R                                 Input data management in health care simulation Industrial methodology for discrete event simulation applied in a health care system   Master of Science Thesis in the Master Degree Programme, Systems, Control and Mechatronics SOFIA HOLMBLAD MALIN IVARSSON Department of Product and Production Development Division of Production Systems CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden, 2012 Input data management in health care simulation Industrial methodology for discrete event simulation applied in a health care system   SOFIA HOLMBLAD MALIN IVARSSON                         Department of Product and Production Development CHALMERS UNIVERSITY OF TECHNOLOGY Gothenburg, Sweden 2012 Input data management in health care simulation Industrial methodology for discrete event simulation applied in a health care system SOFIA HOLMBLAD MALIN IVARSSON © Sofia Holmblad and Malin Ivarsson, 2012. Examiner: Anders Skoogh anders.skoogh@chalmers.se Department of Product and Production Development Chalmers University of Technology SE-412 96 Gothenburg Sweden Telephone +46 (0)31-772 1000 Reproservice, Chalmers Gothenburg, Sweden 2012 Abstract   In   2010   the   Swedish   government   introduced   a   health   care   reform   called   Vårdvalet.   This   lead   to   a   higher   level   of   competition   between   different   health   care   centres   and   their   desire   to   be   more   effective   increased.   In   industry,   DES-­‐simulation   is   often   used   to   investigate   how   different   changes   affect   a   system,   to   get   the   best   improvements   to   system   performance   for   the   lowest   cost.   This   method   for  modelling   flows   is   now  becoming  more   common   in   the   health   care   sector.   Input   data   management  is  the  most  time  consuming  part  of  a  DES-­‐project,  so  if  it  can  be  reduced  the  total  time   and  cost  for  DES-­‐projects  decrease.  Due  to  this,  an  input  data  management  methodology,  originally   developed   for   industrial   use,  will   be  evaluated   in   a  health   care   context   in   a   case   study   at   Sörhaga   health   care   centre   in  Alingsås.   The   evaluation  will   investigate   how  well   the  methodology  works   in   DES-­‐projects   for  health  care  systems  and  what  modifications   it  might  need  to  be  more  suitable   for   such  projects.   Sörhaga   health   care   centre   experiences   problems   with   their   waiting   room,   which   gets   crowded   during   their   emergency   reception,   also   called   triage   reception.  Another   problem  with   the   triage   is   that   it   usually   does   not   end   when   planned,   which   puts   a   lot   of   pressure   on   the   personnel   who   sometimes  have  to  skip  their  lunch  or  work  overtime  to  catch  up.   From   the   simulation  model   of   Sörhaga   health   care   centre   it  was   seen   that   the   physicians   and   the   district   nurses   working   during   the   triage   are   alternating   bottlenecks.   Several   concepts   were   then   developed   to   try   to   cope  with   this   problem.   The   concepts   were   analysed   and   four   concepts,   two   including   minor   changes   and   two   including   more   extensive   changes,   were   recommended   to   the   health  care  centre.  The  health  care  centre  should  chose  one  of  the  more  extensive  changes  but  both   of  the  two  minor  changes  should  be  implemented.   The  evaluation  of  the  input  data  management  methodology  showed  that  the  methodology  is  suitable   for  DES  project  in  health  care  systems.  It  is  believed  that  by  using  this  methodology  can  the  total  time   consumption  for  DES  project  in  health  care  be  reduced  and  that  this  will  lead  to  that  simulation  will   become  a  greater  option  for  optimising  patient  flows.  This  will  in  turn  result  in  more  effective  health   care  systems  with   less  waiting  times  and  thus  shorten  the  course  of  the  diseases  which   in  the   long   run  will  give  a  healthier  population.               Foreword   This  report  is  the  result  of  our  master  thesis  performed  at  the  Department  of  Product  and  Production   Development  at  Chalmers  University  of  Technology  during  the  spring  of  2012.  The  project  has  been   performed  in  cooperation  with  Sörhaga  health  care  centre  in  Alingsås  and  ÅF  AB.   We  want  to  thank  Catharina  Persson  along  with  all  the  other  personnel  at  Sörhaga  health  care  centre   for  helping  us  with  such  enthusiasm.  We  also  want  to  thank  our  two  supervisors  Anders  Skoogh  at   Chalmers   and   Nils   Bengtsson   at   ÅF   AB.   Thanks   also   to   those   at   ÅF  who   have   helped   us  with   DES   modelling   tips.   Finally  Floda  health   care  centre  deserves   thanks   for  having  us  and   showing  us  how   they  work.     Gothenburg,  June  2012:   Sofia  Holmblad  and  Malin  Ivarsson     Table  of  Contents   1   Introduction  ......................................................................................................................................  1   1.1   Purpose  ......................................................................................................................................  1   1.2   Project  aim  ................................................................................................................................  2   1.3   Problem  formulation  .................................................................................................................  2   1.4   Delimitations  .............................................................................................................................  2   1.5   Short  description  of  the  health  care  centre  ..............................................................................  3   1.6   Report  outline  ............................................................................................................................  3   2   Theory  ...............................................................................................................................................  5   2.1   Lean  ...........................................................................................................................................  5   2.2   Simulation  in  health  care  ...........................................................................................................  6   2.3   Theory  of  constraints  .................................................................................................................  8   2.4   Improvement  strategy  for  health  care  ......................................................................................  9   3   Method  ...........................................................................................................................................  13   3.1   Input  Data  Management  Methodology  ...................................................................................  14   3.2   Evaluation  of  input  data  management  methodology  ..............................................................  17   3.3   Improvement  of  the  health  care  central  .................................................................................  18   4   Current  system  ...............................................................................................................................  23   4.1   Description  ..............................................................................................................................  23   4.2   Results  from  model  of  current  system  ....................................................................................  26   4.3   Analysis  of  the  current  system  ................................................................................................  27   5   Improvement  concepts  for  the  health  care  centre  ........................................................................  29   5.1   General  concepts  .....................................................................................................................  29   5.1.1   Drop-­‐in  to  lab  during  triage  ..............................................................................................  29   5.1.2   Special  booking  to  lab  .......................................................................................................  29   5.1.3   New  schedule  ...................................................................................................................  29   5.2   Triage  team  concepts  ..............................................................................................................  30   5.2.1   First  concept  .....................................................................................................................  30   5.2.2   Second  concept  ................................................................................................................  31   5.3   Other  concepts  ........................................................................................................................  31   5.3.1   Advices  ..............................................................................................................................  31   5.3.2   Back-­‐up  starts  early  ..........................................................................................................  32   5.3.3   Extended  triage  ................................................................................................................  32   6   Results  ............................................................................................................................................  