What the Autism Life-Expectancy Headline Leaves Out

A 38-year-old autistic patient brings a headline to a follow-up appointment. She has read that autistic people die roughly 14 years earlier than everyone else. She has been calculating what that might mean for seeing her daughter grow up.

“Does this mean I probably won’t make it to retirement?”

The psychiatrist starts to explain that the finding concerns a population. The patient has a more immediate question: “Am I supposed to plan my life differently?”

In this fictional composite, a number has become a forecast, and the forecast is beginning to organize the patient’s future.

The clinician owes her an explanation of the number and a useful response to the health concerns behind it. The research contains more than a lifespan estimate. It also describes causes of death that deserve attention, provided we distinguish the recorded outcomes from our hypotheses about why they occurred.

The September 10 JAMA Network Open research letter analyzed 2,048,046 Medicaid beneficiaries identified with autism spectrum disorder. It estimated life expectancy at birth at 64.9 years, with a 95% confidence interval of 64.6 to 65.1 years.

The reported difference depended on the comparison:

Each estimate answers a different comparison question. Neither isolates years of life caused to be lost by autism.

The authors acknowledge that Medicaid enrollment selects for income and disability, and that autistic beneficiaries may differ from autistic people outside the program. Extending the estimate to every autistic American exceeds the population studied.

The clinically usable sentence includes the group studied and the group used as a reference. Leaving either out invites a different question: “What does an autism diagnosis do to a person’s lifespan?” A large population comparison cannot, by itself, answer that causal and personal question.

Choose the comparator because it fits the question being asked. Choosing the smaller gap to reassure a patient, or the larger one to strengthen an advocacy message, puts the intended effect ahead of the meaning of the result.

The narrow confidence interval reflects statistical precision within the analysis. Applicability to a particular patient remains a separate issue. Likewise, subtracting the two gaps produces a difference between reference estimates; it does not divide the result into an “autism effect” and a “Medicaid effect.”

The Office for National Statistics distinguishes period life expectancy from the experience of a birth cohort. A period life table combines mortality rates at different ages during a specified period. It describes what would happen if those rates continued across a lifetime. It uses people of different ages to construct that summary, rather than observing one group of newborns until everyone has died.

Life expectancy at birth also answers a different question from remaining life expectancy at a person’s current age. Someone who has reached 38 has already survived the earlier ages represented in the birth estimate. Subtracting 38 from 64.9 would not give a valid estimate of her years remaining.

The supplement clarifies the construction. The researchers pooled autism-group data from 2000 through 2020 and used 2010 reference data for the overall comparison. The Medicaid comparison included beneficiaries regardless of autism status. Infants were excluded from the analytic enrollee population, and corresponding national infant-mortality rates were substituted in the life tables because of limitations in the Medicaid data.

The study’s own estimates changed over time: from 63.9 years in 2000 through 2004 to 65.6 in 2015 through 2019, an increase of 1.7 years. In 2020, the estimate fell 4.1 years relative to 2015 through 2019. The comparison gaps persisted.

That movement helps explain what a period estimate measures. It reflects mortality conditions in the periods being studied, and may also reflect changes in who is identified in the data. It supplies neither a fixed biological limit nor a guaranteed trajectory of improvement.

For the patient asking about retirement, the explanation can be brief:

“This number summarizes mortality in a particular population using historical data. It cannot calculate how many years you have left. Your age, health, circumstances, and future care all matter to the conversation we need to have.”

Diagnosis-based databases raise a further question: who had a diagnosis recorded, and who did not?

An independent UK matched-cohort study by O’Nions and colleagues estimated life expectancy from age 18 and analyzed people with and without recorded intellectual disability separately. Both groups had shorter estimated life expectancy than their matched comparison groups. The estimated gaps were larger in the group with recorded intellectual disability, particularly among women.

The authors also emphasized incomplete diagnosis ascertainment. Diagnosed adults, especially older adults, may have greater support needs and more co-occurring conditions than autistic adults overall. They cautioned that this could overstate the average gap. Those limitations coexist with evidence of premature mortality among diagnosed people.

That concern raises a question for the US analysis: how well do the older ages in a historical claims database represent the people receiving autism diagnoses today? If older adults with fewer support needs are less likely to have a recorded diagnosis, the mortality rates available to the life table may describe a selected group.

This is a possible source of bias, not a finding established by the Medicaid paper. Its supplement does not provide the age-by-diagnosis-history and support-needs breakdown needed to test that explanation. We cannot assume that its older participants were predominantly people diagnosed decades earlier, or calculate how much this issue changed the estimate.

The UK results also argue against an easy reassurance: “You work, speak fluently, and live independently, so this is not about you.” Absence of recorded intellectual disability did not eliminate the group-level disparity in that study. A person’s actual health concerns deserve assessment regardless of how independent they appear.

Equally, someone who needs substantial daily support deserves a future that is larger than a population statistic. Individualization means learning what is true for the person, including what has been difficult for them to communicate or obtain care for.

On September 15, demographer Vladimir Canudas-Romo questioned whether the analysis adequately accounted for age at diagnosis. On September 17, lead author Guohua Li defended the approach, citing autism’s early developmental origin, unavailable diagnosis-age data, and comparability with conventional life tables.

The exchange leaves a methodological disagreement on the record. Neither comment independently settles the validity of the estimate. It does, however, make the question being estimated worth keeping visible: a conventional population life table and survival after a recorded diagnosis are different quantities.

A psychiatrist can explain that distinction without resolving the demographic dispute during a follow-up visit. The immediate task is to prevent a population estimate from becoming a personal deadline, then examine what the rest of the findings can contribute to care.

