Postanalysis: Medical Decision Making

Key Points

Laboratory results must undergo a two-step postanalytic review for analytic correctness (using delta checks, linearity ranges, etc.) and for clinical significance for the patient (applying critical values, reference ranges, pretest and posttest probability, etc.).
Reference intervals are most commonly defined as the range of values into which 95% of nondiseased individuals will fall. This definition implies that 5% of nondiseased individuals can have laboratory results outside the reference range.
The ability of a test to discriminate disease from no disease is described by the sensitivity and specificity of the test. Sensitivity is the probability of a positive result in a person with the disease (true-positive rate). Specificity is the probability of a negative result in a person without disease (true-negative rate).
Screening tests require high sensitivity so that no case is missed. Confirmatory tests require high specificity to be certain of the diagnosis.
Altering a test cutoff has a reciprocal effect on sensitivity and specificity. A cutoff can be lowered to include all patients with disease (100% sensitivity), but this reduces the specificity (i.e., increases false-positives).
Receiver operator characteristic (ROC) curves plot the true-positive rate (sensitivity) versus the false-positive rate (1-specificity) and graphically present the range of sensitivities and specificities at all test cutoffs. If two tests are compared, the more accurate test is closer to the upper-left-hand corner of the ROC curve.
The likelihood ratio of a test refers to the ratio of the probability of a given test result in the disease state over the probability of the same result in the nondisease state. The likelihood ratio of a test changes as the cutoff value defining disease and nondisease is varied.
Predictive value describes the probability of disease or no disease for a positive or negative result, respectively. The predictive value of a positive test increases with disease prevalence.
The Bayes theorem uses information about test characteristics (sensitivity and specificity) and disease prevalence (pretest probability) to obtain the posttest probability of disease given a positive test. Similarly, it can be used to determine the posttest probability of no disease given a negative test.
Evidence-based medicine is a process by which medical decisions can be made using as many objective tools as possible; it integrates the most current and the best medical evidence with clinical expertise and patient preferences.
Clinical practice guidelines (CPGs) can be used to decrease practice variability, establish monitoring parameters, improve health, increase safety, and decrease cost.
Systematic reviews of clinical effectiveness are transparent, scientifically rigorous, and reliable syntheses of literature that evaluate specific medical interventions and are often used to inform groups developing CPGs.

Every time a clinical laboratory produces a test result, the value must undergo a two-step postanalytic evaluation process. The result needs to be assessed for analytic correctness and for clinical significance. It is often assumed that these two tasks can be easily divided between the performing laboratory, which is responsible for determining analytic correctness, and the clinical team, which is responsible for evaluating the clinical meaning of the results. However, significant overlap is seen in the responsibilities for these tasks. Although the laboratory performs most of the review of laboratory results for analytic reliability by using techniques such as delta checks, flagging of questionable results, moving averages, and linear ranges, it is incumbent on the clinician to review every laboratory result with regard to the patient’s clinical situation and to question the analytic reliability of implausible results. When evaluating the clinical significance of a laboratory result, one of the most important factors to consider is the reference range. In most settings, the laboratory determines reference ranges, with varying degrees of input from the clinical staff. Postanalytic decision making is therefore a shared responsibility of the laboratory and the clinical staff; it behooves both groups to maintain close communication to optimize every part of the process. The purpose of this chapter is to discuss postanalytic review of laboratory data and its use in medical decision making and to provide tools that can be used to objectively interpret results.

Assessment of Analytic Correctness of Results

Alarms and Flags

Modern diagnostic laboratories often analyze large numbers of samples using highly automated instruments. Many results are released into patients’ electronic medical records without prior review by a laboratory employee (e.g., autoverification). To prevent the release of erroneous results, most laboratories utilize a variety of “flags” or alarms. Automated analyzers can flag specimens that require additional or repeat testing before results are released (i.e., reported) by specialized middleware (see Chapter 6 ) or by the laboratory information system. Flags can indicate a problem with the specimen (e.g., the presence of an interfering substance) or an issue with the result (e.g., a numeric value outside the analytic range of the method, or the need for confirmation by an additional assay).

