1995 Institute: Frontiers in Laboratory Practice Research

Many strategies have been proposed for the setting of quality goals in laboratory medicine. Traditional strategies are based on the use of reference intervals, opinions of clinicians, the state of the art, opinions of experts, assessment of the effect of error on clinical characteristics of tests, and biological variation. All have advantages and disadvantages but the use of data on biological variation appears to have many telling merits. The database is large and within-subject biological variation appears generally constant. Goals based on biology are available for bias as well as imprecision and the same theories can be used to generate goals for drug assays. A variety of proposals has recently been advanced for general models involving analyzing the effects of errors on diagnostic efficiency and sensitivity and a model for the allowable difference between two methods; it is of much interest that all of these generate quality goals based upon fractions of biological variation. Although in the U.S., it appears as if the advent of CLIA’88 has caused many concerns, in Europe there has been much recent rediscussion on goal setting. Approaches based on biology are much favored and, in essence, recommendations are that: desirable imprecision is less than or equal to one-half of the average within-subject biological variation, desirable bias is less then or equal to one-quarter of the group (withinplus between-subject) biological variation, and desirable difference between methods (or drift during monitoring) is less than one-third of the average within-subject biological variation. It is important to recognize that imprecision and bias must be considered together and that, when imprecision approaches the goal, bias must be small, and vice versa. Disseminating goals must be more actively pursued by all involved in laboratory medicine including journal editors and referees, industry, and organizers of PT and EQA schemes. Introduction Quality management processes, including aspects of health care. However, to use such the essential components of control, quality tools correctly, it is necessary first to assessment, assurance, planning, define the standards of quality required. improvement, and audit, have pervaded Perusal of the literature might suggest that manufacturing and service industries in both problems are few because there are many private and public sectors, including all 252 1995 Institute: Frontiers in Laboratory Practice Research papers, reviews, and book chapters dealing documented. with the generation and application of quality goals in laboratory medicine . It does 1 appear, however, that many still have difficulties in deciding the standards which Traditional approaches for setting goals should be attained by laboratories, ideally for for precision (often used for total error, all performance characteristics, but however) are based on use of (I) reference particularly for the important reliability values, (ii) the opinions of clinicians, (iii) the characteristics of precision and bias state of the art, (iv) views of expert (accuracy). There are plausible reasons for individuals and groups, (v) analysis of the this including the facts that: effect of error on the clinical utility of tests, tests are used in many clinical situations advantages and disadvantages which have and it might be that there is no single set been discussed in detail previously and are of goals which would make a method only summarized here . Reference intervals suitable for all purposes, are available for most quantities and the there are many recommendations in the reference interval chosen to set goals are literature and it might be difficult to empirical, and reference intervals depend on choose the most appropriate, the precision and bias of the analytical new recommendations continue to appear, statistical technique used for data reduction. suggesting that there is no professional The opinions of clinicians have been mainly consensus on the topic, obtained by questionnaire involving clinical there appears to be no evidence that between two results (or a result and a patients have been harmed by current reference limit) are not due only to random performance, analytical error as generally supposed but in countries which have legislation analytical variation, the use of the median involving proficiency testing, the focus result satisfies only half of the respondents, might be simply on achieving the and the probability with which decisions are standards required to pass, and made is not always P < 0.05. The state of industry does not appear to use laboratories, changes with time, laboratories professionally set goals as major may adopt special techniques in the analysis considerations either in development or of samples circulated in proficiency testing marketing. (PT) or external quality assessment (EQA) In view of the apparent lack of ubiquitous usually derived, and the matrix of the use of numerical quality goals, these will be samples may not be the same as samples briefly critically reviewed here and then the from patients. The views of expert widely accepted current recommendations individuals and groups, although interesting, Traditional strategies for setting quality goals and (vi) biological variation. All have