International Journal of Health Care Quality Assurance Patient results and laboratory test accuracy Alexander Katayev James K. Fleming

Article information:

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

To cite this document: Alexander Katayev James K. Fleming , (2014),"Patient results and laboratory test accuracy", International Journal of Health Care Quality Assurance, Vol. 27 Iss 1 pp. 65 - 70 Permanent link to this document: http://dx.doi.org/10.1108/IJHCQA-09-2012-0092 Downloaded on: 30 January 2016, At: 15:45 (PT) References: this document contains references to 4 other documents. To copy this document: [email protected] The fulltext of this document has been downloaded 132 times since 2014*

Users who downloaded this article also downloaded: S. Kanitvittaya, U. Suksai, O. Suksripanich, V. Pobkeeree, (2010),"Laboratory quality improvement in Thailand's northernmost provinces", International Journal of Health Care Quality Assurance, Vol. 23 Iss 1 pp. 22-34 http:// dx.doi.org/10.1108/09526861011010659 Khwanjai Wangkahat, Somboon Nookhai, Vallerut Pobkeeree, (2012),"Public health laboratory quality management in a developing country", International Journal of Health Care Quality Assurance, Vol. 25 Iss 2 pp. 150-160 http:// dx.doi.org/10.1108/09526861211198317 Cecilia Mercieca, Sara Cassar, Andrew A. Borg, (2014),"Listening to patients: improving the outpatient service", International Journal of Health Care Quality Assurance, Vol. 27 Iss 1 pp. 44-53 http://dx.doi.org/10.1108/ IJHCQA-03-2012-0033

Access to this document was granted through an Emerald subscription provided by emerald-srm:382916 []

For Authors If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.

About Emerald www.emeraldinsight.com Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online products and additional customer resources and services. Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation. *Related content and download information correct at time of download.

The current issue and full text archive of this journal is available at www.emeraldinsight.com/0952-6862.htm

Patient results and laboratory test accuracy

Patient results and laboratory test accuracy

Alexander Katayev and James K. Fleming Department of Science and Technology, Laboratory Corporation of America Holdings, Elon, North Carolina, USA

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

Abstract

65 Received 19 September 2012 Revised 8 November 2012 Accepted 17 December 2012

Purpose – Traditional quality control materials used for monitoring the clinical laboratory test accuracy might be non-commutable with patient samples and may not detect systematic errors. The aim of this paper is to describe a method to monitor inter-instrument bias using result distributions that are independent of the control’s commutability. Design/methodology/approach – Serum calcium data collected within a laboratory network were assessed. A reference interval was calculated using a computerized, indirect Hoffmann’s algorithm using all data across a laboratory network without excluding any results. Results outside the reference interval were considered as the zero-bias distribution. Three allowable bias levels were then calculated to determine the corresponding shift in abnormal results for each bias level in both directions from the zero-bias distribution. The observed result distributions in three laboratories within the network were compared for bias performance after one year of the reference interval study. Findings – Performance levels for bias were: minimum allowable , 1.27 percent; desirable , 0.85 percent; and optimal , 0.42 percent. Zero bias result distribution above and below the reference interval for calcium was 3.92 percent and 2.53 percent respectively. All three laboratories performed within the desirable allowable bias level. Originality/value – Bias-monitoring process using patient result distributions allows managers to: assess systematic error between laboratory instruments; improve laboratory quality control; and strengthen patient risk management. Keywords United States of America, Quality control, Bias, Analytical performance, Laboratory tests, Reference interval Paper type General review

Introduction Constantly analyzing laboratory test performance that meet clinicians’ expectations, assure test accuracy and interpretation is crucial for care quality. Ideally, laboratory personnel establish accurate population-based reference intervals (RI) and keep analytical bias and imprecision under tight control within an optimal goal based on medical needs and biology. All quality control (QC) material used to maintain this performance has to be commutable with patient samples. However, what is achievable by laboratory technology may not be what is desirable for medical care. This discrepancy is not well communicated between laboratory and medical staff, and analytical quality performance goals are often set arbitrarily with achievable limits in mind. Recently, considerable effort was made to link analytical and clinical goals by establishing analytical goals based on biological data and optimizing performance regarding medical strategy (Klee, 2010). Assessing analytical bias is traditionally The authors thank Mark Sharp for collecting the laboratory data.

