Penn Alcohol Craving Scale (PACS)

1. the frequency of thoughts about drinking
2. the intensity of the urge to drink
3. the duration of craving episodes
4. the perceived ability to resist drinking
5. an overall self-rated craving level

PACS Example Item

The Significance of Cravings

Clinical Applications of the PACS

PACS Scoring

Each item is rated on a 7-point scale, where 0 represents the complete absence of craving (e.g., ‘never thought about drinking’, ‘no urge’) and 6 represents the most severe level (e.g., ‘thought about drinking nearly all of the time’, ‘would not be able to resist’). Response anchors vary across items but follow the same severity progression. The PACS total score is the sum of all five items, yielding a possible range of 0 to 30. Higher scores indicate greater alcohol craving.

PACS Results Table

PACS Craving Severity Categories

Descriptive severity labels are based on the response anchors’ meanings at each score level (i.e., what a typical respondent at that score would endorse across items). Based upon the PACS total score, descriptive severity labels are (with Murphy et al.’s (2014) original descriptors in brackets):

• 0–14: No/Minimal (Absent). Craving is absent or minimal. The average response falls in the “nonexistent”, “slight”, or “mild” range across items.
• 15–20: Moderate (Subclinical). Craving is present at a moderate level. The average response corresponds to “moderate urge” or “moderately difficult” to resist.
• 21–30: Elevated (Present). Craving is clearly present and potentially clinically significant. The average response corresponds to “strong urge” or “very difficult” to resist.

These categories are content-based interpretive guideposts derived from the response scale semantics, not empirically validated diagnostic thresholds. They should be interpreted as describing the general level of craving severity rather than as clinical classifications. In clinical practice, a score of 15 or above may warrant further attention to craving management, while scores of 21 or above suggest craving at a level where the respondent reports strong urges and significant difficulty resisting.

PACS Total Score Graph

PACS Percentile Context

To provide additional interpretive context, percentile rankings are derived from a treatment-seeking heavy-drinking reference sample (Murphy et al., 2014). These percentiles indicate a respondent’s position relative to other treatment-seeking heavy drinkers. For example, a total score of 15 corresponds to approximately the 26th percentile in the treatment-seeking sample, indicating that roughly one-quarter of treatment-seeking heavy drinkers scored at or below this level. A score of 21 corresponds to approximately the 63rd percentile.

PACS Horizontal Graph

PACS Change Score Interpretation

The PACS is primarily intended for progress monitoring, and score changes across administrations provide clinically informative data about treatment response. A minimally important difference (MID) of 3 points is used to evaluate whether a change in total score is clinically meaningful. This threshold is derived by applying the widely used convention that a meaningful change equals approximately half a standard deviation (Norman et al., 2003; Turner et al., 2010), applied to the treatment-seeking sample SD of 5.99 (Murphy et al., 2014).

PACS Total Score Over Time Graph

PACS Graphs

On first administration, a bar chart displaying the PACS total score with craving severity shading in the background is presented. Clinical percentiles (a treatment-seeking heavy drinking sample) are shown on the right-hand side of the graph. A horizontal bar chart displays the PACS total score in comparison to a non-treatment-seeking (heavy drinking) sample and a treatment-seeking (heavy drinking) sample. Multiple administrations of the PACS will produce a line chart plotting the raw total score over time, with background shading indicating the craving severity categories.

PACS Development

The PACS was originally developed and validated in a sample of 147 alcohol-dependent outpatients participating in a combined naltrexone and psychotherapy trial (Flannery et al., 1999). The PACS was developed to address the need for a brief, psychometrically sound measure of alcohol craving. Items were designed to capture the key phenomenological dimensions of craving (frequency, intensity, duration, resistance, and overall level), while remaining theoretically neutral with respect to the underlying nature of craving.

PACS Reliability

Internal consistency of the PACS is excellent. The original development study reported Cronbach’s α = .92 with item-total correlations ranging from .80 to .92 (N = 147; Flannery et al., 1999). Subsequent validation studies have consistently replicated this finding: α = .89 in a Polish clinical sample (N = 510; Chodkiewicz et al., 2016), α = .95 in a Taiwanese clinical sample (N = 160; Ko et al., 2024), and α = .80–.91 across three German clinical sites (Ns = 107–440; Nakovics et al., 2023).

