A Meta-analysis to Evaluate the Predictive Validity of the Braden Scale for Pressure Ulcer Risk Assessment in Long-term Care

Although it is among the most commonly used pressure ulcer risk assessment tools, the Braden Scale may lack strong predictive validity when used in the long-term care setting. A meta-analysis was conducted of English-language articles published in the PubMed database and Web of Science from the indices’ inception through July 2015 to assess the predictive validity of the Braden Scale for pressure ulcers in long-term care residents.Search terms included pressure ulcer, pressure sore, bedsore, decubitus, long-term care, nursing home, skilled nursing facility, hospice, and Braden. Data extracted from the publications included sample and setting characteristics and predictive value indices. The pooled sensitivities, specificities, diagnostic odds ratios (DOR), and constructed summary receiver operating characteristic (SROC) curves were calculated. Eight studies (2 prospective cohorts and 6 cross-sectional studies) with 41 489 residents met selection criteria for inclusion in the analysis. The pooled sensitivity and specificity were 0.80 (95% CI: 0.79-0.81) and 0.42 (95% CI: 0.42-0.43), respectively, yielding a combined DOR of 5.66 (95% CI: 3.77-8.48). The area under the ROC curve (AUC) was 0.7686 ± 0.0478 (95% CI: 0.6749-0.8623), and the overall diagnostic accuracy (Q*) was 0.7090 ± 0.0402 (95% CI: 0.6302-0.7878). Significant heterogeneity was noted among the included studies; Q value was 302.54 (P = 0.000), and I² for pooled sensitivity, pooled specificity, and pooled DOR was 97.4%, 98.7% and 96.4%, respectively. Meta-regression analysis showed no heterogeneity was noted among Braden scale cut-offs (P = 0.123) and pressure ulcer prevalence P = 0.547). The evidence showed the Braden Scale has moderate predictive validity and low predictive specificity for pressure ulcers in long-term care residents. The development and testing of new risk assessment scales for this population is warranted.

Pressure ulcers are common in long-term care residents.¹ According to the 2004 National Nursing Home Survey² (NCHS), 159,000 nursing home residents (11%) in the United States had a pressure ulcer, most commonly Stage II. According to a case review,³ long-term care residents with pressure ulcers suffer pain, disfigurement, and decreased quality of life, and their risk of illness and death increases. A retrospective cohort study⁴ (N = 1539) shows residents in long-term care with pressure ulcers had a relative risk for dying of 1.45 (95% confidence interval [CI]: 1.30-1.65) as compared to those without ulcers after adjusting for clinical and functional status. The same study also shows litigation involving pressure ulcers is costly; residents realized some type of recovery against the facility in 87% of the cases (verdicts for the resident plus settlements) and were awarded amounts as high as $312 million in damages. 

Risk assessment is the first step in pressure ulcer prevention. The Braden Scale, the most widely used risk assessment scale, evaluates skin breakdown in 6 domains: sensory perception, moisture, activity, mobility, nutrition, and friction and shear.⁵ A systematic review⁶ showed the Braden Scale has optimal validation and the best sensitivity/specificity balance (57.1%/67.5%, respectively) and accurately predicts pressure ulcer risk (odds ratio = 4.08, CI 95%: 2.56-6.48). Many studies have examined the use of the Braden Scale for pressure ulcer prevention in long-term care. In a prospective cohort study, de Souza et al⁷ found the Braden Scale to have good predictive validity in elderly, long-term care residents, with sensitivity of 75.9% and 74.1%, specificity of 70.3% and 75.4% in cutoff scores of 18 and 17, and area under the receiver operating characteristic curve (AUC-ROC) of 0.79 and 0.81 at the first and last assessments. Braden and Bergstrom⁸ also reported a good predictive validity in long-term care, with sensitivity of 79%, specificity of 74%, 54% predictive value of a positive test, 90% predictive value of a negative test, and 75% correct classification rate. However, a prospective cohort study by Ramundo⁹ found the Braden Scale has limited predictive ability in long-term care. At a score of 18, the sensitivity of the tool was 100%, but the specificity was only 34%. The sample size in these studies was small; thus, use of the Braden Scale in long-term care remains controversial. 

The aim of this meta-analysis was to assess the predictive validity of the Braden Scale for pressure ulcers in long-term care residents.