Flags for Problem Specimens

Many automated instruments can measure the amount of sample present in a collection tube and flag samples that contain amounts that are inadequate for a reliable analysis. The laboratory will have to identify another tube containing an adequate sample volume or will request the collection of a new sample. Another frequent cause of inadequate samples is the presence of high concentrations of interfering substances in the specimen, most commonly lipids (lipemia), hemoglobin (Hb; hemolysis), paraproteins (gammopathies), or bilirubin (icterus). The mechanism for this interference is dependent on the substance and the analytic method. For example, in spectrophotometric assays, lipids interfere mainly by increasing light scatter (turbidity); in assays using ion-specific electrodes for measurement, lipids will affect results by solvent exclusion. A more detailed discussion of interference mechanisms is provided in Chapters 3 and 28 . Most commercial assays will list concentrations of interfering substances above which assay results are no longer valid. Visual inspection is often an adequate means of assessing the presence of unacceptable concentrations of interfering substances. A technologist may immediately flag samples that are grossly hemolyzed or icteric as inappropriate for analysis. However, automated analyzers are also able to detect troublesome levels of interfering substances even when they are not apparent to the laboratorian at the macroscopic level. Automated systems can measure the concentrations of bilirubin, lipid, and hemoglobin in samples and can report the degree of interference as an index ( ; ). If the index exceeds a given threshold, then the sample is flagged as problematic and should be rejected or rerun after sample treatment to remove the interfering substance. Serum bilirubin and Hb levels have been shown to correlate very closely with interference indices, but because of the chemical heterogeneity of serum lipids, lipemia indices do not correlate as well ( ).

Flags for Specimens That Require Additional Analysis With Another Method

Some laboratory technologies are screening methods used to analyze large numbers of samples, rapidly report results on most samples, and identify potentially abnormal samples that require retesting with a more labor-intensive method. Automated cell counters are the paradigm of such instruments. These instruments can often analyze more than 100 samples per hour in a highly automated fashion; samples that are normal or that show only quantitative abnormalities (e.g., increased or decreased percentage of lymphocytes, low platelet counts, low red cell counts) can be reported immediately, and samples that could potentially contain qualitative abnormalities (e.g., atypical lymphocytes, platelet clumps, red cell fragments) are flagged for preparation of a blood smear and further evaluation. The flags are generally based on forward- and side-scatter and impedance measurements that provide information about size and nuclear complexity/granularity of the cells, and on special stains that help identify the potential presence of immature cells ( ). The sensitivities and specificities of these flags show poor discriminatory power; clinical judgment is needed if suspicion of an underlying hematologic abnormality is high ( ; ).

Flags for Problematic Results

An analyte concentration outside the validated linear range is another common problem affecting samples. Generally, package inserts of commercial assays will provide end users with an estimated range within which an increase in signal is linearly related to an increase in the analyte concentration. The laboratory may validate this range or may establish its own acceptable linear range when the assay is introduced. Analyzers, middleware, or the laboratory information system will identify and flag samples in which the measured analyte values fall outside the linear range. If the analyte falls above the linear range, many instruments can automatically dilute and reanalyze the sample. In some cases, a manual dilution may be necessary, or the information that the result is higher than a certain value may be sufficient for the requesting clinician. For example, patients in diabetic ketoacidosis will have glucose measurements over 1000 mg/dL, far exceeding the linearity of most analyzers. These samples will be flagged, diluted by a predetermined factor, and then rerun before reporting. If an analyte concentration falls below the linear range, the sample is usually reported as “less than the limit of detection.”