International Journal of Health Care Quality Assurance Vol. 27 No. 1, 2014 pp. 65-70 q Emerald Group Publishing Limited 0952-6862 DOI 10.1108/IJHCQA-09-2012-0092

IJHCQA 27,1

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

66

conducted by testing QC material, proficiency samples or by performing multi-instrument comparison studies. Quality control material remains the laboratory quality management program’s backbone and when the predefined goals for bias are met, it is assumed that patient results are consistent. However, a recent study reported that non-commutability of QC materials owing to “matrix effects” is a significant issue for many analytical methods. The numerical relationship of results from QC material and patient samples is often different and QC data may not reflect changes in patient sample measurements. This observation may also be applied to proficiency testing samples and other artificial materials including standards and calibrators (Miller et al., 2011). Methodological shifts that result in analytical bias are especially important for measurands (particular quantities subject to measurement) with a narrow RI like calcium and when a result is close to the RI’s upper or lower limit. In this situation, medical strategy may be critically affected. Reasons for shifts are well known to laboratory professionals, analytical reagent and instrument manufacturers. They include lot-to-lot changes in calibration assignment, antibody affinities, raw materials, instrument aging and environmental changes. Controlling bias in those situations may be achieved by performing preventive maintenance procedures, recalibration and other methods. However, these are performed only if excessive bias is detected in a timely manner. Negligible biases in the same direction will cause a significant drift over time. When bias troubleshooting methods do not work, there are only two options: eadjust the calibration to eliminate bias and keep the existing RI; or change the existing RI by performing a comprehensive RI study. The latter is rarely done owing to RI re-evaluation complexity and cost. Patient care is dependent upon a medical strategy that relies on RI accuracy that is not affected by analytical bias. This article describes a simple and reliable analytical bias monitoring method using patient result distribution independently from QC material commutability that will maintain selected quality goals. Materials and methods Our study used nationwide test data stored in the Laboratory Corporation of America information system (www.labcorp.com). Patient test results for serum calcium were chosen to describe the monitoring analytical bias method. However, any applicable measurand (such as any routine chemistry test) can be used. Patient test results for the RI and bias assessment studies were processed as described by Katayev et al. (2010). Our study protocol was exempt under the institutional review board’s existing regulations (i.e. ethical approval was not required). Calcium was measured on Roche/Hitachi MODULAR (Roche Diagnostics, Indianapolis, IN, USA) in all laboratories using the same reagent lots, calibrators and controls. Results Analytical conditions during the initial RI study should be considered a zero bias (or benchmark) method performance. Bias during benchmarking will not influence clinical outcomes if medical decisions are based on results from standardized instruments within the laboratory network. The RI for calcium was calculated using a modified computerized indirect Hoffmann method (Katayev et al., 2010). The study used data from 11 regional laboratories within the network. A total of 58,452 test results from

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

patients 18-60 years old were processed by the RI software. No results were excluded. The RI for calcium was 2.175 mmol/L-2.550 mmol/L (central 95 percent) with a median of 2.375 mmol/L. To use the same RI for any standardized instrument network, the quality specifications for bias should meet one of three performance levels calculated from biology: minimum; desirable; or optimal (Fraser, 2001). Mathematically, those levels are expressed as: Minimum allowable performance:

B , 0.375 (CVi2 þ CVg2)1/2.

Desirable performance

B , 0.250 (CVi2 þ CVg2)1/2.

Optimal performance:

B , 0.125 (CVi2 þ CVg2)1/2.

Where: B is allowable analytical bias, CVi is within-subject and CVg is between-subject biological variation of measurand (Fraser, 2001). Using published data for biological variation for calcium (Fraser, 2001) and the calculated median from the RI study, the performance levels for bias were: Minimum allowable:

, 1.27% (^ 0.03 mmol/L at median).

Desirable:

, 0.85% (^ 0.02 mmol/L at median).

Optimal:

, 0.42% (^ 0.01 mmol/L at median).

Test results that fall outside the RI will increase when bias escalates in either direction and the percent increase may be calculated (Fraser, 2001). With zero bias, results above and below the RI for “normal” subjects using 95 percent confidence interval should be 2.5% þ 2.5% ¼ 5.0%. For optimal, desirable and minimum allowable bias in the “normal” population, the maximum percentage of results outside the reference limit in the bias direction will be 3.3 percent, 4.4 percent and 5.7 percent respectively. For the reference limit opposite to the bias direction, it will be 1.8 percent, 1.4 percent and 1.0 percent respectively (Fraser, 2001). Using data from the benchmark RI study, results falling above and below the RI for the entire “normal” and “abnormal” population may be determined (Table I). Next, we calculated the expected distribution above and below the RI for each bias threshold applied in both directions (Table I and Figure 1). Laboratories serving comparable populations should have similar distributions above and below the RI and that the abnormal-population percentage is relatively constant. Data were collected within a ten day period, one year after the RI study in three regional laboratories to assess analytical bias using percentages above and below the RI for calcium. Total results included: 93,767 for laboratory 1; 172,235 for 2; and 71,208 for laboratory 3. No results were excluded. The observed distribution outside the RI and performance assessment based on quality goals for bias in those laboratories is presented in Table I. The distribution above and below the RI fell within either the “Desirable” or “Optimal” limits. Since the network quality goal for this measurand was established to meet “Desirable” performance level, all laboratories met the quality standard. Discussion Our bias assessment method using patient result distribution permit bias monitoring that is independent from control material commutability. When used in conjunction

