Page 1 
  Practical 
  d'écriture 21st 
  century 
  Interview with Anne-Laure, statistician at the Inami 
  Page 2 
  Page 3 
  'Chaque Graph is like a novel for me ' 
  she said one day. 
  'Faire Graphics, it's way 
  d'apercevoir worldwide.  Through'norme' it 
  looking for patterns, and it is verified that deviates, 
  the reasons we seek.  Sometimes there are reasons 
  which are entirely legitimate, sometimes they are 
  more questionable ... For this to be done well, it 
  would have all the ground is covered and the 
  jurisdiction explicit, but the reality is 
  always a little more complicated than the 
  BOM! ' 
  Interview with Anne-Laure, Statistician. 
  Directed by Constant An Mertens, Brussels, June 
  2013. 
  1 
  Page 4 
  2 
  Page 5 
  Some kinds of novels in the analysis 
  data 
  The detective story 
  Anne-Laure: A colleague example examines 
  providers who bill the Global Medical Record. 
  Normally, you can not charge if you saw the 
  patient, and more than once a year.  For 
  provider, we see that in his billing invoice 
  for patients he has never seen, and it has a 
  systematic in billing, but it is not known 
  which yet.  Often our work is this: it 
  looking at how the crime was committed.  We play 
  riddles in the data, we look for the life that is 
  behind.  It's like when you trace a 
  Crime scene: data track and you 
  looking to reconstruct the story.  You play cat and 
  Mice with the data, it is fun, you're in 
  game. It's fun to do, I love the novels 
  police, and has that side'I'll coincer'.  We play 
  smarter with the provider. 
  In fact we always try to decode the data, 
  find in the data logic Due fraud, and if 
  was something wrong, you can go check on 
  field, for patients, for example. 
  The novel Genesis 
  Anne-Laure: In data analysis more 
  complex, the goal is to reopen the data 
  3 
  Page 6 
  give them meaning.  I prefer because the story 
  is a little more complicated: you can compare it to 
  when someone dies and where you keep the 
  memory of his presence.  The data is somewhat 
  like life the memory of someone.  From the 
  Data you can réévoquer, reinvoke the history of 
  no one.  That leaves more room for the interpretation 
  the method of "whodunit" is what you recreated 
  richer, more complex. 
  We made ​​an analysis, for example, in the 
  Gastroenterology, in this case and we look 
  very openly if there are weird things.  To 
  get there you are forced to rebuild everything.  The 
  Doctors who specialize in oncology, for 
  example, not all the same practice as 
  not oncologist.  For one, it is quite 
  normal to gastroscopy month to some 
  patients, while for others, rather it is expected 
  gastroscopies repeated more than once a year. 
  It's difficult to understand the rules of a 
  field when you are outside this area.  He has a lot to 
  things in billing for which you do not know 
  if it's normal or not.  You always need 
  the story around, context, a doctor who has a 
  view of the field, which is the translation. 
  Another example was the question of how 
  can reduce expenses on medicines with 
  the special permission of the medical officer.  He had a will 
  action on certain groups of drugs and our 
  role was to see what had been the effect of this action 
  on reducing expenses.  Behind every 
  4 
  Page 7 
  group of drugs we analyzed, you have 
  different stories, which are also in relation to the 
  Specific pharmaceutical companies.  Each 
  drug has its peculiarities. 
  So we track, to see how the action made 
  changed things, but to be able to interpret 
  differences and really understand what is happening, 
  we have to go back to the context.  You can the 
  Compare accounting.  You can not 
  understand the reasons for spending watching 
  billing, but from acceptable as you can all re-open, 
  re-narrate.  As if you open a book and each 
  Page is an anomaly that you see and you're trying to 
  figure out.  On every page you turn, you're more 
  deep in the questioning.  Or, 
  like when you meet a new person, the 
  you start printing, and then you're gonna ask 
  questions to understand the family, ideas, 
  particular story.  This induces problems 
  d'interprétations, it's 
  messy 
  . 
  In the case of these drugs, we stayed with 
  many questions: what costs fall 
  because the drugs have gone generic 
  or because there are fewer patients?  If treatments 
  are for chronic conditions, how to make 
  reduce a budget?  How to measure a change 
  monthly if there patients receiving boxes 
  Drug three months ...? 
