YOHA: Starting to talk about databases

Published on August 25th, 2011


* * * * * * * * * * * *

With their project Data Visualisation as Documentary Graham Harwood and Matsuko Yokokoji (YoHa) critically investigate the gap between the perception of the wider public about the social reality created by data and databases. They explore how database systems transform those that commission, use them as well as those that it makes subjects. Furthermore, Yoha tries to find out in what way the information generated and used by different protocols, services and facilities have an affect on the formation of the data body in the health system. By examining the National Health Services in Liverpool, organising workhops, and discussing with Dutch midwives and obstetricians they took the public health record and mapped out the authorities, polices, and rules that hold it in place and give it meaning. By focussing on the underlying code of the database they trace the meaning of the database. Whereas semantic analyses mostly look at the final image of the database, a picture of the database, it is the underlying code that makes the database and gives it meaning. Leaving from the assumption that relations are never neutral and that they always play a role between statements, nothing is arbitrary and everything is interlinked, they aim to find the point where agency is occurring.

A conversation between Graham Harwood (GH), Matsuko Yokokoji (MY), a Dutch midwife (DM) and Annet Dekker (AD)

part of NetArtWorks: The Database, the Datacloud and the Public (Data)Domain

SKOR, Amsterdam 25 August 2011

 

BACKGROUND TO THE PROJECT
GH: initially, Liverpool Primary Care Trust (PCT), a division of the UK’s National Health Service, asked us to look at their various data sets. Compared to national mortality statistics, people in Liverpool die roughly fifteen years earlier than those living in some other cities. They asked us to think about these differences and how pieces of data could be used to encourage people to reflect on their wellbeing. Usually the PCT’s methods are very strict where they say ‘you must do this’, but of course this is a useless approach for most teenage girls at risk of pregnancy.

Also, people in the primary care trust working with data and public health frequently realise that their efforts are ignored. They find it difficult to get the population interested. So the brief was to begin to find a way to get young mothers to think about the kinds of information the primary care trust collects, and to see if we could encourage reflection on this to improve wellbeing. Our job was to create a kind of space outside the discipline of the database, and medicine, where all these people could come together and think about these data sets.

That is what we’ve been doing in Liverpool. And now we are interested in seeing how this type of investigation might work in the Netherlands, because the systems share similarities but are also quite different. What we want to talk to you about are databases and births and hopefully discover how things connect up.

WHY BIRTH AND PERINATAL DATA?
MY: I believe that when people think about a database they usually envision a grouping of numbers that might be interpreted in a certain way to provide information. But if you think about it, a database is actually all types of things and is not limited to a single authority – it can be a hospital, the government, etc. These institutions store information that is used to influence individual lives. We’d like to see if we could simplify the subject and work out how a database can affect an action. We’re using the action of a midwife, because in some ways this job has clear and defined tasks, namely to ensure that the birthing process does not endanger the mother or the baby. A lot happens around that process, of course, so a great deal of data is gathered and the data has a purpose.

GH: Yes, we were particularly interested in data sets relating to birth, because it is the first place where you are recorded in information systems – in effect you are given a databody – which is really interesting, as it articulates our initial relation to power and governance.

ART METHODOLOGY
GH: To place this ‘research’ in the context of art is appropriate because art is a space that no one owns; we can imagine it as a non-discursive space. Dealing with a topic such as this using art as the framework is interesting, as you can gather a group of young mothers to have a conversation with people from the statistics department, and create a space that is not professionalised or medicalised. Art in this context provides a space without hierarchies or boundaries and people tend to open up more easily in such a situation.

CARE IN THE NETHERLANDS
MY: Yesterday we had a brief discussion with midwives from the Birth Centre (Geboortecentrum) in Amsterdam, and we were fascinated by the difference between the types of information that the mother is asked to provide in the Netherlands compared to the UK. In the UK demographics is based on your lifestyle and where you live. Something that surprised us about Holland is that the midwife asks people about their education.

