Subepidermal moisture detection of heel pressure injury: The pressure ulcer detection study outcomes

Abstract

We examined subepidermal moisture (SEM) and visual skin assessment of heel pressure injury (PrI) among 417 nursing home residents in 19 facilities over 16 weeks. Participants were older (mean age 77 years), 58% were female, over half were ethnic minorities (29% African American, 12% Asian American, 21% Hispanic), and at risk for PrI (mean Braden Scale Risk score = 15.6). Blinded concurrent visual assessments and SEM measurements were obtained at heels weekly. Visual skin damage was categorised as normal, erythema, stage 1 PrI, deep tissue injury (DTI) or stage 2 or greater PrI. PrI incidence was 76%. Off‐loading occurred with pillows (76% of residents) rather than heel boots (21%) and often for those with DTI (91%). Subepidermal moisture was measured with a device where higher readings indicate greater moisture (range: 0‐70 tissue dielectric constant), with normal skin values significantly different from values in the presence of skin damage. Subepidermal moisture was associated with concurrent damage and damage 1 week later in generalised multinomial logistic models adjusting for age, diabetes and function. Subepidermal moisture detected DTI and differentiated those that resolved, remained and deteriorated over 16 weeks. Subepidermal moisture may be an objective method for detecting PrI.

1. INTRODUCTION

Heel pressure injuries (PrIs) are the result of damage to the tissues from mechanical forces of pressure, shear, and strain, compressing and deforming tissues between the bony prominence of the calcaneous and the external surface. Several anatomical features make the heels more prone to PrI, and specifically deep tissue injury (DTI)1, 2 than some other areas of the body. First, when a person is lying supine, the posterior aspect of the heel has limited subcutaneous fat volume and includes the stiff Achilles tendon, making the site susceptible to pressure which, if unrelieved, can result in tissue breakdown and necrosis.3, 4, 5 Second, the heel consists of skin overlying avascular fat and reticular dermis present in fibrous septa compartments and supplied by blood vessels including the subdermal plexus, periosteal plexus, and connecting vessels between the reticular dermis and the calcaneus bone.4 The septa create reasonably avascular fat sections which may be vulnerable to pressure‐induced tissue damage in a manner similar to compartment syndrome.4 Third, the contact surface area on the posterior heel is small and there is limited fat tissue available in the posterior heel when pressure is exerted directly on the calcaneous bone. Finally, there is only one muscle in the heel, a thin layer of panniculus carnosus muscle that along with the subcutaneous fat is likely involved in PrI development.3, 4 As noted, some of the heel tissues are stiffer and less elastic than in other areas of the foot and body.3, 4, 5 There is evidence from finite element modelling that stiffer soft tissue is less capable of deforming and alleviating the stress and strain from pressure, contributing to increased PrI susceptibility.5 In addition, the fat pad is subjected to a variety of different stresses and strains as the position of the resting heel changes6, 7.

Heel PrIs are common among vulnerable elders and special populations, with incidence rates among hospitalised elders reported between 6% and 30%8, 9; among persons undergoing orthopaedic surgery, incidence is about 15%10, 11 and in critical care units, incidence has been reported at 9%.12 Heel PrIs are estimated to account for 24% of all PrIs in hospitals, 12% of all medical device‐related PrIs among hospitalised patients and 23% of all PrIs in nursing homes (NHs).1, 13, 14

PrIs are commonly classified according to the level of visible tissue damage and skin discoloration.15 Stage 1 PrI is identified by non‐blanchable erythema (ie, redness) on intact skin. Stage 2 PrIs exhibit wounds involving the epidermal and dermal tissues; unstageable stage 3 and stage 4 ulcers involve visible damage involving the full thickness of the dermis with deeper tissues and structures observed; DTI presents on the skin surface as a blood‐filled blister or maroon/purple skin discoloration. While two‐thirds of PrIs in elders are in stage 1 or 2, 9% of all PrIs are classified as DTI with about one‐third of all DTIs presenting on the heels.15, 16, 17

Heel PrIs present with characteristics that are different from PrIs observed at other anatomical locations. Heel PrIs may present as fluid‐filled or blood‐filled blisters or black dry eschar areas that over time re‐epithelialize under the eschar.17 The severity of a heel PrI does not always reflect the degree of pressure applied.3, 4 DTIs and stage 3 and 4 full‐thickness PrIs are not necessarily caused by more pressure than superficial lesions (which may also involve friction forces as a contributing factor).4, 5

The skin colour of heels may also be different than other areas of the body. As noted above, change in skin colour is a key component of PrI classification. We have shown that, among patients with darker skin tones, colour changes on areas of the trunk can be difficult to detect.18 However, the heel has limited melanocytes and so skin tone may not be as important a factor in visually detecting heel pressure damage as with other anatomical sites.18

Detection of pressure damage early on is critical for viable tissue rescue. Yet detection continues to be focused on observation of skin discoloration that occurs when damage has already progressed within tissues. Specifically, others have demonstrated that inflammatory changes with tissue oedema can occur from 3 to 10 days before visible skin breakdown at the heel19, 20 and the time of injury precedes DTI from 3 to 5 days.20 Such changes offer an opportunity for a biophysical measure of subclinical, or pre‐stage 1, pressure‐induced tissue damage. Detecting of damage to the tissue before it is visible as skin discoloration could lead to more effective prevention practices and less severe skin damage.

One such biophysical measure is subepidermal moisture (SEM). SEM is a term coined in early studies of the relationship between oedema and PrI development to describe oedema present in tissues below the stratum corneum in order to differentiate from the moisture on the surface of the epidermis.21, 22, 23, 24, 25, 26, 27, 28 The instruments available to measure SEM do not differentiate the level at which the oedema occurs in the tissues except as a reflection of the depth of tissue interrogation. In our prior studies, we demonstrated that a hand‐held device for SEM measurement identified early skin damage (erythema or stage 1 PrIs) at the sacrum and buttocks locations in NH residents, persons with dark skin tones, and persons with spinal cord injury.21, 22, 23, 24, 25 Further, among NH residents, SEM was higher (eg, increased oedema and inflammation) when skin damage was not visible at assessment but was detected as erythema or stage 1 PrI 1 week later.21, 22 SEM values predicted 26% of the occurrences of such damage to the skin the following week.22

This study examines the relationship of SEM to pressure‐induced tissue damage at the heels, which was not done in our prior work. The specific aims were:

1. To describe the relationship of SEM to pressure‐induced tissue damage at the heels over sequential weeks in NH residents.

