Home / Resources / Psychological Measures / PSQI
Pittsburgh Sleep Quality Index:
Scoring and Interpretation
The Pittsburgh Sleep Quality Index (PSQI) is a self-report measure of sleep quality and disturbance over the past month that converts 19 questions into seven component scores and a global score from 0 to 21 (Buysse et al., 1989). Two stages define PSQI scoring, from 19 raw answers to seven 0 to 3 components and then to the global score. Alongside them come the original cut-off above 5, SPSS and R syntax for the final sum, the 1989 validation figures and an example of how to report the global score.
What the PSQI Measures
The PSQI assesses self-reported sleep quality and the kinds of disturbance a person had during the previous month. Unlike a plain Likert sum, it folds 19 questions with different response formats into seven scored components, and those are added to give a global index of sleep difficulty (Buysse et al., 1989).
It is used in clinical, psychiatric, sleep-medicine and health-outcomes research because it allows subjective sleep quality, latency, efficiency and related disturbances to be compared across groups and over time. It is not a polysomnographic assessment and does not diagnose a particular sleep disorder. Problems tend to arise in the scoring step: adding raw answers directly, and misreading times that cross midnight.
Structure of the PSQI
- Items: 19 self-rated questions, with five optional bed-partner questions that do not enter the global score.
- Components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication and daytime dysfunction.
- Response formats: some answers are frequency categories and others are clock times or durations, which the official algorithm converts into component categories.
- Component scale: each component is scored from 0 to 3.
- Total: the global score is the sum of the seven components, from 0 to 21, with higher scores meaning worse subjective sleep quality.
- Reverse scoring: the conventional computation does not use reverse-coded item numbers.
The instrument itself is available from the University of Pittsburgh Department of Psychiatry and the American Thoracic Society outcome-measures directory.
How to Score the PSQI
- Apply the University of Pittsburgh scoring key to the raw questionnaire responses. Sleep duration, time in bed and habitual sleep efficiency each have prescribed calculations, and times that cross midnight need checking.
- Create the seven keyed component variables psqi_c1 to psqi_c7, each ranging from 0 to 3, in the official order. These component scores, not the 19 raw answers, are the inputs to the code below.
- Sum the seven components to obtain psqi_global, which ranges from 0 to 21. The code does not prorate: if a required component cannot be computed, the global score stays missing unless a justified, prespecified alternative applies. Describe it in the methods section.
* First apply the official scoring worksheet to the 19 raw answers.
* Illustrative syntax: psqi_c1 to psqi_c7 are already computed component scores (0 to 3).
COMPUTE psqi_global = SUM.7(psqi_c1, psqi_c2, psqi_c3, psqi_c4, psqi_c5, psqi_c6, psqi_c7).
EXECUTE.# data: one row per respondent; psqi_c1 to psqi_c7 are components already scored 0 to 3
# with the official key. Do not pass the 19 raw responses to this sum.
cols <- paste0("psqi_c", 1:7)
data$psqi_global <- rowSums(data[cols], na.rm = FALSE)Interpreting PSQI Scores
| Score band | Interpretation |
|---|---|
| 0 to 5 | At or below the original threshold for poor sleep quality |
| 6 to 21 | Above the original threshold for poor sleep quality (PSQI above 5) |
Source: Buysse et al. (1989); American Thoracic Society (n.d.).
Higher global scores indicate worse sleep quality. In the original patient and healthy-control sample, a global score above 5 separated poor sleepers with sensitivity of 89.6% and specificity of 86.5%. Those estimates describe that validation setting and not every contemporary population. The seven components show where a group’s sleep difficulty comes from.
Summing the 19 data columns without first computing the official components gives an invalid result, because the raw questions use different response formats. Confirm the scoring worksheet and the handling of the medication and bed-partner questions, and note that agreement with objective measures such as polysomnography or actigraphy cannot be assumed.
Psychometric Properties
| Property | Published finding | Source |
|---|---|---|
| Internal consistency | Cronbach’s α = .83 across the seven components in the original evaluation | Buysse et al. (1989); American Thoracic Society (n.d.) |
| Test-retest | Correlation of approximately r = .85 reported in the initial work | Buysse et al. (1989); American Thoracic Society (n.d.) |
| Validity evidence | Differentiated healthy good sleepers from patients with insomnia or depression | Buysse et al. (1989) |
| Diagnostic discrimination | Global score above 5: sensitivity 89.6% and specificity 86.5% in the original sample | Buysse et al. (1989) |
| Components | Seven domains intended to capture different aspects of past-month sleep | Buysse et al. (1989) |
These are source-specific results and cannot be assumed to hold as reliability or accuracy estimates in a new sample.
Using the PSQI in a Study
Use the 0 to 21 global score for between-group and repeated-measures questions about subjective sleep quality, and show the component profile to explain what drives a difference. Floor effects or marked skew in the global score should shape the model choice, for example a rank-based or robust approach when the distribution cannot support a linear model. A design with two groups measured before and after an intervention suits a Mixed ANOVA when its assumptions hold, while Generalized Estimating Equations (GEE) handle repeated measures with unbalanced follow-up and give population-averaged effects.
Report the scoring worksheet and the questionnaire version you used. If PSQI scores are related to an objective sleep measure, Correlation Analysis (Pearson & Spearman) summarizes the association with a confidence interval. Because the components come from a compound scoring algorithm and not from 19 interchangeable ordinal items, Reliability Analysis (Cronbach’s Alpha) applies to the seven components and should be run on your own sample. No minimal important change should be asserted without a setting-specific anchor study.
Related field: Neuropsychology & Neuroscience
Reporting the PSQI
Sleep quality was assessed with the Pittsburgh Sleep Quality Index (PSQI; Buysse et al., 1989). The 19 self-rated responses were converted, using the published algorithm, to seven component scores (0 to 3) that were summed to a global score (range = 0 to 21), with higher scores indicating worse sleep quality. Missing responses were handled by [rule], and internal consistency of the seven components was [α = XX].
References
American Thoracic Society. (n.d.). Pittsburgh Sleep Quality Index. https://site.thoracic.org/assemblies/srn/sleep-related-questionnaires/psqi
Buysse, D. J., Reynolds, C. F., III, 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. https://doi.org/10.1016/0165-1781(89)90047-4
University of Pittsburgh, Department of Psychiatry. (n.d.). Scoring instructions for the Pittsburgh Sleep Quality Index. https://www.psychiatry.pitt.edu/sites/default/files/inline-files/PSQI%20Article.pdf
Frequently Asked Questions
Can I simply add up the 19 PSQI answers?
No. Convert the raw responses into the seven components with the official key and sum those components. The questions use different formats, so a raw sum has no meaning.
What does a PSQI score above 5 mean?
It is the poor-sleep threshold from the original validation. Treat it as a screening classification for comparable populations, not a diagnosis of a sleep disorder.
Do the bed-partner questions count toward the score?
No. The optional bed-partner or roommate questions sit outside the seven components and the global total, so keep them in separate variables.
Can the PSQI show response to treatment?
It is often analyzed longitudinally for that purpose. Interpret change alongside other patient outcomes and population-specific evidence, since no universal threshold marks a meaningful improvement.
NEXT STEP
Need PSQI Components
Scored and Checked?
OnlineSPSS is a statistical consultancy for researchers. Share your design and dataset, and a statistically trained Research Coordinator reviews it. The scope and one fixed project fee are confirmed before work begins.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014