Total Quality Control: Evolution of Quality Management Systems

SYMPOSIUM Total Quality Control: Evolution of Quality Management Systems James O. Westgard, PhD, and Patricia L. Barry, BS, MT(ASCP) An understanding ofindustrial total quality control and the evolutionary process experienced for quality systems in business and industry provides guidance for the development ofquality management systems in health care laboratories. Quality systems in laboratories are built on quality laboratory practices, have a strong statistical process control or quality control component, and have recently been extended to include broader monitoring for quality assurance. Laboratories need to develop a well-organized quality improvement component and need to place more emphasis on quality planning. All these components, working together, provide the necessary system for quality management. From the Department of Pathology and Laboratory Medicine, University of Wisconsin (Dr Westgard); and Clinical Laboratories, University of Wisconsin Hospitals and Clinics (Dr Westgard and Ms Barry), Madison, WI53792. he industrial model for total quality control (TQC) was introduced nearly 30 years ago by Feigenbaum,1 who describes TQC as "the agreed, organization wide, detailed operating work structure of technical, scientific, and managerial procedures for guiding the coordinated actions of the humans, the equipment, and the information of the institution in the best and most practical ways, to assure user satisfaction and reasonable costs of quality."2 This definition provides a broader description of quality control than what is usually understood in health care laboratories. The importance of the word "total" as a modifier of QC is evident. Feigenbaum's TQC includes quality definition and evaluation, quality planning, purchased material evaluation, product control and evaluation, special process studies, quality information feedback, quality information equipment, quality education and training, user quality evaluation, and management of quality. The management aspect "deals with the integration and assurance of satisfactory operation of these quality activities throughout the institution."2 Japanese industry further expands the TQC concept to company-wide quality T control, a name that implies everyone in the company is involved in quality control activities. That expansion explains why participative mechanisms such as "quality circles" and suggestion programs are so important in quality activities in Japanese organizations. The application of TQC in health care organizations is becoming of interest as the health care marketplace becomes more and more competitive. While the cost of health care received much attention in recent years, quality is now becoming a major issue. Currently, the quality of laboratory testing itself is attracting much public attention, causing professional groups such as the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) to encourage further development of quality systems and legislative bodies to mandate improvements in the quality of laboratory testing. Attempts to improve quality systems in health care laboratories should be guided by an understanding of the expected stages of evolution for quality systems, along with recognition of where we are starting and what our ultimate destination is. Current accreditation and regulatory attempts to define quality systems provide insufficient guidance because they Laboratory Medicine June 1989 377

In understanding quality systems in health care laboratories, thefirststage of "inspection" corresponds to the review of laboratory test results to assess whether they agree with the patient's diagnosis and condition. This form of quality system was in place in the 1950s, began to give way to QC in the 1960s, and was replaced by QC during the 1970s evolution to stage 2 as would be expected in the industrial TQC model. During the 1980s, broader quality assurance programs have been developed evolution to stage 3. Little has been done to provide in-depth training and education in quality, so by Sullivan's model, most health care laboratories have progressed no further than much of American industry. Fig 1. Components of a quality management system. lack this perspective. There is a lack of understanding of what "quality" is, how it can be defined and measured, how it must be managed, and who is responsible for quality. There is little recognition of the capabilities and limitations of present quality management techniques and therefore little guidance for making real improvements in quality and quality management. Unfortunately, the wealth of experience in other industries seems to be ignored because of the belief that "we are different." Lessons learned elsewhere somehow cannot be applied in health care. This belief ignores our fundamental commonality with all other industries we are all organizations of people who must accomplish the missions of our institutions. While we may differ in the goods and services we provide, we have the same kinds of problems in managing our resources to accomplish our missions. Managing quality is about managing an organization to accomplish its mission, whatever that mission may be. Quality management must be central in all management activities and decisions, not a separate committee function as found in many health care organizations. Quality management is fundamentally a management philosophy, supported by principles, procedures, tools, techniques, and resources. It is