ADHD Diagnosis Linked to 71% Higher Risk of Death in Meta-Analysis of 15.9 Million People

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental condition whose traits, including inattention, hyperactivity, and impulsivity, can affect daily life, from school and work to personal relationships. In clinical and educational settings, attention tends to focus on challenges such as sustaining concentration, carrying out plans, and controlling impulses. In recent years, however, large-scale epidemiological studies have begun to suggest that ADHD may also be associated with long-term health outcomes and mortality risk. It has been suggested that people diagnosed with ADHD may have a shorter life expectancy than those without a diagnosis, and identifying the factors behind this has become an important research question.

On October 19, 2026, a research team from University College London (UCL) and the University of Liverpool published the results of a systematic review and meta-analysis in the journal JAMA Network Open. The paper, by Lucy Corrigan, Lauren Heywood, Professor Joshua Stott, Dr. Amber John, and colleagues (accepted August 24, 2026; DOI: 10.1001/jamanetworkopen.2026.39448), examined data from 19 observational studies conducted across three continents, covering a total of 15,920,970 people (36.1% female). Some studies followed participants for more than 20 years. The analysis found that the hazard of all-cause death was about 70% higher in people diagnosed with ADHD than in comparison groups without a diagnosis. By cause of death, risk was elevated not only for external causes such as suicide and accidents but also for natural deaths from physical illness.

However, these results do not prove that ADHD itself shortens lifespan. All of the analyzed studies were observational, and no causal relationship has been established in which ADHD directly causes early death. Likewise, the association between medication and lower mortality risk seen in the study does not prove that the drugs have a direct protective effect. What matters is distinguishing what these statistics represent from how far they can be interpreted.

{
  "type": "bar",
  "title": "死因別に見たADHD診断群の死亡ハザード比（HR）",
  "unit": "ハザード比",
  "categories": ["全死亡", "自殺", "事故・外傷", "身体疾患（自然死）"],
  "series": [
    { "name": "統合ハザード比", "values": [1.71, 1.89, 1.55, 1.27] }
  ],
  "caption": "Corrigan et al. (2026), JAMA Network Open。ADHDのない対照群を1.0とした場合の相対ハザード比。いずれもP < .01で統計的に有意。",
  "source": "JAMA Network Open (2026). DOI: 10.1001/jamanetworkopen.2026.39448"
}

What a Hazard Ratio of 1.71 Means in Data on 15.92 Million People

The central measure in this study is the pooled hazard ratio (HR), which compares mortality risk over the follow-up period. For the association between ADHD and all-cause mortality, the analysis found a statistically significant hazard ratio of 1.71 (95% confidence interval: 1.31–2.24, P < .001). This means that at any given point during follow-up, deaths occurred about 1.71 times as readily in the ADHD group as in the comparison group. It does not mean that the probability of dying within a fixed period is itself 1.71 times higher.

Cause-specific analyses also showed significant associations in several categories. The hazard ratio for death by suicide was 1.89 (95% CI: 1.21–2.95, P = .005), about 89% higher than in comparison groups. For deaths from intentional or unintentional injury, the ratio was 1.55 (95% CI: 1.22–2.00, P < .001), indicating roughly 55% higher risk, including deaths from traffic accidents and falls. Natural deaths from physical illness, including cardiovascular and metabolic disease, also showed a hazard ratio of 1.27 (95% CI: 1.16–1.38, P < .001), an increase of about 27%. This cause-specific analysis does not, however, confirm the same proportional increase for each individual physical disease.

To understand these figures, relative risk must be distinguished from absolute risk. A hazard ratio expresses the relative difference in death rates during the observation period; it does not directly indicate an individual's probability of dying in the future. In younger people, whose baseline mortality is low, a 1.7-fold relative risk may translate into a comparatively small increase in the absolute number of deaths. In middle-aged and older people, whose baseline mortality is higher, the same proportional increase could have a larger impact. Because this meta-analysis did not present absolute mortality risk for each comparison group in a unified form, the results cannot be applied directly to an individual patient's life expectancy or probability of death.

Not all 19 studies were used to calculate the pooled hazard ratios. Twelve studies were included in the quantitative meta-analysis: nine for all-cause mortality, five for suicide, four for intentional or unintentional injury, and three for natural death. The remaining seven were excluded from the quantitative analysis and covered in a descriptive review, because their outcome measures did not match, they addressed outcomes that could not be combined with other studies (such as COVID-19 deaths), or too few studies were available for pooling. Data extraction was carried out independently by two researchers, but study quality was assessed by one, which leaves some methodological limitations.