Patient results and laboratory test accuracy 67

Table I. Quality goals for bias and performance assessment example for calcium, using patient results 3.33

4.43

5.73

3.22

2.82

2.42

, 20.42/20.01

2 0.85/20.02

, 21.27/20.03

1.83 2.53

3.92

0/0 (0)

4.72

1.43

5.82

,0.85/0.02

,0.42/0.01

1.03

7.12

,1.27/0.03

8.15

7.25

6.55

6.45

6.55

7.25

8.15

3.14

Observed % of results above RI

2.58

5.72

Laboratory 1 Total % outside RI

Observed % of results below RI

Within the benchmark

Comment

5.82

Observed % of results above RI

1.30

7.12

Laboratory 2 Total % outside RI

Observed % of results below RI

Within desirable bias

Comment

4.24

Observed % of results above RI

2.03

6.2

Laboratory 3 Total % outside RI Observed % of results below RI

Within optimal bias

Comment

Notes: aBias threshold level acceptability criterion (optimal, desirable, or minimum allowable) to be determined by individual laboratory based on the risk management assessment; bCalculated percent bias/absolute bias in mmol/L that corresponds to the listed bias threshold level

Zero bias (benchmark) Optimal negative bias Desirable negative bias Minimum allowable negative bias

Optimal positive bias

Minimum allowable positive bias Desirable positive bias

Bias threshold levela

68

Calculated specifications for bias using different performance levels Total Expected % Expected % % of results of results outside %/mmol/L at above RI below RI RI Medianb

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

IJHCQA 27,1

Patient results and laboratory test accuracy

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

69

Figure 1. Relationship between bias and the percentage of the abnormal results for calcium

with traditional QC procedures, it controls performance, detects shifts in patient results in a timely way and applies corrective actions such as instrument-maintenance, recalibration or adjusting calibration set points. If the sum of percentages above and below the RI exceeds what is the expected, it may indicate that laboratory staff need to perform a multi-instrument comparison study using patient samples because instruments may exhibit biases in the opposite direction. This method may be especially useful when laboratory staff within a network are using different generations of the same instruments or undergoing reagent or calibrator lot changes. Our method is limited by the need for statistically significant observations and standardization across laboratories. It is the laboratory staff’s responsibility to determine acceptable performance for each measurand. Results distribution monitoring is a simple and reliable method for controlling the analytical bias, independent of control commutability and encourages proactive measures to avoid erroneous medical decisions.

IJHCQA 27,1

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

70

Conclusion and future applications Our analytical performance monitoring method may be easily implemented at a low cost in any laboratory network or individual laboratories that use multiple standardized instruments. Considering how widely advanced laboratory information systems are used, result distributions can be monitored in real time and shifts in the distribution may be detected without compromising integrity and accuracy. Traditional QC events are only performance snapshots and overall quality assurance will undoubtedly improve when laboratories have tools for continuous monitoring. References Fraser, C.G. (2001), Biological Variation: From Principles to Practice, American Association for Clinical Chemistry, Washington, DC. Katayev, A., Balciza, C. and Seccombe, D.W. (2010), “Establishing reference intervals for clinical laboratory test results: is there a better way”, American Journal of Clinical Pathology, Vol. 133, pp. 180-186. Klee, G.G. (2010), “Establishment of outcome-related analytic performance goals”, Clinical Chemistry, Vol. 56 No. 7, pp. 14-22. Miller, W.G., Erek, A., Cunningham, T.D., Oladipo, O., Scott, M.G. and Johnson, R.E. (2011), “Commutability limitations influence quality control results with different reagent lots”, Clinical Chemistry, Vol. 57 No. 1, pp. 76-83. Corresponding author Alexander Katayev can be contacted at: [email protected]

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

This article has been cited by:

Downloaded by University of Queensland At 15:45 30 January 2016 (PT)

1. Martin Risch, Dominik W. Meier, Benjamin Sakem, Pedro Medina Escobar, Corina Risch, Urs Nydegger, Lorenz Risch. 2015. Vitamin B12 and folate levels in healthy Swiss senior citizens: a prospective study evaluating reference intervals and decision limits. BMC Geriatrics 15. . [CrossRef] 2. James K. Fleming, Alexander Katayev. 2015. Changing the paradigm of laboratory quality control through implementation of real-time test results monitoring: For patients by patients. Clinical Biochemistry 48, 508-513. [CrossRef]

Patient results and laboratory test accuracy.

Traditional quality control materials used for monitoring the clinical laboratory test accuracy might be non-commutable with patient samples and may n...
194KB Sizes 3 Downloads 3 Views