INTRODUCTION

More than 650 pressure injury (PI)-related articles were added to the literature in 2022, marking an extremely productive year for PI clinicians, researchers, and educators. Not surprisingly, it is a very diverse literature that encompasses the breadth of topics needed to remain current in the field. This diversity is also seen in where PI articles originate. High-quality papers are increasingly published not only from the areas typically associated with PI research (such as the US, Europe, and Australia) but also from Asia and South America. Although the many implications of the COVID-19 pandemic for PIs remain a frequent topic in this new literature, the authors also noticed an uptick in the number of smaller clinical trials being published, something that seemed noticeably absent early during the height of the pandemic. The authors expect that this diversity in the PI literature will continue in the coming years.

For this fourth annual curated review of the PI literature, the authors reviewed every abstract identified through a PubMed search using the terms pressure ulcer, pressure injury, or decubitus ulcer limited to articles in English published in 2022. The authors selected six articles based on expert opinion-considering articles' methodological rigor (ie, preferring high-quality evidence such as RCTs), innovation or novelty, and likely interest-while ensuring a diversity of topics. These are not necessarily the six "best" articles from the past year, but rather the authors believe that they represent important additions to the field.

Articles were selected to represent a range of topics including PI epidemiology, risk factors, detection, and healthcare quality. The authors considered not only original research but also meaningful systematic reviews. To avoid potential bias in article selection, the authors avoided ones they authored or that were published in the present journal, Advances in Skin & Wound Care.

For each article, the reference is first provided in bold text. Next, the authors summarize the key findings from the article. Finally, the authors offer a commentary on why they believe the article is important or innovative.

ARTICLE 1

Zhang H, Ma Y, Wang Q, Zhang X, Han L. Incidence and prevalence of pressure injuries in children patients: a systematic review and meta-analysis. J Tissue Viability 2022;31:142-51.

This systematic review and meta-analysis aimed to quantify the incidence and prevalence of PIs among hospitalized children. Eligible articles were cross-sectional or cohort studies that reported PI incidence and/or prevalence in hospitalized children and also specified the age of children. Studies of PIs in the context of single diseases that might not be representative were excluded. A detailed literature search was completed using multiple databases from their inception to July 20, 2020, and there was no restriction based on language. Starting from a pool of 6,672 articles, after elimination of duplicates and ineligible studies, a total of 30 articles were included in the systematic review and 24 in the meta-analysis.

Selected articles were mostly at low risk of bias. The resulting sample consisted of 251,501 pediatric patients across 11 different countries. The estimated pooled incidence of PIs in all participants was 13.5% (95% CI, 10.5%-16.5%). In the five studies that evaluated only neonates, the incidence was 15.1% (95% CI, 12.3%-17.8%). The estimated pooled prevalence was 12.2% (95% CI, 8.0%-16.3%). There was heterogeneity among the studies; incidence was higher in ICUs compared with other settings and lower in Europe compared with other regions. The occiput and sacrococcygeal areas were the most affected areas but accounted for approximately only one-third of the PIs. The authors concluded that the incidence of PIs in children is higher than what has been previously reported. Further research is needed to better understand which children are at highest risk and which preventive measures are most effective.

Comment

In their 2010 review, Kottner et al1 highlighted the methodological limitations of the existing literature on PI frequency in children while estimating an overall incidence of 7% and an incidence of 26% in the pediatric ICU. This new article by Zhang et al now includes many new studies with more than 150,000 children-and the results are alarming. Their findings suggest an incidence almost double that reported by Kottner et al1 and higher than that reported in adults.2 Clearly, PIs are a major issue in children.

Having surveyed the PI literature for 4 years now, it is surprising how few articles address this problem given its magnitude. Both 2007 and 2019 National Pressure Injury Advisory Panel white papers3,4 emphasized the lack of evidence-based research on which to base clinical practice, stating that most protocols for children are extrapolated from adult practices. Many years later, their call for further research to optimize PI prevention and treatment in children still rings true. Only once we know more is progress likely to be made in reducing this rate.

ARTICLE 2

Cao Y, DiPiro ND, Krause JS. Staying pressure injury free: the role of modifiable behaviors. Arch Phys Med Rehabil 2022;103:2138-44.

The purpose of this cross-sectional survey was to identify potentially modifiable behavioral factors in people with chronic spinal cord injury (SCI) that are associated with being free of PIs in the previous year. The survey sample was identified through a single specialty hospital and two state-based surveillance systems. Study participants were at least 1 year post-SCI and had some residual impairment. Participants were invited to complete a self-reported assessment that captured key demographics, as well as information on weight, diet, alcohol use, prescription medication misuse, nonmedical substance use, cigarette smoking, and exercise. Sixty percent of potential participants responded to the survey, and complete information was available from 3,817 individuals.