The definition of ADHD also varied between studies. Most relied on clinical diagnostic codes, but some used self-reported diagnosis history, questionnaire-based assessment of ADHD traits, or polygenic risk scores (PRS), which quantify the effect of related genetic variants. These differences mean that the range and background of the people classified as having ADHD differ from study to study. The pooled hazard ratio should therefore be understood not as a fixed value applying uniformly to everyone with ADHD, but as an estimate statistically summarizing associations observed in different study populations.

High Variability Between Studies Points to Complex Underlying Factors

A key point in interpreting this meta-analysis is the high heterogeneity between studies. Heterogeneity indicates how much the results of the pooled studies differ from one another, and it is assessed with a statistic called I². For all-cause mortality, I² reached 98% (Q = 479.36, P < .001), showing extremely large variability between studies.

An I² above 75% is generally regarded as high heterogeneity. A value of 98% means that estimates differed substantially across studies and that most of this variation cannot be explained by chance sampling error alone. Possible contributors include differences in national health systems, age composition of participants, length of follow-up, methods of diagnosing ADHD, and how confounding factors were adjusted for. However, the I² value alone cannot tell us which factors had how much influence.

Cause-specific analyses also showed high heterogeneity: I² was 93% for suicide and 87% for intentional or unintentional injury. By contrast, for natural death from physical illness, I² was below 1% (Q = 1.94, P = .37), indicating little variation between studies. The increased risk of natural death was thus observed relatively consistently across different studies. Only three studies contributed to this analysis, though, so there are not enough data to conclude that the same degree of increased risk occurs in all populations.

To understand the medication findings, it is also helpful to compare them with a 2024 Swedish nationwide study that is included in the meta-analysis. The two differ in important ways in design and assessment, but because the Swedish study is among those pooled in the 2026 meta-analysis, they are not independent confirmations of each other. Their features are summarized below.

The research team cites differences in national diagnostic criteria, access to health care, help-seeking behavior, and societal understanding of and stigma toward ADHD as possible sources of heterogeneity. In 15 of the 19 studies, men outnumbered women, so the sex composition of the study populations was also skewed. Because ADHD symptoms in women may be overlooked or diagnosed late, data dominated by men may not fully capture health risks specific to women.

The degree to which studies adjusted for socioeconomic status (SES) and co-occurring psychiatric conditions also varied. Because the number of studies was limited and information on their characteristics was not uniform, the team could not carry out meta-regression or subgroup analyses to examine the causes of heterogeneity in detail. A sensitivity analysis excluding studies that used ADHD polygenic risk scores, however, produced results broadly consistent with the main analysis. This suggests that, at least, studies using genetic risk scores did not drive the pooled results.

ADHD Medication Use Linked to 25% Lower Mortality Hazard, but Causation Remains Unclear

From a public health perspective, the finding of greatest interest in the paper is the association between ADHD medication use and mortality risk. Among people diagnosed with ADHD, comparing those receiving drug therapy with those who were not, the all-cause mortality hazard ratio for the medicated group was 0.75 (95% CI: 0.64–0.87, P < .001). This means that the hazard of death was about 25% lower in medication users than in non-users. The medications covered include methylphenidate, amphetamine-based formulations, and atomoxetine. This pooled analysis, however, did not present the absolute difference in mortality risk between medicated and unmedicated groups.

This result alone might make it seem that ADHD medication prevents early death. But the research team cautions against interpreting the association directly as a causal effect of drug therapy. Only 3 of the 19 studies were included in the medication meta-analysis, so the available evidence is more limited than in the overall analysis of ADHD and mortality. Heterogeneity across these three studies was I² = 55% (Q = 4.38, P = .11), a moderate level of variability. Because the number of studies is small, the analysis may also lack the statistical power to detect heterogeneity.

Of further importance are "confounding by indication" and "residual confounding," which accompany observational research. People who receive drug therapy and those who do not may differ in many ways besides medication use. For example, people who receive prescriptions on an ongoing basis may have easier access to health care and receive regular checkups and health management. Socioeconomic conditions and the support systems available for continuing treatment may also differ. Conversely, if the non-medicated group includes many people who cannot take the drugs because of serious comorbidities or contraindications, that in itself could contribute to the difference in mortality between the two groups.

