Muddled answers on routine questionnaires were linked to later dementia in eight aging studies
Across eight long-running studies with more than 76,000 older adults, people whose questionnaire answers were inconsistent or unusual were more likely to develop dementia in later years. The links were modest and the study is observational; the authors see it as a research tool, not a screening test.
Large studies of aging ask their volunteers to fill in long questionnaires about mood, personality, quality of life and much more. Researchers usually care about what people answer. A team at the University of Southern California looked instead at how they answer: whether they skip questions, contradict themselves, or give a pattern of answers that doesn't hang together.
Across eight studies and 76,350 people, those whose answers were more erratic or inconsistent were more likely to develop dementia (a loss of memory and thinking severe enough to disrupt daily life) in the following years. The associations were modest, and as in every observational study, they show a link, not a cause.
Dementia usually builds up over many years before anyone diagnoses it. Filling in a questionnaire takes reading, working memory and the ability to stick to the same answer format from one question to the next. So small slips in that task might be an early, everyday sign of trouble. An earlier study had found such a link in one American study, the Health and Retirement Study. The new question was whether it holds up in very different groups of people.
What they did
The team pooled data from eight long-running studies of aging in Australia, the US, England, Ireland, Sweden and across Europe. Everyone included was 50 or older, free of dementia at the start, and had at least one later check of their dementia status. The studies ranged from 600 to 38,376 people, and the longest followed people for 23 years. In total the team analysed almost 5 million individual answers.
From each person's first questionnaire, the team computed five kinds of unusual answering, using only sets of questions that measure the same thing (for example, several items on well-being):
- Skipped questions (on average, 3.3% of items were left blank).
- Random errors: answers that jump around within a set of related questions.
- Outlier patterns: a profile of answers very unlike everyone else's.
- Incompatible answers: combinations that don't fit the expected structure of a scale, measured with several statistical scores.
- Extreme answering: a habit of always picking the ends of the scale.
They then checked, in each study separately, who later developed dementia, taking age, sex and education into account and allowing for the fact that some people died first. Finally, they combined the eight results in a meta-analysis, a statistical way of pooling studies.
What they found
- Four of the five patterns were linked to later dementia. For each step up in a pattern (one standard deviation, a statistical unit of spread), the risk of dementia was 6% higher for skipped questions, 25% higher for random errors, 31% higher for outlier patterns, and 22% to 25% higher for incompatible answers.
- Extreme answering was not linked to dementia. The authors read this as a sign that the signal comes from trouble keeping answers coherent, not from personal style.
- Pooled over all measures and studies, the link was modest but statistically significant: about 17% higher risk.
- Some of the signal went beyond memory tests. When the team also took into account people's scores on standard thinking tests at the start, the links became weaker, but random errors, outlier patterns and incompatible answers still predicted dementia.
- The links were stronger in people under 75 and in more educated people. The authors suggest that answer patterns may be most informative early on, before ordinary tests pick up a decline.
- There was little difference between men and women.
What to keep in mind
- Dementia was defined differently in each study: some used test scores plus reports from relatives, some test scores alone, some self-reported diagnoses. These were research classifications, not clinical diagnoses, and the share of people who developed dementia ranged from 1.38% to 39.00%.
- The results varied a lot between studies in size, although they mostly pointed the same way.
- Drop-outs and deaths were handled differently across studies and may have hidden some cases.
- It is unclear what the answer patterns reflect. Problems with attention or planning, poor eyesight or hand movement, or simply losing interest in a long questionnaire could all play a part.
- The links are modest. The authors say plainly that the main value is for research, not as a stand-alone screening test for individuals.
- The study was not preregistered (its analysis plan was not published in advance).
Why it matters
Many large studies of aging hold decades of questionnaires but have little detailed testing of thinking skills. This work suggests that the questionnaires themselves carry a faint early signal of dementia, which researchers could use at no extra cost or burden to the volunteers. It does not mean that skipping a question on a form is a warning sign for any one person. The authors also point out that today's online surveys record more, such as how long each answer takes, which might add further information in future studies.