Comparison of Air-Fluidized Therapy with Other Support Surfaces Used to Treat Pressure Ulcers in Nursing Home Residents—Part 2.
Methods and Procedures
The NPULS study data collection instrument. A multidisciplinary team composed of pressure ulcer experts including physicians, clinicians, academicians, dietitians, physical therapists, pharmacists, outcomes researchers, and risk managers, developed the NPULS data collection instrument to evaluate factors affecting the development and/or healing of pressure ulcers. The team based the instrument on their clinical expertise, information from the literature, and existing pressure ulcer guidelines.1,55,56
Data from the NPULS study. Data for each resident included more than 500 variables related to the medical history, physical examination, laboratory studies, and radiological examinations collected over the duration of the study. Demographic information included age, gender, and race. Data related to the medical history included resident diagnoses, illness severity, activities of daily living, cognitive ability, mobility, and level of continence. Pressure ulcer data included Braden Scale assessments; total number of ulcers per resident; ulcer tissue type; amount of exudates; and the site, stage, and surface area of each pressure ulcer. Skilled caregivers used standard clinical practice methods to measure pressure surface areas by maximum width and length. Variables related to the nutritional assessment included weight, body mass index, albumin, and meal intake. Each resident’s discharge disposition, including an emergency room visit, hospitalization, or death, was recorded when indicated. Data for support surfaces were coded according to CMS codes for Group 1, 2, and 3 support surfaces.
Nineteen data abstractors were trained to use the data collection instrument and the CSI® software system. To ensure reliability, results from trainees were compared to results from experienced abstractors. A reliability check required an overall 95% agreement with experienced abstractors for at least 95% of the assessed criteria.
Comprehensive Severity of Illness (CSI®). The CSI® measures the patient’s total burden of illness using data from the ICD-9-CM codebook. The CSI® maps each of the 12,775 ICD-9-CM diagnoses and more than 2,100 CSI® severity criteria to more than 5,500 criteria sets. Severity criteria are based on objective findings from the physical examination such as temperature and blood pressure, laboratory values, and radiological examinations. Signs and symptoms then are ranked according to their severity — a zero represents the absence of signs and symptoms of illness, while the highest score (Level 4) denotes catastrophic signs and symptoms likely to result in organ failure or death. The CSI® generates a continuous score; each abnormal finding adds points to the overall score. Higher numbers indicate a greater level of illness. The final CSI® score represents a sum of total illness (no maximum value) reflecting the patient’s principal diagnosis and the severity levels of secondary diagnoses, including comorbidities and complications.52,57,58
The CSI® includes a disease component and a nutritional component. Elements included in the nutritional component include serum albumin, lymphocyte count, recent weight loss, body mass index, and criteria related to malnutrition in ICD-9-CM codes 260 to 269.99.57 The long-term care component of the CSI system was used in the NPULS study.
General NPULS population. The NPULS study data were obtained from the medical records of 2,486 adult residents from 109 skilled facilities throughout the US that belonged to either one of five long-term care companies or from one specialized pressure ulcer care center. Residents included in the study were either at risk for or had an existing pressure ulcer. Three months of data were collected for each resident 18 years of age or older with a length of stay ≥14 days during the period from February 1, 1996, through October 31, 1997. Overall data from the NPULS study have been published.52
Population included in Support Surface Analysis. Residents included in this study, a subset of the general NPULS population, had to have at least one pressure ulcer documented in the medical record. Pressure ulcers were defined as lesions caused by unrelieved pressure that damages underlying tissue, usually located at the site of bony prominences.1 Caregivers from the nursing homes were instructed to assess and stage pressure ulcers according to the AHCPR guidelines.1 Additionally, to be included in this analysis, residents had to have been on one of three support surfaces described by the CMS as: Group 1 (overlays and replacement mattresses such as dry pressure mattresses, gel pressure pads, and air and water pressure overlays); Group 2 (non-powered overlays, powered low-air-loss overlays and mattresses, powered low-air-loss reducing beds, and alternating pressure surfaces); or Group 3 (air-fluidized therapy).31
Data analyses. Data were analyzed to assess the healing rate of pressure ulcers using two approaches — the first described as resident level and the second as episode level analyses. Additionally, subsets of Group 1 and Group 2 surfaces were compared to assess potential differences in healing rates on the different surfaces. Group 1 surfaces were categorized as alternating pressure, foam, or water or gel surfaces. Group 2 surfaces were categorized as powered air floatation (low-air-loss), powered pressure-reducing, powered air overlay, or non-powered advanced pressure surfaces.
Resident level analyses. The resident level analysis assessed healing rates from one pressure ulcer per resident, using the largest ulcer when residents had multiple ulcers. Residents had to remain on the same support surface for at least 5 days to be included in this analysis; the assessment period ended if the resident left the support surface for more than 24 hours.
