Screening, Diagnostic Testing and Early Detection

Introduction and Context

Screening and diagnostic testing are central to population health, clinical decision-making and health-care management. They are related but not interchangeable activities. Screening applies a test to people who do not have recognized symptoms, with the purpose of identifying those who may have a disease or elevated risk. Diagnostic testing is undertaken when symptoms, signs, history or a screening result create a clinical suspicion of disease. Early detection is the broader objective: identifying disease at a stage when treatment can improve outcomes, reduce complications or prevent transmission.

A test is not inherently “good” or “bad.” Its usefulness depends on the population, disease prevalence, consequences of missed or false diagnoses, available confirmatory tests, treatment effectiveness, costs, and the capacity of the health system to complete follow-up. A highly accurate test can produce little health benefit if positive patients cannot obtain diagnosis or treatment. Conversely, a moderately accurate test may be valuable when disease is serious, treatment is effective and confirmatory services are accessible.

For health-care managers, screening is therefore a programme rather than merely a laboratory procedure. A programme includes policy, eligibility criteria, invitation, testing, communication, referral, diagnosis, treatment, monitoring and quality improvement.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Distinguish screening, diagnostic testing, case-finding and surveillance.
2. Construct and interpret a 2 × 2 diagnostic-testing table.
3. Calculate sensitivity, specificity, predictive values and likelihood ratios.
4. Explain how prevalence and thresholds affect test interpretation.
5. Interpret ROC curves and select practical cut-off points.
6. Evaluate benefits, harms, overdiagnosis, ethics and cost-effectiveness.
7. Assess operational feasibility and quality of a screening programme.
8. Apply these principles to managerial and public-health decisions.

1. Screening, Diagnosis and Early Detection

Screening is appropriate when the target condition is sufficiently important, has a recognizable early or latent phase, and early intervention improves meaningful outcomes. Screening differs from routine clinical assessment because the screened population is generally asymptomatic or not seeking care for the target condition.

Case-finding is a targeted form of detection in which clinicians test people with risk factors or relevant symptoms, even if the disease has not yet been diagnosed. Surveillance is the continuous or periodic collection and interpretation of health data to monitor disease occurrence, trends and control measures. Surveillance may use test results but is not itself a diagnostic service.

A screening pathway usually contains the following stages:

1. Define the eligible population and interval.
2. Invite or offer testing in an equitable manner.
3. Perform a validated initial test.
4. Communicate results clearly.
5. Refer positive or indeterminate results for diagnostic confirmation.
6. Initiate evidence-based treatment or prevention.
7. Track completion and outcomes.
8. Monitor benefits, harms, equity and cost.

A central principle is that a screening test is not a final diagnosis. A positive result usually means “further assessment is warranted,” not “disease is confirmed.”

2. The 2 × 2 Framework

Test performance is evaluated by comparing test results with a reference standard, sometimes called the gold standard. The reference standard may be histopathology, culture, imaging, expert clinical diagnosis or long-term follow-up. It is important to recognize that reference standards can themselves be imperfect.

A true positive is a person correctly identified as having the condition. A false positive is a person without the condition who tests positive. A false negative has the condition but tests negative. A true negative does not have the condition and tests negative.

Sensitivity

Sensitivity is the probability that a person with disease has a positive test:

The denominator is all people who truly have disease. Sensitivity is also called the true-positive rate. A highly sensitive test produces few false negatives, assuming the reference standard and study design are valid.

Specificity

Specificity is the probability that a person without disease has a negative test:

The denominator is all people who are disease-free. Specificity is the true-negative rate. A highly specific test produces few false positives.

Sensitivity and specificity are conditional probabilities based on true disease status. They do not directly tell a patient the probability that their positive or negative result is correct.

Predictive values

The positive predictive value (PPV) is the probability that disease is present when the test is positive:

The negative predictive value (NPV) is the probability that disease is absent when the test is negative:

Predictive values depend strongly on disease prevalence. In a low-prevalence population, even a good test may produce more false positives than true positives. Sensitivity and specificity are usually more transportable across populations, although they can also vary with disease spectrum, test quality and the chosen threshold.

