perceived risk of developing diabetes in early adulthood: beliefs about inherited and behavioral...

12
Journal of Health Psychology 17(2) 285–296 © The Author(s) 2011 Reprints and permission: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1359105311412840 hpq.sagepub.com Introduction Diabetes is a major health concern in the United States and throughout the world (Bruno and Landi, 2011). In the United States, diabe- tes is the seventh leading cause of death con- tributing to over 230,000 deaths each year (Molinaro, 2011). It is therefore concerning that diabetes incidence is projected to grow from eight cases per 100,000 in 2008 to 15 cases per 100,000 by 2050 (Boyle et al., 2010). Perceived Risk of Developing Diabetes in Early Adulthood: Beliefs about Inherited and Behavioral Risk Factors across the Life Course Justin B. Dickerson 1 , Matthew Lee Smith 1 , Erica Sosa 2 , E. Lisako McKyer 3 and Marcia G. Ory 1 Abstract This study aimed to determine how college students perceive their risk of developing diabetes over their life course, with specific emphasis on their beliefs about the influence of inherited versus behavioral risk factors. A bivariate ordered probit regression model was used to simultaneously predict perceived risk for 10-year absolute risk of diabetes and lifetime absolute risk of diabetes. Ten-year and lifetime absolute risk were both increased when the respondent self-identified with a race/ethnicity other than non-Hispanic white (β = 0.42, p < .001 and β = 0.33, p = .004, respectively), and when the respondent had an increasing number of family members with diabetes (β = 0.33, p < .001 and β = 0.45, p < .001, respectively). Beliefs linking behavioral risk factors to perceived risk of developing diabetes across the life course were not statistically significant. The absence of significant association between perceptions of behavioral risk as factors for developing diabetes and perceived risk for diabetes over the life course supports the need for educational interventions about behavioral and genetic causes of diabetes among the college-aged population. Keywords diabetes, late adolescence, risk factors, risk perception 1 Texas A&M Health Science Center, USA 2 University of Texas at San Antonio, USA 3 Texas A&M University, USA Corresponding author: Justin B. Dickerson, PhD Student, Health Services Research, Department of Health Policy & Management, Texas A&M Health Science Center, School of Rural Public Health, 1266 TAMU, College Station, Texas 77843, USA. Email: [email protected] 412840HPQ XX X 10.1177/1359105311412840Dickerson et al.Journal of Health Psychology Article at Texas A&M University - Medical Sciences Library on March 26, 2015 hpq.sagepub.com Downloaded from

Upload: independent

Post on 22-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Journal of Health Psychology17(2) 285 –296© The Author(s) 2011 Reprints and permission: sagepub.co.uk/journalsPermissions.navDOI: 10.1177/1359105311412840hpq.sagepub.com

Introduction

Diabetes is a major health concern in the United States and throughout the world (Bruno and Landi, 2011). In the United States, diabe-tes is the seventh leading cause of death con-tributing to over 230,000 deaths each year (Molinaro, 2011). It is therefore concerning that diabetes incidence is projected to grow from eight cases per 100,000 in 2008 to 15 cases per 100,000 by 2050 (Boyle et al., 2010).

Perceived Risk of Developing Diabetes in Early Adulthood: Beliefs about Inherited and Behavioral Risk Factors across the Life Course

Justin B. Dickerson1, Matthew Lee Smith1, Erica Sosa2, E. Lisako McKyer3 and Marcia G. Ory1

AbstractThis study aimed to determine how college students perceive their risk of developing diabetes over their life course, with specific emphasis on their beliefs about the influence of inherited versus behavioral risk factors. A bivariate ordered probit regression model was used to simultaneously predict perceived risk for 10-year absolute risk of diabetes and lifetime absolute risk of diabetes. Ten-year and lifetime absolute risk were both increased when the respondent self-identified with a race/ethnicity other than non-Hispanic white (β = 0.42, p < .001 and β = 0.33, p = .004, respectively), and when the respondent had an increasing number of family members with diabetes (β = 0.33, p < .001 and β = 0.45, p < .001, respectively). Beliefs linking behavioral risk factors to perceived risk of developing diabetes across the life course were not statistically significant. The absence of significant association between perceptions of behavioral risk as factors for developing diabetes and perceived risk for diabetes over the life course supports the need for educational interventions about behavioral and genetic causes of diabetes among the college-aged population.

Keywordsdiabetes, late adolescence, risk factors, risk perception

1Texas A&M Health Science Center, USA2University of Texas at San Antonio, USA3Texas A&M University, USA

Corresponding author:Justin B. Dickerson, PhD Student, Health Services Research, Department of Health Policy & Management, Texas A&M Health Science Center, School of Rural Public Health, 1266 TAMU, College Station, Texas 77843, USA. Email: [email protected]

412840 HPQXXX10.1177/1359105311412840Dickerson et al.Journal of Health Psychology

Article

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

286 Journal of Health Psychology 17(2)

If current incidence and mortality trends con-tinue, diabetes prevalence will reach 33 per-cent by 2050 (Boyle et al., 2010). Diabetes also has a high financial cost to health systems throughout the world (Clarke et al., 2010). In the United States, these costs are predicted to exceed $190 billion by 2020 (Molinaro, 2011).

