Using the Braden subscales to assess risk of pressure injuries in adult patients: A retrospective case‐control study

Abstract

The aim of this study was to compare the pressure injury risk predictability between the individual Braden subscales and the total Braden scale in adult inpatients in Singapore. A retrospective 1:1 case‐control design was used from a sample of 199 patient medical records. Clinical data were collected from a local university hospital's medical records database. The results showed that, among the six subscales, the activity subscale was the most sensitive and specific in predicting pressure injury (PI). However, the overall results showed that the Braden scale remained the most predictive of PI development in comparison with the individual subscales. The study also found that, among the Singaporean patients, the Braden cut‐off score for PI risk was 17 compared with the current cut‐off score of 18. Therefore, it may be relevant for local tertiary hospitals to review their respective Braden cut‐off scores as the study results indicate an over‐prediction of PI risk, which leads to unnecessary utilisation of resources. The hospital may also consider developing a PI prevention bundle comprising commonly used preventive interventions when at least one Braden subscale reflects a suboptimal score.

1. INTRODUCTION

Pressure injuries (PIs) are defined as localised damages to the skin and/or underlying soft tissue, usually over a bony prominence, and develop as a result of intense and/or prolonged pressure or pressure in combination with shear.1 Implications of PI development may include serious medical complications such as secondary systemic infections or even death.2 Global incidences of PI range from 4.7% to 32.1%3 and are classified into four stages of severity according to international guidelines.4 PI occurrence is frequently cited as an outcome indicator for the quality of care provided by health care institutions.5, 6 High incidences of PI have been observed in environments where organisational strategies for prevention are inadequate or non‐existent.7 Therefore, it is necessary to ascertain more effective methods to recognise patients at risk of developing PIs as most are preventable on timely risk assessment and preventive measures.8

Risk assessment scales (RASs) are designed based on known risk factors to assess individuals' risk of PI.9 Cut‐off scores are used to distinguish between various risk categories, which are then used to guide planning of the type and intensity of preventative interventions.10 Given that patients have varying susceptibility to PIs, it is thus important to determine their risk through the use of RASs. Among the 40 validated RAS tools, including the Norton Scale and Waterlow Scale, the Braden scale has been shown to provide the best risk predictability of PI in adults.11, 12, 13, 14 The Braden subscales rate patients on six risk factors that are deemed significant in PI development, including: sensory perception, nutrition, activity, mobility, moisture, and friction and shear.10 However, the Braden scale has not been proven effective in reducing PI incidences.15, 16 This shortcoming may be because of the administration of interventions being recommended based on the total score instead of specific risk factors identified in the subscales.10, 11 A case study by Gadd and Morris17 demonstrated that patients who had low Braden subscale scores did not receive specific preventive interventions as their cumulative score indicated that they were not at PI risk.

Furthermore, there appears to be a weak correlation between the Braden scale scores and the implementation of appropriate preventive interventions.10, 18 Further analysis demonstrated that, most of the times, the preventive interventions implemented were not specific to the subscale scores, suggesting that nurses may have initiated preventive measures solely based on their clinical knowledge.10, 17, 19 Moreover, a randomised controlled trial conducted by Saleh et al16 observed no difference between the validated Braden scale and nurses' clinical judgement in terms of PI risk assessment suggesting the redundancy of the scale. This highlights the importance of viewing subscale scores as independent risk assessment tools as they are able to demonstrate the need for individualised preventive interventions even when the cumulative Braden score indicates otherwise.17, 20

Past studies have demonstrated that individual Braden subscales yield substantial predictability in PI development among adult patients.10, 21 For example, the mobility subscale was examined in various trials and was found to yield considerable construct validity, as well as sufficiency as a tool by itself to plan preventive interventions.22 Given the paucity of evidence that describes the use of Braden subscales as a risk assessment tool, it is timely to conduct a local study to examine the predictabilities of the individual subscales against the original Braden scale in PI risk prediction among adult patients in Singapore acute care settings. In 2017, the PI incidence in the study hospital was reported to be 0.22%.23 The Braden scale is used as a standard tool to assess PI risk in all patients to plan preventative and management interventions.