Even when such differences are statistically adjusted for, it is difficult to remove all confounding from observational data. For instance, even if continued medication use is associated with better health, it cannot be fully distinguished whether that reflects the effect of the drug or an environment that allows continuous medical care. The hazard ratio of 0.75 should therefore be treated not as a direct measure of the mortality-preventing effect of drug therapy, but as a statistical association between medication use and better survival outcomes.

In UCL's official announcement, lead author Lucy Corrigan described the findings this way: "These findings offer early evidence that medication may play an important role in supporting the health of people with ADHD. However, the studies underlying this are few, and further research is needed to clarify the association." She also stressed that medication is only one part of support, and that comprehensive, long-term treatment and support are important for many people with ADHD. The results are thus an important clue for considering the role of medication, but they do not guarantee that taking medication reduces mortality risk by 25%.

Comparison With the 2024 Swedish Nationwide Study

One important study helps in understanding the association between medication and mortality risk shown by the meta-analysis: the Swedish national registry study by Lin Li and colleagues, published in JAMA on March 12, 2024 (JAMA. 2024;331(10):850–860, DOI: 10.1001/jama.2024.0851). The study covered 148,578 people aged 6 to 64 who were newly diagnosed with ADHD between 2007 and 2018, with a median age at diagnosis of 17.4 years. To reduce bias when estimating treatment effects from observational data, the team used a method called "target trial emulation."

Target trial emulation is a method in which, even without conducting an actual randomized controlled trial, the starting point, eligibility criteria, and treatment group definitions of a hypothetical trial are specified in advance, and observational data are analyzed accordingly. Li and colleagues compared starting ADHD medication within 3 months of diagnosis with not starting it, and evaluated mortality risk over the following two years. The design aims to reduce bias arising from differences in treatment start times and patient backgrounds, compared with simply comparing death rates between medicated and unmedicated people. Treatment initiation was not actually randomized, however, and the study remains observational.

The analysis found that the group that started medication had an all-cause mortality hazard ratio of 0.79 (95% CI: 0.70–0.88), significantly lower than the group that did not. By cause of death, an association was seen for external causes such as suicide and accidents, with a hazard ratio of 0.75 (95% CI: 0.66–0.86). For natural death from physical illness, the hazard ratio was 0.86 (95% CI: 0.71–1.05), and no statistically significant difference was found. In other words, the reduction in all-cause mortality in this study was explained mainly by the association with fewer external-cause deaths.

Li and colleagues' study is especially informative because it reports absolute as well as relative risk. Over the two-year follow-up, all-cause mortality was 48.1 per 10,000 in the group that did not start medication and 39.1 per 10,000 in the group that did. The risk difference between the groups was −8.9 per 10,000 (95% CI: −17.3 to −0.6): a relative reduction of about 21%, but an absolute difference of about 9 per 10,000 over two years. Presenting the relative reduction alongside the absolute difference in deaths makes the scale of the result easier to grasp.

The hazard ratio of 0.75 from Corrigan and colleagues' 2026 meta-analysis and the 0.79 from Li and colleagues' study give similar estimates of the association between medication use or initiation and lower all-cause mortality. As noted above, however, Li and colleagues' study is included in Corrigan and colleagues' meta-analysis, so the two cannot be taken as independently reaching the same conclusion. Moreover, while the Swedish study assessed treatment initiation within two years of new diagnosis, the meta-analysis pools studies with differing age groups, follow-up periods, and definitions of medication use. The similarity of the estimates helps in understanding the direction of the results, but careful verification is still needed before concluding that a long-term mortality-preventing effect of the drugs has been established.

Background on Shortened Life Expectancy and Public Health Challenges

The link between ADHD and survival was not first reported in this study. Professor Joshua Stott, a co-author of the paper, and colleagues on the UCL team published a study in the British Journal of Psychiatry in 2025 using the large UK electronic health records database CPRD (DOI: 10.1192/bjp.2024.199). That study compared the life expectancy of adults diagnosed with ADHD with that of a comparison group without a diagnosis. It estimated a reduction of about 6.78 years (95% CI: 4.50–9.11) in men and about 8.64 years (95% CI: 6.55–10.91) in women. These values, however, were calculated from data on the diagnosed population studied and do not mean that everyone with ADHD will lose the same number of years.

