Pre-operative Waterlow score and outcomes after kidney transplantation

Pre-operative risk prediction tools undertaken prior to major surgery can support counselling and consent processes and aid the development of targeted intra- and post-operative care. Several scoring systems have been established and validated in the context of general surgery with variable sensitivity [1]. One example is the Waterlow score, first introduced in the 1980s as a nursing tool to stratify patients at risk for development of decubitus ulcers (pressure sores) [2]. Using a multisystem approach, with weighted scores based on several variables (Fig. 1), patients are categorised into ‘at risk’, ‘high risk’ or ‘very high risk’ for developing decubitus ulcers. Due to its ease of use, it is commonly utilised by UK hospital staff for this specific purpose and forms part of a standard nursing assessment for acute medical and surgical admissions [3‐5]. However, there is conflicting data with regards to the association between Waterlow scores and development of pressure ulcers, with positive [6] versus negative [7‐9] reports of its predictive ability. Some of the criticism labelled at Waterlow scoring is the inter-rater variability, poor concordance and low sensitivity that limits generalizability.

While the original creation of the Waterlow score was never intended be used as a predictive tool [10], due to its ubiquitous assessment by surgical nurses it has been frequently assessed as a surrogate measure to predict adverse clinical outcomes after surgery. In a systematic review of the general surgery literature, summary evidence suggests high pre-operative Waterlow scores (> 20) correlate with increased risk for morbidity, mortality, length of stay and intensive care unit (ICU) admission [11]. Increasing Waterlow scores are also strongly associated with post-operative infection risk after neck of femur fractures [12]. From a transplant perspective, Khambalia and colleagues have shown high pre-operative Waterlow scores are associated with total length of hospital stay and intensive care unit length of stay following simultaneous pancreas and kidney (SPK) transplantation [13]. No study has explored Waterlow scores in the context of kidney-alone transplantation, with the only solid-organ transplant outcome data in the literature limited to length of stay in the work from Khambalia et al. To date, no universally accepted pre-operative risk prediction tool exists in the context of solid organ transplantation.

Considering its ubiquitous use in pre-operative settings, it is likely Waterlow scores may be perceived as an acceptable surrogate measure to predict adverse outcomes after surgery. In the setting of transplantation, no data exist to show any association between Waterlow scores and adverse clinical outcomes independent of other important factors. Therefore, the aim of this study was to study the association between Waterlow scores on admission for kidney transplant candidates and post-operative outcomes including length of stay, risk of emergency re-hospitalisation within 90-days of surgery and mortality.

In this large single-centre analysis, we observed many kidney failure patients being admitted for kidney transplant surgery have raised Waterlow risk categories when checked pre-operatively as part of routine nurse admission checks. As a surrogate measure for risk stratification, ‘very high risk’ Waterlow group (as a categorical variable) or Waterlow score (as a continuous variable) had an independent association with increase length of stay after transplant surgery. No association was observed between any Waterlow risk group or score with emergency 90-day readmission rates. When analysed alone any Waterlow risk group or Waterlow score was associated with increased risk for post-transplant mortality. However, after adjustment with baseline variables, no association remained with either Waterlow risk groups or Waterlow score and mortality. Our analysis raises some interesting discussion points about the utility and value of using Waterlow risk score, or other surrogate measures of ‘fitness’, suggesting caution in their utilisation as a proxy measure for adverse outcomes that is outside of their intended application.

As a recommended tool for pressure ulcer risk stratification among acute medical and surgical admissions [16], the simplicity of Waterlow scores has made them commonplace in the United Kingdom. However, compliance rates for adherence to Waterlow score charting are variable across reports [11]. One of the major limitations in our study was the significant number of missing data with regards to Waterlow scores pre-operatively, which may be related to inadequate completion pre-operatively or an informatics data capture issue. These are unlikely to be missing at random as illustrated in Table 1, with missing participants tending to be at the low risk end of the spectrum for most baseline demographics. Inclusion after multiple imputation may be skewing the median Waterlow scores artificially higher. Indeed, the analysis performed without including missing data shows even less importance of the Waterlow score and may be more accurate.

The other major issue with the Waterlow score is studies tend to report variable inter-rater reliability due to the presence of subjective variables [17]. This is a concern as it questions the generalisability of the Waterlow score between different assessments and/or assessors.

The greatest issue identified in this analysis is the questionable utility of Waterlow scores as a surrogate measure for identifying kidney transplant candidates at higher risk for adverse post-transplant outcomes. While intuitively the subjective components of Waterlow scoring can render higher grades based upon personal bias, the failure of Waterlow scores to identify adverse post-transplant outcomes independent of other variables (apart from prolonged hospital stays post-operatively for ‘very high risk’ patients) demonstrates its inadequacy as a risk predictor. However, that was never its intended use. Even in its original context of risk stratification of patients at risk for pressure sores, the superiority of Waterlow scores versus rival scoring systems is not clear. In a prospective single centre study, Gurkan et al. compared the predictive capacity of three pressure sore risk tools in 250 patients undergoing abdominal surgery: Norton, Braden and Waterlow risk assessment scales [18]. They observed the Waterlow score demonstrated the best values of predictive validity among the three scales in the assessment of pressure injury risk, with sensitivity (100%), specificity (48.1%), positive predictive value (20.8%) and negative predictive value (100%) demonstrated for the Waterlow score (a cut-off point of 10). However, all three scales had low specificity despite high sensitivity in terms of a good risk prediction. In a narrative review of 26 studies, Charamboulous et al. confirmed the predictive validity of the Waterlow is characterised by high specificity and low sensitivity, with inadequate inter-rater reliability due to lack of clear definitions within categories and differentiating level of user knowledge [17].

