Best way to survey patients without leading them toward a specific answer?

Big one is avoiding loaded language and double-barreled questions — instead of 'How easy and reliable was the device?' split those into two separate items. I'd also lean on Likert scales with a neutral midpoint (5 or 7-point) and always pilot with 5-10 patients before full launch to catch any wording that nudges people. Qualtrics has decent survey bias checklist resources if you want a quick reference.