Hi Salman,
I have found OSM to be the best available free dataset for pedestrian network analysis, but it does require cleaning first. Generally an OSM network requires at the very least planarized lines where there is not a "tunnel" or "bridge" tag (it does not go over or underground- you don't want your trail under a highway to have connectivity to an interchange!). It also usually requires speeds to be imputed off functional class. If your functional class field from OSM is not very dependable, I would suggest not using it. However, if OSM is the best available, maybe try to compare it to another series of datasets?
If you can't find a good dataset to compare against in centerline form, you might be able to use more aggregate geometries to prioritize which parts of the network you QAQC. This could be demographic data sets, or potentially amenity datasets from OSM so that you can prioritize connections around restaurants and other likely destinations that people might want to walk to.
Others might have better advice.
David
David Wasserman, AICP