As a non-randomized controlled pilot study with a convenience sampling design, this study utilizes a single-blind approach. We recruited participants from communities and nursing homes in Taiwan between April and December 2019 through social networking tools, including Line and Facebook. Participants in the experimental and comparison groups were exposed to bright light therapy (2,500 lux) and general lighting, respectively.

The calculated sample size was 20 participants using G*Power 3.1 computer software, and the data were analyzed using analysis of variance with the conditions set to a statistical significance of α = 0.05, β = 0.2 and an effect size of f = 0.39 (McCurry et al., 2005). F tests and repeated measures ANOVA were used to determine within-between interactions.

Participants were selected based on the following inclusion criteria: (1) diagnosis of dementia consistent with the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (Kupfer et al., 2008); (2) age between 60 and 95 years; (3) consent to participate in the study obtained from the participants and their guardians; and (4) willingness to participate in the group. Participants were excluded if they (1) had adverse reactions to light, such as systemic lupus erythematosus, epilepsy, blindness, retinal detachment, and macular degeneration; (2) were on the sleep cycle of an advanced sleep-wake phase disorder; and (3) had a score < 3 points on the Mini-Mental State Examination (MMSE) and were unable to express their own intentions verbally.

Among the 60 individuals invited to participate in the study, 11 did not meet the inclusion criteria, and 14 declined to participate. Accordingly, we included 35 patients and consecutively allocated them to the experimental (n = 17) and comparison (n = 18) groups. Thirteen patients withdrew from the study for the following reasons: unable to attend on their own and unavailability of family members to accompany them to daily therapy (n = 9), pneumonia or leg injuries (n = 2), and inability to integrate into the group (n = 2). Finally, only 22 patients completed the study (Figure 1). Patients in the experimental group received treatment at the light center of National Yang Ming Chiao Tung University. In contrast, patients in the comparison group received light exposure at the institutionalization modalities center. The patients were transported to the experimental site using a bus equipped with special wheelchair equipment or were escorted by family members 5 days a week.

The measurement tools included demographic data and two outcome measures: BPSD were determined using the Neuropsychiatric Inventory (NPI), and cognitive function outcomes were obtained using the MMSE. The NPI data were obtained by interviewing the corresponding caregivers and nursing home staff, while the researchers directly obtained the MMSE data. The data were collected at baseline and in the 5th and 9th weeks from April to December 2019. The data collectors were single-blinded to the participants.

The demographic data included sex, age, educational level, marital status, sleep pattern, dementia type, source, dementia severity, and medication use (benzodiazepines, antidepressants, antipsychotics, and anti-dementia drugs). They were obtained from the primary caregiver and nursing home staff.

According to the reports of caregivers and nursing home staff, sleep and wake times were observed for a week. Additionally, patterns were classified as advanced sleep-wake phase disorder, delayed sleep-wake phase disorder, irregular sleep-wake rhythm disorder, or a sleep-wake rhythm (Burgess and Emens, 2016), and accelerometer-monitored sleep pattern data were recorded using an accelerometer. Compared with sleep logs of older adults with dementia, the between-measurement error value was < 30 min. The accelerometer (XA-5, Taipei, Taiwan) was worn around the wrist to continuously record actigraphic data for ≥ 3 days at baseline using the KY laboratory software package¹.

The defined daily dose (DDD), which is promoted by the World Health Organization (Polk et al., 2007), is the assumed average maintenance dose per day for a drug used for medication in adults. The DDD represents a unit of measurement only and was calculated as the total dose of a drug divided by the DDD. The drug data were collected at baseline and in the 5th and 9th weeks to monitor the effect of drug changes on the study results.

The NPI was developed by Cummings et al. (1994) (see also Connor et al., 2008; Lai, 2014) to assess dementia-related psychopathology and neuropsychiatric behaviors. The NPI includes four domains: emotional symptoms (e.g., dysphoria, anxiety, apathy, euphoria, and irritability), psychiatric symptoms (e.g., delusions and hallucinations), behavioral problems (e.g., agitation, disinhibition, aberrant motor behavior, appetite, and eating abnormalities), and sleep disturbances (e.g., nighttime behavioral disturbances) (Connor et al., 2008). The NPI shows good content validity, concurrent validity, and interrater reliability. The caregivers were asked to rate the frequency of the symptoms of each disturbance on a scale of 1 (occasionally or less than once a week) to 4 (very frequently, more than once a day, or continuously). The rating of the symptom severity was 1, 2, or 3 for mild, moderate, or severe, respectively. The total score ranged from 0 to 144 points, where a higher score reflects a more severe level of BPSD. The NPI subdomains were significantly correlated with the domains of behavioral disturbances and the Hamilton Rating Scale for Depression. Interrater reliability ranged from 93.6 to 100%, depending on the subdomain, and test-retest reliability was also high: r(20) = 0.79 (Cummings et al., 1994; Connor et al., 2008; Lai, 2014; Cloak and Al Khalili, 2021). In this study, the NPI exhibited adequate overall internal consistency (α = 0.66). Cronbach’s α coefficients for the older adults with dementia were 0.56 and 0.75 for the pretest and posttest, respectively.