  We have not saved what was expected, but 
  compared to what would have been spent if we had nothing 
  5 
  Page 8 
  Indeed, much has been saved ... 
  Finally for that matter, was formulated 9 
  calculation methods, 9 possible points of view!  We 
  was clear on the possible parameters and we said: 
  decide what is your strategy.  This returns the 
  person asking the questions to his responsibility: if 
  to account for the complexity of the issue, because 
  there is no way to decrease quickly, for example, and 
  a decrease, it is a complex issue, 
  itself!  And then you have to see it in time, long 
  term. 
  6 
  Page 9 
  Methods 
  Year: How I imagine your work in the 
  concrete? 
  I get the billing data extractions 
  disease insurance.  What I have in 
  Data is the date of the act, the act numbers 
  in the nomenclature of diseases (that is accurate enough, 
  there are 20 000 classification codes, can be 
  see the website of the INAMI), the provider number, 
  Patient national registration number, name, 
  address, invoice amount, the date on which the act 
  is introduced into mutual, the hospital, the location. 
  This is very accounting info, but very 
  Detailed. 
  I get the raw data.  We have access to everything 
  requires control work.  It has access to the 
  Billing of a physician, patients with identified 
  in order to verify the statements of the provider.  And 
  From these data, it restores relationship between 
  things.  There are living at the same address and 
  who have the same last name, it is a link.  There are 
  patients returning to their doctor, it is a 
  link.  There are providers who share a 
  patient base is a link .. 
  Using data completely flattened and 
  detailed, it's up to me to make connections based on 
  variables: same provider or not, same service 
  or not, same place or not, event sequences 
  etc ... 
  7 
  Page 10 
  Year: Is there a way to see the process? 
  Is that at the end of the process can be read 
  in the graphs? 
  Anne-Laure: I realize that I document 
  very few steps, the way my work 
  is moving!  Fortunately, it's all 
  programming, there is a trace in the code.  There 
  regular structure of the SAS code could be: 
  Datastep 
  : You eliminate the data that you do not have 
  need you audited, you grouped in categories, 
  etc .. 
  Sql proc 
  : You select the data and you make it a 
  money, or other computing 
  Histogram 
  : For the variable you are visualizing the'quantité' 
  eg results of the 2011 and / or thou compares 
  with the same calculation for 2008 
  If you do it well, you put the titles, which will be displayed 
  in graphics I export a document 
  word.  Otherwise, it is impossible to find yourself in the 
  code after which graphics just what stage! 
  But it is a habit that takes as and 
  extent, to better document the, delete 
  code parts that were not used to it 
  recover even after a few months. 
  We work in stages.  We made ​​a step, we test it 
  the custody or not. 
  I try to separate the code in the portion of the 
  data preparation (with a stable Datastep) and 
  8 
  Page 11 
  the data analysis, but it does not work 
  like that when you work: it does both 
  in parallel: code & context. 
  The other day a colleague told me, patients who 
  made a d'oncologie analysis, I think they should be 
  removed from the analysis, because it is in patients with 
  a particular profile.  We must therefore change the data 
  departure and eliminate all that.  In addition, sometimes I do 
  calculations on all patients, sometimes on a part 
  only.  Often I do not know very well what I have 
  done, so I do everything to turn, so I'm sure 
  I have not missed a step.  Sometimes it runs 1.5 hours 
  or 2 am, sometimes I change a little trick and I must 
  again.  That said, I do not do this full cycle 5-6 
  Once a year, not more. 
  It takes time and it must be documented!  In 
  Generally, I do not understand anything about my code when 
  I come back over a month later.  Fortunately, more 
  you programs, more is always done. 
  An: What tools do you use? 
  Anne-Laure: We work on databases 
  very large, so you have a large system (SAS), which is 
  complicated to use.  The digital data themselves 
  usually go back two or three years, but sometimes 
  working on data from the late 90s. 
  Then, publish the results in a written report of 
  20 to 30 pages.  I put the questions, assumptions 
  and then the details of the analysis. 
  Year: And the graphs. 
  9 
  Page 12 
  Anne-Laure: Both.  There are people who understand 
  good graphics and other tables. 
