Quality Planning Models
How sophisticated are your quality planning needs? Does your quality take imprecision (CV) and inaccuracy (bias) into account? That's an analytical quality requirement. Does your quality take biological variability of the patient, as well as the physician's clinical decision interval, into account? That's a clinical decision interval. Learn the differences between analytical and clinical quality planning models, as well as the advantages and disadvantages of these approaches.
- The Need for quality planning models
- Planning expenditures via financial budgets
- Take home pay analogy
- Gross income analogy
- Planning quality via laboratory error budgets
- Pictorial description of laboratory error budgets
- Mathematical description of laboratory error budgets
- Conventional total error budget for stable performance
- Analytical quality planning error budget for unstable performance
- Clinical quality planning error budget for unstable performance
- Expanded models used by QC Validator 2.0 program
This lesson provides some theoretical background for the QC planning tools that have been introduced in previous lessons and illustrated in the QC planning applications. I purposely avoided this basic theory on quality planning models in earlier discussions so you could get an overview of the QC planning process, identify the tools that are needed, and see how these tools can be used to select statistical QC procedures that are appropriate for the analytical or clinical quality requirement for the test.
In learning difficult subjects such as statistical QC, it is often easiest to focus on a problem and then learn the statistics as tools that can be used to solve that problem. Statistical tools that are learned this way have a good chance of becoming part of an analyst's toolbox and being used in the future when similar problems are encountered. The statistical tools needed for the validation of QC procedures - power function graphs, critical-error graphs, and OPSpecs charts - should be part of the toolbox, preferably on the computer desktop of anyone responsible for analytical quality management in healthcare laboratories.
How much imprecision and inaccuracy are allowable with a measurement procedure? What control rules and number of control measurements are necessary for a control procedure? These are difficult questions because they depend on the quality you want to achieve. Specific answers depend on the quantitative relationship between the quality required and the performance factors that contribute to the variability of a test result. A quality-planning model is needed to describe that relationship.
The factors of interest include the performance characteristics of measurement procedure (imprecision, inaccuracy), performance characteristics of a control procedure (probabilities for error detection and false rejection), as well as pre-analytic factors (such as within-subject biological variation). Both analytical and clinical quality requirements are of interest, therefore at least two quality-planing models are needed. The analytical model can relate the performance characteristics of measurement and control procedures to the total error allowable, which is often defined by proficiency testing criteria for acceptability. The clinical model can relate both analytical and preanalytical characteristics to a medically significant change or decision interval, which is a customer- oriented requirement that depends on the medical application and interpretation of a test result. [See the discussion of Quality Goals, Requirements, and Specifications as well as example lists of the analytical quality requirements and clinical quality requirements.]
The concept of a quality-planning model can be understood by analogy with a budget . A budget is a management tool for planning expenditures and allotting certain amounts of the available resources to different categories of expenses. Our experiences with budgets are not always pleasant because, as individuals, we often find our expenses can easily exceed our resources. The same happens to be true for individual tests in laboratories. While our budgeting experiences may not always have been pleasant, the knowledge gained can prevent some much more unpleasant consequences, such as bouncing a check or reporting invalid test results.
You have a job that pays a certain salary and gives you a monthly check that you can spend - let's call this NMavail for the Net Money available. You may spend your money for many things, such as (A) variable living expenses for food, clothing, etc., (B) fixed expenses for rent or house payment, gas and electricity, telephone, car payment, etc., and (C) long term savings for a down payment on a home, contingency fund for emergency expenses, college fund, retirement investments, etc. For this situation, a personal budget could be formulated as follows:
NMavail = Avariable + Bfixed + Csavings
The idea of this budget is to plan your expenditures so that A+B+C will not exceed your take home pay, assuming you want to be fiscally responsible and maintain a balanced budget. In preparing this budget, you will find that NMavail is well defined, you have relatively good data on what you have observed for your variable and fixed expenses, and you will need some discipline if you are to save money to achieve your hopes and desires.
Although a personal budget is often constructed based on take-home pay, it could also be based on gross income. For example, consider the situation if you were running your own business as an independent laboratory consultant. You would then need to construct your budget around the gross money expected (GMexpected), as illustrated below:
GMexpected = Avariable + Bfixed + Csavings + Ddeductions
Additional deductions must be budgeted, for example, category D may include business expenses, federal income taxes, social security taxes, health insurance, etc. A budget based on gross income is more complicated because there are more categories or variables that must be included.
