Context
Approximately 15% of patients undergoing vascular surgery are diabetic. These patients face longer hospital stays, a higher incidence of complications, and significant variability in how their insulin is managed during the perioperative period. National guidelines provide clear standards for intravenous insulin infusion management in surgical patients, but formal measurement of adherence to those standards at the ward level remained limited.
I saw an opportunity to quantify that gap and, more importantly, to test whether introducing continuous glucose monitoring into the perioperative workflow could close it.
What I Did
I co-led a two-cycle clinical audit measuring adherence to 11 national guideline standards for perioperative insulin management. The first cycle ran for six months and gave us a baseline. We were doing well in most areas, but there were clear gaps, particularly around glucose monitoring frequency.
Between cycles, we introduced five changes. Most were process fixes: adjusting medication timings, adding routine checks at key handover points, extending insulin overlap windows. But the one that interested me most was the introduction of continuous glucose monitoring sensors for eligible patients.
CGM is well established in outpatient diabetic care. It is not standard in surgical inpatient settings. I wanted to see whether giving clinical staff access to real-time glucose trend data, rather than isolated finger-prick readings every few hours, would change anything.
What Changed
The second cycle, covering 45 consecutive patients over six months, showed improvement across nearly every domain. Compliance rose from 8 to 9 out of 11 domains at the good threshold.
But the more interesting finding was behavioural. CGM users showed meaningfully higher monitoring compliance than non-CGM users. The technology did not just produce better data. It changed how the team interacted with the data. Nurses started reading trend arrows rather than waiting for threshold breaches. Consultants began reviewing glucose trajectories during ward rounds rather than reacting to isolated spikes.
The shift was from reactive to anticipatory. That mattered more than the percentage improvement.
Why This Stuck With Me
This project sits at the start of a thread that runs through most of what I have built since. The sensor technology was not new. The guidelines were not new. What was new was applying an existing tool in a context where it had not been tried, measuring the result properly, and showing that the real value was not in the device itself but in the workflow change it enabled.
That pattern, taking proven technology and finding the right clinical context for it, is what I keep coming back to. It is the same instinct behind CarePass, behind DischargeKit, and behind most of the health tech work I care about. The technology is rarely the hard part. Adoption is.