Is Your Parent's Dementia Being Missed? A Singapore Study

You've noticed the repeated questions, the missed appointment, the story told twice in one phone call. You're telling yourself it's normal aging, and then Googling to check if that's actually true. Underneath both of those is the real question: if something were wrong, would anyone actually catch it?

Singapore has an answer, because one study has actually measured this, twice, a decade apart. The IMH-led WiSE study found dementia prevalence among older adults barely moved between 2013 and 2023, a change so small it isn't statistically significant. But the share of dementia cases going undiagnosed dropped sharply, from 7 in 10 in 2013 to about half in 2023. That's the number that should change how you read your parent's memory lapses, not because dementia got rarer, but because it got harder to miss.

Here's what the study actually found, why the "improvement" isn't the one you'd guess, and what a single number can and can't tell you about your own parent.

The Study That Actually Measured This

The Well-being of the Singapore Elderly (WiSE) study is run by Singapore's Institute of Mental Health, led by Assoc Prof Mythily Subramaniam, and it's the only Singapore-specific study that has measured dementia prevalence and diagnosis rates twice using the same method a decade apart. The comparison was published in *Alzheimer's & Dementia* in February 2025: Subramaniam et al., "Prevalence of dementia in Singapore: Changes across a decade", DOI 10.1002/alz.14485.

Both waves used the same 10/66 protocol, an internationally validated diagnostic tool built specifically for population surveys outside the wealthy countries where most dementia research happens. WiSE 2013 surveyed 2,565 Singapore residents aged 60 and above, plus 2,421 of their informants. WiSE 2023 surveyed 2,010 older adults and 1,798 informants. Same sampling method, same questionnaires, same diagnostic criteria, run independently a decade apart on a fresh, randomly selected sample.

The Two Waves, Side by Side

Two numbers moved. Only one of them moved in a way you can actually trust.

The Prevalence Number Didn't Really Change

Dementia prevalence went from 10.0% to 8.8%, a 1.2 percentage point drop. The study's own authors call this non-significant, and the confidence intervals show why: 8.7–11.5% in 2013 overlaps almost entirely with 7.3–10.7% in 2023. Statistically, that's not a decline. It's noise around a number that hasn't moved.

The paper is explicit about this. Comparing the two waves, the researchers reported "a non-significant decrease of 1.2 percentage points, or 12% (95% CI: −1.1 to 3.5), was observed (p-value = 0.298)." A p-value of 0.298 is nowhere near the 0.05 threshold researchers use to call a result real. Their own conclusion states it plainly: "the lack of a significant change could mean that there is no change in the prevalence of dementia."

So read this correctly: dementia in Singapore's older population is roughly as common today as it was in 2013, not rarer, not more common, just about the same, within the margin of what this kind of survey can detect. If you were hoping the data would tell you dementia is becoming less likely, it doesn't say that. If you were worried it's becoming more common, it doesn't say that either.

The Number That Did Move: Diagnosis

Here's what's sharp instead of flat. The share of dementia cases going undiagnosed dropped from 70.6% in 2013 to 51.5% in 2023, a statistically significant change (p = 0.002). The researchers define "undiagnosed" precisely: someone who met the 10/66 criteria for dementia during the survey but had never been told by a doctor that they had dementia or a memory problem.

In 2013, roughly 3 in 10 people who met the clinical criteria for dementia had actually been diagnosed. By 2023, that had climbed to roughly half. The study's authors put it directly: "about 50% of individuals with dementia were diagnosed by a doctor as having dementia/memory problems, a sharp contrast to the previous study, where only approximately 30% of those assessed as having dementia during the survey were diagnosed by a doctor." They attribute the shift to the memory clinics added to primary care since 2013 and to Singapore's broader dementia-awareness push, including the CREST outreach teams that work with at-risk older adults in the community.

Even at 51.5%, Singapore isn't unusual. A separate systematic review of 23 international studies found a pooled global rate of undetected dementia at 61.7% (95% CI 55.0–68.0%), with detection rates worse in China and India than in Europe and North America. Singapore in 2023 sits meaningfully better than that global pooled average, but "about half go undiagnosed" is still the honest headline, not "diagnosis is solved."

