Preventive Medicine Reports 27 (2022) 101768

Contents lists available at ScienceDirect

Preventive Medicine Reports

journal homepage: www.elsevier.com/locate/pmedr

The joint effect of multiple health behaviors on odds of diabetes, depression

Madison Sheffield a, *, Carol Lewis b
a
California Department of Public Health, 1616 Capitol Ave, Sacramento, CA 95814, USA
b
University of Florida, Psychology Department, 4037 NW 86 Terrace, Gainseville, FL 32606, USA

A R T I C L E I N F O A B S T R A C T

Keywords: This study examines the relationship between health determinant behaviors themselves, and their subsequent
Multiple health behavior change combined relationship with chronic illness (diabetes/impaired glucose regulation, depression). While numerous
MHBC studies have proven the benefits of engaging in more healthy behaviors, the question has not been answered
Depression
whether the effect of multiple healthy behaviors together is greater than the sum of the effects alone.
Diabetes
Health promotion
The study design is cross-sectional, using data on the adult population from the 2017 California Health
Chronic illness Interview Survey (CHIS).1 A total of 21,116 participants were included in final analyses. We used multivariable
Primary prevention adjusted logistic regression to calculate odds ratios for diabetes and for depression at each subsequent level of a
Mental health healthy lifestyle index (HLI). We also calculated the adjusted odds ratios between adjacent levels of the index.
Mental illness The odds of having depression and, separately, of having diabetes each decreased with each additional healthy
Impaired glucose regulation lifestyle behavior, with three of five depression ratios significant at p < 0.05, and four of five significant for
Dysglycaemia diabetes. The magnitude of the association between the HLI level and odds for disease declines exponentially
with each additional healthy lifestyle factor, contrary to the hypothesis, for depression, but fits the hypothesis for
diabetes. Our results are important for health promotion, suggesting that even one healthy behavior may
dramatically decrease the odds for having depression, regardless of the type of healthy behavior chosen. Our
results also show an association between lower prevalence of depression and health behaviors historically only
considered preventive for physical illness.

1. Introduction longitudinal and intervention studies have found that physical activity
has a direct positive impact on mental health outcomes, there is also
The burden of chronic illness on individuals and on society is well- evidence that physical activity may promote prosocial behavior and
documented (Centers for Disease Control and Prevention. Health and relatedness, which in turn have direct impacts on long-term health
Economic Costs of Chronic Disease. Published, 2019). The Centers for (World Health Organization, 2020; Harold et al., 2013). Likewise, recent
Disease Control and Prevention has stated that the majority of chronic studies have determined that loneliness is as impactful on early mor­
illness can be traced back to only a few high-risk behaviors: poor diet, tality as smoking is, and one of the pathways to explain this may be that
inactivity, and tobacco and alcohol use (Centers for Disease Control and diet quality and physical activity levels can suffer as a result of social
Prevention. How You Can Prevent Chronic Diseases. Published, 2019). A isolation (Holt-Lunstad et al., 2015; Conklin et al., 2014; Whitelock and
growing body of research examines how these healthy behaviors may Ensaff, 2018; Chen et al., 2018).
beget each other (Fleig et al., 2014, 2015; Nair et al., 2015; Priebe et al., Recent research on “multiple health behavior change” (MHBC2) ex­
2017). A prime example of this relationship is illustrated in research plores the efficacy of changing several high-risk behaviors simulta­
showing that a full night of sleep is both directly beneficial for chronic neously in order to achieve improved habit-formation and health impact
disease prevention and also increases the likelihood of engaging in other (Amireault et al., 2018). While it is clear from numerous studies that
health-promoting behaviors such as exercise, maintenance of a healthy engaging in more healthy behaviors rather than less confers greater risk
diet, and emotional regulation (Atkinson and Davenne, 2007; Dashti reduction for many chronic illnesses, the question has not been
et al., 2015; Goldstein and Walker, 2014; Greer et al., 2013). While answered whether the effect of multiple healthy behaviors together is

* Corresponding author.
E-mail addresses: madisonsheffield@gmail.com (M. Sheffield), carollewis@ufl.edu (C. Lewis).
1
CHIS: California Health Interview Survey.
2
MHBC: multiple health behavior change

https://doi.org/10.1016/j.pmedr.2022.101768
Received 29 April 2021; Received in revised form 9 March 2022; Accepted 13 March 2022
Available online 15 March 2022
2211-3355/Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
M. Sheffield and C. Lewis Preventive Medicine Reports 27 (2022) 101768

