Introduction

This project is focusing on mapping healthcare facilities and mental health services in New York City using Tableau. The idea was generated from the willingness to find out if citizens in NYC have adequate access to healthcare help. For this lab on mapping, I chose to work with data on NYC healthcare centers, specifically a dataset from 2011 on Health and Hospitals patient care in New York City by council districts, which also contained data on the facility name, facility type, zip code, borough, etc. Meanwhile, I sourced this dataset through NYC OpenData as the shapefile. I was interested in creating a visualization that demonstrated different facility types mapped in districts as well as counting how many types within the categories of location, residential, and council.

Background

As I was exploring the datasets on NYC Open Data, one file from Political and Administrative Districts caught my attention. The geographies represent the post-redistricting political boundary changes implemented by the NYC Board of Elections to reflect 2020 Census changes in population. A corresponding version of Geosupport Desktop Edition™ (Release 22A1) that reflects these political boundary changes is also available for download from the website.

When I look for other visualizations that use the same dataset, I encountered a report on HITConsultant which dive deep into the analysis of finding over a third of the U.S. population lives in a county where there is less than adequate access to pharmacies, primary care providers, hospitals, trauma centers, and/or low-cost health centers. (Resource) Inspired by the purpose of creating a health map that using released data from the City of New York Government , from which we can introduce a wide range of national datasets (e.g. census) and local health datasets to support a greater focus on evidence based planning and decision making.

Goal

Inspired by the visualizations of healthcare deserts, and also “Mapping Healthcare Deserts: 80% of the Country Lacks Adequate Access to Healthcare“, I decided to explore the total distribution and per capita distribution of Health Center Facility Type and Mental Health Service in NYC to find out:

1. The top and last three of total distribution on health center type/mental health service based on NYC council district;
2. The top and last of total distribution on health center type/mental health service based on NYC borough;
3. The top and last three of per capita distribution on health center type/mental health service based on NYC council district;
4. The top and last of per capita distribution on health center type/mental health service based on NYC borough

Map visualization will be applied for a better view of distribution on location and by dimension, along with that, I’ll use bar chart for a better comparison between districts and borough and I’ll use bubble chart and tree chart for audience to figure out the largest group within health center facility types in NYC.

Materials and Dataset

For this map dashboard in Tableau, I utilized three datasets and two shape files. The datasets are “NYC Health + Hospitals patient care locations – 2011”, “Mental Health Service Finder Data”, and “New York City Population by Borough, 1950 – 2040”, and the shapefiles are “NYC Borough Boundaries” and “NYC City Council District”. NYC Health + Hospitals is the largest municipal health care system in the country serving more than one million New Yorkers every year. From these datasets, the City’s public health care delivery system provides trauma, emergency, medical, mental health, and substance abuse services across the five boroughs. This is a list of the public hospitals, skilled nursing facilities, and some community-based health centers that are part of the NYC Health + Hospitals system as of 2011.

I downloaded three datasets as CSV file, after I tried to link the dataset “NYC Health + Hospitals patient care locations” file in Tableau, there’s an error with the value where the council district appears as Null instead of strings or numbers. I used OpenRefine to transform the file to google sheet then linked it in Tableau after I cleaned up and organized the data(Figure 2). For the dataset “Mental Health Service Finder Data”, the original dataset is not in a borough format but in city format. I used R to transform and categorize all city data to boroughs(Figure 3).

Methods and Design Process

1.Joining Sheets in Tableau

In order to map in Tableau, we need shapefiles and dataset files. By joining them through certain relationships, the information on a dataset file could be visualized on a map. Tableau also supports joining two spatial data sources using their spatial features (geography or geometry). You can only create spatial joins between points and polygons. For a better and intuitive view which will meet the goal of the project, I’ll use two shapefiles on Council districts and on Borough to see the distribution of the healthcare centers. (Figure 3)

For the first shape file “nycc”, there’s a “Coun Dist” column contains the council districts data, by connecting the column with the Council district column in the NYC health center dataset, it turned out Council district map the data into matching categories. When clicked into “nycc” file, it allowed me to join two shapefiles using intersect on “Geometry”.

2. Creating the Bottom Map

Distribution of health care on map is the best way to look at the overall density of facilities and services. Two base maps of the visualization is NYC City Council District(Clipped to Shoreline) with the Council District as categorized to dimension overlaid to top(Figure 4), and NYC Borough(Clipped to Shoreline) with Borough overlaid to top(Figure 5). By dragging count of my dataset to the basemap, it’s obviously showcase the scale from 1 – 5 using gradient color. Green was selected for the district color based on general color association with the topic of healthcare center, and purple was on the association with the theme of mental health. I also created related map based on borough for the use to add in datasets layers in the next step.

3. Adding Layers and Creating Dot Map for total distribution

Based on the category map, I’d like to visualize the total health center facility type and mental health service of NYC in a visually straightforward way. Adding a layer of Facility Type upon the shape file achieved this goal and by adding in the coordinate information, it formed a dot map. Selecting the colors and opacities were crucial for the layers, which are meat to be viewed one layer at a time over the base map. Shades of red were used for the healthcare type because they are opposite to the green council district area on the color wheel. I choose colors that are generally associated with the variable that they represent, for example, red for acute care hospital, pink for child health center, purple for Diagnostic & treatment center and yellow for nursing home, all set to 80% opacity to create enough distinction. It was also important to consider the way the colors of dots would look overlaid on the shades of green in the base map to be able to recognize locations.

