Developing for Healthcare - .NET vs J2EE?

An anonymous reader asks: "Our small southern shop (an eleven man team) is about to commence development on some medical software geared for physician's offices and hospitals. Since we have never developed in this area before (our primary source of income comes from developing software for regional transportation offices of the government) we are at loss for the reigning technologies. The two technologies we are considering are J2EE and .NET. What are the opinions of the Slashdot crowd? Surely others have developed for the monstrous healthcare industry. Thanks!" "My senior manager recommended using .NET. His argument is that most desktops he has seen in hospitals already run Windows, the development time will be cut down for this small to mid-size project, rich desktop clients are possible and there will be no application server costs involved. He also contends that .NET has more templates and abilities than J2EE (which is simply 'web targeted' in his opinion.

Half of my coworkers are with him and the other half have suggested using J2EE due to portability--we do not want to cut off any potential sources of income with an already dwindling future. Has GNU/Linux become widespread in the healthcare industry that we should consider developing for it too? What about Mac OS X server?

The last problem we have is winning over the IT staff hearts. They are the ones who ultimately give the go ahead to purchase the software. Java gets a bad rap for being slow, while Microsoft (and by extension) .NET has the shadow of being insecure. How can you possibly win?"

Comments

J2EE and webapps

[ip_fired]

In my opinion, I think that webapps are the way to go with a project like this. I say this because we currently have an old legacy application that uses Borland C++ Builder, and it is very difficult to manage the all of the versions being used by staff members. As we've moved the functionality over to a web portal, we've been able to spend less time distributing executables and more time developing features.

A webapp has the strength that all you need is a web browser to view the content. When you update the webapp, all clients are updated instantly, without having to push something out to them, or making them download something. This saves a lot of stress and headache.

There are many technologies out there for writing a database oriented project like this a lot easier. These include

• Hibernate for object relational mapping and cross compatibility with most major databases (ie, develop on MySQL, deploy on Oracle, Informix, whatever you want).
• Spring which manages your mappings and helps maintain consistency across your data connections and helps you abstract your business logic, keeping it out of the actual pages. It also integrates with . . .
• Struts which gives you a great Model-View-Controller framework to practice good development and good security.

Apache Tomcat, JBoss, and many of the other java based application servers are supported on many different platforms. You can run it on Windows, Linux, Mac OSX, and virtually any hardware provider you want.

I've been working now on this project for 6 months, and I have to say, I love the structure of this way of working on webapps far more than the common hodge-podge of php or perl. This isn't to say you can't program a nice, easy to maintain app in these languages, just that there is a nice framework with examples and plenty of books to do it in Java.

well...

[domenic v1.0]

Both .NET and J2EE are good platforms for developing and hosting business applications. Both support n-tier architectures via client- and server-side component models for assembling enterprise applications. This allows for use of either fat or thin user interfaces with both platforms. However, .NET and J2EE are far from identical, and each platform has distinct strengths.

The primary strength of .NET is its use of multiple programming languages running on a single platform. This eliminates the programming language as a barrier for adoption. Further .NET strengths include ease of use and speed of development as programmers might be transitioning from COBOL or C. (In contrast, transitioning to Java usually takes greater skill in object orientation.)

J2EE takes .NET's multiple programming-language/ single-platform paradigm and turns it around: The primary strength of J2EE is its use of a single programming language capable of running on multiple platforms. This eliminates the platform as a barrier for adoption. Further J2EE strengths include vendor neutrality and the active involvement of a large, open-source community.

Legacy applications: Can either platform make integration easier?
Healthcare providers are moving from traditional reliance on paper-based records and isolated legacy systems. Most now embraced more systematic electronic exchange of patient data, exchanged both within and across organizations such as other providers, diagnostic and laboratory test facilities, managed care organizations, etc. Creating and maintaining patient records requires the ability to extract and integrate patient data from a range of "legacy" document and system formats. legacy applications constitute an even more difficult integration category. Consider a host application that was developed in-house 20 years ago, and written in COBOL. No J2CA resource adapters are available, nor will they be available, to help integrate these applications in a J2EE environment.

From an integration standpoint, JDBC is actually more promising than J2CA. This API provides access to virtually any data source, from relational databases to spreadsheets and flat files. With a JDBC driver, all corporate data can be connected, even in a heterogeneous environment. In addition to its support for actual relational databases, JDBC can also support emulated relational models based on legacy information sources. But to do this, JDBC requires an integration product that can map the legacy-application functions to emulate a relational database model.

The .NET platform, with its dependence on Microsoft BizTalk Server, has the same drawbacks for legacy-application integration as it does for packaged-application integration. So, despite the very real integration potential of .NET and J2EE, both platforms have their associated limitations. And when it comes to legacy-application integration, neither platform can complete the job on its own.

Be aware that all vendors of application integration products do not provide equal support for .NET and J2EE. Many vendors have built their solutions exclusively around one platform or the other, heedless of the fact that both .NET and J2EE will still be the strongest presence in the years to come. Buying an integration product too hastily can tie your organization to a development platform they might not have chosen otherwise.

