Abstract Objectives There is currently a lack of an outcomes-based definition of critical values for the pediatric population. This has contributed to a highly heterogeneous critical value reporting practice between laboratories. Methods Anonymized results were extracted from a laboratory information system for 10 biochemistry tests. The probability of high-dependency/intensive care unit admission (as a proxy for adverse outcomes) for each individual laboratory concentration was calculated and adjusted to fit using a polynomial function to model the probability trend. The laboratory value that intersected the 90% probability trend line was considered the critical value threshold. Results The critical value thresholds for the serum analytes were sodium (mmol/L: <131, >148), potassium (mmol/L: <2.4, >6.4), bicarbonate (mmol/L: <13, >37), chloride (mmol/L: <91, >115), urea (mmol/L: >12), creatinine (μmol/L: >129), glucose (mmol/L: >17.2), total calcium (mmol/L: <1.9), magnesium (mmol/L: <0.6, >1.2), and phosphate (mmol/L: <0.4, >2.6). Conclusions This study described an approach to derive contemporary pediatric critical value thresholds. Critical values, Pediatric, Children, Outcomes based, Evidence based, Panic value, Alert value A critical (or panic) laboratory value is a laboratory test result that represents a pathophysiologic state at such variance with normal as to be life-threatening if an action is not taken quickly and for which an effective action is possible. 1From a more operational perspective, they are also defined as laboratory results that are associated with 90% probability of death if left untreated. 2The laboratory practice of critical value reporting is highly heterogeneous. A Canadian survey on pediatric critical value reporting found that the thresholds differ significantly between laboratories. 3For example, the lower critical threshold of potassium ranged from 2.5 to 3.0 mmol/L while the upper critical threshold ranged from 5.9 to 8.0 mmol/L. The large difference in critical value thresholds cannot be explained by interlaboratory bias (which is generally <5%) or biological variation (between- or within-subject variation, both <2%). 4,5These differences exist despite industry guidance such as the Clinical & Laboratory Standards Institute (CLSI GP47: Management of Critical- and Significant-Risk Results) and International Organization for Standardization 15189. Given the relatively small between-subject biological variation and interlaboratory bias, it is possible that a significant proportion of residual heterogeneity in practice was due to a lack of evidence base for the critical value thresholds. Consequently, most laboratories determine these thresholds individually, with or without clinician consultation. This is particularly true for the pediatric population as there are inherent difficulties in deriving laboratory thresholds due to resource, ethical, and operational challenges. However, there has been recent recognition that data-mining (indirect) approaches can produce such data. These include reference intervals and biological variation data that would otherwise be difficult to derive directly from children. 5-7 In this study, we derived outcome-based critical value thresholds for the pediatric population using a large hospital data set from a tertiary pediatric hospital in Australia. Materials and Methods Mortality during hospitalization is, fortunately, uncommon in the pediatric population. As such, we defined admission into the high-dependency unit (HDU) or intensive care unit (ICU) as an adverse outcome in lieu of death for the purpose of this study. By this definition, the critical values derived in this study will flag children who require immediate medical attention and are at high risk for HDU or ICU admission (which are clinically important outcomes in the pediatric population). The Women’s and Children’s Hospital is a tertiary care specialist center serving a population of 2 million people from South Australia, Northern Territory, Western Victoria, and New South Wales regions in Australia. This study was exempted from local ethics board review. We extracted from the laboratory information system anonymized results performed by South Australia (SA) Pathology