Symptoms and Diagnosis.

In this post I will present the argument for establishing a unified database which tracks observed symptoms of patients presenting at any participating entity. The goal of such a system would be to build a very large body of data about diseases as expressed as symptoms, not just as diagnoses.

Why?

    The Bad Doctor

It’s of course now well known that Dr Patel in Bundaberg was a bit wonky.

One of the problems with the Dr Patel case is in how one could identify his incompetence. One way to do so is by gross statistical comparison. Far more of Dr Patel’s patients died than patients of other doctors at the same hospital or elsewhere. The problem is that such a measure is retroactive, even tragically so. Innocent people must die in order to uncover the incompetence of a given doctor.

Record the Symptoms

I suggest that instead of simply recording diagnoses, all patients who present should have their symptoms recorded. I understand that this currently occurs to greater and lesser degrees at various places. My proposal is for a single system to be developed into which all such data can be fed.

By tracking symptoms as the primary focus of data collection, the process of oversight is greatly simplified. Learned committees or more senior doctors can quickly review symptoms and the diagnoses entered for them. Even more significantly, data mining and medical expert systems can be used to quickly root out incompetent doctors. This might save lives.

The benefits continue. Such a body of data would be invaluable for medical research. To take one obvious example, it will allow us to roughly estimate disease levels over time, including for diseases not known at the time of data capture. Put otherwise, Joe Bloggs might present today with symptoms X, Y and Z. Today those symptoms may be mysterious, tomorrow an entirely new disease may discovered which neatly explains them. Such a system would provide researchers a way to estimate previous incidence of the disease, by searching for patients who presented with X, Y and Z. Again it might save lives by alerting relevant doctors that a new potential diagnosis has emerged from research and that they should contact their patient.

There will probably be uses and research possible which I cannot forsee. I’m confident that they would be very beneficial.

It also goes without saying that such a system could improve the portability of patient histories. As patients move from doctor to doctor, they could authorise their new doctor to unlock their history as stored by the central service. This would also benefit doctors by cutting down dramatically on their current administrative overheads.

Problems

There are a number of problems with this proposal.

Privacy

In the first instance there is privacy. This is not too difficult, however. Any such system complies to Australian privacy laws by seeking patient consent. Furthermore patient identities would be stored separately from their symptom data. The symptoms database would refer to anonymous ID numbers; the details relevant to those ID numbers would be stored separately. The real problem of privacy will be for patients who refuse to allow their details to be entered. Such refusals will skew the data, possibly quite seriously.

The only solutions I can see would either be to require information is a condition of service (unpopular), or to get non-participants to fill out a survey on why they have not given permission, in an attempt to give some data with which to adjust the main body of the database.

Cheaters

It’s possible that doctors who understand the way this database operates will try to “game the system”. This could range from attempts to cover up medical negligence to murderous psychopathology.

The truth is that it is hard to stop such cheating in system which relies on trustworthy “clients” (in this case, the doctors). The chief way to do so is to back up the system with other methods of detecting cheaters. One way to do so is the current method of tracking deaths of patients. Another way is to use data mining to learn more about data patterns of known-good versus known-bad doctors. Computers can pick up subtle and unintuitive connections that humans might miss. Some program of random spot checks would also need to be instituted.

What Kind of Institution?

Who will run this show? That’s the next big question. As a raving loony libertarian my instinct is to keep government out of it. I envisage that some sort of non-profit industry association could do the work. With contributions by hospitals, universities, doctor’s surgeries and the like it should be viable to operate such a system.

Why not a for-profit operation? There are some factors which mitigate against it. Patients may be reluctant to contribute to the enrichment of a private firm, and a profit-making firm would invite competitors and fragment the database across multiple providers.

This entry was posted in Geekery. Bookmark the permalink.