Worked example: prevalence and predictive value

Suppose 10,000 asymptomatic people are screened for a disease with prevalence 1%. The test has sensitivity 90% and specificity 95%.

There are 100 people with disease and 9,900 without disease. Expected results are:

• TP = 90% of 100 = 90
• FN = 10% of 100 = 10
• TN = 95% of 9,900 = 9,405
• FP = 5% of 9,900 = 495

Therefore:

The negative result is highly reassuring, but most positive results are false positives and require confirmation. This is not evidence that the test is useless; it demonstrates why prevalence and diagnostic follow-up matter.

3. Likelihood Ratios and Bayesian Interpretation

Likelihood ratios express how much a test result changes the odds of disease. They are particularly useful for combining a pre-test probability with a test result.

The positive likelihood ratio is:

The negative likelihood ratio is:

A large LR+ provides strong evidence for disease. An LR− close to zero provides strong evidence against disease. Unlike predictive values, likelihood ratios do not directly depend on prevalence, although they may vary across settings.

Bayes’ relationship can be expressed using odds:

Worked example

Assume a patient has a pre-test probability of 20%. A test has sensitivity 90% and specificity 80%.

A positive result raises probability from 20% to approximately 53%, but does not establish the diagnosis. Further testing may be required.

4. Thresholds, Continuous Tests and ROC Curves

Many tests produce continuous results, such as blood pressure, glucose concentration, tumour-marker level or imaging score. A threshold or cut-off classifies results as positive or negative. Changing the threshold changes sensitivity and specificity:

• Lowering the threshold generally increases sensitivity but reduces specificity.
• Raising the threshold generally increases specificity but reduces sensitivity.

The appropriate threshold depends on the consequences of errors. If missing disease is dangerous and treatment is safe, a lower threshold may be reasonable. If false positives cause invasive procedures or major anxiety, a higher threshold may be preferred.

A receiver operating characteristic (ROC) curve plots sensitivity against the false-positive rate at multiple thresholds. The false-positive rate is:

The area under the ROC curve (AUC) summarizes discrimination across thresholds. An AUC of 0.5 indicates discrimination no better than chance; an AUC of 1.0 indicates perfect separation in the study population. AUC is useful for comparing overall discrimination, but it does not select the clinically appropriate threshold and does not measure calibration, patient benefit or cost.

Figure: An illustrative ROC curve. Each point on the curve represents a different decision threshold.

Threshold selection should consider clinical utility, not only statistical performance. Decision-curve analysis and net-benefit methods can help compare testing strategies across plausible risk thresholds, but they require explicit assumptions about the relative harms of false positives and false negatives.

5. Screening Programme Evaluation

Evidence of benefit

The strongest evidence for screening usually comes from randomized trials or well-designed quasi-experimental studies assessing outcomes such as disease-specific mortality, all-cause mortality, serious complications, quality of life and health inequities. Improved five-year survival alone can be misleading because it may reflect earlier diagnosis without extending life—a phenomenon known as lead-time bias.

Length-time bias occurs because screening is more likely to detect slowly progressing disease, which may have a better prognosis even without screening. Overdiagnosis occurs when screening detects disease that would never have caused symptoms or death during the person’s lifetime. Overdiagnosed cases can inflate incidence and survival statistics while exposing people to treatment harms.

Benefits and harms

Potential benefits include reduced mortality, prevention of complications, earlier less-invasive treatment, reduced transmission and informed risk management. Potential harms include false positives, false reassurance, anxiety, unnecessary biopsies or procedures, radiation exposure, treatment complications, labelling, opportunity costs and inequitable access.

A programme should report absolute effects, not only relative effects. For example, a 20% relative reduction may represent a reduction from 10 deaths per 1,000 people to 8 per 1,000—an absolute reduction of 2 per 1,000. Both measures are informative, but they answer different questions.

Programme indicators

Managers should monitor:

• participation and coverage;
• positivity rate;
• referral and diagnostic completion;
• time from positive test to diagnosis and treatment;
• cancer or disease detection rate;
• interval disease after a negative screen;
• false-positive and false-negative rates;
• complications and patient-reported outcomes;
• stage distribution where relevant;
• equity by geography, income, sex, ethnicity, disability and other protected characteristics;
• cost per person screened and cost per health outcome gained.