at the Women’s and Children’s Hospital site for 10 biochemistry tests between October 1, 2011, and March 28, 2017. The tests were performed on the Roche Cobas 6000 platform (Roche Diagnostics, Basel, Switzerland). Patients who were younger than 6 months and older than 18 years were excluded from further analysis. Children who are younger than 6 months undergo significant physiologic adaptation to life ex utero and should be considered separately. However, our data set did not contain sufficient data for meaningful analysis of this important cohort. Furthermore, we assumed that age-related dynamic changes in children aged 6 months to 18 years are negligible for the purpose of this study. The remaining patients were then grouped into those who were admitted only to the pediatric general ward throughout their hospitalization and those who had been admitted into the HDU or ICU. For the latter, their lowest and highest laboratory results 24 hours prior to HDU or ICU admission were recorded in two separate data bins and used to derive the lower and upper critical value thresholds, respectively. This procedure ensured that the laboratory results at greatest deviation from normal physiology were included. The 24-hour timeline was selected to maintain the temporal relationship between the laboratory results and the adverse outcomes (HDU or ICU admission). For patients who had only a single laboratory result 24 hours prior to HDU or ICU admission, the same laboratory value was entered into both data bins to ensure equal data density between the two bins. Subsequently, the probability of HDU/ICU admission for each individual laboratory concentration was calculated using the following formula based on the Bayesian theorem: P(y=transferred to ICU/HDU |x=lab values)=P(x=lab values|y=transferred to ICU/HDU)⋅ P(y=transferred to ICU/HDU)P(x=lab values) This calculation was performed for all laboratory tests at fixed concentration intervals Figure 1. The calculated probabilities were then plotted against the concentration intervals and adjusted to fit using a polynomial function to model the probability trend. The laboratory value that intersected the 90% probability trend line was considered the critical value threshold. This above analysis was performed using Python Numpy (http://www.numpy.org), Scipy (https://www.scipy.org), and Pandas packages (http://www.pandas.pydata.org). Figure 1 View large Download slide View large Download slide View large Download slide View large Download slide Scatterplots and fitted polynomial probability trend lines of the tests. The laboratory value that intersected the 90% probability trend line was considered the critical value threshold (lower, left panels; upper, right panels). A, Sodium. B, Potassium. C, Bicarbonate. D, Chloride. E, Urea. F, Creatinine. G, Glucose. H, Total calcium. I, Magnesium. J, Phosphate. Figure 1 View large Download slide View large Download slide View large Download slide View large Download slide Scatterplots and fitted polynomial probability trend lines of the tests. The laboratory value that intersected the 90% probability trend line was considered the critical value threshold (lower, left panels; upper, right panels). A, Sodium. B, Potassium. C, Bicarbonate. D, Chloride. E, Urea. F, Creatinine. G, Glucose. H, Total calcium. I, Magnesium. J, Phosphate. Results In total, 235,890 patient records were extracted from the laboratory information system. Of these, 20,808 patients who stayed only in the general ward and 655 patients who were transferred from the general ward to ICU/HD wards were included in the final analysis. The mean (SD) age of the patients was 9.7 (5.31) years, with a female to male ratio of 10,245 to 11,215, and three had unspecified sex. The scatterplots and fitted polynomial probability trend lines of the tests are shown in Figure 1. The proposed lower and upper critical value thresholds derived from this study compared with the reference intervals and critical value thresholds currently in use at the hospital for 10 biochemistry tests are provided in Table 1. The flag rates of the current and proposed critical value thresholds for these tests are compared in Table 2. Table 1 Summary of the Proposed Lower and Upper Critical Value