concerned with mak- ing quality a primary objective in people's work, educating and training people to do their jobs and achieve the quality objectives, planning and improving their jobs so the work can be done correctly and efficiently, and guiding, supporting, and recognizing people so they will continue to achieve the quality objectives. volution of Industrial Quality Systems Sullivan3 has described the development ofjapanese industrial quality systems in terms of the following seven stages: Stage 1 Inspection Stage 2 Statistical QC or process control Stage 3 Broader quality assurance Stage 4 Education and training in quality Stage 5 Optimization of design for robustness Stage 6 Optimization of design for cost Stage 7 Deployment and integration of quality into all the activities of the organization According to Sullivan, much of American industry has developed stages 1 through 3, whereas leading American companies that are competitive with Japanese industry are working on stages 4 through 7. E 3 7 8 from Laboratory Medicine June 1989 Downloaded https://academic.oup.com/labmed/article-abstract/20/6/377/2642040 To provide additional perspective on the TQC model, Juran's description of quality management as a "quality trilogy" is useful.4 According to Juran, the three essential components of a quality system are QC, quality improvement, and quality planning. Juran's QC encompasses Sullivan's stages 2 and 3, both control and assurance; quality improvement includes education and training in quality (stage 4), with an added emphasis on problem solving to eliminate long standing, chronic problems; and quality planning encompasses the design stages 5 and 6. Stage 7 integration is accomplished by having all three components working together throughout the organization. Juran describes QC as the conduct of operations to assure that quality goals are met under routine operating conditions. Quality improvement is the process for breaking through to new levels of performance, achieving distinctly superior performance to what was achieved in the past. Quality planning is the preparation to meet quality goals, with the end result being the development of a process capable of meeting goals under routine operation. Although planning should precede control, and improvement should not be necessary if processes are properly planned in thefirstplace, the only component that is usually in place in most organizations, including health care laboratories, is QC. Development of better quality systems requires addition of the quality improvement and quality planning components. The order of addition, according to Juran, should be quality improvement then quality planning. Juran recom-

mends doing quality improvement to gain the experience necessary to understand quality planning. 140- r omponents of Quality Systems in Health Care Laboratories With this background, an ideal quality system for a health care laboratory would be made up of the following components: quality laboratory practices (QLP), process control or quality control (QC), quality assurance (QA), quality improvement (QI), and quality planning (QP). QC and QA will be considered separate components because of existing practices, but it should be recognized that the demarcation of QC from QA will be difficult and subject to some disagreements. All these components, added together, constitute a quality management system (QM), as shown in Fig 1. Quality Laboratory Practices Good policies, procedures, processes, and people are the essential starting point for providing a quality testing service. No amount of control, assurance, improvement, or planning can make up for the lack of quality practices as the foundation for achieving quality in routine operations. Quality laboratory practice includes qualified staff, adequate facilities and resources, adequate staffing for workload, procedure manuals for all processes and their maintenance, initial and ongoing training, etc. Uldall5 has discussed "good laboratory practice" in his review of quality systems. Standards and checklists are provided by the JCAHO and the College of American Pathologists. Qc QC is commonly understood in health care laboratories to refer to statistical quality control or statistical process control. QC charts are in widespread use. Each measurement procedure in a laboratory is usually monitored by a statistical control procedure. Common control procedures include Levey-Jenning's charts6 and Westgard multirule QC.7 Less common procedures include cumulative sum, mean and range, trend analysis, and patient data QC algorithms. The purpose of these QC procedures is to detect potential problems. Production is then stopped to prevent erroneous results from being reported, the process is fixed and restarted, and the test samples in pre- 100 120-80 BLOOD GOS REQUESTS 100- UJ» S92 >LJ 80- a fcj60- - 3 Mlnutas 44 Minutes > 9 Mlmrtas «27 Mlnutas «37 Mlnutas u 60 UJ Q_ 40 CC 3 40(-1 20 30 n* 40 50 MINUTES Fig 2. Histogram display of turnaround time for a la ogy, with permission). tory test (from Eggert et ^Informatics in Pathol- viously rejected analytical runs are reanalyzed. Quality is finally achieved, but at the additional expense of correcting the problem and repeating the analyses. The evolution of QC procedures is itself interesting, especially in comparison to the evolution of measurement procedures or analytical systems. Levey-Jenning's charts were introduced in clinical laboratories in the 1950s, prior to the introduction of automated analytical