Not a knowledge problem, a system problem: why we keep prescribing Z-drugs.
The evidence for CBT-i is clear. The prescriptions keep coming. The fault isn't with clinicians but with the system they're asked to operate in.
A large cohort study published in The BMJ offered a sobering look at the downstream consequences of benzodiazepine and Z-drug prescribing in pregnancy and specifically, elevated risk of psychiatric disorders in children. The science is meticulous. The clinical implication is serious. Yet the framing of the problem in most commentary was familiar and, I'd argue, incomplete: clinicians aren't following the guidelines.
We've known for decades that Cognitive Behavioural Therapy for Insomnia (CBT-i) is the gold-standard intervention for chronic insomnia. We know that Z-drugs treat the symptom rather than the condition. We have further evidence, that the long-term consequences of chronic sedative use can extend to the next generation. However, Z-drugs remain overprescribed globally.
First-order thinking and second-order consequences
Shane Parrish describes first-order thinking as taking the action that solves the immediate problem without examining what it produces downstream. A patient presents in distress, sleepless, exhausted. A prescription ends the consultation. The first-order effect is relief, for the patient and the clinician alike. The second-order effect, dependency, tolerance, the hardening of an acute problem into a chronic one, arrives later, and often invisibly.
Treating chronic insomnia with Z-drugs is like applying a sticking plaster to a fracture. The surface is covered. The underlying pathology of sleep-related anxiety continues. Conditioned arousal, the learned association between bed and wakefulness, these are the mechanisms driving chronic insomnia, and a sedative addresses none of them. What it does do is arrive quickly, require no training, and fit inside a stressed ten-minute appointment.
The system produces the behaviour
We tend to frame overprescribing as a knowledge problem. If clinicians understood CBT-i better, they'd use it more. I don't think this is true. Most GPs understand perfectly well that CBT-i is superior and that Z-drugs are harmful in the long run. The barrier is not knowledge but it is that the system makes prescribing the default and therapy the exception.
CBT-i requires multiple sessions, a trained therapist, and a patient with the motivation and capacity to do difficult psychological work. Digital CBTi is patchy in uptake, availability and efficacy. In a system rationing every minute, recommending CBT-i can feel like pointing someone toward a bus that doesn't run. The prescription is not a failure of clinical knowledge. It is a rational response to an irrational system.
The transition from acute to chronic insomnia is frequently preventable. Prevention requires something in genuinely short supply: continuity. When a patient is seen each time by a different clinician, there is no one tracking the trajectory. No one noticing the repeat prescription that began as a crisis measure six months ago has become routine.
Deprescribing is a relational act
This is the part of the problem I rarely see discussed. Deprescribing Z-drugs is not simply a mechanical act of tapering a dose or substituting a formulation, but it is a relational one. Guiding a patient away from the perceived quick fix of a sedative and toward the harder work of CBT-i requires clinical capital: accumulated trust, knowledge of the patient's history, stressors, temperament, and comorbidities.
A GP who has known and been involved in the life of a patient for five years has that capital. Continuity is not a luxury feature of general practice. It is the mechanism by which complex behavioural change becomes possible.
If we want to reduce Z-drug dependency and implement gold-standard sleep medicine at scale, we must address the system that makes prescribing rational. The evidence from Cho et al. tells us what to do medically. The question of how we redesign primary care to make continuity and CBT-i delivery genuinely possible is harder, and it is the one we have not yet seriously answered.
This essay is developed from a rapid response published in The BMJ, May 2026, in response to: Cho et al. Benzodiazepine or Z-hypnotic use during pregnancy and risk of psychiatric disorders in children. BMJ. 2025.View original rapid response →
