Professional Documents
Culture Documents
Measurement Article
Received: November 11, 2018; Editorial Decision Date: March 25, 2019
Abstract
Background and Objectives: Cannabis use among older adults is on the rise. Despite growing interest in the topic, there
exists a paucity of standardized measures capturing cannabis-specific attitudes among older adults. Using data from a
survey of older Coloradans, we create two scales that separately measure medical and recreational cannabis attitudes. We
also examine how these two attitudes relate to individual-level characteristics.
Research Design and Methods: We assess reliability using Cronbach’s alpha and item-rest correlations and perform
confirmatory factor analyses to test the two attitude models. We conduct a seemingly unrelated regression estimation to
assess how individual characteristics predict medical and recreational cannabis attitude scores.
Results: Twelve indicators combined into two valid and reliable scales. Both scales had a three-factor structure with affect,
cognition and social perception as latent dimensions. For both scales, fit indices for the three-factor model were statistically
superior when compared with other models. The three-factor structure for both scales was invariant across age groups. Age,
physical health, and being a caregiver differentially predicted medical and recreational cannabis attitude scores.
Discussion and Implications: Medical and recreational cannabis attitude scales can inform the development and evaluation
of tailored interventions targeting older adult attitudes that aim to influence cannabis use behaviors. These scales also
enable researchers to measure cannabis-specific attitudes among older adults more accurately and parsimoniously, which in
turn can facilitate a better understanding of the complex interplay between cannabis policy, use, and attitudes.
Keywords: Scale construction, Substance use, Factor analysis
The passage of medical and recreational marijuana legisla- Milavetz, Shane, & Arora, 2017; Stoner, 2016) has drawn
tions by several states has shifted the emphasis to examining attention to the need to understand their attitudes as well.
determinants of use, particularly the role of cannabis- Despite this growing interest, there is a surprising pau-
specific attitudes. Although a majority of this work has fo- city of standardized cannabis attitude measures to guide this
cused on adolescents and young adults (Bachman, Johnson, line of investigation. Although some studies on cannabis
& O’Malley, 1998; Keyes et al., 2011), increasing rates use among older adults have included various attitude-
of cannabis use among older adults (Kaskie, Ayyagari, related measures (Black & Joseph, 2014; Choi, Marti,
© The Author(s) 2019. Published by Oxford University Press on behalf of The Gerontological Society of America. All rights reserved. e232
For permissions, please e-mail: journals.permissions@oup.com.
The Gerontologist, 2020, Vol. 60, No. 4 e233
DiNitto, & Choi, 2018; Han et al., 2017; Salas-Wright One commonly held perspective portrays attitudes
et al., 2017), little evidence exists concerning the reliability as a superordinate construct with affective and cognitive
and validity of these items. Variability in content has also factors as underlying dimensions (Breckler & Wiggins,
resulted in a lack of clarity regarding the actual construct 1993; Eagly & Chaiken, 1993; Van der Pligt, Zeelenberg,
being measured. Moreover, prior research on the topic has van Dijk, de Vries, & Richard, 1997). The affective dimen-
largely operationalized the attitude construct from a uni- sion has traditionally been postulated to include the set of
dimensional perspective where it is not possible to simul- physiological responses, positive and negative feelings and
taneously hold both positive and negative evaluations of emotions an individual associates with an object (Breckler
the substance depending on purpose of use. Although this and Wiggins, 1993; Crites, Fabrigar, & Petty, 1994). The
may not be as relevant for adolescents and young adults, it cognitive dimension is assumed to be composed of know-
is conceivable that older adults are ambivalent: They may ledge regarding an object’s attributes or traits (Breckler &
disapprove of recreational cannabis and regard it as risky, Wiggins, 1993; Crites et al., 1994). In empirical analyses of
use of marijuana should be made legal or not?; Moeller & use of cannabis for medical purposes as legitimate, while
Woods, 2015; Nielsen, 2010). recreational cannabis use was perceived more variably,
with many associating it with societal risks and others
comparing it favorably against alcohol. To develop a con-
Methods cise set of items, a preliminary analysis of survey responses
from 80 focus group participants was conducted to help
In this article, we develop two scales capturing medical
with item-reduction and response scale adjustment. These
and recreational cannabis attitudes among older adults. In
analyses led to a pool of 22 attitude questions under
light of the foregoing discussion, we hypothesize that these
the following headings: “attitudes on medical use (7
scales will comprise three latent dimensions: affect, cogni-
questions),” “attitudes on recreational use (6 questions),”
tion, and social perception. Below, we describe the sample,
and “general attitudes about marijuana (9 questions).”
