A blinded clinical study using a subepidermal moisture biocapacitance measurement device for early detection of pressure injuries

Abstract This study aimed to evaluate the sensitivity and specificity of subepidermal moisture (SEM), a biomarker employed for early detection of pressure injuries (PI), compared to the “Gold Standard” of clinical skin and tissue assessment (STA), and to characterize the timing of SEM changes relative to the diagnosis of a PI. This blinded, longitudinal, prospective clinical study enrolled 189 patients (n = 182 in intent‐to‐treat [ITT]) at acute and post‐acute sites (9 USA, 3 UK). Data were collected from patients' heels and sacrums using a biocapacitance measurement device beginning at admission and continuing for a minimum of 6 days to: (a) the patient developing a PI, (b) discharge from care, or (c) a maximum of 21 days. Standard of care clinical interventions prevailed, uninterrupted. Principal investigators oversaw the study at each site. Blinded Generalists gathered SEM data, and blinded Specialists diagnosed the presence or absence of PIs. Of the ITT population, 26.4% developed a PI during the study; 66.7% classified as Stage 1 injuries, 23% deep tissue injuries, the remaining being Stage 2 or Unstageable. Sensitivity was 87.5% (95% CI: 74.8%‐95.3%) and specificity was 32.9% (95% CI: 28.3%‐37.8%). Area under the receiver operating characteristic curve (AUC) was 0.6713 (95% CI 0.5969‐0.7457, P < .001). SEM changes were observed 4.7 (± 2.4 days) earlier than diagnosis of a PI via STA alone. Latency between the SEM biomarker and later onset of a PI, in combination with standard of care interventions administered to at‐risk patients, may have confounded specificity. Aggregate SEM sensitivity and specificity and 67.13% AUC exceeded that of clinical judgment alone. While acknowledging specificity limitations, these data suggest that SEM biocapacitance measures can complement STAs, facilitate earlier identification of the risk of specific anatomies developing PIs, and inform earlier anatomy‐specific intervention decisions than STAs alone. Future work should include cost‐consequence analyses of SEM informed interventions.

grams. 1 The use, benefits, and limitations of risk assessment tools (RATs) are well documented elsewhere [11][12][13][14][15] 11 Clinical judgment of nurses, informed by risk tools and STA, however, "achieved inadequate capacity to assess PU risk" 12 and suffered from "high inter-examiner variability." 12,13 Clinical judgment has a sensitivity of 50.6% and specificity of 60.1%. 13 Because of the skill dependency of STA, correct identification of a stage I pressure ulcer has been observed as low as 60% in a diverse group of 1452 nurses. 14 Latency of visual and palpable signs do not address nonvisible cues of tissue damage: PI often occur without prior visual and palpable cues appearing in time to prevent them. In patients at high risk for pressure ulcer formation, nonblanchable erythema can develop in as little as 2 hours from injury. 15 Due to the subjective, latent nature of STAs, there is a clear need for an objective, point-of-care tool for diagnosing or assessing developing PIs on at-risk patients.
A change in subepidermal moisture (SEM) due to local edema or accumulation of interstitial fluid is a biomarker of a developing PI and precedes the appearance of visible or palpable skin changes by approximately 3-10 days. [16][17][18][19][20][21] Electrical biocapacitance of human tissue varies with interstitial moisture content; therefore, SEM changes manifest as changes in biocapacitance, which can be measured by point-of-care devices. 19,22,23 A biocapacitance measurement device 24 that has previously shown inter-operator and interdevice reliability in detection of SEM changes (R ≥ 0.8), as well as agreement with ultrasound detection of hypoechoic lesions consistent with PIs, supporting the potential utility of this device. 19,21,25 Independent studies suggest that a biocapacitance measurement is a useful adjunct to clinical STAs, particularly for prediction of developing PIs, and the data, when used to initiate anatomy-specific interventions may reduce the incidence of hospital-acquired PI. 18,20,[26][27][28][29] The comparative sensitivity and specificity of biocapacitance measurements in detecting PIs relative to the current reference standard of clinical STA is however, not well characterized. This longitudinal clinical study was conducted to evaluate the sensitivity and specificity of assessing SEM using a biocapacitance measurement device in accordance with regulatory guidelines for new diagnostic tests, 30 and further, to characterize the timing of SEM changes compared to the diagnosis of PI by STA scales. Subjects were tracked from enrollment with no observed PI to discharge. Discharge was precipitated by any of three events: (a) confirmation of a pressure injury by an expert "Specialist";