systems. The charts can be considered firstgeneration QC. They were designed for ease of use manually and were well suited to the manual methods that constitute first-generation measurement procedures. When applied to later generations of automated measurement procedures, however, there may be serious limitations. When control limits are set as the mean ± 2s, an appreciable "waste" of process output may result because of false rejections of analytical runs, even though the process is working as well as expected. Changing to 3s control limits reduces the false rejection rate, but also reduces the error detection capability, providing less assurance that the quality of the process is satisfactory. Multirule QC procedures designed to maintain low false rejection and high error detection rates can be considered second-generation QC. False rejections are minimized by careful selection of control rules; error detection is maximized by using groups of control rules in parallel to increase sensitivity. Like Levey-Jennings QC, multirule QC was used initially as a general QC design applied to the many different measurement procedures in a laboratory. Multirule procedures are well suited to second-generation measurement systems, which are the first-generation automated systems. Individualized multirule designs based on medical requirements represent thirdgeneration QC. Stable automated systems that are calibrated infrequently (thirdgeneration measurement procedures) require changes in the approach to quality control. The establishment of quality goals for individual laboratory tests leads to individualized designs based on the medical requirements for analytical quality. In third generation QC, these individualized designs are often made qualitatively, based on experience and general recommendations. Individualized multistage multirule procedures based on cost-effectiveness represent fourth-generation QC. A quantitative design process is employed to consider the many factors and characteristics that are important in determining performance.8 Analytical goals are defined, medically important errors are calculated, Laboratory Medicine June 19S9 379

error detection and false rejection probabilities are determined, the observed or expected frequency of errors is considered, and the effect of different designs on the defect rate and test yield of the process are assessed by use of quality-productivity models. Most health care laboratories today employ third- and fourth- generation analytical systems, with fifth-generation systems beginning to be introduced. Laboratories should likewise be applying thirdand fourth-generation QC procedures and beginning to develop fifth-generation procedures. Even in QC, which is a strength of quality systems in health care laboratories, there is much room for improvement and development. C (A * < B " v- C i_ a> <D X3 -O CD E 7S Z. T. 600-1600- 1400-1200- 1000-800- 400-200^ 0 -l ICU 911/ 356 1 H H 215 / H H 11 MSurg Nturo P*ds 207 Hes Oncol OP Trwis 12 9 5 2 I Lab Othw Ext Curt -90-80 -70-60 50.6-50 OR -40-30 -20-10 o c mul a < co CD o CD 3 r-»- 0D CD Quality Assurance (QA) Satisfactory quality depends on many critical factors affecting the laboratory testing process and many performance characteristics related to the test result and the services provided. Important factors that must be considered include proper patient preparation, proper specimen acquisition, correct specimen processing and sample preparation, sensitive and specific measurement procedures, proper analytical protocols, accurate recording of test results, and clear and understandable reports. Performance characteristics include accurate specimen collection, precision and accuracy in the performance of the test, turnaround time to satisfy the urgency for the testing service, and proper interpretation of test results. The purpose of Q A is to provide measures of the many factors and characteristics that are important for assuring that customers' quality requirements are achieved and also for identifying problems that limit their achievement. QC provides measures of analytical quality, but the many other factors and characteristics need to be monitored, at least periodically, to assess performance of the overall testing process and to detect when and where problems are occurring. For example, turnaround time is an extremely important performance characteristic that needs to be carefully monitored. Simple studies are possible with manual records, or elaborate computer programs can be implemented to provide extensive on going measures. 9 The data collected can be used to describe the general performance that is observed, such as shown in Fig 2, or an exceptions report Fig 3. Pareto diagram of the source of specimen mislabeling and misidentification problems. o. CO w O I. CU a -J z CD C CU.o <u CO i_ Surg ICU TLC M«d ICU Inter C»r«Fig 4. Pareto diagram with further breakdown of the for intensive care units. can be generated to identify problems with individual specimens and individual tests that do not meet specified time requirements. Both approaches are acceptable under present QA practices, but the first provides a description of the general level of performance for a test, while the second provides a mechanism for identifying individual problems and gaining control of day-to-day performance. Many other monitors can be implemented, such as specimen mislabeling and misidentification, the number of specimen mislabeling and misidentification problems specimens recollected, reorders of tests by physicians, orders for redundant tests, duplicative orders for tests, changes in test results once entered in the laboratory record, incident and occurrence reports, surveys of nurses and physicians, surveys of patients, etc. Problems identified by such Q A monitoring may be difficult to solve. Consider specimen mislabeling and misidentification, which are chronicproblemsforvirtually every laboratory. Each laboratory probably monitors how many problems 380 Laboratory Medicine June 1989