33   6.1   Input  data  management  methodology  ...................................................................................  33   6.2   General  concepts  .....................................................................................................................  35   6.3   Triage  team  concepts  ..............................................................................................................  37   6.4   Other  concepts  ........................................................................................................................  41   6.5   Current  strategy  ......................................................................................................................  44   7   Analysis  ...........................................................................................................................................  47   7.1   Input  data  management  methodology  ...................................................................................  47   7.2   General  concepts  .....................................................................................................................  47   7.3   Triage  team  concepts  ..............................................................................................................  47   7.4   Other  concepts  ........................................................................................................................  48   7.5   Current  strategy  ......................................................................................................................  49   7.6   All  concepts  .............................................................................................................................  49   8   Discussion  .......................................................................................................................................  51   8.1   Input  data  management  methodology  ...................................................................................  51   8.2   DES-­‐project  ..............................................................................................................................  52   8.3   Current  system  ........................................................................................................................  53   8.4   General  concepts  .....................................................................................................................  53   8.5   Triage  team  concepts  ..............................................................................................................  54   8.6   Other  concepts  ........................................................................................................................  56   8.7   Current  strategy  ......................................................................................................................  56   8.8   Comparing  all  concepts  ...........................................................................................................  57   8.9   Summary  .................................................................................................................................  57   9   Conclusion  ......................................................................................................................................  59   References  ............................................................................................................................................  62             1     1 Introduction   In   2009   the   Swedish   government   decided   to   introduce   the   health   care   reform   [Vårdvalet],   which   means  that  from  2010  is  every  person  allowed  to  choose  what  district  health  care  centre  they  want   to  be  listed  at  (Sveriges  riksdag,  2009).  This  puts  more  pressure  on  the  health  care  centres  to  meet   patients’  expectations  to  be  highly  competitive  as  more  patients  yield  more  money  from  the  county   council  (Socialstyrelsen,  2010).  To  stay  competitive  patients  should  be  received  and  taken  care  of  as   quickly  as  possible,  which  means  that  queues  and  lead  times  have  to  be  reduced.  Even  though  some   money   is   received   for   each   patient   are   the   health   care   centres’   resources,   such   as   personnel   and   premises,  limited  and  still  have  to  suffice  for  a  possible  increase  in  number  of  patients.   The   health   care   centre   in   this   study,   Sörhaga   health   care   centre   in   Alingsås,   has   attracted   many   patients   due   to   its   good   reputation.   As   a   result   the   health   care   centre   now   has   to   admit   more   patients   than   it   was   originally   designed   for.   Due   to   this,   there   is   now   a   major   problem   with   the   waiting  room,  which  sometimes  gets  so  crowded  that  patients  are  sitting  on  the  floor.  Also,  there  are   many   patients   and   little   personnel   during   the   emergency   reception,   which   leads   to   long   waiting   times  for  the  patients  and  the  personnel  sometimes  have  to  skip  lunch  or  work  overtime  to  catch  up.   This   leads   to   both   unsatisfied   patients   and   personnel,   which   in   turn   can   lead   to   economic   losses.   There   is  a   risk  of  an   increased  amount  of  patients  as   the  health  care  situation  might  change   in   the   region.   Thus,   the   health   care   centre  will   not   only   have   to   be   able   to   handle   the   patients  who   are   already  listed  but  also  new  patients  and  still  meet  the  expectations  of  both  patients  and  personnel.   To  deal  with  this,  the  health  care  centre’s  director  wants  to  try  an  approach  normally  used  in  industry   called   Discrete   Event   Simulation   (DES),   which   is   becoming   more   common   within   the   health   care   sector  (Centre  for  health  care  management  –  University  of  British  Columbia,  2011).  It  can  provide  a   decision   basis   to   improve   operations  management   but   the   result   from   a   DES   simulation   depends   heavily  on  the  quality  of  input  data.  The  data  collection  phase  is  also  the  most  time  consuming  part   of  a  DES  project  (Skoogh,  2008)  and  can  be  even  more  time  consuming  if  data  is  not  collected  in  an   organised   way.   Therefore,   this   study   will   examine   if   an   input   data   management   methodology   (Skoogh,  2008),  originally  developed  for  industrial  use,  can  be  applied  in  the  medical  sector.   1.1 Purpose   The  purpose  is  to  contribute  to  making  DES-­‐simulation  a  greater  option  for  analysing  patient  flows  in   health   care   by   evaluating   the   input   data  management  methodology   by   Skoogh   (Skoogh,   2008).   A   good   methodology   for   input   data   management   reduces   time   and   cost   for   DES-­‐projects.   Skoogh’s   methodology  is  originally  designed  for  industrial  projects  and  might  need  some  modification  to  fit  for   health   care  projects.   This  methodology  will   therefore  be  applied   in  a   case   study  at   Sörhaga  health   care  centre  to  see  what  such  modifications  might  be.   The   high   number   of   listed   patients   at   Sörhaga   health   care   centre   is   causing   concern   as   it   causes   crowding  in  the  waiting  room.  Another  concern  is  the  stress  level  for  the  health  care  personnel,  who   feel  that  the  situation  is  unsustainable.  The  second  purpose  of  this  project  is  therefore  to  provide  the   health   care   centre   director   with   a   decision   basis,   of   how   to   take   best   actions   to   meet   the   high   demands.   2       1.2 Project  aim   The  aim  is  to  apply  and  evaluate  the  input  data  management  methodology  in  a  case  study  to  find  out   how  well   it   works   for   DES-­‐projects   in   health   care   systems.  Modification   suggestions   shall   also   be   given  which  can  make  the  methodology  more  suitable  for  health  care  applications.   The  aim  of   this  project   is   also   to  provide  Sörhaga  health   care   centre  with  a   solution  suggestion   to   their  problems  regarding  the  high  demands  on  the  personnel  and  crowded  waiting  room  caused  by   the  high  amount  of   listed  patients.  The  aim   is   that   the  developed  concepts  will  handle   the  current   amount  of  patient  and  meet  the  health  care  centre’s  goals.   In  short  terms,  the  following  will  be  done  for  the  health  care  centre:   • Model  and  analyse  the  current  situation  at  the  health  care  centre.   • Develop  solution  concepts.   • Model  the  concepts.   • Analyse  and  optimise  the  most  promising  concepts.   • Present  the  best  ones  to  the  health  care  centre.   1.3 Problem  formulation   The   general   question   is:   Is   the   Input  Management  Methodology   for   DES-­‐projects   usable   for   DES-­‐ projects   in   health   care   systems   and   how   can   it   be   modified   to   be   more   suitable   for   health   care   systems?   The   combination   of   the   input   data   management   methodology   and   a   DES-­‐project   is   applied   at   Sörhaga  health  care  centre.  This  combination  will  evaluate  how  the  triage  can  be  organised  to  meet   the  triage  goals  with  the  current  number  and  type  of  triage  patients.  