Work Ability Index (WAI)

Structural validity Moderatecontestedevidence form: mixed

indirect evidence from German, Spanish, Brazilian, Iranian and Chilean working samples; no UK confirmatory data located.

Dimensionality is genuinely contested and the original single-score (unidimensional) scoring assumption is not well supported. A representative German confirmatory analysis (n=3968) found two correlated factors, a 'subjective work ability and resources' factor and a 'health-related' factor, fitting better than one factor (Freyer 2018). An earlier German multi-occupational study (n=324) likewise rejected both a one-factor and an orthogonal two-factor model, favouring a two-dimensional correlated-factor solution in which only five of seven items loaded unambiguously (Martus 2010). Spanish health-centre workers (n=1184) showed EFA support for one factor but better CFA fit for two factors (Mateo Rodriguez 2021), a pattern echoed in Spanish aeronautical workers (Gonzalez-Dominguez 2024) and a Brazilian online-application study (Pucci 2024). Three-factor solutions have also been reported in Iranian (Abdolalizadeh 2012) and Chilean/Spanish (Bascour-Sandoval 2020) samples, while a recent large Brazilian re-evaluation (n=3051) reported a unidimensional structure but with only marginal fit (RMSEA 0.125, TLI 0.900, CFI 0.882) (Martinez 2025). The weight of evidence favours a two-factor structure over the historical single sum score, but solutions vary by sample and language.

Sub-grades (evidence differs by subgroup):

• {"subgroup": "one-factor (historical sum-score assumption)", "grade": "Low", "note": "generally not supported by CFA in representative samples"}
• {"subgroup": "two correlated factors", "grade": "Moderate", "note": "most frequently replicated solution (subjective work ability vs health-related)"}

Convergent and discriminant validity Moderatewell-establishedevidence form: mixed

indirect convergent evidence mostly from translated versions in non-UK working samples.

The WAI correlates positively with self-rated general health and health-related quality of life, consistent with its construct. In Spanish working individuals it correlated significantly with the SF-36 v2 (Bascour-Sandoval 2020), and in Spanish aeronautical workers it was highly correlated with self-assessed health status (Gonzalez-Dominguez 2024). Among Iranian construction workers the WAI explained about 46 per cent of the variance in WHOQOL-BREF quality of life, very close to the single-item score (Mokarami 2021). Convergence with the SF-36 has also been used as evidence of criterion/convergent validity in Iranian healthcare workers (Abdolalizadeh 2012). Discriminant (known-groups) validity is consistently demonstrated: WAI separates groups differing in sickness absence and individual/work characteristics (Martinez 2025).

Criterion validity: reference standard Lowthinevidence form: mixed

indirect no independent diagnostic reference standard exists; comparators are self-report health measures in non-UK samples.

There is no diagnostic gold standard for work ability, so reference-standard criterion validity is assessed chiefly against established health measures. The WAI correlates moderately to strongly with the SF-36 and WHOQOL-BREF (Abdolalizadeh 2012; Mokarami 2021; Bascour-Sandoval 2020), and its own dimensions include physician-diagnosed disease counts. However, these are health-status comparators rather than an independent diagnostic reference standard, so this criterion is better regarded as convergent evidence; a true reference-standard anchor for work ability does not exist.

Criterion validity: organisational Highwell-establishedevidence form: canonical

indirect for UK generalisability (Finnish, Dutch, Swedish and German cohorts) but directly in occupational settings against genuine work outcomes such as disability pension and sickness absence.

This is the WAI's strongest evidential domain: it prospectively predicts work-relevant outcomes in large occupational cohorts. Among 11,537 male construction workers the full WAI discriminated future disability pension well (AUC 0.78, 95% CI 0.75 to 0.80) and was adequately calibrated (Roelen 2014). In 5251 Finnish municipal employees the hazard ratio for disability pension was 5.0 (95% CI 4.4 to 5.6) for poor versus good/excellent WAI (Jaaskelainen 2016). In a Swedish general-population sample the full WAI predicted long-term (90-day) sickness absence over four years with AUC 0.79 (95% CI 0.76 to 0.82), outperforming its individual items (Lundin 2017). Among 1331 Dutch office workers each one-point lower WAI raised the odds of 15 or more sick days (OR 1.27, 95% CI 1.21 to 1.33; AUC 0.77 for 15+ versus 0 days) (Reeuwijk 2015), and in 3660 employees WAI was inversely associated with frequent (OR 0.85), long-term (OR 0.79) and combined (OR 0.74) sickness absence (Notenbomer 2015). Higher WAI was protective against work nonparticipation (long-term sickness absence, early retirement, unemployment) in a representative German cohort (n=2426) (Conway 2023), predicted long-term sickness absence across a ten-year follow-up (Palmlof 2019), predicted application for disability pension after vocational rehabilitation (Bethge 2013), and was associated with sickness absence in young employees (Kujala 2006) and with disability more broadly (Alavinia 2009). One caveat: because a WAI dimension counts recent sickness absence days, prediction of absence outcomes is partly non-independent of the predictor.

Internal consistency Moderatewell-establishedevidence form: mixed

indirect alpha estimates from translated versions in non-UK occupational and nursing samples.

Cronbach's alpha for the total WAI typically falls in the 0.7 to 0.8 range: 0.79 in Iranian healthcare workers (Abdolalizadeh 2012), 0.71 in Croatian nurses (Smrekar 2020), and 0.787 (McDonald's omega 0.819) in 3051 Brazilian nurses (Martinez 2025). Values are acceptable but modest, which is expected because the WAI is a heterogeneous composite (partly a formative index of distinct health and work-demand indicators) rather than a homogeneous reflective scale, so alpha is of limited interpretive value for this instrument.

Test-retest reliability Moderatewell-established

indirect retest evidence from Dutch and Iranian occupational samples; no UK data.

Full-scale WAI test-retest reliability is adequate. A four-week study in elderly construction workers found stable group means and 66 per cent exact category agreement without reporting an ICC (de Zwart 2002); Iranian validations report ICCs of 0.92 for the total score (Abdolalizadeh 2012) and above 0.70 for each dimension (Adel 2019). Retest reliability of the single first item (Work Ability Score) is treated separately and is more variable; see the WAS record.

Measurement invariance Absent (a finding about the literature)untestedevidence form: canonical

indirect no formal invariance analyses located in any population.

Formal measurement-invariance testing (configural, metric, scalar) across sex, age, occupation, language or time was not located for the WAI in in the latest review pass. Several studies report factor structures in different countries and note mean score differences by age (younger workers scoring higher), but cross-group score differences are not evidence of measurement invariance, and the varying factor solutions across samples (Martus 2010; Freyer 2018; Mateo Rodriguez 2021) suggest invariance cannot be assumed. This is a genuine gap.

Responsiveness and MIC Absent (a finding about the literature)thinevidence form: canonical

indirect no responsiveness/MIC study of the full WAI located.

Responsiveness and minimal important change for the full 7-item WAI were not established in the retrieved literature; longitudinal WAI studies use it as a predictor rather than an outcome, and no anchor-based MIC for the total index was located. Responsiveness and MIC evidence exists for the single-item Work Ability Score (in clinical rehabilitation samples) and is reported in the WAS record.