Sunday, August 15, 2010

Predicting catastrophic health events - noninvasively and short term

This post was chosen as an Editor's Selection for
"I've just picked up a fault in the AE35 unit. It's going to go 100% failure in 72 hours". These were famous words of the almighty computer HAL in "2001: A Space Odyssey". Few of us believe too much in software forecasts - be it weather, earthquakes or computer hard disk failures. Yet, we all know that sometimes it works. And such systems are very valuable, assuming they continuously improve.
Word cloud generated from QS #15 intros
Wellness is always a common theme of the many presentations and experiments of Quantified Self enthusiasts - citizen scientists quantifying everyday life in order to improve it.  "Health" was the most frequent word among the two-word introductions of QS #15 attendees, followed by variations of sleep and happiness, and ways to measure and understand it.
Are health failures predictable?  Nobody argues they are, although most available services  - such as genomic testing - provide only long term predictions.  Biomarkers that scientists are interested in are obtained through invasive or costly interventions - like blood proteins or intracranial electroencephalograms. Technologies such as Mobile Cardiac Outpatient Telemetry™ (MCOT™) developed by CardioNet  are based on real time EKG focused on heart rhythm abnormalities that can't be detected in small 24 or 48 hour windows. These relatively rare events are correlated with symptoms and used for diagnostics.
Your husband just died, … here’s his black box
(from Gordon Bell's presentation)
Perhaps in the future every one of us will leave a black box holding all the truth about our health, helping next generations to better maintain, test and repair their bodies. Many lives have already contributed to the understanding of causes and effects such as the link between cholesterol levels, diet and heart attack. But what about the short term prediction horizon,  like the 72-hour window of HAL or 30 seconds to 15-to-45 minutes warning by seizure dogs?
Could computer luminary Gordon Bell predict his heart attack if he wore his heart monitor strap while bicycling?  A cardiologist would say "no way", but maybe sometimes it's better to keep quantifying and experiment despite of what medical establishment has to say?
Data from the Women’s Health Initiative study (129,135 postmenopausal women observed over a period of nearly eight years) show that a woman’s resting heart rate may be a good indicator of her risk for a heart attack. Hsia and others (2009) found that women whose resting heart rate is more than 76 beats per minute are significantly more likely to have a heart attack than women with resting heart rate less than 62 beats a minute. This risk factor is particularly strong for women between the ages of 50 and 64, less so for women over the age of 65. (Trial NCT00000611)

Of course, women have different kinds of heart attacks than men do. They are more likely to die from a spasms of heart and the blood vessels leading to the heart, and are more likely to complain of fatigue and sleep disturbances in the weeks and months leading up to a heart attack. Sleep disturbances and decreased physiological differences between day and night are increasing heart attack and stroke risks for males too, at least for shift workers. Exercise tests  - known for their false positive results - have better cardiovascular prognostic value for smokers with high cholesterol than healthy men.

Finger arterial pulsatile volume changes or finger blood flow - related to blood pressure - is another medium-to-long term predictor of pending cardiac events. A simple, noninvasive finger sensor test called EndoPAT  can predict major cardiac events such as a heart attack or stroke for people who are considered at low or moderate risk.

AngelMed Guardian System (inventor: Dr. Tim Fischell) is an early warning system - telling 24 hrs to a week before if heart attack is coming. Average time between the cardiac event and hospital arrival is 3 hrs, and about 5 hrs before the surgery starts. By that time 90% of muscles could be dead, so 5 or 10 minute warnings could save lives.  Unfortunately, Guardian is highly invasive - sized as an iPod, it is implanted directly into the patient's chest, monitoring electrogram  waveforms and other crucial heart-signal data 24-7.

Many portable oximeters and ECG are already in the market - and even though consumers are complaining about noise and difficulties in getting readings of diagnostic intervals, the technologies will continue to improve and self-quantifiers will be finding solutions for better health.

Hsia, J., Larson, J., Ockene, J., Sarto, G., Allison, M., Hendrix, S., Robinson, J., LaCroix, A., Manson, J., & , . (2009). Resting heart rate as a low tech predictor of coronary events in women: prospective cohort study BMJ, 338 (feb03 2) DOI: 10.1136/bmj.b219

Mayo Clinic (2009). High resting heart rate could predict heart attack in women. Mayo Clinic women's healthsource, 13 (7) PMID: 19498326