The overall age- and sex-standardized mortality ratio was 1.44 (95% confidence interval, 1.41 to 1.46). In other words, approximately 44% more deaths occurred than expected using the 2010 Medicaid population’s age- and sex-specific rates.

Selected underlying causes showed substantial relative excesses:

Relative excess and observed deaths answer different questions. Influenza has the largest ratio among these selected rows; pneumonia has the most recorded deaths. Recorded deaths also include those expected at reference rates, so neither column measures excess deaths directly. Neither column, on its own, tells a clinician which problem is most important for a particular patient. These are selected categories, not a complete inventory or a ranking of contributions to the life-expectancy gap.

The next distinction is between an outcome and its pathway. A malnutrition code does not establish avoidant/restrictive food intake disorder. An accidental drowning code does not reveal whether wandering, a seizure, medication effects, or another circumstance was involved. Pneumonia and pneumonitis due to solids and liquids are separate rows; combining them as medication-related aspiration deaths would add a causal claim the analysis did not test.

The useful response is to turn those possibilities into questions that can be answered for the person in care.

NICE’s adult-autism guidance provides an independent foundation for that work. It advises attention to under-recognized physical disorders, food preferences and sensory needs, communication, and participation in decisions. It also advises considering physical and mental conditions when evaluating changes in behavior.

Swallowing and alertness. When relevant to the history, ask about difficulty swallowing, coughing or choking with meals, and troublesome sleepiness. Review symptoms alongside the medication regimen and its timeline. Independently of the mortality study, risperidone’s prescribing information warns about swallowing dysfunction and aspiration associated with antipsychotics, as well as somnolence and potential cognitive or motor impairment. Those warnings support attention to symptoms during prescribing; they do not establish that risperidone caused the respiratory deaths in this dataset.

Nutrition and eating. Ask about changes in intake, food range, weight, and the practical ability to obtain and prepare acceptable food. Determine whether a reported difficulty calls for medical, nutritional, or eating-disorder assessment. NICE specifically advises attention to unusual food preferences and, when necessary, support for referral to a general practitioner or dietitian. A death-certificate category should prompt curiosity about the patient’s circumstances, rather than supply a psychiatric diagnosis.

Injury and safety. Start with any actual history of falls, near-drowning, getting lost, or seizures. Where a concern exists, clarify the circumstances, the person’s priorities, and any current safety plan. Medication-related impairment belongs in that review when clinically relevant. A population association alone supplies no basis for presuming that every autistic adult wanders or needs restrictions on independent activity.

Infection prevention and access. The respiratory findings give a reason to ask whether the patient can obtain routine preventive care and prompt assessment of new physical symptoms. Clarify who reviews vaccination needs and whether appointments or communication have become obstacles. The study did not test a vaccination strategy or estimate how many of these deaths a specific intervention would prevent.

These are proposed applications of the findings alongside existing guidance and prescribing information. Decisions follow the patient’s symptoms, history, preferences, and clinical assessment.

Access becomes concrete at the point of follow-through. If a swallowing assessment is indicated, can the person arrange and tolerate the visit? Would written information in advance help them describe the problem? Does a sensory adjustment make the appointment usable? Involve a supporter when the person wants that help, while continuing to address the patient directly.

“Discuss with primary care” leaves one essential task unfinished: establishing what happens next. Where coordination is agreed and appropriate, identify the concern, the receiving clinician, the patient’s role, and when its status will be checked.

The opening question deserves a direct answer before a list of possible health concerns. Ask what the headline has come to mean for the patient. Has it changed her plans, increased anxiety, or made an existing symptom feel more frightening?

In the fictional composite, the psychiatrist might close with:

“I understand why that headline frightened you. It does not tell us when you will die, and I would not use it to plan your retirement or your daughter’s future. The study does raise health concerns worth understanding. Let’s identify whether any apply to you and agree on what needs attention.”

A useful note can record the patient’s interpretation, the explanation provided, any symptoms or access barriers identified, and the agreed follow-up. If the encounter reveals no new concern requiring action, it can say that without manufacturing a risk estimate or a referral.

The headline compresses several different tasks into one number: describing a disparity, explaining its causes, predicting a person’s future, and deciding what care should change. Each requires its own evidence. The clinical opportunity is to give the patient a more accurate understanding of the research and a specific next step when one is needed.

Li G, DiGuiseppi CG, Blanchard A, Russell MT, Ing C. “Autism Spectrum Disorder and Life Expectancy Among Medicaid Beneficiaries.” JAMA Netw Open. 2026;9(9):e2633251. doi:10.1001/jamanetworkopen.2026.33251. Includes Supplement 1 and the September 15 and September 17, 2026 comments.

Office for National Statistics, “Period and cohort life expectancy explained”.

O’Nions and colleagues, “Estimating life expectancy and years of life lost for autistic people in the UK: a matched cohort study”, The Lancet Regional Health - Europe, doi:10.1016/j.lanepe.2023.100776.

NICE, “Autism spectrum disorder in adults: diagnosis and management,” CG142, particularly recommendations 1.1.1 through 1.1.9, 1.1.15, 1.2.10, and 1.5.1.

RISPERDAL prescribing information, revised May 2026, sections 5.8, 5.10, and 5.12, via DailyMed.

Sources reviewed September 23, 2026.

Educational Disclaimer: The Psychiatric Record provides general educational information for psychiatric and mental-health professionals. Content does not constitute medical, legal, regulatory, compliance, billing, or other professional advice; does not establish a standard of care; and is not a substitute for independent professional judgment. Readers should consult the requirements that apply to their own practice. This article does not estimate any individual’s lifespan or provide an individualized screening, prevention, or treatment plan. The clinician and patient described are fictional composites and do not depict any real clinician or patient.