| Study design
(b) discharge from the facility after a minimum of 6 days of observation; or (c) satisfying the maximum enrollment of 21 days. The prospective design tested two objectives; the first, that measured changes in the SEM biomarker are associated with the later manifestation of a PI (stage 1, stage 2, deep tissue injury) to 70% sensitivity and 55% specificity with 95% confidence compared to the reference standard of clinical STA. Second, that the SEM biomarker gives notification of such changes prior to an expert skin assessed PI manifesting at the skin's surface.
A subtle distinction of consequence is that the first objective was based on the detection of a biomarker (SEM) of the later onset of a Although using different descriptions of the delta calculation, the definitions of the SEM delta as >0.5 or Δ ≥0.6 are mathematically equivalent), went on to develop signs of a pressure injury. Further, the use of the device identified changes to the SEM biomarker on average 3.9 days earlier than the nurses' STAs confirming a PI. 20 Study participants provided informed consent in writing. Three categories of clinical staff were organized to execute the trial at each site. Each site's principal investigator, or their formally appointed designee, oversaw the conduct of the entire trial for their site. "Specialists," comprising experienced wound care specialists with accreditation in wound and ostomy or tissue viability care performed risk, skin, and PI diagnoses. Diagnoses by Specialists formed the "gold standard" of the existence or absence of a PI. It was against these gold standard diagnoses that device readings were evaluated. SEM biocapacitance measurements were performed by a separate group of, "Generalists" who received training in operation of the device by the device manufacturer and were tested for proficiency prior to initiation of the study. Generalists were clinical practitioners, generally Registered Nurses, Surgical Assistants, and Medical Doctors and were prohibited for the purposes of the trial from participating in wound care related clinical decisions for enrolled patients. Specialists and Generalists' data were blinded from each other. Each site's protocol compliance and data integrity were overseen and audited by a, "Study Coordinator." All data were loaded to Medrio, an electronic data management system (Medrio Inc., California, USA), and locked; consistent with good clinical practices. A full post hoc audit of all data points was conducted by a Regulatory Body; four minor audit findings were noted relating to two sites, all of which were immediately able to be corrected. No findings affected the integrity of the data. No audit findings were noted for the Sponsor.
Withholding "standard of care" prevention or treatment interventions from enrolled subjects would have been unethical. Interventions were applied to all patients consistent with facility Standard of Care protocols. No interventions were permitted to be withheld nor were any interventions triggered by device readings. Data about such interventions were collected at every visit to allow assessment of potentially confounding effects.
Etiological studies show that early pressure damage does not always manifest into a visible PI. 19,[31][32][33] Researchers of early stage PIs and PI biomechanics demonstrated the inherent reversible nature of early pressure damage. 19,[31][32][33] Reversibility and self-resolution are known phenomena PI. 19,[31][32][33] Some early damage will progress to a PI and some will reverse back to a healthy state, depending on a variety of factors including a patient's overall health and whether an intervention is taken to alleviate pressure and or shear. Moreover, some pressure damage can be (a) stable, (b) not progressing, or (c) reversing. 31 Ultimately this means that increases in SEM delta values will not always lead to a PU, but a PI will be preceded by a change in SEM. Complications to end-point analysis from these etiological realities were considered during trial design and in interpretation of the data.
Eligible study participants were patients ≥55 years of age who could be followed for at least six consecutive days and were determined to be at risk of developing PIs based on a score from validated risk assessment scales (Braden scale, < 15; Waterlow scale, ≥ 10; or Norton scale ≤18), poor nutrition, poor mobility (chair or bed-bound), or completion of a recent medical procedure requiring subsequent immobilization (eg, surgery). The minimum follow-up period of 6 days was chosen to provide a sufficient window of time for PI development, which evolves over approximately 1-5 days. 34,35 Patients were excluded from study participation for any of the following: existing PIs; broken skin at either the sacrum or heels; moisture lesions or incontinence-associated dermatitis (IAD); and biomechanical or other limitations preventing protocol-driven assessments.
Use of the device over broken skin was contraindicated, hence, in part, the exclusion of IAD.
The study focused entirely on heels and sacrum; anatomies accounting for more than 50% and up to 87% of reported PIs. 23