Our results are important as published literature extrapolating use of Waterlow score from risk assessment of pressure sores to adverse post-surgical outcomes has seemingly favourable evidence. In a systematic review of published evidence in the surgical literature, Nayar and colleagues identified 4 relevant studies (n= 505) exploring morbidity and mortality associated with Waterlow score in surgical patients [11]. The studies included general, vascular, transplant and orthopaedic surgical settings. A high Waterlow score demonstrated a statistically significant association with increased morbidity, length of stay, need for intensive care admission and mortality. Furthermore, this was a more accurate predictor for adverse post surgery outcomes compared to others scoring systems in routine practice (Portsmouth Physiological and Operative Severity Score for the enumeration of Mortality [P-POSSUM] or American Society of Anesthesiologists [ASA]. However, these handful of studies have several limitations including small cohort sizes, heterogenous cut-offs for dichotomous Waterlow scores and lack of adjustment against baseline characteristics suggesting publication bias that should be interpreted with caution.

Rather than relying on non-validated tools, better measures for risk stratification are warranted to identify kidney transplant candidates at high risk for adverse outcomes. One possibility is assessment for frailty, a syndrome of accelerated ageing across multiple physiological systems and characterised by increased vulnerability to stressors, which has been studied in the context of surgical interventions including kidney transplantation. While associated with several adverse outcomes after kidney transplantation [19], several limitations exist in its clinical application. Firstly, frailty can be defined using many tools and it is unclear which is best from a plethora of available options [20]. This is important as frailty scoring tools vary in their requirement of time, resource, and equipment. Secondly, frailty is a dynamically changing status and evolves with time for patients living with kidney failure [21]. Therefore, a quick and simple tool that can be utilised prior to time-pressured kidney transplant surgery will have most utility and clinical application. Further work in this area is important to understand how assessment of frailty can be used for pre-operative risk stratification.

A more fundamental question is what benefit such risk stratification has for kidney transplant candidates. Unless changes to surgical intervention or post-operative care will occur, which currently lacks any evidence-base, apart from risk counselling it serves no additional benefit for the success of that kidney transplant in the peri- and post-operative period. By contrast, it may persuade some transplant professionals to deny the opportunity for kidney transplantation to some kidney failure patients identified as high-risk who will still have survival benefits from proceeding with kidney transplantation versus remaining on dialysis. In the absence of any tailored intervention to attenuate risk, the aim and purpose of any risk stratification before kidney transplant surgery requires careful thought for utility. While we currently lack robust evidence-based intervention to demonstrate ‘prehabilitation’ of frail kidney transplant candidates has clinical benefits, encouraging work is happening in this area. McAdmas-DeMarco et al. conducted a single-arm intervention trial (with a historical control as comparison) in 18 kidney transplant candidates using centre-based prehabilitation involving weekly physical therapy sessions [22]. Physical activity based on accelerometry improved by 64% (p = 0.004) and length of hospital admission was shorter for 5 participants who received transplantation compared to historic demographic-matched controls (5 versus 10 days respectively; RR = 0.69; 95% CI 0.50–0.94; p = 0.02). However, while high satisfaction was reported by the 18 participants, it is important to note 84/106 eligible kidney transplant candidates decline to participate (with 6/24 consented participants not actually participating). Therefore, feasibility on a wide scale will require more engagement from patients to be worthwhile. Our analysis also suggests if such a strategy is employed, rather than using surrogate measures like Waterlow score intended for other purposes, we must adopt validated risk stratifications measures that are amendable to intervention with a stratified management plan.

Limitations of our work must be appreciated for an accurate interpretation of the data. The study was retrospective in nature, and so is prone to all the shortcomings of such analyses, including the inability to establish causation, and the potential effects of unmeasured or intangible confounders. As previously highlighted, we had a sizable proportion of missing data with regards to pre-operative Waterlow scores which is a major limitation. While we undertook a multiple imputation model to overcome this, the limitations of this statistical approach must be acknowledged [14]. In addition, we have highlighted the literature regarding inter-rater variability for pre-operative Waterlow scores. This is important to appreciate in our dataset as it spans 13.5 years and scores are likely to have been conducted by a heterogenous group of nurses.

To conclude, in this large single-centre study, we observed over half of kidney transplant candidates admitted for transplant surgery with available data had elevated Waterlow risk scores. As a nursing tool designed to risk stratify acutely admitted patients for pressure sores, as a surrogate measure for physiological fitness in isolation it is associated with prolonged length of stay after transplantation and mortality (but not emergency 90-day readmissions post-operatively). However, after adjustment against baseline demographics, most of these associations are lost. Our data suggest Waterlow scores have no utility as a risk stratification tool for adverse post-transplant outcomes outside its intended use. Further research is required to validate risk stratification measures to identify high-risk kidney transplant candidates, although the purpose for adopting such pathways should be to modify peri- and post-operative care to improve outcomes rather than deny the opportunity for kidney transplantation to kidney failure candidates.