  Personally, I like the squatter plot, which 
  use 1 point per person on two axes.  They 
  allow more complexity, they make 
  developments finer, more visible.  They show 
  variability between individuals.  Sometimes it sends a 
  mail providers, in the kind of'la limit 
  normal is 10%, you are 15%, there should be 
  careful, it is costing the health insurance ... 'In 
  They usually take them into account.  With one 
  chart, you can show how each individual 
  the modified group (or not!) behavior. 
  The view of the choir 
  Year: From your calculations you find points 
  specific view, you could choose one, and 
  not leave the choice to the other .... Do you have 
  d'appliquer not want your knowledge 
  other areas that the Inami? 
  Anne-Laure: I've done Garance, an association 
  Brussels to fight against all kinds of 
  violence.  I worked on the results of their 
  of assessment questionnaires.  There were questions 
  open, the horror quantitative because you're 
  have to recode all the answers.  We had 200 
  or 300 responses to two different questions.  When you 
  open questions, you have all the voices 
  individual which are represented, each with their 
  differences, their specificity, while in the 
  10 
  Page 13 
  closed questions, there are 2 or 3 possible voice and you 
  abstracts answers.  In open questions 
  Garance, it was impossible. 
  A question like,'Qu'est it was hard for 
  the self defense course? ', led to answers 
  depended on the experience of each.  One of 
  responses was for example, someone'taper 
  knowing that we can break his leg, because I 
  am aware that I can do mal'. 
  There was wealth in their questionnaires.  They 
  asked two short responses open to all 
  world, that is a little to much.  In their 
  If, in my opinion, it could be better to ask 
  much like a small number, for example by 
  individual interviews with some participants, 
  rather than a questionnaire to everyone, because 
  the exploitation of a questionnaire is necessarily 
  simplifying .. But it's still difficult choices, 
  they also want to have variables that indicate the 
  satisfaction of each participating ... hence the choice 
  d'avoir open questions in a questionnaire 
  general anyway.  I almost never have to 
  open questions in my job. 
  An: Open questions therefore allow 
  no collective narrative? 
  Anne-Laure: If They allow choral narration. 
  as Svetlana Alexievitch does in his book 
  There 
  Supplication 
  or 
  Voices from Tsjernobil 
  .  The author 
  interviewed hundreds of people on the tragedy 
  of Tsjernobil and it makes something that is not 
  11 
  Page 14 
  more journalistic, that's choir and you can hear all 
  voice of all people, rewritten.  The 
  perspectives, which is serious or not, depend on 
  experiments.  You follow a person or another, but you 
  are immersed in a collective situation. 
  An: You work so exclusively with the 
  result of personal choice questions 
  multiple, often ask me great 
  problems ... How not to complete the box 
  'varia'?  The typical example is that your kind 
  it is male or female?  How you as 
  anyone so sensitive to that, how 
  you POSITIONED you in relation to this? 
  Anne-Laure: It's complicated.  There is a book very 
  interesting is written in relation to this, 
  Sorting 
  things out 
  Susan Leigh Star. 
  Here it is more of a nomenclature problem.  You 
  all legislation found on the website of the INAMI and you 
  see that the acts referred many are defined so 
  blurred.  The boundaries are blurred.  Once you have a flaw 
  in the nomenclature, often a part of 
  providers will benefit ... 
  The other problem is the problem of the identification. 
  Events are described in the billing and 
  each individual shall have a number, otherwise we 
  can not identify it.  When providers or 
  patients do not have a number, it's a mess.  By 
  example, there are in smoking cessation which have no 
  number because they are not doctors, but 
  psychologists.  So, I have a'fourre-tout' category 
  12 
  Page 15 
  which spends 200,000 euro with people I do 
  can not identify, it does not suit me at all! 
  To be precise in the calculations, it would 
  cover all the bases, numbering, identifying all 
  the world!  And here we touch a discussion 
  problematic, comparable to that of monitoring, 
  which has the same problem of legislation. 
  The health insurance generates a billion lines 
  Billing per year!  It was therefore a lot of data, 
  Due enormous wealth but are underused 
  Data analysis.  For the moment, these data 
  are widely used for analysis of 
  spending, but they could be used to open 
  more complex analyzes, and give meaning and 
  Content ... Gradually, users are more 
  numerous: KCE (Kennis Centrum Expertise), the IMA 
  (Intermutualiste Agency) develop analyzes 
  targeted, atlases, etc ... but it's still a work in 
  progress! 