DM: Yes, but perhaps I should clarify this because we distinguish between primary and secondary care in the Netherlands. If you require primary care, you go to a general midwife, as you saw yesterday. A general midwife takes care of pregnant woman who don’t show signs of complications and can give birth at home. If there is a potential risk, the woman is hospitalised – this is secondary care. Primary care is more personal. For example, there is more time for the intake, and more information is gathered that provides greater insights into people’s social environments. This is so that we can get as complete a picture as possible about issues such as (possible) domestic violence, their education, etc. In secondary care, where I work, there is less time to talk about the social context, and we focus more on the complication and the medical case as such.

MY: Is the information that midwives in primary care gather more complex?

DM: No, it’s more personal.

MY: It is more detailed?

DM: Yes, and more private.

MY: Do they keep their own databases?

DM: Yes, and its all privacy protected.

GH: So, primary care is where the mother gives birth at home with a midwife in attendance, and in secondary care the mother is hospitalised?

DM: Yes, if there is a complication the mother goes to the hospital, and if not, she stays at home with the midwife.

MY: Where do most people go?

DM: I don’t know the exact percentages and I know that some things are changing. There are about 200,000 births in the Netherlands each year, and the percentage of home births is rapidly decreasing. I believe the figure now is about 25 per cent of all births, whereas it used to be 30 per cent, maybe more. Increasing numbers of women are being admitted to hospital to give birth. So, from primary care they go to secondary care because a risk has been diagnosed.

MY: The kind of data that you deal with is quite medicalised, and this medicalisation is costly. Somebody in the hospital says that this woman has to spend four days in bed, and these are the costs. If you are diagnosed in the UK, a coder takes your diagnoses and assigns a code to it, which allocates a certain amount of money to the PCT.

DM: It‘s the same here. When a woman is hospitalised, the gynaecologist enters the codes.

MY: Do midwives still deliver the babies in the hospital?

DM: Yes, unless there is a complication that requires a vacuum, forceps, or a cut – only a doctor can perform these procedures.

In the Netherlands, data relating to births is divided across two databases. One is for primary care (LVR1) and the other is for secondary care (LVR2). Then there is Perinatreg, which is an institute that does a great deal of analysis and demographics. They send their annual reports to everyone in the field – I also get a copy each year – and they also put them online, so they are accessible to everyone.

DATA COLLECTING
MY: So, basically, when a child is born, it is the midwives who collect and enter the data in the database. There are a few different ways to store the data. Do you use other methods to collect data?

DM: We use paper and digital forms, and we have special software, which is designed for this purpose, because we collect specific data.

MY: Are the forms that you use already divided into categories?

DM: Yes, they are always categorised.

MY: Are the categories relevant to you? When you fill in the form, does what you write down differ from the actual situation with the mother and child? Do you think the form is appropriate? Have you ever felt that certain things are missing?

DM: Of course, the gaps occur when you describe, for example, the position of the baby’s head in the pelvis. Then you’re talking about obstetrics, and in a way this means that you’re talking about the content. General information such as the time the baby was born and the mother’s age are usually filled in consistently.

GH: Does having these various pre-determined categories in the database mean that you focus on information that is appropriate to those categories? If you have a pre-determined category, a container, say, you’re always looking for things that fit into that container. All the fields in the database of births define what the birth process is, and anything outside of that no longer belongs to the birth machine. Now, I’m also wondering if the system can grow or change? How would it – or you – do that?

DM: It could grow, because midwives are often confronted with the problem that they cannot always fill in all the information they acquire because it is not covered by any of the categories. If this happens we go to the coordinator, who contacts the software maker, who adds another category.

MY: Do you think that your opinions feed into the system?

DM: Yes, because I know that if I have a certain experience with something that isn’t covered in the form, my colleagues will understand what I write down, and then others from our department will also recognise it. It takes time, but it is possible to make changes, and the options to do this are becoming simpler. The system is becoming more and more flexible.

MY: Something else… there are specific categories and certain places where the midwife or the doctor can record their findings, and they also have a certain role and different perspectives, but how do these abstract numbers and categories relate to your experience? Categories are never definite.

GH: Yes, but the data always trivialises our experiences; it simplifies them. You have this intense experience with someone that is distilled into a number of discrete answers. It doesn’t adequately reflect the experience, because we have all these different sensations, and the database does not operate in that way.

DM: It does, because the space allocated for additional notes continues to expand, which is reflected in the hard copy.

GH: That is interesting. But there are different ways of interpreting notes. There is an algorithm that analyses your notes, and then looks at a potential pathology based on them. Is that something you are aware of?