Study participants had a mean age of 48 years and were an average of 12 years post-SCI, and 63% were nonambulatory. Overall, 67% of the participants reported being PI-free in the year prior to the survey. People who were PI-free were more likely to be ambulatory, had higher incomes, were non-Hispanic Whites, and were women. Four of the eight behavioral factors examined were associated in multiple logistic regression models with being PI-free. The odds of being PI-free were higher in those reporting an excellent, healthy diet (OR, 1.69; 95% CI, 1.15-2.48); doing planned exercise at least once per week (OR, 1.27; 95% CI, 1.09-1.47); and with self-reported average weight (OR, 1.37; 95% CI, 1.10-1.70) or overweight (OR, 1.32; 95% CI, 1.06-1.64). A higher frequency of prescription medication use for depression, spasticity, and pain were each associated with significantly lower odds of being PI-free. Alcohol and tobacco use, nonmedical substance use, and using prescription medications for a purpose other than intended were not associated with being PI-free.

The authors concluded that there are behavioral factors associated with being free of PI. Interventions targeting these behavioral factors, such as promoting healthy eating and regular exercise, could promote optimal health while also preventing PIs.

Comment

Much of the literature on PI development has focused on understanding the complex interplay between intrinsic and extrinsic factors that predispose individuals to PI development.5 Rarely do we ask the related opposite question: How are some high-risk people successful in avoiding PIs? This is certainly a highly relevant question among people with SCI, at least one-third of whom are likely to be impacted by a PI.6

The present manuscript is from a team that already has an extensive record examining behavioral factors associated with consequences of PIs.7 They now expand on their previous work by looking at a larger sample and solely considering the presence or absence of a PI. Perhaps not surprisingly, people who engage in healthy behaviors around nutrition and exercise were less likely to have had a PI. Although it may be that eating healthy results in a better nutrition status and skin that is more resilient in the face of pressure, this study is cross-sectional and cannot address causality. Eating healthy and doing more exercise may just be associated with other "healthy" behaviors such as more frequent repositioning.

A recent study has emphasized, though, how few patients with SCI actually engage in the "healthy" behavior of regular repositioning; only 25.6% of participants reported repositioning themselves every 2 hours when in bed at night.8 Although further research is required to better understand these associations, and the use of self-reported information could be an important study limitation, it is hard to argue with the potential benefits of counseling at-risk individuals to follow a healthy lifestyle.

ARTICLE 3

Ousey K, Stephenson J, Blackburn J. Sub-epidermal moisture assessment as a prompt for clinical action in treatment of pressure ulcers in at-risk hospital patients. J Wound Care 2022;31:294-303.

The aim of this study was to assess the use of subepidermal moisture (SEM) technology as a diagnostic tool to assist in the prevention of PIs in at-risk hospital patients. Specifically, do abnormal SEM measurements (defined as a SEM [DELTA] reading equal to or above 0.6) result in changes in clinical therapy? Assessments on patients were collected from 25 facilities in the UK, Ireland, Belgium, and Spain from June 2015 to February 2020. Patients included in the study were 18 years or older, could provide verbal consent, had unbroken skin on their heels and sacrum, and had a standardized risk assessment score at or above medium risk. The SEM assessments were performed on the sacrum and both heels, and the presence of nurse actions, such as a new support surface or use of heel boots, was recorded following the scan.

Overall, there were 1,995 patients assessed a total of 15,574 times with a median of 13 SEM assessments per patient. An abnormal SEM occurred in 83.9% of patient assessments. At individual sites, SEM measurements and the presence of skin reddening were often discordant. For example, among 3,194 assessments of reddened sacral skin, the SEM was abnormal for 1,990 (62.3%). Similarly, in 12,265 sacral assessments without reddening, the SEM was abnormal in 4,698 (38.3%). When the SEM was abnormal, nurse actions occurred 34.8% of the time, compared with 22.0% when there was no SEM prompt. Results were similar for cases without reddening of the skin (34.8% vs 21.1%). In a multilevel multiple logistic regression model considering the SEM result and skin reddening, both a SEM prompt (OR, 1.99; 95% CI, 1.79-2.23) and skin reddening (OR, 1.45; 95% CI, 1.34-1.57) were independently associated with a nurse action. The authors conclude that the integration of SEM technology into existing PI prevention care pathways improves nurse clinical judgment and results in timely, targeted interventions.

