Athens Insomnia Scale (AIS)

Insomnia is a highly prevalent health complaint, estimated at 10% to 30% in the general population, and it remains underdiagnosed and undertreated (Soldatos et al., 2000; Chiu et al., 2016). It is considerably more common in mental health settings. The Athens Insomnia Scale measures how much difficulty a person has had sleeping over the past month, counting only difficulties that occurred at least three times a week. It was written by Constantin Soldatos, who drafted the ICD-10 diagnostic criteria for sleep disorders, and its items map directly onto those criteria (Soldatos et al., 2000). In deriving the predictive values of the AIS, Soldatos et al. (2003) assumed a prevalence of 10% in the general population and 50% among unselected psychiatric patients. Insomnia symptoms are strongly associated with mental illness, including depression (Jahrami et al., 2023). Because structured clinical interviews take time and require trained administration, and polysomnography is not recommended for routine insomnia assessment, brief validated self-report measures are the practical route to identifying insomnia in routine care (Chiu et al., 2016).

AIS Domains

The AIS covers two aspects of insomnia as described by Soldatos et al. (2000).

• Sleep Difficulty (items 1 to 5): difficulty falling asleep, waking during the night, waking earlier than desired, insufficient sleep duration, and poor overall sleep quality.
• Daytime Consequences (items 6 to 8): reduced sense of well-being, impaired physical and mental functioning, and sleepiness during the day.

These are the groupings described by the scale’s author rather than validated subscales. They carry no cut-offs or normative data, and are reported so that night-time difficulty and daytime impact can be considered separately.

Versions of the AIS

Soldatos et al. (2000) describe two versions: the eight-item AIS-8, and a brief five-item AIS-5 for use where only sleep quantity and quality are of interest. NovoPsych implements the AIS-8, and every figure on this page refers to that version.

Clinical Applications of the AIS

The AIS was developed and validated in primary care and psychiatric services (Soldatos et al., 2000, 2003), and suits a variety of mental health settings. Its brevity supports several clinical uses.

• Screening at intake: the AIS identifies sleep complaints reaching the clinical cut-off.
• Progress monitoring: the AIS can be repeated across an episode of care. Cognitive behavioural therapy for insomnia is often recommended for chronic insomnia (Okajima et al., 2020) and is delivered over repeated sessions of sleep review, so the AIS fits naturally into treatment planning and review.
• Distinguishing night-time from daytime difficulties: a client whose difficulty is concentrated at night, with little daytime impact, presents differently from one whose sleep is only moderately disturbed but whose daytime functioning is markedly affected, and the two profiles point toward different emphases in treatment.
• Identifying treatment targets: difficulty falling asleep, waking through the night, and waking too early each suggest a different behavioural emphasis.

Alignment with Diagnostic Criteria

The first five items correspond to criterion A of the ICD-10 definition of nonorganic insomnia, and the requirement that difficulties occur at least three times a week over a month corresponds to criterion B. The last three items address the next-day consequences of disturbed sleep, covering the ICD-10 requirement for distress or interference with daily activities.

ICD-11 replaced the ICD-10 insomnia criteria in 2022, and both ICD-11 and DSM-5-TR set requirements the AIS does not address. Two matter most for interpretation. Both require at least three months of symptoms for their principal insomnia diagnosis, whereas the AIS asks about the past month. Both also require that symptoms occur despite adequate opportunity for sleep, which no AIS item asks about (American Psychiatric Association, 2022; Spiegelhalder & Riemann, 2025). Under ICD-11, poor sleep quality on its own no longer qualifies, since a person must report difficulty initiating or maintaining sleep. In one respect the AIS is stricter than ICD-11 requires: ICD-11 asks for symptoms several times per week, whereas the AIS counts only difficulties occurring at least three times a week. A positive AIS therefore indicates a recent pattern of sleep difficulty warranting further assessment, not a diagnosis.

AIS Scoring

Each AIS item is rated from 0 to 3, with response options worded specifically for that item, and the eight item scores are summed. The total ranges from 0 to 24, where 0 denotes the absence of any sleep-related problem and 24 the most severe degree of insomnia.

Two domain scores are also reported, each expressed as an average item score from 0 to 3, so that a five-item domain and a three-item domain can be read on the same scale.

