PSQI Scoring: How to Calculate and Interpret the Results

The Pittsburgh Sleep Quality Index (PSQI) produces a single global score ranging from 0 to 21, where lower numbers mean better sleep and a score above 5 traditionally flags poor sleep quality. Developed in the late 1980s, the questionnaire asks you to reflect on your sleep over the past month and funnels your answers into seven component scores, each rated 0 to 3, which are then added together.¹ The math is straightforward once you know the rules for each component, but a few steps trip people up. Here is how the whole process works and what your results actually mean.

The Seven Components

The PSQI is built around 19 self-rated items (plus five questions for a bed partner, which are not scored). Those 19 items map onto seven component scores. Each component gets a score of 0, 1, 2, or 3, with 0 representing no difficulty and 3 representing severe difficulty. The seven components are:

Subjective sleep quality: your own rating of how well you slept overall during the past month.
Sleep latency: how long it takes you to fall asleep.
Sleep duration: how many hours of actual sleep you got per night.
Habitual sleep efficiency: the percentage of time in bed that you spent sleeping.
Sleep disturbances: how often things like pain, breathing problems, or nightmares woke you up.
Use of sleeping medication: how frequently you took something to help you sleep.
Daytime dysfunction: how much trouble you had staying awake or staying motivated during the day.

Some components pull from a single question. Others combine two or more items. The scoring for most of them follows intuitive thresholds, but a couple require a bit of arithmetic before you can assign the 0-to-3 rating.

How to Calculate Each Component Score

If you are scoring the PSQI by hand, the first thing to know is that different components use different items and sometimes different scoring rules. Here is a walkthrough of each one.

Subjective Sleep Quality

This is the simplest component. Question 9 asks you to rate your overall sleep quality during the past month as “very good,” “fairly good,” “fairly bad,” or “very bad.” Those four responses map directly to scores of 0, 1, 2, and 3.

Sleep Latency

This component combines two items. Question 2 asks how many minutes it took you to fall asleep each night, and Question 5a asks how often you couldn’t get to sleep within 30 minutes. For Question 2, the raw minutes get converted: 15 minutes or fewer scores 0, 16 to 30 minutes scores 1, 31 to 60 minutes scores 2, and more than 60 minutes scores 3. Question 5a is already on a 0-to-3 frequency scale (“not during the past month” through “three or more times a week”). You add those two sub-scores together, and then the sum gets recoded: a sum of 0 becomes a component score of 0, sums of 1 or 2 become 1, sums of 3 or 4 become 2, and sums of 5 or 6 become 3. This two-step process is the part that most often causes errors during hand scoring.

Sleep Duration

Question 4 asks how many hours of actual sleep you got per night. The scoring thresholds are: more than 7 hours scores 0, 6 to 7 hours scores 1, 5 to less than 6 hours scores 2, and less than 5 hours scores 3.

Habitual Sleep Efficiency

This is the component that requires you to do a little math. Sleep efficiency is the ratio of hours you actually slept to the total number of hours you spent in bed. You calculate it using Questions 1 (usual bedtime), 3 (usual wake time), and 4 (hours of actual sleep). First, figure out how many hours you spent in bed by subtracting your bedtime from your wake time. Then divide your actual sleep hours (Question 4) by that number and multiply by 100. The thresholds: 85% or higher scores 0, 75% to 84% scores 1, 65% to 74% scores 2, and below 65% scores 3.

Sleep Disturbances

Questions 5b through 5j cover nine potential reasons you might wake up during the night, from getting up to use the bathroom to having bad dreams. Each sub-item is scored on a 0-to-3 frequency scale. You sum the nine sub-item scores, then recode that sum: 0 stays 0, 1 through 9 becomes 1, 10 through 18 becomes 2, and 19 through 27 becomes 3. In practice, almost everyone lands in the 1 range here because nearly everyone reports at least occasional nighttime awakenings, so this component rarely drives much variation in the global score.

Use of Sleeping Medication

Question 6 asks how often you took prescribed or over-the-counter medication to help you sleep. The frequency responses (“not during the past month” through “three or more times a week”) map directly to 0, 1, 2, and 3.

Daytime Dysfunction

This component combines Question 7 (trouble staying awake while driving, eating, or engaging in social activity) and Question 8 (trouble keeping up enthusiasm to get things done). Both are on 0-to-3 frequency scales. Their scores are summed, and the sum is recoded the same way as sleep latency: 0 stays 0, 1 or 2 becomes 1, 3 or 4 becomes 2, 5 or 6 becomes 3.

Adding Up the Global Score

Once you have all seven component scores, you simply add them together. The minimum possible global score is 0 (no sleep difficulty on any dimension) and the maximum is 21 (severe difficulty on every dimension). No weighting is applied: each component contributes equally. That equal weighting is one of the design decisions researchers have debated, since a person who takes sleeping medication every night but otherwise sleeps well would score the same on that component as someone who wakes up multiple times and has poor efficiency but never takes medication. Still, the global score has proven to be a robust general indicator of sleep quality across many populations.²

The Standard Cutoff and When It Shifts

In the original 1989 paper, a global PSQI score above 5 distinguished “poor sleepers” from “good sleepers” with high sensitivity and specificity in a sample of clinical and control participants.¹ That cutoff of greater than 5 has become the default in research and clinical practice worldwide. But it is not a universal magic number.

