Recent advances in generative models have accelerated progress in automated report generation from chest X-rays, which can potentially reduce the workload for clinicians. However, existing methods are often biased to the most prevalent reporting style in their training data, overlooking variations in clinical workflows across regions, institutions, and languages that occur in real-world clinical data. To address these limitations, we introduce a report generation framework that customizes reports through in-context learning from style examples. Our approach includes (i) a style classification metric to quantify customization effectiveness, (ii) a report customization method to adapt generated reports to diverse styles, and (ii) a standalone report generation model that enables multi-style, multilingual report generation. Through our results we show current biases in report generation methods and how our report customization strategies mitigate this. By improving adaptability, our method enhances the possibilities for integration of automated report generation into diverse clinical environments.