greater than the sum of the effects alone (Kvaavik et al., 2010; Lv et al., contributed to the 2017 survey, with a total of 21,116 adults included in
2017; Ford et al., 2009; Aleksandrova et al., 2014). To the best of our the present analysis due to missing data on diabetes and/or depression
knowledge, a study has yet to examine MHBC under the hypothesis that diagnosis for 37 individuals.
interaction between healthy behaviors could provide additional benefit
for well-being, beyond the impact of the healthy behaviors acting 2.2. Type II diabetes and depression ascertainment
individually.
We explore the association between multiple healthy behaviors For the purposes of this study, the type II diabetes outcome is broadly
combined and odds of experiencing one of two of the more common defined using a proxy of impaired glucose regulation: by “yes” responses
chronic conditions: type II diabetes, via a glucose regulation impairment to any question determining if a doctor had ever given the respondent a
proxy, and depression. While researchers have explored similar ques­ diagnosis of type II diabetes, double diabetes, or prediabetes (does not
tions in respect to cancer, heart disease, and diabetes, this is the first include gestational diabetes). The depression outcome was defined
study we are aware of to explore the relationship of a mental illness using a score of 13 or above on the validated Kessler-6 scale, which
outcome variable with the same healthy behaviors generally included in screens for general psychological distress, and also measures severity.
MHBC research (Kvaavik et al., 2010; Lv et al., 2017; Aleksandrova Respondents are asked how often in the past 30 days have they felt:
et al., 2014; Investigation, 2009). Primary prevention of mental illness is nervous; hopeless; restless or fidgety; worthless; “so depressed that
less studied than methods of early intervention, crisis stabilization, or nothing could cheer you up”; “that everything was an effort”. Answers
other evidence-based interventions applicable in cases of more severe All, Most, Some, A little, and None of the time are scored 4 to 0 points,
mental illness. We seek to add to the conversation on critical and respectively. A cumulative score of 13 and above is widely used to
effective, low-barrier, daily-use methods for maintaining mental well- categorize “serious mental illness.” (Prochaska et al., 2012) While the
being that are available at the public health scale. Kessler-6 scale is not a clinical depression diagnosis tool, investigations
We use an evidence-based list representing the core set of health of the ability of this scale to predict diagnosis (including depression)
behaviors that are broadly agreed upon by public health experts and are based on the Diagnostic and Statistical Manual of Mental Disorder and
included in previous MHBC studies, as available in the data source: known correlates of mental illness severity substantiate its use for proxy
healthy dietary factors, physical activity, avoidance of tobacco use, measures, such as it is used here for depression (Mitchell and Beals,
community engagement, and water consumption (Kvaavik et al., 2010; 2011).
Lv et al., 2017; Aleksandrova et al., 2014; Investigation, 2009). We
hypothesize that the core list of primary prevention health behaviors for 2.3. Healthy behavior ascertainment
maintenance of mental well-being generally mirrors that for physical
well-being, as indicated in a growing body of research exploring the The list of healthy behaviors was determined based on apriori
etiology of mental illness (Atkinson and Davenne, 2007; Dashti et al., knowledge of chronic disease prevention and on extensive literature
2015; Goldstein and Walker, 2014; Greer et al., 2013; World Health review, and represent a core set of healthy behaviors used across MHBC
Organization, 2020; Harold et al., 2013; Holt-Lunstad et al., 2015; and primary prevention research. The determination of the point at
Conklin et al., 2014; Whitelock and Ensaff, 2018; Chen et al., 2018). which to dichotomize each variable into “healthy” and “unhealthy”
(Note: The variables available for inclusion in this study are naturally behavior was based on bodies of research demonstrating the relation­
limited by those in the data source. The CHIS 2017 survey did not ship between the lifestyle factor and chronic disease, including national
include questions on alcohol consumption or sleep quantity/quality.) public health recommendations. We build upon the body of research
It is worth noting the possibility of co-occurrence of impaired glucose that uses this large dataset, however that comes with natural limitations
regulation and depression. Evidence indicates that the co-occurrence of in the variables used within this study.
these two conditions leads to worse health outcomes and higher health
care costs (Alva and Ikeda, 2020). In addition, these conditions may 2.3.1. Dietary factors
share underlying biological and behavioral causes, thus increasing the This is a composite variable formed from weekly soda consumption
value of determining healthy behaviors that positively impact both (Holt and daily fruit and vegetable intake. Average monthly fruit intake and
et al., 2014). vegetable intake were used to calculate an average combined fruit and
We hypothesize that the odds of having impaired glucose regulation, vegetable intake per day. Respondents who consumed an average of less
and of having depression, are lower with each additional healthy than four sodas per week or consumed five or more fruits and vegetables
behavior exhibited and that the magnitude of the change in odds is per day were considered to have “healthy” dietary factors (U.S. 2015;
greater with each subsequent additional healthy behavior. Malik et al., 2010).
The implications of these findings are pertinent to health promotion
and health education professionals, to individuals who experience 2.3.2. Exercise
mental illness and the professionals that support them, as well as those While extensive research shows a dose–response relationship be­
that work in health systems and policy. tween exercise and disease, the World Health Organization now rec­
ommends getting 150-mintues per week of moderate moderate-intensity
2. Methods aerobic physical activity to prevent chronic illness (World Health Or­
ganization. Physical Activity Fact Sheet, 2020). Thus, exercise was
2.1. Study design and population dichotomized as “healthy” if the respondent had exercised five or more
times in the past seven days (an average of 30 min of per event is thus
The present study examines the adult population from the 2017 assumed, to meet the recommended weekly minimum of 150 min).
California Health Interview Survey (CHIS) (University of California
Canter for Health Policy Research, 2018). CHIS is the largest state health 2.3.3. Smoking
survey in the United States, asking questions on a wide-range of health This variable is dichotomized based on whether respondent has
topics via telephone response (Survey and Design, 2019). The survey is smoked 100 or more cigarettes in their lifetime.
conducted across California on a rolling basis each year (Survey and
Design, 2019). The approval and methodology for this survey of is 2.3.4. In-person community engagement
explained in detail in CHIS resources (Survey and Design, 2019). This variable was also composed of multiple variables related to in-
Approval for the present study was granted by the California Committee person engagement, as available in the CHIS dataset. “Healthy” com­
for the Protection of Human Subjects. A total of 21,153 adults munity engagement was determined if respondents answered “Agree” or