Episode level analyses. The episode level analysis assessed pressure ulcer healing rates using multiple assessments; the time from one assessment to the next was termed an “episode.” Each 7- to 10-day period that residents remained on a support surface generated a separate episode, providing a uniform duration of treatment for episodes of approximately 1 week. Additionally, when residents had multiple pressure ulcers, each ulcer generated a separate episode(s). Therefore, the episode analysis included all pressure ulcers from each support surface used by a resident. For example, a resident with two pressure ulcers assessed on one support surface for 3 weeks, and then on a second surface for 2 weeks, would generate 10 episodes — six episodes from the first surface (2 ulcers x 3 weeks), and four episodes from the second surface. In contrast, using the resident approach, the same resident would generate data limited to measurements from the largest ulcer while the resident remained on the first surface.
Analysis with comparable baseline size. An additional analysis was conducted comparing Group 2 and Group 3 surfaces using episode level data controlled for variability in initial pressure ulcer size by limiting data to Stage III/IV ulcers with comparable baseline sizes.
To control for potential bias from multiple ulcers for each patient and/or multiple episodes per ulcer, hierarchical analyses were performed for both overall episode analyses and for episode analyses used to control variability in baseline size.
Analyses related to secondary objectives. First, the number of hospitalizations and emergency room visits were determined at the resident level. Second, regression analyses evaluated demographic data, pressure ulcer characteristics, and comorbidities that affect pressure ulcer healing using both resident and episode level data. Variables were excluded from the analysis if data were either missing for a substantial number of residents in the subset or if the data were ambiguous.
Statistical methods. Bivariate analyses were used to assess healing rates, while Analysis of Variance and the Duncan Multiple Range Test were used to assess differences between the three groups of support surfaces. Additionally, ordinary least square regression was used to identify the variables most important to pressure ulcer healing. Chi-square testing was performed to compare discrete variables. A significance level of P = 0.05 was adopted.
Results/Findings
Sample size. From the total NPULS study sample, 664 (26%) met study inclusion criteria; namely, they had at least one pressure ulcer with placement on Group 1 (N = 463), Group 2 (N = 119), or Group 3 (N = 82) support surfaces for at least 5 days (see Table 1). The episode analysis included 4,190 assessment periods or episodes — 2,406 on Group 1, 744 on Group 2, and 1,040 on Group 3 surfaces. Comparable baseline sizes between Groups 2 and 3 surfaces occurred for pressure ulcers with an initial area between 20 cm2 and 75 cm2 (P = 0.2816) and included 409 ulcers — 154 on Group 1, 74 on Group 2, and 181 on Group 3 support surfaces (see Table 2).
Demographic data. The average age of all residents was 77.5 years. The mean age on Group 3 support surfaces (67.6 years) was significantly lower than the mean age of residents on Group 1 (79.3 years) or Group 2 surfaces (77.4 years) (P <0.0001). Additionally, significantly more residents on Group 1 (305 out of 463; 65.9%) and Group 2 (76 out of 119; 63.9%) surfaces were women, while more residents on Group 3 surfaces were men (37 out of 82; 45.1%) (P = 0.0016) (see Table 1). Regression analyses failed to show an age or gender impact on healing rates.
Using episode level analyses, the mean age remained significantly lower for residents on Group 3 (67.5 years) as compared with Group 1 (79.6 years) or Group 2 (76.9 years) surfaces. Again, significantly more episodes were assessed for female residents on Group 1 (1,515 out of 2,406; 63.0%) and Group 2 (536 out of 744; 72.0%), as compared to Group 3 (508 out of 1,040; 48.85%) surfaces (see Table 2). Regression analyses did not show an effect due to age, while female gender had a positive association with healing.
Mean ages for the episode analysis with comparable baseline sizes resulted in significant differences between Group 1 (73.0 years), Group 2 (80.4 years), and Group 3 (58.5 years) support surfaces (P <0.0001; Duncan’s Multiple Range Test: each group significantly different from each other). Difference was significant between genders, with more women on Group 2 (55 out of 74; 74.3%) as compared with Group 1 (68 out of 154; 44.2%) or Group 3 (64 out of 181; 35.4%) (P <0.0001) surfaces (see Table 2). Regression analyses did not indicate a significant correlation between mean ages and healing rates, while female gender showed a positive correlation with healing.
Mean weights, height, and race were evaluated at the resident level. No significant differences were found in mean weights for residents on Group 1 (mean = 134.3 lb), Group 2 (mean = 137.5 lb), or Group 3 surfaces (mean = 139.7 lb) (P =0.4144). Although a statistically significant difference in height was found between residents on Group 1 (mean = 5.38 ft), Group 2 (mean = 5.46 ft), and Group 3 (5.46 ft.) (P = 0.0429) surfaces, these differences were not significant using Duncan’s Multiple Range Test. Data for race were only available for 182 out of 664 residents (27.4%); 66.5% (121 out of 182) were Caucasian, 28.6% African American, and 4.9% designated as “other.” Although significantly more African American residents (32.7%) as compared with Caucasian residents (12.4%) (P = 0.0145) were on Group 3 surfaces, the limited sample makes it difficult to draw meaningful conclusions from these data.
