A risk score based on procalcitonin for predicting acute kidney injury in COVID‐19 patients

Abstract Background Acute kidney injury (AKI) has been reported developing commonly in coronavirus disease 2019 (COVID‐19) patients and could increase the risk of poor outcomes in these patients. We design this study to explore the value of serum procalcitonin (PCT) on predicting AKI and construct risk score for predicting AKI in COVID‐19 patients. Methods Patients diagnosed with COVID‐19 and hospitalized in Renmin Hospital of Wuhan University between January 30 and February 24, 2020, were included. The least absolute shrinkage and selection operator (LASSO) regression was performed to identify the strongest predictors of AKI. Multivariate logistic regression analysis was conducted to find independent risk factors for AKI and construct risk score using odds ratio (OR) value of those risk factors. Receiver operating characteristics (ROC) curves were plotted, and area under the ROC curve (AUC) value was calculated to evaluate the predictive value of single PCT level and the constructed risk score. Results Among 389 included COVID‐19 patients, 28 (7.2%) patients developed AKI. LASSO regression showed hypertension, saturation of arterial oxygen (SaO2), PCT, and blood urea nitrogen (BUN) were the strongest predictors for AKI. After multivariate logistic regression analysis, only SaO2 (<0.001), PCT (p = 0.004), and BUN (p = 0.005) were independently associated with development of AKI in COVID‐19 patients. The AUC of single PCT and constructed risk score was 0. 881 and 0.928, respectively. Conclusion PCT level is correlated with AKI in COVID‐19 patients. The efficient risk score consisted of SaO2, PCT, and BUN is readily accessible for physicians to evaluate the possibility of AKI in COVID‐19 patients.

storm developed in COVID-19 patients, which indicates excessively upregulated production of cytokines, could result in the systemic inflammatory response (SIRS) and subsequent multiorgan dysfunction syndrome (MODS). 1,2 In addition to the prominent respiratory failure, other organ dysfunction including liver injury, cardiac injury, and kidney injury were also commonly developed in COVID-19 patients. [3][4][5][6] Moreover, the development of extra-pulmonary organ injuries has been confirmed associated with higher mortality of COVID-19 patients. [7][8][9][10] Acute kidney injury (AKI), whose incidence ranged from 0.5% to 40% in COVID-19 patients, could increase the risk of poor outcome and length of hospital stay. 6,[11][12][13][14] Consequently, identifying COVID-19 patients with high risk of developing AKI during hospitalization and therefore making suitable treatment strategies are essential to improve prognosis of patients.
The procalcitonin (PCT), a precursor of the calcitonin under physiological condition, could be increasingly produced under pathological inflammatory stimulation and is discovered associated with severity of SIRS. 15,16 Previous studies have found that serum PCT level was useful in predicting the development of AKI during hospitalization for many kinds of diseases including critically ill, traumatic injury, influenza infection, and acute pancreatitis. [17][18][19][20][21] However, the specific association between PCT and AKI in COVID-19 patients has not been explored. We designed this study to evaluate the value of PCT on predicting AKI in COVID-19 patients. Furthermore, we constructed a multivariate risk score to grade the risk level of developing AKI in COVID-19 patients.

| Patients
Patients admitted to Renmin Hospital of Wuhan University for COVID-19 from January 30 to February 24, 2020, were eligible for this study. The diagnosis of COVID-19 was confirmed by positive results of COVID-19 RNA in nasopharyngeal swabs utilizing realtime fluorescence reverse transcription-polymerase chain reaction (RT-PCR). Patients transferred from other hospitals and those who lacked clinical or laboratory records were excluded from this study.

| Data collection
We collected relevant clinical and laboratory variables from electronic medical record (EMR) system of Renmin Hospital of Wuhan University. Demographic variables (age, sex) and vital signs on admission (mean blood pressure, heart rate, body temperature and respiratory rate, saturation of oxygen) were collected in this study. History of underlying diseases including hypertension, chronic obstructive pulmonary disease (COPD), cardiovascular disease, diabetes mellitus, liver disease, and malignancy were also collected. Variables of blood routine and blood biochemistry tests including serum PCT level were obtained by analyzing blood samples collected at admission.