A  major  part  of  this  question  is   to  answer:  Can  the  current  staff  manage  the  triage  or  will  more  personnel  have  to  be  scheduled  for   it?   1.4 Delimitations   These  are  the  delimitations  for  the  DES-­‐project:   • The  health  care  centre  takes  care  of  several  different  types  of  patients  during  a  day  but  only   triage  patients,  afternoon  emergency  patients,  semi-­‐emergency  patients  and  revisiting   patients  will  be  modelled  and  analysed.   • Only  activities  at  the  health  care  centre  itself  will  be  taken  into  account.   • The  counsellor  will  not  be  modelled  or  analysed.   • The  district  nurses  will  only  be  analysed  during  the  triage.   • The  medical  secretaries  will  only  be  modelled  while  they  are  at  the  reception  desk.   • The  time  for  a  patient  to  undress  and  dress  is  the  same  with  a  district  nurse  as  with  a   physician.   • All  revisiting  patients  are  scheduled  for  45  minutes  if  they  are  seeing  a  resident  physician,  i.e.   if  only  30  minutes  are  scheduled  in  the  real  system  it  is  changed  to  45  minutes  in  the  model.   • There  are  always  two  district  nurses  scheduled  for  the  triage,  never  only  one  as  it  is  in  the   real  schedule  one  day.   • No  regard  is  given  to  the  fact  that  there  are  usually  more  patients  at  the  triage  during   Mondays.   3       • All  triage  patients  see  a  district  nurse  first.   • No  wild  cards  will  be  modelled,  such  as  patients  who  should  have  been  sent  to  a  hospital   emergency  ward.  This  sort  of  patients  craves  more  time  and  they  are  rare.   • Medical  aspects  will  not  be  taken  into  consideration.   1.5 Short  description  of  the  health  care  centre   At   the   health   care   centre   there   are   four   main   groups   of   personnel:   physicians,   district   nurses,   assistant   nurses   and  medical   secretaries.  When   a   patient   calls   in   the  morning   and   needs   to   see   a   district   nurse   or   physician   the   same   day   they   are   booked   for   the   triage.   The   district   nurses,   who   answer  the  calls,  assess  whether  a  patient  needs  to  come  to  the  health  care  centre  or  not.  If  so,  the   patient  is  told  to  come  sometime  within  a  30-­‐minute  span.  Each  patient  needs  to  call  first  and  have   to   register   at   the   reception   desk   when   they   arrive,   where   a   medical   secretary   is   stationed.   The   patient  will  then  have  to  wait  until  it  is  their  time  to  see  a  district  nurse,  who  examines  them.  Then   the  patient  has  to  wait  again,  unless  it  is  sent  home,  until  it  is  called  for  by  a  nurse  assistant  from  lab   or  by  a  physician.  Either  a  physician  or  a  district  nurse  can  call  for  the  patient  after  a  visit  to  lab.  The   physicians  can  also  send  patients  to  wait  for  lab  tests  but  then  the  physician  will  call  for  them  again.   All  waiting  is  done  in  the  same  waiting  room,  which  often  becomes  crowded  during  the  triage.   1.6 Report  outline   Chapter   1   –   Introduction:  Here   the   purpose,   aim,   problem   formulation   and   delimitations   for   the   project  are  presented.    There  is  also  a  short  description  of  the  health  care  centre  at  which  the  case   study  in  this  project  is  performed.   Chapter  2  –  Theory:  The   theory  used  during   the  project   is  presented  here.   Some   theory   is  usually   used  in  industrial  projects  and  some  theory  specific  for  health  care  projects  presented.   Chapter  3  –  Method:  Several  methods  are  used  during  the  project  and  they  are  presented  here.  The   chapter   is   divided   into   three   parts   where   the   first   part   describes   the   input   data   management   methodology  that   is  evaluated  in  this  project.  The  second  part  describes  how  the  evaluation  of  the   input  data  management  methodology  will  be  made  and  the  third  part  describes  how  the  case  study   is  performed.   Chapter  4  –  Current  system:  A  more  extensive  description  of  how  the  health  care  centre  is  organised   today   is   provided   in   this   chapter.   Also   the   results   and   analysis   from   the   simulation  model   of   the   current  system  are  presented  here.   Chapter   5   –   Improvement   concepts   for   the   health   care   centre:  All   of   the   improvement   concepts   developed  for  the  health  care  centre  are  presented  here.  There  are  eight  concepts  divided  into  three   categories  depending  on  their  characteristics.   Chapter   6   –  Results:  The   results   from   the  evaluation  of   the   input  data  management  methodology   and  from  the  model  runs  of  the  different  concepts  are  presented  in  this  chapter.  There  are  also  some   results  from  model  runs  with  the  current  strategy  but  with  more  personnel  presented  here.   Chapter  7  –  Analysis:  In  this  chapter  the  results  from  the  evaluation  of  the  input  data  management   methodology  and  from  the  model  runs  are  analysed.   4       Chapter  8  –  Discussion:  The  results  and  the  analysis  of  the  input  data  management  methodology  and   of   the  different  model   runs  are  discussed  and  conclusions  are  drawn   in   this   chapter.  The  methods   used  during  this  project  are  also  discussed  here.   Chapter  9  –  Conclusion:  The  final  conclusions  of  this  project  are  presented  in  this  chapter.     5       2 Theory   The   theory   used   during   the   project   is   described   in   the   following   four   sections.   Theories   from   industrial  systems  that  can  be  applied  in  health  care  systems  are  described,  namely  Lean  and  Theory   of   constraints,   and   what   needs   to   be   considered   when   applying   these   theories   in   health   care   systems.  There  are  also  some  theories  that  are  specific  for  health  care  systems,  namely  what  needs   to  be  considered  while  performing  a  simulation  project  in  health  care  systems  and  an  improvement   strategy  specific  for  health  care  systems  called  Genombrott.   2.1 Lean   When  talking  about  Lean,   the  expression  ”Lean   thinking”   is  often  used   instead  of   just   Lean.  This   is   because   Lean   is   not   a  method   that   is   implemented   once   to   optimise   the   system   but   it   is   a   work   philosophy   that   gives   a   long   term  effect   in   a   system   (USIL,   2007)(LERC,   2007).   Lean   is   about  using   current  resources  in  a  better  way,  creating  a  flexible  work  environment,  reduce  lead  times,  increase   quality  and  to  focus  on  the  costumers  desires  (Lean  Concepts  AB).  To  successfully  use  Lean  it  has  to   be  implemented  throughout  the  entire  organisation,  everyone  must  be  on  board  and  engaged  in  the   change  and  remember  that  the  most  important  resource  when  working  with  Lean  is  the  co-­‐workers.   Lean  requires  good  leaders  that  should  work  to  create  a  culture  that  values  candidness,  consensual   trust,  teamwork,  customer  focus  and  education.  The  focus  within  the  organization  should  not  be  on   results  but  instead  on  the  people  and  the  work  environment.  If  something  does  not  work  as  intended   one  should  not  try  to  find  a  scapegoat  but  instead  see  it  as  an  opportunity  to  learn  and  correct  it  so  it   does   not   happen   again.   The   leaders   in   the   organization   should   be   visible   to   the   co-­‐workers,   they   should  listen  to  the  co-­‐workers  and  act  from  it,  it  is  important  that  the  co-­‐workers  feel  attended  to.  It   is  also  important  that  the  co-­‐workers  know  that  their  improvement  suggestions  will  benefit  them  and   not  lead  to  that  someone  losses  their  job  (ibid).   There  are  several  main  principles  in  Lean  that  should  be  kept  in  mind  at  all  times  (ibid):   • Focus  on  the  costumer   • Work  in  a  standardised  way  to  reach  stability  in  the  system   • Each  step  in  the  process  should  be  done  with  high  quality   • The  demand  should  control  the  workflow   • Teamwork  and  commitment   • Continuous  improvements   Another  important  principle  of  Lean  is  to  minimise  waste  by  dividing  time  into  value  adding  time  and   non-­‐value  adding  time  and  then  try  to  remove  as  much  non-­‐value  adding  time  as  possible.  To  make   this  process  easier  there  are  five  steps  that  can  be  followed  (LERC,  2007):   1. Specify  value  from  a  costumer  perspective   2. Identify  all  activities  in  the  value  stream  and  remove  non-­‐value  adding  activities  when   possible   3. Tighten  the  sequence  of  the  value  adding  activities   4. Let  the  costumer  pull  value  from  the  sequence   5. Repeat  step  1-­‐4  until  perfection  is  reached   6       In  a  typical  manufacturing  system,  research  suggests  that  only  about  5%  of  all  activities  add  value  to   the  product,  35%  are  necessary  non-­‐value  adding  activities  and  60%  do  not  add  any  value  at  all  to  the   product.  This  suggests  that  there  is  much  to  gain  by  removing  the  non-­‐value  adding  activities  (ibid).   2.1.1 Lean  Healthcare   There  are  two  main  principles  for  Lean  Healthcare  (SUS,  2009):   • Focus  on  patient  flow.   • Change  the  work  structure  to  allow  different  work  teams  to  improve  their  own  way  of  work   on  a  continuous  basis.   To   achieve   this,   the  managers   in   the   health   care   sector  will   have   to   change  how   they  work.   Their   most   important   task   has   to   be   to   work   for   an   organization   with   constantly   improving   quality,   productivity  and  well-­‐being  for  the  personnel   (SUS,  2009)  (USIL,  2007).  