  An: You work with personal data, 
  name, address, Social Security Number ... 
  What about confidentiality? 
  Anne-Laure: You sign a contract that you can not 
  ask questions that emerge from the field 
  staff, because even when the name and address 
  are removed and that the data are supposedly 
  anonymous, there is always the register number 
  National where you end date of birth, 
  gender, and can possibly afford 
  d'identifier the person.  I find that we are not 
  13 
  Page 16 
  yet sufficiently vigilant about these privacy issues, 
  but this is changing.  The awareness of the 
  how we should protect the data that is 
  use is more prevalent now. 
  An: And how do you POSITIONED you vis-à-vis the 
  Practice Standards is anyway 
  Based on your work?  What is extraordinary, 
  obvious, is about suspect. 
  Anne-Laure: We are all the time confronted with the 
  normality, indeed.  But this does not mean that 
  Exceptions are all out law.  There are some years 
  we looked at the amount of average prescription 
  in drug per patient for GPs.  The 
  doctor with more spending in Belgium is 
  someone who has a haemophiliac patient, who must take 
  many drugs that are very expensive, so it 
  happens to disproportionate budgets, which seem 
  abnormal but are completely legitimate! 
  Furthermore, our work is conditioned by the 
  context.  Data and research related to the 
  Insurance system diseases as constituted 
  right now.  The data is related to the act, and not 
  of hours worked, for example, as 
  England where the medical service is a free service. 
  It has disadvantages as well as advantages: 
  this allows also'connaître' reality acts 
  medical facts. 
  Year: Easy access to major bases 
  data from the institutions of State gives 
  thinking these days, that the activities of 
  14 
  Page 17 
  NSA control are revealed.  What do you think 
  as an activist? 
  Anne-Laure: I kept an article to speak at 
  classes in the ERG.  It is of a woman who sought 
  'cocotte minute' on Google, the same day as his 
  husband sought'sac to dos'.  They had landed the FBI, 
  which is a hundred such descents the year.  In short, 
  you can see everything in the data, but when you 
  are preparing a terrorist attack, you will not find 
  how to do it in Google, or purchase your tools 
  internet.  You do it with a fake valid identity card 
  an internet cafe, or you go to a store 
  and you pay in cash! 
  That is to say, the data are not always 
  sufficient, they are not a complete picture of the 
  reality.  Those who did not escape the NSA are not 
  terrorists!  The control systems want 
  there is more room, but there will always be 
  non-covered areas. 
  The big problem for me, politically, more 
  the data analysis, the Patriot Act.  There 
  possibility of detaining people without evidence is 
  much more serious than the continuous monitoring 
  data.  But maybe it's my job to 
  deform me? 
  D'un other hand, I love Scotland because there are plenty 
  non-covered areas without roads access, not 
  on the map.  At the same time it always creates 
  invisible spaces, also on the internet.  I blog 
  with a password that nobody knows 
  15 
  Page 18 
  the address ... 
  Year: Do you have fantasies about 
  data that you would treat them well?  Or 
  types of narration you could develop 
  with this? 
  Anne-Laure: I think all the new data 
  are always exciting.  It's like when you open 
  a new book.  After that, it is true that everything 
  linked to quantitative and qualitative.  I have a friend who makes a 
  thesis in psychology and when I see how it uses 
  the analysis of texts, the analysis of the interviews for 
  try to find the subtexts of what 
  interviewees tell their story 
  personal ... You can sometimes use techniques 
  quantitative to reveal things that are not 
  always made ​​clear, because they 
  make choices of words they do not know 
  necessarily that time. 
  When I was a summer camp counselor, one 
  was making reports of activities that adults 
  had with us.  People write what they believe to be 
  an objective description, but when you read it, it is not 
  never an objective description.  In fact, when you 
  talking about your life, what you say, what you do not say, the 
  words you use, you choose a formulation from 
  hundred possible.  When I say, for example, I have a dog, 
  I will not say his name, how it is 
  arrived home, why I keep it, but I chose to 
  give that information there.  So it is that it is 
  important to share this information with you, it 
  16 
  Page 19 
  which probably means that my dog is important, 
  and I try to share something with you ... 
