Analysing Worked Examples in Laboratory Medicine

Hello! This lesson examines how a research problem becomes an aim and assessable objectives in five areas of clinical laboratory science. We will use published work to identify realistic laboratory questions, then analyse worked examples. The point is not to master every testing method; it is to recognise whether a proposed study can actually answer the problem it names.

Allow about 40 minutes for the reading, short video segments, and lesson.

A way to analyse any example

A laboratory topic such as blood typing or bacterial identification is not yet a research problem. A problem identifies something consequential that remains uncertain in a defined setting. An aim states the study’s overall purpose. Objectives specify the distinct things investigators will measure or compare to fulfil that aim.

Watch these two short portions of Grad Coach’s Research Aims, Objectives & Research Questions (The Golden Thread). They provide a useful test for the examples that follow: do the proposed measurements address the stated purpose, without quietly changing the population or question?

For each case, look for four links: the local uncertainty, the overall aim, the measurements named in the objectives, and the conclusion those measurements would permit. An objective can be precisely worded yet still be wrong for the study if it measures something other than the stated problem.

Microbiology and hematology: choosing the right outcome

The following review surveys applications across laboratory specialties. Read its short specialty subsections for the tasks researchers investigate—such as interpreting susceptibility-test images or recognising blood cells. The worked problem statements and objectives below are teaching examples constructed from those tasks, not objectives quoted from the review.

Worked example 1: Clinical microbiology

Situation and problem. A hospital is considering automated reading of antibiotic susceptibility-test images for E. coli isolated from urine cultures. The consequential uncertainty is not whether automation exists. It is whether, in this laboratory, its interpretations sufficiently agree with the laboratory’s established reading procedure. A disagreement that reports an isolate as susceptible when the reference reading calls it resistant deserves particular attention.

Illustrative aim: To evaluate automated interpretation of antibiotic susceptibility-test images for E. coli urine isolates at Hospital A.

Illustrative objectives:

1. Determine the proportion of eligible susceptibility-test images that the system can interpret from E. coli urine isolates processed at Hospital A during a defined six-month period.
2. Measure agreement between automated and reference interpretations for a prespecified antibiotic panel, including the frequency of interpretations labelled susceptible by automation but resistant by the reference procedure.
3. Compare the time required to obtain an interpretation using the two reading procedures under the study conditions.

The objectives divide “evaluate” into coverage, interpretation quality, and time. Each has a specimen, comparator, and observable result. The important boundary is that disagreement with a reference procedure is not automatically proof of clinical harm, and a shorter image-reading time is not automatically a shorter time to treatment. Those broader claims would require additional evidence.

Worked example 2: Diagnostic hematology

Situation and problem. Suppose a laboratory wants to use an automated flag to prioritise peripheral blood smears for manual review when blast cells may be present. Its uncertainty is how often the flag misses smears with blasts and how many smears it flags unnecessarily. The review’s examples of blood-cell recognition make this a plausible task; the proposed study below is hypothetical.

Illustrative aim: To evaluate a blast-review flag as a way to prioritise manual examination of adult peripheral blood smears at Hospital B.

Illustrative objectives:

1. Estimate the flag’s sensitivity for detecting smears with blasts and its specificity for smears without blasts, using a prespecified expert microscopy assessment as the reference.
2. Determine the proportion of eligible smears that would require manual review under a defined flag-based rule.
3. Count smears with blasts that the rule would not select for review.

These objectives are linked but not interchangeable. A rule could reduce the number of smears reviewed while missing an unacceptable number of important cases. Conversely, a sensitive rule might flag many smears and save little work. The sample must include flagged and unflagged smears: studying only flagged smears would not reveal what the flag missed. Finally, the study evaluates smear triage, not a stand-alone diagnosis of leukemia.

Both examples show why the intended decision matters. In microbiology, a concerning outcome is a misleading susceptibility interpretation. In hematology, it is a smear with blasts that a proposed review rule would overlook. “Accuracy” alone would not tell either laboratory everything it needs to decide.

Clinical chemistry: analysing an actual study

A published evaluation of an integrated chemistry and immunoassay analyser provides a contrast to our constructed cases. The authors explicitly state an aim: to evaluate the instrument’s performance in a clinical laboratory. Their methods make that aim assessable through precision measurements and comparison with another analyser.

Problem-to-aim link. A manufacturer’s evaluation does not remove the need for a laboratory to verify performance before reporting patient results. That is the practical concern described in the article. The authors’ aim is broad enough to cover multiple assays, but their study is bounded by a particular analyser, laboratory setting, period, and selection of chemistry and immunochemistry assays.

