I am very happy to say that I have been named Executive Director of the Center for Health Information and Analysis. My appointment letter is below.
In response to a question about relative compensation in the public and private sectors, which arose at the February, 2016 meeting of the Massachusetts Group Insurance Commission, I grabbed the most recent wage data from the US Bureau of Labor Statistics, segmented out the Massachusetts data, and created the dashboard below. Because this visualization is wider than my blog’s style sheet allows, you need to access the interactive version on the Tableau Public website. A word of caution: to interact with the dashboard a little patience goes a long way. Hover for a second before you click something, and wait for a second after you click something. There is a lot of data behind this dashboard, and it is being served up across the Internet. Enjoy!
The following dashboard provides a breakdown of 7,759 emails I received over a ten week period in connection with my role as a Commissioner on the Massachusetts Group Insurance Commission (GIC). The GIC is the Commonwealth agency that obtains medical, dental, vision, and other insurance benefits for employees and retirees (and their dependents) of the Commonwealth of Massachusetts and numerous municipalities. As a result of budget pressures, this winter the GIC had to consider making various adjustments to member benefits in order to reduce costs. In anticipation of these discussions and subsequent votes, various groups of public employees and retirees were encourage to contact the Commissioners to express their opposition to any benefit reductions. As the number of emails continued to pile up, I felt that merely knowing the number of messages was a poor way of understanding what for me is an unusual phenomenon – thousands of form emails, none of it spam. And so, since the text of the messages was all identical (save for some very rare personalizations), the metadata associated with each message offered the best source of data for analyzing the email campaign. Happily, this also provided a reason to apply what I have been learning about programming in Python, and so I wrote a script to extract basic metadata from each message and save that data to a file. I then used Tableau to do some basic analysis of this data, and combined it into the dashboard below.
For a while, each time I saw Hans Rosling’s famous GapMinder visualization of GDP and life expectancy for the counties of the world, I would have the desire to construct a similar viz for Massachusetts, with municipalities taking the place of countries. The following is a work in progress, showing nine scatter plots that explore the relationship between poverty at the municipal level and various other socio-economic factors.
Borrowing from Rosling’s visual formulation, the size of each bubble is proportional to population, and the color of each bubble corresponds to the municipality’s geographic region, in this case EOHHS Region. I had originally used counties as the color variable, but it turns out having 14 geographic groupings is less useful than having only 6 colors, and since the topic at hand is poverty, the EOHHS Regions seemed like a logical way to group the municipalities.
All of the data for these scatter plots is taken from the US Census Bureau’s American Community Survey, 2009-2013. Since this is a survey, rather than a census, it involves making projections from small, random samples. Accordingly, these numbers should be thought of as estimates, not as actuals. Selecting an EOHHS Region from the list will highlight the municipalities in that Region in all of the charts. Choosing a bubble in any chart will highlight that municipality’s bubble in each of the charts.
The tabbed visualization below allows interactive access to various data about discharges, charges, and payments for Massachusetts hospitals for the years 2011 and 2012. If the visualization below does not render well, try viewing it on the Tableau Public website here.
This is a first draft effort to pull together some visualizations using the combined 2011 and 2012 IPPS charge and payment data for Massachusetts hospitals. This is very much a work in progress and will be updated as time allows.
As part of the Accountable Care Organization (ACO) Shared Savings Program, CMS has released aggregate spending data to let ACOs calculate their market share if required to do so by the ACO antitrust enforcement policy adopted by the Federal Trade Commission and the Department of Justice.
The CMS data shows Medicare spending at the ZIP Code level for FY2010, CY2011, and CY2012, and comes in three different files – inpatient, outpatient and ambulatory surgical centers, and physician services.
The Inpatient Facility File is broken down by Major Diagnostic Categories and covers facilities paid under the Inpatient Prospective Payment System, Critical Access Hospitals, the Inpatient Rehabilitation Facility Prospective Payment System, Inpatient Psychiatric Prospective Payment System, Long Term Care Hospital Prospective Payment System, Indian Health Service Hospitals, Children’s Hospitals (to extent CMS has data available), Cancer Hospitals and TEFRA Hospitals.
The Outpatient Facility File covers outpatient fee-for-service claims for facilities that include Ambulatory Surgical Centers, Outpatient Prospective Payment Systems facilities, Critical Access Hospitals, Comprehensive Outpatient Rehabilitation Facilities, Community Mental Health Centers, End-Stage Renal Disease facilities, Federally Qualified Health Centers, Outpatient Rehabilitation Facilities and Rural Health Clinics.