2
M. Sheffield and C. Lewis Preventive Medicine Reports 27 (2022) 101768

“Strongly Agree” to one or more of the following statements: people in and of depression, with and without adjustment by the aforementioned
the neighborhood are willing to help each other, people in the neigh­ covariates. Participants with all five healthy factors are the reference
borhood get along, people in the neighborhood can be trusted; and the group. HLI was modeled as a categorical variable, under the assumption
respondent must have also stated that they had completed volunteer from which the hypothesis is based – that the relationship between HLI
work in the community in the past year. This variable is built on a and odds of impaired glucose regulation/depression is not linear.
growing body of research showing that social isolation is as strong a Additionally, we executed contrast analyses to allow for comparison
predictor of early mortality as tobacco use (Holt-Lunstad et al., 2015). between each two sequential levels of the HLI. We also present HLI
modelled ordinally, for comparison.
2.3.5. Water consumption All statistical analyses were performed using the Statistical Analysis
Consumption of sugar-sweetened beverages or other calorie-positive System (SAS) Version 9.4; SAS Institute, Inc., Cary, NC, USA. P-values
beverages is very common in lieu of water consumption, which is were based on two-sided tests, and p < 0.05 was considered statistically
necessary for normal body functioning, thereby dramatically altering significant.
energy balance (Panel on Dietary Reference Intakes for Electrolytes and
Water and Committee, 2005; Dubois et al., 2007; Dennis et al., 2010; 2.7. Theory
Jahns, 2019). A specific recommendation on water intake is not given by
the USDA due to the variability based on body size, water content in the The use of a combined HLI is the most common method for exam­
diet, activity level and environmental factors (Panel on Dietary Refer­ ining multiple health behavior change impact on disease (Kvaavik et al.,
ence Intakes for Electrolytes and Water and Committee, 2005; Jahns, 2010; Aleksandrova et al., 2014; Investigation, 2009). While the method
2019). Thus, we calculated an average amongst the total study popu­ we used in this study is the most common in the field of MHBC research,
lation of the number of “times in the previous day” that water was we urge other studies upon this or similar research questions to consider
consumed, and a healthy level of water consumption was determined as using a model with interaction terms instead. There is no evidence for
greater than this population average. No other measure of water quan­ use of this method in such application at the time of completion of this
tity nor adequacy per individual was captured in the survey data. analysis.

2.4. Covariates 3. Results

It is well established that healthy behaviors often cluster together as 3.1. Descriptive analyses
a result of age, gender, education level, income, and access to other
health-promoting determinants (Marmot, 2005; World Health Organi­ Of the total 21,116 adult study population, the mean age was 46
zation, 2017). These determinants in turn impact glucose regulation and (±0.19) years, and the cohort was 51% women. Overall, 19.3% were
depression outcomes. Thus, we included age, gender, physical disability classified as having impaired glucose regulation, and 10% with
status, and percent of federal poverty level as confounders in all adjusted depression. The prevalence of impaired glucose regulation varied
models. The prevalence of people who are physically disabled in the slightly across genders, with 18% of women and 20% of men classified
extent that they cannot walk is small for all groupings but is important as diabetic, while 11% of women and 8% of men were classified as
for both exercise ability and for disease outcomes. We present descrip­ having depression. The mean for the study population of percent of
tive analyses by race category as well, while echoing calls for improved federal poverty level was 450% (±4%) (equivalent to approximately
methodology for conceptualizing socioeconomic variables (Bowleg, $54,630 annual salary for one person in 2018 US Dollars, $112,950 for a
2019). All covariates were confirmed for confounding relationship in family of four), and that average drops in both the disease categories:
chi-square tests (p < 0.05). 413% (±9%) and 355% (±13%) for impaired glucose regulation and
depression, respectively (Office of the Assistant Secretary for Planning
2.5. Healthy lifestyle index (HLI) definition and Evaluation., 2018). The prevalence of each disease varies between
race categories: for both impaired glucose regulation and depression,
Each healthy lifestyle factor was dichotomized to allow for easy those who identified as White report the lowest prevalence, while
translation of research findings, alignment with current health guide­ impaired glucose regulation prevalence is highest for those who iden­
lines, and to allow for construction into an index variable. Study par­ tified as African American, and Hawaiian Island/Pacific Islanders report
ticipants were given a point for each of the aforementioned five healthy the highest depression prevalence. Percent of study participants who
lifestyle factors exhibited in their responses, with a point range of 0 to 5. engage in 0, 1, 2, 3, 4, and 5 healthy lifestyle behaviors follows a sym­
Thus, the HLI variable was a sum of all “healthy” behavior points, per metrical modal distribution. Further details, as well as percent of the
respondent, with 0 representing least healthy and 5 representing most cohort and of each race reporting each healthy lifestyle factor and each
healthy behaviors. HLI level are provided in Table 1.

2.6. Statistical analyses 3.2. Primary findings

In descriptive analyses we present prevalence of each healthy 3.2.1. Depression

behavior, each level of the HLI, and each disease outcome, for the full Every sequential reduction in healthy lifestyle index level from the
cohort and among women and men, and among seven race categories. reference group (HLI = 5 healthy behaviors) was associated with an
We also present averages and standard deviations for the full cohort and increased odds of depression, after adjusting for covariates. Compared
by disease outcome for age and percent of federal poverty level. We with participants at the HLI level of 5, the adjusted odds ratio for par­
present odds ratios with 95% confidence intervals (CI’s) for each disease ticipants with HLI of 4 was 1.04 (0.72–1.51), 1.14 (0.80–1.62) for 3
outcome modelled against each binary predictor alone, each binary factors, 1.48 (1.03–2.12) for 2 factors, 2.16 (1.45–3.22) for 1 factor, and
predictor adjusted for confounders, all binary predictors in a crude 3.38 (1.89–6.02) for those who reported 0 healthy behaviors. The ratios
model, and all binary predictors while adjusting for all confounders. between 0, 1, 2 compared to 5 were significant at p < 0.05. When
These results are shown in Tables 2(a) and (b). evaluated as an ordinal variable, each additional level of the HLI was
For the primary analysis, we present a full multivariable logistic associated with a 20% lower odds of depression (OR for a one-point
regression model including the binary predictors as the combined HLI increase on the HLI = 0.80, CI: 0.74–0.86). When each level of the
for odds of impaired glucose regulation (shown as “diabetes” in tables) HLI is compared, instead, against the next consecutive level (rather than

