Study design, setting, and population
This descriptive cross-sectional study included residents aged ≥65 years with a diagnosis of dementia. Participants were selected through convenience sampling from multiple LTC facilities in which nurses work. Of the 130 residents who were recruited, 13 were excluded owing to deteriorated conditions, communication difficulties, or refusal to participate. The data were collected between April 9 and May 26, 2018 from five cities in South Korea.
Data collection procedure
We selected seven LTC facilities, based in five cities in South Korea, for data collection and obtained approval from the directors of facilities after we explained the study proposal and data collection procedures. Lists of all eligible residents were provided by the directors of facilities. We initially approached these residents to evaluate their capacity for making decisions, and we obtained written informed consent from those residents who demonstrated an acceptable decision-making capacity. Consent for residents who could not comprehend the purpose of the study and/or make an informed decision to participate was provided by health care proxies such as family members. Approximately 10% of participants personally provided informed consent, while consent for 90% of the participants was provided by their health care proxies. This study was approved by the Institutional Review Board of the Ewha Womans University (No. 148–14).
Before data collection, four research assistants and nine nurses from the chosen LTC facilities completed a 1-h training session on the administration of the assessment instruments. At the end of the training session, the degree of inter-rater reliability was measured, and the result was a kappa value of 0.82, indicating acceptable inter-rater agreement.
The questionnaire collected information on participants’ demographic characteristics, ADLs, cognitive function, physical capability, grip strength, BPSD, and depression. Demographic characteristics included age, sex, marital status, length of stay (in months), comorbidities, and level of LTC service. The level of LTC service was determined by assessing participants’ diseases or symptoms, visual/hearing ability, ADLs, cognition level, behavioral changes, and need for nursing care/rehabilitation/medical devices. LTC service levels ranged from level one, where participants required help in all aspects of daily life, to level five, where participants had dementia with limited functional decline . In this study, levels one and two were defined as severe status, level three as moderate status, and levels four and five as mild status. The nurses in the LTC facilities assessed the following general characteristics, which were measured by minute observations in daily life: ADLs, BPSD, and depression. Cognitive function, physical capability, and grip strength were assessed by two researchers and four trained research assistants. We obtained permissions to use all survey instruments from authors or publisher.
The Korean Activities of Daily Living (K-ADL), a tool developed by Won et al.  based on the Katz Index , was used to assess the participants’ level of ADLs. K-ADL comprises seven questions on dressing, washing the face, bathing, feeding, transferring, toileting, and continence. The scores range from 7 to 21 points with lower scores indicating higher independence. In the study that developed the K-ADL, the Cronbach’s α was 0.937  compared with 0.91 in the present study.
Eating performance was assessed using the item “feeding” in the K-ADL . This single item (“does the participant feed oneself without help?”) is rated on a scale of 1 (“completely independent to feed self”), 2 (“partially dependent and some assistance needed”), or 3 (“completely dependent and needs to be fed”). In this study, 1 point indicated independent eating performance, while 2 and 3 points indicated dependent eating performance.
Cognitive function was measured using the Korean Mini Mental State Examination (K-MMSE) . The K-MMSE is the Korean standardized version of the MMSE developed by Folstein et al. . The total possible score is 30 points, with higher scores indicating higher cognitive function: 20 ≤ MMSE< 24 indicates mild impairment, 10 ≤ MMSE< 20 indicates moderate impairment, and 0 ≤ MMSE< 10 indicates severe impairment . The Cronbach’s α for MMSE was 0.86 , and the sensitivity of the K-MMSE ranged between 0.70 and 0.83 . The Cronbach’s α in this study was 0.90.
The Korean version of the Physical Capability Scale (PCS), translated by Jung et al.  and based on the PCS developed by Resnick et al. , was used to measure participants’ physical capability. This tool comprises 16 questions assessing the extent to which the participant can follow the examiner’s instructions, and it tests the ability and range of the joints in the arms and legs. One point is scored if the participant can perform the instruction, and 0 points are scored if they cannot. A higher score indicates higher physical capability. When the tool was first developed, the Cronbach’s α was 0.83, and it was 0.90 in our study.
A grip strength dynamometer (Model KS-301; Lavisen Co. Ltd., Namyangju, Korea) was used to measure the participants’ grip strength in kilogram-force (kgf). The participant grabbed the instrument with their predominant hand. Grip strength was measured in the sitting position with the engaged arm fully extended and hanging down; however, for patients who could not sit, grip strength was measured in the supine position with the engaged arm extended. Grip strength was measured twice using the same method, and the higher of the two measurements was selected. If there was a substantial difference between the two measurements, the grip strength test was repeated. The range of normal grip strength in the 65–69 age group is 28.2–44.0 kgf (men) and 15.4–27.2 kgf (women) and 21.3–35.1 kgf (men) and 14.7–24.5 kgf (women) in the 70–99 age group (Lavisen Co. Ltd., Namyangju, Korea).
To assess the participants’ level of BPSD, we used the Neuropsychiatric Inventory-Questionnaire (NPI-Q) originally developed by Kaufer et al.  and translated to Korean . For each participant, a nurse indicated the presence of 12 psychiatric symptoms (delusion, hallucination, agitation/aggression, depression/dysphoria, anxiety, euphoria/elation, apathy/indifference, disinhibition, irritability/lability, aberrant motor behavior, night-time behavior, and appetite/eating change) over the prior 4 weeks as present (“yes”) or not present (“no”). If “yes” was indicated, the symptom severity was assessed using a 3-point scale: 1 = mild, 2 = moderate, and 3 = severe. The total possible NPI-Q score ranged from 0 to 36, with a higher score indicating severe BPSD. The reliability of the NPI-Q in the previous study was Cronbach’s α = 0.75 , and in this study it was Cronbach’s α = 0.71.
The Cornell Scale for Depression in Dementia (CSDD), developed by Alexopoulos et al.  and translated into Korean by Kim , was used to assess participants’ depression. For each participant, a trained nurse was asked to describe 19 depressive behaviors observed during the previous week by rating them using a 3-point scale: 0 = absent, 1 = mild or intermittent, 2 = severe, or N/A = unable to evaluate. The total possible score ranged from 0 to 38 points, with a score ≥ 8 points indicating depression. In the study that developed the Korean CSDD , the Cronbach’s α was 0.84, and in our study, it was 0.77.
Data were analyzed using SPSS Version 21.0 (Armonk, NY: IBM Corp.). Participant characteristics were analyzed using descriptive statistics (frequency, percentage, mean, and standard deviation). General characteristics were expressed as frequency, percentage, mean, and standard deviation. The level of ADLs, cognitive function, physical capability, grip strength, BPSD, and depression between eating performance (“independent” vs. “dependent”) groups were compared using Chi-squared (χ2) and t-tests. Logistic regression analysis was performed to identify predictors of independent eating performance. Variables that were potential predictors of independent eating performance (comorbidities, cognitive function, physical capability, and grip strength) were selected from the conceptual framework based on the literature review and results of the bivariate analysis performed in this study.