Lessons Learned from the Open Source Antimicrobial Stewardship Project

When I started as an Antimicrobial Stewardship (AMS) pharmacist I did the SIDP Antimicrobial Stewardship Certificate Program (which was fantastic), and read through the Australian Antimicrobial Stewardship in Australian Health Care book which was critical to setting up our AMS program. Both resources emphasized the benefits of using an electronic platform to support AMS activities.

Early Challenges

At that early stage, I had no idea what I was doing. I had to cover three hospitals, all located over 500 kilometres apart, there was constant staff training due to high staff turnover within in the Northern Territory. I had no system to identify which patients to see each day, both at my base hospital in Darwin and the other two hospitals in Katherine and Nhulunbuy.

I spoke to my manager about potential budget for an electronic AMS decision support system. She said there was no hope. It was a fight to get my position in place, with no hope of getting an electronic system in place.

The Excel Era

With the help of some Excel experts online and friends within NT Health, I built out a very convoluted Excel spreadsheet. It used macros to import and combine data from pathology and our e-Meds platform, identifying antibiotic vs medication interactions, drug-bug mismatches, and drug vs renal function mismatches.

It was a bit clunky, but the spreadsheet worked well. Suddenly, we were identifying all sorts of issues across all the hospitals. I used to strut around the wards, homing in on interactions and mismatches, feeling like an AMS genius (while in reality, it was 20 nested IF statements doing all the work).

We saw a large increase in the number of interventions we were recording, and some minor impacts on our antimicrobial usage, but we were still only ever fixing problems after they occurred.

Failed Attempts

Post prescription review and feedback and JMO and registrar training helped a little, but with our staff turnover, and limited availability of education slots we were always fighting a losing battle. We needed something lasting which could help at the point of prescribing.

We tried developing a few hospital procedures on common AMS issues and eagerly tracked their uptake. They were opened only a handful of times (most likely not by any floor staff at all), and we saw no change in practice after publishing them. It became apparent early on that busy hospital staff did not have the time or inclination to open our AMS guidelines.

We tried to make a laminated AMS support card for people to put on their name badge and lanyards to help guide prescribing, but then had that plan dashed by infection control as it was a vector for spreading infection (and rightly so).

We made posters and strategically put them up in drug rooms and doctors rooms, but didn't see much impact there, as the decision on what to prescribe occurs during the ward round, once you are in the drug room or discharge write up room you are too late...

We needed something which could help guide staff during their ward rounds. It had to be fast and integrated into their workflow. It was time to re-visit the digital AMS decision support system.

Revisiting Digital Solutions

I had never actually seen an AMS decision support system, but I had a good idea of what I wanted in my head.

As a hobby, I developed a basic web app based on the Australian Therapeutic Guidelines. It guided clinicians through a series of questions to point them to the right therapy and included AMS tips and dosing calculators for vancomycin and gentamicin. It was very fast to access and get information from. I knew that would be a critical factor.

Once triple checked and released, we saw some impacts. Our historically high ceftriaxone usage went to its lowest point on record, and you could tell on rounds that practice had started shifting.

We also won a bunch of awards, and the emergency department (when assessing all their clinical applications) voted it as the most usable and useful of the clinical decision support tools available in the ED. This really drilled home the fact that we were on to something (ED consultants are a hard crowd to please).

Open Source Initiative

The AMS spreadsheet and pathways were so successful that we decided to release them as open-source tools to support other AMS services like ours. At its peak, 15 hospitals used the digital pathways tool, and several more used the spreadsheet.

Results were even better elsewhere, with one medium sized hospital demonstrating an 18-25% improvement in antimicrobial prescription appropriateness after implementing the tool, with no changes to funding or any additional FTE assigned.

Challenges and Lessons Learned

Over time, the open-source project started to fail. Each site made their own updates to the pathways, causing maintenance issues as staff rotated. The number of active hospitals reduced from 15 to 7.

User feedback indicated a need for more functionality, such as AMS audits, pathway updates without accessing source code, patient prioritization, and a mobile-friendly, cybersecurity-compliant solution.

It was obvious the old HTML/JavaScript web app was not going to cut it. We needed to start again with a new framework and build something truly fit for purpose from the ground up.

It was time to start work on Clinical Branches.


Written by John Shanks - Antimicrobial Stewardship Pharmacist and Software Developer at Kraken Coding