Construct Validity of the Braden Scale for Pressure Ulcer Assessment in Acute Care: A Structural Equation Modeling Approach
The Braden Scale is the most widely used pressure ulcer risk assessment system in the world. To investigate its construct validity using structural equation modeling (SEM), a secondary analysis of retrospective data of patients admitted to an acute care facility was conducted using the records of 2588 patients who were at risk for pressure ulcers and admitted between January 2013 and December 2013. Data were extracted to an Excel sheet and analyzed, including demographic characteristics (ie, patients age, gender, weight, and disease spectrum), as well as total Braden scores and subscale scores.
The SEM was set according to modification indices suggestion. The original Braden Scale model was supported by χ2(9) = 22.854, CFI = 0.902, GFI = 0.974, root mean square error of approximation (RMSEA) = 0.092, indicating inadequate model fit. After modification according to software indices, χ2(2) = 2.052, CFI = 0.999, GFI = 0.999, RMSEA = 0.020 indicated an acceptable fit of the model (final model). The factor loadings of 6 subscales were all significant (P <.001), with .147 for nutrition, .137 for activity, .167 for friction and shear, .825 for sensory perception, .626 for mobility, and .556 for moisture subscale. The nutrition, activity, and friction and shear subscales were corrected to examine their relationships with other Braden Scale subscales (nutrition with activity [φ -0.063], activity with friction/shear [φ 0.136], and nutrition (φ friction/shear [0.159]). The factor loadings ranged from -0.067 to 0.159. These findings suggest the original Braden Scale has inadequate construct validity for acute care patients and that new risk-predicting scales should be designed based on data mining. Second, according to the factor loadings in the SEM, the most important risk factor in the Braden Scale for this patient population is sensory perception, followed by mobility and moisture. This suggests practitioners should pay particular attention to pressure ulcer prevention when patients have limited sensory perception, mobility limitations, and/or when moisture status is less than optimal.
The Braden Scale for pressure ulcer risk assessment is comprised of 6 subscales: sensory perception, skin moisture, activity, mobility, nutrition, and friction and shear. The assessment instrument was developed utilizing a theoretical model and clinical experience by Bergstrom et al in 1984 and published in 1987.1,2 Subsequently, many studies have investigated its reliability and validity. Intraclass correlation coefficients for reliability range from 0.73 to 0.95 in nursing homes3 to 0.603 to 0.964 for hospitalized patients.4 In terms of validity, 1 meta-analysis5 of 9 publications revealed the pooled sensitivity, specificity, positive predictive value, and negative predictive values were 86%, 38%, 28%, and 93%, respectively, in long-term care; another meta-analysis6 involving 6070 hospitalized patients showed pooled sensitivity was 0.72 (95% CI 0.68, 0.75), pooled specificity was 0.81 (95% CI 0.80, 0.82), and the summary receiver-operating characteristic area under the curve was 0.84 (SE = 0.02). These results demonstrated the Braden Scale had validity in predicting pressure ulcer risk that was better than nursing judgment.
Although interrater reliability and predictive validity have been well investigated, construct validity (ie, whether a scale measures or correlates with the theorized scientific construct it purports to measure) for the Braden Scale has not been as well researched.
Structural equation modeling (SEM) is a family of statistical methods that includes confirmatory factor analysis, path analysis, and latent growth modeling. SEM is widely used in the social sciences because of its ability to isolate observational error from the measurement of latent variables such as pressure ulcer risk.7 SEM already has been used to test conceptual or theoretical models in nursing studies.8,9
The purpose of this study was to investigate the construct validity of Braden scales using SEM.
Design. A secondary analysis of data from the authors’ previous study,10 a retrospective analysis of consecutive patients (N = 2625, mean age 59.8 ± 16.5, 1601 men), was conducted to investigate the construct validity of the Braden Scale using SEM. Inclusion criteria stipulated patients should be 1) at risk for pressure ulcers (eg, received care in the intensive care unit [ICU] between January 2013 and December 2013, bedbound for at least 1 month, >60 years of age, and post cardiothoracic surgery and/or lengthy procedure [>3 hours]); 2) age ≥18 years; and 3) have had total Braden scores and subscores recorded and available in the medical records. Patients with missing Braden score data were excluded. The study was approved by the medical ethics committee of the authors’ hospital.
Setting and sample. This study was conducted in a 3000-bed teaching hospital. Each year, approximately 2000 to 3000 patients who are at risk for pressure ulcer development are treated in the ICU and neurology, geriatric, cardiac surgery, neurosurgery, and orthopedics departments.