2. To describe the progression of heel DTIs.

3. To describe the relationship of observed preventive practices and subsequent pressure damage at the heels.

Replacing current methods of heel PrI detection that rely only on visually observed skin changes with other approaches has the potential to identify PrI damage earlier with the possibility of improved viable tissue rescue and PrI prevention.

2. METHODS

2.1. Subjects and setting

The protocol and methods for this study, approved by The University of California, Los Angeles, Human Subject Protection Committee and NH administration in accordance with facility guidelines, have been reported elsewhere25 but are provided here for ease of understanding. Written informed consent to participate in the study was obtained directly from residents who were able to provide informed consent or from their designated representatives for residents unable to provide consent, with assent obtained from the resident. Analyses for this paper are restricted to participants who completed at least 3 weeks of skin assessments with SEM measures (N = 417) to enable longitudinal analyses (Figure 1).

Figure 1.

Flow of participants through the study.25 Of the 490 (40% consent rate) consented subjects 297 completed all 16 weeks of the study, 42 self withdrew (most common reason was did not want to be checked each week), 10 were withdrawn by the principal investigator (PI) (due to severe agitation and aggression during assessments), 7 died and 14 were transferred, hospitalised or discharged

Residents were recruited from 19 NHs in the greater Los Angeles area. The characteristics of the NHs have been previously reported.25 Briefly, 74% (n = 14) of study NHs were rated as average or below using the overall quality of care rating from California's (http://www.calqualitycare.org/providers/nursing-homes/nursing-homes) and the US Centres for Medicare and Medicaid Services (https://www.medicare.gov/NursingHomeCompare) NH Compare websites. Study NHs were slightly better than state and US national average for mean percent of short stay (3 weeks or less) residents with PrIs and worse than state and US national average for long‐stay residents.

2.2. Medical record data

Research staff extracted information from participants’ medical records and their most recent Resident Assessment Instrument Minimum Data Set (MDS). During the study, the MDS was updated from version 2.0 to 3.0 but the majority of extracted items of interest were not affected. Medical and demographic information was extracted at baseline and additional medical data was abstracted monthly, specifically recording PrI prevention strategies as well as new medical information.

2.3. Braden scale risk assessment scores

To determine risk for PrI development, research staff assessed each participant at baseline and each subsequent month using the Braden Scale for Predicting Pressure Sore Risk (Braden Scale). The Braden Scale is composed of 6 subscales: sensory perception, moisture, activity, mobility, nutrition, and friction and shear, which are summed for a total score ranging from 6 to 23, with lower scores indicating higher risk for PrI development.29, 30 Inter‐observer agreement of the Braden Scale total score was examined with percent agreement and weighted kappa. Of 993 reliability assessments with 2 observers who were randomly assigned over the 16‐week study period, the agreement of 2 research staff for Braden Scale total score resulted in a weighted kappa of .58 with 77.8% of ratings with a difference of 2 or less.

2.4. Visual skin assessment

Trained research staff assessed skin health through direct independent visual assessments each week of the 7 anatomical locations: sacral, right and left buttocks, ischial tuberosities, and heels. This paper reports data on the heel locations only. Participants were positioned in bed on their right side (unless unable to turn to the right), and heels were observed after turning and all trunk locations were assessed (which occurred between 3 and 5 minutes after turning). Areas of visual skin discoloration were palpated for blanchability and any PrIs were assessed using the Bates‐Jensen Wound Assessment Tool (BWAT).31, 32 Following visual skin assessment, a second assessor blinded to the initial assessor's recorded visual skin assessment obtained SEM measures of each heel.

Training for visual skin assessment emphasised stage 1 PrIs and DTI because these conditions may be difficult to detect.33, 34, 35, 36 Erythema was graded as minimal, moderate, or severe discoloration; minimal was defined as pink or slight skin redness, moderate was bright redness in light skin tones and purple in dark skin tones, and severe was dark red to purple in light skin tones and black to blue‐grey colours in dark skin tones. In persons with light skin tones, blanching based on finger palpation, was defined as blanchable or non‐blanchable. Erythema was defined as moderate skin discoloration (bright redness in light skin tones and purple in dark skin tones), with blanching (blanching not considered for dark skin tones). Stage 1 PrIs were defined as moderate skin discoloration (all skin tones), with non‐blanching (non‐dark skin tones). PrIs more severe than stage 1 were classified using the (National Pressure Ulcer Advisory Panel) NPUAP's 2009 system.37 DTI was defined as severe skin discoloration (purple or maroon in light skin tones and black to blue‐grey in dark skin tones) with or without blanching or a blood‐filled blister. For skin that demonstrated none of the above visual characteristics, the assessment was considered normal skin. The presence or absence of scar tissue and dry skin was also recorded.

At all baseline and then randomly selected observation visits, 2 observers obtained independent visual skin assessments 5 to 10 minutes apart. Paired observers were blinded to each other's ratings. Inter‐observer agreement was examined for heels across all skin conditions, with weighted kappa of .78 for the right heel and .76 for the left heel (n = 2124 observation pairs for each heel). As noted above, individual characteristics contribute to visual assessment category and inter‐observer agreement as assessed by simple kappa was calculated for these items: erythema presence (present vs absent) right heel = .75 and left heel = .77; erythema type (blanchable vs non‐blanchable) right heel = .75 and left heel = .77; erythema severity right heel = .61, left heel = .60. Inter‐observer agreement via weighted kappa was calculated for PrI stages 2, 3, and 4: kappas = .95 for both heels. Inter‐observer agreement for presence of DTI, simple kappas = .96 (right heel) and .94 (left heel). Kappa values above .80 indicate near perfect reliability and values between .61 and .80 imply substantial reliability.38 At each assessment, observers also noted presence or absence of pressure redistribution bed support surface, heel protectors, and/or pillows in use for off‐loading heels.

