In this study, we select older adults living in nursing home through a multi-stage, stratified sampling scheme. Firstly, per geographical divisions of mainland China by the Statistical Yearbook of Chinese Health and Family Planning, 2 provinces / municipalities from each of the Eastern, Central and Western Regions of China were randomly picked, totaling 6 provinces (Liaoning, Hainan, Shanxi, Hubei, Sichuan and Shaanxi). Then, 2 cities from each province were conveniently picked with consideration of travel accessibility. Meanwhile we added 6 back-up cities conveniently, forming a total of 18 cities. Thirdly, 3 nursing homes of various sizes (one having < 100 beds, one having 100–300 beds, and one having > 300 beds, respectively deemed as small size, medium size, and large size according to the official website of the Civil Affairs Bureau) in each city were reached out conveniently, totaling 54 nursing homes. Despite our desperate effort, 38.7% selected nursing homes declined our request for survey. In worst case scenario, we failed to obtain permissions from any of the selected nursing homes in 2 provinces (Liaoning and Shanxi). Therefore, we made adjustment to the sampling scheme by switching to other provinces/ municipalities and adding more cities to fill the vacancy. Finally, 38 nursing homes in 13 cities granted permission and were approached for the cross-sectional questionnaire survey which stretched from September 2017 through June 2018. Inclusion criteria of respondents were: 1) ≥ 60 years old ; 2) living in nursing homes for more than 1 month; 3) being consented for the survey. The exclusion criteria were: (1) the score of the Short Portable Mental Status Questionnaire for the older adults was less than 5 points; (2) having severe impairment in communication . We used the self-made demographic and health status questionnaire, the Chinese version of CANE revised by expert consultation, Modified Barthel Index , The Numerical Rating Scale for pain assessment (NRS for pain assessment)  and the 15-item Geriatric Depression Scale  to survey the basic information and health status, needs, activities of daily living (ADL), pain, and depression of the older adults living nursing homes, respectively. The questionnaires were filled out through face-to-face interviews with the respondents by 12 nursing graduate students. Before the survey, the students were given unified training on surveying technique and countermeasures to problems emerging during the survey. The interviews took place in private places such as lounges or the participants’ rooms to minimize distractions and insure the authenticity of the data.
The sample size was calculated according to the formula n = Zα/22×p×(1-p)/δ2 , where α = 0.05, Zα/2=1.96, and δ = 2.5%. Based on the survey results of the Ministry of Civil Affairs of China, the occupancy rate of nursing homes in 2020 was 50% . On the ground of an 85% effective response rate, the sample size of this study was 1808. A total of 2094 questionnaires were sent out and 2063 were effectively received with a response rate of 98.5%.
Camberwell Assessment of Need for the Elderly (CANE)
Camberwell Assessment of Need for the Elderly (CANE)  is a measuring tool that can comprehensively assess the physical, psychological, social and environmental needs of older adults. This questionnaire was developed by Reynolds in 2000. It was first used to survey the needs of older patients with mental disorder, and was later expanded to general older population . CANE consists of three versions, intended respectively for older population, home caregivers, and health care providers. This study chose the older adults version. CANE questionnaire looks into whether the needs of older adults are met or not, and groups them into either of the 4 categories of no needs, met needs, unmet needs, and unclear needs .
This questionnaire enjoys good reliability and validity, and can comprehensively assess the needs of older adults , which has been widely used in foreign countries. The 24-item questionnaire included social, physical, psychological and environmental needs. Each item addresses a specific need, with the response option 0 meaning there is no need, 1 meaning the need is met, 2 meaning the need is not met, and 9 meaning the need is not clear. The survey result of the questionnaire item is presented with the count of 4 categories (no need, met needs, unmet needs and unclear needs), without using its scoring values. The number of total needs is the tally of met needs plus the tally of unmet needs.
After obtaining the authorization of the original author, the research team translated CANE into Chinese, and then invited graduate students majoring in English to back translate the initial Chinese version. After confirming that there was no deviation and ambiguity caused by translation, expert consultation method was used to test the content validity of the Chinese version. Twelve experts (5 males and 7 females) involved in the consultation included managers of nursing homes, managers of hospital geriatric nursing, college researchers on older population. These experts came from Hubei Province and were aged 40 to 57 years old (median: 48 years old). Seven experts had a senior title, 3 had a deputy senior title, and 2 intermediate title (these two were presidents of nursing homes). The expert consultation form adopts a 4-level scoring method: experts give scores as specified by the correlation between the questionnaire items and the research content, 4 meaning very relevant, 3 relevant, 2 general, and 1 irrelevant. The Item-level content validity index (I-CVI) was obtained by dividing the number of experts giving 4 or 3 for each item by the total number of experts participating in the consultation. There is also a comment column in the consultation form for experts to give suggestions on adding, deleting or modifying the items in the tool.