Trends in the rate of diabetes prevalence continue to exhibit racial and ethnic disparities. The age and sex adjusted prevalence rate of diagnosed diabetes among African-Americans (11.0%) and Hispanics (11.0%) is higher than among non-Hispanic whites (5.5%) (Zhang et al., 2009). In addition, the trend of earlier onset of diabetes is becoming a worldwide con-cern (Bruno and Landi, 2011). As such, dispari-ties in diabetes prevalence are likely to persist into the foreseeable future.

Risk factors for diabetes include: high blood pressure; physical inactivity; being overweight or obese; race/ethnicity; and hav-ing a family history of diabetes (Heikes et al., 2008). While family history is a strong predic-tor for developing Type 2 diabetes, this is only the case within American and European cul-tures (American Diabetes Association, 2011). In non-western cultures, family history does not share the same predictive relationship for development of Type 2 diabetes (American Diabetes Association, 2011). The likely reason for this divergence is the difference in lifestyle behaviors such as diet and physical activity habits (American Diabetes Association, 2011) which are known as controllable behavioral risk factors. Such extant research suggests that family history of diabetes among Americans has more to do with shared health-risk behav-iors than genetics.

Understanding perceptions of behavioral and genetic risk is important for guiding diabe-tes prevention among high-risk populations; however, there is much to be learned about how the high-risk early adulthood population per-ceives their risk and the factors that influence those perceptions. While there is limited research directly on this topic, related studies have provided a foundation upon which to build

research hypotheses. The work of Aoun and colleagues (2002) demonstrates the utility of the Health Belief Model (Rosenstock, 1966) in assessing perceived risk of developing diabetes and resulting cues to act among middle-aged men. Similarly, Soo and Lam (2009) point to the connection between physical and mental manifestations of diabetes by examining the role of stress management.

This article examines risk perceptions among college students for developing diabetes during the life course. Given our focus on early adult-onset diabetes, our study uses the term ‘diabetes’ to imply the clinical condition known as Type 2 diabetes mellitus, which accounts for up to 95 percent of diagnosed cases of diabetes mellitus (Molinaro, 2011). Specifically, this article examines what factors influence college students’ risk perceptions for developing diabe-tes – beliefs that diabetes is due to controllable behavioral risk factors or non-controllable risk factors, such as genetic causes of diabetes; their personal characteristics such as race/ethnicity and sex; and family history of diabetes. This article also examines these factors on a tempo-ral basis across both proximal and distal time periods. Our a priori hypothesis was that col-lege students lacked an appreciation for the behavioral causes of diabetes, and as a result, may have an optimistic bias about their risk for developing diabetes during their lifetimes. This assertion was based in large part on prior stud-ies of college students and heart disease risk perception (Collins et al., 2004; Green et al., 2003), although these studies did not attempt to explain the temporal nature of risk perception. Since the college environment represents a unique opportunity to establish healthy lifestyle behaviors (Nelson et al., 2008), this is a critical population to study.

Methods

Participants

A total of 703 college students voluntarily responded to an Internet-based survey, with the

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

Dickerson et al. 287

option to discontinue their participation at any time. Ninety-one respondents were excluded from the study because they reported being personally diagnosed with one of three chronic diseases being studied: cancer; heart disease; and diabetes. The resulting sample contained 612 respondents.

Measures

Dependent variable selection. The dependent variables in the study were respondents’ per-ceived risk of developing diabetes during the next 10 years (‘10-year absolute risk’) or at some point in their lifetime (‘lifetime absolute risk’). These risks were characterized as ‘abso-lute’ because they were not based on a relative or peer comparison. These two questions asked respondents to score their perceived risk of developing diabetes across each of the two risk periods. Each question had five ordinal response choices ranging from ‘no chance’ to ‘certain to occur’. The verbatim text of the 10-year absolute risk question was, ‘How likely are you to get the following disease/health condition within the next 10 years?’ The verbatim text of the lifetime absolute risk ques-tion was, ‘How likely are you to get the follow-ing disease/health condition at any point in your lifetime?’ Based on the work of Hevey and colleagues (2009), our asking relative risk questions prior to the two questions stated above may indicate our approach was best reflective of a type of indirect elicitation, which may minimize unrealistic optimism of respond-ents in self-rating their susceptibility for devel-oping diabetes (Hevey et al., 2009). Further, these questions were located on page nine of the 13-page instrument. Ideally, Lister and col-leagues (2002) suggest such questions be located closer to the front of the instrument to minimize the chance anxiety about health risk questions would influence responses.

Independent variable selection. Nine variables were selected as possible predictors of individ-ual risk assessment for developing diabetes.

