Saturday, May 9, 2015

Uncertain Health in an Insecure World – 39

“49 Reasons, All in a Line…”

“… Some of them good ones, some of them lies.” These Crosby, Stills, Nash & Young song lyrics critiqued the dirty politics of the Richard M. Nixon administration.

The lie of modern data science, according to IBM big data evangelist James Kobeilus, is “overfitting”. He opines that statistical models of historical datasets do not guarantee their future predictive value in the same domain of interest. Data scientists are trained to regularly score their statistical models, but resulting overfitting biases may introduce skewed data that cause predictive modeling failures. Such pseudo-data are often spurious – like the age of Miss America correlating with the incidence of murders by hot steam and hot objects!

This is one just example of how the digital world that tries to help us can hurt us.

Can we make good decisions from sifting through the 2.4 quintillion bits of digital data generated daily from the big data stream in the Cloud? And when an estimated 50% of the digital data generated on individuals and their personal interests is inaccurate, is the aggregated big data a giant solution, or a giant problem?

Who decides what of this data is actually relevant, or “A solution looking for a problem”.
Uber uses Google Map as its customer locator solution. But Uber CEO Travis Kalanick (below) has just put in a $3 billion offer  to buy Nokia’s mapping product, Here, because nobody wants to entrust the future of a data-driven business to the privacy-mining Google platform. 

If patients’ individual little data is being protected by public interest policy, who is assuring that the constraints of privacy and surveillance are balanced in the private sector?

More than half of Kaiser-Permanente patients will willingly trade their healthcare privacy for better care. Who is to say that they don’t have the right to waive their government-given right to privacy? But caution is warranted when the private sector is involved – recall what happened to the Obamacare signup website last fall, when it was revealed that Google and Twitter were sharing enrollees’ personal information.

Subjects enrolled in clinical trials are protected by strict research ethics and rules of conduct designed to guard their confidentiality. But pharma is now encountering & managing the reality that some research subjects are self-identifying on social media, demanding to know whether they received the active drug or a placebo. Does such enlightened self-interest spoil it for others, and could such disclosures render an expensive clinical trial ineligible for regulatory review and new drug approval?

Analytics firms can render big data meaningful, making order of chaos and even predicting the probability of bad (or good) health outcomes. But who makes the call as to the right question for the predictive model to attack… especially when healthcare providers are increasingly viewed by data collectors and analyzers as ‘disintermediaries’, marginalized in the process of care?

Big data analytics do not replace competency.

Big companies, including large healthcare systems, must possess business competencies in order to derive a benefit from what their big data warehouses might reveal. Daily decision-making based on good operating information cannot be replaced by digital data analytics. Customer and patient experiences provide experiential information about system performance at many levels.

When aggregated and thoughtfully analyzed, these little data bytes may provide insights on process improvement that can foster real change. Performance management using big data metrics derived from little data events can be a difference-maker, but only if the organization’s culture allows for people and patients on the ground to understand their roles in making change happen

Kaiser-Permanente makes this happen, but most other healthcare systems fall woefully short. 

Financial services companies use big data analytics to identify prospective ATM machine locations. When it takes me three tries to process an ATM transaction, should somebody at the bank alert my doctor to early signs of Alzheimer’s dementia?

Fitbit’s business success in personal fitness tracking analytics has just prompted it to seek a $100M IPO filing with the U.S. Securities & Exchange Commission (SEC). When I fall below 10,000 steps for three consecutive Fibit days, should the company alert my health insurer of impending type-2 diabetes due to physical inactivity?
Sound absurd? Perhaps...

A Future Watch world of little data collecting and big data analytics is being responsibly applied in the present by progressive employers and healthcare insurers.
In the Square, we can speak freely about the lies – intended and unintentional – that surround this new data movement.

Richard Nixon is no longer the U.S. President, but threats to our personal freedom persist! 

No comments:

Post a Comment