Thresholds Compared With the Critical Value Thresholds Currently in Use at the Hospital and 14 Canadian Laboratories 3for 10 Biochemistry Testsa Biochemistry Test Unit No. in General Ward No. in HDU/ICU Current Lower Critical Value Thresholds Current Upper Critical Value Thresholds Range of Lower Critical Value Thresholds in Canada 3 Range of Upper Critical Value Thresholds in Canada 3 Proposed Lower Critical Value Threshold Proposed Upper Critical Value Threshold Sodium mmol/L 17,772 638 <130 >150 120-127 150-160 <131 >148 Potassium mmol/L 17,755 637 <3.0 >6.0 2.5-3.0 5.9-8.0 <2.4 >6.4 Bicarbonate mmol/L 17,767 638 <15 >35 10-15 35-40 <13 >37 Chloride mmol/L 17,772 638 <90 >115 NA NA <91 >115 Urea mmol/L 17,749 637 NA >12 NA NA NA >12 Creatinine μmol/L 17,752 637 NA >120 NA NA NA >129 Glucose mmol/L 3,871 34 <2.5 >9.0 1.7-3.0 10-30 NA >17.2 Total calcium mmol/L 7,918 338 <1.8 >3.0 1.5-1.9 2.9-3.5 <1.9 NA Magnesium mmol/L 7,492 340 <0.6 >1.30 0.3-0.55 1.2-2.5 <0.6 >1.2 Phosphate mmol/L 7,492 338 <0.5 >3.0 0.3-0.7 2.87-5.00 <0.4 >2.6 Biochemistry Test Unit No. in General Ward No. in HDU/ICU Current Lower Critical Value Thresholds Current Upper Critical Value Thresholds Range of Lower Critical Value Thresholds in Canada 3 Range of Upper Critical Value Thresholds in Canada 3 Proposed Lower Critical Value Threshold Proposed Upper Critical Value Threshold Sodium mmol/L 17,772 638 <130 >150 120-127 150-160 <131 >148 Potassium mmol/L 17,755 637 <3.0 >6.0 2.5-3.0 5.9-8.0 <2.4 >6.4 Bicarbonate mmol/L 17,767 638 <15 >35 10-15 35-40 <13 >37 Chloride mmol/L 17,772 638 <90 >115 NA NA <91 >115 Urea mmol/L 17,749 637 NA >12 NA NA NA >12 Creatinine μmol/L 17,752 637 NA >120 NA NA NA >129 Glucose mmol/L 3,871 34 <2.5 >9.0 1.7-3.0 10-30 NA >17.2 Total calcium mmol/L 7,918 338 <1.8 >3.0 1.5-1.9 2.9-3.5 <1.9 NA Magnesium mmol/L 7,492 340 <0.6 >1.30 0.3-0.55 1.2-2.5 <0.6 >1.2 Phosphate mmol/L 7,492 338 <0.5 >3.0 0.3-0.7 2.87-5.00 <0.4 >2.6 ICU, intensive care unit; HCU, high-dependency unit; NA, not available. aThe number of patients from the general ward and HDU/ICU who contributed to the derivation of these thresholds is also provided. View Large Table 2 Flag Rates of the Current and Proposed Lower and Upper Critical Value Thresholds When Applied to the Entire Cohort of Patients Aged >6 Months to <19 Years Biochemistry Test Flag Rate for Current Lower Threshold, % Flag Rate for Proposed Lower Threshold, % % Difference (Proposed – Current) Flag Rate for Current Upper Threshold, % Flag Rate for Proposed Upper Threshold, % % Difference (Proposed – Current) Sodium 0.42 0.65 0.23 0.23 0.44 0.21 Potassium 0.9 0.1 –0.8 0.86 0.26 –0.6 Bicarbonate 1.62 0.53 –1.09 0.09 0.05 –0.04 Chloride 0.18 0.24 0.06 0.59 0.59 0 Urea NA NA NA 0.63 0.66 0.03 Creatinine NA NA NA 0.62 0.51 –0.11 Glucose NA NA NA 6.48 1.7 –4.78 Total calcium 0.72 1.54 0.82 0.05 0 –0.05 Magnesium 1.25 1.0 –0.25 0.15 0.31 0.16 Phosphate 0.52 0.29 –0.23 0.09 0.31 0.22 Biochemistry Test Flag Rate for Current Lower Threshold, % Flag Rate for Proposed Lower Threshold, % % Difference (Proposed – Current) Flag Rate for Current Upper Threshold, % Flag Rate for Proposed Upper Threshold, % % Difference (Proposed – Current) Sodium 0.42 0.65 0.23 0.23 0.44 0.21 Potassium 0.9 0.1 –0.8 0.86 0.26 –0.6 Bicarbonate 1.62 0.53 –1.09 0.09 0.05 –0.04 Chloride 0.18 0.24 0.06 0.59 0.59 0 Urea NA NA NA 0.63 0.66 0.03 Creatinine NA NA NA 0.62 0.51 –0.11 Glucose NA NA NA 6.48 1.7 –4.78 Total calcium 0.72 1.54 0.82 0.05 0 –0.05 Magnesium 1.25 1.0 –0.25 0.15 0.31 0.16 Phosphate 0.52 0.29 –0.23 0.09 0.31 0.22 NA, not available. View Large Discussion There is a concerted effort to harmonize laboratory practices globally. This includes the postanalytical phase, where there are multiple national, regional, and international initiatives to harmonize laboratory decision limits such as reference intervals. 8-12In this region, the Australasian Association of Clinical Biochemists has recently published a list of proposed harmonized reference intervals for the pediatric population after years of scientific discussion. 