‘Black-box’ medicine promises to revolutionize treatment, but with serious risk

Tejas Choudhary

Every fall, the University Health Services rolls out its annual flu-shot campaign. Hundreds of students are injected with the vaccine across tens of booths around campus. The campaign mirrors the nation-wide drive to bring more people under the vaccination umbrella. But the hard truth remains that more than 50 percent of the adults refuse to take a flu-shot.

To those of us who wonder why so many people are incredulous of the efficacy of modern medical treatment, it might be a surprise that 38 percent of patients with depression, 40 percent with asthma and 75 percent with cancer do not respond to medical treatment. The problem, however, isn’t that medical science hasn’t advanced, but that it’s not used efficiently. The ‘one size fits all’ model of medicine, at its core, is an ambitious attempt to oversimplify biology.

Current medical treatments and drugs need to go through a series of clinical trials to demonstrate drug efficacy. And although the trials have led to advances in drugs and treatment, the drugs are approved broadly, not for a specific group. It is as if all shoes were made size eight to fit maximum people.

Personalized medicine promises to be the solution. The heart of it is this: Since every patient is different, medical treatment should be tailored to the individual patient. The challenge with personalized medicine for decades was the sheer size of computational requirements. With the improved data analysis available today, that challenge can be soon resolved.

A growing number of researchers are now working on using big-data to create “black-box” medicine, where an opaque algorithm would parse through data about the patient’s familial medical history and treatments for a similar problem to find relations that maximize a patient’s chance to feel better. The algorithm would then recommend a drug and the patient would eventually be cured, but no one would know exactly why. Not even the computer.  

A Black-box doctor might seem like a long shot to implementation, but remember that the 2014's Protecting Access to Medicare Act mandates that computerized clinical decision support (CDS), or algorithmic medicine, must be in place in the United States by 2017.

The very idea of a robotic doctor seems scary. Riding in your self-driving car to a futuristic building where a robotic arm takes your blood and vends a bottle of medication is unsettling at best. It takes away all the human feelings associated with seeking help. The cozy waiting-room chairs, the polite nurse and the reassuring doctor. There certainly is a degree of psychological relief they provide.

Privacy, regulation and security also pose additional challenges. Misuse of medical data can create a headache. The data could potentially be used to alter insurance rate for some patients or to create targeted biological weapons. Regulation of the algorithms is another concern that will require addressing. Regulators must clarify who would be responsible for the mishaps resulting from the black-box. The system will also need to be safe and secure. Hacking could give birth to potential weapons. But the most disturbing aspect of the growth of algorithmic medicine is that it is far from mature, yet growing in use. A recent study by the Journal of American Medical Association stated that the CDS doesn’t work over 60 percent of the time.

It is hard to imagine, then, that the technology would develop and be used within a year. Regulators must acknowledge that it will take plenty of efforts to educate the public and gain their trust. The desire for revolutionary innovation has often lead to undercooked solutions. With something as sentimental as treating human lives, it is best we adopt change gradually.

Choudhary is a civil engineering and finance junior from Mumbai.