Sex of the respondent was determined to be important based on previous findings relating to sex-based differences in chronic disease risk perception (Hart, 2005). Considering the known relationships between diabetes and racial/ethnic minorities, a variable was created to indicate which respondents self-identified with a race/ethnicity other than non-Hispanic white. Based on prior studies examining the link between risk perception and family health history (Hart, 2005; Montgomery et al., 2003), the number of immediate family members (i.e. biological and non-biological parents, grand-parents, siblings, aunts, uncles, cousins, and children) with diabetes were aggregated into a composite variable, ‘number of family mem-bers with diabetes’. Similarly, the number of non-family members (i.e. friends, friends’ fam-ily members, classmates, and co-workers) with diabetes was aggregated into a composite vari-able, ‘number of non-family members with diabetes’.

In developing the research question, we were intrigued by the beliefs and communica-tion about family health history between family members as studied by Hastrup and colleagues (1985). To build upon this literature, we wanted to examine the role parents play in their chil-dren’s discussions of family health history, rela-tive to their children’s risk perceptions for developing chronic diseases. As such, a varia-ble describing these relationships was included in our analysis, ‘have discussed family health history with parents’.

Finally, we believed important differences in perceived risk of developing diabetes may exist based on respondents’ perceptions about the causes of diabetes (Sousa et al., 2010). To examine these perceptions, we incorporated four variables which asked the respondent to score how likely their chance was of devel-oping diabetes due to: ‘behavioral causes’; ‘social causes’; ‘genetic causes’; and ‘envi-ronmental causes’. Respondents were pro-vided with 10 ordinal response choices ranging from ‘not at all likely’ to ‘absolutely likely’.

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

288 Journal of Health Psychology 17(2)

Analytical approach – bivariate ordered probit regression

Our study involved an examination of two dependent variables that were significantly related to one another (i.e. Rho = 0.81, X2 = 346.78, p < .001). Estimating two separate ordered categorical models (presumably two ordered probit regression models) would have resulted in biased estimators as discovered by researchers using similar analytic methods (Oshio and Kobayashi, 2010). To account for this potential bias, we utilized a bivariate ordered probit regression model to jointly esti-mate both outcomes simultaneously (i.e. ‘10-year absolute risk’ and ‘lifetime absolute risk’). This approach has been used in several similar situations (de Lapparent, 2008; Weiss, 1993; Yamamoto and Shankar, 2004) where two ordi-nal response variables were correlated and required joint estimation (see Kim (1995) who utilized this method when studying diabetic neuropathy and found it unreasonable to assume two eyes in the same pair behaved independ-ently). Our specific utilization of bivariate ordered probit regression was accomplished through the ‘bioprobit’ command in Stata Release 11 (StataCorpLP, 2011) developed by Sajaia (2006), and based on maximum likeli-hood estimation.

Analytical approach – bivariate ordered probit regression – marginal probability

In addition to reporting coefficients, robust standard errors, and p-values; we report the results of a detailed sensitivity analysis con-ducted on all variables and response choices called ‘marginal probability’ analysis. This met-ric was calculated using techniques described by Sajaia (2006). Marginal probability repre-sents the anticipated change in the dependent variable when a single independent variable takes a different value, while all other variables remain constant. To determine the full effect of the model results, we calculated the marginal

impact of a change in value of each predictor on the response choice for each dependent variable. For example, if the respondent self-identified as being part of a race/ethnicity other than non-Hispanic white, and all other varia-bles remained constant, the marginal change in the percentage of respondents predicted to answer ‘not at all likely’ when asked to assess their perceived 10-year absolute risk percep-tion of developing diabetes would be –35.0 percent (see Table 1). Given a baseline for this response choice of 27.5 percent, this means the model would expect the new predicted percent-age of respondents in this category (‘not at all likely’) to be 17.9 percent (27.5% * (1–0.35)). Since changes in marginal probabilities are based on the coefficients of the bivariate ordered probit regression model predictors, statistical significance of changes in probabili-ties are evaluated by examining the model coefficients and overall model adequacy (Sajaia, 2006).

Calculating marginal probabilities was thought to be important to facilitate interpreta-tion of the data for researchers and practitioners alike. Reporting statistics yielded from the mar-ginal probability analysis complements the raw coefficients from the bivariate ordered probit regression, which does not allow for the same ease of interpretation as other statistical meth-ods such as ordinary least squares regression.

Procedures

Focus groups consisting of college students were used to create the ‘Finding Roots: Exploring Your Family History’ study (Smith et al., 2011). The purpose of the study was to assess the knowledge, beliefs, and perceptions of family history and chronic diseases. In addi-tion to an initial set of focus groups, one round of follow-up meetings with the same focus group participants was held to review the resulting questionnaire. The final instrument contained 60 items organized according to the constructs of the Health Belief Model (Rosenstock, 1966) (i.e. perceived susceptibility,

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

Dickerson et al. 289

severity, barriers, and benefits of health behav-iors). The survey items contained Likert-type scales, checklists, and closed-end response for-mats. Likert-type questions utilized a bipolar scale of response choices. Likert-type questions were scored on an ordinal basis with no weight-ing. Other question types generally yielded

indicator variables, and as such, were scored as zero or one. No weighting was used for scoring any item in the instrument.