DM: What do you mean exactly?

GH: Well, you said there are more and more ways to write notes. If you write your notes, a block is made in the database, which consists only of that text. Then there is another algorithm that looks for the words you used and makes little clouds of these words and predicts what kinds of problems, experiences, or outcomes will occur, based on previous entries and in relation to your notes. I’m not sure, of course, if this happens in your database, but it is another type of analysis. It looks for key words and makes a cloud, which it can compare to other clouds to determine what is potentially going on. What is interesting in the medical field is that doctors use a quite specific set of terms, which makes this sort of analysis comparatively simple.

MY: What happens if words only have a meaning in the context of the sentence or if a mistake is made? Something might be interpreted in the completely the wrong way?

DM: The words are translated into data, which makes it objective. I’m not aware if they do something like that already.

GH: If they aren’t already doing it, they will start to do so soon. A category used to be just a category, nothing more, but with notes they can start to decipher and understand motivations and behaviours, which could potentially benefit medical practice. So, notes can tell much more and are therefore more interesting.

ACCESS AND VIEW IN THE DATABASE?
GH: Does the doctor see different data from you?

DM: I would say that most midwives see the possibilities, and most doctors see the pathology, the things that can go wrong.

GH: And what about the literal view of the data set?

DM: We use the same database. There is no special doctor’s database.

GH: What I’m also referring to is, for example, when the admin people look at the database they see a code for the treatment and the costs, but they won’t see patients’ personal information. So they have a different view in the database. Are you aware that in practice someone else might have a different view of the same data?

DM: Yes, I’m conscious of that, because I know that doctors have different access. I see the surface, the things that I filled in. But I know that with a different code, I can go behind this superficial layer, see what kinds of codes have been assigned, and what kind of money is involved. It must be there, but I don’t know what it looks like.

MY: You mean that the doctor has more access?

DM: Yes, the doctor is responsible. Of course, I have my own responsibilities, but if I’m sued, the doctor is too.

GH: Does the doctor see birth as a machine?

DM: Yes, more than the midwives, I think.

GH: I would perhaps describe it as a machine-like process, in which the database informs the doctor’s mindset. He has experienced the worst situations and, to prevent them from recurring, will respond to the data more than to the real situation, which might be of a more emotional nature than medically complicated?

VALUE OF DATA AND DATA ANALYSTS
GH: The chaos of databases in the UK really amazed me. There is no single database where you can easily track a patient. You have a patient number, which is linked to records of all your hospital treatments, so you can pull up all their records and find out exactly what everything costs. But it’s very difficult and there is always a delay, because the hospital or the general practitioner’s surgery has a different database. All these different databases go to the Office of National Statistics, where they are anonymised and sent back to the primary care trust. So the primary care trust has all the data from all the doctors, all the hospitals and the Office of National Statistics, and then it tries to compare these to approximate what is going on. One of the things we were told was that people who work on the data spend a lot of time just working out what the costs are. They said that they produced data around a kind of median, with a maximum and a minimum. There is a word we use – caveat – referring to the potential problems of each definition within the probability. So they solve that by saying, we think this is true as long as you place it in this context. But then management takes the context away, and at that point it ends up being the one issue that they are looking for, and disregards other complications. Is that something that happens here as well? Put differently, do you believe the data? Is the data that is sent to you true, or is it an interpretation?

DM: If it’s national data then I trust it, but if I read a file by someone else I’m not sure if I can trust it entirely.

GH: So you might read a doctor’s file and think, he hasn’t filled this in very well.

DM: It happens all the time.

AD: That would affect those outcomes as well, but you still put more trust in the national database?

DM: Yes, because it is less detailed and more general. The greater the detail, the less I trust it.

GH: One of the things that is really surprising in the UK is that data analysts don’t have a professional body regulating their ethics. Suppose their boss said, I need you to find another quarter of one per cent – besides quitting their job in protest, there is no ethical body they could turn to if they wanted to make a complaint. I was really shocked by that.

DM: What are the consequences?

