Aim 1. Factors driving health
Identifying factors that drive health: traditional health influencers, social determinants of health, and sense of community health
Of 21 items used for EFA and CFA, we retained 18 items based on model fit and reasonable loading. The three items dropped were smoking, genetic makeup inherited from parents, and having health insurance. Three factors were extracted from these 18 items to represent health drivers: Factor 1 (F1) represents Traditional Health Influencers (THI) and is defined by four indicators; Factor 2 (F2) represents Social and Economic Determinants of Health (SDoH) and is defined by ten indicators; Factor 3 (F3) represents Sense of Community Health (SoC) and is defined by four indicators.
FMM results for factor loadings (Table 1) show how each factor is represented by its composite indicators. Loading values are all moderate to high. Among these, the highest loading indicator for F1 (THI) is access to affordable healthcare, for F2 (SDoH) is housing quality, and for F3 (SoC) is endorsement of the statement my community works together to make positive change for health.
Table 1. Factor loadings for each factor driving health, assuming fixed values across latent classes
Factors and composite indicators
|
Standardized loadings
|
F1: Traditional Health Influencers
|
|
1. Access to affordable healthcare
|
0.66
|
2. Stress
|
0.52
|
3. Knowledge about health
|
0.63
|
4. Personal health practices other than smoking
|
0.51
|
F2: Social Determinants
|
|
5. Having a job
|
0.52
|
6. Neighborhood options for healthy food and exercise
|
0.59
|
7. Amount of social support
|
0.59
|
8. Physical environment such as clean air or water
|
0.65
|
9. Income
|
0.61
|
10. Community safety
|
0.66
|
11. Housing quality
|
0.71
|
12. Education
|
0.58
|
13. Where a person lives
|
0.64
|
14. Race/Ethnicity
|
0.46
|
F3: Sense of Community Health
|
|
15. My community can work together to improve its health
|
0.83
|
16. My community has the resources to improve its health
|
0.69
|
17. My community works together to make positive change for health
|
0.84
|
18. I know my neighbors will help me stay healthy
|
0.78
|
Revealing varied comprehension of drivers of health: Classes of respondents by patterns of indicator and factor endorsement
FMM results for LCA specify “classes” of respondents based on their pattern of endorsement of health drivers. Figure 1 presents the four extracted latent classes. The x-axis presents the indicators in the following order (or see Table 1): indicators 1-4 define THI; 5-14 define SDoH; and 15-18 define SoC.
Each class is represented by a set of color-coded lines. The dots on the lines correspond to conditional probabilities (on the y-axis) of endorsing each indicator given class membership. For example, those in class 3 have approximately a 77% probability of endorsing indicator 1 (“access to affordable health care”) as being an important or very important driver of health.
Each class exhibits a distinctive pattern of indicator and factor endorsement. Class 4 is defined by a near zero percent endorsement of indicators for both THI and SDoH. Classes 1, 2, and 3 all have relatively high endorsement of indicators defining THI (greater than 70% probability for all indicators). Classes 1 and 2 are high endorsers of SDoH (greater than 70% probability for most indicators), but probability of class 3 endorsement of SDoH indicators generally hovers around 50%. Notably, for all 3 groups, indicator 14, which assesses race/ethnicity as a driver of health, was endorsed at a substantially lower percentage for classes 1, 2, and 3 compared to endorsement of other SDoH indicators. In class 2, all other SDoH indicators were endorsed at 70% of the sample or higher, but race/ethnicity was endorsed at approximately 45%. For SoC indicators, classes 1, 3, and 4 endorsed them at low percentages (all indicators under were 15%). Class 2 endorsed SoC indicators at a higher percentage (approximately 15%-25%) compared to the other classes, though these percentages are still low.
Examining the distribution of survey respondents within classes, class 3 is the largest class at 88.3% of the sample. This class represents high endorsers of THI, moderate endorsers of SDoH, and low endorsers of SoC. As the majority of respondents are in class 3, it was used as the reference group in later analyses. All other classes contain very small percentages of respondents: class 1, 5.7%; class 2, 3.2%; class 4, 2.8%. Respondents of class 2 are of particular interest as they are high endorsers of both THI and SDoH, and the highest endorsers of SoC, potentially representing the group with the most comprehensive understanding of the factors that drive health.
Examining variation in class membership by demographics
We examined how these four classes differed by key demographics. Table 2 presents summary statistics of demographics by class.