The MMSE was developed to quantitatively assess dementia-related cognitive function, ranging from 0 for the worst to 30 for the best (Folstein et al., 1975; Graf et al., 2001; Baek et al., 2016). It has good test-retest reliability (0.80–0.95), sensitivity, and specificity in detecting mild dementia. The MMSE is categorized into orientation [e.g., orientation questions, five each for time and place (10 points)], registration [e.g., three-word registration and 1-min recall (3 points)], attention and calculation [e.g., assessed either by serial subtraction of 7 from 100 or serial subtraction of 3 from 20 (5 points)], recall [e.g., three-item recall test for memory (3 points)], and language, visuospatial construction [assessed by a three-stage command, repetition, naming, reading comprehension, and writing (8 points) and copying two intersecting pentagons (1 point)] (Graf et al., 2001). In this study, the MMSE exhibited an adequate overall internal consistency (α = 0.85). Cronbach’s α coefficients for the older adults with dementia were 0.81 and 0.88 for the pretest and posttest, respectively.

Mini-mental state examination scores ≥ 21 indicate the presence of mild dementia, 11–20 indicate moderate dementia, and 0–10 indicate severe dementia (Perneczky et al., 2006).

The demographic characteristics of the participants in the two groups were analyzed using SPSS version 24.0 (IBM Corp., Armonk, NY, United States), with sex, educational level, marital status, sleep pattern, dementia type, dementia severity, and source taken into consideration. These descriptive statistics were performed at baseline and in the 5th and 9th weeks to examine whether the demographic characteristics were affected by the loss of any of the participants. Descriptive statistics are presented as the number of cases (n), percentage (%), mean, and standard error for both the experimental and comparison groups. The primary pretest-posttest analyses were based on the intention-to-treat sample. Non-parametric inferential statistics (chi-square test and Mann–Whitney U test) were used to compare the basic profiles of the two groups before and after the intervention and within groups. The Mann–Whitney U test was used to assess within-group differences in age, NPI and MMSE scores, and medication (benzodiazepines, antidepressants, antipsychotics, and anti-dementia drugs) to compare the baseline values of the two groups. The data were collected at three time points: baseline and after light exposure interventions (in the 5th and 9th weeks). One unit of the drug was defined as the DDD. Considering the possible consequences of repeated measurements, the intervention effects were tested using generalized estimating equations for comparisons between groups (Zeger and Liang, 1986). Improvements in outcomes (NPI and MMSE scores) over time in both groups were analyzed using generalized estimating equations with an exchangeable working correlation matrix. A robust standard error was used to calculate statistical significance. We tested for the main effects of the groups (experimental and comparison) and time points (baseline, 5th week, and 9th week), as well as their interaction (group × time point); a statistically significant interaction effect (group × time point) indicated that the two groups exhibited significantly different changes over time. Benzodiazepine use was included as a covariate in the generalized estimating equations (Rosenberg et al., 2012).

All the participants from the experimental and comparison groups completed profile tracing. Thirty-five participants completed the baseline assessment, and 29 and 22 participants completed the 5th and 9th week assessments, respectively, for a retention rate of 64.7% in the experimental group and 61.1% in the comparison group. The attendance rate was 93% in the experimental group and 75% in the comparison group. No adverse effects were observed (e.g., fall, injury, eye injury, headache, and dizziness) during the 8 weeks of light therapy.

The participants were recruited by a candidate psychiatric nurse who was confirmed by the doctoral supervisor to have met the academic rigor, was briefed on the research plan before implementation, and had received training to ensure the consistency of data collection. The principal researcher met with the first author to ensure that the protocols were being followed, establishing the interrater reliability of the evaluator. The intervention was delivered by the first author, who had completed training before program implementation. They intervened in accordance with standard procedures.

Participant recruitment began after the Institutional Review Board of National Yang Ming Chiao Tung University (Taiwan) approved this study (YM107120F). Additionally, the ethical standards listed in the Declaration of Helsinki were followed throughout the study. The participants could withdraw from the study at any time during the study period and individuals were assured that any potentially identifiable images or data included would not be included in this article. A cooperative research plan for the study was explained to the participants and family members, and both of these groups provided written informed consent that was approved by the Ministry of Legal Consent, and written informed consent was obtained from the participants. No particular ethical issues arose concerning the study design or conduct. Codes were used to represent the names of the participants to ensure privacy and confidentiality.