2.5. Subepidermal moisture measures

Concurrent with visual assessments, SEM was measured weekly at the posterior aspect of each heel on clean dry skin using the Delfin MoistureMeter D (Delfin Technologies, LTD, Greenwich, Connecticut) dermal phase meter. The device detects and measures water below the stratum corneum (the uppermost skin layer and the most influenced by outside surface moisture). Using dielectric parameters, high‐frequency low‐power electromagnetic waves of 300 MHz are transmitted via a coaxial line which terminates in an open‐ended coaxial wand that is manually placed on the skin surface. In the skin, the induced electrical field interacts mainly with water molecules closest to the wand with depth of interaction depending on the diameter of the circular electrode on the wand (in this study at a depth of 2.5 mm).39, 40 The portion of the electromagnetic energy that is not absorbed by tissue water is reflected and measured by the same wand as used for wave transmission. From the properties of the reflected wave, the dielectric properties of the site are determined and displayed in the measuring unit. Among dielectric properties, a dielectric constant is calculated and is directly proportional to the free and bound water (total water) in the skin and tissues.39 SEM values are displayed in relative dielectric constant units. The tissue dielectric constant (TDC) is directly proportional to the amount of water in the tissue and increases with increasing water content and oedema. At around 300 MHz, the electrical properties of free and bound water are nearly identical and thus, the measure is reflective of total water content in the tissue.39, 40 Pure water has a dielectric constant of 78.5 TDC, the dielectric constant of air is 1 TDC, and that of normal skin is approximately 40 TDC.39, 40, 41, 42, 43 The MoistureMeter D demonstrates a coefficient of variation of only 2.8% (readings every 5 seconds for 600 seconds).39, 43

A hand‐held wand was used to take readings at each heel on the posterior aspect. Readings were taken by placing the wand on the skin surface for 5 seconds to measure the impedance value of the skin in TDC units (range 0‐70; higher values indicate presence of more oedema). Values were automatically stored on the hand‐held computer connected to the wand and transcribed at the conclusion of the observation visit. Observations were blinded, with different observers conducting SEM measures and visual assessments. SEM was measured on the heels after all other anatomical locations were completed.

As with visual assessment, duplicate measurements by different observers (n = 42) were conducted at all baseline and randomly selected subsequent visits to assess precision of the device. Agreement was assessed via Bland‐Altman analysis, with a 95% confidence interval defined as mean ± 2 SDs.38, 44 Data for the left heel were more precise than for the right heel (2.5% vs 3.7% of the observations more than 2 SDs below the mean and 2.4% vs 3.4% of the observations more than 2 SDs above the mean). Of the duplicate observations, 75% (N = 1291) were conducted by the 4 primary study staff. Intra‐class correlations were assessed for this subset of observations (right heel = .57, left heel = .59). Both sets of analysis indicate that the SEM measure was moderately precise for heels.

2.6. Statistical analysis

To assess the relationship between visual assessment and SEM, techniques were used that allowed for clustered‐repeated measures, as an observation was defined by week and anatomical location. Data were analysed using observation as the unit of analyses with generalised multinomial logistic modelling.45, 46, 47, 48, 49 Stata version 13 allows for this (a subset of generalised estimating equations), including effects for participant and measurement period (included in all models) to account for the correlated nature of the data.49

The primary aim of the study involved longitudinal analysis of skin damage, so participants with less than 3 weeks of observations are excluded from the analysis data set. Of the 490 enrolled in the study, 417 participants had at least 3 weeks of data. Most participants had 14 or more weeks of data (N = 297); 38 participants had 11 to 13 weeks of data and 82 had 3 to 10 weeks of data.

Heel PrI incidence was calculated as the number of participants who developed erythema or heel PrIs divided by the total number of participants assessed and followed up for at least 3 sequential weeks (n = 417). Incidence was calculated by skin outcome (any level damage, erythema/stage 1, stages 2 to 4 PrI, and DTI) and by ethnicity/racial groups.

Initial bivariate relationships of visual skin assessment outcome (normal, erythema, stage 1 PrI, DTI) were examined with all covariates of interest: Braden Score, functional status, age, gender, provision of preventive measures (eg, pressure redistribution support surfaces, heel protectors, or pillows for off‐loading heels), medical comorbidities (presence of diabetes mellitus, coronary artery disease [CAD], hypertension, peripheral vascular disease [PVD]), and body mass index (BMI). Functional status was examined by creating a dichotomous variable combining MDS transfer, mobility, and urinary incontinence scores where “functional” was defined as MDS transfer, mobility, and urinary incontinence scores less than 3. Covariates with an a priori significance level of P = .10 were included in SEM models. Several covariates were available only for some participants (due to revisions in the MDS assessment form during the study) including BMI and hypertension. Thus, because of the limited data set with these variables they were not included in any of the models.

The frequency of types of preventive off‐loading methods based on direct observations of research staff during weekly assessments was examined. Of particular interest, we explored the relationship of preventive off‐loading heel boots and pressure damage by examining heel boot use prior to development of incident pressure damage. Characteristics of those participants with and without heel boot use were compared across risk for pressure damage at the heels.

The relationship of visual skin assessment and SEM at the same time (concurrent SEM) was conducted with only erythema, stage 1, and DTI skin outcomes, adjusted for covariates identified in the bivariate analyses. (If an open ulcer exists, there is no need for detection regardless of skin discoloration.) To determine if SEM from the prior week was associated with visual skin damage the next week, the outcome was a 3‐level variable with normal skin/erythema, stage 1 PrI, and stage 2 and greater PrI. In all analyses, model assumptions were tested and models modified for any violations. Goodness of fit using McFadden R ² and Negelkerke R ² was assessed and Bayesian Information Criterion (BIC) was used to compare models and to determine the best model fit.46, 47, 48 Finally, we examined SEM for determining progression of DTI sites.

3. RESULTS

Participants were older (mean age of 76.5; SD 14.8 years) and 58% were female (Table 1). The majority of participants were ethnic minorities: 29% African American, 12% Asian American, and 21% Hispanic. About two‐thirds of the participants had light skin tone (61%). Participants were at risk for PrIs, with a mean Braden Scale score of 15.6 (SD 3.2), and were functionally dependent, with 25% of participants totally dependent for bed mobility and 43% requiring extensive assistance with bed mobility. Participant mean length of hospital stay was 1.6 years (25% of participants were newly admitted during study).