You can develop your budget based on either of these approaches (take-home pay or gross income). If the starting point is take-home pay, some components have already been deducted in your budget. If the starting point is gross income, you must allow for certain deductions so you won't overspend in your categories for living expenses. Regardless of which starting point you choose, you should arrive at reliable figures for what you can spend for living expenses and the margin of safety needed to satisfy your goals.
Similar types of budgets can be constructed to relate the errors expected in a testing process to the final quality needed in a test result. We can think of these as error budgets, or quality-planning models. They describe the relation between the pre- analytical and/or analytical factors that contribute to the variation of a testing process and the clinical or analytical outcome requirement for a test result. Analytical factors may include measurement error due to the stable imprecision and stable inaccuracy of the measurement procedure, as well as the sensitivity of the quality control procedure, which describes the capability of detecting changes or unstable performance of the measurement procedure. Pre-analytical factors include the subject's own biological variation, sampling variation in obtaining a specimen that is representative of the subject, and sampling bias due to the nature, handling, and stability of the specimen obtained.
For comparison, a personal budget based on take-home pay is like an analytical quality-planning budget for a laboratory test. Unfortunately, the limit for the budget has not been well defined for laboratory tests in the past, thus there has actually been little planning for quality. Today, however, regulations for proficiency testing often define criteria for acceptable performance in the form of the total errors that are allowable, thus the quality requirements for many tests are now defined in regulations. Categories A and B of variable and fixed expenses correspond to the imprecision and inaccuracy that are the fundamental performance characteristics of the measurement procedure. These characteristics are often well-studied and good estimates are generally available from laboratory data. Category C - the margin of safety to guard against disaster - corresponds to laboratory QC. All laboratories do QC, just like all people have some savings. The question is whether the savings have been properly planned and are adequate for the situation. In laboratory error budgets, like personal budgets, categories A and B tend to preoccupy the budgeting process, causing category C to be neglected.
A personal budget based on gross income is similar to a clinical quality-planning model for a laboratory test. The gross variation experienced by the customer or user of laboratory tests is larger than just the analytical factors. Preanalytical factors, analogous to the items in category D, contribute to the total variation experienced by the physician when interpreting a test result. These preanalytical components need to be deducted before determining the variation due to analytical components.
To understand what is new and different with these quality-planning models, compare the conventional total error budget to the analytical and clinical quality-planning budgets shown in the accompanying figure. Notice that the conventional total error budget that is widely accepted is made up of two expenses - the stable imprecision and stable inaccuracy observed for the method. The analytical quality planning model adds a category for the QC safety margin. The bottom line for the conventional total error budget and the analytical quality planning budget are the same. However, the analytical quality planning model will demand better imprecision and inaccuracy (smaller CVs and biases) in order to provide the margin of safety needed for QC.
The clinical quality planning model includes the additional category for preanalytical factors and the budget is also different in order to cover the additional expenses. It is not clear whether the clinical or analytical budget will be more demanding because the relationship between the clinical and analytical quality requirements is not known. In addition, the size of the preanalytical deductions depends on the specific test of interest, particularly the within subject biological variation that varies quite widely from test to test.
This pictorial perspective may be a sufficient explanation to understand the concept of the quality planning models. These models provide the rationale for mathematical equations, from which the critical-sizes of medically important errors can be calculated and imposed on power curves to provide one of our quality planning tools - the critical-error graph. These models also can be displayed as charts of operating specifications that show the relationship between an analytical or clinical quality requirement, the imprecision and inaccuracy that are allowable, and the QC that is necessary to assure the desired quality is achieved.
For those who want the details, go on to the next article that provides a mathematical description of laboratory error budgets.
- Westgard JO. Error budgets for quality management: Practical tools for planning and assuring the analytical quality of laboratory testing processes. Clin Lab Manag Review 1996;10:377-403.
- Westgard JO, Burnett RW. Precision requirements for cost-effective operation of analytical processes. Clin Chem 1990;36:1629-32.
- National Cholesterol Education Program Laboratory Standardization Panel. Current status of blood cholesterol measurements in clinical laboratories in the United States. Clin Chem 1990;36:1629-32.
- Westgard JO, Wiebe DA. Cholesterol operational process specifications for assuring the quality required by CLIA proficiency testing. Clin Chem 1991;37:1938-44.
- Westgard JO, Hytoft Petersen P, Wiebe DA. Laboratory process specifications for assuring quality in the U.S. National Cholesterol Education Program (NCEP). Clin Chem 1991:37:656-661.