From: Dimensionality of the Pittsburgh Sleep Quality Index: a systematic review
Author and year of publication | Software | Extraction method | Types of Modification index used | Correlation between factors | Standardized Factor loadings | Factors in final model; same/different from EFA | Number of models used in comparative CFA | Reason for the selection of models in comparative CFA | Model fit indices |
---|---|---|---|---|---|---|---|---|---|
Aloba et al. 2007 [31] | NO CFA | – | – | – | – | – | – | – | – |
Babson et al. 2012 [30] | NO CFA | – | – | – | – | – | – | – | – |
Burkhalter et al. 2010 [29] | Mplus version 5.21 | Not reported | Path diagram change | 0.532, 0.773, 0.801 | F1 DURAT = 0.85, HSE = 0.98, SLPQUAL = − 0.51 F2 SLPQUAL = 1.09, LATEN = 0.68, MEDS = 0.92 F3 DISTB = 0.93, DAYDYS = 0.56 | 3, No EFA | 3; 1F-1 3F-2 | Not explained | Non-significant p value of χ2; RMSEA< 0.08–0.05; CFI > 0.95; WRMR < 0.90. |
Buysse et al. 2008 [28] | NO CFA | – | – | – | – | – | – | – | – |
Casement et al. 2012 [35] | Mplus version 5.1 | Mean and variance-adjusted weighted least squares (WLSMV) estimator | Not reported | 0.46, 0.77, 0.81 | F1 DURAT = 0.87, HSE = 0.75 F2 SLPQUAL = 0.75, LATEN = 0.56, MEDS = 0.45 F3 DISTB = 0.74, DAYDYS = 0.43 | 3, No EFA | 4; 1F-1 2F-2 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | χ2/ df < 3, RMSEA < 0.06, WRMR < 0.90, CFI ≥ 0.95, and TLI ≥ 0.96 |
Chong & Cheung 2012 [34] | Mplus version 5 | Not reported | Not reported | 0.522, 0.567, 0.641 | F1 DURAT = 0.73/0.85/0.95, HSE = 0.76/0.84/0.78 F2 SLPQUAL = 0.81/0.59/0.63, LATEN = 0.64/0.64/0.70, DISTB = 0.59/0.40/0.47, DAYDYS = 0.44/0.21/49, MEDS = 0.33/0.35/0.17 | 2, No EFA | 9; 1F-1 2F-6 3F-2 | Partially explained, some of the documented models not used, no reasons given for their omission | SRMR< 0.05; RMSEA < 0.07; CFI > 0.95 |
Cole et al. 2006 [22] | Not reported | Maximum likelihood extraction on the covariance matrix, & multivariate non-normality smoothed by bootstrapping | Lagrange Modification index with change in path diagram | 0.42, 0.82, 0.75 | F1 DURAT = 0.76, HSE = 0.91 F2 SLPQUAL = 0.89, LATEN = 0.67, MEDS = 0.43 F3 DISTB = 0.67, DAYDYS = 0.52 | 2, 3 | 2; 2F-1 3F-1 | Comparison between originally proposed 1F model & outcome of EFA Fit indices for 1F model not reported | RMSEA≤0.06; CFI ≥ 0.90; GFI ≥ 0.90; AGFI≥0.90; LOWER χ2, BIC (difference of at least 10 between two models) |
Gelaye et al. 2014 [44] | Stata version 12.0 software | Maximum likelihood estimation | Not reported | 0.46, 0.26, 0.36, (0.53, 0.40, 0.10) | F1 DURAT = 0.79/0.73/1.0/0.6, HSE = 0.43/0.78/0.21/0.57 F2 SLPQUAL = 0.81/0.58/0.61/0.67, LATEN = 0.47/0.35/0.34/0.53, DISTB = 0.47/0.51/0.54/0.38, DAYDYS = 0.49/0.51/0.5/0.39, MEDS = 0.25/0.25/0.14/0.28 | 2, 2, 2, 3, same | Not performed | Not explained | SRMR ≤0.08; RMSEA ≤0.06; CFI ≥0.95 |