2. AIMS OF THE STUDY

The aims of the study were to:

1. examine which of the demographic and clinical data are the good predictors of PI risk;

2. examine the predictabilities of the Braden scale and each of the subscales;

3. examine the sensitivity and specificity of Braden scale and subscales at various cut‐off scores; and

4. determine the Braden scale cut‐off score for the Singaporean patients.

3. MATERIALS AND METHODS

A retrospective case‐control study was conducted using 199 patients' medical records in a public university hospital in Singapore. The sample was matched according to a 1:1 ratio to ensure comparability of various risk factors on the outcome.24, 25 The variables selected for matching included: age, gender, length of hospitalisation stay, stay in ICU/high‐dependency unit, and presence of surgery (including the types of surgery) throughout the patients' hospitalisation. Inclusion criteria of the medical records included patients above 18 years old on admission, admitted to the hospital for at least 24 hours, and discharged from 1 June 2015 to 31 March 2017. Patients who acquired a PI of at least Stage 1 severity during their hospitalisation were classified as a case for the study. For controls, patients who did not acquire any PI, incontinence‐associated dermatitis, or broken skin throughout their stay were included. Exclusion criteria of the medical records were patients who were admitted into the hospital with existing PIs or admitted into obstetrics and gynaecology units.

3.1. Data collection

Electronic medical records of the cases and controls were accessed through the Hospital Occurrence Report (eHOR hospital clinical systems). Two outcome measures were used in this study: the Braden scale and a specific data collection instrument to extract the relevant items in the medical records. The Braden scale was first developed in the United States and consists of six subscales: Sensory Perception, Moisture, Nutrition, Activity, Mobility, and Friction and Shear, which are scored according to the patients' conditions to measure their level of risk of PI development.26 The scores are classified into five risk categories—not at risk,19, 20, 21, 22, 23 at risk,15, 16, 17, 18 moderate risk,13, 14 high risk,10, 11, 12 and very high risk (≤9)—whereby the highest risk category is strongly associated with PI development.26 Typically, the recommended Braden cut‐off score in predicting PI development among adult patients is 18.9, 13 The scale has also yielded optimal validation among adult patients, presenting with sufficient sensitivity, specificity, and reliability.21, 27

To aid the retrieval of data, a data collection instrument from a study by Mordiffi28 was revised and used as a reference in this study. The items in the tool were constructed based on risk factors of PIs, as well as clinical knowledge related to the management of disease. It comprises of two separate forms. Form A consists of 17‐item demographic information (ie, age, gender, race, and smoking status) and admission assessments (ie, diagnosis, comorbidities, and Braden scores). Form B consists of 3‐item characteristics of hospital‐acquired PIs, inclusive of the location and staging of the PI as categorised by the International National Pressure Ulcer Advisory Panel (NPUAP)/European Pressure Ulcer Advisory Panel (EPUAP) Pressure Ulcer Classification System.4 The data collection tool was formatted in an ordinal manner (ie, 1‐Yes, 2‐No) to allow ease of data entry later on in the study.

3.2. Data analysis

Statistical analysis of the study data was performed using SPSS version 24.0.29 The demographic and health status‐related independent variables of the patients (age, length of hospitalisation stay, and length of surgery) were presented in the mean and SD using independent samples t test, while categorical variables (gender, smoking status, race, ICU stay, presence of surgery, clinical discipline, presence of comorbidities, types of comorbidities, level of consciousness on admission, and pre‐arrival setting) were presented in frequency and percentage using the chi‐square (χ ²) test. The frequency of the staging and location of PI occurrences for cases were also reported to provide a context of PI development in the study site. The effect of independent variables on the mean Braden summative score was assessed using independent t tests. The relationship between risk factors predictive of PI development and the Braden summative score were evaluated using binary logistic regression analysis.30 The odds ratio (OR), level of significance (P‐value), 95% confidence intervals (CI), and Cox and Snell's indexes were reported in the regression analysis. Diagnostic accuracy tests using sensitivity and specificity analyses were carried out to ascertain the optimal cut‐off scores for the Braden scale and its subscales.31 For all statistical analyses, the level of significance (α) was set at P ≤ 0.05.

4. RESULTS

4.1. Demographic and clinical characteristics

The mean age of cases was 68 (±17.1) years, whereas controls were 64 years old (±16.4). Of the cases, 60.0% were males, and 68.0% were Chinese. Within the controls, 59.6% were males, and 70.7% were Chinese. More than half of the samples were non‐smokers (cases = 70.7%; controls = 70.0%) (Table 1). From the following variables, only race yielded statistical significance (χ ² = 9.519, P = 0.049) in terms of PI development (Table 1).

