Can You Trust That Medical Headline? A Guide for Families

Can You Trust That Medical Headline? A Guide for Families

Watch on YouTube

The episode is about medical literacy, not eldercare specifically

A headline says a treatment cuts risk in half. A supplement is tied to a lower rate of dementia. A product is “recommended by nine out of ten doctors.” A family member sends the article and asks, “Should we do this?”

Episode 9 of the AI and Healthcare Podcast explains how to slow that decision down. In the conversation, recorded June 2, 2026, Noah Vandal asks Dr. Joseph Yoon how physicians evaluate medical studies and health reporting.

The episode is a broad discussion of medical literacy rather than eldercare specifically. This companion guide applies its lessons to families supporting older adults, who may be balancing multiple conditions, medications, side effects, costs, function, and personal goals.

The answer is not to dismiss every headline. It is to identify what the evidence actually supports before applying a population-level result to one person.

Association can be an important clue without proving cause

An observational study looks for patterns without randomly assigning people to an intervention. It may find, for example, that people who drink coffee have a different measured rate of dementia from people who do not.

That association can be worth studying. It does not prove coffee caused the difference. The groups may also differ in sleep, income, activity, age, social connection, healthcare access, or other factors that affect the outcome.

The coffee and dementia example in the episode is illustrative, not a result from an identified study. Its lesson is broadly useful: “associated with” does not mean “prevents,” “causes,” or “will work for your parent.”

Families should be especially careful when an observational result becomes a recommendation to buy a supplement, eliminate a food, or change a routine. The study may support a question for the healthcare team without supporting the advertised action.

The same distinction is important when reading metabolic-health claims like those discussed in [Episode 5 on insulin resistance and healthy aging](/resources/podcast-episode-05-insulin-resistance-healthy-aging).

Randomization strengthens evidence without creating certainty

In a randomized controlled trial, chance determines which intervention participants receive. That process helps distribute known and unknown differences across groups on average, making it easier to test whether the intervention contributed to the outcome.

A control group provides a comparison. A placebo may help separate the treatment's effects from expectations or other aspects of receiving care. Blinding can reduce the chance that participants, clinicians, or outcome assessors influence results because they know the assignment.

These methods are powerful but not universal. A diet, surgery, exercise program, or caregiving intervention may be impossible to hide from participants. Withholding effective care can be unethical. A long-term study may also struggle when participants stop following the assigned plan.

Randomization does not make a trial infallible. Families should still ask how many people completed it, how many events occurred, how outcomes were measured, how long follow-up lasted, and whether the participants resemble the person they support.

Ask for absolute numbers before reacting to a percentage

Relative risk describes the proportional difference between groups. Absolute risk describes the actual difference in event rates.

Episode 9 uses a hypothetical change from two events per 1,000 people to one event per 1,000. That is a 50% relative reduction because one is half of two. The absolute reduction is one event per 1,000 people, or 0.1 percentage point.

“Cuts risk by 50%” is therefore incomplete without the original risk. But the opposite shortcut is also wrong: a small absolute difference is not automatically meaningless. Preventing a rare, severe outcome may be worth substantial effort, while a larger change in a minor outcome may matter less.

The decision depends on the seriousness of the outcome, uncertainty, possible side effects, treatment burden, cost, follow-up period, and the person's baseline risk. Those are exactly the details a clinician can help place in context.

Evidence must fit the older adult being supported

Even a strong study may not answer a particular family's question. Before applying its result, compare the study population with the older adult involved:

- Were people of a similar age included? - Did participants have similar health conditions and medications? - Were frailty, cognition, mobility, or kidney and liver function relevant? - Was the measured outcome meaningful to the person's daily life? - Did the study last long enough to observe benefits and harms? - Would the treatment burden fit the person's goals and preferences?

This is not a reason to exclude older adults from evidence-based care. It is a reason to be honest about how directly the available evidence applies.

The boundary becomes even more important when cognition is changing. Our [family guide to dementia, memory, and AI](/resources/podcast-episode-07-dementia-ai-family-care) discusses why support decisions should preserve the person's voice and keep people responsible for diagnosis, treatment, and safety.

One positive study is not the whole evidence base

A result can be statistically significant and still be a false positive, an overestimate, or specific to one set of conditions. Replication asks whether other researchers can obtain a compatible result with comparable methods.

The episode discusses a large psychology replication project in which many selected findings did not reproduce under the project's criteria. It also refers to an effort involving selected preclinical cancer research. The description closely matches a report from Amgen researchers who said they confirmed 6 of 53 findings, although the company and paper are not named in the transcript.

Publication bias can make the visible record look stronger than the full evidence. Positive, surprising results may be more likely to be submitted, published, promoted, and shared than null or negative studies. A family may therefore encounter five articles about the apparent success and none about the unsuccessful attempts.

Look for independent replication, trial registration, systematic reviews, clinical guidelines, and the complete evidence record—not only the newest paper.

AI can summarize a paper but cannot make the care decision

This is an application for Good Company readers, not a claim made in the episode: an AI assistant can help define study terms, turn a paper into plain language, or organize questions for an appointment. It can also omit a limitation, confuse relative and absolute risk, or invent a source.

A useful AI summary should link to the original material, distinguish the paper's findings from its own explanation, state uncertainty, and make it easy to verify important numbers. It should not tell an older adult to start, stop, or change treatment based on a headline.

The boundaries in our guide to [AI assistants in senior care](/resources/ai-assistant-boundaries-senior-care) apply here: the tool can support understanding, while healthcare professionals and the person receiving care remain responsible for clinical decisions.

Use a family checklist for the next dramatic headline

Before acting, ask:

- Can we find the original study rather than only the news story? - Was it observational, randomized, or another kind of research? - How many people participated, and how many actual events occurred? - What were the starting and ending risks? - Were harms and treatment burdens measured alongside benefits? - How uncertain was the estimate? - Does the study population resemble our family member? - Has another team reproduced the result, or is it included in a systematic review? - Who funded the study, and were conflicts disclosed? - What does the person's healthcare team think the evidence means for them?

The takeaway from Episode 9 is calibrated confidence. A dramatic headline can point toward useful research, but it is not a care plan. Read beyond the percentage, look for the full evidence, and bring the decision back to the older adult and the people responsible for their care.

Common questions

Does an observational study prove that a treatment or habit works?

No. It can show an association and provide an important clue, but other differences between the groups may explain the result. An observational study does not by itself prove that changing the habit or treatment will change an individual's outcome.

Is a randomized controlled trial always definitive?

No. Randomization reduces confounding on average, and controls or blinding can reduce other biases. A trial may still be limited by adherence, attrition, measurement, event counts, follow-up, or a study population that does not resemble the older adult considering the intervention.

What does a 50% risk reduction mean for a family decision?

Ask for the starting and ending rates. A change from two events per 1,000 people to one per 1,000 is a 50% relative reduction but an absolute reduction of one event per 1,000. Its importance depends on the outcome, uncertainty, side effects, cost, follow-up, and the person's baseline risk.

How should a family respond to a dramatic health headline?

Find the original study and look at its design, event counts, absolute effects, harms, uncertainty, population, replication, funding, and conflicts. Then ask the person's healthcare team whether the evidence applies to their health, medications, goals, and circumstances before changing care.