The questionnaire was then converted to a web-based survey instrument and field tested (n = 77). The web-link to the survey instrument was then provided through electronic mail to

Table 1. Descriptive Statistics of Sample (n = 612).

Male (N = 227)

Female (N = 385)

Total (N = 612)

X2 p

Age 5.242 0.630 18 23 (10.1%) 37 (9.6%) 60 (9.8%) 19 52 (22.9%) 70 (18.2%) 122 (19.9%) 20 61 (26.9%) 125 (32.5%) 186 (30.4%) 21-24 85 (37.4%) 138 (35.8%) 223 (36.4%) 25+ 6 (2.6%) 15 (3.9%) 21 (3.4%) Member of a Race/Ethnicity other than Non-Hispanic White

4.689 0.030

Yes 62 (27.3%) 76 (19.7%) 138 (22.5%) No 165 (72.7%) 309 (80.3%) 474 (77.5%) Have Discussed Family Health History with Parents

15.835 0.001

Yes 193 (85.0%) 364 (94.5%) 557 (91.0%) No 34 (15.0%) 21 (5.5%) 55 (9.0%) Number of Family Members with Diabetes

9.027 0.108

0 123 (54.2%) 174 (45.2%) 297 (48.5%) 1 62 (27.3%) 146 (37.9%) 208 (34.0%) 1+ 42 (18.5%) 65 (16.9%) 107 (17.5%) Number of Non-Family Members with Diabetes

4.353 0.500

0 149 (65.6%) 229 (59.5%) 378 (61.8%) 1 39 (17.2%) 86 (22.3%) 125 (20.4%) 1+ 39 (17.2%) 70 (18.2%) 109 (17.8%)

Male (N = 227)

Female (N = 385)

Total (N = 612)

t p

Perceived Cause of Diabetes Susceptibility (1 = “Not at All Likely”, 10 = “Absolutely Likely”) Behavioral 7.51 (± 2.872) 8.01 (± 2.643) 7.82 (± 2.739) −2.193 0.029 Social 3.58 (± 2.890) 3.84 (± 3.094) 3.74 (± 3.020) −1.008 0.314 Genetic 7.34 (± 2.706) 7.59 (± 2.475) 7.50 (± 2.564) −1.167 0.244 Environmental 3.82 (± 2.920) 4.08 (± 3.045) 3.98 (± 3.000) −1.020 0.308

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

290 Journal of Health Psychology 17(2)

academic advisers and department heads across the university to distribute to their students at their discretion. This process was used because it best enabled the researchers to deliver the survey to as many college students as possible. The survey took respondents approximately 15 minutes to complete. This study was approved by the Texas A&M University Institutional Review Board.

Results

Sample characteristics

Table 1 reports the descriptive statistics of the sample. The majority of the sample was under the age of 21 (60.1%) and female (62.9%). Nearly a quarter of the sample self-identified with a race/ethnicity other than non-Hispanic white (22.5%).

The predictors exhibited a wide degree of variability. More than half the sample identi-fied at least one family member with diabe-tes (51.5%). A higher percentage of female respondents (54.8%) than male respondents (45.8%) identified at least one family member with diabetes. Approximately 38 percent of respondents also identified a non-family mem-ber with diabetes. The majority of respondents (91.0%) had discussed family health history with their parents. More females reported hav-ing such discussion (94.5%) than males (85.0%). This difference was statistically sig-nificant (X2 = 15.83, p < .001).

Using a one (‘not at all likely’) to 10 (‘abso-lutely likely’) scale of susceptibility to diabetes, respondents attributed their highest risk of developing diabetes to behavioral causes (7.82 ± 2.73) and genetic causes (7.50 ± 2.56). Males attributed a lower perceived risk of developing diabetes due to behavioral causes (7.51 ± 2.87) than females (8.01 ± 2.64). This difference was statistically significant (t = –2.19, p = .029). Similar to the findings of Park and colleagues (2010), we do not believe the way the suscepti-bility questions were framed (i.e. focused on the negative aspects of disease susceptibility versus

a positive focus on the benefits of disease avoid-ance) impacted the results.

Perceived 10-year absolute risk of developing diabetes

Table 2 reports the results of the bivariate ordered probit regression model where the dependent variables of perceived 10-year and lifetime absolute risk of diabetes were jointly estimated. Respondents who self-identified as being part of a race/ethnicity other than non-Hispanic white (β = 0.42, p < .001) and reported increased numbers of family members with dia-betes (β = 0.33, p < .001) were more likely to report higher perceived levels of absolute 10-year risk of developing diabetes relative to their counterparts.

Perceived Lifetime Absolute Risk of Developing Diabetes

Respondents who self-identified as being part of a race/ethnicity other than non-Hispanic white (β = 0.33, p = .004) and reported increased numbers of family members with diabetes (β = 0.45, p < .001) were more likely to report higher perceived levels of lifetime absolute risk of developing diabetes relative to their counterparts.