Table 1.

Demographic and clinical profiles of cases and controls (n = 199)

		Outcome		Statistics	P‐value
		Cases (n = 100) n (%, SD)	Controls (n = 99) n (%, SD)
Gender	Male	60 (60.0%)	59 (59.6%)	χ 2 = 0.003	0.954
	Female	40 (40.0%)	40 (40.4%)
Race	Chinese	68 (68.0%)	70 (70.7%)	χ 2 = 9.519	0.049 *
	Malay	24 (24.0%)	11 (11.1%)
	Indian	5 (5.0%)	10 (10.1%)
	Othersa	3 (3.0%)	8 (8.1%)
Smoking status	Non‐smoker	70 (70.0%)	70 (70.7%)	χ 2 = 3.000	0.392
	Current smoker	13 (13.0%)	15 (15.2%)
	Ex‐smoker	17 (17.0%)	12 (12.1%)
	Missing data	0 (0.0%)	2 (2.0%)
Age		67.99 (17.084)	64.22 (16.446)	t = −1.585	0.115
Length of hospitalisation stay (days)		68.45 (±101.008)	35.86 (±40.011)	t = −2.998	<0.001**
ICU stay	Yes	25 (25.0%)	24 (24.2%)	χ 2 = 0.015	0.901
	No	75 (75.0%)	75 (75.8%)
Presence of surgery	Yes	42 (51.2%)	40 (48.8%)	χ 2 = 0.052	0.819
	No	58 (49.6%)	59 (50.4%)
Length of surgery (hours)		1.25 (±2.268)	0.90 (±1.1410)	t = −1.322	0.188
Clinical discipline	Cardiac/cardiothoracic vascular surgery	16 (53.3%)	14 (46.7%)	χ 2 = 5.835	0.756
	Respiratory	9 (42.9%)	12 (57.1%)
	Gastroenterology	9 (64.3%)	5 (35.7%)
	Nephrology	4 (40.0%)	6 (60.0%)
	Oncology/haematology	9 (42.9%)	12 (57.1%)
	Neurology/neurosurgery	13 (41.9%)	18 (58.1%)
	General medicine	16 (50.0%)	16 (50.0%)
	General surgical/trauma	15 (62.5%)	9 (37.5%)
	Orthopaedics	6 (50.0%)	6 (50.0%)
	Othersb	2 (66.7%)	1 (33.3%)
Level of consciousness on admission	Awake	72 (45.3%)	87 (54.7%)	χ 2 = 12.158	0.007*
	Confused	4 (80.0%)	1 (20.0%)
	Drowsy	8 (53.3%)	7 (46.7%)
	Unresponsive	16 (84.2%)	3 (15.8%)
Presence of comorbidities	Yes	68 (68.0%)	87 (12.1%)	χ 2 = 46.325	<0.001**
	No	0 (0.0%)	12 (12.1%)
Pre‐arrival setting	Home	56 (56.0%)	52 (52.5%)	χ 2 = 15.796	0.003*
	Nursing home/community hospital	5 (5.0%)	4 (4.0%)
	Acute hospitals/clinic	6 (6.0%)	23 (23.2%)
	Othersc	14 (14.0%)	13 (13.1%)
	Not documented	19 (19.0%)	7 (7.1%)

^* Significant at P < 0.05.

^** Significant at P < 0.01.

^a Others include Eurasian, Arabic, Filipino, and Bangladeshi.

^b Others include Psychiatry and ENT.

^c Include from workplace and public areas.

Length of hospitalisation stay (P < 0.001, t = −2.998), level of consciousness (P = 0.007, χ ² = 12.158), presence of comorbidities (P < 0.001, χ ² = 46.325), and pre‐arrival setting (P = 0.003, χ ² = 15.796) were significantly associated with PI development. Cases had a longer average length of hospitalisation stay of 68.5 days, while controls stayed for 35.9 days. Most of the cases (n = 72, 45.3%) and controls (n = 87, 54.7%) were observed to be awake. In terms of pre‐arrival setting, more than half of the cases (n = 56, 56.0%) and controls (n = 52, 52.5%) were admitted from home. Cases were mainly admitted under cardiothoracic surgery (n = 16), whereas most controls were admitted under neurosurgery (n = 18) (Table 1).