3
M. Sheffield and C. Lewis Preventive Medicine Reports 27 (2022) 101768

Table 1 odds ratios for depression decreases approximately exponentially with

Descriptive analyses. each additional healthy lifestyle behavior. In other words, there is less
Diabetes Depression change in chance of depression as each additional healthy behavior is
added. Odds ratios, 95% CI’s, beta coefficients and p-values for full
Total Yes No (%) Yes No (%)
(%) (%) models are presented in Table 3.

Participants 21,116 4,715 17,032 1,798 19,014

(19) (81) (10) (90)
3.2.2. Type II Diabetes/Impaired glucose regulation
Socio-demographic When compared to the group who reported 5 healthy behaviors and
characteristics adjusted for covariates, all ratios between fewer healthy behaviors and 5
Age (Mean, Standard 46 58 44 37 48 are statistically significant, except comparing 0 to 5. Compared with
Deviation) (0.19) (0.41) (0.21) (0.53) (0.20)
participants at the HLI level of 5, the adjusted odds ratio for participants
Women (%) 51 18 82 11 89
Men (%) 49 20 80 8 92
Percent of Federal 450 (4) 413 (9) 459 (5) 355 460 (5)
Poverty Level (Mean, (13)
Standard Deviation)
Race (%) – – – – –
Hispanic 36 20 80 11 89
White 41 17 83 9 91
African American 6 29 71 9 91
American Indian/ 0.4 23 77 12 88
Alaskan Native
Asian 14 18 82 7 93
Hawaiian/Pacific 0.4 18 82 13 87
Islander
Two or more races 2 20 80 21 80
Cannot walk % (Total) 0.4 – – – –
(1 8 4)

compared to 5 healthy factors), the odds ratios are: 1.56 (0.93–2.62) for
0 healthy behaviors versus 1, 1.47 (1.13–1.91) for 1 versus 2, 1.30
(1.05–1.60) for 2 versus 3, 1.09 (0.87–1.37) for 3 versus 4, and 1.04
(0.72–1.51) for 4 vs 5. Table 4 and Fig. 1 show that the magnitude of the Fig. 1. Odds ratio diagram – Depression.

Table 2
Precursor models.
(a) Single Variable Predictor Models.

Odds of Disease Diagnosis

Single Predictor Variable Models

Variable Crude Adjusted

Ref ¼ Healthy Behavior (0) OR 95% CI Beta p-Value OR 95% CI Beta p-Value

Depression Eating 1.71 1.40–2.10 0.27 <0.0001 1.55 1.25–1.92 0.22 <0.0001
Exercise 1.12 0.95–1.31 0.05 0.04 1.09 0.93–1.29 0.04 0.29
Smoking 1.66 1.42–1.93 0.25 <0.0001 2.66 2.22–3.12 0.48 <0.0001
Comm. Eng. 1.01 0.86–1.17 0.003 0.95 0.99 0.84–1.17 − 0.004 0.92
Water Drinking 0.90 0.77–1.05 − 0.05 0.17 0.98 0.84–1.15 − 0.01 0.82

Diabetes Eating 0.71 0.57–0.87 − 0.17 0.001 0.78 0.63–0.97 − 0.13 0.02
Exercise 1.27 1.12–1.43 0.12 0.0001 1.26 1.11–1.43 0.11 0.0004
Smoking 1.51 1.35–1.70 0.21 <0.0001 1.07 0.95–1.22 0.03 0.28
Comm. Eng. 1.28 1.14–1.44 0.12 <0.0001 1.11 0.98–1.25 0.05 0.11
Water Drinking 1.00 0.90–1.12 0.001 0.97 0.79 0.70–0.89 − 0.12 0.0001

(b) Multivariable Predictor Models

Odds of Disease Diagnosis

Multivariable models
Variable Crude Adjusted

Ref ¼ Healthy Behavior (0) OR 95% CI Beta p-Value OR 95% CI Beta p-Value

Depression Eating 1.66 1.36–2.04 0.25 <0.0001 1.42 1.14–1.77 0.18 0.002
Exercise 1.13 0.96–1.33 0.06 0.13 1.12 0.95–1.33 0.06 0.17
Smoking 1.62 1.39–1.89 0.24 <0.0001 2.59 2.18–3.07 0.48 <0.0001
Comm. Eng. 0.95 0.81–1.12 − 0.02 0.54 0.97 0.82–1.14 − 0.02 0.69
Water Drinking 0.86 0.74–1.01 − 0.07 0.07 0.98 0.83–1.15 − 0.01 0.76

Diabetes Eating 0.65 0.52–0.79 − 0.22 <0.0001 0.78 0.63–0.97 − 0.12 0.02
Exercise 1.27 1.13–1.44 0.12 <0.0001 1.29 1.13–1.46 0.13 0.0001
Smoking 1.53 1.37–1.72 0.21 <0.0001 1.08 0.95–1.23 0.04 0.23
Comm. Eng. 1.29 1.15–1.45 0.13 <0.0001 1.12 0.99–1.27 0.06 0.07
Water Drinking 0.98 0.87–1.10 − 0.01 0.67 0.77 0.68–0.88 − 0.13 <0.0001