Sensitivity analysis

Table 3 illustrates the sensitivity analysis of the bivariate ordered probit regression model. To determine the most relevant predictors when determining changes in 10-year and lifetime absolute risk perception, it is impor-tant to (1) consider the statistical significance of the predictor, and (2) examine the magni-tude of the change to the original predicted values. Using this criterion, we identified the most significant predictor; whether a respond-ent self-identified with a race/ethnicity other than non-Hispanic white.

Respondents self-identifying with a race/ethnicity other than non-Hispanic white

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

Dickerson et al. 291

experienced a higher level of perceived risk for developing diabetes over their life course. In the 10-year risk period, sensitivity analysis indicated only 17.9 percent of these respond-ents were predicted to fall into the first response choice category (‘not likely at all’) versus 27.5 percent when race/ethnicity was not a factor. Conversely, 2.5 percent of those self-identifying with races/ethnicities other than non-Hispanic white were predicted to fall into the fifth response choice category (‘abso-lutely likely’) versus 1.4 percent when race/ethnicity was not a factor. Results were similar in the lifetime risk period. Only 9.3 percent of those self-identifying with a race/ethnicity other than non-Hispanic white were predicted to fall into the first response choice (‘not likely at all’) versus 14.3 percent when race/ethnicity was not a factor. Conversely, 7.3 percent of those self-identifying with a race/ethnicity other than non-Hispanic white were predicted

to fall into the fifth response choice category (‘absolutely likely’) versus 5.0 percent when race/ethnicity was not a factor. While the vari-able measuring the number of family members with diabetes was statistically significant in both risk periods, Table 1 illustrates how its impact on predicted risk perception was negligible.

Discussion

Perceived risk of developing diabetes accurately reflects race/ethnicity trends

Our results indicate respondents identifying with a race/ethnicity other than non-Hispanic white were more likely to perceive their risk of developing diabetes in both the 10-year and lifetime absolute risk periods as greater than their non-Hispanic white counterparts.

Table 2. Bivariate Ordered Probit Regression Model.

Ten-Year Risk Lifetime Risk

Coefficient Robust Standard Error

P-Value Coefficient Robust Standard Error

P-Value

Dependent Variable CorrelationWald Test of Rho = Chi2(1) = 346.78, p = 0.000 Rho = 0.81Independent VariablesSex 0.019 0.095 0.839 −0.042 0.091 0.647Member of a Race/Ethnicity other than Non-Hispanic White

0.422 0.116 0.001 0.334 0.115 0.004

Have Discussed Family Health History with Parents

−0.311 0.182 0.087 −0.211 0.177 0.233

Number of Family Members with Diabetes

0.330 0.050 0.001 0.450 0.052 0.001

Number of Non-Family Members with Diabetes

0.034 0.046 0.459 0.038 0.046 0.418

Perceived Cause of Diabetes Susceptibility - Behavioral

0.037 0.027 0.174 0.015 0.023 0.504

Perceived Cause of Diabetes Susceptibility - Social

0.007 0.019 0.721 0.024 0.019 0.204

Perceived Cause of Diabetes Susceptibility - Genetics

0.019 0.028 0.482 0.047 0.025 0.064

Perceived Cause of Diabetes Susceptibility - Environmental

0.024 0.020 0.233 −0.006 0.020 0.743

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

292 Journal of Health Psychology 17(2)

Tabl

e 3.

Biv

aria

te O

rder

ed P

robi

t R

egre

ssio

n M

odel

Sen

sitiv

ity A

naly

sis

- M

argi

nal P

roba

bilit

ies.

Per

cent

of S

ampl

e by

Sel

f-R

epo

rted

Ris

k P

erce

ptio

nM

argi

nal C

hang

e in

Ris

k Pe

rcep

tion

Giv

en a

Cha

nge

in a

Sin

gle

Inde

pend

ent V

aria

ble

Valu

e*

Base

line

Ten-

Year

Ris

k Pe

rcep

tion

Scor

esBa

selin

e Li

fetim

e R

isk

Perc

eptio

n Sc

ores

Not

at

Abs

olut

ely

All

Like

ly

L

ikel

yN

ot a

t

Abs

olut

ely

All

Like

ly

L

ikel

y

12

34

51

23

45

27.5

%43

.3%

22.1

%5.

7%1.

4%14

.3%

36.1

%29

.8%

14.8

%5.

0%

Mar

gina

l Cha

nge

in R

isk

Per

cept

ion

Sco

re B

ased

on

a C

hang

e in

……

.

Mar

gina

l Per

cent

age

Cha

nge

in T

en-Y

ear

Ris

k Pe

rcep

tion

Scor

esM

argi

nal P

erce

ntag

e C

hang

e in

Life

time

Ris

k

Perc

eptio

n Sc

ores

Not

at

Abs

olut

ely

All

Like

ly

L

ikel

yN

ot a

t

Abs

olut

ely

All

Like

ly

L

ikel

y

12

34

51

23

45

Sex

M

ale

1.4%

0.0%

−1.