Comment

Suggestions that SEM measurements may help in the early detection of PIs go back now over 15 years,9 and detailed descriptions of this mechanism of action have been published.10 In assessing the use of any new technology on clinical practice, such as the introduction of SEM measurement, one often considers three related questions. First, does the new technology provide additional diagnostic information? Second, does this diagnostic information lead to changes in therapy? Finally, do the changes in therapy result in improved patient outcomes? The current manuscript by Ousey and colleagues expands on prior studies11 in addressing the first two questions, using a large, multinational sample. In many cases in which the SEM scan was positive, there was no visible skin reddening, suggesting that the SEM scan is providing additional diagnostic information for individuals at risk of further skin damage. It remains uncertain how to interpret the diagnostic information in those cases with skin reddening but a negative scan. Are they not at risk of further PI development? There was also no mention of skin tone, which could impact the detection of "reddened" skin and the utility of SEM measurement.

The diagnostic information also had a sizable impact on therapy, with the likelihood of a change almost doubled in those with a positive scan. These results suggest that clinicians are finding SEM assessments useful in their practice and in making decisions for care. However, many of the SEM assessments were positive and did not result in a nurse action. The reason for this is unclear, but perhaps it relates to the multiple assessments done for each patient. New actions would not be taken each day.

Ultimately, though, we care most about the impact on PI incidence, and in particular, whether therapies initiated in response to a positive SEM scan prevent the development of stages 3 and 4 PI. The present manuscript provides no information on this question, but other studies have shown reduced PI rates following the implementation of SEM scanning.12,13 These were mostly pre-post studies with no control group, so the strength of this evidence is not high. Obtaining further high-quality data on the efficacy of SEM assessments would be welcome.

ARTICLE 4

Levy JL, Lima JF, Miller WM, et al. Machine learning approaches for hospital acquired pressure injuries: a retrospective study of electronic medical records. Front Med Technol 2022;4:926667.

Machine learning is increasingly being used to identify predictors of hospital-acquired PIs (HAPIs). This study aimed to evaluate the predictive performance of different machine-learning techniques compared with more traditional statistical approaches. Data used in this study were from Dartmouth Hitchcock Medical Center, April 2011 to December 2016. Patients had to be 18 years or older, have a hospital stay of at least 3 days, and have at least three recorded Braden Scale measurements. Investigators considered HAPI present when there was a related International Classification of Diseases code, supporting nurse documentation, and a consultation from the wound care team. Potential predictors of HAPI were selected from variables in the electronic medical record but did not include laboratory results. Individual Braden subscales were used as predictors, considering both the lowest recorded score on that subscale and the average subscale score from all Braden assessments prior to the PI.

The study included a total of 57,227 hospitalizations with 241 HAPI cases. A logistic regression model and 11 machine learning models (including Decision Trees, Random Forest, and XGBoost) were each evaluated using the area under the receiver operating characteristic curve (AUC or c statistic). The contribution of each individual predictor to the predicted probability of a HAPI was determined using Shapley Additive Explanations. Models varied considerably in their performance; the worst models, k-nearest neighbors and Decision Trees, had an AUC of 0.75 and 0.76, respectively. Logistic regression performed better than or equal to all of the machine learning models with an AUC of 0.91.

Using Shapley Additive Explanations, all models found friction, average mobility, and a patient's diet (NPO) to be important predictors. In addition, the logistic model found low nutrition, average activity, and average moisture scores from the Braden Scale to be highly important predictors; in the two best machine learning models, XGBoost and Random Forest, a history of smoking was of high importance. This study concluded that real-time predictive models for developing HAPI can be integrated into the electronic medical record. In doing so, it is important to compare machine learning models with more traditional statistical approaches, and disagreements among models in what constitute important predictors must be reconciled.

Comment:

Although potentially new predictors of PI development continue to be described in the literature, the actual tools for identifying adults at high risk for PI development have undergone little change over the past 30 years. These tools, such as the Braden and Norton Scales, were developed based on decades of clinical experience and have repeatedly demonstrated predictive validity.14 This not only allows the more efficient use of preventive interventions, but also helps clinicians target those specific risk factors that place individuals at greatest risk. However, it has also been long recognized that models based on empirical data and statistical analyses tend to outperform even experienced clinicians.15 We have eagerly awaited the application of new techniques in information sciences to the field of PI and expected that it would greatly enhance our ability to identify at-risk individuals. A recent systematic review identified 23 such studies that used machine learning to predict HAPIs.16 That review article not only highlights some methodological challenges in performing machine learning but also shows how studies all identify different risk factors for PI development. It is unclear, then, what are the specific patient characteristics we should target for interventions.

