Dear Editor,
In a recent issue of the Journal of Wound, Ostomy, and Continence Nursing, Sullivan and colleagues1 aimed to modify the Norton Scale for Pressure Sore Risk to improve its predictive power in critically ill patients. We commend the authors in this area of inquiry and wished to share our findings from a study with similar aims. In our pervious study,2 we reported intraclass correlation (ICC) values for Norton Scale items as 0.595 (95% CI, 0.426-0.764) for "physical condition" and 0.681 (95% CI, 0.526-0.822) for "incontinence." Based on these findings, we sought quantitative indicators to replace some ambiguous items defined in the score to improve the instrument's reliability. In this study,1 the authors identified quantitative indicators to modify the Norton Scale, such as body mass index less than 19 kg/m2 or more than 40 kg/m2, albumin level less than 2.4 g/dL, blood glucose level more than 180 mg/dL, Svo2 or Scvo2 less than 60% for 5 minutes, Spo2 less than 90%, and hemoglobin level less than 7.7 g/dL. They reported a Cronbach [alpha] of 0.944, reflecting excellent internal reliability, and an ICC of 0.933 (95% CI, 0.911-0.950), indicating excellent interrater reliability. They also found the Pearson correlation coefficients between the 5 new optimized Norton subscales and the total score were both over 0.85, indicating good predictive validity. We applauded the authors for their achievement in improving the reliability and predictive validity of their new optimized Norton Scale. However, we believe additional predictive validity analysis is needed.
Predictive validity for a risk assessment scale commonly includes evaluation of discrimination power and calibration power. Discrimination power is the ability of the scale to correctly separate the subjects into different groups. In this study,1 the discrimination power is the ability to correctly predict patients with or without pressure injury using the optimized Norton Scale in the critical care setting. Common assessments used in discrimination for predictive classification include sensitivity, specificity, positive predictive value, negative predictive value, likelihood ratios for positive and negative tests, receiver operating characteristics (ROC) curve, and area under the ROC (AUC). Previous studies3 showed that a Norton Scale score of 16 provided the best balance between sensitivity (0.76) and specificity (0.75), and a score of 17 could also be considered as the cutoff point for sensitivity (0.90) and specificity (0.67). A systematic review and meta-analysis4 showed a cumulative sensitivity of the Norton Scale of 0.75 (95% CI, 0.70-0.79), cumulative specificity of 0.57 (95% CI, 0.55-0.59), and ROC AUC of 0.82 (SE = 0.05) while Q* value was 0.75 (SE = 0.05). The authors used some quantitative indicators to replace ambiguous items for the new optimized Norton Scale, resulting in improved reliability. Nevertheless, we do not yet know the discrimination power of the optimized Norton Scale. Specifically, we wish to know the sensitivity and specificity at different cutoff points and the ROC AUC in predicting pressure injury. The AUC should be in the region of 0.97 or above; an AUC of 0.93 to 0.96 is very good; 0.75 to 0.92 is good; but an AUC less than 0.75 indicates deficiencies in its predictive accuracy.5
Calibration power is another aspect of predictive validity; it is defined as the degree of correspondence between an estimated probability produced by the scale and the actual observed probability. Common assessments used in calibration for confidence of fit include misclassification rate, Pearson's [chi]2, interrater agreement with kappa value, or Hosmer-Lemeshow statistic. In our pervious study,6 we evaluated the calibration power of the Braden Scale for Pressure Sore Risk for pressure injury development and found that the predicted pressure injury incidence did not fit well with the observed incidence ([chi]2 = 42.154, P = .000 by Braden Scale scores; and [chi]2 = 17.223, P = .001 by Braden Scale risk classification). We also investigated the calibration power of our 2 newly constructed pressure injury risk scales.7,8 Nevertheless, we did not find any study investigated the calibration power of Norton Scale in predicting pressure injury risk. We also want to know the calibration power of this new optimized Norton Scale for correspondence between the estimated pressure injury probability and the actual observed pressure injury probability.
In summary, though the new optimized Norton Scale has clearly improved the scale's predictive reliability, we assert more predictive validity analysis is needed. Specifically, we wish to know the discrimination and calibration power of this scale in predicting pressure injury development in the critical care setting.
Hai-Yan Shi, MSN
The People's Hospital of Rugao, and Nantong University Affiliated Rugao Hospital
Nantong City, Jiangsu Province, PR China
Hong-Lin Chen, MD
Nantong University Nantong City, Jiangsu Province, PR China
REFERENCES