Questions of digital data

Today there was an excellent article in ArsTechnica about the cost benefit of going to digital records within the health care provision sector of the USA. The study found that there was limited cost benefits in such a move overall within the health care sector and that mving to such a system could actually increase costs. I found a comment from one of the readers equally interesting:

"Devilbunny

I'm a physician.

"Electronic medical records" is a nebulous thing. You see, some things are already electronic - most in-hospital records, for example, have a computer system that accesses labs, X-rays (and their readings), and maybe the dictations of operative notes, admissions and discharges, and so forth. At my hospital, all cardiology procedures are dictated/typed in, and so are available, as are copies of all EKGs. *If* someone gets their care from our hospital and its associated physicians, it's extremely easy to get all the basic information in one place.

But we still use paper charts for inpatients, because doctors and nurses are busy, and paper can easily be shared among a lot of people. A chart can be opened up as long as there are thirty inches of space on a desk and then closed and put up as soon as a brief note is written; a computer takes up almost as much linear space but can't be moved. You can take out a sheet of paper, find a quiet corner to write in, and bring it back; can't easily do that with a computer.

For sharing data between institutions, there's another concern: what format do you ship it in? The problem with computerization is that the format that is best from the perspective of the practitioner is not necessarily the one that gives you the best gains. For example: one of the VA's greatest pluses is that the computers automatically assess a wide variety of patient health issues and issue reminders when you log in. Mr Jones needs a flu shot (because he's older than 65 and has emphysema), Mr Smith needs a colonoscopy (over 50, none within last 10 years), Mr Wilson needs a blood sugar level, etc., etc., ad nauseam. Now, this can be entered by data entry clerks at the VA, because the VA is big enough to afford to pay a few secretarial salaries in order to improve the efficiency of the whole place. However, a small private practice can't, and so the physician has to spend time entering all those risk factors. Often, this means that the software will attempt to make this part of the history-taking process; choose from an array of checkboxes and drop-down lists which ones need to be tracked. This is then automatically transferred to the list of diagnoses.

But notes written by a computer often read like notes written by a computer, rather than notes written by a human. So the record that results from these systems is almost unreadable, even though it contains all of the same information as a well-written history note. A well-written history, on the other hand, makes very clear sense to another practitioner but has no real advantage other than accessibility over a paper note. It also takes a lot less time to write. (It makes no sense from an efficiency standpoint to have highly-paid physicians acting as $10/hr data entry clerks.)

So which style do you choose? The list-and-checkbox system tracks diagnoses well, and it does a decent job of making itself understood once you know how to read through the data presentation. The block-of-text is generally easy to read, but with another significant failing over paper: it can be very, very difficult to find the note you want. The VA note system is admirably egalitarian - a note from a physician and a note from a dietitian are the same, just notes. But this means that the really important notes - those admission and discharge notes - are lost in a vast sea of notes. Every time a patient calls their nurse with an issue, that nurse has to document the problem and the solution. When a needy patient meets a dedicated nurse, the result is a mountain of individually entered notes over the course of a shift, while the same thing in paper is easily recognized as having occurred in one shift (because it's on one sheet of paper) in the nurses' notes (which are separate from physicians' notes, so we don't end up having to surf through each others' paper).

Now try to figure out what you do when a patient who comes from one system needs to make records available to another.

Computerization made a difference in most businesses because they had automatable processes - billing, for example. Those automatable processes in health care are a lot fewer. The need for non-text data storage - Xrays, CTs, ophthalmology notes (multi-color drawings for most if not all patients), vascular surgery notes - makes health IT expensive and hard to do well.

In short, the reason we don't have this solved is that the problem is very hard, there are a lot of judgment calls about what the best way to solve the problem is, and most of the benefit only exists if there is a massive network effect."

So what then does this show us. In my mind it shows that data itself is only interesting if we give it meaning. Part of any digital medical records project needs to involve questions about what it is they want from the data. Too often this question is ignored and the result is a system that doesn't meet the needs of end users as this commenter points out:

"Rick Weinhaus MD

One of the major problems with the current generation of Electronic Health Record (EHR) software is that it is barely usable or unusable -- the software is not based on how the human brain takes in, processes, and organizes information. In other words, the entire fields of cognitive science and of human factors have largely been ignored in the high-level design of EHRs.