Most  of  today’s  health  care   managers’   lack   the   knowledge   needed   to   carry   out   this   change   and   it   is   therefore   important   to   provide  education  (USIL,  2007).   When  it  comes  to  the  patient  flow  it  should  be  improved  by  creating  work  teams  that  include  all  the   competence   needed   to   treat   the   patient   so   that   the   patient   can   be   treated   immediately.   This   has   proven   efficient;   one   example   is   that   the   time   to   assess   a  woman  with   a   lump   in   her   breast  was   reduced  from  42  days  to  2-­‐3  hours  (ibid).   The  co-­‐workers  are,  off   course,   still   the  most   important   resource  and   it   is   they  who  have   to   impel   improvements  in  the  organization.  It  can  be  seen  as  co-­‐workers  in  health  care  today  have  two  jobs:   one  is  to  provide  care  to  the  patient  and  the  other   is  to   improve  their  way  of  work  with  respect  to   the  patient  (SUS,  2009)  (USIL,  2007).   2.2 Simulation  in  health  care   It  can  be  rather  difficult  to  perform  simulation  studies  in  the  health  care  sector  for  several  reasons,   some  of  them  being  (Hakes,  1994):   • Historically  there  has  been  low  motivation  to  control  costs  in  the  health  care  sector.   • Health  care  managers’  find  their  current  methods  for  decision  making  sufficient.   • Simulation  is  unfamiliar  and  can  sometimes  be  seen  as  dehumanising,  which  is  deterring.   • Simulation  requires  a  high  technical  knowledge  and  it  can  thus  be  hard  to  understand.   In   spite   of   this   the   health   care   sector   is   starting   to   use   simulation  more   and  more   due   to   higher   demands  on  cost  control.  Simulation  is  a  suitable  decision  making  tool  due  to  its  ability  to  handle  the   highly   stochastic   nature   of   disease   processes   and   the   high   complexity   in   and   between   different   subsystems.  Still,   there  are  several  tactical   issues  concerning  simulation   in  health  care  that  need  to   be  considered  when  performing  a  simulation  project  for  the  health  care  sector,  namely  the  degree  of   model   complexity,   definitions  of   input  distribution,  model   validation   and   interpretation  of   findings   (Lowery,  1996).   Since  it  takes  quite  a  lot  of  time  and  effort  to  perform  a  simulation  project  it  is  desirable  not  to  spend   more  time  than  necessary  on  building  a  model.  Because  of  this  the  complexity  of  a  model  should  be   as   low  as  possible  but   still  high  enough   to  give  proper  answers  of  how  to  address  problems   in   the   system.   It   is   better   to   first   build   a   simpler   model   and   then   later   on,   if   proven   necessary,   add   7       complexity  to  the  model.  This  can  in  most  cases  be  done  without  wasting  any  time  by  first  building   the  simpler  model.  In  a  health  care  application  one  example  can  be  the  patient  case-­‐mix  that  can  give   a  high  complexity  to  the  model  if   it   is  divided  into  an  unnecessary  number  of  categories.  The  key  is   not   to   divide   the   case-­‐mix   patient   into  more   categories   than  what   is   needed  with   respect   to   the   objectives  of  the  model.   If  the  objective  is  to  investigate  the  effect  of  a  change  in  the  patient  case-­‐ mix  with  respect  to  different  departments  such  as  surgery  or  medicine  than  the  patients  should  only   be  divided  into  these  categories  and  no  more  (ibid).   When  building  the  model  there  are  several  areas  that  might  be  experienced  as  difficult  by  someone   used  to  perform  simulation  projects  in  industry.  The  patients  might  need  to  be  given  care  in  another   order   than   how   they   arrived   and   are   instead   given   a   priority.   A   patient   with   a   life   threatening   condition   will   have   to   be   given   priority   over   patients   with   non-­‐life   threatening   conditions   even   though  the  physician  already  has  started  to  treat  a  patient  with  lower  priority.  A  similar  problem  can   occur  if  an  emergency  operation  has  to  be  performed  on  a  patient.  The  recovery  room  might  be  full   and   a   patient   from   the   recovery   room   will   have   to   be   moved   to   a   ward   to   make   room   for   the   emergency  patient.   It  might   also  be  difficult   to  model   a   reunion  of   e.g.   lab   samples  with   the   right   patient  or  when  a  patient  is  leaving  a  room  and  later  should  return  to  the  same  room.  It  can  also  be   just  as  difficult   to  model   random  room  reservations,   i.e.  when  any  patient  can  be  examined   in  any   room.  The  use  of  multiple  resources  at   the  same  time  can  also  be  tricky  to  model  or  when  several   activities   are   performed   for   one   patient   at   the   same   time.   Last   but   not   least,   it   is   important   to   remember  that  patients  are  not  the  only  persons  in  a  waiting  room.  Many  patients  bring  companions   who  might   follow   the   patient   or   stay   in   the  waiting   room  during   the   patients   visit;   this   should   be   taken  into  account  when  creating  a  model  (Nordgren,  2012).   A  health  care  simulation  model  often  need  input  data  such  as  arrival  time  for  patients  and  the  service   time   for   different   resources   (e.g.   physicians,   nurses,   rooms,   beds,   equipment,   etc.).   This   data   can   then  be   represented  either  as  a   theoretical  or  as  an  empirical   input  distribution   in   the  model.  The   advantage  of  an  empirical  distribution   is   that   it   fits   the  data   set  but  has   the  disadvantage   that   the   entire   distribution  will   have   to   be   changed   while   testing   different   values   of   the   input   data.   For   a   theoretical  distribution  only   the  distribution  parameters  will  have   to  be  changed,   the   shape  of   the   distribution  can  be  assumed  to  remain  the  same,  but  on  the  other  hand  it  might  not  fit  the  data  as   good.   It   is   important   to  weigh  these   factors   to  each  other  when  deciding  what   type  of  distribution   that   should   be   used.   It   can   also   be   seen   in   the  model   validation   if   the   accuracy   of   the   input   data   distributions  are  enough,  if  not  it  is  better  to  use  an  empirical  distribution  (Lowery,  1996).   Many   clinicians   and   administrators   still   doubt   the   capability   to   simulate   the   high   complexity   and   randomness  in  a  health  care  system  even  though  this  is  way  simulation  was  chosen.  Due  to  this  it  is   important  to  properly  validate  the  model  and  demonstrate  it.  Validation  of  a  model  can  be  done  by   comparing  a  sample  of  model  observations  to  a  sample  of  real  observations  given  that  they  have  the   same   set   of   input   conditions.   This   can   however   be   rather   difficult   to   accomplish   in   a   health   care   system   because   of   the  many   rapid   changes   in   health   care   today.   Even   though   the   validation   of   a   model   fails,   i.e.   the   simulation   model   does   not   accurately   model   the   real   system,   the   simulation   model  might  still  be  useful.  Such  as,  if  the  purpose  for  building  the  simulation  model  was  to  compare   different  alternatives  to  each  other  and  not  to  give  absolute  answers  of  how  the  system  will  behave   when  implemented  (ibid).   8       Once   the   model   is   validated   different   experiments   can   be   performed   to   provide   answers   to   the   desired  questions.  There  are  however  some  common  misunderstandings  when  clinicians  and  health   care  managers   review   the   results   from   the  model.   First,   they   often   want   the  model   to   provide   a   single  answer  of  how  to  optimise  the  system,  which  a  simulation  model  does  not  give.  Instead  it  gives   answers   to   “what   if”   questions,   which   require   a   lot   of   time   and   effort   to   review,   and   gives  more   information  about  how  different  input  parameters  affect  each  other.  Secondly,  they  often  think  that   the  simulation  model  can  predict  the  future,  i.e.  predict  how  the  input  parameters  will  change  in  the   future,   which   is   not   possible   to   do   with   the   model   either.   What   is   possible   though   is   to   do   a   sensitivity  analysis  of   the   input  parameters   to  provide   information  about  how  much  a   change  of  a   variable  affects  the  system  and  this  can  be  used  while  deciding  what  parameters  that  should  be  given   extra  attention.  To  avoid  misunderstandings  it  is  important  to  explain  early  in  the  simulation  project   what  a  simulation  can  and  cannot  do.  A  simulation  project  can  also  help  the  decision  making  process   in  other  ways  then  just  by  the  results  from  the  model.  It  allows  all  the  persons  involved  in  the  project   to  understand  how  the  whole  system  works  and  therefore  prevents  disagreements  over  assumptions   that  do  not  correspond  to  how  the  system  actually  works.  Due  to  this   it   is  not  always  necessary  to   perform  a  full-­‐scale  simulation  project  but  it  can  be  enough  to  only  do  some  steps  (ibid).   