HNC 2023 Resident Well-Being Survey Analysis

Guiding Question - How does a sense of community and a “pro-neighbouring” attitude influence overall health?

Table 1

Methods

Data was collected through SurveyMonkey through both online and paper forms. Paper forms were manually entered by research assistants into SurveyMonkey. Data collection ran from October to December 2023.
Data from SurveyMonkey was first cleaned in Excel.

Data cleaning in Excel

Removing all text box-based variables (e.g., “Other” questions)
Manipulating new variables and collapsing existing variables to contain fewer categories
Assigning variable names
Standardising responses where necessary (ie., “five” -> “5”)

Data Analysis

Both cleaned datasets were uploaded into RStudio for further cleaning and preliminary analysis. These included omitting “NA” values and running chi-square tests to determine how different variables were related to each other.
Three outcome variables were decided upon to comprise the “pro-neighbouring” attitude in analysis: closeness of social networks, frequency of conversations with neighbours, and participation in community events. The full rationale for each of these outcomes is explained more fully below.
Ten predictor variables have been preliminarily selected. These are as follows: physical health, mental health, tenure in the building, perceived neighbourhood safety, age, gender, ethnicity, household composition, loneliness, and immigration. These all have been mentioned in other analyses of this topic and had sufficiently high response rates across both surveys to warrant inclusion.

Outcomes

Attendance of Community Meetings

This outcome consists of three levels - “Never or almost never”, “Sometimes”, and “Weekly or more often”. It had a non-response rate of about 12%. This outcome is representative of “social participation” referenced by Sones et al (2021) and can be used as a proxy for this idea in analysis. Social participation is any “… involvement in activities that provide interaction with others in society or the community” (Sones et al, 2021). Given the broad nature of this question in the survey (examples provided in the question of such activity ranged from a hobby group to a board, and interpretation leaves room for even more breadth), this question can have answers that fall anywhere on the 4 levels of social participation provided by Sones et al (2021), referencing Levasseur et al (2010):

interacting with others without doing a specific activity with them
doing an activity with others
helping others
contributing to society

The rationale behind using this outcome is that it provides an insight into a respondant’s social connection stemming from the community, reaching outside the immediate building community. The use of such an outcome also means that this analysis would account for the structural aspect of social connectedness (Sones et al., 2021).

Frequency of Conversations with Neighbours

This outcome consists of three levels - “Never or almost never”, “Several times a year”, and “Several times a week”. It had a non-response rate of about 12%. This outcome is representative of “social integration” mentioned by Sones et al (2021) and can be used as a proxy for this concept in analysis. Social integration is defined as “Participation in a broad range of social ties…including family cohesion; network size and diversity, relationship type (e.g. spouse, parent), social contact frequency.”. In this case, frequency of conversations with neighbours would fall best under social contact frequency.

Furthermore, this concept is also explored as an proxy for peoples’ engagement in social interaction by Muhuri & Basu (2020) when looking to understand what sorts of spaces are most conducive to these interactions. Nguyen et al (2020) took a similar approach when looking at how the design of housing influences the community of a building. In their case, frequency of conversations with neighbours was one part of how they conceptualised a “building community” in their analysis.

Finally, to return to Sones et al (2021), “social integration” is another part of this structural aspect of social connectedness, which is an important part of a sense of community and which provides a strong contender for understanding the desire to engage in social connection in this population.

Close Friends Network

This outcome has three levels - “0”, “Some”, and “6+”. It is an amalgamation of a three-part question (see data dictionary) in the survey that originally had 4 possible answers to each - “0”, “1-3”, “4-6”, and “More than 6”. This naturally led to a high diversity of possible response combinations and the existance of categories with too few people in them to be useful in analysis. To deal with this, these questions were collapsed into one variable, and the four categories became three. To be included in the “0” category of the “Close Friends - All” variable, one must have answered “0” to all three original questions. To be included in the “6+” category of the “Close Friends - All” variable, one must have answered “More than 6” to all three original questions. The “Some” category that remains was the “catch-all” bin for any other possible response combination. It had a non-response rate of about 12%.

This outcome can represent two possible constructs explained by Sones et al (2021): social integration and perceived social support. Social integration has to do with the size of one’s network in this case - the larger the social network experienced by the respondant. However, the nature of the question also gives us insight into how the respondant perceives their social support, since the nature of the original questions (discussing private matters, calling for help, and seeing once per month) give us an understanding of the strength of the social support of a respondant. Using this outcome would allow us to explore another aspect of social connection: the functional aspect. This makes it a particularly strong candidate, since it explores not only the existance and number of social contacts, but also how they create value and meaning for those who hold them. As a final note, the relationship between having close non-kin contacts and loneliness was also explored by de Jong Giervald et al (2015), which is worth a mention since loneliness is a predictor variable being in this analysis.

