Socioeconomic bias in influenza surveillance