Upsala J Med Sci 91: 205-208, 1986 Analytical Quality, Test Quality and Quality Costs Torgny Groth Unit of Biomedical Systems Analysis, Uppsala University, Uppsala, Sweden ABSTRACT This paper discusses briefly analytical quality and i t s determinants design of quality assurance procedures on the basis of quality specifications and quality-cost rrrodels test quality ar.d medical costs external quality assessment afid harmonization of clinical chemical test results the potential role of computers, database and knowledge-based systems in achieving quality goals in decentralized analytical activities. ANALYTICAL QUALITY The analytical quality of a laboratory test result is determined not only b y the quality of the analytical method prcper, but also b y t h e calibration procedure and the control procedure. The saying that 'A chain is no stronger than the weakest link', applies also here, and it is necessary to consider all three links when attempling to optimize the performance of the total analytical procedure. Analytical quality is generally expressed in terms of analytical imprecision and analytical bias. This may be appropriate during stable analytical performance. However, the existence of analytical dis- turbances resulting in e . g . systematic shifts in baseline and increases in the inherent random e r r o r , alsc influences on the analytical quality of reported test results. The need of a quality control procedure depends on the fre- quency of occurencc and the magnitude of these disturbances (11). For instance, i f analytical disturbances occur very seldom or never at all, or if the magnitude is negligible compared to allowable analytical e r r o r s , we do not actually need a quality control system. Or) the other hand, i f we do not know the nature of analytical disturbances we require an effective quality coritrol procedure to find out. 205 DESIGN O F QUALITY ASSURANCE P R O C E D U R E S Given the limits for allowable analytical e r r o r s we can estimate critically. sized analytical disturbances, i. e. how large systematic shifts and how large increases in the inherent random e r r o r we can accept without jeopardizing the quality goal. Recommendations of allowable analytical e r r o r s in various clinical situations have been published ( c f . 5,6), but much work still remains to be done to assess the quality requirements in primary care. Such quality goals derived from medical needs form a rational starting point for a quanti- tative design of quality assurance procedures ( 8 ) . Computer simulation techniques a r e then very helpful to evaluate various statistical control rules and to determine the optimal number of controls and calibration standards for a particular analytical method (1). A suitable optimization criterion would be to minimize the quality-costs of the procedure ( 1 2 ) . This formulation of the problem considers costs related to r e r u n s caused by t r u e and false rejects of analytical r u n s , requests for r e r u n s in connection with false (and true) accepts, in additon to costs for calibration and control material, training of personnel etc. The multi-rule Shewhart procedure published as a "proposed selected method" in clinical chemistry (91, was designed to be applicable also in small laboratories without computer support. It is not very likely, how- e v e r , that primary care doctors will plot these types of control charts manually. Microcomputers a r e required for implementation of working quality control procedures i n these settings (10). More complex trend analysis techniques could also be easily applied for continuous monitoring of critical changes in accuracy and imprecision. Colour-coded displays would fascilitate the interpretation and communication of actual control status to non-labora- tory personnel. TEST QUALITY AND MEDICAI, COSTS A clinical chemistry test is not only an analytical determination but also a decision criterion. The quality of a test therefore also depends on the choice of the discriminatory limit. It should be realised that the conventional reference limits for healthy individuals provided by most central laboratories do not generally constitute optimal discriminatory limits. Such limits can only be calculated after specification of the relative importance to avoid m i s - classification of different types (falsely positive and falsely negative out- comes) in order to minimize costs for misclassification. Costs and incon- veniences related to misclassification may, to the extent they are quanti- fiable, be used a s measures of the importance to avoid misclassification. These costs include "financial cost" for personnel, investigations and care etc. , "health costs" related to mortality, morbidity, pain and anxiety in connection with diagnostic investigations and therapy. 