A recent report from the National Research Council came to a similar conclusion. The report found that currently implemented healthcare IT programs often ‘provide little support for the cognitive tasks of the clinicians or the workflow of the people who must actually use the system . . . [and] do not take advantage of human-computer interaction principles, leading to poor designs that can increase the chance of error, add to rather than reduce work, and compound the frustrations of executing required tasks.’

The current generation of EHR software does a poor job of presenting information visually, thereby not taking advantage of the highly developed processing capabilities of the human visual system.
Secondly, for EHR technology to be truly useful for physicians, the software design must correspond to the physician’s mental model of the patient. I believe the mental model most physicians implicitly use is that the patient is a dynamic organism whose health changes over time. The temporal sequence of past, inter-related events is the best way to comprehend current health issues."

Social Impacts of data

One of the most important parts of the sociology of data is how this way of thinking is impacting on how we live as human beings.

The question remains what kind of society are we living in now? Postmodernism? Post-structuralism?

Or Information Society?

The theory of the information society holds that information and knowledge becomes a significant economic force such that it replaces traditional industries (Wikipedia).

If this theory is true then what impact does this have on the way we think and act? And how much are we really an information society.

SIDE NOTE: One thing I have noticed in my own job is that the amount of information we collect has increased. In comparing the records from the 1990's to those from the late half of the 2000's it is quite clear that the organisation I work for has increased the amount of information it collects and stores. I wonder how much will ever be read again? Maybe we need an information auditor we examines what information is actually being used.

Data visualisation

Today I came across the wonderful Information is Beautiful blog. It it written by David McCandless who calls himself an independent visual and data journalist (What a cool title!!!!). He has created wonderful data visualisations such as 'Reduce your chances of dying in a plane crash' or 'Mountains out of mole hills'. 
What this shows is that data itself can be manipulated and changed according to how we perceive it. This is not to say that data visualisation is bad. On the contrary good data should give us more information about the world around us and further our knowledge of it.

Data visualisation reimnds me of the statistics courses I did at university. I can remember my lecturer drumming into us that the type of data dictates what form the data should be shown in (must did out that course book).
 

Looking at data from an information system view

The discipline of information systems theory is one academic field that is concerned with issues of data. Importantly it focuses on the interaction between processes, mainly business, and technology (Wikipedia).

As one possible area that may provide some framework for a broader understanding, that is one that goes beyond just technological and business uses, further research and reading must be done in this area. I have been reading Philosophical Frameworks for Understanding Information Systems by A. Basden (2008).

If anyone knows of any good books as starting points into this theoretical framework please email or leave a comment.

The meaning of data: Version 1.0

As this project had been evolving over the last couple of days. I have been really really struggling with a suitable definition of data. Since my whole project is based on challenging our notions of data some time and effort must be given to coming up with a reasonable definition of what it is.

According to wikipedia :

Data is the plural of datum and means 'something given'. It is the lowest level of abstraction from which information and knowledge are derived.

In this definition data is seen as the raw stuff that is used to develop information and knowledge about things.

Some questions to ponder with this broad definition.....

Can knowledge and information be data too?
Is all data equal?
Human created versus natural?
Is data useful on its own?

My current job involves working with large amounts of people's data. Specifically I am involved in digitising this data so that the paper copies may be destroyed. This data goes back to early 1990 when I was just 9 years old.

So has started a fascination with what data is. I am on a journey to answer some questions.....

What is data?
Why do we keep data?
What does data do?

Clearly all data is not the same. Nor does it have the same value between different people. My personal data such as my age, DOB, address, email address, phone number are all pretty public information. But my data, which I freely put onto the internet into such services as Google or Facebook, is this my data too. Do I own it or do I give that away? Should I worry that companies can use my data for profit.

But I believe these questions go beyond this to a broader question of systems of data.

Should they be integrated?
Be portable?
Who owns them?

It is my hope that this blog will become a place to start a conversation about data as something that impacts on us all the time