Hita-Contreras et al. 2014 [43] | NO CFA | – | – | – | – | – | – | – | – |
Ho et al. 2014 [42] | Mplus version 7.11 | Robust maximum likelihood estimator | Error-term correlation | Not applicable | F1 DURAT = 0.59, HSE = 0.60, SLPQUAL = 0.84, LATEN = 0.61, DISTB = 0.61, DAYDYS = 0.56, MEDS = 0.36 | 1, same | 4; 1F-2 2F-1 3F-1 | Partially explained, some of the documented models not used, no reasons given for their omission | Insignificant χ2-test; CFI & TLI ≥0.95; RMSEA≤0.06; SRMR≤0.08; Lower BIC |
Jiménez-Genchi et al. 2008 [27] | NO CFA | – | – | – | – | – | – | – | – |
Jomeen & Martin 2007 [26] | Mplus version 3 | Weighted least-square with mean and variance correction estimator (WLSMV) | Not reported | Not reported | not reported | 2, No EFA | 7; 1F-1, 2F-6 | Not clear | CFI & TLI > 0.90, RMSEA< 0.08–0.05, WRMR< 0.90 & Insignificant χ2 |
Koh et al. 2015 [41] | FactoMineR in R | Not reported | Not reported | (0.27, 0.64, 0.89); (0.39, 0.72, 0.90) in 2 sample groups | F1 DURAT = 0.68/0.60, HSE = 0.72/0.67 F2 SLPQUAL = 0.72/0.63, LATEN = 0.63/0.60 F3 DISTB = 0.37/0.52, DAYDYS = 0.51/0.42, MEDS = 0.40/0.26 | 3/3, 3/3, same | 4; 1F-1 2F-1 3F-2 | Not explained | GFI > 0.90; AGFI> 0.90; CFI ≥ 0.95 RMSEA < 0.08–0.05; LOWER χ2, BIC (difference of at least 10 between two models), CAIC |
Kotronoulas et al. 2011 [25] | NO CFA | – | – | – | – | – | – | – | – |
Lequerica et al. 2014 [40] | SPSS Statistics 21 with AMOS | Not reported | Not reported | 0.87, 0.85 | F1 DURAT = 0.68, HSE = 0.51, LATEN = 0.68 F2 DISTB = 0.73, DAYDYS = 0.66, MEDS = 0.25 | 2, same | 5; 1F-1 2F-3 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | Non-significant p value of χ2; CFI ≥ 0.95; NNFI≥0.95 RMSEA < 0.06 |
Magee et al. 2008 [24] | SPSS version 15 with AMOS version-7 | Not reported | Not reported | 0.73 | F1 DURAT = 0.68, HSE = 0.62 F2 SLPQUAL = 0.76, LATEN = 0.61, DISTB = 0.46, DAYDYS = 0.52, MEDS = 0.23 | 2, different | 6; 1F-2 2F-2 3F-2 | Partially explained, some of the documented models not used, no reasons given for their omission | χ2-test lower, non-significant values; RMSEA ≤0.05; CFI, GFI, & AGFI > 0.90 |
Manzar et al. 2016a [17] | SPSS 16.0 with amos | Maximum likelihood extraction with bootstrapping to smooth non-normality | Not reported | Not applicable | F1 DURAT = 0.74, HSE = 0.32, SLPQUAL = 0.74, LATEN = 0.63, DISTB = 0.43, DAYDYS = 0.41, MEDS = 0.40 | 1, 2 different | 2; 1F-1 2F-1 | Comparison between outcome(s) of EFA | Non-significant Bollen–Stine bootstrap χ2 p values, Non-significant p value of χ2; χ2/df < 2; RMR ≤ 0.05; CFI ≥ 0.95; RMSEA < 0.05; GFI & AGFI> 0.9; AIC = lesser value indicated a better fit |