1%−

1.9%

−2.

6%3.

9%1.

1%−

1.0%

−2.

7%−

4.4%

Fe

mal

e−

0.8%

0.0%

0.6%

1.1%

1.7%

2.3%

0.6%

−0.

6%−

1.6%

−2.

7%M

embe

r of

a R

ace/

Ethn

icity

oth

er t

han

Non

-His

pani

c W

hite

Ye

s−

35.0

%−

2.9%

29.6

%57

.9%

78.3

%−

35.0

%−

12.3

%9.

4%29

.1%

46.9

%

No

9.3%

1.5%

−7.

8%−

18.7

%−

32.5

%8.

7%3.

7%−

1.7%

−8.

0%−

18.2

%H

ave

Dis

cuss

ed F

amily

Hea

lth H

isto

ry w

ith P

aren

ts

Yes

3.0%

0.2%

−2.

4%−

4.8%

−7.

1%2.

7%0.

8%−

0.7%

−2.

1%−

3.3%

N

o−

29.3

%−

3.0%

24.2

%51

.4%

81.8

%−

25.4

%−

8.9%

6.3%

20.8

%38

.2%

(Con

tinue

d)

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

Dickerson et al. 293

Mar

gina

l Cha

nge

in R

isk

Per

cept

ion

Sco

re B

ased

on

a C

hang

e in

……

.

Mar

gina

l Per

cent

age

Cha

nge

in T

en-Y

ear

Ris

k Pe

rcep

tion

Scor

esM

argi

nal P

erce

ntag

e C

hang

e in

Life

time

Ris

k

Perc

eptio

n Sc

ores

Not

at

A

bsol

utel

yA

ll Li

kely

Lik

ely

Not

at

A

bsol

utel

yA

ll Li

kely

Lik

ely

12

34

51

23

45

Num

ber

of F

amily

Mem

bers

with

Dia

bete

s

+10

% C

hang

e in

Num

ber

of F

amily

M

embe

rs−

2.0%

−0.

8%1.

7%5.

8%12

.8%

−2.

5%−

1.9%

−0.

2%3.

4%12

.2%

Num

ber

of N

on-F

amily

Mem

bers

with

Dia

bete

s

+10

% C

hang

e in

Num

ber

of F

amily

M

embe

rs−

0.2%

0.0%

0.2%

0.4%

0.6%

−0.

3%−

0.1%

0.1%

0.3%

0.5%

Perc

eive

d C

ause

of D

iabe

tes

Susc

eptib

ility

- B

ehav

iora

l

+10

% C

hang

e in

Sco

re−

3.1%

−0.

2%2.

5%4.

9%7.

3%−

1.6%

−0.

5%0.

4%1.

3%2.

2%Pe

rcei

ved

Cau

se o

f Dia

bete

s Su

scep

tibili

ty -

Soc

ial

+

10%

Cha

nge

in S

core

−0.

3%0.

0%0.

2%0.

5%0.

7%−

1.2%

−0.

4%0.

3%1.

0%1.

9%Pe

rcei

ved

Cau

se o

f Dia

bete

s Su

scep

tibili

ty -

Gen

etic

s

+10

% C

hang

e in

Sco

re−

1.6%

−0.

1%1.

3%2.

5%3.

7%−

4.7%

−1.

7%1.

2%3.

9%7.

0%Pe

rcei

ved

Cau

se o

f Dia

bete

s Su

scep

tibili

ty -

Env

iron

men

tal

+

10%

Cha

nge

in S

core

−1.

0%−

0.1%

0.8%

1.7%

2.7%

0.3%

0.1%

−0.

1%−

0.3%

−0.

5%

*Cal

cula

tions

are

mut

ually

exc

lusi

ve. A

ssum

es v

alue

s fo

r ot

her

inde

pend

ent

vari

able

s re

mai

n th

e sa

me.

Not

e -

Mar

gina

l cha

nges

do

not

sum

to

1 be

caus

e of

the

nat

ure

of s

imul

tane

ous

mod

el s

peci

ficat

ion

with

the

var

iabl

es.

Tabl

e 3.

(C

ontin

ued)

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

294 Journal of Health Psychology 17(2)

By examining the marginal probabilities in Table 3 for perceived 10-year risk, it can be determined that approximately 60 percent of those identifying with a race/ethnicity other than non-Hispanic white were predicted to fall into the two lowest risk response catego-ries. Conversely, 74 percent of non-Hispanic white respondents were predicted to fall into the two lowest risk response categories. Despite the literature critiquing the ability of college students to accurately perceive their risk for chronic diseases such as heart disease (Collins et al., 2004; Green et al., 2003), the perceptions of respondents in our study reflected congruence with prevalence data along race/ethnicity lines (Zhang et al., 2009). Although our outcome variables were differ-ent than Ho and colleagues (2007) who stud-ied perceived health status among chronic disease and diabetes patients as young as age 25, our research was not able to establish the same relevance of age and sex to our research questions. This could be due to the geo-graphic differences in the studies (i.e. our study was conducted in the United States while the study performed by Ho and col-leagues was conducted in Hong Kong).