Search strategy. Two (2) databases, PubMed and Web of Science, were searched for relevant English-language entries from database inception through July 2015. The search terms included pressure ulcer, pressure sore, bedsore, decubitus, long-term care, nursing home, skilled nursing facility, hospice, and Braden. The search strategy in PubMed was (“pressure ulcer” [MeSH Terms] OR “pressure ulcer*” [title/abstract] OR “pressure sore*” [title/abstract] OR “bedsore*” [title/abstract] OR “decubitus” [title/abstract]) AND (“long-term care” [title/abstract] OR “nursing home” [title/abstract] OR “skilled nursing facility” [title/abstract] OR “hospice” [title/abstract]) AND “Braden” [TW]. The search strategy in Web of Science was TS= (pressure ulcer* OR pressure sore* OR bedsore* OR decubitus) AND TS = (long-term care OR nursing home OR skilled nursing facility OR hospice) AND TS = Braden. The “related articles” function in PubMed and the “citing articles” or “cited articles” function in Web of Science also were utilized to broaden the search. 

Inclusion and exclusion criteria. Studies were considered for inclusion when they met the following criteria: 1) they discussed assessing the predictive validity of Braden Scale for pressure ulcer risk in long-term care residents; 2) they provided true positive (TP), false positive (FP), false negative (FN), and true negative (TN) for predicting pressure ulcer risk at different Braden Scale cut-offs; 3) they provided sensitivity and specificity, positive likelihood ratio, and negative likelihood ratio, and TP, FP, FN, and TN were calculated using a web-based program (ie, diagnostic test calculator, available at: https://araw.mede.uic.edu/cgi-bin/testcalc.pl). Two (2) authors independently judged study eligibility while screening the citations. Disagreements were resolved by a third author.        Data extraction and quality assessment. A spreadsheet developed by the authors was used to extract and record data on first author, year of publication, country in which the study was carried out, study design, year of participant enrollment, location, number of residents, gender and age of the participants, when in the pressure ulcer process the Braden Scale was used, pressure ulcer prevalence, Braden Scale cut-off, and the predictive TP, FP, FN, TN.

The Quality Assessment of Diagnostic Accuracy Studies¹⁰ (QUADAS) tool was used to assess the quality of the predictive validity studies. The QUADAS included 14 items. Each item of QUADAS is answered with “yes”, “no”, or “unclear”. Two (2) authors independently assessed study quality. Disagreements were resolved by a third author.

Statistical analysis. The overall pooled sensitivity, specificity, and diagnostic odds ratio (DOR), with 95% CI, were estimated using DerSimonian and Laird’s¹¹ random-effects model. In addition, summary receiver operator characteristic (SROC) analysis was performed to examine the interaction between sensitivity and specificity and to quantify test performance using the area under the curve (AUC) and overall diagnostic accuracy (Q*) value.¹² Heterogeneity was analyzed by Cochran’s Q test and I² statistic; a P value <0.05 by Cochran’s Q test and I² >50% indicated substantial heterogeneity. Meta-regression analysis was conducted to explore the heterogeneity. All analyses were performed using Meta DiSc 1.4 (version 0.6; accessed at www.hrc.es/investigacion/metadisc_en.htm).¹³ 

Eligible studies. Eight (8) cross-sectional or prospective cohort studies^7,8,14-19 (total sample 41 489 residents) met inclusion criteria for meta-analysis. Figure 1 shows the flow diagram of study selection. The included studies were conducted in 7 countries: Brazil, The Netherlands, Germany, the United States, Canada, Saudi Arabia, and Australia. The participants were residents of nursing homes, long-term care facilities, tertiary care hospitals, veterans’ administration medical centers, and skilled nursing facilities. The pressure ulcer diagnoses all were based on the National Pressure Ulcer Advisory Panel²⁰ (NPUAP) guide and included Stage I through Stage IV. Pressure ulcer prevalence ranged from 6.4% to 30.1%. The Braden scale cutoff ranged from 17 to 20. The details of all 8 studies are shown in Table 1. The quality assessment results according to the QUADAS list for the individual studies can be found in Table 2. All 8 included studies met 80% “yes” in all 14 items.

Predictive validity in long-term care facilities. The predictive validities of the included studies are listed in the Table 3. The pooled sensitivity of the studies was 0.80 (95% CI: 0.79-0.81; χ² (11)= 420.74, P = 0.000), shown graphically in the forest plot in Figure 2A. The pooled specificity was 0.42 (95% CI: 0. 42-0.43; χ² (11) = 859.80, P = 0.000) (see Figure 2bB), and the pooled DOR was 5.66 (95% CI: 3.77 to 8.48) (see Figure 2C). The overall weighted AUC was 0.7686 ± 0.0478 (95% CI: 0.6749-0.8623), and the Q* value was 0.7090 ± 0.0402 (95% CI: 0.6302-0.7878). The SROC curve is shown in Figure 2D. 

