It looks like you're dealing with polymorphic relations in SQL. Right now I don't have a set of tables to test with, but you should be able to do the join on the server-side like this (untested):

SELECT m.*,
    coalesce(sb.somecol, sd.somecol, tb.somecol, td.somecol) AS someco
+l
FROM KPI_Master m
LEFT JOIN SDCCHBLOCKING sb ON m.NE = sb.NE AND m.KPI = 'SDCCHBLOCKING'
LEFT JOIN SDCCHDROP     sd ON m.NE = sd.NE AND m.KPI = 'SDCCHDROP'
LEFT JOIN TCHBLOCKING   tb ON m.NE = tb.NE AND m.KPI = 'TCHBLOCKING'
LEFT JOIN TCHDROP       td ON m.NE = td.NE AND m.KPI = 'TCHDROP'
[download]

This way you should only need to fetch n rows from the database -- where n is the amount of rows you have in KPI_Master -- with only a single round-trip. This should also simplify your code to something like this:

my $dbh = DBI->connect('dbi:ODBC:driver=Microsoft Access Driver (*.mdb
+, *.accdb);dbq=D:\Project\Project\BSC_KPIMonitoring\CHN\E_BSC_KPIMoni
+torDB.mdb', undef, undef, { RaiseError => 1});

my $query = <<EOF;
SELECT m.BSC, m.KPI, m.THRESH,
    coalesce(sb.somecol, sd.somecol, tb.somecol, td.somecol) AS someco
+l
FROM KPI_Master m
LEFT JOIN SDCCHBLOCKING sb ON m.NE = sb.NE AND m.KPI = 'SDCCHBLOCKING'
LEFT JOIN SDCCHDROP     sd ON m.NE = sd.NE AND m.KPI = 'SDCCHDROP'
LEFT JOIN TCHBLOCKING   tb ON m.NE = tb.NE AND m.KPI = 'TCHBLOCKING'
LEFT JOIN TCHDROP       td ON m.NE = td.NE AND m.KPI = 'TCHDROP'
EOF

my %kpis = (SDCCHBLOCKING => 'SB', SDCCHDROP => 'SD',
            TCHBLOCKING   => 'TD', TCHDROP   => 'TD');

my $sth = $dbh->prepare($query);
$sth->execute;

while (my @row = $sth->fetchrow_array()) {
    my ($bsc, $kpi, $thresh, $somecol) = @$row;

    if ($somecol > $thresh) {
        print join(",", $kpis{$kpi}, $bsc, $thresh, "/$somecol");
    }
}
[download]

Other points of interest for performance might be reducing the amount of columns you get: don't SELECT *, but SELECT theOneColumnYouNeed