College students’ perceived risk of diabetes is driven by non-controllable factors

The most alarming finding in this study is the lack of a statistically significant relationship between perceived risk of developing diabetes and the contribution of behavioral risk factors. While college students report the importance of behavioral risk factors in developing diabe-tes (see Table 1), that belief was not associated with their own personal risk assessment for developing diabetes. This finding is no differ-ent than other recent studies (Copeland et al., 2009; Schwartz et al., 2010) which have tried to identify the sources of unfounded optimism of college students with respect to the conse-quences of their often high-risk health behaviors.

In addition, this finding is consistent with much of the established social psychology and medical literature regarding optimistic biases and perceived health risks (Allen and Blumenthal, 1998; Azzarello, 2007; Weinstein, 1984, 1987) In this study, not only were behav-ioral causes of diabetes susceptibility not sta-tistically significant in predicting levels of perceived risk of developing diabetes, but instead, the two factors that were statistically significant (i.e. identifying with a race/ethnic-ity other than non-Hispanic white, and having an increased number of family member with diabetes) were both considered to be non-con-trollable factors. This is worth noting because not only does it suggest college students have minimized their perceived diabetes risk as attributed to behavioral factors, it also sug-gests they base their risk assessment primarily on non-controllable risk factors, lessening their likelihood of engaging in preventive behaviors. As discussed in this study, such a view is not only without support in the clinical data, it is also a view with substantial societal implications in terms of economics, morbidity, and ultimately mortality.

Study limitations

This study is unique in its examination of how risk perceptions are influenced across different points of time in the life course. However, one notable shortcoming is the study does not link perceptions to health status. This is an opportu-nity for a future study. Second, while the bivari-ate ordered probit regression model is an appropriate and unique tool for this data set, it does not report a metric such as the coefficient of determination to allow us to see how much of the variation in risk perceptions is explained with the independent variables. Next, the meth-ods for administering the survey instrument did not allow for a calculation of response rate. Finally, the study was focused on a single col-lege campus which may or may not be repre-sentative of the wider college population.

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

Dickerson et al. 295

Conclusion

As researchers, administrators, and policymak-ers continue to examine ways to improve the health of adolescent and young adult popula-tions, it is important to study cognitive percep-tions alongside clinical factors. This study is among a handful of efforts to illustrate how a population of college students perceived their risk for developing a specific disease along a racial/ethnic dimension. This finding is differ-ent than the prevailing view that college stu-dents rarely understand their personal health risks. Future research is needed to better under-stand why differences exist in the accuracy of the perceived risk of developing diabetes, and how such differences can be overcome. The work of Eiser and colleagues (2002) reminds us how risk assessment and changes based on behavioral motivations are multifaceted, not only in terms of clinical issues such as the dif-ferentiation between Type 1 and Type 2 diabe-tes patients, but also in terms of the complex interactions that occur between patients and cli-nicians. Issues such as these should be further studied among early adult populations.

Competing InterestsNone declared.

ReferencesAllen JK and Blumenthal RS (1998) Risk factors in

the offspring of women with premature coronary heart disease. American Heart Journal 135(3): 428–434.

American Diabetes Association (2011) Genetics of Diabetes – American Diabetes Association. Available at: http://www.diabetes.org/diabetes-basics/genetics-of-diabetes.html

Aoun S, Donovan R, Johnson L and Egger G (2002) Preventive care in the context of men’s health. Journal of Health Psychology 7(3): 243–252.

Azzarello J (2007) Perceived versus actual risk for diabetes: Unrealistic optimism or needless pes-simism? Diabetes 56: A496–A496.

Boyle JP, Thompson TJ, Gregg EW, Barker LE and Williamson DF (2010) Projection of the year 2050 burden of diabetes in the US adult popula-tion: Dynamic modeling of incidence, mortality,

and pre-diabetes prevalence. Population Health Metrics 8: 29–40.

Bruno G and Landi A (2011) Epidemiology and costs of diabetes. Transplantation Proceedings 43(1): 327–329.

Clarke PM, Glasziou P and Patel A, et al. (2010) Event rates, hospital utilization, and costs asso-ciated with major complications of diabetes: A multi-country comparative analysis. PLoS Medi-cine 7(2): 1–10.

Collins KM, Dantico M, Shearer N and Mossman KL (2004) Heart disease awareness among col-lege students. Journal of Community Health 29(5): 405–420.

Copeland AL, Kulesza M, Patterson SM and Ter-lecki MA (2009) College student smokers’ cog-nitive appraisal of high-risk activities. Journal of American College Health 58(3): 203–212.

De Lapparent M (2008) Willingness to use safety belt and levels of injury in car accidents. Acci-dent Analysis and Prevention 40(3): 1023–1032.