This new article by Levy and colleagues further dampens our enthusiasm for machine learning in PI prediction. Twelve different modeling approaches were used, but the logistic regression model, which has been used for over 30 years in developing predictive models for PIs, was equal to or better than these machine learning approaches. Moreover, the most important predictors in the logistic model were based on the Braden subscales of friction and shear, nutrition, and mobility, whereas the machine learning models devalued nutrition and considered a history of smoking as important. The AUC for the logistic model was 0.91, suggesting that outstanding model performance could be achieved when supplementing information from the Braden Scale with limited other clinical data.

One area of concern is that the rate of HAPIs in this study was very low (0.4%), although this rate is similar to that reported by the CMS. In part, this low rate may reflect the inclusion criterion of a consultation from the wound care team. This could suggest that only the most significant wounds were being detected-although it is true that these are also the wounds we are most interested in preventing. In addition, although the Braden Scale was supplemented with information in the medical record such as cigarette use, body mass index, and NPO status, other important predictors may have been omitted. At least for now, in predicting PIs, it should not be assumed that new is necessarily better.

ARTICLE 5

Chen Z, Gleason LJ, Sanghavi P. Accuracy of pressure ulcer events in US nursing home ratings. Med Care 2022;60:775-83.

Pressure injury rates are a key indicator of nursing home quality and are publicly reported through the CMS Nursing Home Compare (NHC) website. This study examined the accuracy of the PI data reported on the Minimum Data Set (MDS) that is used in calculating the rates reported by NHC. The study sample consisted of people with a Medicare hospital claim for a present-on-admission PI between 2011 and 2017 and who also had both a nursing home MDS discharge assessment within 1 day of hospital admission and an MDS readmission assessment at the same nursing home within 1 day of hospital discharge. The PI was considered a primary diagnosis if it was in the admitting, first, or second diagnosis field of the hospital claim; if it was listed in any diagnosis field after the second, it was considered a secondary diagnosis. The accuracy of the MDS data was described as the percentage of patients with a PI on the hospital claim who also had a PI on one of the MDS assessments, stratified by whether it was a primary or secondary diagnosis, a long- or short-stay resident, and by PI stage.

There were 114,729 hospital claims with a primary PI diagnosis and 293,617 with a secondary diagnosis. Among people with a primary PI diagnosis, only 74.7% were recorded in the MDS. For a secondary PI diagnosis, an MDS record was present in 52.1% of cases. In 29.8% of primary diagnoses and 53.9% of secondary diagnoses, the reported stage from the two data sources differed by more than one stage. Accuracy of the MDS reporting was better for short-stay residents and when the PI was more severe. There were 42,467 short-stay claims filed by nursing homes that included a stage 2, 3, or 4 PI, but the MDS did not record the PI in 7.3% of these cases. Little correlation was noted between PI rates and nursing home star ratings. The authors conclude that PIs are substantially underreported in NHC, and policy changes are required to enhance monitoring of nursing home quality.

Comment

Pressure injury rates are unacceptably high. One of the most basic principles of quality improvement is "if you don't measure it, you can't improve it." The nursing home MDS grew out of recommendations from the 1986 Institute of Medicine report, Improving the Quality of Care in Nursing Homes, calling for the creation of a comprehensive assessment tool that would provide standardized data on all residents.17 The first quality measures based on the MDS were developed in the late 1980s, and the public reporting of quality measures based on MDS data began in 2002.18 The five-star composite rating system has been in place since 2008. Measures of PI have always been an important component of this rating system. Yet there have also been many concerns regarding these quality measures including a lack of risk adjustment,19 potential for gaming of facility-reported data,19 and lack of correlation among the measures, suggesting that they are not capturing a single underlying quality construct.18

Chen and colleagues now remind us of problems with the MDS data on PIs. Very simply, the MDS substantially undercounts PIs. Among nursing home residents with a hospital stay, even when a PI is listed in the admitting, first, or second diagnostic field of the Medicare claims, it is not listed in the MDS 22.4% of the time. If we consider PIs anywhere after the second diagnosis field, 45.0% were missed by the MDS. The stage of the PI also often differed between the Medicare claims and the MDS, although we cannot assume which is more accurate. The lack of a strong correlation between PI rates and star ratings could suggest that the underreporting is not uniform across nursing homes, although other interpretations are also possible. As once again emphasized by the National Academies of Sciences, Engineering, and Medicine in their most recent 2022 report on nursing homes, ensuring the quality and safety of nursing home care should be a priority.21 Is it not about time we get sufficiently accurate data to help make this happen?

