Why is genetic diagnosis so slow in intensive care?
Some serious illnesses in children are caused by a fault in the genes. When a baby is in intensive care, finding out quickly which genetic disease is responsible can point to a targeted treatment and improve the outcome, according to the authors. To do this, doctors read the whole genome, meaning all of the person's DNA.
The problem is not only reading the DNA: it must also be interpreted. That takes highly qualified professionals who compare thousands of genetic variations with the child's symptoms, which are often scattered through the medical record. There are few of them, which holds back wide use of the method.
Every hour counts in a very sick infant. An accurate diagnosis sooner makes it possible to choose a suitable treatment, or to avoid tests and treatments that would not help.
What did the team develop?
The researchers put two building blocks together. First, accelerated sequencing: the genetic material is prepared directly from a blood sample, and reading takes 15.5 hours. Second, a clinical language-processing program reads the child's electronic record and pulls out their clinical signs.
A third step then ranks the possible genetic diseases by comparing the child's signs with those known for each disease, and by judging how harmful each variation in their genome is. The result is a provisional diagnosis that experts can then confirm.
The idea is to replace part of the slow, highly specialised manual work with automatic sorting, so that experts only have to check the result rather than rebuild everything by hand.
What are the results?
The record-reading program pulled out clinical signs with 80% precision and 93% recall. In 101 children with 105 genetic diseases, it matched on average 4.3 expected signs per disease, compared with 0.9 in manual interpretation.
Looking back, in 95 children with 97 diseases, the automatic diagnoses agreed well with human expertise: 97% recall and 99% precision. Prospectively, the platform correctly diagnosed 3 of 7 seriously ill intensive care infants, with a mean time saving of 22 hours 19 minutes. The median overall time was 20 hours 10 minutes.
In other words, the computer found many more useful signs in the record than were used in manual interpretation, which helps point to the right disease.
Key figures
- 15.5 hto sequence the genome from a blood sample
- 97% and 99%recall and precision against human expertise, in 95 children
- 3 of 7intensive care infants correctly diagnosed prospectively, with 22 h 19 min saved on average
What does this change for patients?
In the three infants who were diagnosed, the diagnosis affected treatment, which shows the potential value of the approach. If a result arrives in hours instead of days, the care team can adapt a treatment sooner.
The authors hope that this automation will allow fast sequencing to spread to more intensive care units, where specialists to interpret the results are currently in short supply.
Faster answers can also spare families days of uncertainty, although this study did not measure that.
What are the limits of this study?
The real-life part covers only 7 infants, and the platform found the answer for just 3 of them. It gives a provisional diagnosis that experts must confirm. We cannot conclude that there is a health benefit on a larger scale.
The retrospective results rely on cases whose diagnosis was already known. Larger studies in several hospitals are needed.
It also does not say how the method would perform in hospitals with different record systems, nor how much it would cost to run routinely.
A baby in intensive care: keeping track, with Mima
This research study shows that rapid genome sequencing can sometimes shed light faster on the cause of a serious illness in a newborn. It is not available everywhere, and the decision rests with the team.
In these very hard moments, Mima can help a parent note the explanations received and the questions for the team, so nothing is forgotten. One possible question: "Could a rapid genetic test be offered to my child?" She never replaces the care team.
Frequently asked questions
What is whole-genome sequencing?
It is the reading of all of a person's DNA, meaning all of their genes. It lets doctors look for a variation that causes a genetic disease.
What does the AI do in this study?
It automatically reads the medical record to pull out the child's clinical signs, then helps rank possible genetic diseases. Experts then confirm.
Is the diagnosis final?
No. It is presented as provisional. Experts must confirm it before any decision.
Is this method available everywhere?
Not yet. The study is a demonstration on a limited number of children. The authors hope it will help fast sequencing to be adopted in intensive care.
The original study
Clark MM, Hildreth A, Batalov S, et al. “Diagnosis of genetic diseases in seriously ill children by rapid whole-genome sequencing and automated phenotyping and interpretation” Science Translational Medicine, April 2019. DOI : 10.1126/scitranslmed.aat6177
Read the full study in Science Translational MedicineThis article summarises a scientific study and is not medical advice.