Difference between revisions of "AY1516 T2 Inpatient Meals Survey ProjectOverview Methodology"
Poojat.2012 (talk | contribs) |
Poojat.2012 (talk | contribs) |
||
| Line 67: | Line 67: | ||
The user has the option to select and view the overall divergent graph for any of the months for which the survey has been uploaded. Apart from the overall view, we are also drilling down across different dimensions of the data. These dimensions include: | The user has the option to select and view the overall divergent graph for any of the months for which the survey has been uploaded. Apart from the overall view, we are also drilling down across different dimensions of the data. These dimensions include: | ||
| − | + | *Wards | |
| − | + | *Types of patients (Private or Subsidized) | |
| − | + | *Types of patient diets (Normal diet or Special diet) | |
Moreover, time series analysis is used to view and analyze the survey data across different months. The user can drill down to a specific attribute and see the trend of its ratings across several months. | Moreover, time series analysis is used to view and analyze the survey data across different months. The user can drill down to a specific attribute and see the trend of its ratings across several months. | ||
Revision as of 18:26, 28 February 2016
|
|
|
|
|
| Project Description | Data | Methodology |
|---|
Current Method
MRC currently conducts these surveys using Qualtrics and then uses SPSS software to analyze the data. The main statistical measure used to analyze the data is the mean of the ratings given for each attribute. They drill down to analyze the average rating by wards, dietary preferences, types of patients and other dimensions. However, as discussed before in the report, this method is flawed and does not provide the best results. This is primarily because the distance between the ratings are not measurable and cannot be interpreted.
The mean can fall in a decimal and a mean of say 3.4 or 3.6 is not telling much of the story that the client will want to read. What can one make of the data when the mean comes to a decimal point between 3 and 4? Say for example the mean is 3.4, the report would say that the patients are satisfied on an average, but then again this is a mean what if this mean arrived due to a lot of people rating 5 and just a lesser number than that rating 1. Is the analysis correct then? No, because there are a lot of people who are not at all satisfied and their needs are overlooked.
Different graphs are generated and then insights are collaborated. Below are some screenshots of the analysis done by MRC on the survey data for October 2015 and November 2015.
Overall Satisfaction by Ward Class
Overall Satisfaction by Patient Type
Service Gap Analysis
Analysis by Patient Ward
Images source: MRC Mediacorp Oct 2015-Nov 2015 Survey analysis
New Method
The main objective is to provide the users with a dashboard which provides visualizations that appropriately represents the data and can be interpreted easily. The dashboard is built using HTML, CSS and Java and the visualizations are made using Javascript and D3.js
The dashboard provides upload functionality where the user can upload the survey data that he/she wants to analyze. This ensures that the dashboard is dynamic and can be used for analysing future surveys. The application is linked to a MySQL database which keeps record of the data that is uploaded. This helps the user to store earlier surveys and to analyze the attributes across months and record the change in pattern.
The statistical measure used to represent the data is frequency of each rating for each factor as a percentage of the total population of interviewees. The basic graph shows the frequency of the ratings across each attribute for all patients using a divergent bar graph for both importance and satisfaction. As discussed earlier, this graph is easy to understand and interpret and provides an overall view of the ratings across each attribute.
The user has the option to select and view the overall divergent graph for any of the months for which the survey has been uploaded. Apart from the overall view, we are also drilling down across different dimensions of the data. These dimensions include:
- Wards
- Types of patients (Private or Subsidized)
- Types of patient diets (Normal diet or Special diet)
Moreover, time series analysis is used to view and analyze the survey data across different months. The user can drill down to a specific attribute and see the trend of its ratings across several months.
Apart from analyzing importance and satisfaction separately, it is also imperative to analyze the service gap for each attribute. Service gap represents the difference in the satisfaction and importance rating for each attribute. Bigger the difference in the importance and satisfaction, more attention needs to paid to that attribute.
All the above analysis is represented using Divergent bar graph which can be easily implemented by using D3.js.
Apart from that, the dashboard also provides the relationship between the different attributes based on the ratings provided by the patients. This is visually represented using parallel sets. The picture below shows an example of a parallel set. D3.js helps build an interactive parallel set to see the relationship between the ratings across different attributes.
The dashboard is a simple, easy-to-use application which provides interactive graphs with appropriate filters for in-depth analysis. The graphs are easy to read and interpret and provide accurate representation of Likert scale data.
