This cross-sectional observational study, titled “Integrating Systematic Data of Geriatric Medicine,” was conducted from September 2019 to October 2020 to identify strategies for promoting healthy aging. Older adults living in the community were recruited from the Songshan District of Taipei City and from Chang Gung Health and Culture Village in Taoyuan City. Participants were recruited through random sampling and screened according to the Health Check-up Process. Inclusion criteria were: (1) age ≥ 60 years; (2) at least one visit to Chang Gung Memorial Hospital within the year before recruitment; and (3) residency in Taiwan for > 180 d during that year. Exclusion criteria were: (1) significant organ-system abnormalities; (2) severe autoimmune disorders; (3) ongoing cancer treatment at recruitment; (4) antibiotic use within the preceding month; (5) Ascertain Dementia 8 score ≥ 2, Mini-Mental State Examination score ≤ 26, or Geriatric Depression Scale score ≥ 5; (6) physician-diagnosed dementia, stroke, Parkinson’s syndrome, Parkinson’s disease dementia, epilepsy, brain tumors, or history of psychotic disease, major depressive disorder, anxiety disorder, bipolar disorder, alcohol or drug abuse, or significant head trauma with loss of consciousness; (7) substantial hearing, vision, or cognitive impairment interfering with clinical assessment; and (8) frailty preventing standing or walking. Only participants with complete PSG assessments were included in the present analysis. This study received approval from the Institutional Review Board of the Chang Gung Medical Foundation, Taiwan (approval number 201900702A3), and it was carried out with the principles of the Declaration of Helsinki and written informed consent from all participants. As a pilot, the study aimed to establish preliminary processes and data collection under resource constraints, and therefore targeted an initial enrollment of 100 participants able to complete the program.

Participants’ demographics, medical histories, polysomnographic indices, and responses to questionnaires on psychological distress, loneliness, and cognitive functioning were recorded. Relevant medical records were also reviewed during interviews.

The PSQI is a self-administered 19-item questionnaire evaluating self-reported sleep quality over the preceding month. Its items generate seven component scores: sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction. Each component is scored from 0 to 3, and their sum yields a global score ranging from 0 to 21, with higher scores indicating poorer sleep quality. The PSQI demonstrates excellent internal consistency (α = 0.83). Its clinical and psychometric properties have been extensively validated, showing a sensitivity of 89.6% and specificity of 86.5% for detecting sleep disorders at a cutoff score of 5 (14).

Participants underwent standardized home-based PSG using the portable Philips Alice 6 lDe monitoring system (Philips Healthcare), which records EEG, EOG, chin EMG, airflow, thoracoabdominal respiratory effort, oxygen saturation (SpO₂), body position, and snoring. Sleep stages were scored in 30-s epochs according to the American Academy of Sleep Medicine (AASM) scoring criteria (15). The following PSG metrics were calculated: total sleep time (TST, the duration from sleep onset to final awakening minus wakefulness periods); SE, the percentage of time asleep relative to total time in bed; SoL, the interval from lights off to sleep onset; WASO, time awake after initial sleep; non-REM stages 1, 2, and 3 (percentage of total sleep time); REM sleep (percentage of total sleep time); and the arousal index (average arousals per hour of sleep). Overnight home-based polysomnography (PSG) was performed using a portable PSG system. Installation and equipment setup were completed by a professional PSG technician with the assistance of a project staff member. Scoring was performed by sleep technicians with > 10 years of experience, and final interpretation was conducted by a sleep specialist. Participation in multicenter studies has validated the reliability of our PSG scoring and reporting, ensuring consistency with international best practices (16, 17).

Each participant’s psychological distress, loneliness, and cognitive functioning were assessed using validated scales addressing different mental domains. The 17-item Hamilton Rating Scale for Depression (HAM-D) evaluates depression by assessing core symptoms, sleep habits, daily functioning, anxiety, and delusions. A score ≤ 7 generally indicates remission (18). The 14-item Hamilton Rating Scale for Anxiety (HAM-A) measures anxiety severity, distinguishing between psychological and physical symptoms, with a score ≤ 7 reflecting no or minimal anxiety (19). The Chinese clinician-administered Apathy Evaluation Scale (AES) includes 18–72 points, with higher scores indicating greater apathy. It shows strong internal consistency (α = 0.90) and test–retest reliability (0.86–0.88). A score > 34 typically indicates apathy (20, 21). The triggers of suicidal ideation inventory (TSII) measures factors associated with suicidal ideation in older adults, including emotional distress, family and financial concerns, and self-worth issues. It demonstrates good internal consistency (α = 0.78), with a cutoff score of 2 used to detect triggers (22). The revised 20-item Chinese version of the University of California, Los Angeles Loneliness Scale measures perceived loneliness, with total scores ranging from 20 to 80; higher scores indicate greater loneliness (23–25). Finally, the Everyday Cognition Scale (ECog) is a 39-item informant-rated questionnaire designed to assess daily cognitive functioning over the past 10 years. It evaluates six cognitive domains—memory, language, visuospatial abilities, planning, organization, and divided attention—using a 4-point scale, with higher scores indicating greater functional decline (26).

