DATA-MINING AS A TOOL TO PREDICT DOMESTIC VIOLENCE


Domestic violence being a serious problem in New Zealand. In 2006 police recorded 71,000 family violence related incidents. One in four women and just under one in five men would be the victim of domestic violence at some stage in their lifetime, according to an article identified at Stuff.co.nz dated 21 December 2008. Newly released crime statistics show a 13 percent jump in the number of domestic violence incidents that have been reported to police in the last year. There was growth in every area of family violence, including murder. Information exclusively obtained by 3 News also shows more women are seeking financial assistance to escape (Source: 3 News 1 October 2009). A number of the victims will end up with their GP or in an emergency room, yet overall the diagnosis of domestic abuse is often not diagnosed and a it is estimated that two thirds assualts go unreported. Would it not be a great thing if the risk of domestic abuse could be predicted on the basis of available data so as to enable the right parties to start asking questions instead of waiting for a report? Wouldn’t that potentially prevent a lot of harm? Well, a group of researchers at Harvard University created the first computer model to detect the risk of abuse at home.

It is a known fact that domestic abuse often goes unnoticed at doctors’ visits and patients often even try to hide abuse. Ben Reis, Harvard pediatrician and computer scientist as well as the designer of the new model, tigether with his colleagues, tapped into a public U.S. database containing six years of medical history for around half a million people. The data was fed into a predictive computer model that subsequently calculated the abuse risks linked to different diagnoses such as burns, sprains or mental disorders.

The advantage of such an approach is that physicians would no longer have to trawl through extensive medical records in what limited time they have, instead they would be presented with a single graphic: the so called “risk gel.” The graph shows the patients medical history as a collection of colored bars, with green bars meaning no risk an red bars meaning a risk that the diagnosis is statistically linked to abuse. The computer subsequently calculates the combined abuse risk.

When determined as high the physician is alarmed about that and made aware that a face-to-face meeting with the patient may be called for. The system is meant as a screening system. Using the system the researchers were able to make a diagnosis of abuse two years prior to that diagnosis by physician and the system was able to pick up signs not revealed by the patient yet. Less than 20% of the patients ffalgged as high risk cases by the system turned out to have actually been diagnosed with abuse. The researchers pointed out that data is far from up to speed when it comes to diagnosing abuse and prediuct that the task will get easier as electronic medical records are becoming more common. Previous trials with screening have shown that the screening is not the end-all and that adequate training is required in approaching patients/victims with the suspicions. The researchers hope to have a full fledged system ready, including user interface, in for years.

The full article including links to more information is available at Wired Science

Reblog this post [with Zemanta]

4 thoughts on “DATA-MINING AS A TOOL TO PREDICT DOMESTIC VIOLENCE

  1. Writing from Seattle, WA. . .

    VERY interesting stuff. There is additional research here by Group Health about the increased health care costs associated with exposure to abuse in affluent households.

    There’s a parallel push to have domestic violence characterized as a pre-existing condition thereby allowing insurance companies to avoid paying for treatment.

    • Dear Anne Caroline,

      It is funny you should mention the idea of pre-existing condition because that is what I could foresee as an issue with these new tools. I could imagine such a test as detailed by the researchers could lead to differentiation in policy rates or conditions. I am not sure whether domestic violence could be used as an argument to deny settlement of claims without it being a condition to be disclosed. At the same time I could imagine that being a victim could be a condition that would require disclosure, but you end up in some mixed up legal issues.

      The victim tries to hide it from the surroundings and even the GP yet the insurance company knows. Subsequently the insurance company is aware of specific crimes having been committed against one of their policy holder (and very like by another of their policy holders) and may be obliged to notify the authorities about that if only on the basis of ethical arguments. But how does such an obligation align with confidentiality.

      All in all quite a complex issue.

      Could you maybe post some further references or links in relation to this matter for me to do some further research.

      • You raise some VERY interesting points. I’ve worked in corporate settings on medical plan cost containment as a human resources professional and on insurance defense as an attorney. You’ve picked up on layers of complexity that I didn’t. Impressive.

        My hope is that insurance companies will perceive this tool as an instrument of preventive medicine rather than a way to evade paying claims. I might not be disabled today if somebody had intervened early on my behalf. The costs for my care now are absolutely staggering. And, the devastation to my career has been enormous. It could have all been prevented.

        One of the things that Group Health discovered is that most women who experienced abuse didn’t comprehend or appreciate the abusive dynamic in their relationships. If someone has been conditioned that abuse is love, abusive behavior is perceived to be “normal” ~ the women have never experienced respectful relationships so, therefore, don’t understand or appreciate what’s happening until the fists start flying or the guns go off. Often, of course, it is too late then for intervention. Expensive too.

        Here’s a link to a post I did on “The Impact of Domestic Abuse on Health” ~ it has embedded links that will take you to the studies I know about: http://annecarolinedrake.com/2009/04/29/impact-of-domestic-abuse-on-health/

        Here’s a link to US Senator Patty Murray’s (D-WA) web site on the legislation she introduced on October 1:
        http://murray.senate.gov/news.cfm?id=318516

        Here’s a link to a frequently cited blog post that will take you to Huffington Post’s blog on the issue:
        http://www.feministe.us/blog/archives/2009/09/15/domestic-violence-as-a-preexisting-condition/

        Please feel free to e-mail me privately if you have more questions. I’m quite intrigued by how this might play out personally as well as professionally. And, I will probably take the liberty of writing a post this week about what you’ve discovered.

        Your site, BTW, is quite impressive.

      • Thank you for the great comments and links. I guess this offers a great opportunity to further expand on this subject as it will have substantial relevance for New Zealand as well. Over here domestic violence would be covered under the so called Accident Compensation Corporation scheme.But … this scheme applies to incidents only and with the current predictive technology one could argue that at a certain point specific conditions are pre-existing or the result of (psychological) degeneration and therefore no longer covered. A harsh but potentially true possibility.

        I will most certainly go over all the links for further research.

        And my compliments for your site, I love great quality content and it looks your site has no shortage of that. I will get it on my blog roll.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s