4.2. Braden scale score and subscale scores

Almost half (47.0%) of the cases stayed in hospital between 22 and 80 days. The approximate time period taken for patients to acquire a PI appears to be 1 month (μ = 34.33 ± 67.451). Of 100 cases, 74.0% developed Stages 2 and 3 PIs. Majority (31.0%) of the cases developed PIs at the sacrum. The records of PI risk assessment showed that, overall, the total Braden scale score and subscale scores were found to be statistically significant in PI development (Table 2).

Table 2.

Total Braden score and subscale scores of cases and controls (N = 199)

	Outcome
	Cases (n = 100) n (SD)	Controls (n = 100) n (SD)	Statistics	P‐value
Total Braden score on admission	16.10 (±3.555)	19.34 (±3.114)	t = 6.844	<0.001*
Sensory perception subscale score	3.25 (±0.957)	3.73 (±0.620)	t = 4.179	<0.001*
Moisture subscale score	3.38 (±0.722)	3.71 (±0.479)	t = 3.770	<0.001*
Mobility subscale score	2.56 (±0.880)	3.17 (±0.756)	t = 5.261	<0.001*
Activity subscale score	1.94 (±1.081)	2.89 (±1.044)	t = 6.274	<0.001*
Nutrition subscale score	2.60 (±0.711)	3.11 (±0.621)	t = 5.404	<0.001*
Friction‐shear subscale score	2.31 (±0.615)	2.74 (±0.465)	t = 5.536	<0.001*

^* Significant at P < 0.01.

4.3. Demographic and clinical predictors of PI development

Demographic factors such as “age” and “race” were observed to be statistically non‐significant as predictors of PI development. Both length of hospitalisation stay and presence of comorbidities were non‐statistically significant as independent predictors of PI development. Within the subgroups of level of consciousness and unresponsive patients, they were found to be six times (OR = 6.444) more prone to developing PIs than patients who were awake (P = 0.004, 95% CI: 1.806‐22.995). Patients who were admitted from other acute hospitals or clinics were found to be 76.8% (OR = 0.232) less likely to develop a PI compared with those who were admitted from public areas/workplaces (P = 0.018, 95% CI: 0.075‐0.783 (Table 3).

Table 3.

Demographic and clinical predictors of PI development

		Odds ratio (B)	95% CI	P‐value	Cox & Snell R 2
Age		1.014	0.997‐1.031	0.116	0.013
Race	Chinese	1.943	0.467‐8.082	0.361	0.051
	Malay	4.364	0.918‐20.742	0.064
	Indians	1.000	0.173‐5.772	1.000
	Eurasians	0.000	0.000	0.999
	Othersa
Length of hospitalisation stay (days)	1‐21	0.981	0.432‐2.225	0.963	0.002
	22‐80	1.161	5.23‐2.580	0.713
	81 and abovea
Level of Consciousness	Confused	4.833	0.528‐44.211	0.163	0.064
	Drowsy	1.381	0.478‐3.991	0.551
	Unresponsive	6.444	1.806‐22.995	0.004*
	Awakea
Presence of Comorbidities	No	0.000	0.000	0.999	0.273
	Yesa
Pre‐arrival setting	Home	1.000	0.430‐2.326	1.000	0.080
	Nursing home/community hospitals	1.161	0.255‐5.286	0.847
	Acute hospitals/clinics	0.242	0.075‐0.783	0.018*
	Not documented	2.520	0.799‐7.954	0.115
	Othersa

Abbreviations: CI, confidence interval; PI, pressure injury.

^* Significant at P ≤ 0.05.

^** Significant at P < 0.001.

^a Reference group.

4.4. The Braden score and subscale scores as predictors of PI development

The Braden scale and all the subscales were found to be independent predictors of PI development. The Braden summative score on admission was significant in predicting PI development (P < 0.001). Patients who were assessed for PI risk were 24.2% (OR = 0.758) less likely to develop a PI than those who were not assessed (Table 4).

Table 4.

