Big data in medicine: five examples of successful and unsuccessful use
It has become commonplace that so-called big data will soon revolutionize medical science. After all, a lot of medical information about many people: the results of CT-MRI, genetic and other tests, electronic medical records, data on work, lifestyle, favorite foods - can not help but find the causes of disease and treatment. The question is when this long-awaited breakthrough will take place. We’ve picked up a few examples of what big data is all about.
Twitter against fever
One of the earliest examples of the use of what can be called big data is the prediction of dengue outbreaks via Twitter. Dengue fever comes to Brazil almost every year like the flu in Ukraine. But epidemics start every time in a new place. It is impossible to react quickly: weeks pass from the outbreak of the disease to the corresponding reaction of state medicine. In 2009, two Brazilian research institutes set up a program to analyze the incidence of tweets. In the general message flow, the program catches the word denge and geographical references. By analyzing the structure of the phrase, the program determines whether it is about the personal experience of the author of the tweet, or he, for example, discusses the policy of the Ministry of Health in this area. In 2009, the program analyzed 2,500 tweets and was able to predict in which areas there will be outbreaks of the epidemic. The work continued in the following years. In 2017, a group of scientists summed up this tweet . In 2016, for example, there were 475,000 relevant tweets with the word denge, and 74% of them were geographically linked. The study confirmed that the analysis of tweets allows not only to get an express picture of the spread of fever, but also to predict an outbreak a maximum of two months before it begins. However, a full-fledged program for the use of this sea of information in Brazilian health care has not yet been established.
Big data and diet
This study of carbohydrates is the champion in the number of participants- was published in August last year. In the first part, the object of observation were 15,400 Americans aged 45 to 64 years. They all answered questions about their diet at the beginning of the study and six years later. Observation of them lasted 25 years. As a result, it turned out a simple thing: and low-carbohydrate diet (when the body receives less than 40% of energy), and rich (more than 70%) reduce life expectancy. The best middle ground is to get 50-55% of energy from carbohydrates. A 50-year-old study participant who consumed such a reasonable amount of carbohydrates could hope to live another 33 years. A low-carbohydrate diet promised a citizen of the same age 29 years of age. Rich - 32 years.
In the second part of the study, these results were tested on a larger number of participants and at the same time examined the question of what to replace missing carbohydrates. The number of participants was 432 thousand people from 20 countries. More precisely, it was a meta-study that combined the results of eight separate studies. Conclusion: if you have to replace carbohydrates with something, then proteins and fats of vegetable origin (nuts, etc.), not animal. Of course, the idea itself is not very new - but what is important here is not novelty, but the probative value of big data.
Intestine in numbers
The Japanese have shown how big data should be used. They studied the risks of right-sided hemicolectomy - removal of the right colon (colon - the main part of the colon). This work is very indicative of the amount of data and the unpredictability of the result. Researchers collected data on 1.2 million cases of surgery in 3,500 Japanese hospitals during 2011 and identified among them all cases of right-sided hemicolectomy. They turned out to be 19,070. Data analysis was entrusted to a computer. He calculated the mortality rate (2.3% in conventional surgery and 6% in emergency surgery), and then on the basis of records in case histories independently identified as many as 26 risk factors (insufficient blood clotting, renal failure, previous peripheral vascular disease, etc.). etc., etc.) and calculated the contribution of each of them to the operation. The authors state that their work, which at that time had no precedent in terms of size and rigor of approach, determined the path of surgery in the coming years. However, only in a narrow segment of medicine.
Farewell, hearing, hello, depression
A very important plus big data - the ability to relate parameters that are usually analyzed and considered only separately. In January 2019, a study was publishedabout the connection of hearing problems with other diseases. The object of the study were 154,414 people aged 50 and older who consulted a doctor about hearing loss. It turned out that the risk of developing dementia and depression in such people within five years after treatment is 50% and 40%, respectively, higher than in people who did not complain of hearing. In addition, the financial side of the issue was studied: if a person has hearing problems, but no action has been taken, the cost of treatment in the next ten years is 46% higher than the average at this age.
IBM treatment: fast but with errors
IBM is working hard to break into medicine with its IBM Watson supercomputer. This is the most powerful machine, consisting of several dozen servers and aimed at understanding natural human language. The medical unit of the program, Watson Health, or - in free translation - Dr. Watson, is trying to get a computer to help doctors make decisions. It is assumed that a machine with all the information about the patient and the disease can at least speed up this process. The results are still twofold. On the one hand, according to IBM itself ,In analyzing genomic information to more accurately determine the method of treating glioblastoma, Dr. Watson did a job in ten minutes that would cost a living expert 160 hours. On the other hand, it has recently become known that the supercomputer often makes mistakes literally in the same place. For example, when bleeding advises a drug that can aggravate this bleeding.
