Key takeaways
- Hospital-acquired infections are a serious risk. One common infection, CLABSI, can cost hospitals up to $65,000 per case.
- Technology can drive down infection rates. Using AI to surface critical data can reduce medical errors and improve patient outcomes.
- Agentic AI has vast untapped potential in healthcare. AI agents can perform numerous routine tasks, freeing up time.
By Christine Landry, Global Healthcare Lead, Kyndryl Consult
Intensive care units in hospitals can be stressful, hectic places. In a medical emergency, all nurses and other healthcare professionals on shift may need to dive in to help a patient, which means putting other tasks, like cleaning catheters or changing dressings, on hold. Ideally, they will return and finish the job after the emergency subsides. But if there is a shift change, the next care team may not be aware that a scheduled cleaning was skipped, increasing the rate of infections.
Infections acquired in hospitals can be quite serious — jeopardizing both patient health and the reputations of healthcare facilities, which can face litigation and incur financial penalties from regulators. Fortunately, technology can help.
Recently, a large hospital in Texas struggled with high rates of infections related to central venous catheters. While standard intravenous catheters, or IVs, are inserted into our hands or arms and removed within hours or days. Central venous catheters, or central lines, are placed in a large vein in the neck, chest or groin to deliver drugs such as chemotherapy or to draw blood. Central venous catheters can remain in place for weeks or months.
A central line bloodstream-associated infection, or CLABSI, develops when a virus or bacteria is introduced to the patient’s bloodstream through the central line. That can happen either during the initial insertion of the line, when it's being used to administer drugs, or from the dressing covering the line. Because patients with central lines are already ill, it can be difficult to catch CLABSIs, and they can result in serious diseases like sepsis and infective endocarditis. Despite major efforts to reduce its spread, about 30,000 cases of CLABSI occur annually in the United States.
30,000
There are about 30,000 cases of CLABSI annually in the United States.
$65,000
A single case of CLABSI can extend a patient’s ICU stay by one to two weeks and incur additional hospital costs of up to $65,000.
$1 million
If a hospital is found negligent, it may be required to pay malpractice settlements exceeding $1 million each.
A case of CLABSI can increase a patient’s time in the ICU by one or two weeks and cost a hospital as much as $65,000, according to JAMA. And because the illness is regarded as preventable, the hospital is prohibited from billing the patient or Medicare for the additional costs. Hospitals with high rates of CLABSI can face additional financial penalties from Medicare, and if found negligent, may pay malpractice settlements exceeding $1 million each.
Hospitals have rigorous procedures to prevent CLABSI, with a set schedule for changing dressings and flushing IV lines. They also require nurses to clean the access point, or hub, whenever a new IV bag is hung (hence the mantra “scrub the hub”). For some patients in intensive care who require several IV bags every hour, a medical team may need to clean the hub 15 to 20 times per shift. When an emergency derails that schedule, errors can creep in.
The next step in improving patient safety is to introduce agentic AI into the process. Instead of running a data tool on a set schedule, an AI agent can continuously examine the records of every patient and compare them to the hospital’s policies and treatment schedules.
While quality improvement initiatives have decreased the incidence and costs of health care-associated infections (HAI), much more remains to be done. Here is where technology can help. Trusted technology partners can work with hospital systems to develop compliance reporting systems that draw data from the hospital’s electronic medical charting system and surface it so the shift manager is alerted to gaps in cleaning schedules. With this system in place, the Texas hospital drove down infection rates from 1.32 per 1,000 patients to .55 — a 60% reduction, according to Kyndryl data.
The next step in improving patient safety is to introduce agentic AI into the process. Instead of running a data tool on a set schedule, an AI agent can continuously examine the records of every patient and compare them to the hospital’s policies and treatment schedules. Oversights can be identified and rectified immediately.
Agents can also analyze dozens of risk factors that could lead to CLABSI — such as the patient’s history, the age of the IV equipment, and biomarkers revealed in blood work — and predict which patients are at the highest risk of infection. Their charts could then be flagged for special monitoring or increased preventative care.
More advanced systems can include sensors, such as cameras, that autonomously detect whether a healthcare worker is performing the scheduled hygiene tasks. If the agent observes that a nurse has not cleaned the hub, it can send a reminder to their mobile device or trigger a reminder at the patient’s bedside. The system can also be programmed to only record silhouettes to preserve the privacy of the patient and caregiver.
AI is already proving itself useful for streamlining the dozens of administrative tasks that can absorb the attention of healthcare professionals and keep them from spending time with patients. AI can perform routine paperwork functions, retrieve medical records and transcribe doctor-patient conversations. Advanced uses include analyzing patient test results and diagnosing disease. Healthcare professionals are enthusiastic about its potential, with 78% saying it will expand their capacity to serve more patients, according to Philips.
But like any AI system, it requires human oversight and judgment. Patients, many of whom are in vulnerable conditions, are justifiably concerned about the role technology plays in their care. The same Philips report found just 59% of patients are confident AI can improve healthcare. It’s incumbent on healthcare systems to introduce systems deliberately to help ensure all members of the care team understand what AI can and can’t do, and to explain to patients the advantages of automation. None of it works, of course, without a baseline commitment to patient health and satisfaction, and that always starts with the humans who provide the care.
