Some children diagnosed with ADHD (attention-deficit/hyperactivity disorder) still have symptoms as adults, and others do not. Adult daily life also varies a lot between people. A new peer-reviewed study tested whether computer models could predict these outcomes. The models did little better than chance.
What the researchers did
The team used machine learning. This means computer programs that look for patterns in data and use them to make predictions. The data came from a study that followed people for 18 years. The researchers assessed 133 adults from this group, with an average age of about 30. Three in four were men. The models tried to predict two things: whether ADHD persisted into adulthood, and how well people functioned in daily life in the long term.
What they found
Prediction accuracy was judged with a score that runs from 0.5 to 1. A score of 0.5 means the model does no better than tossing a coin, and 1 means perfect prediction. For whether ADHD persisted, the best model reached only 0.58. For daily functioning, the models explained almost none of the differences between people. The authors conclude that long-term outcomes of ADHD are complex and hard to predict.
What remains uncertain
The group was small. With only 133 people, computer models have little data to learn from, so a larger study might do better. Most participants were men, so the results may not apply to women. The summary available to us does not list all the information fed into the models. The study shows that these particular models could not predict outcomes in this group. It does not show that outcomes can never be predicted, and it says nothing about any single person's future.