The Braden score and subscale scores as predictors of PI development

		Odds ratio (B)	95% CI	P‐value	Cox & Snell R 2
Braden score on admission		0.758	0.689‐0.833	<0.001**	0.187
Sensory perception	Completely limited	10.370	1.240‐86.703	0.031*	0.087
	Very limited	3.704	1.352‐10.146	0.011*
	Slightly impaired	2.963	1.366‐6.426	0.006*
	No impairmenta
Moisture	Often moist	8.522	1.045‐67.467	0.045*	0.068
	Rarely moista
Activity	Bed bound	9.917	4.087‐24.064	<0.001**	0.165
	Chair bound	3.333	1.325‐8.386	0.011*
	Walk occasionally	1.631	0.685‐3.882	0.269
	Walk frequentlya
Mobility	Immobile	15.429	3.056‐77.903	0.001**	0.122
	Very limited	5.829	2.451‐13.859	<0.001**
	Slightly limited	2.236	1.058‐4.728	0.035*
	No limitationa
Nutrition	Inadequate	10.459	3.717‐29.434	<0.001**	0.119
	Adequate	3.095	1.238‐7.741	0.016*
	Excellenta
Friction‐shear	Problem	4.171	2.295‐7.582	<0.001**	0.111
	No apparent problema

Abbreviations: CI, confidence interval; PI, pressure injury.

^aReference group.

^* Significant at P ≤ 0.05.

^** Significant at P < 0.001.

4.5. Sensitivity and specificity of Braden scale at various cut‐off scores

Five cut‐off scores15, 16, 17, 18, 19 were derived from Mordiffi's28 thesis and tested in this study. A cut‐off score of 17 was found to achieve the best balance between sensitivity (0.75) and specificity (0.68) values (Table 5).

Table 5.

Sensitivity and specificity of the Braden scale at various cut‐off scores

Cut‐off score	True positive	False negative	True negative	False positive	Sensitivity	Specificity
≤15	46	54	86	13	0.46	0.87
≤16	55	45	79	20	0.80	0.55
≤17	68	32	74	25	0.75a	0.68a
≤18	70	30	64	35	0.65	0.70
≤19	82	18	53	46	0.82	0.54

^aThis set of value is the best specificity‐sensitivity balance among the listed Braden cut‐off scores.

4.6. Sensitivity and specificity Braden subscales at various cut‐off scores

For the Sensory Perception subscale, a cut‐off score of 3 was identified, with a sensitivity value of 0.46 and specificity value of 0.81. Similarly, the Moisture subscale yielded a cut‐off score of 3, with sensitivity and specificity values of 0.51 and 0.72, respectively. For the Mobility subscale, the best cut‐off score was derived to be 2, which yields a sensitivity value of 0.46 and a specificity value of 0.84. A cut‐off score of 2 was derived for the Activity subscale, with sensitivity at 0.68 and specificity at 0.69. The Nutrition subscale also obtained a cut‐off score of 2, with sensitivity at 0.48 and specificity at 0.87. In the Friction‐shear subscale, a cut‐off score of 2 was derived, and the sensitivity and specificity values were observed to be 0.61 and 0.73, respectively. It appears that the Activity subscale score best predicts PI development, followed by Friction and Shear (Table 6).

Table 6.

Sensitivity and specificity of the Braden subscales at various cut‐off scores

Braden subscales	Cut‐off scores	True positive	False negative	True negative	False positive	Sensitivity	Specificity
Sensory perception	≤2	22 (22.0%)	78 (78.0%)	92 (92.9%)	7 (7.1%)	0.22	0.93
	≤3	46 (46.0%)	54 (54.0%)	80 (80.8%)	19 (19.2%)	0.46a	0.81a
Moisture	≤2	8 (8.0%)	92 (92.0%)	98 (99.0%)	1 (1.0%)	0.08	0.99
	≤3	51 (51.0%)	49 (49.0%)	71 (71.7%)	28 (28.3%)	0.51a	0.72a
Mobility	≤2	46 (46.0%)	54 (54.0%)	82 (82.8%)	17 (17.2%)	0.46a	0.84a
	≤3	86 (86.0%)	14 (14.0%)	36 (36.4%)	63 (63.6%)	0.86	0.37
Activity	≤2	69 (69.0%)	31 (31.0%)	67 (68.4%)	31 (31.6%)	0.68a	0.69a
	≤3	88 (88.0%)	12 (12.0%)	34 (34.7%)	64 (65.3%)	0.88	0.35
Nutrition	≤2	41 (41.0%)	59 (59.0%)	85 (85.9%)	14 (14.1%)	0.41a	0.87a
	≤3	93 (93.0%)	7 (7.0%)	25 (25.3%)	74 (74.7%)	0.93	0.26
Friction‐shear	≤2	61 (61.0%)	39 (39.0%)	72 (72.7%)	27 (27.3%)	0.61a	0.73a

