Determining the mechanisms that drive the evolution of protoplanetary disks is a necessary step to understand how planets form .
Here we measure the mass accretion rate for young stellar objects with disks at age > 5 Myr , a critical test for the current models of disk evolution .
We present the analysis of the spectra of 36 targets in the \sim 5-10 Myr old Upper Scorpius star-forming regions for which disk masses were measured with ALMA .
We find that the mass accretion rates in this sample of old but still survived disks are similarly high as those of the younger ( \sim 1 - 3 Myr old ) star-forming regions of Lupus and Chamaeleon I , when considering the dependence on stellar and disk mass .
In particular , several disks show high mass accretion rates \gtrsim 10 ^ { -9 } M _ { \odot } /yr while having low disk masses .
Furthermore , the median values of the measured mass accretion rates in the disk mass ranges where our sample is complete at a level \sim 60 - 80 % are compatible in these three regions .
At the same time , the spread of mass accretion rates at any given disk mass is still > 0.9 dex even at age > 5 Myr .
These results are in contrast with simple models of viscous evolution , which would predict that the values of the mass accretion rate diminish with time , and a tighter correlation with disk mass at age > 5 Myr .
Similarly , simple models of internal photoevaporation can not reproduce the observed mass accretion rates , while external photoevaporation might explain the low disk masses and high accretion rates .
A partial possible solution to the discrepancy with the viscous models is that the gas-to-dust ratio of the disks at \sim 5-10 Myr is significantly different and higher than the canonical 100 , as suggested by some dust and gas disk evolution models .
The results shown here require the presence of several inter-playing processes , such as detailed dust evolution , external photoevaporation and possibly MHD winds , to explain the secular evolution of protoplanetary disks .