13However, critical value reporting remains a difficult area for such activity due to a persistent lack of evidence that is crucial to inform and guide such discussions. This study defined the critical value thresholds for a list of common biochemistry analytes for the pediatric population that are supported by contemporary laboratory data, clinical practice, and statistical analysis. To properly define critical value thresholds, one should observe the clinical outcome associated with an abnormal laboratory result without any medical intervention so that the clinical impact can be objectively assessed without confounding factor or bias. Clearly, this is unacceptable in today’s research and medical practice. As such, it is reasonable and logical to use data-mining techniques to obtain such data. Overall, the critical value thresholds determined in this study agree well with those that are currently in use in the laboratory. In some instances, the critical value thresholds obtained in this study are more extreme than those currently in use in the laboratory (eg, serum creatinine). Consequently, the proposed critical value thresholds identify fewer children compared with the current thresholds. In part, this may be related to the continuous change in pediatric biochemistry profiles during growth and development, 5-7which are not well accounted for in this study since children within a wide age group (age >6 months and <19 years) were included in the analysis. It will be important to examine age-partitioned critical values using a larger data set to determine whether further refinements are necessary. This is particularly true for analytes that have a significant sex difference and/or age-related changes, such as urea, creatinine, and phosphate. 14 This study failed to determine the lower critical value threshold for glucose. This was because the probability of HDU/ICU admission did not rise steeply enough to breach the 90% probability threshold. In other words, fewer children with a low glucose measurement were subsequently admitted into the HDU/ICU. This may be related to hypoglycemia secondary to prolonged fasting in sick children. Typically, treatment of hypoglycemia is instituted immediately upon detection and reversed in the general wards, thus negating the need for HDU/ICU admission. Another possible explanation for this observation may be that pediatric clinicians have a higher tolerance for low glucose measurements. The concentration at which glucose is considered low is generally accepted as 2.6 mmol/L outside the neonatal population, where debate continues. The critical value thresholds for some tests are asymmetrical, being more extreme at one end of the scale. For example, the physiologic midpoint for potassium can be considered 4.0 mmol/L. The upper critical value threshold is 6.4 mmol/L (2.4 mmol/L away from the midpoint), whereas the lower threshold is 2.4 (1.6 mmol/L away from the midpoint). This finding raises the possible need for reconsideration of the current practice of defining critical thresholds symmetrically around a physiologic midpoint (usually the midpoint of the reference intervals), particularly for analytes that have narrow biologic variation. The asymmetry distribution of the critical value may imply that the risk of adverse outcome is disproportionately larger at one end of the laboratory result. There are several limitations to this study. First, admission into the HDU/ICU was used as the adverse outcome instead of death for the definition of critical value in this study. This may have the effect of attenuating the critical values. However, it can be argued that very high risk (≥90% probability) of admission into the HDU or ICU should be considered a severe outcome, which requires immediate medical attention in the pediatric population. Nevertheless, the goal for a critical value could be considered to avoid an ICU admission and treating before such a move is required. As such, the critical value thresholds established here should be interpreted more conservatively. They are limited to provide only an upper (or lower) limit for a critical value (meaning that the critical value should be no higher [or lower] than the 90% probability point). A second limitation relates to the number of children who had laboratory results 24 hours prior to HDU/ICU admission. This may be improved by expanding the timeline to 48 hours prior to HDU/ICU admission. However, it