The following are our restatements of assessable objectives, not a numbered objective list quoted from the paper:

1. Estimate repeatability and within-laboratory precision for the selected assays, using repeated measurements and reporting variation for each analyte.
2. Compare results from the new analyser with results from the existing analyser using remnant patient serum specimens.
3. Judge measured performance against stated analytical specifications, rather than relying only on whether two sets of results are strongly correlated.

Why include the third objective? Two instruments can give results that rise and fall together yet differ systematically in their reported values. Correlation addresses how closely values track one another; it does not, by itself, establish agreement adequate for a clinical decision. The article reports comparisons relevant to this distinction.

The findings illustrate a disciplined interpretation. The authors report strong correlations for the compared assays, yet albumin and calcium on the new analyser did not meet the minimum biological-variation-based precision specifications applied in the study. They also report differences between methods for some assays. Thus, “the analysers correlate well” is narrower than “their results can always be used interchangeably.”

Notice a second alignment issue: the introduction discusses the potential for automated instruments to improve turnaround time, but the precision and method-comparison results we examined do not demonstrate a turnaround-time improvement. A motivating benefit should not be reported as an achieved finding unless the study measured it. The published evaluation covers both chemistry and immunochemistry; our next example uses a different question within diagnostic immunology.

Immunology and transfusion science: defining the reference and the limits

Worked example 4: Diagnostic immunology

The review describes work on classifying patterns in HEp-2 cell immunofluorescence images. These patterns are used in autoimmune-disease testing. This is not the same question as measuring serum hormone concentrations with an immunoassay, even though both fall within laboratory immunology-related practice.

Situation and problem. A laboratory considering automated pattern labels does not know how closely they match its trained readers’ interpretations, particularly for less common patterns.

Illustrative aim: To assess the agreement of automated HEp-2 immunofluorescence pattern labels with expert interpretations for positive slides processed at Laboratory C.

Illustrative objectives:

1. Measure overall agreement between automated labels and a prespecified expert interpretation for eligible positive slides collected during a defined study period.
2. Describe disagreement separately for each pattern category with enough observations to permit a meaningful estimate.
3. Determine the proportion of slides that the system marks as indeterminate and would therefore require expert review under a prespecified rule.

Here, the reference interpretation needs careful definition. For example, two trained readers might interpret slides independently, with disagreements adjudicated before their combined judgement is used as the reference. Agreement with readers assesses pattern classification; it does not establish that the system diagnoses an autoimmune disease. If a pattern is rare, a high overall agreement could also conceal poor performance on that particular category. The second objective prevents that important distinction from disappearing into a single summary number.

Worked example 5: Transfusion science

The review also identifies ABO typing as a laboratory task. In transfusion science, a poorly resolved blood group is a safety concern, so an evaluation must specify how discrepancies are handled—not merely how many straightforward samples a system labels correctly.

Situation and problem. Hospital D is considering whether an automated ABO testing workflow can support routine reporting. Its uncertainty concerns agreement with independently resolved blood groups and the frequency of cases that still need manual investigation.

Illustrative aim: To evaluate automated ABO typing for routine patient samples at Hospital D while identifying results that require manual resolution.

Illustrative objectives:

1. Determine agreement between automated ABO results and an independently established reference group for consecutive eligible patient samples collected during a defined period.
2. Record the number and types of samples with discrepant or uninterpretable forward- and reverse-typing results that require manual resolution.
3. Estimate the proportion of samples that would be withheld from automatic reporting under prespecified safety rules.

These objectives evaluate a proposed reporting workflow, not permission to release every automated result. A study containing mostly straightforward samples could have excellent overall agreement but too few rare discrepancies to evaluate them confidently. Investigators would need to acknowledge that limitation or design additional, separately analysed testing of challenging samples. Neither approach should disguise how uncommon problems differ from routine-case performance.

What the five examples reveal

Across the specialties, a well-aligned study does more than attach measurable verbs to a topic. It chooses measurements that match the decision behind the problem:

The pattern you can carry into a written report—النص المكتوب, “the written text”—is a short, coherent sequence: state the setting and consequential uncertainty; give one overall aim; then write objectives whose population, specimen, comparator, and measurements stay consistent with that aim. Label a constructed example as illustrative, and distinguish what a published study measured from benefits it merely anticipated.

The next step is to choose one laboratory topic of your own and develop a coherent problem, aim, and set of objectives for it. These five analyses provide models for that task, but also a warning: an impressive result is useful only if it answers the question the study set out to investigate.