High participation is not sufficient. A programme with excellent attendance but poor follow-up may generate harm without benefit.

6. Cost-Effectiveness and Resource Allocation

Screening consumes staff time, laboratory capacity, equipment, information systems and follow-up resources. Cost-effectiveness analysis compares incremental costs with incremental health outcomes between a screening strategy and an alternative.

The incremental cost-effectiveness ratio is:

where strategy 1 is the new programme, strategy 0 is the comparator, and effects may be measured in life-years gained or quality-adjusted life-years (QALYs). Costs and outcomes should be measured from a clearly specified perspective and discounted over time when appropriate. The ICER is meaningful only when the comparator, time horizon, outcome measure and assumptions are transparent.

Economic models should include test costs, invitation, staff, confirmatory investigations, treatment, adverse events, patient time, transport and downstream consequences. Sensitivity analysis should examine uncertainty in prevalence, participation, test accuracy, treatment effectiveness and overdiagnosis.

Cost-effectiveness is not the same as affordability. A programme may be cost-effective but unaffordable within the current budget. Conversely, a programme may be affordable but provide poor value relative to alternatives.

7. Ethics, Governance and Communication

Screening raises ethical questions because it recruits people who may feel well into a pathway that can produce harm. Ethical practice requires respect for autonomy, informed choice, proportionality, confidentiality and justice.

Information should explain the purpose of screening, the possibility of false positives and negatives, overdiagnosis, downstream procedures, alternatives and the option not to participate. Communication should use absolute risks where possible and avoid implying that screening guarantees prevention or cure.

Governance requirements include validated assays, documented standard operating procedures, staff competency, equipment calibration, external quality assessment, secure records, audit and adverse-event reporting. Personal health data must be protected, with appropriate access controls and lawful data use. Algorithms used in risk assessment require monitoring for bias, drift, explainability and unequal performance across population groups.

Equity must be designed into the programme. Digital invitations alone can exclude people with limited connectivity or literacy. Accessible communication, language support, mobile services, transport assistance and community partnerships may be necessary. Managers should distinguish equal provision from equitable provision: people with greater barriers may require additional support to achieve comparable opportunity.

8. Technology and Artificial Intelligence

Point-of-care testing can shorten turnaround time and extend services into primary care or remote settings, but its accuracy must be assessed under real operating conditions rather than only in specialized laboratories. Digital imaging and AI-assisted interpretation may improve consistency or workload management. However, performance can deteriorate when devices, populations, disease prevalence or clinical workflows differ from the development setting.

AI should be evaluated for discrimination, calibration, clinical utility, subgroup performance, cybersecurity, automation bias and effects on workload. A model with high AUC may still be poorly calibrated, meaning its predicted risks do not correspond to observed risks. Implementation should include human oversight, clear responsibility for errors, prospective monitoring and a process for model updating or withdrawal.

9. Illustrative Case: Community Colorectal Cancer Screening

Background

A regional health authority considers introducing a biennial stool-based screening programme for adults aged 50–74 years. The region has rising colorectal cancer incidence, limited colonoscopy capacity and substantial geographic inequity.

Problem

Leaders initially focus on the analytical sensitivity of the stool test. However, the programme may create a large number of positive results requiring colonoscopy. Participation is expected to be lower in rural communities and among people with limited literacy.

Analysis

The authority models prevalence, test performance, participation, colonoscopy capacity and treatment availability. It finds that a positive screening result is not diagnostic. It therefore estimates the number of positive participants, diagnostic completion, procedure complications and cancers detected. It also examines whether delayed colonoscopy could increase anxiety and reduce benefit.

A pilot identifies three operational risks: invitation letters are poorly understood, colonoscopy appointments are concentrated in the capital, and positive results are not consistently tracked. Equity analysis shows lower participation in two deprived districts.

Approach and solution

The programme adopts plain-language and multilingual invitations, telephone and community-based support, reminder systems, a central results registry and reserved diagnostic capacity. A positive result triggers navigator contact and a defined referral interval. Performance dashboards report participation, positivity, diagnostic completion, time to colonoscopy, cancer stage and outcomes by district and demographic group.