4
M. Sheffield and C. Lewis Preventive Medicine Reports 27 (2022) 101768

Table 3 comparison—in line with the hypothesis—only the contrast between 4

Unadjusted, adjusted, and contrast HLI models. and 5 is statistically significant.
(a) Unadjusted HLI model
4. Discussion
Healthy Lifestyle Index (HLI) Models

HLI Level Crude

In this study, we used a sample of 20,000 + adults living in California
Ref ¼ 5 Healthy OR 95% CI Beta p-Value to examine the combined associations of healthy dietary factors,
Behaviors adequate physical activity, avoidance of tobacco use, in-person com­
Depression 0 vs 5 2.41 1.38–4.20 0.64 0.00 munity engagement, and/or water consumption with odds of having a
1 vs 5 1.52 1.05–2.21 0.18 0.07 diagnosis of impaired glucose regulation, and separately, of having ev­
2 vs 5 1.2 0.84–1.70 − 0.06 0.47
idence for depression. Our findings are consistent with previous studies
3 vs 5 1.02 0.72–1.44 − 0.22 0.01
4 vs 5 0.94 0.65–1.34 − 0.31 0.001 that have found that greater engagement in healthy behaviors is asso­
ciated with lower prevalence of chronic illness. Longitudinal studies
have shown that engaging in 4 healthy behaviors, compared to 0 or 1,
Diabetes 0 vs 5 1.69 1–2.85 − 0.03 0.89
1 vs 5 2.44 1.84–3.25 0.34 <0.0001 can significantly increase life expectancy and decrease cause-specific
2 vs 5 2.07 1.60–2.68 0.18 0.01 and all-cause mortality (Ford et al., 2009).
3 vs 5 1.86 1.44–2.39 0.07 0.30 The findings of our study, which solely present association via odds
4 vs 5 1.71 1.31–2.23 − 0.01 0.85
ratio, are also consistent with temporal studies that have found an
(b) Adjusted Model apparent diminishing of returns in risk reduction associated with each
Healthy Lifestyle Index (HLI) Models additional healthy behavior. Contrary to our hypothesis, the magnitude
of the change in odds of chronic disease decreases with each additional
HLI Level Adjusted
healthy behavior. (The odds of impaired glucose regulation associated
Ref ¼ 5 Healthy OR 95% CI Beta p-Value
with each additional healthy behavior do follow a trend similar to the
Behaviors
hypothesized association, however, only the contrast model comparison
Depression 0 vs 5 *3.38 1.89–6.02 1.22 <0.0001 between 4 and 5 healthy behaviors is statistically significant). Our re­
1 vs 5 *2.16 1.45–3.22 0.77 0.0001
sults for depression mirror those in similar studies in that the odds ratios
2 vs 5 *1.48 1.03–2.12 0.39 0.03
3 vs 5 1.14 0.80–1.62 0.13 0.48 comparing 0 through 4 healthy behaviors to 5 healthy behaviors
4 vs 5 1.04 0.72–1.51 0.04 0.83 decrease approximately exponentially. In fact, some studies have
observed that particular combinations of 2 healthy behaviors presented
Diabetes 0 vs 5 1.01 0.57–1.79 0.01 0.97 similar reductions in risk to a combination of all 4 healthy behaviors
1 vs 5 *1.39 1.03–1.86 0.33 0.03 (Ford et al., 2009). One study, however, found that the combination of
2 vs 5 *1.39 1.70–1.82 0.33 0.01 physical activity and healthy diet yielded a hazard ratio smaller than the
3 vs 5 *1.41 1.09–1.83 0.34 0.01
expected hazard ratio based on the individual factors alone (Ford et al.,
4 vs 5 *1.47 1.12–1.93 0.39 0.01
2009). This finding also supports the health promotion idea that
(c) Contrast Model
achieving a single positive health behavior change is the most difficult
Healthy Lifestyle Index (HLI) Models but potentially the most rewarding for long-term health. While the
HLI Level Contrast finding that a few simple healthy actions can have measurable impact is
good news for health education efforts, it runs contrary to our hypoth­
Ref ¼ 5 Healthy OR 95% CI Beta p-
Behaviors Value esis that interactions between healthy behaviors themselves could have
far greater impacts on odds for chronic disease. Still, it is clear that more
Depression 0 vs 5 1.56 0.931–2.619 0.45 0.09
1 vs 5 *1.47 1.125–1.910 0.38 0.001
combined healthy factors are associated with the smallest odds/risk/
2 vs 5 *1.30 1.053–1.600 0.26 0.01 hazard ratios for chronic illness, throughout the literature (Aleksan­
3 vs 5 1.09 0.870–1.370 0.09 0.45 drova et al., 2014; Kvaavik et al., 2010; Ford et al., 2009; Gopinath et al.,
4 vs 5 1.04 0.720–1.505 0.04 0.83 2010; Babey et al., 2016).
By exploring the relationship between well-established healthy be­
Diabetes 0 vs 5 0.73 0.425–1.256 − 0.31 0.26 haviors and mental well-being, our results augment the fields of MHBC
1 vs 5 0.99 0.814–1.213 − 0.01 0.95 and primary prevention. Our results indicate that healthy behaviors
2 vs 5 0.99 0.846–1.155 0.01 0.89
historically only associated with lower prevalence of physical illness are
−
3 vs 5 0.96 0.809–1.136 − 0.04 0.62
4 vs 5 *1.47 1.122–1.927 0.39 0.01 also associated with lower prevalence of mental illness. These findings
are important in a time of rising mental illness prevalence and increased
need for upstream, primary prevention measures, as well as effective
Table 4 treatment for mild cases of mental illness. In addition, due to the high
Ordinal model. rate of co-occurrence of mental and physical illness, these results should
inspire further exploration of multiple health behavior change in the
Adjusted ordinal model Ref ¼ 5 OR 95% CI Beta p-Value
Healthy Behaviors
context of addressing these two spheres of wellness simultaneously.