| Study procedures
Participants were followed for at least 6 and up to 21 days unless a pressure injury occurred in the intervening period, in which case they were discharged per protocol. No patients were followed after developing a PI in the study. Each day, participants were assessed for risk of a PI by Specialists using one of three established scales, depending on facilities' existing standard of care: the Braden Scale for Predicting Pressure Sore Risk, 38 the Waterlow Pressure Ulcer Prevention/Treatment Policy score, 39 or the Norton Pressure Sore Risk-Assessment Scale Scoring System. 40 Skin temperature, erythema, edema, consistency (induration/hardness) in relation to surrounding tissue moisture, turgor, and health, as well as patient-reported pain, were assessed in accordance with established guidelines. 1,41 Incidents of PIs were staged as described elsewhere. 1 SEM biocapacitance measurements were obtained from the sacrum ( Figure S1) and heels ( Figure S2). Results reflective of the underlying distribution of moisture depend on multiple readings from areas over and immediately contiguous to the anatomy. Six readings were therefore obtained from the sacral area and at least three readings were obtained from each heel. 42

| Device
The principles underlying the use of biocapacitance as an indirect measure of localized SEM and edema have been reviewed previously. 19,43 Biocapacitance varies with interstitial fluid content. Measurements of tissue biocapacitance may be used as an indicator of SEM. 24 The SEM Scanner (device) ( Figure

| Statistical analysis
This study was powered to detect at least 70% sensitivity and 55% specificity of the device compared to the reference standard of clinical STA, with 95% confidence. A total of 189 patients were enrolled, of which 96.3% (n = 182) patients were listed as intent-to-treat (ITT) Figure S4. Within the 12 sites included in the study, the trials were completed in sites of service shown in Table 1. an additional sensitivity and specificity analysis was conducted using the Bootstrap method. The bootstrap method was applied by sampling, with replacement, from the original data set. Sampling was done on a per subject basis such that all records for a randomly chosen subject were extracted. One thousand data sets were generated using this method, each with the same number of subjects as the original data set. 44 A post hoc receiver operating characteristic (ROC) curve analysis was conducted to augment the description of diagnostic accuracy.
ROC curves estimate and report all combinations of sensitivity and specificity the test is able to provide. 45 Results are expressed as a statistic, the "area under the curve" (AUC), a value ranging from 0 to 1 (ie, 0%-100%) where 0.5 (ie, 50%) represents randomness. Values above 0.5 trend toward increasing diagnostic certainty for the test.
The measure for secondary endpoint ("time to detection") was the number of days between pressure ulcer diagnosis by clinical judgment of the Specialist and the first day of measuring a delta (Δ) value ≥0.6.

| Limitations of the evaluative rubric of sensitivity and specificity
The classic approach 30 UK, respectively). Seven participants' data were not analyzable, resulting in an ITT population of 182 ( Figure S4). The removal occurred before data analysis, which was performed following the a priori established statistical analysis plan. The reason for removal of six subjects' data was necessary because the delta values for each SEM reading point had been entered erroneously and could not be corrected by study staff or site monitors post hoc. One subject with an existing pressure ulcer (exclusion criteria, 1) was also erroneously enrolled into the study for 1 day and therefore did not meet eligibility. This was also a protocol deviation which was noted, reported and rectified by the principal investigator at the site.
With the exception of one site that used the Waterlow Scale, all study sites used the Braden scale for risk assessment as part of routine care. Demographic data are shown in Table 2. Applying PI staging per published guidelines, 1 a total of 26% (n = 48/182) of the ITT population were diagnosed with a PI (Table 3), and the incidence by anatomical site was 11% (n = 48/437 (Table 5). Table 4 provides a baseline skin profile of the enrolled subjects. Over 85% of the ITT population with available data were identified as having normal skin color and, a small percentage showed evidence of redness at the sacrum (12.60%), left heels (6.67%), and right heels (4.88%). Over 99% of the ITT indicated that the skin was blanchable at day 0. Enrollment data also show a majority of subjects were observed as having normal skin temperature, edema, induration, moisture, turgor, and health with no gross observations of bruising, cracked skin, abrasions, dehydration, pain; and moisture lesions were absent at the sacrum.