The new JAMA Network Open study builds on this earlier research on differences in life expectancy and seeks, using international data, to clarify which causes of death are involved in the elevated mortality risk. In UCL's announcement, Professor Stott explained: "In our earlier research, we found that life expectancy is reduced in people with ADHD. This study aims to understand what lies behind that worrying statistic." He also noted that risk was elevated not only for suicide and injury but also for death from physical illness, indicating that the health inequalities faced by people with ADHD are wide-ranging. The elevated mortality risk, in other words, cannot be explained solely by accidents linked to impulsivity or inattention.

The research team raises several possibilities for why risk of death from physical illness is higher. One is an association with behaviors that affect health, such as smoking, heavy drinking, and substance use. Prior research has also pointed out that physical health problems such as cardiovascular and metabolic disease are more common among people with ADHD. These factors may affect long-term health and, as a result, lead to higher mortality risk.

Another important issue is access to health care and continuity of treatment. Attention and executive function traits associated with ADHD can make it difficult to manage appointments, attend regular visits, or maintain medication schedules for chronic conditions. If systems and support for using health services do not accommodate these traits, people may fail to receive necessary treatment or may drop out partway through. Financial hardship, social isolation, and stigma may compound this and contribute to worsening health.

These explanations, however, are not causal pathways confirmed by this meta-analysis. The data cannot be used to calculate separately how much factors such as smoking, drinking, health care access, and treatment continuity contribute to mortality risk. What the research team presents are hypotheses to be tested, based on prior research and clinical knowledge. Future work will therefore need not only to identify factors associated with elevated mortality risk but also to examine what kinds of support can narrow these health inequalities.

Co-senior author Dr. Amber John stresses that for people with ADHD to live long, healthy lives, it is important that they have equitable access to evidence-based assessment and treatment. Professor Stott likewise explained that ADHD is not a condition that inevitably shortens life, and that with appropriate support and treatment, long and healthy lives are possible. He also noted that although ADHD diagnosis rates have risen in recent years, diagnosis may not have kept pace with actual prevalence, particularly among older age groups. In considering mortality risk, what is needed is not only treatment after diagnosis but also systems for identifying people who need support and connecting them to ongoing medical care.

Unresolved Confounders and the Danger of Overgeneralizing Mortality Risk

While this large study shows the need to take a broader view of the health problems of people with ADHD, it also calls for caution in how the statistics are received. In particular, it is not appropriate to interpret the figure of "about 70% higher all-cause mortality risk" as determining the future of an individual diagnosed with ADHD. It is a relative measure calculated from observational studies in different countries, age groups, and health care settings, not a prognosis reflecting an individual's health, living environment, comorbidities, or treatment and support. Communicating the information without this distinction risks fueling unnecessary anxiety and social stigma about ADHD.

Among the study's limitations, confounding by comorbidities and socioeconomic conditions has not been fully removed. People with ADHD may have co-occurring psychiatric conditions such as depression, bipolar disorder, and substance use disorders, which may influence mortality risk. Socioeconomic hardship, lack of access to health care, and long-term stress may also shape health. This analysis alone cannot clearly separate how much of the elevated mortality risk is related to ADHD traits themselves and how much is attributable to comorbidities and the social environment.

In future research, it will be important to standardize, as far as possible, the definition of ADHD, the classification of causes of death, and the methods for adjusting for confounders. With heterogeneity between studies this high, it is necessary not only to present a single pooled hazard ratio but also to examine in detail how the strength of the association varies by age, sex, comorbidity, and health care setting. For the effects of medication, verification that limits the influence of confounding is needed, such as more refined target trial emulation or analyses that exploit changes when treatment is started, stopped, or resumed. Even with such methods, the limitations of observational research do not disappear entirely, and results from multiple research approaches will need to be evaluated carefully.

What this study has shown is that there is a non-negligible association between ADHD and mortality risk, and that the background extends beyond suicide and accidents to a wide range of issues, including physical illness and access to health care. There is no shortage of challenges that can be addressed to improve outcomes: environments that prevent accidents, early intervention in mental health crises, and health care systems that support ongoing management of chronic disease. How much such interventions can reduce mortality risk, however, remains to be tested. What matters is not to define the lifespans of people with ADHD by statistics, but to determine which medical care and social support are effective in narrowing the health inequalities that have been identified.