Factors   commonly   used   to   measure   performance   of   a   health   care   system   are   utility   for   different   resources   like   personnel,   rooms,   equipment   and   machines   (Brenner,   2010).   Other   measures   are   patients’   total   time,   also   called   throughput   time,   and   number   of   waiting   patients   in   queues   (Thorwarth,  2009).  Performance  can   in  some  cases  be  calculated   into  costs,  where  different  events   can  increase  or  decrease  money  earned  per  patient  (Swisher,  2002).   2.3 Theory  of  constraints   The   purpose   of   the   theory   of   constraints   is   to   improve   system   performance   through   change   (Dettmer,  1997).  To  accomplish  this,  three  questions  have  to  be  answered:   • What  to  change?  (What  is  constraining  the  system?)   • What  to  change  to?  (How  should  we  change  it?)   • How  to  cause  the  change?  (How  should  it  be  implemented?)   When  answering  these  questions  several  principles  should  be  considered  (ibid):   • Systems  as  Chains   –  A   chain   can  never  be   stronger   than   its  weakest   link   and  nether   can  a   system.   • Local  Optima  vs.  System  Optimum  –  The  sum  of  all  local  optima  is  not  the  same  as  the  whole   system’s  optimum.  If  all  resources  in  a  system  are  performing  at   its  maximum  capacity,  the   whole  system  is  still  not  performing  as  good  as  possible.   • Cause  and  Effect  –  An  event  at  one  place  in  a  system  can  cause  effects  in  another  place  in  the   system.   • Undesirable  Effects  and  Core  Problems  –  Almost  all   issues  seen  in  a  system  are  not  the  real   problems  but  merely  the  problems’  undesirable  effects.  It  can  seem  like  fixing  an  undesirable   effect  will   solve   the  problem  but  unless   the   core  problem   is   fixed   there   is   a   large   risk   that   new  undesirable  effect  will  arise  instead.   • Solution  Deterioration  –  No  matter  how  good  a  solution  seems  to  be  when  implemented,  it   will   still   deteriorate   over   time.   To   maintain   a   good   efficiency   of   a   solution   continuous   9       improvements   are   necessary.   When   improving   a   system   it   can   have   some   inertia,   even   though  a  problem  already  has  been  solved  once  it  does  not  mean  it  cannot  come  back.   • Physical   Constraints   vs.   Policy   Constraints   –   Physical   constraints   are   often   rather   easy   to   identify  and  adjust  but  most   constraints  are  policy   constraints.  Policy   constraints  are  often   complex  and  hard   to  both   identify  and  adjust  but  when  adjusted   it  generally   improves   the   system  to  a  much  larger  extent  than  a  physical  constraint.   • Ideas  are  not  solutions  –  An  idea  cannot  solve  any  problems  unless  it  is  implemented.   To  make   the  process  of   continuous   improvement  easier   and   to   get   a  positive  effect   in   the   system   performance  from  every  improvement  five  sequential  steps  have  been  developed  (ibid):   1. Identify  the  system  constraint   Where  is  the  weakest  link  in  the  system?  Is  it  physical  or  is  it  policy?   2. Decide  how  to  exploit  the  constraint   How  to  get  the  most  capacity  out  of  the  constraint  without  expensive  changes?   3. Subordinate  everything  else   Tune  the  rest  of  the  system  to  allow  the  constraint  to  operate  at  its  maximum  capacity.  Then   check  if  the  constraint  still  is  constraining  the  system.  If  not,  continue  to  step  five  otherwise   go  to  step  four.   4. Elevate  the  constraint   Take  whatever  actions  needed  to  break  the  constraint.  This  involves  major  changes  to  the   system  that  can  require  large  investments  in  money,  time  and  energy.   5. Go  back  to  the  first  step,  but  watch  out  for  inertia   The  cycle  never  ends;  the  system  always  has  a  constraint.  It  can  either  be  a  completely  new   constraint  or  an  already  broken  constraint  that  has  reoccurred.   2.4 Improvement  strategy  for  health  care   There   is  a  Swedish   improvement  strategy  called  Genombrott   (Landstingsförbundet,  1998),  which   is   built  on  a  model  called  Breakthrough  Series  developed  by  The  Institute  of  Health  Care  Improvement   (IHI)  in  Boston,  USA.  The  idea  of  this  improvement  model   is  that  there  is  knowledge  that  is  not  administered  enough   within  the  health  care  today.  The  participants  should  learn   what  changes  will   lead  to  an   improvement  of   the  system   by   systematically   testing   small   changes   and   document   their   effect  on   the   system.  There  are   four  headstones   to   follow   then   using   the   Genombrott   model.   First   is   Tom   Nolan’s  method   for   improvement,   the   Plan-­‐Do-­‐Study-­‐Act   cycle,   see   figure   2.1.   Second   is   a   structured   working   process,   which   structures   both   time   and   content   of   the   project,  that  should  be  followed.  Third  is  that  there  should   be   people   and   organizations   with   different   experience   working  towards  a  common  goal,  allowing  an  exchange  of   knowledge,  within  the  project  group.  The  fourth  and  most   important   headstone   is   the   pressure   and   will   of   change   within   the  organization,  which   is   something   the  manager   of  the  organization  will  have  to  accomplish  (ibid).   Figure   2.1:   The   Plan   -­‐   Do   -­‐   Study   -­‐   Act   cycle   developed   by   Tom   Nolan   which   is   used   in   the   Genombrott   model.   10       The  work  towards  improvement  should  start  with  establishing  the  goals.  They  should  be  challenging   and  clear  to  give  pressure  to  change  and  to  make  it  clear  that  it  will  not  be  possible  to  reach  these   goals  with   the   current  work  method.  By   clear,   it  means   that   the  goals   should  be  numeric   i.e.   they   should   be  measurable   in   numbers.   To   be   able   to   know   if   a   change   leads   to   an   improvement   it   is   important  to  know  the  value  of  the  goal  variables  before  the  change  was  implemented  and  the  value   after.   It   is  also  important  to  have  some  balancing  measures  from  other  parts  of  the  organization  to   make  sure   that   the  change  will  not  deteriorate   something  else,   the  problem  should  not  be  moved   from  one  part   of   the   organization   to   another.   All   changes   that   are   to   be   tested   should   follow   the   PDSA-­‐cycle   (Plan-­‐Do-­‐Study-­‐Act)   they  should  thus   first  be   implemented   in  a  small   scale  and  then  be   evaluated  before  it  is  decided  whether  they  should  be  implemented  in  the  whole  organization  or  not.   To   get   the   process   of   change   started   there   are   several   changing   concepts   to   get   ideas   from.   The   concepts   are   divided   into   three   strategies:   “Match   capacity   to   demand”,   “Influence   demand”   and   “Change  the  system”,  and  can  be  seen  in  table  2.1  (ibid).     11           Table   2.1:   The   changing   concepts   from   Genombrott.   They   are   divided   into   three   different   categories:  Match  capacity  to  demand,  Influence  demand  and  Change  the  system.  The  changing   concepts  are  supposed  to  help  to  get  the  process  of  ideas  of  changes  started.   Match  capacity  to  demand Influence  demand Change  the  system Predict  demand  with  higher  accuracy Confine  material-­‐  and  instrument   selection Do  several  work  tasks  in  parallel Level  off  variations  in  demand Watchful  waiting  may  reduce  demand Elaborate  and  use  care  programs  for   common  problems Adapt  manning  to  predictable  needs Coordinate  patient  visits Minimize  the  number  of  elements  and   involved  persons  in  the  process Identify  and  remove  bottlenecks Standardize,  establish  routines  and  care   programs  for  common  problems   Synchronize  different instances Work  up  accumulated  queue Triage-­‐  correct  evaluations and  handling  from  the beginning "Attract  patients"  instead  of  that   previous  link  are  "pushing  patients   away" Use  existing  resources  flexibly Practice  knowledge  based  medicine Reduce  distance  between  different   functions  in  the  process  to  ease   communication Keep  alternative  options  of  action   prepared  if  the  planning  bursts Move  demand  on  a  care  service  to   another  care  instance Automate  when  possible Preempt  demand  by  fulfilling  the  need   before  it  arises Work  up  understanding  for  the  system   to  make  persons  within  the  organization   work  towards  the  same  goal Promote  self-­‐care  and  engage  the   patient  more  in  its  own  care  and   treatment Create  more  work  stations,   plan  premises  and  equipment  so  they   can  be  used  flexibly Use  persons  with  specialist  knowledge   to  what  they  are  good  at Move  tasks  in  the  process  that  could  be   better  executed  somewhere  else  or  by   someone  else Strategy Ch an gi ng  co nc ep t 12           13       3 Method   There  are  several  different  approaches  that  could  have  been  used  to  analyse  the  health  care  centre   system.  One  such  approach   is  queuing  theory,  which   is   fairly  easy  to  use  and  can  predict  how  well   improvements  will  work  out.  However,   it   cannot   forecast   short   term  crowding.  Another  method   is   regression-­‐based  analysis,  which  major  advantage  is  its  ability  to  predict  short  term  crowding  and  it  is   easy  to  use.  A  major  drawback  is  though  that   it  cannot  be  used  to  analyse  improvements  very  well   (Wiler,  2011).  