Cognitive Sovereignty Self-Audit for Pharmacists

1. When Lexi-Interact or Epic AI flags a drug interaction, how do you decide whether to contact the prescriber?

I contact the prescriber whenever the system shows a flag. The alert tells me the interaction severity. I usually contact the prescriber for medium and high severity alerts, but I skip low-severity ones the system marks routine. I check the patient's renal function, age, and current dosing before deciding. Sometimes I find the interaction matters less than the alert suggests. I assess the clinical context first, then use the alert as a starting point. I routinely override low-risk alerts and occasionally challenge high-severity ones based on patient factors.

2. A patient comes to collect their prescription. The dispense decision support in your system approves it. How do you verify it is safe to hand over?

The system approved it, so I scan the barcode and dispense. That check is done. I glance at the patient profile for obvious problems, then dispense if nothing jumps out. I check dose, frequency, and indication against the patient's age and weight. I verify the system recommendations match my own assessment most of the time. I independently check dose, indication, and patient factors before looking at the system recommendation. I treat the system output as a prompt to think harder, not as a final answer.

3. When counselling a patient on a new medication, what guides what you tell them?

I use the AI-generated counselling text from my dispensing system or ChatGPT. It covers the main points. I follow the system suggestions but add a sentence or two about their specific condition if I remember to. I ask what the patient already knows and what worries them, then use that to shape my counselling. I reference the system guidance to fill gaps. I listen to the patient's concerns first, then counsel based on what matters to them. The AI suggestions are background research only if I need them.

4. You notice your team is dismissing low-level alerts from your interaction checker without reading them. How do you respond?

Alert fatigue is normal. As long as serious cases are caught, occasional misses are acceptable. You mention it in a team meeting but do not investigate whether real interactions are being missed. You audit a sample of dismissed alerts over two weeks to see if any represent real clinical risks. You discuss findings with the team. You run a formal review of alert patterns, identify which types are genuinely low-risk, and recalibrate your team's response thresholds. You track whether dismissal behaviour changes.

5. A clinical decision support system recommends a different dose than the one prescribed. The patient has mild kidney disease. What do you do?

The system knows dosing rules. I follow its recommendation and contact the prescriber to change it. I look up the patient's estimated glomerular filtration rate and check whether the system's suggestion matches standard guidance. I calculate expected drug clearance using the patient's kidney function and other factors. I decide independently whether the system is right. I work through the pharmacokinetics myself without consulting the system first. I only use the system to double-check my own conclusion.

6. Over the past year, have you had to make a difficult dispensing decision where the AI system did not guide you?

No. The systems handle nearly all scenarios I encounter. I rarely need to think beyond what they show me. Rarely. Maybe once a month I face a situation the system does not cover well. Regularly. A few times a month I encounter prescriptions where clinical context matters more than the system logic. Frequently. I take independent action on complex cases at least weekly. The systems help, but my judgement is the final step.

7. When you counsel a patient on side effects, how much do you customise the message to their specific situation?

I read the system or AI summary of common side effects. That is what most patients need to know. I mention the most common side effects and add one or two that apply to the patient's age or conditions. I ask what side effects concern them most, then focus on those. I explain which are common for their age and kidney function. I assess their risk profile first, then counsel on the side effects most likely for this patient. I tell them what to watch for and when to call.

8. If your AI systems were suddenly unavailable for a day, how confident would you be in your dispensing judgement?

Very uncomfortable. I would make fewer decisions and refer most cases to others. I could manage routine scripts but would hesitate on anything complex. I would manage well. I might take longer on difficult cases but would feel confident in my decisions. I would work normally. The systems speed me up, but my clinical knowledge is the real foundation.

Cognitive Sovereignty Self-Audit for Pharmacists

Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You