U.K. Primary-Care Eczema Algorithm: Diagnosis Plus Treatment Codes in 200 Records

Evidence summary: Diagnostic-accuracy research asks a different question from treatment research: not whether a therapy works, but whether a rule identifies the intended patients without systematically missing others. This review examines 200 randomly selected children and adults from a primary-care EHR database and keeps diagnostic accuracy separate from treatment claims and causation.

Electronic records are not clinical examinations

Electronic records introduce a different problem. A U.K. primary-care algorithm combined diagnosis and treatment codes [8]. Swedish administrative algorithms performed differently in children and adults [9], Canadian researchers found large prevalence differences across 23 definitions [10], and the Born in Bradford cohort emphasized local code availability [11]. Record-based definitions can scale to hundreds of thousands of people, but scale does not remove misclassification.

Codes are created during care, not for research. Missing codes can mean no disease, no visit, no access, no documentation or care outside the captured system. Repeated treatment codes can improve confidence but also select for people who received prescriptions. The best definition therefore depends on whether the goal is surveillance, comparative safety, prognosis or recruitment.

Why validation across countries and ages matters

Geographic validation also matters. A multicenter Chinese study compared criteria in adults and older adults [12]; an updated Korean rule was tested against earlier Korean and Japanese criteria [13]; an Iranian study reported high specificity but very low sensitivity for the tested rule [14]; and Ethiopian research found different estimates when ISAAC questions and U.K. criteria were applied [15]. Telephone administration of ISAAC questions was separately studied in schoolchildren [16], while a population-based parent questionnaire addressed children through age two [17].

Differences across studies should not be reduced to stereotypes. They may reflect age distribution, severity, language, climate, referral patterns, clinician training, competing diagnoses and how the reference standard was applied. A locally adapted rule should be validated again rather than assumed equivalent after translation.

Association, classification and causation

A validation study estimates classification performance. It does not show that a diagnostic feature causes eczema, that a negative result proves absence, or that a positive result predicts a particular treatment response. When later cohort studies associate eczema with another outcome, misclassification at the case-definition step can dilute, exaggerate or even reverse that association. Transparent definitions are therefore part of causal caution.

What this changes in practice and research

For researchers, the practical consequence is to prespecify the case definition, time window, setting and reference standard; report the exact numerator and denominator behind each accuracy estimate; and run sensitivity analyses with stricter and broader definitions. For clinicians and families, a survey or database label should be treated as a prompt for appropriate assessment, not as a substitute for examining the skin and considering lookalikes. A person can have genuine eczema even if one research rule is negative, and a positive questionnaire does not rule out contact dermatitis, scabies, psoriasis, infection or another diagnosis.

A high specificity may be useful when false positives are costly, whereas a high sensitivity may be preferable for initial screening. Neither choice is automatically “better.” Researchers should explain the consequence of each error type and whether the chosen threshold was decided before seeing outcomes.

Important limitations

Several recurring limitations apply to this evidence. Referral samples can overrepresent severe disease. Small validation sets produce imprecise confidence intervals. Observer training and knowledge of prior diagnoses can influence the reference assessment. Translation can alter meaning. Treated disease may be invisible on the study day. Prescription codes may reflect access and clinician preference rather than biology. Race and ethnicity are social and demographic descriptors, not biological diagnostic shortcuts; validation across diverse populations is necessary precisely because morphology, access, language and sampling differ.

No included study proves that one universal criterion set is optimal for every infant, child, adult and older adult. External validation, calibration and periodic updating remain necessary as documentation systems and clinical populations change.

Safety boundaries for families and adults

Seek medical assessment when a baby has a new widespread rash, poor feeding, fever, blistering or rapidly worsening skin; when any person develops painful clustered blisters, eye involvement, pus, spreading redness or systemic illness; or when a rash is persistent, atypical or not responding as expected. Do not use an online checklist to start or stop prescription corticosteroids, calcineurin inhibitors, PDE4 inhibitors, JAK inhibitors, antibiotics, antivirals or systemic medicines. Infant, pregnancy, breastfeeding, older-adult and immune-compromised care require individualized advice.

Where #003 fits—and does not fit

Moisturizing is supportive skin care, not a diagnostic test. Dr. Arsy Moisturizer #003 may be considered as a fragrance-free moisturizing option for dry, sensitive or eczema-prone skin, but it was not the study product in the research reviewed here. #003 does not diagnose, treat, cure or prevent eczema, and no diagnostic-accuracy finding can be transferred to it. Product use should not delay evaluation of infection, allergy, a medication reaction or a changing rash.