Same methods were applied when creating the dot map of NYC mental health service, since the dataset includes individual mental health service and I’d like to have a general insight from the visualization, so I marked all addiction related mental health service to orange, and all retreatment related mental health service to blue.

4. Creating bar chart, bubble chart and tree chart for per capita distribution

To compare and contrast per capita distribution by council district and borough, we need to use the calculation function in “Calculated Field” in Tableau to create a new column. Applying “Count(NYC Health Hospitals Records)/(2010 Boro Population)” to generate the ratio of health center located in each borough/council district. I created three per capita bar chart to visualize the per capita distribution of mental health service by borough, per capita distribution of health center by council district and borough. For better comparison, I created bar chart of total distribution and it will give us very interesting result when comparing with each others in NYC.

Other than that, I’m interested to see which facility type in NYC occupies the largest proportion compared to others. Hence, I created bubble chart and tree chart as below.

In the end, instead of putting every map and chart on the same poster, I created three dashboards to showcase similar categorized information and by adding an interactive arrow button, viewers can be navigated through each dashboard and jump back to the first map dashboard quickly.

UX Research Protocol

I plan to recruit two participants for in-person user testing after the first draft of the dashboard has been created. During the session, users will be asked to think out loud. The tasks are designed to evaluate the information structure, visual/appearance, and interactivity of the chart. 5 tasks were given to them to complete, followed by post-task questions:

1. Try to find a section on the chart that interests you, and describe the information you find.
   • What did you see at first sight of exploring the map/chart?
   • Why does the information interests you?
2. Explore each dashboard of the visualization, and describe the information you find.
   • What do you think of the information?
   • Do you think the slides show of the dashboards make sense to you?
   • Does the color stand out and make sense to you?
3. Suppose you are planning to, try exploring the map in the first dashboard.
   • Did you find the information you want?
   • Does the coloring make sense to you?
4. Suppose you are planning to, try exploring the chart in the second dashboard.
   • Did you find the information you want?
   • Does the coloring make sense to you?
5. Suppose you are planning to, try exploring the chart in the third dashboard.
   • Did you find the information you want?
   • Does the coloring make sense to you?
6. Compare total distribution chart and per capita distribution chart of the visualizations, and describe the information you find.
   • Can you describe the information you see from the chart?
   • Is there any information you feel is missing in each visualization?
7. Check the bubble chart and the tree chart, and describe the information you find.
   • Does the color code make sense to you?
   • Does the information provide meet your expectation?

Findings and Results

Visualization

From the council district and borough map, we can tell from the color of the scale that there are more facility types in most of the districts in Manhattan, especially lower Manhattan and upper Manhattan. Also there are more healthcare facility types in Brooklyn, especially around Bedford, Crown Heights, Ocean Hill, New Lots, etc. There are fewer health care facility types on Staten Island and also on the east side of New York, as well as Bay Bridge in Brooklyn. Apart from that, top two of council district with the most health facility is district 1 and district 36, the borough that owns the most health centers is brooklyn which has 25 health facilities and child care center occupies the most.

From the bar chart(Figure 9&10), we can tell that if calculated by total distribution, Manhattan has the most mental health services while Brooklyn owns the most health center facilities, Staten Island has the least mental health services as well as the least health center facilities. However if we focus on the health resource of individuals and involve in the population, Staten Island owns the most per capita health center facility resources compared to other boroughs, and the second most mental health service resource next only to Manhattan.

UX Research Results

According to the ux research protocol, I recruited two participants to test on the visualizations. The first participant is 25 years old female and is majoring in art and animation, the second participant is 34 years old male and is working as a web engineer. They all gave positive impression and feedback on the dashboard and also provided some thoughts,suggestions and confusions as below:

1. One participant was confused about council district number when looking at the second dashboard, he’d like to see the council district map near to the bar chart for better comprehend of the visualization;
2. It took a while for them realize the bar chart is for the purpose of comparison between total distribution and per capita distribution;
3. Two participants think it made more sense to look at the per capita distribution of health resource in NYC and they’re surprised to see the different results of Staten Island;
4. One participant mentioned that on the map dashboard of “NYC Health Center Facility Type by Council District” map, though there’s scaled color from light green to dark green to show the amount of the facility type, she thought to make it as gray if the count is zero will help with better understanding;
5. When it comes to the bubble chart and the tree chart, they all prefer to see tree chart to figure out the most health center in NYC is children health center. One participant questioned that to use pie chart to show the proportion of different types of the health center;
6. Color use of the map and the categories makes sense to both participants and they agreed that the interpretation of the visualization is efficient and effective.

Recommendations

Based on the findings and feedbacks collected from the ux research, the recommendations will be applied on the visualizations in the future as below:

1. For better differentiate between total and per capita distribution, the change of the title will be applied to each map and chart;
2. To create another dashboard of distribution by council district along with the bar chart of per capita distribution and total distribution;
3. Implement pie chart to present the proportion of the health facility type;
4. Come up with a method to change the corresponding color to gray when it comes to zero count on the map

Reference

1. report on HITConsultant
2. NYC Health + Hospitals patient care locations – 2011