A validation study in Indian university students, for example, found that the optimal cutoff for distinguishing students with and without sleep problems was a score above 6, not above 5. Sensitivity at that threshold was 75% and specificity was about 89%.³ The Japanese version of the PSQI settled on a cutoff of 5.5, and it correctly identified patients with insomnia, depression, generalized anxiety, and schizophrenia with sensitivity above 80% in each group.⁴ A study of Korean pregnant and postpartum women used a higher threshold of 10 to define “very poor” sleep quality, and even at that bar, about 80% of participants qualified during late pregnancy.⁵

The general advice: if you score 5 or below, the PSQI considers your sleep quality “good.” A score of 6 to 10 often lands in a moderately poor range. Scores above 10 suggest fairly significant sleep difficulty. But the exact cutoff that best separates healthy sleepers from problematic ones depends on the population, the language version of the questionnaire, and sometimes the clinical condition being studied.

Why the Component Scores Matter on Their Own

Researchers and clinicians tend to report the global score because it is convenient, but collapsing everything into a single number can obscure useful information. Two people with a global score of 8 could have very different sleep problems: one might lie awake for an hour every night (high sleep latency) but sleep soundly once asleep, while another falls asleep instantly but wakes repeatedly (high sleep disturbances) and feels exhausted by midday (high daytime dysfunction).

A study of healthy older adults found that objective sleep measures correlated with some component scores but not others, and that relying on the global score alone masked those relationships.⁶ For clinical purposes, looking at which components are elevated can point toward different interventions. High sleep latency suggests difficulty initiating sleep, which might respond well to relaxation techniques or cognitive behavioral therapy for insomnia. High daytime dysfunction paired with adequate sleep duration might raise questions about sleep apnea or another disorder that disrupts sleep quality without shortening it.

How Reliable Is the PSQI Across Different Groups

A questionnaire’s internal consistency tells you whether its items hang together as a coherent measure rather than pulling in random directions. Across a broad range of populations, the PSQI’s internal consistency has generally held up well. An early evaluation found an overall Cronbach’s alpha of 0.80, with moderate to high correlations between global and component scores.² Studies in pregnant women reported alphas in the 0.70 to 0.75 range, which is considered acceptable.⁷

The picture gets murkier in certain populations. Among very old adults (centenarians), the alpha dropped to 0.68, and it improved to 0.78 only after the medication-use and daytime-dysfunction components were removed.⁸ A study of healthcare workers during the COVID-19 pandemic found that internal consistency improved when the daytime-dysfunction subscale was dropped.⁹ That particular component seems to be the weakest link in the chain, perhaps because daytime tiredness can stem from many causes beyond poor nighttime sleep.

The PSQI also struggles with shift workers. A study of long-haul truck drivers in South Africa found an overall alpha of just 0.42, well below the threshold considered reliable. It dropped even further among the subgroup of night-shift workers. The authors recommended that shift-worker studies either validate the PSQI in their specific sample or use questionnaires designed for irregular schedules.¹⁰ This makes intuitive sense: the PSQI asks about “usual” bedtimes and wake times over the past month, and someone alternating between day and night shifts does not have a single usual pattern.

What the PSQI Captures Compared to Objective Sleep Measurement

One of the most persistent questions about the PSQI is whether it lines up with what happens when you actually track sleep with a device. The short answer is: sometimes, and only loosely.

A community-based study found that neither the PSQI nor the Epworth Sleepiness Scale correlated with actigraphy or polysomnography measures. Both questionnaires separated people with different psychological symptoms and subjective sleep complaints, but neither predicted what the devices measured.¹¹ A longitudinal study tracking women from pregnancy through 12 months postpartum found that self-reported and actigraphic total sleep time were correlated but only at a low-to-moderate level, and that actigraphy consistently estimated longer sleep times than the women reported. Sleep efficiency and sleep latency showed even weaker agreement between the two methods.¹²

A separate study in the general population found that people overestimated their habitual sleep duration by about 20 to 30 minutes when reporting on questionnaires, compared to what polysomnography and actigraphy recorded.¹³ Meanwhile, research in older adults concluded that the PSQI and actigraphy should not be treated as interchangeable, and that best practice is to use both subjective and objective measures together.¹⁴

This does not mean the PSQI is wrong or useless. Subjective sleep quality captures something real that devices miss: how you experience your sleep. Feeling rested matters clinically, even if a wrist sensor says you slept seven hours. The mismatch just means you should not interpret a low PSQI score as proof that nothing is objectively wrong with your sleep, or a high score as confirmation that a sleep study would find a disorder.