Related Dr. Arsy research reviews

• How eczema is diagnosed and why there is no single blood test
• Tape strips, optical imaging and AI for eczema diagnosis
• U.K. Working Party hospital validations
• Diagnostic criteria across 27 validation studies
• U.K. criteria in 11,340 adults
• Infant facial eczema diagnosis in 5,967 babies
• Home skin-barrier measurement validation
• Understanding the POEM symptom score

Why this specific validation question matters

The central issue in U.K. Primary-Care Eczema Algorithm: Diagnosis Plus Treatment Codes in 200 Records is transportability. A rule can be statistically persuasive in the sample that produced it and still misclassify people elsewhere. Clinical presentation changes with age, treatment, disease activity, body site and the mix of competing diagnoses. Survey wording changes what families remember; database structure changes what researchers can see. For that reason, this article treats the reported estimates as properties of a defined study process, not permanent properties of eczema itself. The responsible next step is replication in an independent population, with blinded reference assessment, prespecified thresholds and results stratified by age, severity and relevant care setting. That approach makes the evidence more useful without pretending that a checklist, questionnaire or code set can replace individualized clinical judgment.

Sources

1. The U.K. Working Party diagnostic criteria: derivation of minimum discriminators.
2. Population validation of the U.K. diagnostic criteria.
3. Strengths and limitations of U.K. criteria in adults.
4. Danish questionnaire development and validation.
5. Atopic Dermatitis Screening and Evaluation Questionnaire pilot.
6. Self- and caregiver-report multicenter validation.
7. Validation of recalled childhood eczema in adults.
8. U.K. primary-care EHR algorithm validation.
9. Swedish healthcare-database patient-identification algorithms.
10. Canadian primary-care EMR case-definition validation.
11. Born in Bradford EHR allergic-disease case definitions.
12. Chinese adult and older-adult diagnostic-criteria validation.
13. Updated Korean diagnostic-criteria clinical validation.
14. Iranian diagnostic-criteria validation.
15. ISAAC and U.K. criteria validation in Ethiopian children.
16. ISAAC telephone questionnaire validation in children.
17. Parental questionnaire validation in 476 children through age two.
18. American Academy of Dermatology atopic dermatitis clinical guidelines.
19. American Academy of Dermatology: diagnosis and treatment overview.
20. NICE guideline CG57: atopic eczema in under-12s.
21. NHS: atopic eczema.

What this research question means

This article focuses on 200 randomly selected children and adults from a primary-care EHR database. The design was a physician-questionnaire validation of diagnosis and prescription code combinations. The central finding was that a diagnosis code plus repeated skin-directed treatment codes achieved a strong positive predictive value. The most defensible interpretation is that stricter definitions improved confidence among identified cases but excluded many true cases. That interpretation is deliberately narrower than saying the instrument “diagnoses eczema” in every person. It describes performance under a specific protocol, population and reference standard.

Time windows matter. Point prevalence requires active disease now; one-year prevalence permits recent but inactive disease; lifetime prevalence depends heavily on recall. A rule optimized for current flexural dermatitis may undercount a person whose eczema is controlled on the examination day. Conversely, a broad lifetime question may include remote or misremembered rashes.

How diagnostic validation should be read

Age and phenotype matter. Infant eczema may involve cheeks and extensor surfaces before classic flexural patterns appear. Adults can have hand, head-and-neck, nummular or later-onset phenotypes. Older adults may have xerosis, drug eruptions, scabies or cutaneous lymphoma in the differential. A single historical criterion can therefore perform differently across the life course.

The original U.K. Working Party program reduced 31 candidate observations to a compact set of discriminators in a blinded derivation exercise [1], then tested the rule in a population setting [2]. Later adult work showed that performance depended strongly on current activity and comparison with psoriasis [3]. A Danish child questionnaire demonstrated how wording, history and a severity component can be adapted for population work [4].

Questionnaires, recall and proxy reports

Other approaches broadened the format. A self-administered screening questionnaire combined symptom and image questions [5]; a U.S. multicenter study compared self and caregiver reports with expert examination [6]; and a record-linked study showed that adults' recall of childhood eczema could overestimate prevalence [7]. These studies do not make one instrument universally superior. They show why the intended age, time window and decision must be specified before selecting a tool.

A diagnostic rule is not a blood test and the reference standard is not perfect. Dermatologist judgment integrates morphology, distribution, chronicity, itch, personal and family history, treatment response and competing diagnoses. Validation studies therefore measure agreement with an operational reference, not access to an infallible truth. This distinction prevents accuracy statistics from being interpreted as biological certainty.

Educational information only; it is not a diagnosis or individualized medical advice.