A  health  care  centre   is  a  complex  system  with  many  different  paths   for  patients  and   every   patient   traveling   through   the   system   is   unique.   This   makes   it   unsuitable   to   use   a   simple   evaluation  method  such  as  linear  analyse  with  an  Excel  spread  sheet,  which  is  more  suitable  for  less   complex   systems   with   fewer   paths.   Instead   agent   based   simulation   (Siebers,   2010)   can   be   an   alternative   which   focuses   on   modelling   the   entities   and   their   interactions   in   the   system,   i.e.   the   patients,   personnel   and   rooms.   This   seems   applicable   since   almost   all   entities   in   the   system   are   unique  but  there  are  no  textbooks,  methodologies  or  frameworks  available  for  this  method  (Siebers,   2010).   Thus,   it   is   hard   to   use   for   a   novice   and   it   was   therefore   not   used   in   this   project.   Instead   Discrete   Event   Simulation   (DES)   was   used,   which   is   a   well-­‐known   method   and   there   is   a   lot   of   available  information.  This  simulation  method  focuses  on  modelling  a  system’s  dynamics  and  not  so   much  its  entities  (Siebers,  2010).  DES  simulation  is  good  at  predicting  how  well  improvements  affect   a   system  and  can  also  predict   short   term  crowding.  The  drawback  with  DES   is   that  a   lot  of   special   knowledge  is  required  to  develop  and  use  a  model  (Wiler,  2011).   A  DES-­‐project  structure  by  Banks,  from  now  on  referred  to  as  Banks’  model  (Banks,  2010),  has  served   as  the  main  structure  of  the  DES-­‐project.  Banks’  model  has  been  a  good  support  for  the  project  as  it   reflects   a   natural   way   to   perform   a   DES-­‐project.   There   is   another   structure   for   DES-­‐projects   by   Kelton,  which  is  similar  to  Banks’  model  but  it  is  not  as  detailed  and  there  is  no  data  collection  step   (Kelton,  2007).  Banks’  model  does  therefore  give  better  support  for  this  form  of  project.  Figure  3.1   shows   the   relationship   between   Banks’   model,   the   evaluation   of   the   input   data   management   methodology   and   the   case   study   at   the   health   care   centre.  While   all   steps   but   the   last   of   Banks’   model   has   been   performed   the   focus   of   the   case   study   has   mainly   been   to   model   the   current   situation  at  the  health  care  centre,  analyse  it,  create  concepts  and  evaluate  the  concepts  to  provide   decision  support  for  the  health  care  centre  director.   14         3.1 Input  Data  Management  Methodology   The  input  data  management  is  built  on  13  steps  and  is  created  to  assure  that  no  important  activities   towards  having  sufficient  and  representative  input  data  are  forgotten.  The  methodology  is  believed   to  give  great  improvements  to  DES-­‐projects,  especially  for  project  groups  with  little  experience.  This   methodology   is   mainly   aimed   to   DES-­‐projects   in   industry   and   can   be   seen   in   figure   3.2,   (Skoogh,   2008).     Identify  and  Define  Relevant  Parameters:  In  the  first  step  the  system  needs  to  be  examined  in  order   to   determine   all   parameters   required   in   the   model.   An   extensive   understanding   of   the   system   is   important   because   this   task   is   not   as   simple   as   it   might   seem.   The   level   of   detail   should   not   be   Figure  3.1:  Relationship  between  the  different  parts  of  the  DES-­‐project.  The  DES-­‐project  structure   by  Banks   is   shown   in   the  middle   in   light   grey  boxes.   To   the   left   are   the  main  steps  of   the   case   study  and  to  the  right  is  the  evaluation  of  the  input  data  management  methodology.   15       greater  than  stated  by  the  problem  definition  and  what  is  required  for  the  project.  How  parameters   will  be  measured  and  used  in  the  model  also  have  to  be  defined  to  avoid  confusion  later.     Specify   Accuracy   Requirements:   The   second   step   includes   defining   how   accurate   the   data   parameters  have  to  be.  More  data  points  gives  better  accuracy  but  to  be  efficient  it  is  important  to   identify  how  important  the  different  parameters  are  compared  to  each  other.  Parameters  describing   parts   of   a   system   that   are   constraining   other   parts   are  more   important   to   focus   on,   as   they   have   greater  influence  on  the  output.   Identify   Available   Data:   Data   collection   is   generally  one  of  the  most  time  consuming  steps   of  a  DES-­‐project  and  a  lot  of  time  can  be  saved   if   there   is   previously   gathered   data   which   can   be  used.  This  step  includes  identification  of  such   data,   how   and   where   it   can   be   accessed.   This   should  result  in  a  list  of  available  data  and  how   it  can  be  retrieved.     Choose   Methods   for   Gathering   of   Not   Available   Data:   Suitable   gathering   methods   need   to  be   chosen   for   data   that   is   not   already   available.  There  are  two  types  of  data,  the  first   can   be   collected   and   the   second   cannot   be   collected.   The   first   type   is   the   most   time   consuming  to  collect,  as  this  type  of  data  often   requires   a   lot   of   manual   work.   Therefore   the   most  common  method  is  to  manually  clock  the   processes   with   a   stopwatch.   Some   data   like   cycle  times  might  be  collectable  from  e.g.  code   for   PLC   or   NC   machines.   This   is   though   not   possible   if   the   system   does   not   exist   or   if   no   data   can   be   collected,   instead   estimates   will   have   to   suffice.  Valid  estimates   can  be   created   by  discussions  with  system  experts  and  machine   vendors   or   by   examining   similar   systems   or   historical   data.   Care   should   be   taken   when   gathering  data  by  clocking  working  humans  due   to   several   reasons.  Workers   can   be   influenced   by  knowing  that  someone   is  clocking  them  and   cause   the  Hawthorne  effect,   i.e.   they  will  work   in   another   way   than   they   usually   do.   Humans   are  also  more  unpredictable  than  a  mechanised   system  but  may   still   be  described  by   statistical   distributions.   Figure  3.2:  Input  data  management  methodology   (Skoogh,  2008).   16       Will  All  Specified  Data  Be  Found?:  If  the  previous  steps  result  in  that  all  required  data  can  be  found   and   at   a   sufficient   accuracy,   next   step   can   be   started.   If   not,   the   previous   steps   have   to   be   re-­‐ evaluated.  There  is  always  a  risk  that  accuracy  requirements  have  to  be  re-­‐evaluated  if  it  later  turns   out  that  some  data  is  not  measureable  after  all.   Create  Data  Sheet:  All  data,  both  raw  and  analysed,  should  be  stored  in  a  spread  sheet  or  database,   common   to  all   project  members.  No  data   should  be   stored   in  model   interfaces  or   temporary  data   sheets.  Data  risk  to  be  deleted  and  a  lot  of  data  analysing  risk  having  to  be  redone  if  not  stored  in  a   specific  location.   Compile  Available  Data:  Already  available  data  should  be  collected  here.  Some  data  may  already  be   ready   for   validation  while   the   rest   needs   to   be   filtered   and   analysed  before   it   can  be   validated.   If   everything   is   in   order   the   required   amount   of   data   points   specified   earlier   can   be   extracted.   Collectable  data  does  often  need  to  be  calculated  to  a  desired  value,  e.g.  calculating  processing  times   from   timestamps,   and   filtered   from   faulty   or   non-­‐representable   data   points.   When   this   step   is   completed  the  result  should  be  a  list  of  data  points  ready  for  calculating  a  suitable  representation.   Gather  Not  Available  Data:  This  step  aims  to  convert  not  available  data  into  available  data  ready  for   calculating  data  representations.  Collectable  data  should  be  collected  as  decided  earlier   in  “Choose   Methods  for  Gathering  of  Not  Available  Data”  and  non-­‐collectable  data  should  be  estimated.  There  is   a  high  risk  that  this  step  is  the  most  time  consuming  step,  especially   if  many  parameters  are  not  of   the   available   data   category   and/or   if   cycle   times   are   long   as   more   than   200   data   points   are   preferable.   Prepare   Statistical   or   Empirical   Representation:   In   this   step   non-­‐constant   data   need   to   be   represented   in   some   way.   There   are   basically   four   different   options:   statistical   distributions,   empirical  distributions,   traces  or  bootstrapping.  While   the   last   three  can  be  calculated   fairly  easily,   the   first   is  more  difficult   to   calculate.   There  are  however   tools   like  ExpertFit  or   Stat::Fit  which   can   calculate  and  analyse  statistical  distributions.   Sufficient  Representations?:  This  is  a  difficult  step  where  the  representations  from  the  previous  step   have  to  be  confirmed  to  be  good  enough.  A  goodness-­‐of-­‐fit  test  at  the  0.05-­‐level  can  be  applied  but  it   is  hard  for  a  representation  to  pass.  Therefore  the  level  can  be  set  differently,  reflecting  the  accuracy   requirements.  A  graphical  test  can  be  performed  between  the  original  data  and  the  representation  to   see  how  well  the  representations  correspond  to  the  data.   