This is what I came up with, which took waaay too long, the rules are hard to remember

#!/usr/bin/perl --
use strict; use warnings;
use XML::Rules;
use Data::Dump qw/ dd /;

my $ta = XML::Rules->new(
    qw/ stripspaces 8 /,
    rules => {
'Outpatient_Services' =>  'no content',
  'Outpatient_Service' => 'as array no content',
#~     'Outpatient_Clinic' => 'content by AM_Key',
    'Outpatient_Clinic' => sub {
#~ $rule->( $tag_name, \%attrs, \@context, \@parent_data, $parser)
#~ my ($tagname, $attrHash, $contexArray, $parentDataArray, $parser) =
+ @_;
        my $amk = $_[1]->{AM_Key} ;
        return unless $amk;
        { $amk => 1 };
    },
#~         _default => sub { $_[0] => $_[1]->{_content} },
        _default => 'content',
        'Outpatient_Clinic_Special' => undef,
    },

);

my $ref = $ta->parsefile( 'pm1002448.xml' );
dd $ref;
use YAML(); print YAML::Dump( $ref);
__END__
{
  Outpatient_Services => {
    Outpatient_Service => [
      { AM01 => 1 },
      { AM01 => 1 },
      { AM02 => 1 },
      {},
      { AM04 => 1 },
      {},
    ],
  },
}
---
Outpatient_Services:
  Outpatient_Service:
    - AM01: 1
    - AM01: 1
    - AM02: 1
    - {}
    - AM04: 1
    - {}
[download]