All statistical analyses were conducted using IBM SPSS Statistics for Windows, version 27.0. The Kolmogorov–Smirnov test assessed the normality of continuous data, and results showed the survey data were not normally distributed (p < 0.05) (27, 28). Accordingly, Spearman correlation analysis was used to compare the PSQI subscales with PSG indices. Participants were also categorized using a PSQI cutoff score > 5 to indicate poor sleep.

Logistic regression analyses examined associations between subjective poor sleep and objective sleep indicators, psychological distress, loneliness, and cognitive function. Subjective poor sleep status (poor vs. non-poor) was the dependent variable, while objective sleep measures, psychological distress, loneliness, and cognition served as independent variables. Model 1 adjusted for demographic factors (age, sex, and education). Model 2 additionally adjusted for psychological distress and loneliness. Model 3 further incorporated cognitive function, based on Model 2.

Stepwise regression analysis was performed to explore associations between PSQI subscales and indices significantly related to subjective poor sleep in Model 3. Each analysis included the same independent variables: three demographic factors (age, sex, and education level) and the seven PSQI subscales. Demographics were included as covariates to control for potential confounding. Statistical significance was defined as a two-tailed p-value < 0.05.

Between September 2019 and October 2020, 128 participants were enrolled in the Integrating Systematic Data of Geriatric Medicine study, which investigated strategies for promoting healthy aging. Sixteen were excluded for not meeting the inclusion criteria. Of the 112 eligible participants, 9 (8.0%) withdrew, 5 (4.5%) did not complete the scales, 8 (7.1%) did not complete the PSG, and 1 (0.9%) had incomplete PSG records, resulting in 23 exclusions. Thus, 89 participants (79.5%) completed both PSG and all questionnaires and were included in the final analysis (Figure 1). Table 1 summarizes participants’ demographics, sleep profiles, and clinical characteristics. The mean age was 73.4 years (±6.99), with 39 men (43.8%) and 50 women (56.2%). The mean average PSQI global score (6.35 ± 3.84) exceeded the cutoff, whereas mean scores for HAM-A (4.17 ± 3.84), HAM-D (2.53 ± 2.62), AES (25.47 ± 6.85), and TSII (0.52 ± 0.78) all fell within normal ranges.

The analysis of the correlation between PSQI subscales and PSG indices revealed significant associations, underscoring links between self-reported and objective sleep measures (Table 2). PSQI sleep quality was significantly correlated with PSG WASO (rs = 0.272, p = 0.010). PSQI sleep latency was associated with PSG sleep efficiency; (rs = −0.258, p = 0.015), SoL; (rs = 0.285, p = 0.007), and WASO (rs = 0.285, p = 0.007). PSQI sleep disturbances were correlated with PSG SE (rs = −0.285, p = 0.007) and WASO (rs = 0.255, p = 0.016). PSQI use of sleeping medication was correlated with PSG N2 sleep duration (rs = 0.363, p < 0.001), N3 sleep duration (rs = −0.308, p = 0.003), and REM sleep (rs = −0.336, p = 0.001). Finally, PSQI global scores correlated with PSG WASO (rs = 0.216, p = 0.042) and N2 sleep duration (rs = 0.225, p = 0.034).

Logistic regression analysis defined poor sleep status according to PSQI scores. In Model 1, PSG SE, SoL, and WASO were significantly associated with poor sleep. After adjusting for psychological distress and loneliness in Model 2, PSG SE, SoL, WASO, and depression remained significant. In Model 3, which additionally adjusted for cognitive function, the same variables—PSG SE, SoL, WASO, and depression—remained significantly associated with poor sleep. The results are listed in Table 3.

Building on the dominant indicators identified above, stepwise regression models further clarified indices of subjective sleep condition. Poor sleep is defined by the PSQI dichotomous method. Education, age, PSQI disturbance, and sex were significantly associated with PSG SE, explaining 13.4, 7.5, 7.2, and 4.8% of its variance, respectively. Education and PSQI latency were significantly associated with PSG SoL, explaining 13.4 and 8.7% of the variance, respectively. PSQI quality, education, and sex were significantly associated with PSG WASO, explaining 9.3, 5.5, and 5.1% of the variance, respectively. Finally, PSQI latency and PSQI quality were significantly associated with depression, explaining 20.9 and 4.2% of the variance, respectively. Results are summarized in Table 4.

This study has limitations. First, the relatively small sample size (n = 89) may restrict the generalizability of the results to broader populations. The results should be interpreted cautiously, and larger cohorts are warranted to validate and extend these findings. Second, the cross-sectional design precludes establishing causal relationships between PSQI components, PSG metrics, and outcomes related to psychological distress, loneliness, and cognitive functioning. Longitudinal studies are required to clarify the temporal dynamics among these variables. Third, reliance on self-reported measures, including the PSQI and questionnaires for psychological distress, loneliness, and cognition, may introduce reporting bias and affect measurement accuracy. Besides, only one night of home-based PSG was recorded. The potential impact of first-night effects cannot be fully excluded. Although home-based assessment may reduce unfamiliarity-related disturbance, two-night recordings are recommended for more stable estimates of sleep architecture in the future. Finally, although age and sex were considered, other potential confounders, such as socioeconomic status, medication use, and comorbidities, should also be incorporated into future analyses.