The answer, the researchers argue, is that healthcare-seeking for cognitive symptoms is rarely a solo act. When an older adult begins forgetting appointments, repeating questions, or losing their way in familiar places, the decision to visit a clinic typically emerges from a negotiation between the person experiencing the symptoms and a spouse, adult child, or other close family member. The older adult may minimize the problem, attribute lapses to normal aging, or fear a dementia diagnosis. The caregiver may notice the changes first, worry privately, and then weigh how to raise the subject without causing offense. Existing measurement instruments, the team noted, focus largely on decisional capacity or on engagement with treatment once a diagnosis exists. None were designed to capture the pre-diagnostic decision-making process from both perspectives simultaneously. That gap meant that researchers and clinicians had no standardized way to identify where, precisely, families get stuck on the road to a memory clinic.

To build the new instruments, the research team followed a rigorous multi-phase scale development protocol that is considered the gold standard in psychometric science. The process began with a concept analysis of healthcare-seeking decision-making, combined with a systematic review of the literature, to define the theoretical territory the scales needed to cover. The team then conducted qualitative interviews with 19 dyads, each consisting of an older adult screened as having MCI and their family caregiver, between December 2024 and January 2025. These interviews allowed the researchers to hear, in the participants’ own words, how symptoms were first noticed, what doubts and disagreements arose, and what finally prompted or delayed a clinic visit. From this material, an initial pool of candidate items was drafted, each anchored in the lived experience of the very population the scales are intended to serve.

Refinement came next through a Delphi consultation, a structured method in which a panel of experts independently rates and comments on draft items over successive rounds until consensus emerges. Two rounds of Delphi consultation were conducted with 17 multidisciplinary experts, who evaluated the relevance, clarity, and coverage of each candidate item. Their feedback trimmed and sharpened the item pool, ensuring that every question that survived reflected both theoretical soundness and practical relevance to the clinical realities of cognitive aging. Only after this exhaustive preparatory work did the team move to the decisive phase: large-scale psychometric testing in the field.

Between May and July 2025, the researchers surveyed 447 older adults screened as having MCI and their family caregivers, recruited from two tertiary hospitals, two secondary hospitals, and six communities across China. This deliberately mixed recruitment strategy, spanning both hospital and community settings, was designed to capture families at different points on the healthcare-seeking pathway, from those already engaged with specialist care to those who had never discussed the problem with any professional. The sample was randomly split into two halves, a technique that allows exploratory factor analysis on one portion of the data and confirmatory factor analysis on an entirely independent portion, providing a far more stringent test of a scale’s underlying structure than analyzing the full sample at once.

The psychometric results converged on an elegant architecture. Both instruments, the Healthcare-Seeking Decision-Making Scale for patients (HSDM-P) and the parallel version for caregivers (HSDM-C), each contain 36 items organized into six dimensions: Symptom Perception and Recognition, Decision-making Dilemma, Decision-making Help, Decision-making Conflict, Decision-making Balance, and Decision-making Results. Exploratory factor analysis on the first half of the sample identified a six-factor solution explaining 73.769 percent of the variance for the patient version and 72.246 percent for the caregiver version, figures that indicate the six dimensions capture the overwhelming majority of what the items measure. Confirmatory factor analysis on the second half then tested whether this structure held in fresh data, and it did. The patient version yielded a chi-square to degrees of freedom ratio of 1.833, a comparative fit index of 0.924, a Tucker-Lewis index of 0.917, a root mean square error of approximation of 0.053, and a standardized root mean square residual of 0.051. The caregiver version performed comparably, with a chi-square to degrees of freedom ratio of 1.820, a comparative fit index of 0.941, and a root mean square error of approximation of 0.053. In the language of measurement science, values of this kind indicate that the six-factor model fits the observed data well for both members of the dyad.

