Turnover Intention Scale-6 (TIS-6, Roodt)
Identity
Version: 6-item short form (TIS-6), derived from Roodt's original 15-item Turnover Intention Scale (TIS-15, 2004)
Structure: 6
Original citation: Roodt, G. (2004). Turnover Intention Scale (TIS-15), unpublished document, University of Johannesburg; short form validated as the TIS-6 by Bothma, C.F.C. & Roodt, G. (2013), SA Journal of Human Resource Management, 11(1), a507, doi:10.4102/sajhrm.v11i1.507
Steward / publisher: Instrument developed by G. Roodt (University of Johannesburg); TIS-6 short form published open access by AOSIS in the SA Journal of Human Resource Management. No separate commercial licensing body.
Constructs claimed
Turnover intention: the self-reported cognitive and affective strength of an employee's intention to leave (or remain with) the employing organisation, treated as the proximal antecedent of actual turnover behaviour.
Evidence
Structural validity Moderatewell-establishedevidence form: mixed
indirect . Founding EFA in a South African ICT workforce; CFA support in Hungarian elderly-care and (with item reduction) Moroccan nurse samples; no UK working-adult CFA located.
The founding validation of the six-item form (Bothma & Roodt 2013) reported a single-factor solution on a large South African ICT sample (n = 2429) using principal-axis factoring, with item loadings in the range of roughly 0.73 to 0.82 on one factor, consistent with a unidimensional turnover-intention construct. Confirmatory replication came later and in translation: the Hungarian version among elderly-care workers (n = 269) supported an acceptable single-factor confirmatory factor model (Németh et al. 2024). The unidimensional structure is not universal across languages, however: an Arabic adaptation among Moroccan nurses and midwives (n = 304) did not sustain the six-item structure and was refined to a four-item version (the A.TIS-4) to achieve acceptable fit (Ammari et al. 2026), which signals that the six-item structure may not hold in every linguistic and occupational context. Overall the one-factor model is reasonably well supported, but the primary evidence is exploratory factor analysis in the founding study with confirmatory support arriving mainly through non-UK translations.
Convergent and discriminant validity Moderatewell-establishedevidence form: canonical
indirect . Correlations obtained in South African ICT and Hungarian elderly-care samples; no UK working-adult convergent evidence located.
In the founding study (Bothma & Roodt 2013) TIS-6 scores correlated in theoretically expected directions with adjacent constructs, most strongly and positively with work alienation (r about 0.73), positively with emotional exhaustion (r about 0.56) and depersonalisation (r about 0.37), and negatively with work engagement (r about -0.58), work-based identity (r about -0.56), reduced personal accomplishment (r about -0.20) and task performance (r about -0.13). Translated versions reproduce this pattern: the Hungarian version correlated with burnout components (r about 0.51 and 0.42) and with workplace stress (r about 0.57 and 0.31), and discriminated between groups differing in burnout severity (Németh et al. 2024). The very high correlation with alienation (about 0.73) is a discriminant-validity caution, indicating substantial construct overlap rather than clean separation. The pattern is consistent but the coefficients derive from non-UK samples.
Criterion validity: reference standard Not applicable (category difference)untestedevidence form: canonical
indirect . Not applicable to the construct: no health or diagnostic reference standard exists for turnover intention.
Turnover intention is an organisational-behavioural construct, not a health or clinical state, so there is no diagnostic or health reference standard against which the TIS-6 could be benchmarked. No study located in the latest review pass attempted validation of the TIS-6 against a diagnostic or health gold standard, and none would be expected. The instrument's meaningful criterion evidence is organisational (prediction of actual turnover), recorded separately. This field is therefore a category consideration rather than a genuine evidence gap.
Criterion validity: organisational Moderatethinevidence form: canonical
indirect . Single large prospective study in a South African ICT workforce; strong internal design but non-UK, single-organisation, and not independently replicated on actual turnover.