GH: If someone managed to establish how many people had survived hypertension, they could unlock more government money. The emphasis is then on the statistical analysis to find another fraction of a percentage anywhere in the figures so they can get more money. It could involve getting to a floating-point number, adding just one more digit on one side or the other to find that percentage. There isn’t a body or a union that could back you up if you wanted to say to your boss, I’m not doing that because it’s unethical. This is in the UK, but it could also be true for the Netherlands.

DM: I wouldn’t know.

GH: From another point of view: if a woman experienced complications, would there be a cost analysis of the different options – could we make a cost saving if we advised a caesarean section immediately?

DM: That’s a nasty way of looking at it. The money we receive for a caesarean differs from that for a vaginal birth. If the doctor is afraid of complications, then it could…

TRUTH PRODUCTION IN AND THROUGH THE DATABASE
AD: To continue, the database also takes into account the condition of the woman who is giving birth. As you said earlier, if she’s experiencing a lot of pain and can’t cope with it, what would you advise for the next time – to have a natural vaginal birth, or a caesarean?

DM: When a woman returns for her second delivery and her first delivery was traumatic because it took a long time and ended in a caesarean anyway, then the doctor is more likely to say, I understand your traumatic experience, what do you want? If the woman says, I no longer have confidence in a vaginal birth, she will often choose a caesarean and that will also be allowed. But it also depends on the doctor.

MY: That means you have to be aware of that risk.

GH: Yes, but the risk is calculated. It is medicalised and coded in former experience, or in tests this mother had, which show as an anomaly.

DM: Exactly. And this anomaly can appear during pregnancy or pre-pregnancy. The primary care midwife’s job is to see where the risks are.

GH: What is really interesting is that we have these ideas about calculating risk, and ideas about diagnoses, which are presented in such a way that you have to believe the truth that is being produced by the data. You have to say, yes, we are taking all these measurements and this is true for us, which is why we have to have the statistics. In a way the databases are also producing these truths, do you believe that as well?

DM: We create them by entering the data.

GH: But do they produce their own kinds of truth, for example, if comparisons are made between people?

DM: It depends on what I click. I have many options.

GH: Yes, but that is only in the record you are filling in, but if you bring lots of records together, they could be interpreted to create other truths. If you demonstrate that six woman were given this particular drug with this particular outcome, isn’t that a new kind of truth that is being revealed by the records?

DM: Yes, but you would still have to conduct proper research.

GH: What about your role in this truth-making process. You are empowered by the database, but the database also produces power that can alter your practice. It is not a completely human process. Do you feel to some degree that you are an appendage to the machine of truth making, and the database.

DM: I’ve never looked at it like that. I think you’re crediting the machine with more power than it has from my experience. But it is true that once you have written down certain data, for example, about a previous birth, it has – or can have – major consequences for a subsequent birth.

AD: But only for the same individual…

DM: Yes.

AD: Can it also have consequences for other people?

DM: Yes. For example, a woman gives birth in our hospital. The head is born, but the baby is stuck – we call this a shoulder dystocia. It means that the shoulder is trapped beneath the pubic bone, and you only have a few minutes to release the shoulder so that the baby’s lungs can open up. These manoeuvres can be difficult and potentially cause a lot of harm to the woman, or they can be light. After this has happened you write it down in the woman’s file, and a possible consequence is that the next time she gives birth it will be by caesarean section.

MY: So, to avoid possibly harming the woman and/or baby you have to make a quick decision if the baby gets stuck? It’s all recorded. And that data forms the basis for decisions in the future.

DM: Exactly.

MY: Because of this data they will perform a caesarean if it happens again.

DM: Shoulder dystocia is a grey area: It can be very complicated, or only require a few manoeuvres for the baby to emerge quickly, but the database doesn’t discriminate. Doctors are always sensitive to their liability, and will strongly advise operating, thereby medicalising the pregnancy and possibly harming the woman even more.

AD: How will this affect other women?

DM: It will only affect other women who had a similar complication. Because I know that what I write in the database could affect the way someone is treated in the future, I might change the way I phrase the information so the consequences are less severe.

GH: You are aware that the machine is producing a kind of truth that can act on that person, so you try to be as ethical and careful as you can.

DM: My perspective is also very much informed by the fact that I was a primary midwife. I believe in physiology, and in giving birth vaginally. If I was a doctor who had seen lots of horrible things, I might think differently and decide that a woman should never give birth vaginally again, because I’m concerned that the birth of her next baby will be similarly complicated.

