Why Government IT is Expensive and Archaic

Disclaimer: I do not work for the government, and my rant below is based on my very limited exposure to how IT works at the US government setting.

Why Government IT is Expensive and Archaic? I think, this can be a very long discussion, but I do have a quick answer:  standards imposed by government and used by the government regulated industries. I have a very little experience with these types of standards, but they make me cringe every time I have to deal with them. Bellow I briefly describe my encounter with them.

When I was just a college student, I joined a (very) small IT company sitting next to the University campus. I started just as an intern during my 3rd year of college, and I was working fulltime a year later. At the same time, our team was tasked with making a piece of software for a private company to be used for tracking medical services provided to students at public schools throughout the state. This piece of software was to replace an older one, and we had a very strict set of requirements: “Make it work and look the same but better”.  Such requirements, along with many decisions a college-student-turned-software-engineer had to make, shaped how the system is working right now. Of course, it was not just me making the product, but nevertheless, my “brilliant” ideas slipped in and became what the system is today.

One part if the system is responsible for billing the medical services tracked in the system to Medicaid. And this is where the interactions with government IT has started. Government agency cannot just provide a secure API for the software developers like myself to use. No, it has to hire a major business to implement a standard commonly used for Medical transaction. And here we are, in the 21st having to adhere to the Electronic Data Interchange (EDI) X12, a standard designed in the 1970s to transfer medical (among other EDI uses) data between computer systems. Don’t get me wrong, standards are good, they make different systems work together flawlessly… But that is until you start looking at the standard. For the system, we had to implement only one transaction type at that moment, so what can be difficult about it? The difficulty started with a 700 page manual just for constructing the transaction request. The manual is accompanied by an errata and an errata of errata. In addition, there is a 100 page companion guide that specifies procedures specific for the state Medicaid.

So how easy is the standard itself? Maybe it is not that complicated to work with and the manuals are just full of fluff? Well, EDI is a textual format, so in theory a person can read the file to see all the data. Manuals even provide the example of how pieces of it should look like: SV1HC:99211:2512.25UN111✽✽1:2:3✽✽Y~

Easy enough to read? Nope, so you are back at the manual studying what everything means. For example, SV1 is the header for section describing the professional (medical) service provided, HC code describes what type of codes to follow next, 99211 is the service code followed by code modifiers and modifiers of modifiers and so on. Somewhere in there is how much to charge Medicaid and how many units of service have been provided. But on the bright side, it has cool delimiters: stars, colons and tildes. To top it off, each field can be flexible in size or can be restricted to some number of characters or to a certain set of values, and to find this you consult the manual once again.

What if you make an error? Not a problem, the response to the request comes back as an EDI file, equally cryptic, that describes what went wrong. And we are back to the manuals, counting stars and tildes to check if you send all the data and whether it was in the right format and right order.

It is also worth mentioning that EDI format is used for HIPAA protected medical data, but it has no security built in, everything is plaintext and with the help of manual anyone can read it.  The transmission of the requests, however, is carried out over a secure channel. In my case, I am sending batched requests for processing and the only way to transmit those is by using an SFTP. Needless to say it also becomes my responsibility to pull the response files from the SFTP server once requests are processed, and since there is no strict guarantees on when the processing is completed for each batch, I just do it periodically and eventually collect the responses. How the EDI files are stored on the system side is all up to the engineers designing such system, and as users we can only hope it has passed HIPAA compliance checks and secures the data at rest.

Obviously the archaic standards, like EDI, work well in practice. After all, despite my “brilliant” ideas implemented in other parts of the system, the EDI layer was relatively problem-free. But the standards definitely can be better and new standards, if developed, can provide improved security,  they can be easier to work with, and they will reduce the costs of developing new software. However, I suspect the costs of switching to new standards are just too high for the existing infrastructure, forcing the industry to create a bigger and bigger gap between what government agencies require and what is modern, efficient and secure.

I no longer work for the company, but I still maintain the software. And I was dreading the time a client asks for changes to how the system interacts with Medicaid. It seems like this time is upon me now, and in the summer I will be looking at more manuals for EDI X12 requests needed to implement new features.