Validation of Chest Radiograph Reading and Recording System (CRRS) in Three Sites in Indonesia

One of the main challenges in Tuberculosis (TB) management is the lack of tools for early and accurate diagnosis. Chest X-ray remains one of the most convenient ways for TB diagnosis. Even though chest X-ray result can be obtained quickly, it requires accurate interpretation from radiologist. Chest Radiograph Reading and Recording System (CRRS) can be used to standardize the interpretation process for TB diagnosis. This study aims to determine the validity of CRRS in hospital and primary health center. This is a cross-sectional observational study in Cipto Mangunkusumo Hospital, Jatinegara, and Matraman Healthcare Centers. The study was conducted from May to November 2015. Participants were recruited with consecutive sampling. The eligibility criteria were age ≥ 15 years old, new or previously treated patients with current TB treatment ≤ 1 months, and having symptoms consistent with TB, or having a contact history with patients with smear positive or having extra-pulmonary TB. Recruited patients underwent anamnesis, physical examination, morning sputum examination, and chest X-ray examination. Acid-Fast bacilli smears and culture of the morningsputum was conducted in Microbiology Laboratory of Faculty of Medicine, Universitas Indonesia. Chest X-ray examination was conducted in Cipto Mangunkusumo Hospital. The radiologists filled the CRRS form that contains the scoring based on four features: Score 2 is given for Upper lobe large opacity (> 1 cm), cavity in any location, and adenopathy in any location, and score 1 for unilateral pleural effusion. The totalnumber of the score was then categorized using different cut-offs (≥ 1, ≥ 2, ≥ 3, and ≥ 4) to evaluate the validity of the method compared with the sputum examination. Two radiologists read the same chest X-ray, and when there was a disagreement, a senior radiologist made the final decision. The agreement between the two radiologists was determined using the Cohen's kappa. As many as 210 chest Xx-ray results were readusing CRRS method. The highest Cohen's Kappa coefficient between the two readers was 0.84 for large opacity feature, while the lowest was 0.52 for lymphadenopathy. The validity of the scoring with cut-off ≥ 1 showed the highest accuracy of 76% with sensitivity 76.2%; specificity 61.4%; positive predictive value 68.6%; and negativepredictive value 63.6%. The lowest accuracy showed on the cut-off ≥ 3 and ≥ 4, with accuracy of 60.8%. On each radiological feature, the best accuracy (almost 80%) was showed by large opacity feature. The interpretation using CRRS method has good inter-reader agreement. Keywords: Tuberculosis, CRRS, validity, agreemen