The flow chart of participant selection is shown in Figure 1. Thirty-five participants completed the baseline assessment, and 29 and 22 participants completed the 5th and 9th week assessments, respectively. The chi-square test was used to analyze the participants’ demographic characteristics (Table 1). No statistically significant differences were observed in the demographic characteristics (educational level, marital status, sleep pattern, dementia type, dementia severity, and source at baseline or in the 5th and 9th weeks) between the experimental and comparison groups. This finding indicated a homogeneous distribution of participants. Conversely, a statistically significant difference was observed in sex, with male participants comprising 5.9 and 33% of the experimental group and comparison group, respectively (P < 0.05); most of the participants were female. The Mann–Whitney U test was used to assess the baseline age, NPI and MMSE scores, and medication, and the results showed no significant differences in age, medication, and MMSE scores. However, a statistically significant difference (NPI mean of 36 fractions for the experimental group and 21 fractions for the comparison group; P < 0.006) was observed in the NPI scores between the two groups. The experimental group had more serious BPSD than the comparison group at baseline.

The NPI is categorized into assessments of the emotional state, psychiatric symptoms, behavioral disturbances, and sleep disturbances (Figures 5A–D). Regarding the specific neuropsychiatric behavioral domains of the emotional state, psychiatric symptoms, and sleep disturbances, continuous, promising improvements were observed in sleep disturbances (Figure 5D). The experimental group showed significant improvement in sleep disturbances (Wald’s test = 3.9, P < 0.002; Wald’s test = 10.0, P = 0.04) from baseline to the 5th and 9th weeks, which was greater than that in the comparison group.

The experimental group showed a greater mean improvement than the baseline; the highest value was observed for sleep disturbances, followed by psychiatric symptoms, emotional state, and behavioral disturbances. However, the greatest mean improvement in the comparison group was observed for sleep disturbances, followed by emotional state, behavioral disturbances, and psychiatric symptoms. For both the experimental and comparison groups, the domain with the maximum change was psychiatric symptoms (Figure 5B).

The MMSE is categorized into orientation, registration, attention and calculation, recall, and language visuospatial construction (Figures 6A–E). Regarding the specific cognitive function domains, a significant change in the slope reflected a better status in terms of orientation, registration, attention calculation, recall, and language visuospatial construction. Among these domains, orientation showed the greatest improvement (Figure 6A). The experimental group showed a significant improvement in orientation (Wald’s test = 10.1, P = 0.001; Wald’s test = 8.73, P = 0.003) from baseline to the 5th and 9th weeks, which was greater than that of the comparison group.

The experimental group showed greater improvements in attention and calculation, followed by orientation, recall, language visuospatial construction, and registration compared with the baseline value. However, the greatest improvements in the comparison group were observed for attention and calculation, followed by recall, registration, language visuospatial construction, and orientation. For both the experimental and comparison groups, the domain with the maximum change was orientation (Figure 6A).

These results were consistent with those reported by Fernandez et al. (2018). The authors found that light is effective at modulating sleep and circadian rhythms via an indirect pathway and directly affects mood without affecting the sleep pathway or circadian rhythmicity (Rosenberg et al., 2012). We found that bright light therapy is effective at modulating sleep via an indirect pathway or even by directly affecting psychiatric symptoms. Our results for general cognition are consistent with those reported in the literature (Yamadera et al., 2000; Riemersma-van der Lek et al., 2008; Chiu et al., 2017; Mitolo et al., 2018; Missotten et al., 2019; Rubiño et al., 2020). However, the results for BPSD are inconsistent with those presented in the literature (Dowling et al., 2007; Riemersma-van der Lek et al., 2008). The difference may be attributed to the use of ambient light that is more friendly to participants than a lightbox (Dowling et al., 2007; Riemersma-van der Lek et al., 2008), which restricts freedom of movement. The differences in duration and dosages, namely, 2,500 lux in the present study and 1,000 lux in the literature (Riemersma-van der Lek et al., 2008), may also explain the discrepancy. Based on previous studies, we assumed that the factors listed below were important when measuring the effects of light therapy on BPSD and cognitive function. (1) Bright light intensity with horizontal illumination of 2,500 lux (van Maanen et al., 2016; Chiu et al., 2017; Missotten et al., 2019) and within a 45° visual field (McCurry et al., 2005, 2011). Current meta-analytic evidence suggests that light therapy is effective at a threshold of ≥ 2,000 lux light intensity (Chiu et al., 2017). Thus, we used bright light (horizontal illumination of 2,500 lux within a 45° visual field). This angle allows bright light to enter evenly and effectively without causing discomfort to the participants. (2) At least 60 min of exposure/day from 9 am to 10 am (Skjerve et al., 2004; Burgess and Emens, 2016; Missotten et al., 2019). (3) The use of group-based ambient light therapy to promote acceptability and patient adherence to treatment (van der Wardt et al., 2017; Vseteckova et al., 2018). Bright light exerted a significant effect, suggesting that 4 weeks is the most effective protocol. This result was consistent with previous findings revealing that light therapy for at least 4 weeks was effective (Figueiro et al., 2014; Figueiro, 2017). In the literature, the lack of illumination for the most effective light exposure intervention duration did not produce consistent conclusions. In the present study, 4 weeks of bright ambient light exerted a significant effect. Light exposure interventions require that treatment acceptability and patient treatment adherence be considered (van der Wardt et al., 2017). Ambient light therapy is achieved by installing high-intensity and low-glare overhead lighting. This approach relies on passive exposure with consistently maintained effectiveness parameters, permitting persons with dementia to engage in their normal activities without restricting their movement, and removing the interaction bias inherent in lightbox studies in this population. The effectiveness of light exposure interventions strongly depends on the patient’s adherence throughout the treatment (van der Wardt et al., 2017; Vseteckova et al., 2018).