Depression 1.25 1.16–1.34 0.22

5. Limitations
<0.0001
Diabetes 1.01 0.96–1.06 0.01 0.79

Our results should be taken in the context of the limitations pre­

with HLI of 4 was 1.47 (1.12–1.93), 1.41 (1.09–1.83) for 3 factors, 1.39 sented. Most notably, the cross-sectional nature of the data introduces
(1.06–1.82) for 2 factors, 1.39 (1.03–1.89) for 1 factor, and 1.01 significant difficulty in understanding the temporality of the indepen­
(0.57–1.79), for those who reported 0 healthy behaviors. Evaluated dent and dependent endpoints tested here, and makes determinations of
ordinally, each additional level of the HLI was associated with a 0.7% causation impossible. This issue is clear in the descriptive results that
decrease in odds of impaired glucose regulation (OR for a one-point exhibit poor dietary factors and low water consumption are associated
increase in HLI = 0.993, CI: 0.94–1.04). While the contrast model with a decreased odds of impaired glucose regulation diagnosis, contrary
shows increasingly larger beta coefficients at each subsequent level of to hypotheses and public health recommendations. The fact that these

5
M. Sheffield and C. Lewis Preventive Medicine Reports 27 (2022) 101768

associations are also not observed in the odds for depression may indi­ 8. Research Data
cate that a past glucose regulation impairment diagnosis prompted
recent changes in diet or altered reporting on diet, while this temporality Due to the sensitive nature of the questions asked in the California
is not captured in the survey design. The outcome variables are likely Health Interview Survey, which was used for this study, survey re­
under-reported by this data source. Prediabetes and depression are spondents were assured raw data would remain confidential and would
chronically underdiagnosed, and persistent stigma around mental illness not be shared.
increases the likelihood for under-reporting (Corrigan, 2004; Sartorius,
2007). Finally, though it is not common practice in similar MHBC CRediT authorship contribution statement
studies, there is evidence to support applying weights to each behavior.
Diet, exercise, and tobacco use have been proven to have very strong Madison Sheffield: Conceptualization, Methodology, Formal anal­
relationships with chronic disease prevention. However, there is also ysis, Resources, Data curation, Writing – original draft, Writing – review
growing evidence that both objective and subjective loneliness are as & editing, Visualization, Supervision, Project administration, Funding
impactful on mortality as tobacco use, and while not included in this acquisition. Carol Lewis: Writing – original draft, Writing – review &
study due to the exclusion from the original data source, sleep has editing.
profound impacts on health behaviors and risk for chronic illness. Thus,
we posit that as more research explores the impacts of these as-yet Declaration of Competing Interest
underestimated health behaviors both on chronic disease and on each
other, their importance may become clearer. The authors declare that they have no known competing financial
interest or personal relationships that could have appeared to influence
5.1. Strengths the work reported in this paper.

To the best of our knowledge, this is the first study that has explored
Acknowledgements
the associations of interactions between healthy behaviors and their
subsequent impact on chronic illness. We used a method well-
This work was made possible by the California Epidemiologic
established in the literature base and parsimoniously used variables
Investigation Service (Cal-EIS) Fellowship Program, funded and
from an existing data set to examine health behaviors with strong evi­
administered by the California Department of Public Health. The pro­
dence base in chronic disease prevention. The use of CHIS data enables
gram is supported by the Centers for Disease Control and Prevention’s
statistically relevant results and increased generalizability due to the
Preventive Health Services Block Grant. We would like to thank the
large and diverse sample population and the survey’s robust design,
following for their significant contributions to this study. Dr. Jessica
implementation, and data processing.
Nuñez de Ybarra was a constant support throughout the design, anal­
MHBC research has historically sought to examine the potential
ysis, and completion of the study. Without her, this project could never
psychological benefits of changing multiple health habits simulta­
have been. She is indeed a champion of the next generation of public
neously, as opposed to one habit at a time. In focusing on MHBC, we use
health professionals, and I will never forget her support and mentorship.
a framework proven to contribute to primary prevention and health
In addition, Dr. Adam Readhead generously provided extensive tech­
improvement. Our study expands the investigation to an exploration of
nical assistance and expertise in the modeling process, for which we are
how healthy behaviors beget each other, and the overall impact on
indebted to him. Thank you also to Dr. Amiee Sisson for her continued
chronic illness. Furthermore, the results of this study add a new
devotion to the development of the next generation of epidemiologists,
perspective to the field of MHBC and to health promotion, in the
through her hard work on the Cal-EIS program, and her unwaivering
expansion of understanding the association of health behaviors and
support of all fellows and their dreams.
mental wellness. In addition, this study is the first we can find to
investigate the association of MHBC on mental illness.
References

