Predicting Stroke Patient Outcomes at Discharge

Every year, around 15 million people around the world have a stroke. This serious condition leads to about five million deaths and another five million people living with disabilities related to strokes. Rehabilitation services are very important for helping stroke patients recover and become independent.

Before being discharged from the hospital, stroke patients should go through a formal assessment to check their daily activities, communication skills, and ability to move around. The results of these assessments need to be used to help plan their care and discharge process. When deciding where a patient will go after their stay, doctors consider how well they might recover as a key factor. This means that predicting a patient’s condition at the time of discharge based on their initial assessment can greatly help in creating personalized rehabilitation plans.

#Prediction Tools

Recent reviews have shown that various tools have been created to predict different outcomes for stroke patients, including their independence, ability to use their arms, walk, and swallow. However, many past studies have focused on predicting just one outcome at a time. In a clinical setting, it is essential to predict a patient’s overall condition, which should include multiple outcomes at discharge rather than just one. For instance, a patient might be able to walk independently but still need help with communication or thinking skills. Unfortunately, previous studies have not provided a comprehensive model to predict these overall patient characteristics. Therefore, there is a need for a new prediction model that can assess the overall situation of stroke patients when they leave the hospital.

#Using Latent Class Analysis

To address these issues, we used a method called latent class analysis (LCA). This method helps sort patients into different groups based on characteristics that are not easy to measure directly. LCA is useful when the differences among patients are influenced by various underlying factors. With this analysis, we can identify classes of overall patient characteristics, such as where they will go after discharge, how well they can move, and how long they stayed in the hospital. We can then create a model to predict which class a patient belongs to based on their initial assessment.

#Study Background

Our study looked at data from the Japan Association of Rehabilitation Database. We obtained information from patients admitted to hospitals between January 2005 and March 2016. Since all data were anonymized, we did not need to ask for individual consent. The research was approved by the ethics committee at Kanagawa University of Human Services.

We collected information on patients who had strokes, including their age, type of stroke, how severe their condition was, and the type and length of rehabilitation they received during their stay. A total of 33,657 patients’ data were gathered in 2016, involving 80 hospitals. Specifically, we focused on data from 10,270 stroke patients admitted to acute care hospitals.

#Inclusion and Exclusion Criteria

To be included in this study, patients had to meet certain criteria:

1. They must have had an acute stroke and been admitted to one of the hospitals in the database between 2005 and 2015.
2. They had to be at least 18 years old.
3. They needed to have received some form of rehabilitation during their hospital stay.

Patients with the following conditions were excluded:

1. Those without age data at admission.
2. Those without information about their discharge destination.
3. Those younger than 18 years.
4. Patients who did not receive rehabilitation during their stay.
5. Individuals who were admitted more than seven days after their stroke occurred.
6. Patients who received more than nine units (or 180 minutes) of rehabilitation each day.
7. Individuals who stayed in the hospital for fewer than one day or more than 179 days.
8. Patients missing data on specific assessments at admission or discharge.
9. Patients missing data on all FIM items at both admission and discharge.

#Outcome Variables at Discharge

After reviewing past studies, we selected specific outcome variables to assess at discharge using LCA. These included:

• Functional abilities (daily activities)
• Cognitive Functions (thinking and understanding)
• Upper extremity function (ability to use arms)
• Length of stay in the hospital
• Discharge destination (where the patient goes after leaving the hospital)

The Functional Independence Measure (FIM) was used to assess functional abilities at discharge, which includes 18 items that show how much help a person needs to perform daily activities.

Cognitive functions were also assessed using the cognitive subscale of the FIM, focusing on comprehension, expression, and social interaction. These factors are crucial, as difficulties in understanding or communicating can significantly affect a patient’s recovery.

Upper extremity function was measured using scores from the National Institutes of Health Stroke Scale (NIHSS), which assesses the severity of a patient’s condition. The length of stay varied by stroke severity and a patient’s needs. Discharge destinations were defined as home, other hospitals, rehabilitation facilities, death, or other outcomes.

#Predictor Variables

The study also considered certain predictor variables that could influence patient outcomes at discharge. These included:

• Age
• Functional abilities
• Comprehension skills
• Upper extremity function
• Type of stroke
• Amount of rehabilitation received during hospitalization

Previous studies suggested that better initial functional abilities and lower age increased the chances of being discharged home. We considered several factors from past research to make our prediction model as accurate as possible.

#Statistical Analyses

To develop the prediction model using LCA, we followed a two-step method. In the first step, we conducted LCA with our selected outcome variables to identify patient conditions at discharge. The best model was chosen based on specific criteria to ensure it was the most effective.

In the second step, we predicted patient class membership at discharge using the predictors selected in the first step. We also addressed any missing data during the analysis.

#Patient Characteristics

After applying our criteria, we included 6,881 patients in the study. The average age of patients at admission was 73.7 years, and the average length of stay was 29.5 days. Overall, there were improvements in functional scores from admission to discharge.

#Latent Classes of Patient Characteristics at Discharge

Using LCA, we classified the 6,881 patients into different classes based on their discharge outcomes. We examined models with 1 to 12 classes to find the best fit. The study found that the nine-class model was the most interpretable and clinically relevant.

Each class reflected different conditions of patients at discharge, with Class 1 representing those with the mildest conditions and Class 2 indicating the most severe conditions.

#Predictors of Class Membership

All predictors were significant in predicting outcomes at discharge. The highest odds of belonging to specific classes were linked to upper extremity function scores measured at admission. The amount of daily rehabilitation had varied effects depending on patient condition.

#Model Application

Our prediction model successfully identified overall patient characteristics in acute stroke patients at discharge. It can guide healthcare providers in offering the most suitable interventions for each individual.

For example, if a patient is 85 years old with a severe disability on one side of their body and receives a certain amount of rehabilitation, the model can predict that this patient is likely to need assistance with daily activities and communication upon discharge.

#Conclusion

In conclusion, our study developed a model to predict overall patient conditions at discharge, focusing on the appropriate rehabilitation intensity in stroke patients. However, further validation of this prediction model is essential before it can be used in clinical settings. Future research could improve understanding of how different factors influence rehabilitation outcomes based on individual patient characteristics.