• Sleep Difficulty (5 items; items 1 to 5)
• Daytime Consequences (3 items; items 6 to 8)

AIS Cut-Off and Descriptors

A total score of 6 or more is a positive screen for insomnia. Three descriptor categories are applied to the total score, anchored to that validated cut-off. The second boundary, at 10, is a NovoPsych descriptor gradation rather than a second cut-off.

In a general population, where roughly one person in ten has insomnia, a score of 6 or above carries a positive predictive value of 41%, so a positive result at that level is close to an even chance. At 10 or above it rises to 88% (Soldatos et al., 2003). This is the range where the AIS is most specific, with about 99% of people without insomnia scoring below 10. That confidence comes at the cost of sensitivity, which is 52% at the same point, so many true cases sit lower on the scale. Screening therefore continues to work from 6, while 10 describes a result that is strongly supported.

A positive screen indicates that further clinical assessment is warranted, not that a diagnosis of insomnia is established. Results should be interpreted alongside clinical interview findings, and where the presentation suggests another sleep disorder such as obstructive sleep apnoea or a circadian rhythm disorder, referral for specialist sleep assessment may be appropriate.

Interpreting the AIS Domain Scores

Because the domains have no cut-offs, they are interpreted relative to one another rather than against thresholds. A higher Sleep Difficulty score with a lower Daytime Consequences score suggests that disturbed sleep is not yet producing marked functional impact during the day. A higher Daytime Consequences score with a comparatively low Sleep Difficulty score suggests that daytime impairment is disproportionate to the reported night-time disturbance, which may point toward causes other than insomnia, including other sleep disorders, mood disturbance, or medication effects. High scores on both indicate a presentation in which night-time difficulty and daytime impairment are both prominent.

Individual item responses are worth reviewing alongside the domain scores, since each item describes a specific and separately treatable aspect of sleep.

AIS Graphs and Change Over Time

On a first administration the report presents two graphs: the total score on a bar chart spanning the 0 to 24 range with the three descriptor zones shaded and the clinical cut-off marked, and the two domain scores on a 0 to 3 average item scale against a background gradient beginning at 0.75, the cut-off expressed per item.

When two or more administrations are available, both charts become trajectories across administration dates on the same scales, so values stay comparable between views.

The interpretation also reports the change in total score between the initial and current administrations, and highlights any movement across the clinical cut-off separately, since a change from a positive to a negative screen represents a change in screening status rather than simply a change in score. Change is reported as a raw score difference only. No reliable change estimate has been published for the AIS, so none is applied.

Development of the AIS

The AIS was developed following the publication of the ICD-10 diagnostic criteria for insomnia, as a tool to help clinicians assess insomnia severity against those criteria (Soldatos et al., 2000). It was drafted in English and translated into Greek for the validation study following World Health Organization translation guidelines, with a bilingual expert panel, an independent unilingual panel, and back-translation.

Validation was conducted on 299 participants, all recruited in Greece: 105 primary care patients presenting with insomnia not attributable to an obvious underlying cause, 100 psychiatric outpatients, 44 psychiatric inpatients, and 50 non-patient controls. The sample averaged 46.0 years of age (SD = 15.7), ranged from 18 to 79 years, and was 58% female (Soldatos et al., 2000). The AIS has since been validated in more than ten languages and is among the most widely used insomnia instruments internationally (Jahrami et al., 2023).

AIS Reliability

Internal consistency. In the original validation, Cronbach’s alpha for the total score was .89 (N = 299), and alpha changed little when any single item was removed (range .83 to .89; Soldatos et al., 2000). A reliability generalisation meta-analysis pooling 25 studies and 97,158 participants reported an average alpha of .84 (95% CI .81 to .86), with clinical samples averaging .87 and community samples .81, though heterogeneity between studies was high (Jahrami et al., 2023). For the English-language version specifically, Manzar et al. (2022) reported an alpha of .86 in 563 nurses.

Test-retest reliability. The original study found a correlation of .89 over a one-week interval (N = 194), with individual item correlations ranging from .70 to .86 (Soldatos et al., 2000). Pooling 12 studies, Jahrami et al. (2023) reported an average test-retest coefficient of .86 (95% CI .80 to .92).

