Pets and Children's Anxiety: the Bond Mattered More Than the Species

Somewhere tonight a parent has two tabs open. One is a shelter listing. The other is a school counselor's email about a nine-year-old who has started asking, most evenings, whether everyone in the house is definitely going to be fine. The week's coverage says pets are good for children's mental health, and some of it goes further and says it depends on the animal, which turns a vague reassurance into a decision: a dog, then, or would a cat do. The parent is trying to answer a species question. The evidence underneath is answering a different one.

The paper being covered is by A. Gadomski, M. B. Scribani, N. Tallman, N. Krupa, P. Jenkins and L. S. Wissow, published in BMC Pediatrics in 2022 as Impact of pet dog or cat exposure during childhood on mental illness during adolescence: a cohort study. It is a follow-up. The same group had earlier studied 643 children aged four to eleven and found that those with pet dogs had lower anxiety scores, a result that traveled well. This time they went back to see what had happened to those children by adolescence, and the answer is more interesting than the first one was.

This is a retrospective cohort study, which means nobody was assigned a pet. The team merged its earlier study database with electronic medical record data, then used ICD-9 and ICD-10 codes to identify the common mental health diagnoses — anxiety, depression, ADHD — recorded between the original enrollment and 27 October 2021. Proportional hazards regression compared how long it took young people with and without pets to receive a diagnosis. Record review covered 571 young people with a mean age of 14 and a range of 11 to 19; 53 percent were male, 58 percent had a pet dog at baseline, and mean follow-up ran 7.8 years.

Between April 2020 and October 2021 the researchers also interviewed parents and young people about two things the medical record cannot hold: how much time the young person had actually spent with a pet across those years, and how attached they had been to it. That produced three separate exposures rather than one — having a pet dog at baseline, cumulative exposure to a dog or cat during follow-up, and level of attachment — against three outcomes: an anxiety diagnosis, any mental health diagnosis, and a mental health diagnosis accompanied by a psychotropic prescription. Splitting the exposure that way is what makes the study worth reading, because the three do not agree.

Of the 571 young people, 191 received a mental health diagnosis over the follow-up period. Anxiety was the most common at 99 diagnoses, or 52 percent of those diagnosed; ADHD accounted for 61, or 32 percent; depression for 21, or 11 percent; and 10, or 5 percent, had combined diagnoses. This is a study with enough events in it to say something, which is not true of every paper that reaches a headline.

Having a pet dog at baseline, after adjustment for significant confounders, was associated with a lower risk of any mental health diagnosis: a hazard ratio of 0.74, P = .04. But it was not associated with anxiety specifically, and not with a diagnosis carrying a psychotropic prescription. That is a quiet and important failure. The original finding that made this line of work famous was about anxiety, and in the longer follow-up the dog association did not hold for the outcome it was supposed to explain. Parent-reported cumulative exposure to a pet dog, among the 241 young people (42 percent) reached for interview, came in at the same hazard ratio of 0.74 but at P = .06, which is to say it did not clear the conventional threshold.

What did hold was attachment. Cumulative exposure to the pet a young person was most attached to, whether that animal was a dog or a cat, was associated with a lower rate of anxiety diagnosis at a hazard ratio of 0.57, P = .006, and with a lower rate of any mental health diagnosis at 0.64, P = .013. Those are the strongest associations in the paper, and the species distinction has disappeared from them entirely. The published abstract reports these hazard ratios with P values but without confidence intervals, so how precisely each one is estimated is not visible from it — a hazard ratio of 0.57 could sit inside a wide interval or a narrow one, and that difference matters for how much weight it carries.

A sustained, close bond with an animal over the better part of a decade is not a thing that happens to a household at random. It requires, at minimum, that the family kept the pet, could afford it, stayed housed with it, and did not fall apart around it. It also requires a child with the temperament and the circumstances to form the bond. Every one of those conditions plausibly predicts better adolescent mental health on its own. The authors adjusted for significant confounders, and adjustment is worth something, but no amount of it settles which way the arrow points between a stable childhood and a deep attachment to a pet.

The attachment measure carries a second problem: it was collected at the end, in interviews conducted between April 2020 and October 2021, asking parents and adolescents to characterize years of accumulated exposure and closeness in retrospect. Memory of how much a child loved a dog is not independent of how that child turned out. And only 42 percent of the cohort was reached for those interviews, so the attachment findings — the headline findings — rest on a subset who may differ systematically from the 58 percent who were not.

Because it reframes the question people are actually standing in the kitchen with. The advice market sells species: which breed, which animal, best pets for anxious kids. This cohort suggests that the variable doing the work is not the animal at all but the duration and closeness of the relationship — a fact about how a family lives alongside a creature over years, not about what it chose at the shelter. That is a harder finding to act on, which is probably why it is the one that got flattened in transit.

It is also, by the standards of this literature, a reasonably serious piece of evidence. Most work on pets and child wellbeing is cross-sectional and rests on questionnaire scores collected once. This one has a mean of 7.8 years of follow-up and endpoints that are coded clinical diagnoses rather than self-reported mood, which is the kind of design difference clinicians weigh when deciding how much of a literature to believe. What it describes is a pattern across a few hundred children in one health system. It does not forecast what one particular animal will do for one particular anxious nine-year-old, and the authors do not claim that it does.

The most interesting result in this paper is the one that did not survive the trip to the headlines. The dog finding — the original, famous, anxiety finding — came back weaker: children with a pet dog at baseline were no less likely to be diagnosed with anxiety over the following years, and the one dog association that did reach significance was for any diagnosis at all. What held, and held harder, was attachment, pooled across species. That is a real result and it points somewhere uncomfortable for anyone hoping to solve a child's anxiety by acquisition, because attachment is not a thing a family purchases. It accumulates across years that also contain everything else about that household, which is precisely why it is so hard to disentangle from the household itself. Read strictly, this is a study about the durability of a relationship wearing a study about pets as a costume, and the most defensible reading of it is that the animal is the occasion for the bond rather than the cause of the benefit. That is less purchasable than a breed recommendation, and considerably more honest.