Table: 2. Demographic characteristics of respondents by class*
|
Class 1
|
Class 2
|
Class 3
|
Class 4
|
Percent of total sample
|
5.7%
|
3.2%
|
88.3%
|
2.8%
|
Age mean (SD)
|
57.0 (15.7)
|
44.4 (15.1)
|
54.2 (15.8)
|
43.3 (16.7)
|
% Male
|
39.3%
|
64.1%
|
46.2%
|
50.8%
|
% Married or living with partner
|
55.6%
|
55.9%
|
65.4%
|
54.3%
|
% Unemployed
|
7.9%
|
5.2%
|
4.6%
|
12.1%
|
% Some college or more
|
67.2%
|
46.3%
|
74.0%
|
39.2%
|
% Household size of 1 person
|
23.4%
|
22.1%
|
20.3%
|
20.1%
|
% Living in rural area
|
17.4%
|
25.7%
|
13.1%
|
10.6%
|
% Living in large city
|
69.2%
|
39.0%
|
54.0%
|
61.8%
|
% Has insurance
|
92.4%
|
93.4%
|
94.3%
|
85.0%
|
Household income in dollars
|
|
|
|
|
Less than 10,000
|
2.5%
|
22.8%
|
3.5%
|
16.6%
|
10,000 to 24,999
|
13.2%
|
33.1%
|
10.3%
|
17.6%
|
25,000 to 49,999
|
21.3%
|
23.5%
|
21.0%
|
23.6%
|
50,000 to 74,999
|
18.6%
|
10.3%
|
19.0%
|
16.1%
|
75,000 to 99,999
|
11.4%
|
2.2%
|
13.3%
|
8.0%
|
100,000 or more
|
33.3%
|
8.1%
|
32.9%
|
18.1%
|
Race
|
|
|
|
|
Black
|
24.2%
|
33.1%
|
7.4%
|
18.1%
|
Hispanic
|
19.4%
|
27.2%
|
12.2%
|
20.6%
|
Non-Hispanic White
|
51.3%
|
29.3%
|
74.2%
|
52.8%
|
Asian
|
4.1%
|
5.2%
|
3.0%
|
4.5%
|
Other
|
1.0%
|
5.2%
|
3.2%
|
4.0%
|
*Note: Class refers to classes identified by the latent class analysis
Class 1
LCA results indicated that Class 1 respondents were, on average, high endorsers of both THI and SDoH as drivers of health (though not SoC). Compared to other classes, class 1 respondents were least likely to be male (39%) and most likely to be living in large cities (69%). They had the second highest percentage of respondents with at least some college (67%) after class 3 (74%). They had higher household incomes compared to other classes, i.e., they have the highest percentage of respondents in the highest income category ($100,000+, 33%) and the lowest percentage of respondents in the lowest income category (less than $10,000, 2.5%). Class 1 had the second highest percentage of Black respondents (24%) after class 2 and non-Hispanic whites were the majority (51%) in this class.
Class 2
Class 2 respondents were identified in the LCA as having the broadest comprehension about what drives health, as evidenced by high endorsement of THI and SDoH, as well as a distinctly higher endorsement of SoC, compared to other classes. Class 2 respondents were more likely to be male (64%), living in rural areas (25.7%), and least likely to be living in large cities (39%). They had lower household incomes compared to other classes, i.e., they had the highest percentage of respondents in the lowest income category (less than $10,000, 22.8%) and the lowest percentage of respondents in the highest income category ($100,000+, 8%). Class 2 had the highest percentages of Black (33%) and Hispanic (27%) respondents of all classes and class 2 is the only class in which non-Hispanic whites were not the majority (29%).
Class 3
LCA results indicated that most of the sample was represented in class 3 (88%) and they were high endorsers of THI but had lower likelihood of endorsing SDoH and SoC. Class 3 had the highest percentages of married respondents (65%) and respondents having at least some college (74%) education. About one-third of them were in the highest household income category, similar to class 1. A vast majority of respondents in this category were non-Hispanic white (74%), which is higher than all other classes.
Class 4
Class 4 were low endorsers of all three drivers of health. They had the highest percentage among all classes of unemployed respondents at 12.1%. They had moderate to low household incomes with 58% falling into the lowest three income categories ($49,999 and lower). Non-Hispanic whites were the majority (53%) in this class.
Aim 2. Association between understanding of health drivers and health equity views
Examining differences in perceptions of the importance of health equity, we observed differences by demographics and class membership across equity beliefs (Table 3).