According to experts, the fact is that when teaching Dr. Watson at IBM often used not real patient stories, and synthetic images based on them. As a result, Watson’s knowledge was detached from reality. IBM has something to work on, and the big data is where to grow.
Illustration: pixabay.com
Read also: Will the transplant system work in Ukraine?
Twitter against fever
One of the earliest examples of the use of what can be called big data is the prediction of dengue outbreaks via Twitter. Dengue fever comes to Brazil almost every year like the flu in Ukraine. But epidemics start every time in a new place. It is impossible to react quickly: weeks pass from the outbreak of the disease to the corresponding reaction of state medicine. In 2009, two Brazilian research institutes set up a program to analyze the incidence of tweets. In the general message flow, the program catches the word denge and geographical references. By analyzing the structure of the phrase, the program determines whether it is about the personal experience of the author of the tweet, or he, for example, discusses the policy of the Ministry of Health in this area. In 2009, the program analyzed 2,500 tweets and was able to predict in which areas there will be outbreaks of the epidemic. The work continued in the following years. In 2017, a group of scientists summed up this tweet . In 2016, for example, there were 475,000 relevant tweets with the word denge, and 74% of them were geographically linked. The study confirmed that the analysis of tweets allows not only to get an express picture of the spread of fever, but also to predict an outbreak a maximum of two months before it begins. However, a full-fledged program for the use of this sea of information in Brazilian health care has not yet been established.
Big data and diet
This study of carbohydrates is the champion in the number of participants- was published in August last year. In the first part, the object of observation were 15,400 Americans aged 45 to 64 years. They all answered questions about their diet at the beginning of the study and six years later. Observation of them lasted 25 years. As a result, it turned out a simple thing: and low-carbohydrate diet (when the body receives less than 40% of energy), and rich (more than 70%) reduce life expectancy. The best middle ground is to get 50-55% of energy from carbohydrates. A 50-year-old study participant who consumed such a reasonable amount of carbohydrates could hope to live another 33 years. A low-carbohydrate diet promised a citizen of the same age 29 years of age. Rich - 32 years.
In the second part of the study, these results were tested on a larger number of participants and at the same time examined the question of what to replace missing carbohydrates. The number of participants was 432 thousand people from 20 countries. More precisely, it was a meta-study that combined the results of eight separate studies. Conclusion: if you have to replace carbohydrates with something, then proteins and fats of vegetable origin (nuts, etc.), not animal. Of course, the idea itself is not very new - but what is important here is not novelty, but the probative value of big data.
Intestine in numbersThe Japanese have shown how big data should be used. They studied the risks of right-sided hemicolectomy - removal of the right colon (colon - the main part of the colon). This work is very indicative of the amount of data and the unpredictability of the result. Researchers collected data on 1.2 million cases of surgery in 3,500 Japanese hospitals during 2011 and identified among them all cases of right-sided hemicolectomy. They turned out to be 19,070. Data analysis was entrusted to a computer. He calculated the mortality rate (2.3% in conventional surgery and 6% in emergency surgery), and then on the basis of records in case histories independently identified as many as 26 risk factors (insufficient blood clotting, renal failure, previous peripheral vascular disease, etc.). etc., etc.) and calculated the contribution of each of them to the operation. The authors state that their work, which at that time had no precedent in terms of size and rigor of approach, determined the path of surgery in the coming years. However, only in a narrow segment of medicine.
Farewell, hearing, hello, depression
A very important plus big data - the ability to relate parameters that are usually analyzed and considered only separately. In January 2019, a study was publishedabout the connection of hearing problems with other diseases. The object of the study were 154,414 people aged 50 and older who consulted a doctor about hearing loss. It turned out that the risk of developing dementia and depression in such people within five years after treatment is 50% and 40%, respectively, higher than in people who did not complain of hearing. In addition, the financial side of the issue was studied: if a person has hearing problems, but no action has been taken, the cost of treatment in the next ten years is 46% higher than the average at this age.
IBM treatment: fast but with errors
IBM is working hard to break into medicine with its IBM Watson supercomputer. This is the most powerful machine, consisting of several dozen servers and aimed at understanding natural human language. The medical unit of the program, Watson Health, or - in free translation - Dr. Watson, is trying to get a computer to help doctors make decisions. It is assumed that a machine with all the information about the patient and the disease can at least speed up this process. The results are still twofold. On the one hand, according to IBM itself ,In analyzing genomic information to more accurately determine the method of treating glioblastoma, Dr. Watson did a job in ten minutes that would cost a living expert 160 hours. On the other hand, it has recently become known that the supercomputer often makes mistakes literally in the same place. For example, when bleeding advises a drug that can aggravate this bleeding.
According to experts, the fact is that when teaching Dr. Watson at IBM often used not real patient stories, and synthetic images based on them. As a result, Watson’s knowledge was detached from reality. IBM has something to work on, and the big data is where to grow.
Illustration: pixabay.com
Read also: Will the transplant system work in Ukraine?