Disproportionality　based　methods　are　widely　used　in　detecting　occupational　hazard　signals.　Yet,　deciding　on　a　proper　background　is　hard　because　there　are　subgroups　vulnerable　to　given　exposures,　and　in　many　cases　such　subgroups　can　not　be　defined　by　job.　This　paper　presents　an　interactive　way　to　do　disproportionality　analysis　on　occupational　health　examination　data.　It　firstly　uses　multi-dimensional　scaling　to　project　data　from　a　subspace　to　a　two-dimensional　space,　which　is　further　divided　into　rectangle　cells.　Then　each　cell　is　served　as　a　background　to　evaluate　disproportionality　for　potential　hazard　record(PHR),　which　is　jointly　defined　by　a　cell　and　a　company.　The　PHRs　with　high　disproportionalities　are　checked　interactively　to　select　interesting　ones.　The　method　has　been　used　to　analyse　an　occupational　hearing　loss　data,　and　has　found　an　interesting　potential　signal　for　further　analysis.