The mathematician Alan Turing famously devised a test to answer the question, "Can machines think?"1 He reasoned that imitation is indistinguishable from the real thing. Our laboratory information systems and instruments do a lot of thinking for us, relieving us from performing calculations such as INR, creatinine clearance and common ratios. In the case of the latter, it's what we do with these calculations that separate techs from the machine.
Clinicians use laboratory test results in four key ways, according to professor of Pathology, Frank Wians Jr, PhD, MT(ASCP), DABCC, FACB. These include hypothesis deduction, pattern recognition, medical algorithms and rifle vs. shotgun approach. Pattern recognition specifically compares a cluster of results to discriminate between competing diagnoses. But interpreting lab testing has pitfalls. He writes, "Quantitative values for all analytes are affected by both imprecision (ie: lack of reproducibility) in the measurement of the analyte and intra-individual variation over time in the concentration of the analyte."
A physician has to decide, in other words, if a changed value is a critical difference.2
In a sense, laboratory technicians and technologists do the same. A bench tech uses information about the specimen, instrument and results (preanalytical, analytical and postanalytical phases of testing) to decide if results are reliable or warrant action before reporting. Bench techs use hypothesis testing and pattern recognition to decide if results are true, not unlike a physician establishing a diagnosis.
Understanding how individual test results relate is an important part of verifying results on the bench. These common ratios can flag sampling or systematic errors that affect quality or give the bench tech confidence that results fit a clinical picture.
Here are three common ratios to consider:
The BUN/creatinine ratio is normally 10-20. An increased ratio may be caused by a decreased blood flow to the kidneys in dehydration, congestive heart failure, or urinary tract obstruction. It can also be increased from an increased protein concentration e.g. G.I. bleeding. It may be decreased due to liver disease or malnutrition. The ratio is normal in most cases of chronic renal disease.3 4
The AST/ALT ratio is normally less than 1. An increased ratio may suggest liver disease. One study indicates a positive predictive value of 69 percent.5 A liver biopsy is indicated to confirm cirrhosis, especially if hepatitis C is suspected.5 In patients with high alcohol consumption, a ratio greater than one suggests advanced liver disease.6
Click to view figure 1
Click to view figure 2
Click to view figure 3
The ratio of hemoglobin to hematocrit is normally a third. This ratio is routinely reported as a percentage with a complete blood count as the MCHC (Mean Corpuscular Hemoglobin Concentration). Low values may indicate iron deficiency and certain cases of thalassemia, while high values may indicate red cell abnormalities or abnormal hemoglobin synthesis.7 8 Since values depends on the structure of the hemoglobin protein within the red cell, this ratio (MCHC) is an excellent quality indicator.
These are only a few. Other ratios to consider include albumin/globulin, chloride/sodium, urine calcium/creatinine, and urine protein/creatinine. Your patient population may suggest unique ratios dependent on gender, age or other factors. Clinicians use these as diagnostic clues, and your laboratory can also use them.
Your laboratory already looks for patterns in quality assessment: single or multiple points that exceed limits, lot-to-lot variation and delta check variation. Monitoring common ratios extends this ability to seek intra-analytical variation within a panel. Patterns of variation can signal analytical errors.
Your information system or middleware can calculate and flag the above, for example:
- Flag BUN/creatinine ratio <3 or >30
- Flag AST/ALT ratio >1
- Flag MCHC (Hemoglobin/Hematocrit ratio) <28 or >37
Your laboratory may already use rules of thumb such as these, but they lack specificity. For example, your AST/ALT ratio is more likely to be elevated with abnormal than normal enzyme values. Your information system can refine your rules to make them most useful to the bench tech. Here are suggestions:
- In addition to reference texts, use data mining. Your information system can run ad hoc reports of patient values by age, gender, admission status, and other variables to let you see what your patients are actually running. This data can be used to increase the specificity and sensitivity of flagging.
- Automate the process. The more calculations done at point of verification, the better. Many systems can display "no print" calculations and other data to techs only to be used on the bench; if this isn't possible a "For Lab Use Only" section on the report with more obscure flagging works. In the CPSI system, for example, the above numbers can be translated into binary flags that change an interpreted flag into a simple instruction e.g. ASTALT for technologists.
Consider the Figures showing data from our laboratory at Penobscot Valley Hospital that illustrate relationships between ratios and other variables. Figure 1 shows AST/ALT ratio plotted against AST values with a normal range indicated by vertical lines, visually suggesting the following rule of thumb:
Flag if AST <60 U/L and AST/ALT ratio >1
A BUN/creatinine ratio is less linear as seen in Figures 2 and 3, which plots significantly different slopes depending on the BUN value. To increase the sensitivity of the flag, try a slightly different rule of thumb:
Flag if predicted creatinine (using y = mx + b) does not match reported creatinine for BUN < 41 or BUN > 40 mg/dL
Finally, since the amount of hemoglobin in a red blood cell reflects its structure and function, data mining of MCHC values may be less rewarding. A single cutoff (eg: greater than 37) may be indicated to review the sample for plasma abnormalities.
Using knowledge of common ratios, data mining and information system automation, quality outliers can be flagged for your techs that indicate different patterns of variation. Tailoring cutoffs and other variables to your patient population can make review of these deviations more fruitful, saving technologist time and providing better patient care.
Scott Warner is lab manager at Penobscot Valley Hospital in Lincoln, ME.
1. Turing A. Computing machinery and intelligence. http://www.loebner.net/Prizef/TuringArticle.html. 1/22/14.
2. Wians F. Clinical laboratory tests: which, why, and what do the results mean? http://labmed.ascpjournals.org/content/40/2/105.full. 1/23/14.
3. Lab Tests Online. BUN common questions. http://labtestsonline.org/understanding/analytes/bun/tab/faq/. 1/23/14.
4. Quest Diagnostics. BUN/Creatinine ratio. http://www.questdiagnostics.com/testcenter/TestDetail.action?ntc=296. 1/23/14.
5. Reedy DW et al (abstract). Serum HCV-RNA, ALT levels and AST/ALT ratio do not predict severity in chronic hepatitis c virus infection. http://www.natap.org/2000/7thcroi/25rpt21400.html. 1/27/14.
6. Nyblom H et al (abstract). High AST/ALT ratio may indicate advanced alcoholic liver disease rather than heavy drinking. http://www.ncbi.nlm.nih.gov/pubmed/15208167. 1/27/14.
7. Family Practice Notebook. Mean corpuscular hemoglobin concentration. http://www.fpnotebook.com/hemeonc/lab/MnCrpsclrHmglbnCncntrtn.htm. 1/27/14.
8. Insiripong S et al (abstract). Comparison of hematocrit/hemoglobin ratios in subjects with alpha-thalassemia, with subjects having chronic kidney disease and normal subjects. http://www.ncbi.nlm.nih.gov/pubmed/24050107. 1/27/14.
Figure 1: AST/ALT Ratio
Figure 2: BUN/Creatinine Ratio, low BUN
Figure 3: BUN/Creatinine Ratio, elevated BUN