The I-CVI of Chinese version CANE ranged from 0.75 to 1, and the average S-CVI was 0.93, indicating the tool enjoys good content validity. The authority coefficient of the experts ranged from 0.77 to 0.97, signifying that the consultation results were accurate and credible. After consulting with the experts, the researchers deleted item 5 (taking care of others), 20 (drinking), and 23 (paying your own bills). An initial sample of 201 respondents from 4 nursing homes in Wuhan city participated in the pilot survey from July to August 2017, in which, item 1 (accommodation) was deleted as all participants rated the need as “met”. At last, 20 items covering diet, self-care and daily activities were retained. The Cronbach’s alpha coefficient of the Chinese version CANE in this study was 0.74.
The Short Portable Mental Status Questionnaire, SPMSQ
Compared with similar measurement tools of mental status for older adults such as MMSE, the Short Portable Mental Status Questionnaire  is simpler and takes less time, which is suitable for screening in older population. The questionnaire has 10 questions and the respondents were scored based on the tally of correct answers. A tally of 0–2 correct answers was deemed as having very severe cognitive impairment, 3–5 as severe cognitive impairment, 6–7 as mild cognitive impairment and 8–10 as having normal cognitive function. Although the Chinese version of SPMSQ has been widely used in older adults, there are few studies exploring the threshold value. A Singapore study with a majority of Chinese participants found that the sensitivity (78%) and specificity (75%) of SPMSQ were better when the threshold value was “5 or fewer correct answers”, and the older population having less than 6 years of education could make 4 correct answers . Therefore, this study specified that older adults making “5 or more fewer correct answers” were deemed as having severe cognitive impairment. The scale has good reliability .
Demographic and health status questionnaire
The researchers developed the demographic and health status questionnaire by themselves. The demographic items included organizational type, staffing, room type, age, gender, ethnicity, religion, education, marital status, having living children or not, length of time of living in nursing homes (years), living condition before admission, income (yuan/month). The health status items included eyesight, hearing, number of diseases, sleeping status, skin conditions, occurrence of accidents in the past 30 days, all of which were based on participants’ self-report.
Modified Barthel Index
The Modified Barthel Index was used to assess 10 activities of daily living (ADL): grooming, bathing, feeding, toilet use, stair climbing, dressing, bladder, bowels, walking or wheelchair movement, and transfer (chair to bed and back) . The total score of the scale was 100, with a score ≤ 40 indicating severely dependent, 45–60 moderately dependent, 65–95 slightly dependent, and 100 completely independent. The Cronbach’s alpha coefficient for the ADL in this study was 0.85.
The Numerical Rating Scale (NRS) for pain assessment
The NRS for pain assessment divides pain into 0–10 points: the higher the score is, the more severe the pain is. The NRS for pain assessment has a good acceptance among older population . A score of 0 indicates no pain, 1–3 mild pain, 4–6 moderate pain and 7–10 severe pain. The NRS for pain assessment has been tested by various studies and has good reliability .
The 15-item Geriatric Depression Scale (GDS-15)
GDS-15 has been widely used in the screening of depressive symptoms of older adults and has good reliability and validity . The 15-item scale assesses how older adults felt over the past week, with “yes” (scoring 1) or “no” (scoring 0) responses. Five of the questions used reverse scoring. The score ranges from 0 to 15, with higher scores suggesting more severe depression. A systematic review has shown that the Chinese version of the scale for older adults takes 4 or 5 points as the cut-off point mostly . In this study, we used a score of 5 or above to specify the existence of depression, in an effort to facilitate the comparison between similar studies. The Cronbach’s alpha coefficient for the GDS-15 in this study was 0.80.
Epidata 3.1 was used for data entry. SPSS 25.0 was adopted for statistical analysis. The counting data were described by frequency and percentages; the measurement data were examined for normality with Shapiro-Wilk test, with normally distributed data being described by mean ± standard deviation, and non- normally distributed data being described by median (interquartile range, abbreviated as IQR).
Univariate Logistic regression analysis was used to test whether there were differences in unmet needs among research participants with different characteristics. Hierarchical logistic regression analysis was then employed to analyze the influencing factors of unmet needs of the elderly nursing home residents. The criteria for including and excluding variables were αin = 0.05 and αout = 0.10.