Reliability, the degree to which a scale produces consistent measurements, was assessed with three complementary statistics, and all fell within acceptable to strong ranges. For the patient version, Cronbach’s alpha ranged from 0.843 to 0.914 across dimensions, McDonald’s omega from 0.895 to 0.937, and split-half reliability from 0.787 to 0.912. The caregiver version showed similarly robust figures, with Cronbach’s alpha between 0.884 and 0.915, McDonald’s omega between 0.910 and 0.936, and split-half coefficients between 0.829 and 0.890. Content validity, judged by expert ratings, was exceptionally high: the item-level content validity index ranged from 0.882 to 1.000 for the patient version and 0.941 to 1.000 for the caregiver version, while the scale-level average content validity index reached 0.986 and 0.994 respectively, far above the conventional 0.90 threshold for a new instrument.

The team also examined convergent and discriminant validity, asking whether items intended to measure the same construct actually cluster together and whether the six dimensions are statistically distinguishable from one another. Average variance extracted ranged from 0.538 to 0.748 for the patient version and 0.534 to 0.839 for the caregiver version, with composite reliability values between 0.860 and 0.918 and 0.889 and 0.954 respectively. Critically, the square root of the average variance extracted exceeded the correlations between factors for every pair of dimensions, satisfying the Fornell-Larcker criterion and confirming that each dimension measures something distinct. Finally, domain-level correlations with external healthcare-seeking measures fell in the expected directions, negative correlations ranging from 0.133 to 0.322 in magnitude for the patient version and positive correlations from 0.148 to 0.230 for the caregiver version, offering preliminary evidence that the scales behave as theory predicts when placed alongside conceptually related instruments.

What makes this work potentially consequential is its dyadic design. By producing matched patient and caregiver versions with parallel dimensions, the scales allow researchers and clinicians to compare, within a single family, how the older adult and the caregiver each perceive symptoms, experience dilemmas, seek help, encounter conflict, weigh options, and evaluate outcomes. Divergent scores between the two members of a dyad could flag families at risk of prolonged delay, for example when a caregiver perceives clear symptoms that the older adult does not recognize or acknowledges, or when conflict over whether to consult a doctor runs high. The authors recommend that interpretation focus primarily on the six domain scores rather than a single total, which turns the instrument from a blunt summary into a diagnostic map of where the decision-making process stalls. Such information could directly inform family-centered assessment and targeted interventions, from communication counseling to structured decision support, at the stage when intervention can still alter the course of cognitive decline.

The researchers are careful to frame their findings as preliminary. The validation sample was drawn from a single country, and the scales were tested in Chinese-language settings, so cross-cultural adaptation and further validation in independent and more diverse samples remain necessary before widespread adoption. The study was approved by the Ethics Committee of Guizhou Medical University and conducted in accordance with the Declaration of Helsinki, with written informed consent from all participants, and the authors declare no competing interests. Even with those caveats, the arrival of psychometrically sound, dyad-matched measures of healthcare-seeking decision-making represents a genuine advance for the field of cognitive aging. If the critical window of mild cognitive impairment is to be used rather than lost, understanding the family conversation that precedes the clinic visit, and measuring it with precision, may prove to be one of the most practical tools yet added to the early detection toolkit.

Subject of Research: Development and psychometric validation of dyadic healthcare-seeking decision-making scales for older adults with mild cognitive impairment and their family caregivers

Article Title: Development and validation of healthcare-seeking decision-making scales for older adults with screen-detected mild cognitive impairment and their family caregivers

Article References: Li, Y., Luo, T., Wang, Q., Liu, X., Li, J., Liu, X., Feng, Z., & Luo, Z. (2026). Development and validation of healthcare-seeking decision-making scales for older adults with screen-detected mild cognitive impairment and their family caregivers. BMC Nursing. https://doi.org/10.1186/s12912-026-05470-6

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05470-6

Keywords: mild cognitive impairment, healthcare-seeking, decision-making, scale development, psychometric validation, family caregivers, older adults, dementia early detection, nursing research, Delphi method, factor analysis, China