Pairwise Analyses

As a first step, chi-squared tests were performed with all outcome and predictor variables. This was performed to better understand the influence (albeit unadjusted) of the predictors on the outcomes. These were performed on a dataset that combined both housing providers, which provided a greater sample size. Furthermore, most of the outcomes and predictors were collapsed to house fewer categories. This was done to improve the accuracy of any tests run on this data, since it consolidated categories that were too small to run meaningful tests.
Moving forward, we would like to employ a backwards stepwise logistic regression model, starting with all possible predictor variables included and systematically eliminating those predictors that are deemed to not influence the outcome. The code for this model is proposed to be as follows:

library(readxl)
Cleaned_Both <- read_excel("~/Documents/Cleaned - Both.xlsx")

library(MASS)

Cleaned_Both$neighbour_convo<-factor(Cleaned_Both$neighbour_convo, ordered=TRUE, levels=c("Never or almost never", "Several times a year", "Several times a week"))

convo_model <- polr(neighbour_convo ~ phys_health + mhealth + tenure + safety_neighbourhood + lonely_all + web_social + age_senior + gender + eth_all + ls_all + imm_status, data = Cleaned_Both, method = "logistic", Hess=TRUE)

summary(convo_model)

## Call:
## polr(formula = neighbour_convo ~ phys_health + mhealth + tenure + 
##     safety_neighbourhood + lonely_all + web_social + age_senior + 
##     gender + eth_all + ls_all + imm_status, data = Cleaned_Both, 
##     Hess = TRUE, method = "logistic")
## 
## Coefficients:
##                                   Value Std. Error t value
## phys_healthNeutral              -0.1920     0.2272 -0.8448
## phys_healthPoor                 -0.2566     0.3037 -0.8449
## mhealthNeutral                   0.1777     0.2653  0.6700
## mhealthPoor                      0.1205     0.4074  0.2959
## tenure11+ years                  0.3054     0.2411  1.2667
## tenure4-10 years                 0.1998     0.2353  0.8490
## safety_neighbourhoodUnsafe      -0.2389     0.2982 -0.8010
## lonely_allOften                 -1.4057     0.4470 -3.1450
## lonely_allSometimes             -0.4576     0.2051 -2.2316
## web_socialMore than once a day   0.4330     0.2190  1.9767
## web_socialNever or almost never -0.1238     0.2373 -0.5219
## age_seniorUnder 65               0.1110     0.2355  0.4713
## genderMale                       0.2951     0.1927  1.5310
## eth_allWhite                     0.5405     0.2540  2.1279
## ls_allMore than one             -0.2323     0.2560 -0.9075
## imm_statusImmigrant              0.5569     0.2468  2.2568
## 
## Intercepts:
##                                            Value   Std. Error t value
## Never or almost never|Several times a year -1.1329  0.3744    -3.0258
## Several times a year|Several times a week   1.3341  0.3748     3.5593
## 
## Residual Deviance: 844.3679 
## AIC: 880.3679 
## (276 observations deleted due to missingness)

exp(coef(convo_model))

##              phys_healthNeutral                 phys_healthPoor 
##                       0.8253294                       0.7736536 
##                  mhealthNeutral                     mhealthPoor 
##                       1.1944989                       1.1280891 
##                 tenure11+ years                tenure4-10 years 
##                       1.3571038                       1.2211712 
##      safety_neighbourhoodUnsafe                 lonely_allOften 
##                       0.7875037                       0.2451975 
##             lonely_allSometimes  web_socialMore than once a day 
##                       0.6328027                       1.5418427 
## web_socialNever or almost never              age_seniorUnder 65 
##                       0.8835219                       1.1173704 
##                      genderMale                    eth_allWhite 
##                       1.3432409                       1.7168983 
##             ls_allMore than one             imm_statusImmigrant 
##                       0.7927099                       1.7451927

Cross-Tabulation for Loneliness Subquestions and Friend Network Subquestions

From the table below, we can observe that…

92.2% answered “Hardly Ever” to feeling left out if they answered “Hardly ever” for the two other questions
70.1% answered “Often” to feeling left out if they answered “Often” for the two other questions
71% answered “Sometimes” to feeling left out if they answered “Sometimes” for the two other questions

##                                                loneliness_leftout Hardly ever Often Sometimes Total
## loneliness_isolated loneliness_lack _companion                                                     
## Hardly ever         Hardly ever                                           288     4         6   298
##                     Often                                                   3     0         0     3
##                     Sometimes                                              40     3        15    58
## Often               Hardly ever                                             1     3         1     5
##                     Often                                                   2    34         6    42
##                     Sometimes                                               1     3         6    10
## Sometimes           Hardly ever                                            14     0        10    24
##                     Often                                                   5     3        13    21
##                     Sometimes                                              17     3        81   101