| SEM results
True positive, True Negative, False Positive, and False Negative classification results for the ITT population are shown in

| Safety
The measure of analysis was the percentage of device-related adverse events reported in the study. Adverse events in five enrolled subjects (2.6%) were reported. Four of the five were categorized by the principal investigators as unrelated to the study, with the remaining one reports of adverse events were related to the use of the device or from prevention or treatments of PIs.  Figure S5).

| Characteristics of non-PI skin changes
Visible or palpable changes in skin characteristics identified during STAs were documented for all participants (Table 6). Some changes indicative of PI (eg, redness or warmth) were documented but were not associated with a diagnosis of PI. Red and nonblanchable skin 9 3 6 Red and blanchable skin 69 24 45 Other signs of skin changes (no PI) Changes in skin temperature 58 14 44 Changes in skin firmness 13 3 10 Note: Frequency of skin changes observed in cases in which pressure injury was ruled out during clinical skin and tissue assessment. Assessments were performed by wound care specialists with accreditation in wound and ostomy or tissue viability care. Abbreviations: PI, pressure injuries.  This study did not meet the targeted endpoint for specificity of at least 55%. Two hundred and sixty-one (n = 261) positive SEM deltas were classed as false positives resulting in a specificity approaching 33%. Given the comparison of nonvisual to visual skin damage, it is not surprising that a lower specificity is observed in this study. The Investigators considered a range of explanations for these specificity results.
1. Clinical judgment as an inadequate index value: Risk assessment tools have good predictive capacity, moderate sensitivity and specificity but have highly variable inter-rater reliability. 12,13,47 Clinical judgment has moderate sensitivity and specificity but has poor predictive capacity and inter-rater reliability. 48 This study utilized clinicians with a high level of experience in STA (to represent as the "gold standard" comparator); nonetheless, such assessments are subjective and not the error-free requirement of an index value necessary for complete evaluation of a new diagnostic test. A true assessment of the biomarker is classed as a negative PI result in this test if the PI did not manifest.
2. Reversibility and self-resolution before the damage threshold is reached: Research by Oomens et al showed, "…tissue damage is initiated at a cellular level" and, "unloading the tissue will restore the supply of oxygen and nutrients to the tissue" to return tissue to homeostasis. 33 Further, Halfens 31 noted subgroups of patients with grade 1 pressure ulcer when they were evaluated over multiple days: 22.1% resolved, 22.1% deteriorated, 35.3% unchanged, and 20.6% disappeared (thought to be an initial misdiagnosis or resolution).
Although it is not possible to distinguish between damage that will and will not reverse, application of the right interventions provided before the damage threshold is reached results in tissue "resetting" to "normal homeostasis," namely the restoration of oxygen supply and nutrients to the tissue and removal of waste products. 31

| CONCLUSIONS
Latency between the initial onset of pressure damage and the subse- Even though not all anatomies exhibiting elevated SEM deltas will proceed to eventually develop a PI, it is important for health-care providers to be aware of the early warning signs so they can take risk- Sensitivity F I G U R E 1 Receiver operating characteristic curve for performance of the study device relative to the gold standard of skin and tissue assessment. Receiver operating characteristic curve illustrating diagnostic sensitivity and specificity of the investigational device in detecting pressure injury. AUC, area under the curve. CI, confidence prevention activities (eg, anatomy-specific offloading) on the basis of positive SEM deltas as an adjunct to clinical judgment, rather than solely on risk and visual STAs have shown material reductions in the incidence of PIs. 26 As noted earlier, further analyses on prevention bundles and cost-consequences are still necessary in future work.

ACKNOWLEDGMENTS
Authors listed were Principal Investigators at each of their respective named sites. Each reviewed drafts of the manuscripts. The authors wish to acknowledge the following members of the study teams at individual sites who provided technical help but did not meet the requirements for authorship.