"Although Congress instructed the Attorney General in 1994 to compile and publish annual statistics on police use of excessive force, this was never carried out, and the Federal Bureau of Investigation does not collect these data either."(https://en.wikipedia.org/wiki/Police_use_of_deadly_force_in_...)
In 2013, Fatal Encounters (https://www.fatalencounters.org/our-visualizations/) was started by a single individual to collect statistics of police killings each year. Interestingly, part of the inspiration was an article from 2011, talking about how police may be killing more disabled people than necessary: https://www.pressherald.com/interactive/maine_police_deadly_...
In 2014, Deadspin (https://deadspin.com/were-compiling-every-police-involved-sh...) began a crowdsourcing effort to document police killings. Soon after, The Washington Post followed suit (https://www.washingtonpost.com/graphics/national/police-shoo...).
In 2015, The Guardian started collecting its own statistics in The Counted (https://www.theguardian.com/us-news/series/counted-us-police...).
Mapping Police Violence correlates and visualizes statistics from three different databases (https://mappingpoliceviolence.org/aboutthedata/). An updated 2017 report is here: https://policeviolencereport.org/
Whether the trend of police killing people is increasing or decreasing depends on what numbers you pick, as they vary in collection methods and sources. But one thing that hasn't changed is that black people are disproportionately affected. The 2017 PVR notes "Black people were more likely to be killed by police, more likely to be unarmed and less likely to be threatening someone when killed."
I did not see statistics specifically related to whether someone was disabled. However, the 2017 PVR notes that "Most killings began with police responding to suspected non-violent offenses or cases where no crime was reported". In addition, "1 in 5 people with a gun were not threatening anyone when they were killed. [..] If police did not kill people who were not posing a threat with a gun, there would have been 638 fewer deaths this year — a 57% reduction."