Establishing diagnostic reference levels for pediatric chest radiography at a dedicated children's hospital in Hong Kong: A comparison of single criteria and hybrid grouping methods.

OBJECTIVE: To establish local diagnostic reference levels (LDRLs) and achievable doses (ADs) for pediatric chest anteroposterior/posteroanterior (AP/PA) radiography at a dedicated children's hospital in Hong Kong, and to compare the performance of single criteria and hybrid patient grouping methods for characterizing dose variability within the heterogeneous pediatric population. METHODS: Patient demographic and dose-area product (DAP) data from 5,508 pediatric chest AP/PA examinations performed between 2021 and 2023 were retrospectively extracted from the local automated dose-monitoring system. The sampled patient population, with ages ranging from 0 to 18 years, comprised a diverse demographic profile of weights from 2.26 to 80 kg, body mass index [BMI] from 8.8 to 39.9 and antero-posterior [AP] thicknesses between 9 and 30 cm. They were grouped using four single-criteria methods (age, weight, BMI, and AP thickness), two age-based hybrid methods (age & weight, age & BMI) with a minimum of 20 cases per subgroup per year and an AP thickness-based hybrid method (AP thickness & BMI) with a minimum of 15 cases per subgroup per year. For each group, AD and LDRL were defined as the median (Q2) and 75th percentile (Q3) of DAP distribution, respectively, and exponential DRL curves were also fitted to the pooled data, where goodness of fit was assessed using the coefficient of determination (R²) to address the correlations between patient grouping metrics and the dose indicator. RESULTS: The developed LDRLs generally increased with patient size across all grouping methods, from 0.11 dGy·cm² to 0.78 dGy·cm². Among the fitted DRL curves for the four single-criterion grouping methods, BMI (R² = 0.995 for Q3) and AP thickness (R² = 0.983) showed the strongest dose-size correlations, followed by weight (R² = 0.931), whereas age alone showed the weakest correlation (R² = 0.811). Hybrid age & weight and age & BMI groupings further improved model performance for most categories by revealing within-age dose heterogeneity, while the hybrid AP thickness & BMI grouping was constrained by its logistical impracticality and possible dose masking effect. When compared with published European guidelines on Diagnostic Reference Levels for Pediatric Imaging and national DRLs, HKCH LDRLs were generally comparable or lower across most size groups but were 46.4-54.5% higher for patients with age < 4 years when compared with the German national DRLs, revealing the need for further local optimization in particular areas. CONCLUSION: Comprehensive LDRLs and ADs for pediatric chest AP/PA radiography were established for a major Hong Kong pediatric centre, demonstrating good alignment with international reference levels while highlighting the potential for further dose optimization in the pediatric population. Weight-based grouping offers a practical primary metric for local dose monitoring, whereas BMI, AP thickness, and hybrid age-size approaches provide added value for detailed dose analysis and identification of optimization targets, particularly in the heterogeneous pediatric subgroups. PLAIN LANGUAGE SUMMARY: Children vary greatly in size as they grow, so it is important to use the right amount of radiation when taking chest X-rays. This study reviewed >5,500 chest X-rays from a children's hospital in Hong Kong to develop local radiation dose benchmarks and compare different ways of grouping children for dose monitoring. This study found that radiation doses generally increased with patient size, and that grouping children by weight, body size, or a combination of factors provided a better understanding of dose patterns than age alone. This matters because accurate dose benchmarks can help healthcare teams identify opportunities to reduce radiation exposure while maintaining high-quality imaging for children.