Table 1.

Demographic and functional characteristics of participants with 3 or more observation weeks

Characteristics	Overall (N = 417)	No skin damage (N = 98)	Skin damage (N = 319)
Agea	76.5 (14.8)	71.9 (15.1)	77.8 (15.1)
Female	243 (58%)	51 (52%)	191 (60%)
Ethnicity/racea
African American	122 (29%)	51 (52%)	71 (22%)
Asian American	50 (12%)	5 (5%)	45 (14%)
Caucasian	156 (37%)	23 (23%)	133 (42%)
Hispanic	89 (21%)	19 (19%)	70 (22%)
Skin tone (based on Munsell value)a
Dark	68 (16%)	27 (28%)	41 (13%)
Medium	93 (22%)	32 (32%)	61 (19%)
Light	256 (61%)	39 (40%)	217 (68%)
MDS bed mobility scorea (n = 413)	2.7 (1.2)	2.6 (1.3)	2.7 (1.1)
MDS transfer scorea (n = 414)	3.1 (1.0)	2.9 (1.2)	3.1 (1.0)
MDS urinary incontinencea (n = 407)	2.4 (1.5)	2.2 (1.7)	2.5 (1.4)
MDS bowel incontinencea (n = 406)	2.5 (1.5)	2.3 (1.6)	2.5 (1.4)
Braden scale for predicting pressurea sores total score (n = 405)	15.6 (3.2)	16.3 (3.5)	15.3 (3.1)

Abbreviation: MDS, minimum data set.

Unless otherwise noted, means and SDs are presented. MDS bed mobility transfer score where 1 = independent, 4 = dependent. MDS urinary incontinence, bowel incontinence scores where 1 = continent, 4 = incontinent all the time. Braden scale for predicting pressure sores range 6 to 23, where 6 = high risk and 23 = no risk.

^aDifference between those with no heel damage compared with those with damage (t test for differences or chi‐square), P < .001.

Participants with any level of heel damage were similar to the overall sample. However, most participants with DTI were female (70%), about half were African American (48%), and they were generally at higher PrI risk (mean Braden Scale score = 14.1, SD 2.8).

Incidence of heel erythema and all PrI stages over the 16 weeks was 76% (n = 318 with any level skin damage). Incidence of PrI (all stages, no erythema) was 30% (n = 126), with 5% (n = 21) incidence of stage 2 or greater PrIs and 25% (n = 105) incidence of stage 1 PrI. Of the 105 stage 1 heel PrIs during the study, 93% (n = 98) had visible damage that persisted or progressed to other PrI stages during the study and for the remaining 6% (n = 7) damage resolved within 3 weeks. Heel DTI incidence was 5% (21 participants with 24 DTIs); another 16 DTIs (from 12 participants) were present at baseline (8% prevalence).

SEM was similar for right and left heels for those participants with no heel damage over the entire 16 weeks of the study (mean 29.2 TDC, SD 5.9 and mean 29.0 TDC, SD 5.8 right and left heel, respectively). SEM at the heels was significantly lower when DTI was observed, compared to those with no damage observed (P = .007, Table 2). Compared to DTI, SEM was significantly higher for erythema (both P = .005).

Table 2.

Distributional characteristics of subepidermal moisture measure for heels by visual skin assessment

	Visual skin assessment
	No damage		Erythema		Stage 1 PU		Stage 2 + PU		Deep tissue injury
Anatomical site	N	TDC Mean (SD) Med/P75	N	TDC Mean (SD) Med/P75	N	TDC Mean (SD) Med/P75	N	TDC Mean (SD) Med/P75	N	TDC Mean (SD) Med/P75
Right heel	3753	29.07 (6.91) 28.9/33.6	1001	30.79 (6.72) 31.1/35.2	126	29.83 (7.95) 29.5/35.3	51	28.3 (9.08) 26.5/34.1	97	25.69 (7.66) 23.9/30.1
Left heel	3583	28.85 (6.82) 28.6/33.3	1053	30.61 (7.19) 30.2/35.0	168	30.54 (7.87) 30.4/35.5	41	31.29 (7.81) 31.0/37.8	98	27.58 (7.76) 27.3/32.7

Abbreviations: Med, Median; N, number of assessment observations (multiple weeks for each participant); P75, 75th percentile; PU, pressure ulcer; TDC, tissue dielectric constant.

3.1. Bivariate relationships between visual skin damage, SEM and covariates of interest

Bivariate analyses were modelled for right and left heels with visual skin assessment as the outcome (normal skin, erythema, stage 1 PrI, stage 2+ PrI, DTI). The covariates that were significantly related to visual skin assessment for heels included age (P < .01 all skin outcomes), CAD (P < .01 stage 2+ PrI and DTI), diabetes (P < .01 erythema, stage 1 PrI), functional status (P < .10 erythema, stage 1 PrI and DTI), and use of preventive off‐loading (P = .01 stage 2+ PrI and DTI). Table 3 presents the bivariate results for those covariates related to visual skin outcomes. Covariates of gender and PVD were not significantly related to visual outcomes. Braden Scale was significant for DTI (P < .05) for the left heel only and there was a significant difference in Braden Scale scores (total and all subscales) for those participants with heel PrI damage compared for those with heel PrI damage (total Braden 15.3 (3.1) and 17.0 (3.3), for those without heel PrI P < .001). Thus, Braden Scale was included in the initial models but was not significant and was dropped from the final models. CAD and diabetes were strongly correlated; thus, only diabetes was included in the models as CAD was not available for all participants due to the transition from the MDS 2.0 to 3.0 during the study. There was no significant difference for mean SEM values for those with versus without each of the following covariates: cancer, CAD, diabetes, hypertension, and PVD.

Table 3.