^aThis set of value is the best specificity‐sensitivity balance of each Braden subscale cut‐off score.

5. DISCUSSION

While various authors have claimed that PIs typically develop within the first 2 weeks of hospitalisation.32, 33 The cases in this study stayed for twice as long before the occurrence of PI. It was surprising that most PIs in our study were Stage 2 or unstageable considering that they were from the general wards, with more stable conditions, and should therefore be less prone to acquiring severe PIs. The prevalence of Stage 2 injuries may be attributed to the under‐detection of PIs in their initial development, suggesting inconsistencies in risk asssessment.34, 35

The most frequent anatomical location of PIs was found over the sacrum. This finding is consistent with existing evidence that purports a majority of the PIs are over the sacral area, as it receives the greatest pressure when patients are in dorsal position, and the lower half of the body because of the presence of large bony prominences and the uneven distribution of body weight in these areas.36, 37, 38

5.1. Predictors of PI development

Level of consciousness, pre‐arrival settings, the Braden subscale scores, and the full scale score on admission were shown to be significant as independent predictors of PIs. Our findings showed that “unresponsive” patients were about six times more likely than patients who were “awake” to acquire a PI. This is consistent with studies by Agrawal and Chauhan5 and Bhattacharya and Mishra,36 who reported that impaired levels of consciousness prevented patients from detecting and relieving the discomfort from prolonged pressure, increasing their risks of developing PIs. Furthermore, we observed that patients who were admitted from other hospitals were less likely than those who were admitted from home to acquire PIs, contradicting previous studies that purport the significance of admissions from other acute hospitals.33, 34, 36 It remains relevant for clinicians to understand admissions from a nursing home as an independent predictor of hospital‐acquired PIs in order to allow timely implementation of preventive and treatment measures once patients arrive at the hospital.37

5.2. Braden score and subscale scores as predictors of PI development

A statistical mean difference was observed in the total Braden score on admission between cases and controls, insinuating that higher Braden scores are eventually associated with PI development. This is consistent with other study findings, whereby the first Braden scoring of newly hospitalised patients were predictive of PI outcomes.33, 34 Generally, all the Braden subscale scores were found to be independent risk predictors of PI development. In patients with limited sensory perception, they are prevented from changing their posture in the presence of constant pressure, impeding blood flow and eventually leading to ischaemia of the skin tissues.4, 38 Patients who are frequently exposed to moisture from urine, stools, or wound drainage are also more prone to impaired skin integrity.32 In patients with altered mobility, their abilities to change and control their body positions are diminished, increasing their potential for exposure to prolonged and intense pressure.39 Therefore, patients with changes in mobility levels are more vulnerable to develop PIs and entail more intensive and specific nursing interventions than the rest. Similarly, patients with reduced levels of activity are predisposed to PI risk as they may experience prolonged duration and intensity of pressure on the bed.40

It is also paramount for patients to receive adequate nutritional support so that metabolic disturbances and deteriorating health may be prevented, mitigating their risk of PI development.41 Therefore, the nutrition subscale score may be used as an adjunct nutrition screening tool to detect malnutritioned patients. Moreover, PIs were observed in patients who obtained lower scores on the Friction‐Shear subscale, suggesting that patients with mere friction and shear problems may be overlooked in terms of nursing care. The lack of appreciation of friction and shear as PI variables may be partially attributable to nurses underestimating the degree of shear or not recognising the damaging effect of combined shear and pressure in some patients.42

