The pattern continues indeed except the RHS becomes more and more involved as you increase $r$. At any rate, here is what we get for the case $r=4$: $$\sum_{k=0}^n\binom{n}k\binom{n}{k+a}\binom{2k-n+a}4= \frac{(a^2-n^2-6n+11)\,(a-n)\,(a+n)}{8\,(2n-1)\,(2n-3)}\cdot \binom{2n}{n+a}.$$ Another sampler for $r=5$: $$\sum_{k=0}^n\binom{n}k\binom{n}{k+a}\binom{2k-n+a}5= -\frac{(a^2-n^2-2n+5)\,(a-n)\,(a+n)}{4\,(2n-1)\,(2n-3)}\cdot \binom{2n}{n+a}.$$ These kinds of problems can be proved by the automated tools of [Zeilberger's algorithm][1] which relies on [Wilf-Zeilberger pairs][2]. If you are able to use the Maple symbolic software then [download the algorithm][3] to make use of it. These days, this package is also part of the routines in Maple as well as Mathematica. [1]: https://mathworld.wolfram.com/ZeilbergersAlgorithm.html [2]: https://en.wikipedia.org/wiki/Wilf%E2%80%93Zeilberger_pair [3]: https://sites.math.rutgers.edu/~zeilberg/tokhniot/EKHAD