Comparisons of subsets of Group 1 and Group 2 surfaces were performed at the resident and episode levels. At the resident level, no significant differences were found between Group 1 surfaces categorized as alternating pressure (N = 16); foam (N = 350); or water/gel (N = 83) (P = 0.41) or between Group 2 surfaces categorized as powered air floatation (low-air-loss) (N = 62); powered pressure-reducing (N = 35); powered air overlay (N = 12); or non-powered advanced pressure surfaces (N = 16) (P = 0.91). Similarly, no significant differences were found at the episode level for Group 1 surfaces categorized as alternating pressure (N = 99); foam (N = 1,806); or water or gel surfaces (N = 445) (P = 0.26) or for Group 2 surfaces categorized as powered air floatation (low-air-loss) (N = 343); powered pressure-reducing (N = 214); powered air overlay (N = 131); or non-powered advanced pressure surfaces (N = 88) (P = 0.13).
Pressure ulcer characteristics.
Number of ulcers/resident. Of the 664 residents, 338 (50.9%) had a single pressure ulcer; 166 (25%) had two ulcers; 70 (10.5%) had three pressure ulcers; 36 (5.4%) had four ulcers; two groups of 17 (2.6% each) had five or six ulcers; five had seven ulcers (0.75%), four had eight ulcers (0.6%); one had nine ulcers (0.15%); and 10 had 10 to 14 ulcers each. Approximately 92% of the residents had four or fewer pressure ulcers, while about 8% had more than four pressure ulcers.
Residents placed on Group 3 surfaces (mean = 3.5 ulcers) had significantly more pressure ulcers than those on Group 2 surfaces (mean = 2.5 ulcers), while residents on Group 2 surfaces had more pressure ulcers than those on Group 1 surfaces (mean = 1.8 ulcers) (P < 0.0001); each group differed significantly from each other using the Duncan’s Multiple Range Test) (see Table 1).
Mean initial size and stage of pressure ulcers. The mean initial size of pressure ulcers, using the resident level analysis, showed significantly larger ulcers for residents on Group 3 (56.5 cm2) as compared to Group 2 (22.2 cm2) or Group 1, (11.3 cm2) surfaces (P <0.0001). Similar, although less pronounced, differences in mean initial ulcer size were seen at the episode level — ulcers were largest on Group 3 (23.4 cm2) as compared to Group 2 (12.1 cm2) or Group 1 (7.8 cm2) surfaces. Analyses for comparable baseline size (20 cm2 to 75 cm2) were limited to Stage III/IV pressure ulcers. Mean initial pressure ulcer sizes showed no statistically significant difference between ulcers on Group 2 (36.6 cm2) and Group 3 (38.7 cm2) surfaces (P = 0.2816) or between those on Group 1 (34.9 cm2) and Group 2 (P = 0.3667) surfaces, while initial size between Group 1 and 3 surfaces differed significantly (P = 0.0159) (see Table 2).
Using resident level analysis, residents on Group 1 surfaces had significantly more Stage I/II ulcers (335 out of 463; 72.6%) than those on Group 2 (54 out of 119; 45.4%) or Group 3 (18 out of 82; 22.0%) surfaces (P = 0.0001). In contrast, residents on Group 3 surfaces had significantly more Stage III/IV/eschar ulcers (64 out of 82; 78.05%) than residents on Group 2 (62 out of 119; 52.1%) or Group 1 (118 out of 463; 25.5%) surfaces (P <0.0001) (see Table 1). Similarly, using episode data, significantly more Stage I/II ulcers were on Group 1 surfaces (1,562 out of 2,406; 64.5%) than on Group 2 (352 out of 744; 47.3%) or Group 3 (413 out of 1,040; 39.7%), surfaces. In contrast, significantly more Stage III/IV ulcers were on Group 3 surfaces (624 out of 1,040; 60.0%) than on Group 2 (381 out of 744; 51.2%) or Group 1 (817 out of 2,406; 34.0%) surfaces (P <0.0001) (see Table 2).
Braden Scores. Analyses used the first recorded Braden score that corresponded with a documented pressure ulcer measurement date. Because the timing of Braden score charting did not correlate consistently with ulcer measurements used to determine healing rates, values were only present for 400 of the 664 (60.2%) residents. Braden scores, at the resident level, were generally worse for residents on Group 2 (mean = 12.6) or Group 3 surfaces (mean = 12.9) than those on Group 1 surfaces (mean = 13.7) (P = 0.0047). Similarly, at the episode level, Braden scores were significantly lower for episodes when residents were on Group 2 (mean = 12.2) or Group 3 (mean = 11.7) as compared with Group 1 (mean = 13.5) surfaces (P <0.0001). Only 511 out of 4,190 (12%) episodes had a Braden score charted during the episode.