Therefore, you should look to application-integration vendors that are closely aligned with--and can articulate a long-term strategic relationship with--both .NET and J2EE technologies. That means checking to ensure that a solution offers interfaces such as .NET Class Libraries, COM Objects, ASP, ASP.NET, and .NET web services to support your .NET-based development efforts as well as interfaces like JavaBeans, JSP, and Java web services to support your J2EE-based development efforts.

Coming from the healthcare industry...

[databyte]

I work for a company that just landed a large install to a large hospital system. Here are some points to consider.

My experience...

1) Don't worry about the run-time in terms of a desktop requirement. Since you'll have to install your software, you'll have the opportunity to install the VM.

2) Getting security clearance to run on a hospital system will largely be dependent on your application architecture and not on your application framework. Do you depend on certain ports or services, certain databases, file shares, web services, etc.

3) A majority (overwhelming majority) of these systems do run Windows. If you find Linux in the mix, it'll usually be for very specific systems (PAX/DICOM) or back-end work (your mainframes and such). Due to the large number of small vendors in healthcare, a ton of applications are installed to the clinican's desktop. (And may have simple architectures or older frameworks like VB, Delphi, and even Terminal/Telnet). Everything from intergrated hospital wide systems to what a doctor uses on his own box to manage X, Y, and Z. [hint: look-up CCOW]

4) Integration with an EMR is important, and again, most are Windows based. If you want to feed data into "the system", be prepared to work with the system. EMRs are heavy on HL7, light on Web Services and XML.

5) Depending on your architecture, your GUI/interface and your "server" can be running on different systems, platforms, and/or frameworks. Probably not the best route but it is possible.

General knowledge and opinion...

1) Mono (.NET) for non-Windows applications is just as viable as Java for multi-platform use. You could do .NET AND Mono deployments (or just pick one).

2) Hospitals pay huge $$$ for integrated systems. As such, they will purchase hardware dedicated to your implementation. If your worried about running on their existing OSX servers, don't be - sell them a Dell running Windows 2003 or a black box running RH. Your software will be more expensive than the boxes. You could also just include $2-10K in hardware costs as part of the PO.

I highly recommend going to a trade-show or just talking to vendors "as a 100-300 bed hospital". You can see what others do by being a customer.

Not saying you have to do what everyone else does - but it does help to sell something that won't take a lot to get going.

Hospital Technologies

[Anonymous Coward]

FYI, here is a list of common technologies you will see floating around a typical community hospital:

operating systems
digital HP Unix
AIX
VAX/VMS
Solaris
Windows 3.1, 95, 98
Windows NT, 2000, XP

programming languages
Mumps (ala the 'M' database)
CCL (fourth generation SQL-esque language)
SQL (a very little bit)

databases
oracle, oracle, oracle
some SQL

First of all, you have got to understand your (potential) client. Therefore, you need to understand FDA regulation of healthcare equipment, and of the HIPPA regulations. Because of FDA regulations regarding healthcare equipment, it often takes years (between 3 and 5, on average) for a piece of medical equipment to be designed, tested, run through trials, and (hopefully) approved. On the other hand, hospitals have a very steady stream of revenue, charge lots of money, generally have massive amounts of cash flow, and have major assets (like CT and MRI scanners, for instance). Therefore, it's common for a hospital to invest $1M for a new CT or MRI scanner at least every 10 years. However, that scanner had better be a workhorse for the hospital.

So, what this means for you as a developer is that most hospitals still have legacy systems in production that date back 10 to 20 years. As a systems administrator for the department of radiology at a local community hospital, my production environment involves:

> a Maxifile VMS/VMX terminal server running an 'M' database programmed in 'MUMPS'
> 6 oracle database servers which 100+ non-admin end-users log onto each day
> a remote-hosted oracle database which sits in Kansas (ala Cerner Corporation), which everybody at the hospital must access via a Citrix box.
> a nuclear magnetic resonance imaging scanner running VAX/VMS
etc. etc.

So, to make a long story short, if you want to develop products for the healthcare industry, you had better start thinking about how to interface with products from Oracle, Citrix, Cisco, Microsoft, and so forth.

Also, Linux is generally avoided in hospital settings. It requires too much computer knowledge (remember that hospital workers are healthcare providers, not information technology developers). More importantly, it's usually impossible to hold the developers accountable with Linux, and it's generally very difficult to obtain support contracts. All things considered, I have 1 linux machine in production, 200+ windows machines, and about 30 unix variations (AIX, VAX, HP Unix, DEC, Solaris).

So, all this being said, I think that you're barking up the wrong tree and not necessarily asking the right questions. Personally, I would suggest J2EE over .NET.

If you really want to start asking the right kind of questions, you could start with questions like 'should we start by developing an HL7 sniffer or a HIPPA auditing & compliancy application'? You would then find, either way, you will need to know Cisco (TCP/IP stacks), Oracle (SQL database connectivity), and HIPPA legal concerns (e.g. need for cryptography). In the end, you'll realize that you can do it in either J2EE or .NET, and that the question is kind of a moot point.

anyhow, just my $0.02