Literature Review Week 4
Article 1
Article Title: Exploring the medical decision-making patterns and influencing factors among the general Chinese public: a binary logistic regression analysis.[1]
Authors: Yuwen Lyu, Qian Xu, Junrong Liu
Problem: - Researchers are seeking to understand top driving factors behind decisions made about healthcare and medical issues. The population of interest is the general Chinese public. - Previous research in this field has identified two main types of medical decision making: unilateral and collaborative decision making - Unilateral decision making means there is one main entity making the medical decision, such as a single patient, a patient’s family, or a doctor. Previous research shows that patient families have a very strong influence over a patient’s medical decisions. - Collaborative decision making means there are two or more parties involved in the decision making process. Three subgroups are defined: doctor group, doctor- patient group, patient- family group, and patient-doctor-family group. - There is a lack in research in this field. More needs to be known about factors that play a role in medical decision making processes. This study’s results will be generalizable to China as well as nations around the world.
Solution: - Use binary logistic regression to classify points into two categories: unilateral decision making (value of 1), or collaborative decision making (value of 0) - This model is ideal because it takes into account variable interactions. Also used often in the medical field - The equation of the model is given. It is in the form of the log odds of the desired event happening.
Data: - 2696 data points with attributes including age, education, occupation, family situation, religion, economic status and medical payment methods - Data was collected via survey and included only residents of China from 31 provinces - The data was gathered by the researchers that wrote this study - A power analysis was conducted to determine how many data points would be needed in order to have reliable results after statistical analysis. A G-power test showed that only 2040 valid data points were needed
Results: - Survey results showed that 30% of responses were categorized as unilateral decision making, while 70% were categorized as collaborative decision making. The top category of unilateral decision making was doctor- led decisions, while the top category for collaborative decision making was patient- doctor- family decisions. - Significant predictors were identified with p-values less than 0.05. Significant predictors of unilateral decision making were gender, education level, family status, and religious beliefs. Different occupations also significantly predicted unilateral decision making. - Odds ratios are given for some predictors, with researchers stating that certain categories of specific predictors are x.xx times more likely than the reference group to be a unilateral decision maker. - The significance of the regression model’s intercept is interpreted, and it is significant. This means when all variables are at their reference levels, there is a low likelihood of the outcome variable taking on a value of a unilateral decision making process. - The goodness of fit test used in this study is McFadden’s R-squared value. The value was0.065, and researchers state that this value shows a good fit of the model. It is explained that R squared values for studies in the social sciences are rarely ever close to a perfect fit.
Conclusions: - The researchers discuss why there are contrasting results from this study versus studies in Western countries. They identify different cultural values in different geographic regions, which ultimately lead to different medical decision making processes. - Results are discussed more in depth, with researchers attempting to identify causes behind the correlations that were identified.
Limitations: - The study’s data is solely from China, so results may not be generalizable to global populations. - The binary logistic regression model may not be complex enough to account for the complexities of all the predictors involved in healthcare decision making. Researchers suggest using more complex models in future research.
Article 2
Article Title: Predictors of hospital admission when presenting with acute on chronic breathlessness: Binary logistic regression.[2]
Authors: Ann Hutchinson, Alastair Pickering, Paul Williams, Miriam Johnson
Problem: - People presenting to the emergency room with breathlessness often do not require hospitalization and can be sent home. Only an average of 50% to 67% of these patients require admittance to the hospital. This research focuses on patients presenting with breathlessness to the ER and seeks to identify significant predictors of hospitalization of these individuals. Doctors and emergency department staff would benefit from identifying predictors present in individuals that would need to be admitted to the hospital.
Solution: - Researchers used a binary logistic regression analysis to identify the predictors most strongly correlated with patients with breathlessness being admitted to the hospital from the ER. - First, in order to identify which predictors to include in the binary logistic regression, researchers analyzed 48 total predictors individually. They conducted a separate univariate analysis for each variable to see which were most significantly correlated with hospitalization from the ER. Only seven predictors were significant, and of those, only five were included in the final model (due to eliminating variables with strong multicollinearity).
Data: - Data was collected from a single hospital from December 2015 to May 2015. - Only 171 datapoints are included, as only 171 patients with acute breathlessness consented to have their survey used for research - Predictors included: demographics, preexisting medical conditions, severity of breathlessness, and other vital signs - To determine which predictors were most important, researchers used a stepwise backward elimination process, and excluded one predictor at a time. It was determined that only five predictors were needed. - After univariate analysis, researchers constructed a binary logistic regression model incorporating all of the selected predictors.
Results and Conclusions: - Results are presented with an odds ratio for every predictor in the model. The odds of being admitted to the hospital increased by a certain factor for every one unit increase in a specific predictor. - The odds of admission to the hospital were positively correlated with age, talking to a doctor about symptoms, and the presence of preexisting heart conditions. The odds of being admitted to the hospital were negatively associated with blood oxygen levels. - The researchers state that this study is meant to only be exploratory and the results should not be used for making future predictions. Rather, the results should be used to aid in identifying strong predictors so these variables can be included in future similar studies. - Results of this study are compared to results of other studies, and the findings are consistent. A consistent predictor of hospital admission from this study and other studies includes older age. Another common predictor is tachycardia, but this study did not include this information.
Limitations: - The study’s data is very limited. The data is supplied from just one hospital, and there were only 171 data points analyzed. The results of the study are therefore not as generalizable to global populations as a study analyzing broader populations. - The data also did not include responses from patients with very severe breathlessness because their state of health was too severe for them to be able to complete a survey. So the results of this study do not reflect patients with extreme symptoms.