistical significance that eliminated randomness as the cause for the variation in rates. A detailed review of DRG 89 major morbidity and mortality cases followed. The review included analysis of pneumonia KCFs; however, detailed information about pneumonia subpopulations was not yet available. The review ruled out imperfect data and physician errors as causes for the higher mortality and major-morbidity rates.

Flow-chart and cause-and-effect analyses of the process for treating pneumonia patients at Forbes were developed. These analyses revealed inconsistencies throughout the process, from admission in the emergency department through collection of sputum to administration of antibiotics. One of the major findings was that specimen cultures were often being done on saliva, not sputum, resulting in normal flora findings. This was forcing physicians to order antibiotics empirically without identifying the infecting organism.

The review also found that time between a patient's arrival in the emergency department and initiation of antibiotics varied significantly. In some cases, antibiotics were not administered for four or five hours. The longest delays occurred when antibiotics were not begun in the emergency department and the process of transferring the patient to the floor and contacting the attending physician further postponed the first dosage. The delays were especially significant because 90 percent of the patients diagnosed with community acquired

pneumonia were admitted through the emergency department (figure 5, page 35).

Physicians at Forbes were writing appropriate orders, but the orders were not being consistently executed. Over an eight-month period, the hospital's quality improvement team used the information gleaned from the process analysis to change the process for executing physicians' management plans.

Protocols were instituted, including a respiratory-care sputum-induction protocol to provide physicians with information needed to make an accurate diagnosis of any respiratory infection within 24 hours of admission. To assure that specimens were sputum, not saliva, the laboratory was instructed to gram stain all specimens. The protocol also stated that antibiotics would be ordered and the first dose given before the patient left the emergency department, provided a sputum culture had been obtained. If not, antibiotics would be held until the sputum was collected, but not longer than four hours.

After only 10 months, the hospital's overall pneumonia mortality rate had dropped significantly-from 10.2 percent to 6.77 percent. This mortality rate is benchmark for pneumonia.(1) The average length of stay for pneumonia patients dropped from 10.4 days to 9.1 days. Average charges per pneumonia case dropped nearly $1,000. In the low-risk group, charges dropped even more.(1) These results are especially significant in light of two factors--DRG 89 patients are predominately Medicare patients, and, during the same period that pneumonia charges dropped at Forbes, overall charges at the hospital increased five percent.

When information on KCF high risk and low-risk pneumonia subpopulations became

available in 1992, Forbes reviewed the data and found that its improved mortality rate was due entirely to better outcomes among only 60 percent of its pneumonia population--its low-risk patients. This prompted the hospital to begin a detailed analysis of clinical processes for its high-risk pneumonia patients. That study is part of a clinical bench marking project now in progress. Without KCF analysis revealing high-risk and low-risk subpopulations, Forbes may have considered its process redesign complete, when, in fact, the clinical process for 40 percent of its pneumonia patients, the high-risk group, required further analysis and improvement.

Benchmarking's Advantages for CQI

Clinical benchmarking holds four major advantages for CQI. First, in contrast to measuring results against average results, comparisons against the best results will continuously upgrade the meaning of clinical excellence. Just as the Olympic athlete achieves a new world record, then immediately sets out to break it, so will hospitals conducting clinical benchmarking continuously seek to improve the benchmark. This makes clinical benchmarking the essence of continuous quality improvement.

The second advantage of benchmarking for CQI is that comparisons against the best create a sense of excitement that is missing when a hospital's results are compared against average results. This excitement is a critical factor in gaining the enthusiasm and support of physicians. The strongest motivation for change is belief that there is a need for change. Because 90 percent of hospitals produce results that fall within the average range, traditional comparisons against the mean show little gap between hospitals and thus provide little incentive for change. Benchmark comparisons, on the other hand, reveal a clear gap between the average and the best.

The third advantage of clinical benchmarking for CQI is that it enables hospitals to provide evidence to purchasers that clinical processes are being improved toward achieving benchmark results, not just average results. This evidence can give a hospital a critical competitive edge at a time when purchasers, looking for the most value for their dollar, are buying health care based on comparative outcome and efficiency data.

Clinical benchmarking's fourth advantage is that it helps establish priorities in the continuous improvement process. Because limited resources prevent hospitals from remedying all clinical processes at once, quantifying improvement opportunities, setting improvement priorities, and justifying those priorities to purchasers is vital.

The new focus on identifying, explaining, and reducing clinical process variation will fundamentally improve all care. By analyzing outcomes data and applying the knowledge gained to CQI and benchmarking efforts, providers can achieve the quality and cost breakthroughs necessary to succeed in the value-conscious 1990s and beyond.

References

1. These assertions are based on information in the MediQual database.

2. Farr, B., and others. "Predicting Death in Patients Hospitalized for CommunityAcquired Pneumonia." Annals of Internal Medicine 115(6):428-36, Sept. 15, 1991.

3. Subcommittee of the Research Committee of the British Thoracic Society and the Public Health Laboratory Service. "Community-Acquired Pneumonia in Adults in British Hospitals in 1982-1983: A Source of Aetiology, Mortality, Prognostic Factors and Outcome." Quarterly Journal of Medicine, New Series 62, No. 239, pp. 195220, March 1987.

4. Ispahani, N., and others. "An Analysis of Community and Hospital-Acquired Bacteraemia in a Large Teaching Hospital in the United Kingdom." Quarterly Journal of Medicine, New Series 63, No. 241, pp. 427-40, May 1987.

5. Rayner, B., and Greenwood, D. "Community-Acquired Bacteraemia: A Prospective Survey of 239 Cases." Quarterly Journal of Medicine, New Series 69, No. 259, pp. 907-19, Nov. 1988.

6. Fine, M., and others. "Hospitalization Decision in Patients with CommunityAcquired Pneumonia: A Prospective Cohort Study." American Journal of Medicine 89(12):713-21, Dec. 1990.

Further Reading

The following additional sources of information on quality improvement were obtained through a computerized search of databases. Copies of most articles are available from the American College of Physician Executives for a nominal charge. For further information on citations, contact Gwen Zins, Director of Information Services, at College headquarters, 813/287-2000.

Hughes, J. "Total Quality Management in a 300-Bed Community Hospital: the Quality Improvement Process Translated to Health Care." QRB 18(9):293-300, Sept. 1992.

Lasker, R., and others. "Realizing the Potential of Practice Pattern Profiling." Inquiry 29(3):287-97, Fall 1992.

Ziegenfuss, J. "Organizational Barriers to Quality Improvement in Medical and Health Care Organizations." Quality Assurance and Utilization Review 6(4):115-22, Winter 1991.

Zusman, J. "Quality Assurance to Continuous Quality Improvement: Challenges for the Medical Staff." Medical Staff Counselor 6(4):1-7, Fall 1992.