The authority explains that screening may detect cancers earlier but can also lead to false positives, invasive follow-up and detection of lesions that might not progress. Participation remains voluntary.

Outcome

After the first year, participation increases after community outreach, but diagnostic completion remains below target in remote areas. Management therefore reallocates mobile endoscopy capacity rather than simply expanding invitations. The programme is judged on completed care pathways and health outcomes, not on the number of tests distributed.

Learning points

1. Test accuracy does not guarantee programme effectiveness.
2. Follow-up capacity is a clinical safety requirement.
3. Equity and logistics should be assessed before scale-up.
4. Positive screening results require diagnostic confirmation.
5. Monitoring must include harms and patient experience as well as detection.

10. Best-Practice Decision Framework

Before approving a screening programme, managers should ask:

1. Is the disease an important health problem?
2. Is there a detectable phase during which intervention improves outcomes?
3. Is the test sufficiently accurate, acceptable and safe?
4. What are the expected absolute benefits and harms?
5. Is confirmatory diagnosis available within a safe interval?
6. Are effective treatments accessible and affordable?
7. Can the programme reach underserved groups?
8. Are workforce, laboratory, referral and data capacities adequate?
9. Is the programme cost-effective and financially sustainable?
10. How will overdiagnosis, adverse events and unintended consequences be measured?

This framework encourages a whole-system assessment rather than a narrow comparison of sensitivity and specificity.

Challenges and Limitations

Diagnostic accuracy studies may suffer from spectrum bias if participants are much sicker or healthier than those encountered in practice. Verification bias occurs when not all participants receive the reference standard. Incorporation bias occurs when the test being evaluated contributes to the reference diagnosis. Review bias can arise when assessors know the clinical or reference-standard result. Prevalence estimates may not generalize across settings.

A test may perform differently when used by community workers, primary-care clinicians or laboratories with limited resources. Transport, storage, specimen quality and turnaround time can alter real-world performance. Furthermore, a statistically significant improvement may not be important to patients. Screening policies should therefore combine epidemiological evidence, clinical outcomes, economic analysis, ethics and implementation evidence.

Chapter Summary

Screening identifies possible disease in people without recognized symptoms; diagnostic testing evaluates suspected disease. Test performance is summarized using sensitivity, specificity, predictive values and likelihood ratios. Sensitivity and specificity describe results conditional on disease status, whereas predictive values depend strongly on prevalence. Likelihood ratios update pre-test probability and support Bayesian interpretation.

Continuous tests require threshold selection. Lower thresholds generally increase sensitivity and reduce specificity, while higher thresholds do the opposite. ROC curves describe discrimination across thresholds, but clinical decisions must also consider consequences, patient preferences, capacity and cost.

Effective screening programmes require more than accurate tests. They require evidence that early detection improves outcomes, accessible confirmatory diagnosis, effective treatment, quality assurance, equitable participation, ethical communication and continuous monitoring. Managers must account for false results, overdiagnosis, lead-time and length-time bias, opportunity costs and operational feasibility. Technology and AI can assist screening, but only with validation, governance, equity monitoring and human accountability.

Key Terms

Sensitivity: Probability that a person with disease tests positive.

Specificity: Probability that a person without disease tests negative.

Positive predictive value: Probability of disease among those with positive test results.

Negative predictive value: Probability of no disease among those with negative test results.

True positive: Positive test in a person who has the disease.

False positive: Positive test in a person who does not have the disease.

True negative: Negative test in a person who does not have the disease.

False negative: Negative test in a person who has the disease.

Likelihood ratio: Measure of how a test result changes the odds of disease.

Pre-test probability: Estimated probability of disease before the test result is known.

Threshold: Cut-off used to classify a continuous test result.

ROC curve: Plot of sensitivity against false-positive rate across thresholds.

AUC: Area under the ROC curve, summarizing discrimination across thresholds.

Overdiagnosis: Detection of disease that would not have caused harm during the person’s lifetime.

Lead-time bias: Apparent survival improvement caused by earlier diagnosis without delayed death.

