Serious games, conveying educational knowledge rather than merely entertainment, are a rapidly expanding research domain for cutting-edge educational technology. Digital interventions like serious games are great opportunities to overcome challenges in low-and-middle-income countries that limit access to health information, such as social barriers like low-literacy and gender. MANTRA: Increasing maternal and child health resilience before, during and after disasters using mobile technology in Nepal takes on these challenges with a novel digital health intervention; a serious mobile game aimed at vulnerable low-literacy female audiences in rural Nepal. The serious game teaches 28 learning objectives of danger signs in geohazards, maternal, and neonatal health to improve knowledge and self-assessment of common conditions and risks to inform healthcare-seeking behavior. Evaluations consisted of recruiting 35 end users to participate in a pre-test assessment, playing the game, post-test assessment, and focus groups to elicit qualitative feedback. Assessments analyzed knowledge gain in two ways; by learning objective with McNemar tests for each learning objective, and by participant scores with paired t-tests of overall scores and by module. Results of assessments of knowledge gain by learning objective (McNemar tests) indicate participants had sufficient prior knowledge to correctly interpret and respond to 26% of pictograms (coded AA), which is a desirable result although without the possibility of improvement through the intervention. The geohazard module had greatest impact as 16% of responses showed knowledge gain (coded BA). The two most successful learning objectives showing statistically significant positive change were evidence of rockfalls and small cracks in the ground (p = < 0.05). Assessment of knowledge gain by participant scores (paired t-tests) showed the 35 participants averaged a 7.7 point improvement (p < 0.001) in the assessment (28 learning objectives). Average change in knowledge of subdivided module scores (each module normalized to 100 points for comparison) was greatest in the geohazard module (9.5 points, p < 0.001), then maternal health (7.4 points, p = 0.0067), and neonatal health (6.0 points, p = 0.013). This evaluation demonstrated that carefully designed digital health interventions with pictograms co-authored by experts and users can teach complex health and geohazard situations. Significant knowledge gain was demonstrated for several learning objectives while those with non-significant or negative change will be re-designed to effectively convey information.