Table 3. Average partial effects from regressions of demographics (independent variables) on health equity beliefs (dependent variables)
Health equity beliefs (dependent variable)
|
1. Making sure that the disadvantaged have an equal opportunity to be healthy
|
2. It would be unfair if some people had more of an opportunity to be healthy than other people
|
3. Our society needs to do more to make sure that everyone has ‘an equal’/ ‘a fair and just’ opportunity to be healthy
|
N, % endorsing highest response category
|
3200, 44.7%
|
2236, 31.3%
|
2920, 40.8%
|
Class membership (ref= class 3)
|
|
|
|
class 1
|
0.24*
|
0.20*
|
0.25*
|
class 2
|
0.24*
|
0.14*
|
0.24*
|
class 4
|
-0.22*
|
-0.23*
|
-0.30*
|
Age
|
0.001*
|
0.002*
|
0.002*
|
Education (ref= high school diploma)
|
|
|
|
Less than high school
|
0.05
|
-0.01
|
0.04
|
Some college, associate's degree
|
0.06*
|
0.06*
|
0.07*
|
Bachelor's degree or more
|
0.11*
|
0.07*
|
0.13*
|
Household size (ref = 2)
|
|
|
|
Household size 1 person
|
-0.05*
|
-0.02
|
-0.02
|
Household size 3 or more
|
-0.01
|
-0.00
|
-0.01
|
Household income (ref= between $75,000 and $124,999)
|
|
|
|
< $24,999
|
0.18*
|
0.12*
|
0.16*
|
$25,000 to $49,999
|
0.06*
|
0.08*
|
0.07*
|
$50,000 to $74,999
|
0.04*
|
0.04*
|
0.04*
|
> $125,000
|
-0.03
|
-0.03
|
-0.03
|
Has insurance (ref = none)
|
0.01
|
-0.02
|
-0.001
|
Being male (ref = female)
|
-0.14*
|
-0.08*
|
-0.09*
|
Married (ref = not married)
|
-0.05*
|
-0.02
|
-0.02
|
Unemployed (ref = not unemployed)
|
0.02
|
0.05*
|
0.07*
|
Race/ethnicity (ref = Non-Hispanic White)
|
|
|
|
Black
|
0.17 *
|
0.16*
|
0.15*
|
Hispanic
|
0.13*
|
0.08*
|
0.09*
|
Asian
|
-0.12*
|
-0.05
|
-0.004
|
Other
|
0.04
|
0.08*
|
0.04
|
Living in rural area (ref = not living in rural area)
|
-0.07
|
-0.04*
|
-0.09*
|
Living in large city (ref = not living in large city)
|
-0.001
|
0.01
|
0.03*
|
Note: *p values <.05.
Table 3 provides the average partial effect of each independent variable on the probability of endorsing each equity belief. A separate logistic regression was run for each belief. For a categorical variable, the average partial effect is interpreted as the average difference in the probability of endorsing the belief item for that category compared to the reference group, where the average is over the possible values of all the other covariates. For a continuous variable, the average partial effect represents the average effect of a one unit increase in the variable on the probability (on a scale of 0 to 1) of endorsing the belief item, where the average is over all the possible values of all the other covariates, including the continuous variable. Intuitively, the average partial effects can be interpreted in a similar manner to the coefficients in a linear probability model. For example, looking at the effect of gender on belief 1, the estimate of -0.14 for gender indicates that on average, the probability that men endorsed belief 1 is 0.14 points lower than the probability that women endorsed this item, where the average is over all the possible values of the other covariates. Men are, on average, 0.14 points less likely to think this item is important compared to women.
Across three health equity beliefs, we observed some consistent endorsement patterns. Respondents who were more likely to endorse all three beliefs tended to be female, older, Black or Hispanic (vs. non-Hispanic white), have more education (some college or Bachelor’s degree compared to those with only a high-school degree), and have lower household incomes (less than $74,999 vs. between $75,000 and $124,999). We also found that respondents who strongly agreed with beliefs 2 and 3 were more likely to be unemployed and living in non-rural areas.
Class membership was examined in the logistic regression. We found the same endorsement response patterns across all three health equity beliefs by class in terms of direction of association and significance. Class 2 on average, had a significantly higher probability (.24 percentage points) of endorsing belief 1 compared to class 3 (the reference group), and in previous analyses, class 2 was found to have the highest comprehension of health drivers overall, strongly endorsing THI and SDoH and endorsing SoC higher than the other classes. Class 2 had similar patterns of significantly higher probabilities of endorsement of the two other beliefs compared to class 3 (.24 and .14, respectively). Members of class 1 had, on average, a significantly higher probability (.24 percentage points) of endorsing belief 1 compared to members of class 3, and this pattern was similar for the other beliefs. Members of class 1 were also strong endorsers of both THI and SDoH in previous analyses. By contrast, class 4 had, on average, a significantly lower probability (.22 percent points) of endorsing belief 1 compared to class 3 and 4’s endorsement patterns and were similar for the other beliefs. They were also weak endorsers of any health driver factors in previous analyses.