This is the instrument's distinctive strength and, unusually for a self-report workplace scale, it rests on prediction of actual turnover behaviour rather than intention alone. In the founding census-based study of a South African ICT company (n = 2429), Bothma & Roodt 2013 compared employees who subsequently left the company (leavers) with those who stayed. Over a four-month follow-up window the TIS-6 significantly distinguished leavers (mean about 5.14) from stayers (mean about 4.13), t(170) about 5.20, p <= 0.001, with a large effect size (partial eta squared about 0.14). Over a four-year window the scale still discriminated leavers from stayers, t(801) about -4.10, p <= 0.001, but the effect was small (partial eta squared about 0.02), showing predictive signal that attenuates as the horizon lengthens. This is a rare, prospective, behaviour-anchored organisational-criterion demonstration. Its limitation is that it rests on a single company in a single national and sectoral context, and no independent replication of the actual-turnover prediction (as opposed to correlational studies of intention) was located in the latest review pass. The wider literature applies the TIS-6 as a predictor or outcome in cross-sectional designs (Kabir et al. 2022; Alshmemri 2025; Kentar et al. 2025) but those do not observe actual turnover.
Internal consistency Moderatewell-establishedevidence form: mixed
indirect . Alpha estimates from South African ICT and Hungarian elderly-care samples; no UK working-adult estimate located.
Internal consistency is consistently acceptable to good. The founding study reported Cronbach's alpha of 0.80 on the South African ICT sample (Bothma & Roodt 2013), and the Hungarian elderly-care validation reported alpha of about 0.83 (Németh et al. 2024). Application studies using the TIS-6 across nursing and academic samples routinely report alphas in the acceptable-to-good band, consistent with a coherent six-item scale. Reliability is one of the better-evidenced properties, although the two coefficients extracted in detail in the latest review pass both come from non-UK samples.
Test-retest reliability Absent (a finding about the literature)untested
indirectness not applicable: no retest evidence located in the latest review pass
No test-retest (temporal stability) study for the TIS-6 was located in the latest review pass. Neither the founding validation nor the translation studies retrieved reported a retest coefficient (ICC, r or kappa) over a defined interval. This is a genuine and material gap: a scale used to detect change or to track intention over time carries no published evidence of its own stability, and the absence should be treated as the finding it is rather than assumed adequate.
Measurement invariance Absent (a finding about the literature)untestedevidence form: canonical
indirect . No formal configural, metric or scalar invariance test located; the one cross-language signal (Arabic four-item refinement) argues against six-item invariance.
No formal measurement-invariance analysis (multi-group confirmatory factor analysis testing configural, metric and scalar invariance across sex, age, occupation, language or time) was located for the TIS-6 in the latest review pass. Du Plooy & Roodt 2013 examined biographical and demographic variables as moderators of the prediction of turnover intention, which concerns differential prediction rather than measurement invariance of the scale itself and does not establish that the items function equivalently across groups. Indirect evidence points the other way in the cross-language case: the Arabic adaptation required reduction to four items to fit (Ammari et al. 2026), suggesting the six-item structure may not be invariant across languages. On the evidence retrieved, formal invariance is untested and cross-language structural equivalence is doubtful.
Responsiveness and MIC Absent (a finding about the literature)untestedevidence form: canonical
indirect . No longitudinal or interventional responsiveness study located; construct used cross-sectionally.
No responsiveness or minimal important change (MIC) evidence was located for the TIS-6 in the latest review pass. The scale is used almost exclusively in cross-sectional designs; no study retrieved examined sensitivity to change following an intervention or over time, and no anchor-based or distribution-based MIC threshold has been established. Responsiveness and MIC are therefore Absent.
Populations, languages and norms
The TIS-6 has been used across a broad range of settings and languages, but the evidence is skewed towards healthcare and non-UK populations. Validated or applied samples include South African ICT employees (Bothma & Roodt 2013), South African nursing and higher-education staff (Du Plooy & Roodt 2010; Takawira et al. 2014), Hungarian elderly-care workers (Németh et al. 2024), Moroccan nurses and midwives in Arabic (Ammari et al. 2026), Saudi nurses (Alshmemri 2025), Bangladeshi nurses (Kabir et al. 2022) and US dental and dental-hygiene faculty (Kentar et al. 2025). No formal population norm tables (reference distributions or cut-scores) were located, although some application studies use ad hoc thresholds. No UK working-adult validation or norms were located in the latest review pass.
