Overview
Before kidney stone surgery with a flexible or semirigid ureteroscope, patients must read and sign a written consent form, but many patients find these forms long and hard to understand. This study tests whether listening to an audio recording of the same consent form, in addition to the usual written form and the surgeon's explanation, helps patients understand the operation better. It also asks whether it matters if the recording is read by a real urologist or produced by artificial intelligence (AI) from a clone of that same urologist's voice.
Adults scheduled for their first ureteroscopy or retrograde intrarenal surgery at three hospitals in Turkey will be randomly assigned to one of three groups: (A) standard written consent only, (B) standard consent plus an audio recording read by a urologist, or (C) standard consent plus the same text produced by an AI-cloned version of the same urologist's voice. All three groups receive exactly the same information, word for word; only the way it is delivered differs. Patients in group C are told that the voice was produced by AI.
The main outcome is a 10-question knowledge test on the content of the consent form, taken immediately after the consent process. Anxiety (State-Trait Anxiety Inventory) and satisfaction are also measured on the same day. No forms are completed on the day of surgery.
Description
Design: multicenter, prospective, parallel-group, three-arm (1:1:1) randomized controlled trial with outcome-assessor and analyst blinding.
Stimuli: The same unified URS/RIRS consent text is used in all arms. Arm B uses a single studio recording read by one urologist (identical at all centers; not the patients' own surgeon). Arm C uses a voice clone of the same urologist generated with Resemble.ai from a separate reference recording (≥5 min, different text), rendered once and frozen (no live AI calls). Both recordings were loudness-matched (-24.0 LUFS), mono, 48 kHz, AAC; duration 716.8 s vs 708.4 s (1.2% difference). Recordings are presented after the standard written consent process and physician explanation are completed, on a tablet with headphones; patients may replay them (replay count recorded) and may ask the physician questions in every arm. Arms B and C are introduced with structurally parallel sentences ("read by a urologist" vs "generated with AI from a urologist's voice").
Randomization: computer-generated, stratified by center (3) and education level (middle school or less / high school or higher), 6 strata, permuted blocks of varying size (block sizes not disclosed to site staff); allocation concealment by sequentially numbered, sealed, opaque envelopes; the person opening envelopes differs from the outcome scorer. Knowledge tests are scored by an assessor blinded to arm; the dataset is transferred to the statistician with masked arm labels.
Hypotheses (fixed-sequence testing, two-sided α=0.05): H1 knowledge score arm C \> arm A; H2 arm B \> arm A (tested only if H1 is significant); H3 arm C vs arm B - descriptive only, adjusted mean difference with 95% CI compared against a pre-specified reference range of ±1.0 point; no equivalence or non-inferiority claim will be made.
Sample size: to detect a 1.0-point difference (SD 2.0, d=0.50) with 85% power, 72 analyzable patients per arm; 80 per arm (240 total) allowing 10% loss. A blinded sample-size re-estimation (pooled SD, arm labels not revealed) is performed once after 120 patients; if pooled SD \>2.3 the sample size is recalculated with the same effect, α and power.
Analysis: ANCOVA for post-consent knowledge score with arm, education, procedure, age and center; secondary ANCOVA for post-consent STAI-S (STAI-S2) adjusted for pre-consent STAI-S (STAI-S1), STAI-T, sex, education, procedure and center. Pre-specified supportive analysis: arm A vs pooled audio arms (B+C). Exploratory: education × arm interaction; physician (consenting doctor) variance. Knowledge-test psychometrics (KR-20, item difficulty, discrimination) are reported from trial data. Reporting follows CONSORT.
Analysis set: the primary analysis is performed in the modified intention-to-treat (mITT) population - all randomized patients analyzed in their allocated arm who have a measured primary outcome. If missing primary outcome data exceed 5%, a sensitivity analysis using multiple imputation is performed; a per-protocol analysis is also reported as sensitivity analysis.
Eligibility
Inclusion Criteria:
- Age 18 years or older
- Scheduled for elective ureteroscopy (URS) or retrograde intrarenal surgery (RIRS)
- Able to read, understand and communicate verbally in Turkish
- Provides written informed consent to participate in the study
Exclusion Criteria:
- Hearing or visual impairment preventing use of the study materials
- Confusion, dementia, psychosis or any other condition affecting capacity to consent
- Emergency surgery
- Previous URS or RIRS
- Current psychiatric treatment or use of anxiolytics
- Health-care professional