206 The optimal value of the discriminatory limit and the predictive value of a test also critically depends on the prevalence of disease in the situation the test is applied. The analytical quality may also influence on the useful- ness of the test. Computer simulation is a powerful technique for studying this problem complex. This is illustrated in ref. ( 3 ) as applied to screening of patients with hyperparathyroidism with use of serum calcium determina- tions. The same technique can also be applied to assess quality requirements in connection with diagnosis and patient monitoring. HARMONIZATION OF TEST RESULTS An external quality assessment program should lead to some concrete measures to decrease the variation between various laboratories and differences between single instruments within a laboratory. This problem of "harmonization" of reported test results could be approached by correction of systemic differences a s estimated from analyses of "accuracy assessment standards" and comparison with definitive o r reference methods. This problem will grow in importance with the increased decentralisation of labora- tory services. The conditions for a successful implementation of such a 'lharmonization program" has been investigated by computer simulation ( 4 ) , and is practiced in full scale in connection, with health screening in France (P. Valdiquik , personal communication). Database management systems ( 2 ) and data communication technology ( 7 ) are necessary prerequisites for a successful achievement. TOOLS FOR IMPLEMENTATION OF QUALITY ASSURANCE PROCEDURES A s mentioned above microcomputers a r e necessary for implementation of working quality assurance procedures in decentralized laboratory activities. Computers w i l l probably be commonly available in primary care centers and the doctor's office for word processing, patient administration e t c . , so there should be no severe economical constraints. First generation quality control packages have been commercially available for some years ( 1 0 ) . The next generation of these packages will provide powerful database management facilities, advanced colour graphics and sophisticated reasoning mechanisms providing advice on how to do in various situations to assure the specified quality. Transmission of quality control data between laboratories for ex- ternal quality assessment and harmonization programs is another important facility of such a workstation. Knowledge-based systems a r e also of great interest for supporting proper test selection and interpretation of test results, considering knowledge on age-, sex- and disease related reference values, optimal discriminatory limits, and the influence of d r u g s on analytical determinations. 207 1. 2 . 3. 4 . 5 . 6 . 7 . 8 . 9 . 1 0 . 11. 1 2 . REFERENCES Groth, T . , Falk, H . & Westgard, J . O . : A quality control simulator for. design and evaluation of internal quality control procedures. Scand J Clin Lab Invest 4 4 (suppl. 172):195-201, 1984. Groth, T., Larsson, O . , Aronsson, T. & de Verdier, C-H. : Data base management systems for evaluation of analytical procedures. Scand J Clin Lzb Invest 44 (suppl. 172):209-213, 1984. Groth, T . , Ljunghall, S . & de Verdier, C-H. : Optimal screening for patients with hyperparathyroidism with use of serum calcium observa- tions. A decision-theoretical analysis. Scand J Clin Lab Invest Groth, T . : Strategies for decreasing interlaboratory variation. Studies by computer simulation. Scand J Clin Lab Invest 4 4 (suppl. 171):331-346, 1984. Hdrder, M. (ed) : Assessing quality requirements in clinical chemistry. Report by a project group initiated by Nordic Clinical Chemistry Project. Scand J Clin Lab Invest 40 (suppl 1 5 5 ) : l - 1 4 4 , 1980. Skendzel, L.P. , Barnett, R . N . & Platt, R . : Medically useful criteria for analytic performance of laboratory tests. A m J Clin Pathol 83:ZOO-205, 1985. ValdiguiC , P.M., Fernet, P. .% Verdier, B . : Transmission of quality control data. In Proceedings of Fifth Int Cont on Computing in Clinical Laboratories, S t u t t g a r t , June 1985. Westgard, J . O . , Groth, T. & de Verdier, C-H.: Principles for developing improved quality control procedures. Scand J Clin Lab Invest 4 4 (suppl 172) : 1 9 - 4 1 , 1984. Westgard, J . O . , B a r r y , P . , Hunt, M., & Groth, T . : A multi-rule Shewhart chart for quality control in clinical chemistry. Clin Chem Westgard, J . O . & Groth T . : Computer systems for implementation of internal quality control procedures. Scand J Clin Lab Invest 44 (suppl Westgard, J.O. & Groth, T . : A predictive value model for quality control. Effects of the prevalence of e r r o r s on the performance of control procedures. Am J Clin Path 80:49-56, 1983. Westgard, J . O . , Hyltoft Petersen, P. & Groth, T.: The quality-costs of an analytical process. Development of quality-costs models based on predictive value theory. Scand J Clin Lab Invest 4 4 (suppl 4 3 : 6 9 9 - 7 0 7 , 1983. 27:493-501, 1981. 172):203-207, 1984. 172):221-227, 1984. 208