Criticisms and controversies
Four honest limitations stand out. First, structural instability across languages: the Arabic adaptation had to drop two items to achieve fit (Ammari et al. 2026), questioning universality of the six-item form. Second, the flagship claim (prediction of actual turnover) rests on a single South African ICT company and has not been independently replicated on observed turnover (Bothma & Roodt 2013). Third, temporal stability (test-retest) and formal measurement invariance are both unevidenced, which is a notable gap for a scale used to track and compare intention. Fourth, discriminant validity is imperfect: the very high correlation with work alienation (about 0.73) indicates substantial construct overlap. The evidence base is also concentrated in healthcare and non-UK samples, limiting direct read-across to a general UK working-adult audience.
References (10)
- Bothma, C.F.C.; Roodt, G. (2013). The validation of the turnover intention scale https://doi.org/10.4102/sajhrm.v11i1.507
- Németh, Z.; Deák, P.; Szűcs, M.; et al. (2024). Validation of the Hungarian version of the 6-item turnover intention scale among elderly care workers https://doi.org/10.1038/s41598-024-66671-0
- Ammari, A.; Hamdoune, M.; Gantare, A. (2026). Psychometric validation of the Arabic version of the Turnover Intention Scale (TIS-6) among Moroccan nurses and midwives https://doi.org/10.1891/JNM-2025-0131
- Kentar, Y.; Boyd, L.D.; Vineyard, J.; et al. (2025). Effort-reward imbalance among dental and dental hygiene faculty and turnover intention https://doi.org/10.1002/jdd.13828
- Kabir, H.; Chowdhury, S.R.; Tonmon, T.T.; et al. (2022). Workplace violence and turnover intention among the Bangladeshi female nurses after a year of pandemic: an exploratory cross-sectional study https://doi.org/10.1371/journal.pgph.0000187
- Alshmemri, M.S. (2025). The relationship between quality of life and nurses' turnover intentions: a cross-sectional study https://doi.org/10.1155/jonm/4951493
- Du Plooy, J.; Roodt, G. (2010). Work engagement, burnout and related constructs as predictors of turnover intentions https://doi.org/10.4102/sajip.v36i1.910
- Du Plooy, J.; Roodt, G. (2013). Biographical and demographical variables as moderators in the prediction of turnover intentions https://doi.org/10.4102/sajip.v39i1.1070
- Takawira, N.; Coetzee, M.; Schreuder, D. (2014). Job embeddedness, work engagement and turnover intention of staff in a higher education institution: an exploratory study https://doi.org/10.4102/sajhrm.v12i1.524
- Bothma, C.F.C.; Roodt, G. (2012). Work-based identity and work engagement as potential antecedents of task performance and turnover intention https://doi.org/10.4102/sajip.v38i1.893
Record notes
Overall confidence is moderate and uneven. The TIS-6's genuine differentiator is organisational-criterion evidence: a large, prospective, census-based study that predicted actual turnover behaviour (leavers versus stayers) at four-month and four-year horizons, which is rare among self-report workplace scales, though it rests on a single South African ICT company and is unreplicated on observed turnover. Internal consistency (alpha about 0.80 to 0.83) and single-factor structure are reasonably well supported, largely through non-UK and translated samples. Test-retest reliability, formal measurement invariance and responsiveness/MIC are all Absent or untested and are recorded as such rather than assumed. The reference-standard criterion field is a category consideration (no health gold standard exists for an intention construct) and is marked Not-applicable. Licence terms could not be confirmed in the latest review pass against the steward's current distribution policy: only search-result links to the AOSIS publication-policies page and the DOAJ record were reachable, not their verbatim licence text, so the licence is recorded as not verified in the latest review pass rather than asserted from the article's own CC BY notice (which rule 7 disallows as a source). The main honesty tension the v0.2 schema still made slightly awkward is that the instrument's strongest property, organisational-criterion prediction, is simultaneously well-designed and thin (one study), which the grade (Moderate) plus status (thin) pairing is used to convey; readers should note both fields together.