Delta Checks

Advances in computer technology have facilitated the storage of data from large numbers of patients and increasingly complex calculations in laboratory information systems. This has made it possible to use patient data for quality control purposes in real time. For example, most laboratories routinely submit the results of certain laboratory assays to “delta checks” before releasing them into the patient record. Delta checks are defined as comparing a current laboratory result with results obtained on a previous specimen from the same patient. Parameters chosen for delta checks should not be subject to large intraindividual variations; for example, many laboratories have delta checks in place for the mean corpuscular volume of red cells. Suggested assays, thresholds, and time intervals between measurements can be found in the literature ( ). Some studies have suggested the comparison of multiple test parameters to decrease the false-positive rate of the delta check; however, few laboratories have implemented such delta checks. Delta checks can detect preanalytic (e.g., mislabeling of specimens) and analytic issues (e.g., aspiration of insufficient sample volume by the instrument sample probe), though the latter are much less common ( ). Laboratories should define procedures for samples that have been flagged by delta checks; protocols usually incorporate repeating the assay, reviewing the specimen identification, and notifying the clinical staff of the possibility of a mislabeled specimen. Refer to Chapter 11 for a more comprehensive discussion of this topic.

Assessment of Clinical Significance of Results

Critical Values

A critical value (also known as a panic value ) is a laboratory result that may represent a life-threatening situation that may not otherwise be readily detectable. It must be reported immediately to a health care provider who can provide necessary medical interventions. Federal law, regulatory agencies, and The Joint Commission require rapid communication of such results; it is one of the most common, recurring Joint Commission Laboratory National Patient Safety Goals ( ). To be sure that results are correctly communicated, regulations require the health care provider to read back the critical value and the patient name. The laboratory then has to document the event, including the name and title of the caregiver who is notified, the time and date of notification, and the read-back by the care provider.

No universally accepted guidelines indicate which assays should have critical values, what the thresholds should be, whether critical values should be repeated before reporting, and what is an acceptable time from result availability to caregiver notification. Although it is generally established that critical values must be called in to a caregiver who has the ability to act on the information, there is no universal agreement regarding the types of caregivers (e.g., physician assistant, registered nurse) who fulfill this definition. This has caused significant variation in procedures related to critical values at different institutions. It is ultimately the responsibility of the medical director of the laboratory to work with clinical colleagues to develop a critical values policy that meets the needs of patients and staff served by the laboratory.

Reference Ranges

Definition of Reference Intervals

Comparison of a laboratory result versus a reference or “normal” range is often one of the most important aspects of medical decision making. Reference intervals are usually defined as the range of values into which 95% of nondiseased (“normal”) individuals will fall; the corollary of this definition is that 2.5% of nondiseased individuals will have laboratory results below the reference range, and 2.5% of nondiseased individuals will have laboratory results above the reference range ( ). For some analytes, the reference range is defined as “less than” or “greater than” a certain value. For example, a prostate-specific antigen (PSA) level of 4 ng/mL is often used to distinguish patients who require no further follow-up (“normal”) from those who require a prostate biopsy (“abnormal”). Some reference ranges have been defined by professional organizations without adherence to the 95% rule. A paradigm of this is the recommendation of American and European cardiology associations: “An increased cardiac troponin concentration is defined as a value exceeding the 99th percentile of a normal reference population” (upper reference limit [URL]) ( ). For other analytes (e.g., cholesterol/lipids), laboratories frequently provide therapeutic target ranges that represent recommendations based on clinical trials and/or epidemiologic studies ( ). Finally, it is common practice to provide therapeutic and/or toxic ranges for drug measurements.

In the future, the efforts to standardize methods should minimize the effects of site-to-site differences in the implementation of methods. With such standardization, each laboratory would not need to establish its own reference intervals unless its patient population was clearly distinct and exhibited a unique range of values—for example, the protein differences seen in various Asian populations ( ). The Joint Committee for Traceability in Laboratory Medicine has established a process for standardization by identifying, reviewing against agreed criteria, and publishing list(s) of Higher Order Certified Reference Materials and Reference Measurement Procedures. One effort made by all U.S. manufacturers of in vitro diagnostics is to reduce interlaboratory variability by calibrating and using materials traceable to the isotope dilution mass spectrometry (IDMS) reference measurement procedure ( ). This allows for comparability of estimated glomerular filtration (eGFR) and calculated creatinine clearance rates across laboratories ( ).

You're Reading a Preview

Become a Clinical Tree membership for Full access and enjoy Unlimited articles

Become membership

If you are a member. Log in here