What Counts as a Meaningful Change in Score

If you take the PSQI before and after some intervention, whether surgery, therapy, or a lifestyle change, how much does your score need to drop before the improvement is real and not just noise? One study in patients recovering from rotator cuff repair calculated that a change of about 4.4 points represented the minimal clinically important difference. The same study found that a score of roughly 5.5 or below at follow-up corresponded to a symptom state most patients considered acceptable.¹⁵ Those numbers come from a surgical population, so they may not apply directly to other contexts. Still, the rough benchmark is useful: a shift of one or two points on the PSQI is probably within normal fluctuation, while a shift of four or more points is starting to reflect a genuine change in sleep quality.

Using the PSQI in Psychiatric Populations

The PSQI was originally designed with psychiatric patients in mind, and it performs well in that context. A Hungarian validation study found that global and component scores differed significantly between psychiatric patients and healthy controls, and the questionnaire converged well with the Athens Insomnia Scale across almost all components.¹⁶ One interesting wrinkle from that study: patients with schizophrenia had lower (better) global scores than patients with depression or anxiety. That may reflect the sedating effects of antipsychotic medication, or it may reflect reduced insight into sleep disturbances. Either way, clinicians interpreting PSQI scores in psychiatric patients need to consider how the underlying condition and its treatment interact with sleep self-report.

In older men, the PSQI differentiated participants with self-reported sleep disorders from those without more reliably than the Epworth Sleepiness Scale did, and it correlated with depressive symptoms and functional limitations in expected ways.¹⁷ The PSQI’s daytime-dysfunction component, in particular, tends to track with depression scales, which makes it both a strength (sleep and mood are tightly linked) and a caution (an elevated PSQI may partly be reflecting mood rather than sleep per se).

Adaptations for Younger Populations

The standard PSQI was designed with adults in mind, and its phrasing and scoring thresholds may not map cleanly onto adolescent sleep patterns. A study adapting the PSQI for Spanish adolescents and young adults (ages 14 to 24) made slight modifications and found that the adapted version measured a single factor with adequate reliability.¹⁸ More recently, researchers developed a six-item brief version of the PSQI for adolescents that adjusted its scoring to account for age-specific sleep recommendations, since teenagers biologically need more sleep than middle-aged adults. The age-adjusted version shifted the cutoff for poor sleep to 6 or above, and about 42% of the adolescent sample crossed that threshold.¹⁹

If you are a parent or clinician using the PSQI with a teenager, the standard adult scoring thresholds will likely overcount “good” sleepers, because a teen getting six and a half hours of sleep would score a 1 on the duration component (the same as an adult), even though six and a half hours is well below what most sleep experts recommend for adolescents.

How Recall Period Affects Accuracy

The PSQI asks about sleep over the past month, which is a long window to remember accurately. Research comparing electronic daily ratings with retrospective recall found that for the purpose of comparing one person to another, recall over 3-day, 7-day, and 28-day windows was reasonably accurate, with correlations around 0.74 to 0.77. However, within the same individual, accuracy dropped when people tried to recall specific nights within the past week.²⁰ The practical takeaway: your PSQI score is a reasonable summary of your general sleep pattern, but it should not be treated as a precise diary of what happened on particular nights. If you need that level of detail, a daily sleep log or wearable device is better suited to the task.

Single-Factor Versus Multi-Factor Models

The original PSQI treats the seven components as if they all measure one underlying thing (global sleep quality), which is why adding them up into a single score makes sense. But some researchers have argued that the data fit better when you separate the components into two or three clusters. One influential study found that a three-factor model, grouping the components into “sleep efficiency,” “perceived sleep quality,” and “daily disturbances,” provided a better statistical fit than the original single-factor structure.²¹

For most people filling out the PSQI in a clinical setting, this debate is academic. You will almost always see the global score reported. But if you are reading sleep research and notice that a study reports PSQI “factor scores” or “subscale clusters” rather than just the global score, this is what they are doing: grouping the seven components into broader dimensions instead of averaging everything together. Neither approach is wrong. The single global score is more familiar and easier to compare across studies; the multi-factor approach can reveal patterns that the global score buries.

Shift Work and the PSQI’s Blind Spot

Nurses working rotating shifts had objectively shorter sleep and lower sleep-stage durations compared with day-shift-only nurses in a study that combined PSQI data with smartwatch measurements.²² The PSQI picked up on some of that difference, but the larger concern for shift workers is structural: the questionnaire assumes a single consistent sleep-wake schedule. Someone who sleeps from 11 p.m. to 7 a.m. three nights a week and from 8 a.m. to 3 p.m. the other four has no sensible answer to “what time have you usually gone to bed.” Reporting an average time obscures the very irregularity that is most damaging to their sleep quality. Until shift-worker-specific questionnaires gain wider traction, clinicians using the PSQI with rotating-shift employees should interpret the scores cautiously and supplement with daily logs wherever possible.