Data collection procedures. Medical records were retrospectively reviewed for demographic characteristics (ie, patients age, gender, weight, and disease spectrum) and Braden total and subscale scores. If more than 1 Braden score was noted, the lowest score was used. These procedures were described previously.10
SEM analysis. The significance of the relationships between the exogenous latent variable (Braden total score, which indicated pressure ulcer risk) and endogenous variables (ie, Braden Scale subcores for sensory perception, moisture, activity, mobility, nutrition, friction and shear) and the amount of variance explained in the endogenous variables were examined. Maximum likelihood estimation (MLE, an approach that utilizes available parameters to determine the likelihood of an event) was to used for discrepancy estimation, and the saturated and independence model was used for computing fit measures with incomplete data. The SEM was set according to modification indices (MI). MI can provide suggestions for modifications that likely will result in a better fit as shown by lower chi-squared values. The proposed model was assessed by widely accepted fit measures, which included χ2 (χ2/df : <3 acceptable, <2 excellent), comparative fit index ([CFI] >0.90 acceptable, >0.95 excellent), goodness of-fit index ([GFI] >0.90 acceptable, >0.95 excellent), root mean square error of approximation ([RMSEA] <0.08 acceptable, <0.05 excellent), and Akaike information criterion ([AIC], the lower the better).11 Factor loading represented how well factors (subscales) explained the variable (pressure ulcer risk). AMOS version 6.0 software (SPSS Inc, Chicago, IL) was used for the SEM analysis.
Patient characteristics. Of the records of 2625 patients, 37 who were <18 years of age were excluded; no patients were excluded for missing Braden Scale/subscale information, leaving 2588 patients (1582 [61.1%] men, 1006 [38.9%] women; mean age 60.0 ± 15.6 [range 18–98] years) to be included in this study. The patients came from 7 clinical departments: neurosurgery, ICU, orthopedics, neurology, respiratory medicine, spine surgery, and cardiothoracic surgery. Mean Braden score was 15.4 ± 2.3 (range 6 to 22).
Model estimation. The original model was supported by χ2(9) = 22.854, CFI = 0.902, GFI = 0.974, RMSEA = 0.092, indicating inadequate model fit (see Table 1). The standardized parameter estimates are depicted in Figure 1 and Table 2. All of the factor loadings of subscales were significant (P <.001), with .193 for nutrition, .148 for activity, .276 for friction and shear, .776 for sensory perception, .694 for mobility, and .567 for moisture.
According to the recommendations generated by the MI suggestion, the category friction and shear was added (φ) to assess its relationship to the nutrition pathway in model 2. Then friction and shear φ activity pathway (model 3), mobility φ friction and shear pathway (model 4), moisture φ nutrition pathway (model 5), moisture φ activity pathway (model 6), activity φ nutrition pathway (model 7), mobility φ activity pathway (final model) were successively added. The model fit parameters improved with each model modification.
In the final model, χ2(2) = 2.052, CFI = 0.999, GFI = 0.999, and RMSEA = 0.020, indicating an acceptable fit of the model (see Table 1). The factor loadings of subscales also were significant (P <.001), with .147 for the nutrition subscale, .137 for the activity subscale, .167 for the friction and shear subscale, .825 for the sensory perception subscale, .626 for the mobility subscale, and .556 for the moisture subscale (see Figure 2 and Table 2). The Braden subscales nutrition, activity, and friction and shear were related with other subscales. The factor loadings ranged from -0.067 to 0.159.
The original Braden Scale showed inadequate model fit according to the fit measures (χ2(9) = 22.854, CFI = 0.902, GFI = 0.974, RMSEA = 0.092), and as such poor construct validity. Moreover, previous studies have shown the Braden Scale only has a moderate predictive validity level for pressure ulcer risk, which has been confirmed by systematic review and meta-analysis.4,5,12-14 In use for more than 30 years, the Braden Scale has been subject to suggestions to revise.12-14 In the past 2 decades, many data mining methods have been used successfully to predict risk with great success,15 implying pressure ulcer risk prediction will be more effective when data mining methods replace traditional theoretical models.
SEM also revealed an explanation for the inadequate validity of the Braden Scale. When the SEM of the Braden Scale was modified according to MI suggestion, the final model indicated an acceptable fit of the model (ie, the χ2(2) = 2.052, CFI = 0.999, GFI = 0.999, RMSEA = 0.020). In the final model, the nutrition, activity, and friction and shear subscales had complex correlations with other subscales. As interpreted, poor nutrition will increase friction and shear, increased activity will increase friction and shear, poor nutrition will decease activity, and low activity is related to limited mobility. These findings indicated nutrition, activity, and friction and shear were not independent risk factors for pressure ulcers.
In the final model, factor loading was 0.825 for the sensory perception subscale, followed by mobility subscale (0.626), moisture subscale (0.556), friction and shear (0.167), nutrition (0.147), and activity (0.137). The factor loadings of 3 subscales (sensory perception [0.825], mobility [0.626], and moisture [0.556]) were >0.5, and the other 3 subscales (friction and shear [0.167], nutrition [0.147], and activity [0.137]) were <0.2, indicating sensory perception, mobility, and moisture are 3 important subscales in the Braden scale. These findings suggest practitioners should pay more attention to pressure ulcer prevention, especially when patients have compromised sensory perception, mobility, and moisture status.