is important to keep the time interval between the laboratory result and the adverse outcome relatively short to maintain the strength of association. Moreover, it fulfills the original definition of critical value, which was framed around “an imminent threat to the well-being of the subject.” Finally, as with any data-mining study, the results are subject to confounders such as intervening medical treatment that cannot be controlled for retrospectively. On the other hand, it can also be argued that the critical value thresholds derived in this manner already incorporate these clinical considerations and, as such, represent a more practical/operational definition instead of a theoretical one. This study described an approach to derive contemporary critical value thresholds for the pediatric population. These results should be interpreted with caution and should be seen as preliminary data that require further validation on more centers with larger data sets. It is strongly recommended that the clinical end users are consulted prior to implementation of any critical value protocol to ensure optimal use of the laboratory information. 15 References 1. Lundberg GD. When to panic over abnormal values. MLO . 1972; 4: 47- 54. 2. Catrou PG. How critical are critical values? Am J Clin Pathol . 1997; 108: 245- 246. Google Scholar CrossRef Search ADS PubMed 3. Gong Y, Adeli K; CSCC Pediatric Focus Group. A national survey on pediatric critical values used in clinical laboratories across Canada. Clin Biochem . 2009; 42: 1610- 1615. Google Scholar CrossRef Search ADS PubMed 4. Perich C, Minchinela J, Ricós C et al. Biological variation database: structure and criteria used for generation and update. Clin Chem Lab Med . 2015; 53: 299- 305. Google Scholar CrossRef Search ADS PubMed 5. Loh TP, Ranieri E, Metz MP. Derivation of pediatric within-individual biological variation by indirect sampling method: an LMS approach. Am J Clin Pathol . 2014; 142: 657- 663. Google Scholar CrossRef Search ADS PubMed 6. Loh TP, Antoniou G, Baghurst P et al. Development of paediatric biochemistry centile charts as a complement to laboratory reference intervals. Pathology . 2014; 46: 336- 343. Google Scholar CrossRef Search ADS PubMed 7. Loh TP, Metz MP. Indirect estimation of pediatric between-individual biological variation data for 22 common serum biochemistries. Am J Clin Pathol . 2015; 143: 683- 693. Google Scholar CrossRef Search ADS PubMed 8. Aarsand AK, Sandberg S. How to achieve harmonisation of laboratory testing—the complete picture. Clin Chim Acta . 2014; 432: 8- 14. Google Scholar CrossRef Search ADS PubMed 9. Tate JR, Johnson R, Legg M. Harmonisation of laboratory testing. Clin Biochem Rev . 2012; 33: 81- 84. Google Scholar PubMed 10. Tate JR, Johnson R, Sikaris K. Harmonisation of laboratory testing. Clin Biochem Rev . 2012; 33: 121- 122. Google Scholar PubMed 11. Koerbin G, Sikaris KA, Jones GR et al. ; AACB Committee for Common Reference Intervals. Evidence-based approach to harmonised reference intervals. Clin Chim Acta . 2014; 432: 99- 107. Google Scholar CrossRef Search ADS PubMed 12. Tate JR, Koerbin G, Adeli K. Opinion paper: deriving harmonised reference intervals—global activities. Ejifcc . 2016; 27: 48- 65. Google Scholar PubMed 13. Tate JR, Sikaris KA, Jones GR et al. Harmonising adult and paediatric reference intervals in Australia and New Zealand: an evidence-based approach for establishing a first panel of chemistry analytes. Clin Biochem Rev . 2014; 35: 213- 235. Google Scholar PubMed 14. Colantonio DA, Kyriakopoulou L, Chan MK et al. Closing the gaps in pediatric laboratory reference intervals: a CALIPER database of 40 biochemical markers in a healthy and multiethnic population of children. Clin Chem . 2012; 58: 854- 868. Google Scholar CrossRef Search ADS PubMed 15. Don-Wauchope AC, Wang L, Grey V. Pediatric critical values: laboratory-pediatrician discourse. Clin Biochem . 2009; 42: 1658- 1661. Google Scholar CrossRef Search ADS PubMed © American Society for Clinical Pathology, 2018. All rights reserved. For permissions, please e-mail: email@example.com
American Journal of Clinical Pathology – Oxford University Press
Published: Apr 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera