I can't vouch for the details of the forecast, but this report predicting 100 million otherwise avoidable deaths due to rising global temperatures by 2030 is a great illustration of why conventional economic cost-benefit analyses of global climate change don't really work. The killer sentence is "more than 90 percent of those deaths will occur in developing countries."
The thing about developing countries is that they're really poor. Prime Minister Sheikh Hasina of Bangladesh notes that "one degree Celsius rise in temperature is associated with 10 percent productivity loss in farming" which is about 2 percent of his country's GDP. But 2 percent of Bangladesh's GDP is a really small number—$2.5 billion—since Bangladesh is so poor. Japan, by contrast, has a GDP of $5.87 trillion. So a policy that cost Japan 0.042 percent of GDP in order to avoid a loss of 2 percent of Bangladesh's output fails the cost-benefit test. Normal moral logic holds this to be a perverse result. Precisely because Japan is so much richer than Bangladesh, output loss in Bangladesh is a much bigger deal—literally a life or death matter for the population. But in the global growth calculus taking countries that are already poor with economies dominated by low-productivity agriculture and pushing them into famine conditions isn't a big deal compared to small reductions in rich country manufacturing and construction activity.