If  the  representation  is  not  sufficient  the   data   collection   and   analyse   will   have   to   be   complemented.   If   there   is   no   way   to   reach   the   requirements,  the  requirements  may  have  to  be  re-­‐evaluated.   Validate   Data   Representations:   This   step   is   difficult   and   consists   of   validating   that   collection,   calculations,  analyses  and  filtering  of  raw  data  have  been  done  correctly.  This  is  due  to  that  the  same   data   is  used   for  validation.  Data  can  however  be  validated  by  production   follow-­‐ups,  ensuring   face   validity  and  following  good  routines  during  the  whole  input  data  management.  Even  though  this  step   is  difficult  a  good  data  validation  will  minimise  the  risk  for  later  iterations  of  previous  steps.  Naturally   this  step  is  more  important  for  parameters  with  high  accuracy  requirements.   Validated?:   If  all  data   representations  pass   the  validation  checks   they  are  now  ready   to  use   in   the   model.  It  is  important  to  remember  that  data  can  still  be  the  problem  if  a  model  validation  fails,  due   17       to  the  difficulty  of  the  previous  step.  Results  from  previous  steps  need  to  be  re-­‐examined  if  this  step   shows  that  some  representations  do  not  fulfil  the  requirements.  However,  this  is  often  a  result  from   errors  in  the  preparation  and  analysis  of  raw  data.   Finish  Final  Documentation:  Besides  from  the  data  sheet  there  are  other  things  that  do  not  fit   in  a   data   sheet   that  have   to  be  documented.  These   things   include  everything   from  assumptions  during   data   collection,   definitions   of   data,   how  data  was   validated   and   results   from   validation   steps.   The   data   sheet   should   be   as   good   as   complete   by   now  as   documentation   should   be  written   during   all   steps  of  the  methodology.  This  step  therefore  results  in  a  complete  data  sheet  and  a  data  report  that   should  end  up  in  the  simulation  project  documentation.   3.2 Evaluation  of  input  data  management   methodology   The   input   data   management   methodology   by   Anders   Skoogh   has   been   applied  for  all  input  data  for  a  simulation  model  of  the  health  care  centre.   Field   notes   have   been   used   during   the   implementation   to   capture   different   aspects   of   the   methodology.   Field   notes   can,   according   to   Mulhall,   be   used   to   record   all   sorts   of   impressions   during   a   field   study.   Notes   can   range   from   how   people   behave,   activities,   processes   and   dialogues   to   special   events,   building   layouts   and   personal   reflections   (Mulhall,   2003).   Field   notes   have   in   this   study   been   used   to   capture   thoughts  and  aspects   that  have  arisen  during   the   implementation.  Some   thoughts  have  arisen  during  the  implementation  while  others  have  arisen   afterwards   when   the   bigger   picture   is   more   distinguishable.   Using   field   notes  makes  it  easier  to  capture  brief  passing  thoughts.  Another  method   that  could  have  been  used  is  memoing.  However,  a  memo  is  usually  more   extensive   than   a   field   note   and   it   is   harder   to   find   time   to  write   them.   Originally,   some   unstructured   interviews   were   planned   but   capturing   dialogues  with   field  notes   resulted   in  sufficient   information  such  that  no   unstructured  interviews  were  needed.  This  also  worked  well  for  capturing   information  about  the  health  care  centre  organisation.   In   order   to   analyse   the   field   notes   has   inspiration   been   taken   from   a   method  called  KJ-­‐Shiba  (Ulrich,  2003).   It   is  a  method  designed  to  analyse   and  understand  problems  and  consists  of  several  steps,  how  many  varies   with  each  version  of  the  method.  With  the  KJ-­‐Shiba  method,  facts  about  a   problem   are  written   on   self-­‐sticking   notes.   All   notes   are   grouped,   those   groups   are   then   in   turn   named   with   headers,   which   are   grouped   and   voted   on   so   that   the   most   important   headers   stand   out.   Major   conclusions   can   then   be   drawn   from   the   important   headers.   For   this   study,   step   four   to   seven,   marked   with   green   in   figure   3.3,   have   been   used.   The   field   notes   have   served   as   input   to   step   four   instead   of   facts   about   a   problem   from   step   three.   This   method   is   far   less   ceremonious   than   performing   a   grounded   analysis   which   would   be   all   too   time   consuming.  Due  to  previous  knowledge  of  the  methodology  it  would  also   not   be   possible   to   perform   a   correct   analysis   with   grounded   theory   Figure   3.3:   The   eight   steps   from   the   used   version   of   the   KJ-­‐Shiba   method.   18       (Gelling,  2011).  The  KJ-­‐Shiba  method  has  the  drawback  however  that  important  details  might  get  lost   in   the  process   as   it   generalises   a   lot.  At   the   same   can   this   be   an   advantage;   it   reduces   the   risk   of   being  drowned  with  details.   3.3 Improvement  of  the  health  care  central   The  health  care  centre  was  simulated  with  a  discrete  event  simulation  model  and  almost  all  data  for   it   was   gathered  manually   by   clocking.   The   resulting  model   was   verified   and   validated   graphically,   with  output  data  and  by  health  care  centre  personnel.  Simulated  concepts  were  then  compared  to   the  model  of  the  current  situation  and  to  each  other.   3.3.1 Discrete  Event  Simulation   It   is  most  common  to  simulate   industry  processes  with  DES-­‐simulation.  One  way  to  model  systems   with  this  method  is  to  model  products  as  loads.  Logic  code  is  written  to  build  the  model  and  a  row  is   executed  when  a   load  reaches   it.  That   is,  each   load  runs  through  the  code  and  depending  on  what   load  type  and  attributes  carried  by  the  load,  each  load  will  run  different  paths  in  the  code.  A  load  can   be  told  to  wait  for  other  events  or  move  into  queues.  Machines  and  other  resources,  like  personnel,   are  modelled  as  so  called  resources.  A   load  can  use  resources  and  make  them  unavailable  to  other   loads,   which   then   have   to   wait.   Patients   can   be   modelled   as   loads   while   personnel,   rooms   and   equipment   can   be  modelled   as   resources   in   a  model   of   a   health   care   application.   Some  problems   arise  when  a  health  care  system  is  going  to  be  modelled  in  this  way  that  seldom  arises  in  an  industrial   case.     The  software  that  has  been  used  to  create  the  simulation  model  is  called  Automod  and  it  was  chosen   due   to   the   fact   that   it   is   ”load   driven”   simulation   software.   This   was   suitable   in   this   specific   application   since   the   patients   can   be   modelled   as   loads.   Also,   two   of   Automod’s   strengths   are   detailed  models  and  operational  decision  support  which  both  are  important  in  this  application.   3.3.2 Input  data  management   All   input   data   regarding   times   have   been   collected   manually   with   stopwatches   by   following   staff   members   while   they   were   working   during   three   weeks.   Times   have   been   clocked   from   the   physicians,  the  district  nurses,  in  the  lab  from  assistant  nurses  and  from  the  reception  desk.  Statistics   of   how  many   patients   that   need   to   undress   etc.   have   been   deduced   from   the   time  measurement   data,   while   statistics   of   paths   patients   follow   through   the   system   have   been   taken   from   patient   journals.  The  schedule  planner  could  provide  personnel  schedules  but  the  number  of  patients  in  the   waiting   room  had   to  be  counted  manually  each  15  minutes.  Note   that  all  data  were  not  meant  as   input  to  the  model  but  as  validation  data  in  order  to  validate  the  model.  Table  3.1  and  3.2  show  how   different  input  data  were  collected  along  with  their  definitions  and  number  of  collected  data  points.   The  number  of  desired  data  points  is  calculated  with  respect  to  the  available  time  for  data  collection,   how  many  that  can  be  gathered  per  day,  how  important  the  parameter  is  and  how  large  the  variation   of   the   parameter   is.   In   an   ideal   case   this   number   would   be   higher.   The   input   data   were   then   calculated  with  ExpertFit   into  statistical  distributions.  Quality  of   the  distributions  was  checked  with   ExpertFit’s   assessments   of   the   distributions   and   by   comparing   original   data   with   the   distributions   graphically.     19           Ta bl e   3. 1:  D at a   sh ee t  o ve r  g at he re d   da ta ,  w he re  th e   ga th er ed  d at a   co m es  fr om ,  h ow  it  is  g at he re d,  it s  d ef in iti on  a nd  fr om  w ha t  p at ie nt  ty pe .   Th e   nu m be r  o f  d es ire d   da ta  is  c al cu la te d   w ith  re sp ec t  t o   th e   av ai la bl e   tim e   fo r  d at a   co lle ct io n.  Im po rt an t  d at a   ha s   a   de sir ed  n um be r  o f  d at a   po in ts  o f  4 0-­‐ 50 ,  l es s   cr uc ia l  d at a   on ly  re qu ire s  2 0-­‐ 30 .   20         Ta bl e   3. 