Response format. Manzar et al. (2022) examined the English version using a Rasch model, which tests whether response options behave as the scoring assumes. Each item’s four options formed a consistent progression, with each step reflecting greater difficulty than the last, and all items fitted the model acceptably.

AIS Factor Structure

NovoPsych scores the AIS as a unidimensional measure, with the total score as the primary index. The original principal component analysis identified a single component accounting for 56.9% of the variance, with all eight items loading between .62 and .82 (Soldatos et al., 2000). Manzar et al. (2022) found that exploratory methods, including parallel analysis, supported a one-factor solution in their sample of 563 nurses.

Confirmatory analyses have at times favoured a two-factor solution separating items 1 to 5 from items 6 to 8 (Elbi et al., 2025). Reported structures range from one to three factors, and the structure recovered varies with the analytic method applied. The two domains reported alongside the total score are therefore presented as the groupings described by the scale’s author, not as validated subscales.

AIS Validity

Convergent validity. The total score correlated .90 with the Sleep Problems Scale in the original validation (Soldatos et al., 2000), .80 with the Insomnia Severity Index in a sample of 1,666 Japanese government employees (Okajima et al., 2020), and between .62 and .81 with the Pittsburgh Sleep Quality Index in older adults (Elbi et al., 2025). A meta-analysis comparing the AIS, ISI and PSQI found no significant difference in diagnostic accuracy between the instruments, and rated the AIS and ISI the stronger of the three on diagnostic properties, sleep domains assessed, and feasibility (Chiu et al., 2016).

Known-groups validity. Participants who met ICD-10 criteria for insomnia averaged 11.05 (SD = 4.89) on the total score, compared with 2.28 (SD = 2.56) for those who did not, a highly significant difference (t = 20.19, p < .001; Soldatos et al., 2003).

Measurement invariance. Gender invariance has been demonstrated at all four levels for the two-factor model in the English version, indicating that the scale functions equivalently for men and women (Manzar et al., 2022). Total scores can be compared directly between men and women.

AIS Clinical Cut-Off

A total score of 6 or more indicates a positive screen and is the cut-off used by NovoPsych. It was derived by Soldatos et al. (2003) against blind ICD-10 clinical diagnosis in the 299-participant validation sample. Each participant was assessed for the diagnosis by two co-investigators conducting an unstructured clinical interview without knowledge of the participant’s AIS score. The cut-off demonstrated sensitivity of 93%, specificity of 85%, and 90% of participants correctly classified. Logistic regression confirmed the same boundary, with the predicted non-insomnia group scoring between 0 and 5 and the predicted insomnia group scoring 6 or above (R² = .78, p < .001).

Soldatos et al. identified 6 and 7 as the two best candidate cut-offs, and noted that the optimal value may vary with setting and with the resources available to assess those who screen positive. Higher cut-offs improve positive predictive value at the cost of sensitivity.

Accuracy in other samples. A meta-analysis pooling four AIS studies reported an average sensitivity of .91 (95% CI .87 to .93) and specificity of .87 (95% CI .68 to .95), closely matching the original derivation (Chiu et al., 2016). One of the replication studies used a structured clinical interview as its diagnostic criterion.

Predictive value by setting. The predictive value of a positive screen depends on how common insomnia is in the setting where the measure is used. In a psychiatric population a score of 6 or more corresponds to a positive predictive value of 86%; in a general population the same score corresponds to 41% (Soldatos et al., 2003). Negative predictive value is high in both settings, at 92% and 99% respectively, so a score below 6 is a strong indication that clinically significant insomnia is absent.

AIS Normative Data

NovoPsych does not report percentiles for the AIS. No English-speaking, Australian or Western general-population normative sample has been published, and the largest adult non-clinical sample available is 1,666 Japanese government employees. The AIS is interpreted against its validated clinical cut-off rather than against norms.

Assess your patients with the Athens Insomnia Scale (AIS) using a NovoPsych account

• For Psychologists & Mental Health Clinicians
• Send Assessments to Patient’s Phone
• Receive Comprehensive Reports
• Access Over 100 Validated Assessments​
• Instant Psychometric Scoring
• Track Symptoms​
• Inform Treatment​