Table 3.

Thematic analysis of the feedback on the tool. Key themes revealed by analysing qualitative feedback from anaesthesiologist questionnaires and interviews.

Theme 1: hypotension prediction as a component of holistic postoperative risk assessment: opportunity for prospective care planning and reflection on intraoperative care
A	‘Perhaps if the model predicts a high risk of PACU hypotension, that suggests I should have managed the anaesthetic differently for this patient, and the model could help me recognise that’.
B	‘Potential uses of this risk model: (i) education of perioperative providers and improved awareness; (ii) potential strategies to prevent PACU hypotension, which may reduce PACU time and discharge to floor; and (iii) prevention of PACU hypotension with potential to prevent adverse events related to effects of hypotension on end-organ function’.
C	‘It would typically take a lot more to be done before we could switch them to an ICU disposition, so the hypotension predictor alone would not move the dial to start talking about getting an ICU bed. It would have to be hypotension plus something else’.
Theme 2: hypotension risk prediction was of perceived value for PACU staff and handover discussion
A	‘I would communicate that this patient is at increased risk and ask them to have a heightened awareness of that possibility and communicate with me if they are starting to see some degree of hypotension. It is reasonable to consider making sure that there is a fluid bolus option in the postoperative orders assuming that is a reasonable first step to treating hypotension. Asking the nurse to keep a closer eye and let me know if it trends downward’.
B	‘Of course, the resource allocation in the PACU can be challenging. If the nurses knew [the risk], then they would need 1–1, dedicated, frequent monitoring of vital signs and assessment. Sometimes, we do that but not with any risk profile in mind. The score can help take it to the next step and define them better (e.g. check complete blood count, etc. Would help us be more proactive; minimise the AUC for blood pressure’.
C	‘If everyone had similar levels of confidence in the model, that would have some value. Some PACU if they have been told that they have been assigned this patient coming out of operating theatre and if they have had time, they will have read over the chart and may have already brought a higher level of preparation to what they are doing. Like everywhere else, there are variations in ability and practice amongst PACU nurses so it might be most helpful for less experienced PACU nurses or nurses that did not have time to read the whole chart and think “oh yeah, when I see these patients, then I typically have to do this” or “I am expecting a longer PACU stay”.
Theme 3: model has greater value when model risk score is unexpected
A	‘It could help with identifying cases that might have an unrecognised high risk for PACU hypotension that I did not identify as being high risk’.
Theme 4: top-ranked feature list insufficient for clinicians to understand discrepancy with model
A	‘It made me—when I was interpreting the indicators suggested a certain amount of weight—I would look at those and it still did not get me closer to confidence one way or the other, so I was not sure whether to throw out my estimate entirely or just trust the model. I suppose if I had enough info about how the model was generated and how deep the data were that were on it—so it kind of just left me just wondering when it was different from my prediction or experience was—who was wrong—I was not convinced I was entirely wrong and I was not convinced the model was entirely right’.
B	‘Would need to see the data and peer review’
Theme 5: use of standard threshold labels may be overly simplistic
A	‘Whilst the risk score is at times helpful, it seems to overpredict minor hypotension in teens and younger women who have normal BP for these age groups, so no treatment would be necessary’.