Study heterogeneity. The Cochran’s Q test value was 302.54 (P = 0.000). I² for pooled sensitivity, pooled specificity, and pooled DOR was 97.4%, 98.7% and 96.4%, respectively. The results showed significant heterogeneity among the studies; meta-regression analysis showed heterogeneity between Braden scale cut-offs (P = 0.123) and pressure ulcer prevalence (P = 0.547).

A meta-analysis found the overall weighted AUC for Braden Scale assessment in long-term care was 0.7686 ± 0.0478 (95% CI: 0.6749-0.8623). An AUC value of 0.5 indicates the test has no discriminatory ability, whereas an AUC value of 1.0 indicates perfect diagnostic capability.²¹ The SROC curve was used in meta-analysis for diagnostic accuracy. An AUC of 0.97 or above indicates excellent accuracy, 0.93 to 0.96 is good and 0.75 to 0.92 is moderate, but an AUC <0.75 suggests obvious deficiencies in diagnostic accuracy.²² Thus, the results of this study indicate the Braden Scale has moderate predictive validity for pressure ulcers in long-term care residents. However, the predictive specificity is low, with the pooled specificity 0.42 (95% CI: 0.42-0.43), suggesting some deficiency of the Braden Scale for pressure ulcer risk assessment in long-term care residents. This result was the same as the study results reported by Ramundo.⁹ In that prospective cohort study involving 48 home care residents, the optimal predictive sensitivity was 100% and the specificity was 34% when Braden scale at cut-off was 18.

The Braden Scale evaluates skin breakdown in 6 domains: sensory perception, moisture, activity, mobility, nutrition, and friction and shear. In the 2004 NCHS,² some characteristics were found to increase the risk for pressure ulcer development in long-term care residents: age 64 years and younger, duration <1 year, and a history of recent weight loss. In a prospective cohort study conducted among 175 nursing home residents, Akca et al²³ found residents who were underweight, had lived at the rehabilitation center for a longer time, and followed feeding regimens had a higher risk of developing pressure ulcers (P <0.001). In a prospective cohort study involving 346 residents, Kwong et al²⁴ reported bedbound or chairbound residents, especially persons with comorbidities (renal failure and stroke), who lived in nursing homes where more nursing assistants than nurses provided care were at higher risk for pressure ulcer development. However, these risk factors are not included in the Braden Scale. 

The Braden Scale design was based on expert consensus.⁵ A recent data analysis of 6 pressure ulcer prevalence studies (2004–2009)²⁵ used Classification Tree Analysis (CHAID) to show all Braden Scale items are not equally important; for residents in long-term care facilities, friction and shear had the strongest association with pressure ulcer prevalence. 

Because a new risk assessment scale for long-term care residents may have better predictive validity than the Braden Scale, some new scales have been designed for this population. The interRAI Pressure Ulcer Risk Scale (interRAI PURS) is based on a logistic model and includes 7 independent variables: impaired in-bed mobility, impaired in walking, bowel incontinence, weight loss, history of resolved pressure ulcers, and shortness of breath. This scale has been shown in a retrospective cohort study²⁶ to have good distributional characteristics, and the c-statistic (a measure of discrimination among models) was 0.708, compared with a c-statistic 0.676 for the Braden scale. However, these results need to be confirmed by other large-sample prospective studies.

The current meta-analysis has some limitations. First, significant heterogeneity was found among the studies. Meta-regression analysis showed the heterogeneity among the studies may be the consequence of different Braden scale assessment times (at admission, 48 to 72 hours after admission, and pre-breakdown), different facilities (long-term care, nursing home, and skilled nursing facility), different study enrollment times, and different countries, among other reasons. Second, because the included studies provided limited information, subgroup analysis according to different Braden scale assessment time, facilities, enrollment time, and country were not conducted. These limitations should be considered when evaluating the results.

A meta-analysis showed the Braden Scale has moderate predictive validity and low predictive specificity for pressure ulcers in long-term care residents. Additional research regarding more suitable scales for predicting pressure ulcer risk in this population are warranted.

Dr. Chen, Dr. Shen, and Dr. Liu are associate professors, School of Nursing, Nantong University, Nantong, Jiangsu, PR China.

Please address correspondence to: Peng Liu, MD, Nantong University, Qi Xiu Road 19# Nantong City, Jiangsu Province. 226001 China PR: email: pphss@126.com.