  This friend identifies channels.  She works with 
  texts work software that allow you to 
  reclassify the life course of the items they 
  tell in the recatégorisant.  But also in this 
  place there, the person speaks in first person ... 
  After you compare between people too. 
  To answer your question, there is a very messy side 
  very messy in the data analysis.  There is 
  always too much data, we never quite put 
  end to this ... So either it's come to make an analysis of 
  data on very localized data, things 
  tiny ... where you can go very in detail. 
  I often have a problem with telling stories 
  because there is question of right and wrong, of 
  the inversion and the subjective.  As a statistician, 
  I try to raise the maximum Other information 
  of reality and world fauna as fair as possible about the aspect 
  the truth.  I'm good to find the stories 
  others and bring them to light, provided 
  I am finally totally absent from the 
  narration.  Or that I have this feeling, because 
  obviously this is not the case.  I try to do 
  scientific and not personal choice. 
  And therefore, it would be a project on data ... of life 
  private, life to me.  Not on the lives of every day, on 
  how you have walked, or how many letters you have 
  17 
  Page 20 
  read ... create information on things to keep by 
  example, the memory, the trace of the people, links. 
  Year: Memories? 
  Anne-Laure: Yes. 
  Year: How would you do that? 
  Anne-Laure: I do not know, that's why I do 
  do not d'ailleurs. 
  There is a d'objectivation side.  When I was a teenager, I had 
  notebooks and I wrote, I wrote ... with this 
  based one, to keep the memory.  And in fact, it has 
  no sense.  The important things you do notes 
  not because they are all up and you get in 
  remember, even twenty years later.  And you write stuff 
  genre, I was in love with this guy, but twenty years 
  after you re-read and you wonder: what was it? 
  But cons, which would take is to remember 
  precious things. 
  If you do not fictionnes your memories, once thou 
  get forgotten, they have no value.  If you describe 
  objectively neutral, an important memory 
  when you saw him finally when you reread the ten 
  years later, it means nothing.  This does not wake up 
  emotion. 
  Pereira Pretends to Tabucchi, for example, is a 
  novel written as a statement.  This is 
  supposedly objective, neutral and policeman.  This is a 
  story set in Lisbon during the period 
  Fascist.  And Pereira claims he did not know this 
  18 
  Page 21 
  gentleman who made ​​the attack, he had nothing to do 
  with him ... It's built for this purpose, but with the 
  d'écrire capacity of fiction.  It's supposedly neutral 
  that is supposedly objective, but in fact, when you read it, you 
  Totally overwhelmed by what is written.  It is very 
  sensitive, very strong.  It is in the literature. 
  I wrote three poems in my life that are far from the 
  against reality but which manage to evoke the great 
  emotion that I lived when I wrote, 
  more than tons and tons of 
  comprehensive and detailed diaries! 
  An: And so how you would approach this project? 
  Anne-Laure: The question of words and data, 
  it is also a different representation of reality. 
  It would be a database which in itself would 
  uninteresting but, when the analysis, regenerates 
  the history and emotion.  It's possible.  It's back 
  all the facts, all the factual side, etc and objectives 
  that you could not not re-tell, revive 
  the story when thou analyzes.  I thought of something 
  like that with timelines ... I can not write 
  the novel, but I can make the database 
  would remake the novel! 
  This is a question to see your behavior and your 
  being as a result of external things 
  (Family influence) and not as a choice, things 
  which découles and you you can replenish you, side 
  Cloning can also be. 
  19 
  Page 22 
  An: We do our own anyway selection 
  data. 
  Anne-Laure: Yes, but we do not all variables. 
  An: This one might imagine, is to choose 
  a particular situation that has marked you and go 
  looking for all the variables you 
  could have had, all the things that were 
  at that time offered, and all the choices you 
  could have done. 
  Anne-Laure: And then, what you've done, who made 
  you are moved you of a point to another. 
  20 
  Page 23 
  Constant Verlag, Brussels (2013). 
  Copyleft: This work is free, you can copy, 
  distribute, and modify it under the terms of the License 
  Free Art http://www.artlibre.org