One sample is not enough, but maybe something like this?

#! perl -slw
use strict;

my %data = (
    1047633 =>
'01.12.199100.00.00003 T8 15 SN Y2001.11.200400095.8000071.8500081.454
+001.11.1994(Anaes.)5001.12.1991Metatarsal, 1 of, treatment of fractur
+e of'
);

my $re_date = qr[\d{2}\.\d{2}\.\d{4}];
my $re_float= qr[\d{5}\.\d{2}];

for my $key ( keys %data ) {
    my @fields = $data{ $key } =~ m[
        ( 20 $re_date $re_float{3} )
        ( 40 $re_date \( [^)]+ \) )
        ( 50 $re_date .* $ )
    ]x;

    print "'$_'" for @fields;
}

__END__
P:\test>401374
'2001.11.200400095.8000071.8500081.45'
'4001.11.1994(Anaes.)'
'5001.12.1991Metatarsal, 1 of, treatment of fracture of'
[download]

To explan the regex:

m[
  ## Capture, starting with '20', one date, and 3x %8.2 floats
  ( 20 $re_date $re_float{3} )

  ## Capture, starting with '40', one date, '(', non-')' to the ')'
  ( 40 $re_date \( [^)]+ \) )

  ## Capture, '50', a date, everything to the end of line.
  ( 50 $re_date .* $ )
]x; ## ignore whitespace.
[download]