Bivariate relationships with visual skin damage

Visual skin outcomes	Braden score N = 5369 Coefficient (P value)	Age N = 5354 Coefficient (P value)	Gender N = 5354 Coefficient (P value)	Functional status N = 5354 Coefficient (P value)	Diabetes N = 5073 Coefficient (P value)	Coronary artery disease N = 1356 Coefficient (P value)	Prevention off‐loading N = 4949 Coefficient (P value)
Erythema Right heel	−.030 (.15)	.019 (<.01)	−.101 (.5)	.29 (.05)	−.38 (.01)	.346 (.84)	.086 (.48)
Stage 1 PrI Right heel	−.049 (.14)	.053 (<.01)	.178 (.57)	.35 (.27)	−.758 (.02)	.281 (.61)	−.348 (.18)
Stage 2+ PrI Right heel	−.197 (.28)	−.049 (<.01)	1.26 (.11)	.237 (.79)	−.988 (.23)	−13.42 (<.001)	.312 (.67)
DTI Right heel	−.103 (.24)	.017 (.36)	−1.06 (.16)	.646 (.36)	−.429 (.54)	−13.63 (<.001)	.616 (.26)
Erythema Left heel	−.030 (.16)	.019 (.002)	−.051 (.736)	.361 (.02)	−.466 (.002)	−.013 (.35)	.67 (.60)
Stage 1 PrI Left heel	−.002 (.96)	.031 (.06)	.054 (.89)	.727 (.03)	−1.09 (.002)	.059 (.04)	−.178 (.53)
Stage 2+ PrI Left heel	−.11 (.40)	−.014 (.50)	−1.72 (.04)	.614 (.5)	−.547 (.56)	−14.694 (<.01)	3.391 (<.01)
DTI Left heel	−.179 (.05)	−.021 (.09)	−.003 (.99)	1.7 (<.01)	−.418 (.43)	−14.749 (<.01)	2.70 (<.01)

Abbreviation: DTI, deep tissue injury.

3.2. Relationship between preventive off‐loading and visual skin damage

Nearly all participants (76%) were observed with pillows for off‐loading heels (n = 318). A significantly higher proportion of those with heel PrI damage were observed with pillows for off‐loading (79%, n = 252; P = .02) compared to those participants with no heel damage (67%, n = 69). Less than half of the participants (41%, n = 172) were on a pressure redistributing support surface and there was no difference in use of pressure redistributing support surfaces between those participants with and without heel PrIs. Heel boots for off‐loading were observed in only 21% (n = 87) of participants, with no difference in use of heel boots for those participants with and without heel PrIs. Of those, only 19% were observed with heel boots for prior to PrI development.

Comparing those participants with no damage on heels during the entire study to those with DTI present or developed on heels, a significantly higher proportion of those with DTI were observed with pillows for off‐loading compared to the proportion of participants with no damage (91% vs 67%, P = .008). There was no difference in proportion of participants with DTI observed on pressure redistributing support surfaces compared to those participants with no damage on heels observed with pressure redistributing support surfaces (64% vs 69%). In contrast to other observed PrI, heel boots for off‐loading were observed in a higher proportion of participants with heel DTI than for participants with no damage on heels (P < .001).

To examine the use of heel boots as an early preventive strategy, we compared participants observed with heel boots for off‐loading prior to visualised incident heel PrIs to participants who used heel boots but with no heel damage during the study. Those participants who developed heel damage were just as likely to have used heel boots as those participants without damage throughout the study. Comparing participants observed with heel boots (regardless of whether or not heel damage occurred) to those participants without heel boots, those with heel boots were at higher risk for PrIs (mean Braden score = 13.9, SD 2.8 compared to 16.1, SD 3.3; P = .001). In addition, there was a significantly higher proportion of participants with heel boots who were dependent for bed mobility (84% vs 77%, P = .002).

3.3. Relationship between SEM and concurrent visual skin damage

SEM was significant for detecting stage 1 damage in both right and left heels (Table 3), with higher values associated with stage 1 PrI. DTI was also significantly related to SEM in left heels, but with lower SEM indicating damage (Table 3). Analyses were conducted with concurrent damage (erythema, stage 1 PrI, and DTI) as the outcome; adjusted models included age, diabetes, and function as covariates. BIC provided very strong support for the full models in both cases (Table 3). For the right heel model, age was associated with erythema and stage 1 PrI (Relative Risk Ratio (RRR) = 1.02, P ≤ .001; RRR 1.05, P ≤ .001, respectively). For the left heel model, age was also associated with erythema (RRR = 1.02, P ≤ .001), diabetes was associated with stage 1 PrI (RRR = .38, P = .008) and functional status was associated with DTI (RRR = 8.11, P = .016).

3.4. Relationship between SEM and future visual skin damage

The relationship between SEM and future damage (to examine the value of SEM as an early marker of damage) was modelled using SEM 1 week prior to visualised skin damage, adjusting for age, diabetes, and functional status. SEM was significantly related to future stage 1 PrI occurrence and DTI (Table 4). Functional status (more impaired status) was significantly related to all levels of skin damage (erythema: RRR = 1.42, P = .03; stage 1 PrI: RRR = 2.08, P = .04; stage 2+ PrI: RRR = 11.39, P = .006; DTI: RRR = 13.63, P = .004). The overall model was significant and with better goodness of fit (McFadden's R ² = .06, Nagelkerke R ² = .11) than the unadjusted model. Left heel model BIC difference was 254 between the unadjusted and the adjusted models, providing very strong support for the adjusted model with age, diabetes, and functional status.

Table 4.

Summary statistics for models of concurrent and next week skin damage at right and left heels using subepidermal moisture measures