Manzar et al. 2016b [15] | SPSS 16.0 with amos | Maximum likelihood extraction | Co-variance, Variance and regression weights | Not applicable | F1 DURAT = 0.363, HSE = 0.374, SLPQUAL = 0.705, LATEN = 0.633, DISTB = 0.501, DAYDYS = 0.406, MEDS = 0.30 | 1, No EFA | 17; 1F-3 2F-8 3F-6 | Most of models of the PSQI reported till 15–02-2015 | Non-significant p value of χ2; χ2/df < 2; RMR ≤ 0.05; CFI ≥ 0.95; RMSEA < 0.05; GFI & AGFI> 0.9; AIC = lesser value indicated a better fit |
Mariman et al. 2012 [33] | SPSS (PASW 17.0) with AMOS module (5.0) | Maximum Likelihood Algorithm | Not reported | 0.64, 0.53, 1.00 | F1 DURAT = 0.9, HSE = 0.78 F2 SLPQUAL = 0.85, LATEN = 0.57, MEDS = 0.18 F3 DISTB = 0.79, DAYDYS = 0.29 But, 3 latent factors shown to load on 1 factor | Second order model, No EFA | 3; 1F-1 2F-1 3F-1 Results for the 2F model not shown | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | Non-significant p value of χ2 (d.f.); GFI > 0.90; AGFI> 0.85; CFI > 0.90; RMSEA< 0.08–0.05; Lower CAIC |
Nazifi et al. 2014 [39] | NO CFA | – | – | – | – | – | – | – | – |
Nicassio et al. 2014 [38] | EQS 6.1 | Maximum likelihood (ML) method | Not reported | 0.65 | F1 DURAT = 0.85, HSE = 0.64 F2 SLPQUAL = 0.89, LATEN = 0.48, DISTB = 0.57, DAYDYS = 0.56 | 2, No EFA | 3; 1F-1 2F-1 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | S-Bχ2; an S-Bχ2/df < 2.0; robust CFI ≥ 0.95; RMSEA≤0.05; Lower & negative AIC |
Otte et al. 2013 [32] | LISREL 8.8 | Weighted least squares | Error term correlation | 0.37, 0.71 in 2 sample groups | F1 DURAT = 0.64, HSE = 0.97 F2 SLPQUAL = 0.86, LATEN = 0.82/0.66, DISTB = 0.66, DAYDYS = 0.5, MEDS = 0.46 | 2, No EFA | 4; 1F-1 2F-1 3F-2 Two 3F models differed with respect to use/non-use of error terms only | Not explained | Non-significant p value of χ2; SRMR ≤0.08; RMSEA< 0.06; CFI ≥ 0.95 |
Otte et al. 2015 [37] | LISREL version 8.8 | Weighted least-squares, none of the indicators showed excessive skew or kurtosis | Not reported | 0.40, 0.73, 0.68 | F1 DURAT = 0.92, HSE = 0.68 F2 SLPQUAL = 0.82, LATEN = 0.57, MEDS = 0.15 F3 DISTB = 0.61, DAYDYS = 0.61 | 3, No EFA | 7; 1F-1 2F-2 3F-3 4F-1 | Not explained | Non-significant p value of χ2; RMSEA< 0.06; CFI ≥ 0.95; |
Rener-Sitar et al. 2014 [46] | STATA version 12 | Diagonally weighted least squares (DWLS) and a “robust” method using the Huber-White sandwich estimator | Not reported | Not applicable | not reported | 1; same in both | Not applicable | Not applicable | SRMR: ≤0.08; RMSEA: ≤0.06; and CFI, TLI: ≥0.95 |
Skouteris et al. 2009 [23] | Structural equation modeling (SEM) | Not reported | Path diagram change | 0.44, 0.59 | F1 DURAT = 0.73/0.85, HSE = 0.91/0.94, LATEN = 0.36/0.39 F2 DISTB = 0.62/0.60, DAYDYS = 0.49/0.62 | Second order model, No EFA | 2; 2F-2 | Compared with model reported in similar population, i.e., pregnant women | CFI & GFI > 0.90–1.0; RMSEA< 0.10 - < 0.05; χ2/df of 2 to 3 (lower is better); lower ECVI |