The Physician File contains all physician fee-for-service claims from physicians with one of the relevant specialty codes (with general practice, family practice, internal medicine, and geriatric medicine combined into the Primary Care category). Claims are attributed/assigned to the category of the physician’s primary specialty.
The following dashboard combines data from all three files (the Inpatient Facility File, the Outpatient Facility File, and the Physician File) for Massachusetts for 2012. Use the drop-down list near the upper right to switch between the three service types.
April 17, 2014 – This is just a quick viz (actually, to my knowledge it is the first viz) of the history of uniform law adoption in the United States. The underlying data, which is mostly clean but still needs a small amount of work, is taken from Table VII of a seven table set of information about acts promulgated by the Uniform Law Commission (formerly known as the National Conference of Commissioners on Uniform State Laws). I have been pulling and cleaning this data with the permission of the ULC in connection with a research project I am collaborating on at the MIT Media Lab’s Human Dynamics Group.
As soon as I started doing data visualizations, I knew I would eventually need to create Massachusetts municipal-level maps. I was born and raised in Massachusetts, my background and interests are in the fields of public policy and data, and a great deal of the government data that interests me is either collected or reported at the municipal level. As part of my mid-career pivot toward data and analytics, I have been using government data sets as fodder for learning R, Tableau, and Python, and much of this data is broken down at the municipal level.
After I had gathered and analyzed some data on municipal demographics, health, finances, and elections I was interested in creating some maps, which is when I discovered that Massachusetts municipal boundaries are not part of the default geographies recognized by Tableau. Fortunately, the Commonwealth of Massachusetts makes a wealth of data, including geospatial data, available online if you know where to find it. It would be nice if there was more consistent and complete use of the state’s Open Data Initiative website, but public resources are exceedingly tight, and making government data easily available in a centralized way is not a mission critical activity for most agencies.
The state’s Office of Geographic Information, also known as MassGIS, is part of an agency I used to work for, the Information Technology Division. MassGIS has an impressive amount of data available for download, a list of which is available here. From this list I was able to find an extremely detailed political boundary datalayer (557,360 rows of latitude and longitude data) that is available in ArcGIS shape files format. The issue for me is that Tableau will import data from ArcGIS, but only if it has been packaged a certain way. Since I don’t have ArcGIS to do the packaging of the MassGIS file, I was dead in the water.
At that point, a woman who shall remain nameless (unless I get her permission to use her name), and whom I’ve met only by email through a connection at the MIT Media Lab, offered to convert the MassGIS files in ArcGIS and send them back to me. I took her up on her kind offer, she did exactly as she said, and the whole thing works just as Tableau said it would. At some point I need to see if R offers a way to bridge between ArcGIS and Tableau for geospatial data, at least until standardized approaches take hold.
Now that I have the municipal boundary data in bare CSV or XLS format the way Tableau likes it, I can join that datalayer to other datasets that include municipality as a field (although that’s trickier than it sounds on the surface). I’m going to start with some basic population information, and then do some maps of prior statewide election results. The Secretary of State’s Elections Division has an excellent searchable database of election results since 1970, and as we enter a gubernatorial election year I think it would be fun, in an informational, non-partisan way, to create some interactive maps of prior statewide election results.
The dashboard below takes public data from the US Census Bureau to show information about the Massachusetts population in 2010, and also about how it has changed since 1980. I hope to update the dashboard a few times to give it more depth and polish, but here is an initial iteration. The top map is color coded to show 2010 population, while the bottom map is color coded to show 1980-2010 population change. Rolling over any municipality brings up additional information.
The following viz is a work in progress. For a while, I’ve been wishing there was an interactive atlas of Massachusetts healthcare data, and in particular I have always wanted to see what the various provider catchment areas in the state would look like if plotted on a map. Having recently obtained the Hospital Summary Utilization data reports for 2005-2010 from the Massachusetts Center for Health Information and Analysis, I was pleased to discover that for 2006-2010 there are tables that lists each hospital’s 40 most frequent zip codes of patient origin, along with the number of patients from each such zip code.
The following map, which uses the 2010 data, is a starting point that I hope to improve as time allows. To use the map, click the name of a hospital from the list to highlight that hospital’s zip codes on the map. The best way to zoom in (although the full-state view works well for most purposes) is to use the rectangle control in the upper left. This zooms and centers on the spot you click. To deselect a hospital, click its name again.
More to come . . .