6. Conclusion Aleksandrova, K., Pischon, T., Jenab, M., Bueno-de-Mesquita, H.B., Fedirko, V., Norat, T.,
Romaguera, D., Knüppel, S., Boutron-Ruault, M.-C., Dossus, L., Dartois, L., Kaaks, R.,
As mental illness incidence increases across the United States, pri­ Li, K., Tjønneland, A., Overvad, K., Quirós, J.R., Buckland, G., Sánchez, M.J.,
Dorronsoro, M., Chirlaque, M.-D., Barricarte, A., Khaw, K.-T., Wareham, N.J.,
mary prevention measures and easily-applicable treatment for mild Bradbury, K.E., Trichopoulou, A., Lagiou, P., Trichopoulos, D., Palli, D., Krogh, V.,
cases are increasingly important. Our results also suggest that engaging Tumino, R., Naccarati, A., Panico, S., Siersema, P.D., Peeters, P.HM., Ljuslinder, I.,
in even one healthy behavior may dramatically decrease the odds for Johansson, I., Ericson, U., Ohlsson, B., Weiderpass, E., Skeie, G., Borch, K.B.,
Rinaldi, S., Romieu, I., Kong, J., Gunter, M.J., Ward, H.A., Riboli, E., Boeing, H.,
having depression, regardless of the type of healthy behavior chosen. 2014. Combined impact of healthy lifestyle factors on colorectal cancer: a large
This can be a powerful message for health promoters and educators European cohort study. BMC Med. 12 (1) https://doi.org/10.1186/s12916-014-
because the impacts of healthy behavior upon mental well-being can be 0168-4.
Alva, M.L., Ikeda, N., 2020. Co-occurrence of diabetes and depression in the U.S. PLoS
immediate. Our study also highlights the need for further investigation One 15 (6), e0234718.
of the psychological and physiological pathways along which healthy Amireault, S., Fong, A.J., Sabiston, C.M., 2018. Promoting healthy eating and physical
behaviors beget other healthy behaviors. Further investigation of the activity behaviors: a systematic review of multiple health behavior change
interventions among cancer survivors. Am. J. Lifestyle Med. 12 (3), 184–199.
interaction between healthy behaviors, and their combined impact on
Atkinson, G., Davenne, D., 2007. Relationships between sleep, physical activity and
chronic disease prevention, should be strengthened with longitudinal human health. Physiol. Behav. 90 (2-3), 229–235.
and randomized control studies, and exploration of this question using a Babey SH, Wolstein J, Diamant AL, Goldstein H. Prediabetes in California: Nearly Half of
California Adults on Path to Diabetes.; 2016.
model with interaction terms.
Bowleg, L., 2019. Perspectives from the social sciences: critically engage public health.
Am. J. Public Health 109 (1), 15–16.
7. Disclosure of funding source Centers for Disease Control and Prevention. Health and Economic Costs of Chronic
Disease. Published 2019. Accessed May 19, 2019. https://www.cdc.gov/chronic
disease/about/costs/index.htm.
This work was supported by the California Department of Public Centers for Disease Control and Prevention. How You Can Prevent Chronic Diseases.
Health, through the California Epidemiologic Investigation Service Published 2019. Accessed May 19, 2019. https://www.cdc.gov/chronicdisease/abou
Fellowship program. t/prevent/index.htm.
Chen Y, Holahan C, Li X. LONELINESS, PHYSICAL ACTIVITY, AND SELF-RATED
HEALTH IN MIDDLE-AGED AND OLDER ADULTS. Innov Aging. 2018;2(suppl_1):967-
967.:10.1093/geroni/igy031.3582.