This study has some limitations. First, it was a retrospective study. Shortcomings included recall bias and missing data points. Second, it was a single-center study, and pressure ulcer incidence can vary among different medical centers. Third, it was a hospital-based study. The construct validity of the Braden Scale in other facilities should be assessed. Prospective multicenter cohorts are needed to confirm the current conclusion.
A retrospective review and analysis of patient pressure ulcer data yielded 2 conclusions. First, the original Braden Scale has inadequate construct validity because the Braden Scale was based on a theoretical model. If data mining was used for building the pressure ulcer risk assessment model, the validity will be increased. Second, according to the factor loadings in the SEM, the most important risk factor in the Braden Scale is sensory perception, followed by mobility and moisture. The practitioner should pay particular attention to pressure ulcer prevention, particularly when patients have compromised sensory perception, mobility, and moisture status.
1. Bergstrom N, Braden BJ, Laguzza A, Holman V. The Braden Scale for Predicting Pressure Sore Risk. Nurs Res. 1987;36(4):205–210.
2. Bergstrom N, Demuth PJ, Braden BJ. A clinical trial of the Braden Scale for Predicting Pressure Sore Risk. Nurs Clin North Am. 1987;22(2):417–428.
3. Kottner J, Dassen T. An interrater reliability study of the Braden scale in two nursing homes. Int J Nurs Stud. 2008;45(10):1501–1511.
4. Wang LH, Chen HL, Yan HY, et al. Inter-rater reliability of three most commonly used pressure ulcer risk assessment scales in clinical practice. Int Wound J. 2015;12(5):590–594.
5. Wilchesky M, Lungu O. Predictive and concurrent validity of the Braden scale in long-term care: a meta-analysis. Wound Repair Regen. 2015;23(1):44–56.
6. Park SH, Choi YK, Kang CB. Predictive validity of the Braden Scale for pressure ulcer risk in hospitalized patients. J Tissue Viability. 2015;24(3):102–113.
7. Marsh HW, Morin AJ, Parker PD, Kaur G. Exploratory structural equation modeling: an integration of the best features of and confirmatory factor analysis. Ann Rev Clin Psychol. 2013;10:85–110.
8. Mallidou AA, Cummings GG, Estabrooks CA, Giovannetti PB. Nurse specialty subcultures and patient outcomes in acute care hospitals: a multiple-group structural equation modeling. Int J Nurs Stud. 2011;48(1):81–93.
9. Kang X, Dennison Himmelfarb CR, Li Z, Zhang J, Lv R, Guo J. Construct validity of the Chinese version of the Self-care of Heart Failure Index determined using structural equation modeling. J Cardiovasc Nurs. 2015;30(3):222–228.
10. Chen HL, Cao YJ, Wang J, Huai BS. A retrospective analysis of pressure ulcer incidence and modified Braden Scale score risk classifications. Ostomy Wound Manage. 2015;61(9):26–30.
11. Raykov T, Tomer A, Nesselroade JR. Reporting structural equation modeling results in psychology and aging: some proposed guidelines. Psychol Aging. 1991;6(4):499–503.
12. Chen HL, Shen WQ, Liu P. A meta-analysis to evaluate the predictive validity of the Braden Scale for Pressure Ulcer Risk Assessment in long-term care. Ostomy Wound Manage. 2016;62(9):20–28.
13. Park SH, Lee HS. Assessing predictive validity of pressure ulcer risk scales — a systematic review and meta-analysis. Iran J Public Health. 2016;45(2):122–133.
14. Park SH, Lee YS, Kwon YM. Predictive validity of pressure ulcer risk assessment tools for elderly: a meta-analysis. West J Nurs Res. 2016;38(4):459–483.
15. Bagherzadeh-Khiabani F, Ramezankhani A, Azizi F, Hadaegh F, Steyerberg EW, Khalili D. A tutorial on variable selection for clinical prediction models: feature selection methods in data mining could improve the results. J Clin Epidemiol. 2016;7:76–85.
Potential Conflicts of Interest: This work is supported by Nantong Municipal Science and Technology Bureau (grant number: BK2013014).
Dr. Chen is an associate professor, School of Nursing, Nantong University, Nantong City, Jiangsu Province, PR China. Dr. Cao is a director of nursing, Qilu Hospital of Shandong University, Jinan City, Shandong Province, PR China. Dr. Shen is an associate professor, School of Nursing, Nantong University. Mr. Zhu is Chief Physician, Department of Hepatobiliary Surgery, The Second Affiliated Hospital of Nantong University and the First People’s Hospital of Nantong City, Jiangsu Province, PR China. Please address correspondence to: Bin Zhu, Department of Hepatobiliary Surgery, The Second Affiliated Hospital of Nantong University and the First People’s Hospital of Nantong City, Haier Lane Road, No.6, Nantong City, Jiangsu Province, PR China; email: firstname.lastname@example.org.