The missingness is *not* random. If low-income patients are more likely to have missing insurance information, dropping those rows silently removes low-income patients from your analysis. Your results now describe only the people with complete records — which may not represent the population you car