Definition and evaluation of the monotonicity condition for preference-based instruments

Area of Science:

• Health Services Research
• Biostatistics
• Epidemiology

Background:

• Preference-based instrumental variable (IV) methods are common in comparative effectiveness research.
• These methods often assume monotonicity (no "defiers") to estimate local average treatment effects (LATE).
• The empirical validity and precise meaning of the monotonicity assumption remain unclear.

Purpose of the Study:

• To clarify the definitions of local and global monotonicity.
• To propose and illustrate a novel study design for empirically assessing the monotonicity assumption.
• To evaluate the impact of monotonicity violations on IV estimates in antipsychotic treatment.

Main Methods:

• Clarified local and global monotonicity definitions.
• Developed a novel survey design involving physicians' treatment plans and preferences for hypothetical patients.
• Conducted a pilot study surveying 53 physicians regarding antipsychotic treatment decisions.

Main Results:

• The pilot study found widespread violations of the monotonicity assumption across nearly all patients.
• Patients could not be reliably classified into standard compliance groups (compliers, defiers, always-takers, never-takers).
• This indicates significant deviations from the assumptions underlying many IV analyses.

Conclusions:

• Preference-based IV estimates require cautious interpretation due to likely bias from monotonicity violations.
• The specific subpopulation (e.g., compliers) for LATE estimates may be poorly defined.
• Supplementing IV studies with empirical surveys to assess monotonicity bias is recommended.