Laboratory Requirements Too many exceptions to requirements euowed Not well communicated \ Policy not pert of hospital administrative manual or nursing manual Long-term or frequent patltnt Specimens collected by non-lab staff Collection procedures not-standardized Inadequate training of non-lab staff Lab Specimen Collection Manual not available to nursing service Not adequately defined Interruptions Specimens not labeled et patient side Person labeling specimen not always person who collected it (non-lab staff) Collection supplies/labels not located In one area on units Nurse has multiple tasks per patlent-contaet Mislabeled. Misidentified Laboratory Specimens Specimen Collection Process Fig 5. Cause-and-effect diagram illustrating potential causes of specimen mislabeling and misidentification problems. occur and prepares monthly and yearly summaries. Periodically, the laboratory attempts to reduce the numbers, often by changing the specimen labeling or specimen acceptance policy. These attempts are seldom successful because the problem generally resides outside the laboratory and is outside its control. Figure 3 shows the source of such specimen problems based on data collected over a 15-month period. Of 1,800 problems, only 12 were due to specimens collected, labeled, and identified by laboratory personnel. Half the problems originated in the intensive care units (ICUs), where laboratory personnel did not draw specimens. Further analysis, as shown in Fig 4, reveals still more information about the location of the problems within the ICU program area. In attempting to solve such problems, a laboratory will typically make changes in the specimen acceptance policy. However, such an approach attempts to "inspect" quality into the product, not eliminate the source of the problem, and leaves the laboratory in the position of rejecting specimens without proper labeling and identification and making users and customers unhappy with its service. Solution of these problems requires that the many potential causes be considered, as shown in Fig 5, the "root cause" be identified, and a change be made to eliminate such errors. Solution requires that laboratory personnel work together with personnel in the nursing units to improve the specimen acquisition process. A related example is the requirement of present accreditation standards that laboratories monitor the "quality and appropriateness" of laboratory testing services. Selection of appropriate tests and appropriate use of testing services are important quality characteristics, but they are characteristics of the test ordering process, which is generally under the control of the physician-users of laboratory services. Current accreditation standards may actually create an adversarial relationship between the laboratory and its customers if the laboratory attempts to enforce changes in test ordering and test utilization. Again, the laboratory cannot "inspect" quality into the process; it must work with the physicians and users to improve the test ordering part of the process. These two examples illustrate a rather widespread problem with current QA practices there are no effective mechanisms for resolving interdisciplinary problems that are identified. Solutions are supposed to somehow occur as a natural consequence of knowing about the problem. Unfortunately, most of the problems identified by such Q A monitoring will be problems that cross departmental lines and cannot be resolved by the laboratory itself. A further difficulty is the practice of comparing measured performance to standards of performance based on what is presently observed or "norms" of some form, rather than what is needed. The number detected is compared to some norm, and no action is taken if performance is no worse than the performance Laboratory Medicine June 1989 381

observed earlier or the performance observed in other organizations. Gambino clearly expresses the fallacy of this approach when he asks how many babies can be dropped in the nursery. Certainly it must happen and "norms" could be established. But, regardless of norms, the problem cannot be tolerated. A solution must be found. Quality Improvement Solving problems that are difficult (such as appropriate use of services) and chronic (such as specimen mislabeling) is actually the domain of quality improvement. QA efforts are inadequate because they are basically detection mechanisms. The need to "break through" to a new and improved level of performance requires that the process be changed and improved. The root cause(s) of the problems must be identified and removed to prevent those problems from occurring in the future. Quality improvement is a systematic approach to problem solving, quite different from the "fire fighting" kind of problem solving that provides a shortterm solution or immediatefix.quality improvement is aimed at difficult and chronic problems that cannot be solved by firefightingmethods. Solutions to these problems usually require a group problem solving effort because the expertise required extends beyond any one individual, or the problem and its solution extend beyond the section, division, or department. Solving these problems requires new tools and techniques to structure the group's efforts and create the synergy that makes the capabilities of the group greater than the capabilities of its individual members. Techniques such as brainstorming, nominal group, pareto diagrams, flow charts, cause and effect diagrams (also known as "fishbone" or "Ishikawa diagrams"), process cause and effect, and forcefieldanalysis are often used, along with other graphical and statistical techniques already commonly used by laboratory analysts. 