Length-time bias: Preferential detection of slower-progressing disease through screening.

Case-finding: Targeted testing of people with symptoms or risk factors.

Screening programme: Organized pathway from invitation and testing through diagnosis, treatment and evaluation.

ICER: Incremental cost-effectiveness ratio comparing additional cost with additional health effect.

Calibration: Agreement between predicted and observed risks.

Review Questions

Short Answer Questions

1. Distinguish screening from diagnostic testing.
2. Define sensitivity and specificity.
3. Why does disease prevalence affect positive predictive value?
4. What does an ROC curve represent?
5. Give two possible harms of screening.

Long Answer Questions

1. Explain the 2 × 2 diagnostic-testing table and derive sensitivity, specificity, PPV and NPV.
2. Discuss Bayesian interpretation using pre-test probability and likelihood ratios.
3. Explain how threshold selection affects sensitivity, specificity and clinical decision-making.
4. Critically evaluate overdiagnosis, lead-time bias and length-time bias.
5. Describe the managerial requirements for implementing an equitable screening programme.

Analytical / Application Questions

1. A screening test is applied to 20,000 people. Disease prevalence is 2%, sensitivity is 85% and specificity is 96%. Calculate TP, FN, TN, FP and PPV, and interpret the result for programme managers.
2. A hospital proposes lowering a test threshold to reduce missed cases. Analyse the likely effects on referrals, confirmatory procedures, costs and patient harms.
3. Design an evaluation dashboard for a regional screening programme. Specify at least six indicators and explain how each would support quality, safety or equity improvement.

Multiple Choice Questions with Answers

Question 1

Which measure is calculated using all people who truly have the disease as its denominator?

A. Sensitivity

B. Specificity

C. Positive predictive value

D. Negative predictive value

Correct Answer: A. Sensitivity

Question 2

In a low-prevalence population, a common consequence of a positive screening result is that it may have:

A. A lower positive predictive value

B. A higher specificity automatically

C. A lower false-positive rate automatically

D. A higher prevalence among the screened population

Correct Answer: A. A lower positive predictive value

Question 3

A test has sensitivity 0.90 and specificity 0.80. What is its positive likelihood ratio?

A. 0.10

B. 0.20

C. 4.5

D. 5.0

Correct Answer: C. 4.5

Question 4

Lowering the threshold for a continuous screening test generally causes which change?

A. Sensitivity decreases and specificity increases

B. Sensitivity increases and specificity decreases

C. Both sensitivity and specificity increase

D. Neither sensitivity nor specificity changes

Correct Answer: B. Sensitivity increases and specificity decreases

Question 5

The false-positive rate is equal to:

A. 1 − sensitivity

B. 1 − specificity

C. Specificity ÷ sensitivity

D. FP ÷ FN

Correct Answer: B. 1 − specificity

Question 6

Which statement about a positive screening result is most accurate?

A. It usually confirms the disease

B. It means no further assessment is needed

C. It generally indicates that diagnostic confirmation is warranted

D. It proves that treatment will improve outcome

Correct Answer: C. It generally indicates that diagnostic confirmation is warranted

Question 7

A programme detects disease earlier but does not postpone death. Which bias can make survival appear improved?

A. Recall bias

B. Lead-time bias

C. Selection concealment

D. Observer expectancy only

Correct Answer: B. Lead-time bias

Question 8

Overdiagnosis refers to:

A. Failure to detect a rapidly fatal disease

B. Detection of disease that would never have caused harm

C. A test result that is technically invalid

D. A diagnosis made without a laboratory test

Correct Answer: B. Detection of disease that would never have caused harm

Question 9

Which is essential for a safe screening programme?

A. High participation regardless of follow-up capacity

B. Diagnostic and treatment pathways for people with positive results

C. Use of only one test without confirmation

D. Reporting relative effects without absolute risks

Correct Answer: B. Diagnostic and treatment pathways for people with positive results

Question 10

The ICER compares:

A. Total prevalence with total incidence

B. Sensitivity with specificity

C. Incremental cost with incremental health effect

D. Positive predictive value with negative predictive value

Correct Answer: C. Incremental cost with incremental health effect