Internationally, a cut‐off score of 18 in the Braden scale is widely accepted across all clinical settings.13 In our study setting, the Braden scale was highly specific yet less sensitive at a cut‐off score of 18, indicating that patients who are truly at risk of developing PIs may not be detected, while patients who are not at risk are accurately identified. A cut‐off score of 17 showed the best balance among sensitivity, specificity, positive predictive value, and negative predictive value. Similarly, Mordiffi's28 study also illustrated a summative Braden score of 17 as the most sensitive (0.56) and specific (0.73) in the prediction of PI risk. Currently, the study site practices a Braden scale risk assessment cut‐off score of 18. Thus, this over‐prediction in risk assessment may lead to difficulty in drawing important conclusions about the capability of the Braden scale in identifying patients at risk successfully.39 Unnecessary nursing interventions may be elicited when PI risk is over‐predicted, which may not be cost‐effective.13, 15 Hence, revision of the cut‐off score of the Braden scale in the study site to better predict for PI risk is warranted.

The study showed that Braden subscale scores have been proven to be strongly related to PI occurrence, and it was possible to identify which of them contributed more or less to the risk of PI. The Activity subscale illustrated the highest sensitivity and specificity, followed by Friction‐Shear subscale. Likewise, patients in the present study had a longer average hospitalisation stay, accounting for their higher scores on the Friction‐Shear and Activity subscales. Friction and shear may occur more prevalently in patients who experience limitations in activity or mobility.38, 42 The significance of Activity and Friction‐Shear subscales in this study reinforces the need for future preventive measures to be focused on mobilisation of patients in order to achieve desirable PI incidences. The Braden scale appeared to be the most predictive form of PI risk assessment compared with all of the subscales. Likewise, a Portuguese study proposed two simplified scales to evaluate their predictive ability against the Braden scale, and the original scale was elucidated to have the best predictive capacity.43 The above findings may conclude that the clinical effectiveness of the scale transcends across countries and settings.27

In summary, while the Braden scale is ultimately not perfect in risk prediction, it should not be further simplified as it possesses the best predictive capability with six conceptualised risk factors (subscales). It is paramount to recognise that the total Braden scale score does not aid nurses in planning for individualised care;42 therefore, Braden subscale scores should also be evaluated and interpreted individually to meaningfully guide care planning through risk factor identification.42, 44

5.3. Limitations of the study

The use of a retrospective approach may have restricted the exploration of other predictors of PIs as existing data were retrieved from the hospital's clinical systems. Future studies may consider using a prospective approach in order to capture the population's exposure to a variety of other risk factors at baseline. Derivation of enhanced evidence may support the suggested Braden cut‐off score from this study. This study was unable to elucidate the effect of preventive interventions on PI development given the short period of time available to conduct the research. Future studies can explore the type and frequency of preventive interventions implemented for patients identified to be at risk by the Braden subscales. In this study, it was not possible to assess the consistency of nurses' risk assessments using the Braden scale as the data were collected retrospectively. Therefore, there appears to be a need to review the inter‐rater reliability of the Braden scale across tertiary hospitals in Singapore.

6. CONCLUSION

The study is a pioneer attempt at addressing the paucity of studies conducted in terms of risk assessment using the Braden subscales in Singapore. From this study, all of the Braden subscales were found to be independent PI predictors. However, it is noteworthy that the Braden scale has the best predictive capability in comparison with all the subscales. The results of the study highlighted that the Braden cut‐off score was 17, which is lower than the established cut‐off score of 18 in Singapore tertiary hospitals. Cut‐off scores are typically influenced by sensitivity and specificity analyses, which in turn are influenced by preventive measures adopted by the institution.12 More research is needed to ascertain the Braden context‐specific cut‐off score, the effectiveness of RAS‐based interventions to prevent PIs, and the accuracy of risk assessment among nurses. Significant predictors of PI among the study population were found to be level of consciousness, pre‐arrival setting, the total Braden score, and the subscales' scores on admission. Nurses are currently trained to trigger preventive measures when the total Braden score is low. Effective prevention requires individualised application of findings from risk assessment using the total Braden score and the subscales to plan for preventive interventions. Therefore, there is a need to review the PI risk assessment protocols across health care institutions. Furthermore, health care institutions may consider developing a PI prevention bundle comprising commonly used preventive interventions.

Lim E, Mordiffi Z, Chew HSJ, Lopez V. Using the Braden subscales to assess risk of pressure injuries in adult patients: A retrospective case‐control study. Int Wound J. 2019;16:665–673. 10.1111/iwj.13078