Eiser JR, Eiser C, Riazi A, Hammersley S and Tooke JE (2002) Interpretations of risk and expectations of change among individuals with types 1 and 2 diabetes. Journal of Health Psychology 7(6): 701–711.

Green JS, Grant M, Hill KL, Brizzolara J and Bel-mont B (2003) Heart disease risk perception in college men and women. Journal of American College Health 51(5): 207–211.

Hart PL (2005) Women’s perceptions of coronary heart disease: An integrative review. Journal of Cardiovascular Nursing 20(3): 170–176.

Hastrup JL, Hotchkiss AP and Johnson CA (1985) Accuracy of knowledge of family history of car-diovascular disorders. Health Psychology 4(4): 291–306.

Heikes KE, Eddy DM, Arondekar B and Schless-inger L (2008) Diabetes risk calculator: A simple tool for detecting undiagnosed diabetes and pre-diabetes. Diabetes Care 31(5): 1040–1045.

Hevey D, French D, Marteau T and Sutton S (2009) Assessing unrealistic optimism: Impact of dif-ferent approaches to measuring susceptibility to diabetes. Journal of Health Psychology 14(3): 372–377.

Ho S, Mak K and Thomas GN, et al. (2007) The relation of chronic cardiovascular diseases and diabetes mellitus to perceived health, and the moderating effects of sex and age. Social Science Medicine 65(7): 1386–1396.

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from

296 Journal of Health Psychology 17(2)

Kim K (1995) A bivariate cumulative probit regres-sion model for ordered categorical data. Statistics in Medicine 14(12): 1341–1352.

Lister A, Rode S, Farmer A and Salkovskis P (2002) Does thinking about personal health risk increase anxiety? Journal of Health Psychology 7(4): 409–414.

Molinaro RJ (2011) Diabetes cases on the rise: Cur-rent diagnosis guidelines and research efforts for a cure. MLO: Medical Laboratory Observer 43(2): 8–15.

Montgomery GH, Erblich J, DiLorenzo T and Bovb-jerg DH (2003) Family and friends with disease: Their impact on perceived risk. Preventive Medi-cine 37(3): 242–249.

Nelson MC, Story M, Larson NI, Neumark-Sztainer D and Lytle LA (2008) Emerging adulthood and college-aged youth: An overlooked age for weight-related behavior change. Obesity 16(10): 2205–2211.

Oshio T and Kobayashi M (2010) Income inequality, perceived happiness, and self-rated health: Evi-dence from nationwide surveys in Japan. Social Science & Medicine 70(9): 1358–1366.

Park P, Simmons R, Prevost AT and Griffin S (2010) A randomized evaluation of loss and gain frames in an invitation to screening for type 2 diabetes: Effects on attendance, anxiety and self-rated health. Journal of Health Psychology 15(2): 196–204.

Rosenstock I (1966) Why people use health services. Milbank Memorial Fund Quarterly 44(3, Pt 2): 94–124.

Sajaia Z (2006) Maximum likelihood estimation of a bivariate ordered probit model: Implementation and Monte Carlo simulations. Stata Journal ii: 1–18.

Schwartz SJ, Forthun LF and Ravert RD, et al. (2010) Identity consolidation and health risk behaviors

in college students. American Journal of Health Behavior 34(2): 214–224.

Smith ML, Sosa ET, Tisone CA and McKyer ELJ (2011) Quality enhancement groups: Integrating qualitative and quantitative research methods for instrument development. Unpublished manu-script.

Soo H and Lam S (2009) Stress management training in diabetes mellitus. Journal of Health Psychol-ogy 14(7): 933–943.

Sousa V, Ryan Wenger N, Driessnack M and Jaber AF (2010) Factorial structure of the perception of risk factors for type 2 diabetes scale: Explor-atory and confirmatory factor analyses. Jour-nal of Evaluation in Clinical Practice 16(6): 1096–1102.

StataCorpLP (2011) Stata release 11. Available at: www.stata.com

Weinstein N (1984) Why it won’t happen to me: Per-ceptions of risk factors and susceptibility. Health Psychology 3(5): 431–457.

Weinstein N (1987) Unrealistic optimism about sus-ceptibility to health problems: Conclusions from a community-wide sample. Journal of Behav-ioral Medicine 10(5): 481–500.

Weiss AA (1993) A bivariate ordered probit model with truncation: Helmet use and motorcycle injuries. Journal of the Royal Statistical Society: Series C (Applied Statistics) 42(3): 487–499.

Yamamoto T and Shankar VN (2004) Bivariate ordered-response probit model of driver’s and passenger’s injury severities in collisions with fixed objects. Accident Analysis & Prevention 36(5): 869–876.

Zhang Q, Wang Y and Huang ES (2009) Changes in racial/ethnic disparities in the prevalence of type 2 diabetes by obesity level among US adults. Ethnicity & Health 14(5): 439–457.

at Texas A&M University - Medical Sciences Library on March 26, 2015hpq.sagepub.comDownloaded from