methods for scale construction and evaluation, as well as
After screening all items independently, a total of 12 items
methods for examining individual-level predictors of med-
Do not Strongly
Medical cannabis use agree agree
I currently believe that use of marijuana for a medical purpose is 5% 3% 15% 18% 59%
acceptable. (N = 265)
When I was 18, I believed that using marijuana for a medical purpose 53% 9% 21% 6% 12%
was acceptable. (N = 256)
Medical marijuana use is risky. (N = 263) 28% 23% 28% 11% 10%
Using medical marijuana leads to the use of harder drugs (N = 262) 47% 16% 16% 6% 15%
The important people in my life have positive attitudes toward using 18% 11% 20% 20% 31%
of the item-rest correlations greatly exceeded the cutoff conducted separately for each scale. For both medical
of .20, demonstrating strong reliability (Kline, 1986). and recreational cannabis attitudes, the data supported a
With the exception of the item assessing acceptability of three-factor structure (Model 1), with dimensions similar
cannabis at age 18, all item-rest correlations were above to those described earlier. In both attitude scales, the two
0.50. Specifically, the smallest item-rest correlation (.34) acceptability items loaded on the affective dimension; the
was for the item on acceptability of medical cannabis riskiness and gateway drug items loaded on the cognitive
when younger among respondents over age 80. (When we dimension; and the significant other attitudes and views on
removed the item on acceptability of cannabis at age 18, legalization loaded on the social perception dimension. The
Cronbach’s alpha improved only marginally [from .86 to standardized loadings for all indicators were above 0.60
.87 for medical cannabis attitudes and from .91 to .92 for on the corresponding dimensions. One exception was the
recreational cannabis attitudes]. Furthermore, the retention item on acceptability of cannabis at age 18. In the medical
of this item is justified on theoretical grounds as it may shed cannabis scale, this item had a standardized loading of 0.45
light on the evolution of cannabis-specific attitudes over reflecting moderate strength. In the recreational cannabis
the life course. Individuals with positive evaluations of can- scale, the standardized loading for this item was 0.56.
nabis acceptability in both young and old ages are likely to Fit statistics generally indicated a good fit of the model
have greater overall affective strength when compared with to the data for both scales (Table 3). The χ2 statistic was
those with a positive evaluation of cannabis acceptability in insignificant in both cases (medical cannabis: χ2 = 9.48,
older age only. Although this is difficult to test empirically p = .12; recreational cannabis: χ2 = 5.34, p = .50). This in-
in this article as we do not have other items to measure dicated that we cannot reject the null hypothesis that there
“affect,” we do find evidence that different groupings of is no difference between the patterns observed in these data
individuals exist. For instance, 17% [15%] of respondents and the theoretically specified model. The comparative fit
held positive evaluations of medical (recreational) cannabis indices of 0.99 in the medical cannabis attitudes case and
acceptability both currently and at age 18. However, 26% 1.00 in the recreational cannabis attitudes case were both
[8%] of respondents held a negative evaluation of medical well above the threshold of 0.90 for indication of a good
(recreational) cannabis acceptability at age 18 but, at the fit compared with the baseline model (Diamantopoulos
same time, a positive evaluation of it at current time. It is & Sigaw, 2000). The RMSEA of 0.04 in the medical can-
possible that overall affective strength varies depending on nabis attitudes scale and 0.01 in the recreational cannabis
group membership. Therefore, the item on cannabis accept- attitudes scale indicated that the model is an adequate fit to
ability at age 18 was retained.) Similar to Cronbach’s alpha the data because for both scales, the value was lower than
coefficients, differences in item-rest correlations across the typical cutoff of 0.05 (Browne & Cudeck, 1993).
