A nonlinear relationship of evoked responses following charge-balanced single-pulse electrical stimulation with varying pulse widths.

Ulcerative colitis (UC) is a chronic inflammatory bowel disease, characterized by cycles of active symptoms followed by periods of remission. Although many patients manage their UC with medication, approximately 20% to 30% of patients will require surgical intervention, in which all or part of their bowel is removed. When medical management fails, patients face a complex and preference-sensitive decision involving trade-offs among various surgical options. These include the option of having reconstructive surgery, such as ileal pouch-anal anastomosis, ileorectal anastomosis, or procedures resulting in a stoma, such as subtotal colectomy or proctocolectomy. Discrete choice experiments (DCEs), a stated preference method, are widely used in the health literature to elicit preferences for treatments. A key advantage of DCEs over other methods, such as time trade-off, is the ability to model data that account for both observed and unobserved heterogeneity across multiple attributes simultaneously. This allows for an exploration of preferences beyond the population average, enabling the estimation of individual or group-level preferences to better understand how different patients make treatment decisions. The surgical decision for UC is preference-sensitive because no single procedure is universally superior for all patients. The choice depends on a patient’s values and treatment history, requiring them to weigh competing treatment factors, such as symptom resolution against drawbacks such as loss of natural bowel function, lifestyle adjustments, and complication risks. Considerable variability exists in how patients trade off these potential advantages and disadvantages., , , For example, some patients may have an overwhelming aversion to a permanent stoma, whereas others may prioritize minimizing bowel frequency. Because some of this preference heterogeneity may be unrelated to observable personal characteristics, ignoring this individual variability can bias model estimates.