Test-retest reliability data for the original English-language PACS have not yet been reported in the literature, though translated versions have provided some relevant evidence. The Chinese PACS-C demonstrated a test-retest correlation of r = .97 over a one-day interval (N = 40; Ko et al., 2024), though this very short retest interval limits its interpretive value for estimating stability over clinically meaningful periods. An adapted version of the PACS for polysubstance use (the Aggregated Drug Craving Scale; Costello et al., 2020) demonstrated a test-retest ICC of .82, providing indirect support for the temporal stability of the PACS measurement framework.

PACS Validity

Factor structure. The original study supported a unidimensional structure (Flannery et al., 1999), which has been confirmed in multiple subsequent analyses (Murphy et al., 2014; Costello et al., 2020). One exception is Nakovics et al. (2023), who reported that a two-factor solution (‘difficulty to resist’ and ‘thoughts about alcohol’) was more stable than the one-factor model in German clinical samples. This two-factor finding has not been replicated in other language versions and may reflect characteristics of the German translation.

Convergent validity. The PACS shows strong convergent validity with other craving measures. Flannery et al. (1999) reported correlations with the Obsessive Compulsive Drinking Scale (OCDS; r values not specified but described as demonstrating convergent validity) and the Alcohol Urge Questionnaire. In a Taiwanese validation study (Ko et al., 2024; N = 160), the PACS-C demonstrated strong convergent validity, correlating highly with both the Visual Analogue Scale for craving (r = .81) and the Yale-Brown Obsessive Compulsive Scale for heavy drinking (obsession r = .81, compulsion r = .79). Discriminant validity was supported by a lack of correlation between PACS scores and non-craving self-report measures in the original study (Flannery et al., 1999).

Predictive validity. The PACS demonstrates predictive validity for relapse. Flannery et al. (1999) showed that PACS scores at week 2 of treatment significantly predicted alcohol relapse during weeks 3–12 via logistic regression. Kharb et al. (2018), in a small Indian sample (N = 30), found that PACS scores at discharge significantly differentiated individuals who relapsed within one month (M = 15.9, SD = 2.51) from those who did not (M = 7.78, SD = 1.56).

Criterion validity. Hartwell et al. (2019) examined correspondence between PACS scores and a structured diagnostic interview for craving (N = 338 non-treatment-seeking heavy drinkers), an important consideration given craving’s status as a formal DSM-5 AUD criterion. A PACS threshold of ≥15 yielded sensitivity of 67%, specificity of 81%, positive predictive value of 49%, and negative predictive value of 90%. The higher threshold of >20 yielded a sensitivity of only 41% with a specificity of 95%. This remains the only known study to date to evaluate the PACS against a structured diagnostic criterion.

PACS Sensitivity to Change

Direct evidence of the PACS’s sensitivity to change in a treatment context comes from the adapted version (ADCS). Costello et al. (2020) reported a large treatment effect (Cohen’s d = −1.54) from baseline (M = 19.6, SD = 8.7) to 12-month follow-up (M = 7.5, SD = 6.9) in an inpatient substance use disorder programme (N = 191). While this was the adapted polysubstance version, the structural equivalence of the ADCS to the PACS (confirmed by CFA) suggests that the PACS measurement framework may be sensitive to treatment-related change in craving, though direct evidence in alcohol-specific samples is needed.

PACS Non-Treatment-Seeking and Treatment-Seeking Sample Data

Two reference samples are available for the PACS total score. The non-treatment-seeking sample (N = 338) comprised non-treatment-seeking heavy drinkers (68% male) with a mean score of 11.1 and a standard deviation of 7.2 (Hartwell et al., 2019). The treatment-seeking sample (N = 104) comprised treatment-seeking heavy drinkers (62% male) who had a mean score of 18.94 with a standard deviation of 5.99 (Murphy et al., 2014). The substantial difference between these samples (approximately 1.2 standard deviations) reflects the expected elevation in craving among individuals who are actively seeking treatment relative to community heavy drinkers.

PACS Thresholds

Murphy et al. (2014) proposed a content-based framework for interpreting PACS total scores, derived from the semantics of the response options. When a respondent’s average item score (total score ÷ 5) is mapped to the response anchors, three interpretive categories emerge: Absent, Subclinical, and Present. NovoPsych uses labels that are functionally equivalent (No/Minimal, Moderate, Elevated) to Murphy’s original labels but provide more descriptively transparent severity language for clinical reporting. Notably, the lower boundary of the Moderate category (score of 15 or greater) receives independent empirical support from Hartwell et al. (2019), who identified this threshold as optimal for discriminating diagnostic-interview-endorsed craving from non-craving (sensitivity = 0.67, specificity = 0.81, NPV = 0.90), providing convergent criterion-validity support for this boundary.

Assess your patients with the Penn Alcohol Craving Scale (PACS) 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​