	Unadjusted					Adjusteda
	Relative risk ratio (95% CI)					Relative risk ratio (95% CI)
Anatomical site by time of visual skin assessment	None vs erythema	None or erythema vs St 1 PU	None, erythema, or St 1 PU vs St 2+ PU	None, erythema, St 1 PU, St 2+ PU vs DTI	Model Wald chi‐square (P value)	None vs erythema	None or erythema vs St 1 PU	None, erythema, or St 1 PU vs St 2+ PU	None, erythema, St 1 PU, St 2+ PU vs DTI	Model Wald chi‐square (P value)	BIC difference
Right heel
Concurrent SEM (TDC)	1.00 (0.99‐1.02)	1.05 (1.01‐1.09)	N/A	0.98 (0.92‐1.04	5.56 (0.13)	1.00 (0.98‐1.01)	1.03 (1.00‐1.07)	N/A	0.98 (0.92‐1.04)	42.90 (<.001)	161.09
Diabetes	—	—	—	—	—	0.77 (0.56‐1.04)	0.55 (0.28‐1.08)	—	0.71 (0.15‐3.26)	—	—
Age	—	—	—	—	—	1.02 (1.01‐1.04)	1.05 (1.03‐1.08)	—	1.01 (0.98‐1.05)	—	—
Functional status	—	—	—	—	—	1.31 (0.97‐1.77)	1.26 (0.66‐2.40)	—	2.65 (0.54‐13.04)	—	—
Next week SEM (TDC)	1.002 (0.98‐1.01)	1.03 1.00‐1.07)	0.94 (0.82‐1.07)	0.97 (0.92‐1.02)	4.89 (0.29)	0.99 (0.98‐1.01)	1.02 (0.98‐1.06)	0.96 (0.84‐1.1)	0.97 (0.91‐1.02)	66.65 (<.001)	165.73
Diabetes	—	—	—	—	—	0.75 (0.55‐1.04)	0.57 (0.29‐1.14)	0.37 (0.07‐2.01)	0.77 (0.16‐3.58)	—	—
Age	—	—	—	—	—	1.02 (1.01‐1.04)	1.05 (1.02‐1.07)	0.93 (0.90‐0.96)	1.01 (0.98‐1.05)	—	—
Functional status	—	—	—	—	—	1.33 (0.97‐1.82)	1.25 (1.02‐1.07)	1.00 (0.16‐6.33)	2.28 (0.45‐11.54)	—	—
Left heel
Concurrent SEM (TDC)	1.00 (0.99‐1.02)	1.05 (1.01‐1.08)	N/A	0.95 (0.90‐0.99)	12.66 (.005)	0.99 (0.98‐1.01)	1.03 (1.00‐1.07)	N/A	0.94 (0.91‐0.98)	58.9 (<.001)	235.95
Diabetes	—	—	—	—	—	0.65 (0.48‐0.89)	0.38 (0.18‐0.77)	—	0.57 (0.16‐2.02)	—	—
Age	—	—	—	—	—	1.02 (1.01‐1.04)	1.03 (0.99‐1.07)	—	0.99 (0.95‐1.02)	—	—
Functional status	—	—	—	—	—	1.41 (1.05‐1.90)	1.98 (0.99‐3.94)	—	8.12	—	—
Next week SEM (TDC)	1.01 (0.99‐1.02)	1.05 (1.01‐1.08)	1.00 (0.93‐1.06)	0.92 (0.88‐0.97)	75.24 (P < .001)	1.00 (0.98‐1.02)	1.04 (1.01‐1.07)	0.99 (0.91‐1.08)	0.92 (0.89‐0.97)	75.24 (P < .001)	254.0
Diabetes	—	—	—	—	—	0.67 (0.48‐0.92)	0.35 (0.17‐0.74)	0.18 (0.04‐0.78)	0.50 (0.15‐1.74)	—	—
Age	—	—	—	—	—	1.03 (1.01‐1.04)	1.03 (0.99‐1.06)	1.00 (0.97‐1.03)	0.97 (0.94‐1.00)	—	—
Functional status	—	—	—	—	—	1.42 (1.04‐1.94)	2.08 (1.03‐4.21)	11.39 (2.03‐64.04)	13.64 (2.30‐80‐83)	—	—

Abbreviations: BIC, Bayesian Information Criterion; CI, confidence interval; DTI, deep tissue injury; PU, pressure ulcer; TDC, tissue dielectric constant.

^aModels adjusted for age, diabetes, and functional status.

The unadjusted and adjusted models for SEM predicting skin damage on the right heel were not significant (Table 3) although SEM appeared to have a relationship with stage 1 PrI (RRR = 1.03, P = .07).

We also examined sensitivity and specificity using SEM values at or above 29 TDC as a threshold for determining stage 1 PrI and SEM values 27 TDC and below as a threshold for determining DTI. The threshold value of 29 TDC provided right heel stage 1 PrI sensitivity of 58.6% and specificity of 47.2% and correctly predicted 47.6% of the outcomes with an area under the curve (AUC) of 0.53 (SE 0.02). Using non‐parametric receiver operating curve (ROC) estimation with bootstrap replications (n = 1000) to examine an estimation sample showed an AUC of 0.55 (SE 0.03). SEM values at or above 29 TDC as a threshold for determining left heel stage 1 PrI provided a sensitivity of 60.8% and specificity of 47.5%, predicting 48.1% of the outcomes with an AUC of 0.54 (SE 0.02) and ROC estimation with bootstrap replications (n = 1000) with an estimation sample showing an AUC of 0.56 (SE 0.03).

Using the threshold of 27 TDC or below for determining DTI provided right heel sensitivity of 49.3% and specificity of 65% correctly predicting 65% of the outcomes with an AUC of 0.57 (SE 0.03). ROC estimation with bootstrap replications (n = 1000) resulted in an AUC of 0.57 (SE 0.05). Left heel DTI sensitivity was 48.3% and specificity 63.4% correctly predicting 63% of the outcomes with an AUC of 0.56 (SE 0.03). ROC estimation with bootstrap replications (n = 1000) resulted in an AUC of 0.55 (SE 0.05).

3.5. SEM for determination of DTI progression

Of the 40 DTIs, 20% (n = 8) resolved to erythema or normal skin; 25% (n = 10) subsequently progressed to full thickness PrIs; and 55% (n = 22) persisted throughout the study. SEM measured at the same week as the DTI site was first visually detected (concurrent SEM) was higher for DTI sites that resolved compared to DTI sites that progressed to PrIs or persisted as DTIs (mean SEM 29.0 TDC vs mean SEM 27.6 TDC DTI progression and 25.4 TDC DTI persisted). Mean SEM was elevated the week prior to DTI observation, compared to contralateral non‐damaged heel (right heel DTI: 29.1 TDC vs 27.9 TDC for contralateral left heel; left heel DTI: 30.4 TDC vs 28.2 TDC for contralateral right heel).