2:  D at a   sh ee t   ov er  g at he re d   da ta ,  c on tin ue d,  w he re  t he  g at he re d   da ta  c om es  fr om ,  h ow  it  is  g at he re d,  it s   de fin iti on  a nd  fr om  w ha t   pa tie nt  t yp e.  T he  n um be r   of  d es ire d   da ta  is  c al cu la te d   w ith  r es pe ct  t o   th e   av ai la bl e   tim e   fo r   da ta  c ol le ct io n.  Im po rt an t   da ta  h as  a  d es ire d   nu m be r  o f  d at a   po in ts  o f  4 0-­‐ 50 ,  l es s   cr uc ia l  d at a   on ly  re qu ire s  2 0-­‐ 30 .   21       3.3.3 Verification  and  validation  of  models   The  model   of   the   health   care   centre   has   been   verified   by   analysing   output   data.  When   an   output   data  point  has  been  unreasonably  far  away  from  other  data,  the  source  of  this  data  has  been  tracked   down   and   corrected.   The   model   has   also   been   verified   graphically.   The   validation   has   been   conducted   by   comparing   output   data   with   data   from   the   real   system.   The   health   care   centre’s   director  has  also  been  introduced  to  the  model  and  agreed  to  that  it  works  as  intended  and  that  the   output  data  are  reasonable.   3.3.4 Triage  goals  and  evaluation  of  concepts   The  health  care  centre  has   four  goals   for   the  triage,  which  should  be  reached  all  average  days,   i.e.   when  no  special  event  occurs.  Such  a  special  event  could  be  the  arrival  of  a  patient  who  requires  a  lot   more  effort  and  maybe  ought  to  have  been  sent  to  a  hospital  emergency  ward  instead.     • The  triage  should  end  at  12:30,  i.e.  the  last  triage  patient  should  not  leave  later  than  12:30.   • There  should  be  no  need  for  a  triage  back-­‐up.   • Ideally  there  should  be  no  more  than  10  patients  in  the  waiting  room  at  the  same  time,  but   the  absolute  maximum  is  15  patients  in  the  waiting  room  at  the  same  time.   • The  triage  should  be  handled  with  the  same  manning  as  today,   i.e.  two  physicians  and  two   district  nurses.   The   goals   have   been   given   different   priorities   with   respect   to   the   health   care   centre’s   needs   and   preferences.  The  highest  prioritised  goals  are  that  the  triage  should  end  at  12:30,  together  with  the   no  need  for  triage  back-­‐up  goal.  After  these  two  is  the  number  of  patients  in  the  waiting  room  goal.   The   lowest   prioritised   goal   is   that   the   other   goals   should   be   reached   with   the   same  manning   as   today.   Along  with  these  goals,  the  throughput  time  for  patients  and  utility  of  physicians  and  district  nurses   have  been  analysed,  i.e.  the  following  parameters  have  been  analysed:   • When  the  last  triage  patient  leaves.   • How  many  triage  patients  the  triage  back-­‐up  physician  handles.   • How  many  patients  there  are  in  the  waiting  room.   • Utility  for  physicians  and  district  nurses.   • Patient  throughput  time.   The   concept   evaluation   has   been   inspired   by   systematic   construction   (Olsson,   1976)   and   three   different  methods  have  been  used.   First   a  method  called   rating   criteria  method   (ibid)  was  used   to   evaluate  how  well  the  concepts  fulfil  the  triage  goals.  The  fulfilment  of  each  goal  have  been  marked   with  yes,  no  or  almost  for  all  concepts  and  the  priorities  of  the  goals  have  been  taken  into  account.   To   complement   this   analysis,   two  methods   based   on   comments   have   been   used.   First   out   was   a   motivation  method  (ibid)  where  arguments  are  given  to  why  a  concept  have  been  or  have  not  been   chosen.   Here   the   throughput   time   and   the   utility   have   been   taken   into   account.   Secondly,   an   advantage/disadvantage  method  (ibid)  has  been  used  to  take  other  aspects,  such  as  social  aspects,   into  account.     22           23       4 Current  system   Here   is   a   description   of   how   the   health   care   centre   is   organised   today   with   schedule,   personnel,   building   layout  and  patients.  There  are  also  simulation   results  and  analysis  of   the  current  situation   here  together  with  comments  and  opinions  from  the  personnel.   4.1 Description   The  current  organisation  at   the  health  care  centre,  only  parts   included   in   the  model  are  described   thoroughly.   4.1.1 Schedule   Every  day  there  is  a  triage  reception,  which  is  a  way  to  categorise  patients  depending  on  how  crucial   their  injuries  or  illnesses  are.  At  Sörhaga,  a  simplified  form  of  triage  is  implemented  by  having  district   nurses   assess  whether   a   patient   has   to   see   a   physician   or   not.   The   triage   begins   at   9:45   and  was   originally  planned  to  always  end  at  12:00.  Nowadays  it   is  unofficially  accepted  that  it  should  end  at   12:30.  That  is,  the  last  patient  should  leave  at  12:30.  Two  district  nurses  starts  the  triage  at  9:45  and   two  physicians   (three  on  Mondays)   starts  working  at   the   triage  at  10:00.  There   is   a   triage  back-­‐up   scheduled  from  12:30  –  13:00  but  this  physician  often  has  to  stay  longer  than  13:00  as  there  are  still   triage   patients   who   have   not   seen   a   physician   yet.   This   happens   even   as   the   two   ordinary   triage   physicians   helps   the   back-­‐up   by   continuing   past   12:30   if   there   are   too   many   patients   left.   The   afternoon  emergency  visits  are  taken  care  of  between  15:00  and  16:00  and  each  visit  are  planned  to   take  either  15  or  30  minutes.  All   other   types  of   visits   to  physicians   can  be  booked  anytime  during   office   hours   depending   on   each   physician’s   personal   schedule.   The   lab   has   booked   appointments   during   the   day   except   between   10:00   to   12:00.   Those   visits   are   booked   in   the   sense   that   the   lab   knows  they  will  come  but  the  patients  still  has  to  pick  a  queue  ticket   for   lab  when  they  have  been   registered.  All  district  nurses  take  phone  calls  until  9:30  every  day  but  after  this  they  can  have  patient   visits   any   time   during   office   hours,   unless   they   are   scheduled   for   the   triage,   depending   on   each   district   nurse’s   schedule.   There   is   always   at   least   one   district   nurse   scheduled   to   take   phone   calls   during  the  day.  A  medical  secretary  always  keeps  the  reception  desk  open  from  7:45  to  16:45.   4.1.2 Personnel   The  health   care   centre   has   in   total   ten   employed  physicians,   six   of   them  are   general   practitioners   while  the  rest  are  resident  physicians.  There  are  also  eight  district  nurses,  four  assistant  nurses  and   four   medical   secretaries.   The   physicians   have   five   main   activities:   administration,   triage,   semi-­‐ emergency   visits,   return   visits   and   afternoon   emergency   visits.   There   are   also   other   activities   like   geriatric   care   located   elsewhere   etc.   The   administration   activity   includes   everything   that   is   not   related   to  a  patient   coming   soon  or   that  has   just   left,   like  calling  patients,  managing  prescriptions,   approving   dictate   transcriptions   and   more.   District   nurses’   tasks   include   among   other   tasks;   answering  phone  calls,  especially  in  the  mornings,  triage,  bandaging  and  measuring  blood  pressure.   Assistant  nurses  are  responsible  for  the  lab,  all  tests  and  analyses.  They  should  also  refill  materiel  in   all   rooms   once   a   week   and   help   physicians   and   district   nurses   when   they   need   assistance.   The   medical   secretaries   are   responsible   for   transcribing   dictates   from   physicians,   the   reception   desk,   scanning  mail,  billing  patients  and  more.  The  secretary  responsible  for  the  reception  desk  transcribes   dictates  when  there  is  no  patient  at  the  reception  desk.   24       4.1.3 Building  layout   The  health  care  centre  building  is  a  part  of  a  larger  building  and  is  connected  to  a  hospital.  There  are   in   total   16   examination/administration   rooms,   physicians   occupy   ten,   district   nurses   occupy   six   rooms.   Each   physician   does   therefore   have   a   room   of   their   own   room,   as   do  most   of   the   district   nurses.   The   lab   room   can  manage   three   patients   at   the   same   time.   Figure   4.1   shows   a   simplified   layout  of  the  health  care  centre,  physicians’  rooms  are  marked  with  a  P,  district  nurses’  rooms  with   an  N,  the  lab  with  L  and  the  reception  is  marked  with  an  R.     The  waiting  room  originally  had  seats  for  nine  patients  but  as  the  waiting  room  gets  crowded  during   the  triage  a  number  of  chairs  have  been  placed  there.  Today  there  are  about  13-­‐15  extra  seats  with   the  additional  chairs,  resulting  in  a  total  of  around  25  seats.  All  of  those  seats  are  not  used  however,   the  original  nine  seats  are  three  three-­‐person  sofas  but  there  are  mostly  only  two  persons  sitting  in   each   sofa.   The  highest   number  of   patients   in   the  waiting   room  has   been   counted   to   25   and   since   many   patients   bring   relatives   everyone   will   not   get   a   seat.   However,   this   has   only   been   counted   during   one   week.   It   has   been   estimated   afterwards   that   at   least   30-­‐40   %   of   the   patients   h