Multivariate Approaches
In preparation for this assignment, read the “Telebehavioral Health (TBH) Use Among Rural Medicaid Beneficiaries: Relationships With Telehealth Policies” article located in the Topic 3 Resources.
Write a 750-1,000-word paper about your selected article. Be sure to include the following in your paper:
- A discussion about the key variables in the selected article
- Identify the validity and reliability reported statistics for the article
- The particular threats to internal validity that were found in the study
- The strengths and limitations of the multivariate models used in the selected article
- A reference and in-text citations for the selected article as well as one additional reference
Prepare this assignment according to the guidelines found in the APA Style Guide
Multivariate Approaches
Student Name
University
Course
Professor Name
Date
The study titled “Telebehavioral Health (TBH) Use among Rural Medicaid Beneficiaries: Relationships with Telehealth Policies” by Talbot et al. (2020) examined the relationship between state Medicaid telehealth policies and the use of telebehavioral health among the beneficiaries of rural fee-for-service (FFS). The study used a cross-sectional design and multivariate data analysis methods. The purpose of this essay is to assess multivariate approaches by examining the variables and covariates in the study, validity and reliability, and the strengths and limitations of the multivariate models.
Variables in the Study
The outcome variable in the study was the use of telebehavioral health while the explanatory variable was telehealth policies. The process of selecting covariates for multivariate analysis was based on the behavioral mode for vulnerable populations. This model identifies the factors that influence the use of health services among underserved populations including people living in rural areas. The main covariates included in the study were state, county, and individual-level covariates. The study also examined the characteristics of the beneficiaries including race/ethnicity, gender, age, and presence of serious mental illnesses such as bipolar disorder, schizophrenia, major depressive disorder, and other psychotic disorders. The FFS beneficiaries were categorized as residing in rural areas that were either nonadjacent or adjacent to towns or cities (Talbot et al., 2020).
Validity and reliability reported statistics for the article
The validity reported statistics for the article were used to make conclusions about the relationship between the variables. The p-value was mostly used to demonstrate the statistical significance of the relationships between variables. A p-value of less than 0.05 indicated statistically significant associations, while a p-value of greater than 0.05 indicated that there were no associations between variables. Other statistics that were used to explain the relationships between telebehavioral health use and the covariates included the odds ratios and the confidence intervals of these ratios. Odds ratios were used as measures of association that quantify the relationship between the explanatory and outcome variables. A ratio with a value greater than 1 indicated greater odds of association, a value equal to one indicated no association, while a value less than one indicated less odds of association. There was no mention of the validity and reliability of the data sources although the primary data source was the Medicaid Analytic eXtract (MAX) that was constructed by the CMS using the Medicaid program data submitted by states (Talbot et al., 2020).
Threats to Internal Validity
Internal validity refers to the degree to which a researcher can be confident that the cause-and-effect relationship that was established during research cannot be attributed to other factors (Patino & Ferreira, 2018). In the study, the main threat to internal validity is the study design. The researchers used the cross-sectional design which does not provide adequate support for making definitive conclusions regarding the cause-and-effect relationship between telebehavioral health use and the explanatory variables. The findings illustrated that telebehavioral health use was higher among FFS beneficiaries suffering from severe mental illness, those living in areas with mental health professional shortages, and those living in rural areas that are not adjacent to metropolitan areas. The cross-sectional study design only examined data collected at one point in time. Notably, since the data was collected in 2011, Fee-For-Service and Telehealth policies have undergone significant change. By 2019, the number of state Medicaid programs working within the FFS environment had declined from 36 to 10 states. Additionally, telehealth policies have evolved. Since more healthcare providers work within the policy environments, they have likely acquired infrastructure that supports telehealth services. Therefore, the factors influencing telebehavioral health use now are likely to vary significantly from the factors determined by the research due to changes and events brought about by the passage of time (Talbot et al., 2020).
Strengths and Limitations of Multivariate Models
The main multivariate model used in the study included generalized estimating equations. The researchers also used contrast analysis to provide additional insight into the nature of the interaction between the variables. According to Pekar and Brabec (2017), the main strength of using the generalized estimating equations model is that parameter estimates are efficient when there is a correct specification of the working correlation structure. Additionally, the parameter estimates always remain consistent even when there is a misspecified working correlation due to the robustness of the model. The main limitation of using the generalized estimating equations is that the parameter estimates are sensitive to any contaminated data and outliers and may fail to give consistent estimators which would lead to incorrect conclusions. In case the working correlation matrix is misspecified, the parameter estimates made may be inefficient. When using the generalized estimating equations model, it is also difficult to use the model selection criterion since the model lacks an objective function.
References
Patino, C., & Ferreira, J. (2018). Internal and external validity: can you apply research study results to your patients? Jornal Brasileiro De Pneumologia, 44(3), 183-183. https://doi.org/10.1590/s1806-37562018000000164
Pekar, S., & Brabec, M. (2017). Generalized estimating equations: A pragmatic and flexible approach to the marginal GLM modelling of correlated data in the behavioural sciences. Ethology, 124(2), 86-93. https://doi.org/10.1111/eth.12713
Talbot, J., Jonk, Y., Burgess, A., Thayer, D., Ziller, E., Paluso, N., & Coburn, A. (2020). Telebehavioral health (TBH) use among rural Medicaid beneficiaries: Relationships with telehealth policies. Journal of Rural Mental Health, 44(4), 217-231. https://doi.org/10.1037/rmh0000160