4. DISCUSSION

In this multi‐ethnic sample of NH residents at risk for PrI, we found a relationship between SEM, a biophysical measure, and the current gold standard of visual skin assessment for PrI at heel locations with higher SEM associated with erythema and stage 2 or greater PrIs and lower SEM associated with stage 1 and DTI damage. The relationship existed when examining concurrent and future visually observed skin damage although the association was small. These findings validate our prior studies using a different device to measure SEM.21, 22, 23, 24 Capacitive devices such as those used in the early studies21, 22, 23 and in the current study determine hydration with electrical methods using 2 approaches, capacitive and impedance‐based capacitive parameters or dielectric parameters. Both approaches share similarities in technique but differ in measurement values and how values are calculated; regardless of approach, higher values indicate increased water in tissues.39, 40 In the early studies of SEM and PrI, SEM was measured with equipment that used bioimpedance methods at low frequencies to evaluate the dielectric properties of the skin at a depth of 0.5 mm.21, 22, 23 The current study used equipment that evaluated the dielectric properties of the skin up to 2.5 mm depth at 300 MHz with capacitance measures. Even with differences in equipment and methods of examining SEM, findings from this study show similar patterns in the relationship between SEM and PrI as shown in the early studies. As in this study, we previously found that higher SEM was associated with increased risk of subsequent (1 week) erythema and stage 2 PrI development at the sacrum among NH residents.21, 22, 25 Previous research in persons with spinal cord injury and the same instrument as used in this study also found a significant difference in SEM in normal skin compared to erythema and/or stage 1 PrI damage at the sacrum.24 The heel anatomical location produced similar findings in this study, with higher SEM values for erythema, and stage 2 and greater PrI compared to normal skin, although the SEM values for heels overall were lower than trunk values.

In contrast to our prior work with NH residents and SEM, in this study SEM was significantly lower for DTI compared to normal skin at both right and left heels. These findings have not been seen in previous research in this area and are of particular interest because of the current discussion in the field related to the pathophysiology of DTI and PrI development. Specifically, the findings that DTI SEM values were lower than normal skin suggest that the pathophysiology model of “bottom to top” damage where damage begins at inner tissue levels and progresses up to the skin19, 50, 51 may be involved in DTI development. As tissue cells die and become non‐viable, the necrotic area is not perfused and has decreased blood flow. Thus, less fluid or water is present in the tissues possibly explaining the lower SEM reflected at the dermal tissue level. Mayrovitz and colleagues found that skin vascular volume and blood flow was associated with changes in SEM with reduced TDC values with decreased vascular volume and increased values with increased vascular volume measured in the forearm skin.52 It is unknown if this is similar for the heel location. Additionally, it is not clear how or if changes in vascular volume as might occur in deeper tissues with DTI also affect the dermal tissues. The findings of lower SEM with heel DTI are similar to others’ findings related to cooler temperature (suggesting less blood flow) of DTI sites that deteriorate compared to normal skin.53 The ability to detect damage below the dermal layer in heels is related to the anatomical structure of soft tissues at the heel. Heel fat pad thickness over the posterior calcaneous (the area covering the insertion of the Achilles tendon) has been reported as 1.1 cm in females and 1.4 cm in males54 using 3‐dimensional in vivo computed tomography measures suggesting that the measurement depth 2.5 mm) used in this study extended only through skin layers and reflecting DTI that has affected the dermal tissues. Alternatively, the pathophysiology of DTI damage may differ from what is currently understood even as SEM may be detecting the outcomes of deeper tissue damage at the dermal tissue layer. The specificity of SEM values for the threshold of SEM values at or below 27 suggests that this technique may be helpful in determining DTI although the AUC values are not high.

Our findings that 20% of the DTIs resolved and only 25% deteriorated into full thickness ulcers is different from studies of adults in acute care. In a retrospective study of 77 adult acute care patients, Sullivan found that 66% of DTIs resolved and only 9% evolved into full thickness ulcers.17 Our findings of 25% deterioration are closer to Richbourg et al's55 prospective study of adult hospitalised patients across 6 hospitals who found 26% DTIs deteriorated to full thickness ulcers. However, they found only 5% of DTIs resolved and our study showed a higher percentage of resolution (20%). Both Richbourg et al's55 and Sullivan's17 study included DTI locations other than heels. No specific information about progression of DTI on the heels is documented and both studies were in acute care hospitals. Cox et al's53 study of thermography use in NH residents with discoloured intact skin (eg, blanchable erythema, stage 1 PrI, or DTI) noted that 45% of participants experienced resolution of the area, which is higher than our findings. The difference in findings may be related to length of follow‐up as the present study followed up participants for 16 weeks and the Cox et al's study followed up participants for 14 days. Additionally, there may be differences in anatomical location as Cox et al's study included locations other than heels. As noted earlier, the anatomy of the heel critically differs from typical truncal PrI sites and this may influence heel skin response to prolonged loading.56

Our study also showed DTI progression and persistence despite NH staff use of preventive interventions (support surfaces, heel protectors, and pillows for off‐loading heels). This may reflect the multifactorial nature of PrI development and natural history as there are multiple studies demonstrating a wide range in PrI progression that are due to multiple factors, not just preventive care measures.57, 58, 59

While preventive off‐loading with pillows occurred for the majority of participants, use of heel boots was not the primary strategy and was used for only 21% of all participants. Others have found that use of heel boots is beneficial in preventing heel ulcers.8, 9, 60, 61 Baath et al showed that use of a heel suspension boot significantly reduced heel PrI incidence when applied early in the care delivery system—beginning with ambulance transport. Similarly, Rajpaul and Acton demonstrated that use of heel boots in an acute hospital adult population resulted in decreased incidence of heel PrIs.60 No studies examining heel boot use specific to the NH population were identified. It may be that heel boot use in NHs is more difficult than acute care hospitals, with problems of poor compliance in wearing the boots or staff concerns with fall risk.61 The lack of heel boot use may be problematic as Wong62 showed that oxygen delivery to the heel decreases after a brief loading of only 15 minutes suggesting that heels should be off‐loaded at all times. In this study, pillows were commonly used for off‐loading. Little data exist to support pillows as effective in preventing heel ulcers and other non‐specific methods of off‐loading heels have not been shown to be as effective as heel boots.63 Further, we were not able to determine if NH staff implemented use of the preventive strategies specifically related to preventing (or treating) heel PrI or if they were implemented as part of general care for all residents (support surfaces and pillows) or implemented because the resident was admitted from an acute care hospital where they had been in use (heel boots). This is an area where further more definitive study is needed.