Sleep is fragmented and disturbed in older adults with dementia. Their sleep problems vary according to dementia subtypes, such as Alzheimer’s disease, Parkinson’s disease, and dementia with Lewy bodies. While various types of dementia have been identified, Alzheimer’s disease (AD) is the most common form (Peter-Derex et al., 2015; Sekiguchi et al., 2017). Sleep disruption is frequently observed in patients with dementia and neurodegenerative disease. Sleep disruption alters sleep-wake timing, destabilizes physiology, and promotes a range of pathologies (e.g., neuropsychiatric diseases, metabolic diseases, neuroinflammation, and immunomodulatory diseases) that are considered associated with abnormal sleep. Circadian rhythms are disordered in elderly individuals; they affect not only sleep but also the endocrine system, leading to diabetes, hypertension, metabolic diseases, cardiovascular disease, brain function, cognition, and a decrease in immunity (Wulff et al., 2010; Salami et al., 2011; Urrestarazu and Iriarte, 2016; Cardinali and Brown, 2020; Wirz-Justice and Benedetti, 2020). These effects of light mediated by ipRGCs involve pacemaker-independent input to the suprachiasmatic nucleus (SCN) (Fernandez et al., 2018; Yanar and Halassa, 2018). The SCN is vulnerable to the aging process, resulting in a decreased volume and decreased cell number sensitivity to input signals such as light. Bright light is the first-line treatment to induce circadian resynchronization and ameliorate sleep disturbances. Bright therapy with morning light is usually superior to evening light (Touitou and Haus, 2000; Chang et al., 2018; Rubiño et al., 2020).

The attendance rate of the participants was 84%. The feasibility analysis showed that the participants in the experimental group were highly engaged during the intervention. The difference may result from the willingness of the participants’ family members to accompany them to the therapy and provide other assistance, as well as our use of social networking tools—namely, Line—to send participants entertaining and interesting videos and to convey to family members the importance of their participation, particularly in accompanying the participants to the therapy. Family members feedback from the participation process throughout the intervention of bright light therapy also improved their sleep quality and established good interpersonal interactions in their homes. Future research on bright light therapy may improve the sleep quality of people with dementia and their caregivers. A retention rate of 64.7% for participants in the experimental group and 61.1% for the comparison group was observed, and no adverse events were reported. Therefore, the distribution of the participants was homogeneous. Conversely, a statistically significant difference in the sex distribution was found. Due to the sex differences, male participants disliked group activities and withdrew at a higher rate than female participants. Future experimental interventions must consider males’ preferences.

This study has several limitations. First, this controlled pilot study is based on a convenience sampling design. This procedure might have affected the generalizability of the results. However, the process of compromising because of the practical limitations of people with dementia makes this batch of data more valuable. A randomized controlled trial should be conducted with this intervention in the future. Second, the experimental group showed a greater attendance rate than the comparison group, a finding that may have also affected the results. Third, the comparison group included more male participants than the experimental group, and the sex ratio of the two groups was significantly different. Thus, we were unable to determine the effects of light exposure on BPSD and cognitive function in older male adults with dementia. Other limitations of this study include the patients’ dementia severity and controlled exposure to natural bright light.

A longitudinal experimental research design is important. Therefore, future research is needed to develop intelligent lighting with variable doses and durations for treating patients with different NPI and MMSE scores by determining the accurate illuminance dose through physiological feedback. We recommend the use of the Clinical Dementia Rating to assess dementia severity in future studies. We suggest that bright light therapy may cause different results for patients with different types of dementia, i.e., Parkinson’s dementia, Alzheimer’s disease, and frontotemporal dementia. These different types of subjects might be included in future studies.