11 Such problem solving requires a focused effort and is best carried out following Juran's project-by-project approach. 12 Aproject team is established to provide the necessary expertise and representation, is trained in group problemsolving tools and techniques, and is charged to make determined efforts to solve one selected problem. According to Juran Steps 1-3 Determine Medical Needs 1 Juran Steps 4-8 Select/Evaluate Diagnostic Test i Select/Evaluate Measurement Procedure I Select/Design Control Procedure l r Juran Step 9 Train and implement Control/Evaluate Routine Operation Fig 6. The planning of analytical processes in health care laboratories inrelation to the steps of Juran's quality planning. 382 Laboratory Medicine June 1989

Juran, quality improvement happens only when individual problems are solved. In Juran's language, a "project" is a problem targeted for solution. Quality improvement, therefore, is an intense, focused problem-solving effort employing a quality team and the tools and techniques necessary for group problem solving. Detailed guidelines for implementing quality improvement are available in Harrington's bookthe Improvement Process.1* This book can serve as a step-bystep procedure manual for developing the quality improvement component. Some translation will be required to adapt the industrial oriented examples to the service setting of a health care laboratory. Quality Planning There would be no problems that need to be solved if adequate processes were implemented in thefirstplace. Selection and design of those processes are the domain of quality planning. Quality planning depends on having a good understanding of customers' quality requirements, translating those quality requirements into specifications, and using those specifications to select, design, evaluate, and validate the processes that are implemented. For example, Juran's systematic approach to planning14 specifies the following steps: "1. Identify who are the customers; 2. Determine the needs of those customers; 3. Translate those needs into our language; 4. Develop a product that can respond to those needs; 5. Optimize the product features so as to meet our needs as well as the customers' needs; 6. Develop a process which is able to produce the product; 7. Optimize the process; 8. Prove the process can produce the product under operating conditions; 9. Transfer the process to the operating forces." This planning approach is applicable to the development of analytical processes in health care laboratories, as illustrated in Fig 6. Initially, the medical needs (customer requirements) must be determined to guide the selection of the diagnostic test, the selection and evaluation of a measurement procedure, and the selection or de- sign of a control procedure. This requires the laboratory to identify the customers (Juran's step 1), determine their needs (step 2), and translate those needs (step 3) into performance specifications to guide the selection and evaluation of measurement and control procedures (steps 4 through 8). Specifications for diagnostic sensitivity, diagnostic specificity, and predictive value are used to guide the selection and evaluation of the diagnostic test. If medical performance is satisfactory, then the analytical quality requirements are specified for both application characteristics (such as sample type, sample size, turnaround time, etc), and performance characteristics (such as linearity, precision, interference, recovery, and accuracy). The customers' requirement in the form of a total allowable error must be translated into error specifications for these specific characteristics. When analytical performance is acceptable, then the control procedure itself must be planned or designed. The sizes of the medically important errors that must be detected by the control procedure must be determined. Performance characteristics such as probabilities for rej ection and average run lengths are evaluated to determine the control rules and the number of control measurements. Once a proper QC design is achieved, analysts can be trained to operate the process, and the analytical process can be implemented (step 9). Routine operation proceeds only with continued monitoring via QC to evaluate the frequency of errors occurring with the process. Quality planning encompasses much more than just the planning of analytical processes. Every repeated activity in the laboratory is a process that should be properly selected or designed and carefully evaluated; thus all activities require careful planning. Quality planning encompasses all management decisions because each and every decision may affect the laboratory's processes and ability to meet the needs of its users and customers, now or in the future. Long-range planning, or strategic planning, is