MEDICALISATION OF THE PROCESS, THE MACHINE, OR, HOW THE DATABASE OPERATES
GH: One of the things that interests me about this process, is that a number of different histories are taking place at the same time. On the one hand, there are women who give birth and all of the knowledge that builds up around that, and on the other there is a medicalisation of that process. The database, I say, is really at the end of the medicalisation process. It’s making a machine of birth.

When we fill out the database, we can think of it as a way of looking at the world, as a kind of gaze. But it also gazes at you, because you fill in the details. And someone else can follow your steps, as such almost policing you through this gaze. But it is also a gaze on the population. This is not limited to a single person; many people do this. We can see that gaze as a kind of surveillance, which instead of looking in only one direction, looks in many directions. Is that something that is true for you, do you recognise this process? The National Health Service in the UK is particularly bad at monitoring itself, but there is an independent organisation that looks at mortality rates. It compares data relating to hospital mortality rates and discovered that the hospital near us had 350 more deaths per year than it should have. This should sound the alarm, which is really good.

MY: It created a scandal, because the hospital hygiene was really bad, and these people died for no reason. The database found it.

GH: Do you have any thoughts about how the database operates as a form of surveillance – perhaps ‘gaze’ is better.

DM: I enter the data, and I know that all our patients’ data produces statistics relating to the number of caesareans or vacuums. This way you can compare, for instance, the number of caesareans performed in different hospitals. This is interesting, as we have this philosophy, or ideal, in the Netherlands to have as few caesareans as possible, because the recovery period is long and difficult. The caesarean rate is much higher in some areas in Holland, particularly in wealthier areas. Why? Maybe it is a specific population that is more demanding. That’s interesting, especially because it costs more. Each form I fill in is dedicated to a single patient, and among other details, I have to enter several names: who was present at the birth, who was responsible, and who did the stitches in the vagina. You have to put your name to it. You stitch a wound, so there is a potential for infection. I always find it a little bit scary, because I’m a bit insecure about my sewing skills. I’m always afraid that at some point there will be complaints about the healing process, and that my name will be connected to it. I’m sure that I’m not the only one who has this insecurity.

GH: It’s almost as if the database polices you in your processes.

Another example: In a way a birth is a bit of a messy, bloody process, and there is this database, which was meant to be an early form of recording messy, bloody births. It was a way of slowing down the process by distilling the experience into a number of discrete elements through the creation of fields in the database. But somewhere along the way, it seems to me, birth has been transformed from being a messy, bloody process into a model, which fragments this experience into a new reality of birth. In other words, the reality used to be you with the mother. But now it’s the database that is real, because that is the one that will act in the world.

DM: That’s absolutely true. The amount of administration that is attached to births or pregnancy takes more and more time that would otherwise be spent on actual care. With the sort of national development that is going on and the discussions about the system we have in the Netherlands, it is really becoming a matter of money. Less attention is being paid to patients. Hospitals want to have more and more births, because they earn them a lot of money.

AD: How does that affect your own work?

DM: It’s getting busier. More women are being referred to us, so more women are ending up in secondary care, which means I have more women to care for during a shift. This is at the expense of the care, and I’m also producing ever more data.

AD: It’s interesting that you produce more data, because that will influence the system. It might even make it more user friendly.

DM: It is always like that. You go from an individual to a group, and from a group to the individual. What is interesting for me is that I collect data; I gain more experience and become more secure in what I do.

GH: That’s an interesting analogy. You describe it as data, but the way you deal with the experience is for the sentient. You use all your senses and all your intelligence, whereas the database gets a very tiny snippet of all the information you have.

DM: True. And when I’m dealing with a file about a previous birth, I add everything I know to it.

GH: Earlier I was talking about you being an appended to the database, at that time you seemed quite hostile to the idea. When we first started talking you were quite separated from the database, but during the conversation we looked at the way it produces truth, how it produces power, or how it’s has become folded up into the birth process , not separate from it. Does this conversation change the way you thought about it?

DM: No, it doesn’t, because as I told you I already feel as if I’m part of a data-production machine, and already know that what I write down has consequences. But it does make me a little more aware of it.


* * * * * * * * * * * *

T O    T O P