Tomfohr et al. 2013 [36] | Mplus version 5.21 | Maximum likelihood estimation | Reported but detail is not clear | Not reported, distinct model with age & gender as co-variates | F1 DURAT = 0.71/0.82, HSE = 0.70/0.72 F2 SLPQUAL = 0.77/0.76, LATEN = 0.64/0.63 F3 DISTB = 0.64/0.70, DAYDYS = 0.56/0.61 | 3, No EFA | 3; 1F-1 3F-2 | Not explained | CFI ≥ 0.90; SRMR ≤0.05; χ2 test of difference (P ≤ 0.01) |
Zhong et al. 2015 [45] | SAS 9.4 | Weighted least squares (WLS) estimation | Not reported | 0.07, 0.36 | F1 DURAT = 0.66, HSE = 0.52 F2 SLPQUAL = 0.47, LATEN = 0.46, DISTB = 0.45, DAYDYS = 0.64 F3 MEDS = 0.48 SLPQUAL = 0.22, LATEN = 0.26 | 3, same | 5; 1F-1 2F-3 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | CFI ≥ 0.90; SRMR< 0.08; RMSEA < 0.06 |
De la Vega et al. 2015 [59] | Not reported | maximum likelihood mean adjusted | Not reported | Not applicable | SLPQUAL = 0.421 LATEN = 0.620 DURAT = 0.656 HSE = 0.567 DISTB = 0.606 DAYDYS = 0.485 | 1, No EFA | 2; 1F-1 2F-1 | Compared with model reported in similar population, i.e., adolescents | S-Bχ2, CFI, RMSEA; cut-off for the indices not reported |
Anandakumar et al. 2016 [67] | No CFA | – | – | – | – | – | – | – | |
Zheng et al. 2016 [51] | Not reported | Not reported | Not reported | 0.34 | F1 DURAT = 0.69 HSE = 0.65 MEDS = 0.15 F2 DISTB = 0.43 DAYDYS = 0.51 SLPQUAL = 0.721 LATEN = 0.63 | 2, No EFA | 4; 1F-1 2F-2 3F-1 | explained, some of the documented models not used, no reasons given for selection and/or inclusion | χ2, GFI, AGFI, RMR, RMSEA, CFI, NFI, NNFI, AIC, CAIC, SBC |
Becker & Jesus 2017 [53] | SPSS 21 and AMOS-29 | Not reported | – | – | F1 SLPQUAL = 0.59 LATEN = 0.76 F2 DURAT = 0.76 HSE = 0.69 F3 DISTB = 0.52 DAYDYS = 0.57 | 3, 2 different | 6; 1F-2 2F-2 3F-2 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | non-significant χ2, RMSEA ≤0.08, CFi, GFI & AGFI > 0.97 |
Benhayon et al. 2013 [61] | No CFA | – | – | – | – | – | – | – | – |
DeGutis et al. 2016 [62] | R | maximum likelihood estimation | Not reported | 0.76, 0.75, 0.45 | F1 HSE = 0.68 DURAT = 0.78 F2 LATEN = 0.70 SLPQUAL = 0.52 MEDS = 0.77 F3 DISTB = 0.56 DAYDYS = 0.78 | No EFA | 4; 1F-1 2F-2 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | χ2/df < 3, SRMR & RMSEA≤0.06, CFI & TLI > 0 .95 |
Yunus et al. 2016 [48] | SPSS 20 | weighted least squares method | Not reported | Not applicable | LATEN = 0.65 SLPQUAL = 0.65 DISTB = 0.49 | 1, No EFA | 4; 1F-2 2F-1 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | CFI, TLI, RMSEA, SRMR cut-off for the indices not reported |
