As there's no such thing as gerrymandering in House races, there's no possibility that an incumbent with low approval ratings would win anyway. Otherwise this data would be statistically suspect.
US House races, US Senate races, and United States Governors races: the results are not hard to find.
And we’re looking for those in which the incumbent was polling below 50% prior to the election. Again, not hard to find.
Adding and subtracting: not hard to do.
Result: like on a math test, what did the calculation say?