A Clinical Prediction Model for Atypical TB Manifestations Among Older Adults

Who

• Population: Older adults aged ≥75 years

• Sample size: 5,651 patients with culture-confirmed pulmonary tuberculosis (aPTB) and atypical symptom presentation

• Subgroups:

  • Group a (Ga): 1,155 patients with aPTB not initially suspected by non-pulmonologists

  • Group b (Gb): 4,496 non-TB comparators within the first 24 hours

• Setting of care: Evaluated initially by non-chest physicians

• Radiology review: 2 radiologists + 1 pulmonologist (blinded)

---

What

• Objective: Development and validation of a TRIPOD-compliant clinical prediction score to identify atypical pulmonary TB (aPTB) in late-elderly patients.

• Key findings:

  • Five independent predictors of delayed aPTB diagnosis were identified:

    1. Age >85 years (strongest predictor)

    2. Hypoalbuminemia (<3.5 g/dL)

    3. Cardiovascular disease

    4. Diabetes mellitus

    5. Predominant lower-lung field involvement

  • A score cutoff ≥7 showed excellent diagnostic performance:

    • AUC: 0.95–0.96

    • Sensitivity: 91–94%

    • Specificity: 97–99%

• Clinical implication: The model reliably detects aPTB even in patients without classic TB symptoms (≈70% had symptom scores ≤1).

---

When

• Derivation cohort: 2000–2020

• Temporal validation cohort: 2021–2023

• Study design: Retrospective analysis

---

Where

• Geographic setting: Chia-Yi, southern Taiwan

• Health system context: TB diagnosis following Taiwan’s tiered strategy using AFB smear, selective PCR, and culture confirmation.

---

Why

• Rationale: Atypical TB presentations are common in the late-elderly, leading to diagnostic delays, especially outside pulmonology settings.

• Existing symptom-based tools underperform in this population, and frailty markers (e.g., sarcopenia, osteoporosis) do not adequately capture risk.

• Early identification is critical to reduce missed diagnoses and improve outcomes in this vulnerable group.

---

How

• Design: Retrospective cohort study with derivation and temporal validation

• Inclusion criteria: Age ≥75 years, WHO/CDC symptom score <5, culture-confirmed TB

• Analysis:

  • Univariate screening (p < 0.05)

  • Multivariate logistic regression with stepwise selection

  • Model robustness assessed via:

    • ROC/AUC

    • Calibration plots

    • Decision curve analysis (DCA)

    • Subgroup analyses (age strata, diabetes status)

• Validation: Independent temporal cohort; no significant AUC differences (DeLong test p > 0.70)

• Reliability: Excellent inter-observer agreement for radiographic assessment (Fleiss’ κ = 0.91)

---

Overall conclusion

The study presents a simple, highly accurate, and externally validated clinical scoring tool for early detection of atypical pulmonary TB in adults aged ≥75 years. Extreme old age (>85 years) emerged as the most powerful predictor, surpassing traditional frailty indicators. Integration of this score into electronic medical records could meaningfully reduce diagnostic delays in non-pulmonology settings.

Source: Yeh, J.J., Chen, J.H., Kuo, Y.L., Tsai, C.H. and Ko, Y.E., 2025. A Clinical Prediction Model for Atypical Tuberculosis Manifestations Among Older Adults. Medicina, 61(10), p.1888.