Qiu et al. 2016 [58] | SAS 9.4 | weighted least squares (WLS) estimation | Error term correlation | 0.68 | F1 HSE = 0.48 DURAT = 0.45 LATEN = 0.44 SLPQUAL = 0.83 F2 DISTB = 0.62 DAYDYS = 0.49 | 2, same | 6; 2F-6 | None | CFI ≥ 0.90, SRMR≤0.08, RMSEA ≤0.06 |
Dudysova et al. 2017 [66] | Not reported | diagonally weighted least squares (DWLS) estimator | Not reported | 0.80, 0.30, 0.16 | F1 HSE = 0.68 DURAT = 0.88 F2 LATEN = 0.70 SLPQUAL = 0.79 MEDS = 0.89 F3 DISTB = 0.32 DAYDYS = − 0.29 | No EFA | 11; 1F-1 2F-6 3F-4 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | non-significant & lower, GFI > 0.90, CFI & TLI ≥0.95, RMSEA ≤0.05 (≤0.08 adequate fit), SRMR ≤0.08 |
Salahuddin et al. 2017 [16] | SPSS -16.0 | maximum likelihood | Error term correlation | Not applicable | Not reported | 1, 1–3 | 5; 1F-4 2F-1 | All based on EFA | RMR & RMSEA ≤0.05, GFI, AGFI ≥0.90, Lesser ECVI, CFI ≥ 0.95, χ2/df ≤ 3 |
João et al. 2017 [57] | SPSS-21.0 | No CFA | – | – | – | – | – | – | – |
Chen et al. 2017 [63] | R 3.1.1 and its package lavaan | Not reported | Used modification indices but details not mentioned | Not reported | Unstandardized loadings Reported | None, No EFA | 1; 3F-1 | Not applicable | CFI & TLI > 0.90, RMSEA < 0.08 |
Khosravifar et al. 2015 [51] | Not reported | Not reported | Not reported | Not reported | Not reported | 2 | 3; 1F-1 2F-1 3F-1 | Based on EFA | Not reported |
Fontes et al. 2017 [49] | STATA version, R, version 3.0.1 | Not reported | Correlation between the PSQI components | Not applicable | HSE = 0.44 DURAT = 0.53 LATEN = 0.54 SLPQUAL = 0.88 MEDS = 0.22 DISTB = 0.42 DAYDYS = − 0.37 | 1, 2 | 2; 1F-1 2F-1 | Based on EFA | non-significant χ2, χ2/df = 2–3, SRMR ≤0.08, RMSEA≤0.07, CFI & TLI ≥ 0.95 |
Guo et al. 2016 [60] | SPSS-22.0 with AMOS18.0 | Not reported | Error term correlation | Not reported | HSE = 0.47 DURAT = 0.52 LATEN = 0.41 SLPQUAL = 0.83 DISTB = 0.35 DAYDYS = − 0.60 | 2, No EFA | 6; 1F-2 2F-2 3F-2 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | χ2/df = 2–5, 0.05 < RMSEA < 0.08, CFI > 0.95, SRMR< 0.05 |
Morris et al. 2017 [65] | SPSS-22.0 | No CFA | – | – | – | – | – | – | – |
Passos et al. 2016 [52] | SPSS-20.0 with AMOS 23.0 | Not reported | Error term correlation | 0.17 | Unstandardized loadings Reported | 2–3, 2 | 3; 2F-2 3F-1 | Based on EFA | SRMR≤0.08, CFI > 0.95, 0.5 < RMSEA> 0.8 |
Zhu et al. 2018 [64] | Stata 13.1 | Maximum Likelihood Algorithm | Not reported | Not applicable | HSE = 0.81 DURAT = 0.75 LATEN = 0.61 SLPQUAL = 0.63 DISTB = 0.46 DAYDYS = − 0.43 | 1, No EFA | 3; 1F-2 3F-1 | Not explained, some of the documented models not used, no reasons given for selection and/or inclusion | non-significant χ2, RMSEA < 0.05, CFI > 0.95, lower BIC, SRMR< 0.06 |