That older age and presence of diabetes were significant covariates in the predictive models supports others who have found a significant relationship between age and diabetes and heel PrIs.57, 58 Delmore et al found diabetes to be a significant and independent risk factor for heel PrI formation in hospitalised patients.57 Similarly, we found diabetes to be significantly associated with heel PrIs among an ethnically diverse NH population. Twilley and Jones found patients with heel PrI were significantly more likely to have peripheral arterial disease (PAD).64 McGinnis et al found the presence of PAD significantly reduced the probability of healing in existing heel PrIs58 and Masaki et al65 showed that heel blood flow decrease in persons with ankle‐brachial index less than 0.8 and transcutaneous oxygen levels below 10 mm Hg when the tissues were loaded. Our findings support these conclusions as CAD and PVD were highly correlated with diabetes and in bivariate analyses CAD was related to heel PrI damage.

That there were differences in the results for right and left heels may be related to the positioning protocol for the study. Participants were observed in a right‐side lying position (unless unable to turn to the right side) with the right heel next to the bed. It may be that obtaining readings from the right heel in this position affected the values or were more difficult to obtain. In future studies, care should be taken to rotate positions so that readings are obtained from both right and left heels when next to the bed.

We thought that other factors might influence SEM and we examined possible alternatives for SEM changes. We found no difference in SEM based on comorbidities of hypertension, cancer, PVD, CAD, or diabetes. In our study reporting the PrI outcomes of the trunk, we also found no difference in SEM based on comorbidities.25 However, as the conversion from MDS 2.0 to 3.0 occurred during the study, we do not have MDS 3.0 measures of all conditions on all residents. Participants were examined at the same time of day (in the morning before activities) in a right‐side lying position (or left‐side lying position); thus, it is unlikely that circadian rhythms or body position influenced SEM values in this study. We did not have data on individual hydration status of participants; however, we are interested in localised inflammation and tissue oedema and it is unclear if hydration status plays a role in local inflammatory response. In a prior study with veterans with spinal cord injury, we noted increased SEM values in persons with high level cervical injury and concomitant generalised oedema, yet the pattern of increased SEM with PrI existed even in these persons.24 Evaluation of conditions that may influence localised tissue oedema and inflammatory response and therefore, SEM values should be further examined.

The population for this study was frail older adults and that may limit generalisability. However, others have shown no effect of age on measures of SEM at the 2.5 mm depth (as was used in this study) but only on more superficial depth measures.66, 67 Further, there was a low correlation between SEM values and age (r = .18 and r = .19 for right and left heel, respectively) in this study although as the mean age was 76.5 this may only indicate that little difference exists between SEM values across older ages.

Our findings that decreased functional status was a significant predictor of heel PrIs is new information but not surprising as decreased activity and mobility have previously been associated with PrI development in general.30 This may be of particular clinical use in terms of risk assessment as the Braden Scale was not significant in the analyses to predict heel damage.

In contrast to our earlier studies, the observation period for this study was only 16 weeks. While significant power was gained through the repeated skin assessments, the period may have been too brief to capture enough incident damage to evaluate the predictive power of SEM among some groups. It may be that an observation period of 6 months is a more robust design for such studies. Over the 3 years of skin observations, the technology for detecting SEM significantly improved. The device used for this study was not specifically designed for high‐volume use in health settings and did not correct for user‐induced pressure of the device against the skin. The high variance associated with the SEM readings may indicate that the device was not sufficiently precise or accurate. Further, application of the device to the heel location is not simple as the heel area is small, bony, and not flat, making it difficult to get full skin contact with the electrodes and this may account for some of the variance as well. Recent studies using devices designed specifically for skin assessments have demonstrated greater interrater and inter‐device reliability and less variability in readings would be an improvement for potential use in clinical settings.27

In addition, all skin assessments were completed by trained research staff. Before SEM measures can be used more widely in clinical practice, it is important to evaluate the use of technology by facility staff. However, the reliability estimates with the high number of observers with limited nursing expertise may be near what would be found with facility staff.

To our knowledge, there are no other studies that have evaluated use of SEM to detect and predict DTI progression and PrIs at the heel location. Based on our findings, not all heel DTIs progress to severe PrIs in the NH setting, many DTI persist in spite of consistent standard prevention efforts, current risk assessment methods may not be adequate for predicting heel PrIs, and SEM detection of DTI may assist in identifying early damage and predicting progression.

For clinicians, a more reliable, objective method of triggering preventive action has substantial value. The potential to predict those DTI which will deteriorate from those that will persist or resolve has important clinical practice ramifications as aggressive interventions for viable tissue rescue could be targeted at those persons with DTI projected to progress at earlier time points.

In sum, SEM measures continue to demonstrate a relationship with PrI development across multiple populations, use of various instrumentation, and with this study, across various anatomical locations. This work suggests technology to measure SEM may be useful for identifying early skin injury that can benefit from intervention. Technology improvements will accelerate this usefulness. Indeed, the small differences in SEM and the large SDs suggest the need for greater precision and accuracy in instrumentation. Newer instruments base PrI detection and prediction on algorithms based on the differences between SEM values obtained over the bony prominence and those obtained at skin sites 2 and 4 cm from the bony prominence. The use of individual delta values (change values) may be more appropriate than use of threshold values (as used in this study) as this approach accounts for individual variation. The use of a delta SEM (change in SEM from the bony prominence to normal skin) approach should be fully examined. Additionally, methods of comparing SEM findings to other means of identifying localised skin and tissue damage such as thermography or ultrasonography should be examined to definitively determine what tissue changes SEM is measuring. Evaluating risk for heel PrI separately from areas on the trunk may be an important distinction for providers to make. In addition, use of effective, targeted prevention practices may result in improved care for NH residents.

Bates‐Jensen BM, McCreath HE, Nakagami G, Patlan A. Subepidermal moisture detection of heel pressure injury: The pressure ulcer detection study outcomes. Int Wound J. 2018;15:297–309. 10.1111/iwj.12869