PSQI — Pittsburgh Sleep Quality Index

The PSQI is the dominant clinical measure of sleep quality. You send it when sleep is part of the formulation — depression with insomnia, generalized anxiety with sleep onset trouble, trauma-related nightmares, CBT-I outcome tracking, or a baseline against ISI for monitoring change. Clients answer 19 items about their sleep over the past month. The original PSQI also includes 5 partner-observer items (snoring, leg movements, apnea, confusion) — Rivet uses the 19-item self-report subset, which is the dominant configuration in clinical practice.

​

What it measures

Sleep quality and disturbance across seven canonical components:

• Subjective sleep quality
• Sleep latency (how long to fall asleep)
• Sleep duration
• Habitual sleep efficiency (time asleep ÷ time in bed)
• Sleep disturbances (waking up, pain, bad dreams, breathing problems)
• Use of sleeping medication
• Daytime dysfunction (staying awake, low motivation)

The canonical algorithm rolls these seven component scores (each 0–3) into a Global PSQI ranging 0–21.

​

When to send it

• Intake for any client where sleep is in the picture
• Monthly or every-6-week monitoring during CBT-I or behavioural sleep work
• As a screening trigger — a poor Global score is the standard prompt for ISI follow-up or sleep-specialist referral
• Pre/post for any intervention where sleep is the primary outcome

The 19 self-report items take 5–10 minutes. Most clients fill it out async in the day or two before session — send it ahead and the result is on the client’s record when they arrive.

​

Scoring and severity cutoffs

Three components are live-scored in the app: The remaining four components (sleep latency, sleep duration, sleep efficiency, daytime dysfunction) and the Global PSQI score require arithmetic that Rivet’s scoring engine doesn’t yet express live in-session — banding continuous values, ratios with banding, chained band-then-sum-then- band sequences. The structured answers are still captured in full; you copy them out for note-taking or compute the Global score against the canonical scoring sheet when you need it.

​

Cutoffs

• Global PSQI > 5 → poor sleeper (Buysse 1989: sensitivity 89.6%, specificity 86.5%)
• Global PSQI > 8 → severe insomnia (Backhaus et al. 2002)

Item 5h (“have bad dreams”) is flagged sensitive in the template so the intake panel keeps the trauma-informed framing on the rest of the form.

​

PSQI vs ISI

The PSQI describes sleep; the ISI grades insomnia. Nineteen items across a month covering quality, how long it takes to fall asleep, how long the client sleeps, how much of their time in bed is spent asleep, what disturbs them, medication and daytime effects — that breadth is why it’s the instrument for characterising a sleep complaint you don’t yet understand, and why it picks up problems that aren’t insomnia at all. The ISI is seven items over two weeks and produces one number, which makes it the easier measure to repeat during treatment. A PSQI above 5 is a sensible trigger to follow up with the ISI or refer on. The practical cost of the PSQI is scoring. Three of its seven components compute in session; the rest, and the global score, need the published scoring sheet — the response exports in full so you have everything you need to do that.

​

Is it free to use?

Yes. The University of Pittsburgh Sleep Research Group permits free clinical, educational and non-commercial research use.

​

Citation

Buysse, D. J., Reynolds, C. F., Monk, T. H., Berman, S. R., & Kupfer, D. J. (1989). The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Research, 28(2), 193–213. Free for clinical, educational, and non-commercial research use per the University of Pittsburgh Sleep Research Group.

​

When not to use it

PSQI asks about the past month — it’s not the right tool for acute sleep disruption in a recent crisis. Use ISI when you want a 2-week window or a faster outcome measure during active CBT-I. Use both together when sleep is the primary treatment target and you want a longitudinal-quality and severity pair.

​

Related articles