In this case study, at the client’s request, I describe a multifaceted analysis of the factors that influence healthcare workers’ “clinical skill evaluation score” and “patient-care management ability score.” Identifying which factors are truly important from among a wide range of candidate variables, and clarifying their degree of influence, provides insights essential for human-resource development and service-quality improvement in the medical setting.

Dr.DataScience combined multiple statistical methods—Pearson correlation analysis, principal component analysis, multiple regression analysis, and tests of group differences—to draw practical implications from complex data. In accordance with our confidentiality agreement, no specific figures or detailed medical background are disclosed.

Background and objective

The client aimed to enhance the expertise of healthcare workers and the quality of patient care, and to that end needed to deeply understand what factors influence the “clinical skill evaluation score” and the “patient-care management ability score.”

In particular, because there are many potential influencing factors—such as healthcare workers’ attributes, experience, training programs taken, and environmental factors in the medical setting—what was required was: to statistically analyze these complex relationships and identify the principal factors with a statistically significant influence on each evaluation score; to appropriately handle statistical challenges such as multicollinearity and ensure the model’s reliability; to evaluate in detail the influence that particular training programs have on the evaluation scores; and, based on these analysis results, to make concrete recommendations contributing to the capability development of healthcare workers and the improvement of medical services.

Data and variables

In this analysis, the following main response and explanatory variables were analyzed using healthcare-worker data.

1. Response variable 1: Clinical skill evaluation score
   • A score evaluating the professional skill level of healthcare workers.
2. Response variable 2: Patient-care management ability score
   • A score evaluating the ability to plan, implement, evaluate, and manage patient care.
3. Main explanatory variables
   • Healthcare workers’ attributes and experience: age group, years of clinical experience, department/division of affiliation, highest level of education/professional qualification, etc.
   • Training-program history: whether an advanced specialty medical training program was taken, history of taking basic medical management training, whether a comprehensive medical innovation training was taken, etc.
   • Environmental factors of the medical setting: the medical-information-system usage environment, the optimization status of clinical space, appropriate staffing of medical personnel, the frequency of information sharing within the team, the degree of involvement in treatment-policy decisions, appropriate protocol adherence, the efficiency of the patient-transport system, support for active involvement in clinical research, etc. These variables were initially divided into many elements, but were dimensionally reduced by principal component analysis and incorporated into the model in an aggregated form.

Analytical methods

In this case, a multi-stage statistical analysis was performed to clarify how the response variables are influenced by multiple factors.

1. Preliminary analysis
   • First, I performed Pearson correlation analysis to grasp the associations among the variables. As a result, variables statistically significantly associated (p<0.05) with the clinical skill evaluation score and the patient-care management ability score were identified.
   • Regarding the environmental factors of the medical setting in particular, because multiple variables were confirmed to be mutually related, I attempted to dimensionally reduce these variables. Using principal component analysis (PCA), I aggregated the many related variables into a small number of “principal components (PCs).” By selecting the top four principal components (PC1/PC2/PC3/PC4) based on eigenvalues and the cumulative proportion of variance explained, I reduced the nine-dimensional data to four dimensions.
2. Multiple regression analysis
   • Using the “clinical skill evaluation score” and the “patient-care management ability score” as the respective response variables, and the explanatory variables extracted above, I performed multiple regression analysis. I also checked the normality of the residuals, a prerequisite of multiple regression, and confirmed that this assumption was satisfied.
   • Among the explanatory variables, some—such as age group, years of clinical experience, and department/division of affiliation—were confirmed to have the problem of multicollinearity (strong correlation among variables). To address this problem, I performed the analysis in three different model patterns, each adopting one of these multicollinear variables.
   • I also referred to variable importance using permutation, a machine-learning method, to select variables that contribute to improving the model’s accuracy.
3. Subgroup analysis
   • To evaluate the influence that particular training programs (the advanced specialty medical training program and the comprehensive medical innovation training) have on the clinical skill evaluation score and the patient-care management ability score, I set subgroups and performed multiple regression analysis. Specifically, I analyzed two patterns: “those who took the advanced specialty medical training program vs. those who did not take the basic medical management training,” and “those who took the comprehensive medical innovation training vs. those who did not take the basic medical management training.”
4. Tests of differences
   • Furthermore, I tested, by two-sided tests, the differences in mean scores between groups for the following patterns: presence or absence of a history of taking the basic medical management training, whether the advanced specialty medical training program was taken, and whether the comprehensive medical innovation training was taken.
   • As a result, for the clinical skill evaluation score, significant differences were found between the groups “those who took the advanced specialty medical training program vs. those who did not take the basic medical management training” and “those who took the comprehensive medical innovation training vs. those who did not take the basic medical management training.”
   • For the patient-care management ability score, significant differences were detected in all of the above patterns.

Overview of the main results and clinical considerations

As a result of this analysis, various factors influencing healthcare workers’ clinical skill evaluation score and patient-care management ability score were clarified. The preliminary analysis showed that age group, years of clinical experience, whether particular training programs were taken, and the frequency of information sharing within the team, among others, were associated with both scores.

In particular, for the wide range of environmental factors in the medical setting, I extracted more essential influencing factors by performing dimension reduction via principal component analysis; however, some aspects remained difficult to interpret as to what combination of environmental factors each principal component specifically represented.

In the multiple regression analysis, I verified multiple model patterns while accounting for the problem of multicollinearity. For example, for the patient-care management ability score, it was suggested that department/division of affiliation and age group may have a statistically significant influence. Furthermore, through the subgroup analysis, it was confirmed that taking the advanced specialty medical training program had a significant positive influence on both scores, suggesting that particular specialty training may be effective in improving healthcare workers’ clinical ability and management ability.

On the other hand, the comprehensive medical innovation training did not show a statistically significant effect. These findings provide concrete implications for optimizing capability-development programs for healthcare workers and for building a more effective medical environment.

Dr.DataScience’s contribution

In this case, Dr.DataScience contributed, through multifaceted statistical analysis, to the client’s important challenge of improving healthcare workers’ evaluation scores. First, to narrow down the influencing factors from among many candidate variables, I performed a preliminary analysis combining Pearson correlation analysis and principal component analysis, and by statistically processing the complex environmental factors of the medical setting, I extracted more essential elements.

Next, in the multiple regression analysis for the clinical skill evaluation score and the patient-care management ability score, I ensured statistical reliability by verifying different model patterns for the challenge of variable groups with multicollinearity, such as age and experience. Furthermore, through subgroup analysis that evaluated the effects of particular training programs in detail, and by testing the differences between groups, I clarified not merely correlations but the effects of concrete interventions.

In this way, by making full use of advanced statistical methods to draw practical insights from complex medical data, Dr.DataScience strongly supported the client in making solid decisions that contribute to medical-personnel development strategies and the improvement of medical-service quality.