subgroups were small within both scales. Table 3 also presents fit indices of two alternative factor
To verify that the items on each of the two scales analytical models. For both medical and recreational can-
tapped into three different attitude components, CFA was nabis attitude scales, χ2 values for Models 2 and 3 were
The Gerontologist, 2020, Vol. 60, No. 4 e237
Table 2. Reliability of Recreational and Medical Cannabis Attitudes Scales, by Age Group and Gender
Item-rest correlations
Table 3. Fit Indices for Confirmatory Factor Models for individuals in the 80+ age category with less than 40
individuals.
Chi-square
As displayed in Table 4, indices showed good model fit
(df; p value) RMSEA CFI
in the two age categories, indicating similar configural pat-
Medical cannabis attitudes scale tern across the two cohorts for both medical and recrea-
Model 1 (three factor) 9.48 (6; .12) 0.04 0.99 tional cannabis attitudes. We then tested for measurement
Model 2 (one factor) 50.44 (9; .00) 0.13 0.93 invariance across age groups by comparing a model with
Model 3 (two factor) 46.65 (8; .00) 0.14 0.94 factor loadings freely estimated with another, where factor
N = 237 loadings were constrained to be equal across the two age
Recreational cannabis attitudes scale categories. As given in Table 5, both models fit the data
Model 1 (three factor) 5.34 (6; .50) 0.01 1.00 quite well. Comparison between these two models showed
Model 2 (one factor) 42.86 (9; .00) 0.12 0.96 that constraints of factor loadings did not significantly
Model 3 (two factor) 42.80 (8; .00) 0.13 0.96 worsen the goodness of fit (medical scale: Δχ2 = 5.16,
N = 244 Δdf = 3, p > .05; recreational scale: Δχ2 = 4.22, Δdf = 3,
p > .05). The results suggested that the constrained model
Note: CFI = confirmatory fit index; RMSEA = root mean square error of ap-
was preferred over the freely estimated one for both scales,
proximation.
revealing cross-age measurement invariance.
Table 4. Fit Indices for the Three-Factor Model in Different potential incongruence in the two types of attitudes for
Age Groups some individuals.
Next, we conducted a SUR estimation predicting med-
Chi-Square
(df; p value) RMSEA CFI
ical and recreational cannabis scores as a function of
various individual-level characteristics. After removing
Medical cannabis attitudes scale observations with missing values on other covariates, a
Age < 72.47 8.40 (6; .21) 0.05 0.99 sample size of 192 individuals was retained. Among these,
Age > 72.47 4.03 (6; .67) 0.01 1.00 about 37% were between the ages of 60–70, 48% between
Recreational cannabis attitudes scale the ages of 71–80, and 15% over age 80. Furthermore, 64%
Age < 72.47 7.46 (6; .28) 0.04 0.99 of the sample were female and about 7% Hispanic. A ma-
Age > 72.47 2.81 (6; .83) 0.01 1.00 jority of this sample reported to be married/partnered and
retired. Approximately 13% of the sample was currently
Table 6. Seemingly Unrelated Regression Estimation of heightened risk for adverse effects such as greater levels of
Medical and Recreational Cannabis Attitude Scores injury and emergency department visits, psychiatric dis-
tress, mental illness, and comorbid substance use (Choi
Recreational
Medical cannabis cannabis attitude
et al., 2018; Han et al., 2017). Attitude scales developed
attitude score score in this article can enable researchers to identify and target
health behavior interventions to “high risk” older adults.