6
M. Sheffield and C. Lewis Preventive Medicine Reports 27 (2022) 101768

Conklin, A.I., Forouhi, N.G., Surtees, P., Khaw, K.-T., Wareham, N.J., Monsivais, P., women. Arch Intern Med. 170 (8), 711. https://doi.org/10.1001/
2014. Social relationships and healthful dietary behaviour: Evidence from over-50s archinternmed.2010.76.
in the EPIC cohort, UK. Soc. Sci Med. 100, 167–175. Lv, J., Yu, C., Guo, Y., et al., 2017. Adherence to a healthy lifestyle and the risk of type 2
Corrigan P. How stigma interferes with mental health care. Am Psychol. 2004;59(7):614- diabetes in Chinese adults. Int J Epidemiol. 46 (5), 1410–1420. https://doi.org/
625.:10.1037/0003-066X.59.7.614. 10.1093/ije/dyx074.
Dashti, H.S., Scheer, F.A., Jacques, P.F., Lamon-Fava, S., Ordovás, J.M., 2015. Short Malik, V.S., Popkin, B.M., Bray, G.A., Despres, J.P., Willett, W.C., Hu, F.B., 2010. Sugar-
Sleep Duration and Dietary Intake: Epidemiologic Evidence, Mechanisms, and sweetened beverages and risk of metabolic syndrome and type 2 diabetes: a meta-
Health Implications. Adv Nutr. 6 (6), 648–659. https://doi.org/10.3945/ analysis. Diabetes Care 33 (11), 2477–2481. https://doi.org/10.2337/dc10-1079.
an.115.008623. Marmot, M., 2005. Social determinants of health inequalities. Lancet. 365 (9464),
Dennis, E.A., Dengo, A.L., Comber, D.L., et al., 2010. Water consumption increases 1099–1104.
weight loss during a hypocaloric diet intervention in middle-aged and older adults. Mitchell, C., Beals, J., 2011. Psychol. Assess. 23 (3), 752–761. https://doi.org/10.1037/
Obesity 18 (2), 300–307. https://doi.org/10.1038/oby.2009.235. a0023288.
Dubois, L., Farmer, A., Girard, M., Peterson, K., 2007. Regular sugar-sweetened beverage Nair, U.S., Collins, B.N., Patterson, F., Rodriguez, D., 2015. Promoting pre-quit physical
consumption between meals increases risk of overweight among preschool-aged activity to reduce cue reactivity among low-income sedentary smokers: a
children. J. Am. Diet Assoc. 107 (6), 924–934. https://doi.org/10.1016/J. randomized proof of concept study. Contem. Clin. Trials 42, 158–166.
JADA.2007.03.004. Office of the Assistant Secretary for Planning and Evaluation. 2018 Poverty Guidelines.
Fleig, L., Kerschreiter, R., Schwarzer, R., Pomp, S., Lippke, S., 2014. ‘Sticking to a healthy Published January 2018. Accessed September 2021. https://aspe.hhs.gov/topics/p
diet is easier for me when I exercise regularly’: cognitive transfer between physical overty-economic-mobility/poverty-guidelines/prior-hhs-poverty-guidelines-feder
exercise and healthy nutrition. Psychol. Health 29 (12), 1361–1372. https://doi.org/ al-register-references/2018-poverty-guidelines.
10.1080/08870446.2014.930146. Panel on Dietary Reference Intakes for Electrolytes and Water, Standing Committee on
Fleig, L., Küper, C., Lippke, S., Schwarzer, R., Wiedemann, A.U., 2015. Cross-behavior the Scientific Evaluation of Dietary Reference Intakes, Board F and N. Dietary
associations and multiple health behavior change: a longitudinal study on physical Reference Intakes for Water, Potassium, Sodium, Chloride, and Sulfate. National
activity and fruit and vegetable intake. J. Health Psychol. 20 (5), 525–534. Academies Press; 2005.:10.17226/10925.
Ford, E.S., Bergmann, M.M., Kroger, J., Schienkiewitz, A., Weikert, C., Boeing, H., 2009. Priebe, C.S., Atkinson, J., Faulkner, G., 2017. Run to Quit: an evaluation of a scalable
Healthy living is the best revenge. Arch. Int. Med. 169 (15), 1355. https://doi.org/ physical activity-based smoking cessation intervention. Ment. Health Phys. Act. 13,
10.1001/archinternmed.2009.237. 15–21.
Goldstein, A.N., Walker, M.P., 2014. The role of sleep in emotional brain function. Annu. Prochaska, J.J., Sung, H.-Y., Max, W., Shi, Y., Ong, M., 2012. Validity study of the K6
Rev. Clin. Psychol. 10 (1), 679–708. scale as a measure of moderate mental distress based on mental health treatment
Gopinath, B., Flood, V.M., Burlutsky, G., Mitchell, P., 2010. Combined influence of need and utilization. Int. J. Methods Psychiatr Res. 21 (2), 88–97.
health behaviors on total and cause-specific mortality. Arch Intern Med. 170 (17), Sartorius, N., 2007. Stigma and mental health. Lancet (London, England) 370 (9590),
1605–1607. https://doi.org/10.1001/archinternmed.2010.303. 810–811. https://doi.org/10.1016/S0140-6736(07)61245–8w.
Greer, S.M., Goldstein, A.N., Walker, M.P., 2013. The impact of sleep deprivation on food California Health Interview Survey. CHIS Design and Methods. Published 2019. Accessed
desire in the human brain. Nat. Commun. 4 (1). September 25, 2019. https://healthpolicy.ucla.edu/chis/design/Pages/overview.
Harold W. Kohl I, Cook HD, Environment C on PA and PE in the S, Board F and N, aspx.
Medicine I of. Physical Activity and Physical Education: Relationship to Growth, U.S Department of Health and Human Services, U.S. Department of Agriculture. 2015-
Development, and Health. Published online October 30, 2013. Accessed May 19, 2020 Dietary Guidelines for Americans, 8th Edition.; 2015.
2019. https://www.ncbi.nlm.nih.gov/books/NBK201497/. [dataset] University of California Canter for Health Policy Research. California Health
Holt, R., de Groot, M., Golden, S., 2014. Diabetes and depresion. Curr. Diabet. Rep. 14 Interview Survey (CHIS) 2017 Confidential Adult Public Use Files. California Health
(491) https://doi.org/10.1007/s11892-014-0491-3. Interview Survey. Published and accessed 2018.
Holt-Lunstad, J., Smith, T.B., Baker, M., Harris, T., Stephenson, D., 2015. Loneliness and Whitelock, E., Ensaff, H., 2018. On Your own: older adults’ food choice and dietary
social isolation as risk factors for mortality. Perspect Psychol Sci. 10 (2), 227–237. habits. Nutrients 10 (4). https://doi.org/10.3390/nu10040413.
Investigation, O., 2009. Healthy living is the best revenge. Arch. Int. Med. 169 (15), World Health Organization. WHO | About social determinants of health. WHO. Published
1355. https://doi.org/10.1001/archinternmed.2009.237. 2017. Accessed November 11, 2019. https://www.who.int/social_determinants/sdh
Jahns L. Water: An Important Part of a Healthy Winter Diet : USDA ARS. U.S. Department _definition/en/.
of Agriculture - Agriculture Research Service. Published 2019. Accessed November World Health Organization. Physical Activity Fact Sheet. Published November 2020.
11, 2019. https://www.ars.usda.gov/plains-area/gfnd/gfhnrc/docs/news-2012/ Accessed November 2020. https://www.who.int/news-room/fact-sheets/detail/ph
water-an-important-part-of-a-healthy-winter-diet/. ysical-activity.
Kvaavik, E., Batty, G.D., Ursin, G., Huxley, R., Gale, C.R., 2010. Influence of individual
and combined health behaviors on total and cause-specific mortality in men and