What "Single Cross-Sectional Survey" Actually Means

Both WiSE waves are single-phase, cross-sectional surveys, not a cohort followed over time. The researchers say this themselves in their own limitations section: "Because the survey has a cross-sectional design, we are unable to establish any temporal relationships between dementia, stroke, and depression."

In plain terms, WiSE didn't track the same 2,565 people from 2013 through to 2023 to see who developed dementia and who didn't. It surveyed one fresh, randomly sampled group of older adults in 2013, and a different, separately sampled group in 2023, and compared the two snapshots. That's a legitimate and rigorous way to track a population-level trend. It's not the same thing as following individuals over time, and it can't tell you anything about any single person's trajectory, including your parent's.

There's a second limitation worth knowing. A doctor's diagnosis in this study was based on what the older adult or their informant reported, not on medical records. The researchers couldn't verify it against administrative health data, because participant identities weren't linked to that data for privacy reasons. The "51.5% undiagnosed" figure is a well-designed estimate from a large, randomly sampled survey. It is not a registry count.

Why We Won't Give You a Combined Number

You might expect the next step here is to multiply Singapore's older-adult population by 8.8%, or by 51.5%, and hand you a headline number: "X thousand Singaporean families have an undiagnosed case of dementia right now." We're not doing that, and it's worth explaining why, because the study's own authors did run that multiplication for their own population-planning purposes, and it isn't the same thing as a claim you can safely reuse.

The paper reports that applying the 8.8% weighted prevalence to Singapore's 2022 resident population aged 60 and above works out to an estimated 73,918 older adults with dementia. That's a legitimate exercise for health system planning, done by the researchers themselves, with full access to the survey's sampling weights and the exact population figures the weights were built against. Reusing a percentage from a stratified, weighted survey against a separate, later population estimate, outside the study's own analysis, isn't the same calculation. It's a new one, and nobody has verified it here.

The same logic applies to the undiagnosed-rate figure. 51.5% is a share of dementia cases within the surveyed sample, not a share of Singapore's whole population, and turning it into "X thousand undiagnosed cases" requires a chain of assumptions about how the 2023 sample's demographics map onto today's actual population, years after the survey closed. We also won't net either WiSE figure against unrelated numbers, like the cost of caregiving reported elsewhere on this site. Diagnosis rate and caregiving cost are different questions, measured in different studies, on different groups, and stitching them together to create a bigger, scarier number is exactly the kind of manufactured statistic this piece is trying to avoid.

What we can tell you, cleanly: dementia prevalence in Singapore's older population hasn't measurably changed in a decade, and the odds that a case gets caught by a doctor have gotten meaningfully better, though still short of the majority you'd want. Both of those are true on their own. Neither needs to be inflated into something bigger to be worth acting on.

What This Means for You, Right Now

The data doesn't tell you whether your specific parent has dementia. It tells you that the system catching cases has gotten better, but still misses about half, and self-diagnosis by Googling symptoms isn't a substitute for either outcome.

If you're noticing repeated changes, not a single forgetful day but a pattern over weeks or months, the useful next step isn't more searching. It's asking your parent's GP for a cognitive screen, or calling the Agency for Integrated Care (AIC) Hotline at 1800-650-6060 (Monday to Friday 8:30am–8:30pm, Saturday 8:30am–4pm) to ask about the nearest memory clinic. Getting assessed doesn't commit you to anything. It just moves your parent from the "undiagnosed" half of that statistic into the half where there's an actual answer, and a plan.

If that assessment does lead to a diagnosis, the next question families usually ask is what care actually costs and what subsidies apply. Our dementia care cost guide walks through the real numbers by care setting, not a median that may not apply to your household.

And if you're thinking further ahead, about what happens financially if a parent's condition progresses to needing daily help, it's worth understanding what CareShield Life does and doesn't cover before you're in the middle of a crisis. You can review that on your own, or if you'd rather talk it through, you may want to speak with CareCompare's MAS-licensed financial adviser partner about your family's specific situation. This is general information, not personalized financial advice. (CareCompare may earn a referral fee if you proceed with a plan or adviser through our tool. This does not affect what you pay.)