Age 71–80 –0.64 (0.84) –2.91** (1.03) For example, interventions educating older adults on po-
Age 80+ –2.51 (1.54) –5.66**(1.87) tential risks of drug interactions and concurrent use of
Female 0.62 (0.88) –1.34 (1.07) multiple substances may be designed specifically for those
Hispanic 0.50 (1.58) 0.09 (1.92) with strong favorable attitudes toward both recreational
Some college 3.37* (1.48) 2.42 (1.80) and medical cannabis. Similarly, other interventions fo-
College graduate 5.32*** (1.51) 5.64** (1.83) cused on health benefits of cannabis may be designed for
Graduate degree 4.92** (1.43) 5.31** (1.73)
Acknowledgments Eagly, A. H., & Chaiken, S. (1993). The psychology of attitudes.
Orlando, FL: Harcourt Brace Jovanovich College Publishers.
The authors acknowledge Emma Cole and Karrey Shannon for their
Festinger, L. (1954). A theory of social comparison processes. Human
assistance with project management. K. Arora planned the study,
Relations, 7, 117–140. doi:10.1177/001872675400700202
helped in survey instrument development, conducted the data anal-
Greene, W. H. (2003) Econometric analysis. Delhi, India: Pearson
ysis, and wrote the paper. S. H. Qualls, J. Bobbit, H. D. Lum, G. M.
Milavetz helped with survey instrument development and imple- Education India.
mentation and provided feedback on the manuscript. J. Croker Han, B. H., Sherman, S., Mauro, P. M., Martins, S. S., Rotenberg, J.,
assisted with data analysis. B. Kaskie conceptualized the survey, & Palamar, J. J. (2017). Demographic trends among older can-
helped in study planning and survey instrument development and nabis users in the United States, 2006–13. Addiction, 112, 516–
implementation, and provided feedback on study results and the 525. doi:10.1111/add.13670
manuscript. Hansen, W. B., & Hansen, J. L. (2016). Using attitudes, age
and gender to estimate an adolescent’s substance use risk.
Journal of Children’s Services, 11, 244–260. doi:10.1108/
and older adults in the United States, 2002–2014. Drug more specific. European Review of Social Psychology, 8, 33–66.
and Alcohol Dependence, 171, 97–106. doi:10.1016/j. doi:10.1080/14792779643000074
drugalcdep.2016.11.031 Wang, M., & Chen, Y. (2006). Age differences in attitude change:
Simons, J., & Carey, K. B. (1998). A structural analysis of attitudes to- Influences of cognitive resources and motivation on responses
ward alcohol and marijuana use. Personality and Social Psychology to argument quantity. Psychology and Aging, 21, 581–589.
Bulletin, 24, 727–735. doi:10.1177/0146167298247005 doi:10.1037/0882-7974.21.3.581
Stacy, A. W., Bentler, P. M., & Flay, B. R. (1994). Attitudes and Wu, L. T., Swartz, M. S., Brady, K. T., & Hoyle, R. H.; NIDA AAPI
health behavior in diverse populations: Drunk driving. Alcohol Workgroup. (2015). Perceived cannabis use norms and can-
use, binge eating, marijuana use, and cigarette use. Health nabis use among adolescents in the United States. Journal of
Psychology, 13, 73–85. doi:10.1037/0278-6133.13.1.73 Psychiatric Research, 64, 79–87. doi:10.1016/j.jpsychires.2015.
Stoner, S. A. (2016) Learn about marijuana: Factsheets: Marijuana 02.022
use by older adults. Retrieved October 22, 2018, from http:// Zellner, A. (1962). An efficient method of estimating seemingly un-
learnaboutmarijuanawa.org/factsheets/olderadults.htm related regressions and tests for aggregation bias. Journal of the