especially important for preventing problems. Such planning should focus on user or customer needs, consider trends that will affect future needs and operations, assess strengths and weaknesses of present operations, and attend to threats and opportunities in the mar- ketplace. The laboratory's mission needs to be clearly stated and a vision of the laboratory communicated through the goals, objectives, and strategies that compose the long-range plan. In all of this, quality must be a central issue. r omment With this perspective on the ex. pected evolution of quality systems and the components of laboratory quality systems, the status of present quality systems in health care laboratories can be more clearly described. Laboratories generally have well-established policies, procedures, processes, and people; thus the QLP foundation is in place. The "inspection" stage is history in most laboratories, having been replaced by statistical QC or process control, a natural evolution according to Sullivan's description.3 Q A activities are increasing, but there is additional work to be done to improve the monitoring and to make good use of the data obtained. Quality improvement and quality planning are not well developed, thus components of quality improvement and quality planning must be added to achieve TQC or a total quality management system. In improving present quality systems, the continued development of Q A needs to be guided by a more careful assessment of what factors and characteristics are critical and what indicators and monitors will provide useful information about laboratory performance and help resolve performance problems. The distinction between QC and QA tends to be arbitrary, at least in Juran's view, where both are grouped together as QC (4), but is a point of some confusion and considerable discussion. Juran emphasizes the selection of subjects (items) for measurement for whatever quality characteristic that is monitored. What is essential is to identify critical factors and performance characteristics, devise and implement monitoring procedures, and obtain quantitative measurements of quality indicators and characteristics. Without such measurements, a laboratory cannot know if quality requirements are being achieved, if process changes are leading to improvement, or if planning activities are preventing problems from occurring. Addition of the quality improvement component requires implementation of a structured group problem-solving ap- Laboratorv Medicine June 1989 383

proach. The existence of problem-solving mechanisms does not necessarily mean that the quality improvement component is in place. The distinction is in having a mechanism for solving difficult and chronic problems by means of quality teams trained in the use of tools and techniques for group problem solving. The clearest evidence of the presence or absence of the quality improvement component is the availability of training in group problem solving and the ready application of the group problem-solving tools. Quality planning, in its optimal form, will follow implementation of quality improvement. Although it goes on in part already, the comprehensive nature of the planning that is required will be learned in the process of making improvements in quality. Quality planning, therefore, will not be fully developed until the approach to improving quality itself becomes better developed. The next major step for most laboratories is quality improvement. Efforts to develop quality systems must start to develop improved mechanisms for solving difficult, chronic problems. Efforts to measure and monitor quality will have little effect if the problems uncovered cannot be solved. Approaches for implementing quality improvement in health care laboratories are now being developed 15 D References 1. Feigenbaum AV:Total Quality Control, 3rd ed. New York, McGraw-Hill International Book Co, 1983. 2. Feigenbaum KM-.Quality Assurance in Health Care: A Critical Appraisal of Clinical Chemistry. Rand RN, Eilers RJ, Lawson NS, et al (eds), Washington, DC, American Association for Clinical Chemistry, 1980. 3. Sullivan LP: The seven stages in company-wide quality control. Quality Progress, 1986(May);77-83. 4. Juran JM: The quality trilogy. Quality Progress, 1986(Aug);19-24. 5. Uldall A: Quality assurance in clinical chemistry. Scand J Clin Lab Invest, 1987;47(suppl 187):l-95. 6. Levey S, Jennings ER: The use of control charts in the clinical laboratory. Am J Clin Pathol, 1950;20:1059-1066. 7. Westgard JO, Barry PL, Hunt MR, et al: A multi-rule Shewhart chart for quality control in clinical chemistry. Clin Chem, 1981;27:493-501. 8. Westgard JO, Barry PL: Cost-Eftective Quality Control. Washington, DC, AACC Press, 1986. 9. Eggert AA, Westgard JO, Barry PL, et al: Automated collection and analysis of turnaround time data from a clinical laboratory computer system. Informatics in Pathology, 1987;2:5-14. 10. Gambino SR: Quality control: Can we achieve error-free work? MLO, 1985(Mar);37-40. 11. Ishikawa K:Guide to Quality Control, 2nd ed. Japan, Asian Productivity Organization, 1982. 12. Juran JM, Endres K-.Quality Improvement - Services. Wilton, Conn, Juran Institute, Inc, 1986. 13. Harrington \i]-the Improvement Process. New York, McGraw-Hill International Book Co, 1987. 14. Juran JMjuran on Planningfor Quality. New York, The Free Press, 1988. 15. Westgard JO, Barry PL: Beyond quality assurance: Committing to quality improvement. Lab Med 1989;20:241-247. 384 Laboratory Medicine June 1989