ARTICLE 6

Dynan L, Smith RB. Sources of nurse-sensitive inpatient safety improvement. Health Services Res 2022;57:1235-46.

The purpose of this study was to examine the association between hospital expenditures for nursing education and staffing on three nursing-sensitive patient safety outcomes: stage 3 or 4 PIs, central venous catheter-related bloodstream infections, and perioperative venous thromboembolism. Study data were obtained from 150 Florida hospitals followed for 12 years from 2007 through 2018, resulting in 1,800 observations. For each hospital-year, investigators reported expenditures for non-wage-related educational activities and for non-physician hospital staff per 1,000 inpatient days. The three outcome events were defined by the Agency for Health Research and Quality Patient Safety Indicators algorithm. Analyses controlled for several hospital characteristics, including ownership, teaching status, and proportion of inpatient days covered by Medicare and Medicaid, as well as for two policy initiatives implemented during this time frame that focused on patient safety. Associations between expenditures and outcomes were examined with a two-way fixed-effect regression model that also considered interactions with the policy initiatives.

Over the 12 years of the analyses, large decreases were seen in all of the patient safety indicators, including an 84% improvement in the PI rate. Expenditures on education increased by 79.3%, and staffing increased by 30.5%. Education expenditures showed no significant association with PIs, whereas a 1-SD increase in staffing was associated with a 31.4% decrease in the PI rate (P < .001). In those hospitals impacted by the policy change, the staffing increase was associated with a 68.5% reduction. Neither education expenditures nor staffing was independently associated with the other two patient safety indicators, but significant interactions were noted. Increased education expenditures combined with policy changes resulted in a 16.6% reduction in catheter-associated bloodstream infections (P < .001) and a 4.6% reduction in perioperative venous thromboembolism (P < .05). These results indicate that tradeoffs exist between funding education and training of staff versus expanding the number of staff when trying to achieve patient safety objectives. What works for PI prevention may not work for other patient safety outcomes.

Comment

Knowledge about PIs is extremely low among nurses and other healthcare professionals. A recent systematic review of nurse and nursing student knowledge on PI prevention found a pooled score on the Pressure Ulcer Knowledge Assessment Test of 51.5%.22 One might expect, then, that educational interventions would result in improved outcomes. Although an old pre- and post-educational intervention study did find reduced PI rates,23 a recent Cochrane review found few randomized clinical trials with limited evidence to suggest that education of healthcare staff can prevent PIs.24 Observational data of nurse staffing levels and PI outcomes show some differences but overall do suggest that higher staffing levels are associated with better outcomes.25-27 These observational data are supported by a recent quasi-experimental study in which hospital units receiving added staffing in the form of an advanced practice nurse had better PI care than units receiving no extra staffing.28 Given limited resources, what is a hospital to do: invest in more nursing education and training, or hire additional staff? Based on Dynan and Smith's analysis, the answer for PIs seems to be hiring more staff.

Interestingly, education was more important than staffing for the other patient safety outcomes examined. Although the reason for these differences is uncertain, perhaps it relates to the nursing time or the level of knowledge and skill that is required to successfully complete these different tasks. High-quality PI prevention is time intensive, and more staff may make all the difference. For managers trying to allocate scarce resources to address many different patient safety concerns, it remains unclear how best to proceed, and decisions may need to balance perceptions of where the greatest problems are. Further studies on how best to use resources to achieve goals are clearly needed.

CONCLUSIONS

Overall, 2022 was another interesting year for PI research and one in which the field saw a large increase in published articles. Although there was a small increase in the number of clinical trials, most were relatively small and did not meet the criteria of sufficient innovation and importance to be highlighted in this review. It is the authors' hope that future years will provide such high-quality studies to further guide PI care. Among the important topics addressed this year are PIs in children, the use of SEM, the role of machine learning in predicting PIs, and the tradeoffs between investing in nursing education versus staffing. In addition, although much can be learned from trying to understand why someone does develop a PI, we should also be trying to learn why some high-risk people do not. As we continue through 2023, we shall see what new ideas in PIs are published.

PRACTICE PEARLS

* Pressure injury incidence and prevalence remain high in children.

* A healthy lifestyle may be important in preventing PIs in people with SCI.

* Subepidermal moisture assessments assist in the identification and management of people at risk of PIs.

* Machine-learning models to predict PI development may not be more useful than more traditional approaches.

* Investing in more staff may result in lower PI rates than investing in nurse education.

REFERENCES