| Statistical analysis
Kolmogorov-Smirnov test was utilized to confirm the normality of included variables. Normally distributed variables were presented as mean ±standard deviation while non-normally distributed variables were presented as median (interquartile range). And categorical variables were shown as counts (percentage). We performed Student's t test and Mann-Whitney U test to compare the difference between two groups of normally distributed variables and two groups of non-normally distributed variables, respectively. Chi-square test or Fisher's exact test were conducted to testify the difference between two groups of categorical variables. The least absolute shrinkage and selection operator (LASSO) regression, which could minimize the collinearity of selected risk factors and avoid overfitting of these factors, was performed to identify predictors with nonzero coefficients. Identifying the strongest predictors from plenty of potential risk factors for targeted outcome with relatively small quantity is another advantage of LASSO regression. Predictors with nonzero coefficients were then included in multivariate logistic regression analysis to explore independent risk factors and construct risk score for predicting AKI in COVID-19 patients. Odds ratio (OR) and 95% confidence intervals (CI) of each risk factors were calculated and presented. The risk score was constructed based on the OR value of each independent risk factor. Receiver operating characteristic (ROC) curves of single predictors and constructed risk score were drawn, and area under the ROC curve (AUC) value of them was calculated. Z test was conducted to verify the difference of AUC between single serum PCT level and the constructed risk score.
Two-sided P value <0.05 was considered statistically significant.

| Baseline characteristics of included COVID-19 patients
A total of 389 patients confirmed with COVID-19 were included in this study. Among them, 28 patients developed AKI during their hospitalization with an incidence of 7.8% (Table 1). We divided patients into two groups according to the occurrence of AKI and then TA B L E 1 Baseline characteristics of included COVID-19 patients grouped by the development of AKI

| The risk score for predicting AKI in COVID-19 patients constructed by multivariate logistic regression
SaO 2 , PCT, and BUN were transformed into binary variables based on their corresponding cutoff value, Then, SaO 2 , PCT, and BUN remained independent risk factors for AKI in multivariate logistic regression analysis and OR of these three variables are shown in Table 3. Corresponding scores of these three factors were assigned according to their β coefficients. Presented as Table 4, the risk score was readily used in clinical practice, which ranged from 0 to 3. The

| DISCUSS ION
AKI is a kind of common organ dysfunction which has been observed occurring in 8.4% to 19.6% hospitalized patients. 22  hypertension and cardiovascular disease than non-AKI group. The fact that mortality of AKI group was higher than that of non-AKI group also indicated AKI development was unfavorable to outcome of COVID-19 patients and suitable treatment decisions should be taken to avoid this detrimental complication.
The mechanism involved in the development of AKI in COVID-19 patients may be multifactorial. Bioinformatic analyses indicated that angiotensin-converting enzyme 2 (ACE2), which is acknowledged as a major receptor of SARS-CoV-2, is highly expressed in renal tissue. 25  Previous studies have confirmed that proinflammatory cytokines, especially the IL-6, play a key role in the pathophysiological process of AKI. [32][33][34] Besides, other common risk factors during viral infection including hypotension, hypoxemia, thrombosis, rhabdomyolysis, and unsuitable use of nephrotoxic drugs may also collectively aggravate the abnormal renal function. 28,35,36 In this study, multivariate logistic regression analysis showed that SaO 2 , PCT, and BUN were three independent risk factors for AKI in COVID-19 patients. The hypoxemia, which sometimes be defined as SaO 2 <90%, has been verified associated with deteriorating renal function. 37,38 A recent study showed COVID-19 patients with AKI whose renal function improved had higher oxygenation index than those did not improved. Therefore, it is logical and effective to include SaO 2 into our constructed risk score. Another constituent in our risk score is PCT, which has been confirmed as a reliable maker for predicting AKI in many kinds of patients including those diagnosed with critically ill, traumatic injury, influenza infection or acute pancreatitis. 17

F I G U R E 4
Patients's percentage of AKI grouped by constructed risk score. The incidence of AKI in low-risk group, medium-risk group and high-risk group was 0.71% (2/282), 9.46% (7/74) and 57.58% (19/33), respectively records of PCT level on admission were excluded from this study so that the selection bias could not be avoided. Secondly, urine volume, an important indicator for diagnosis of AKI, was not recorded in our study. This is partly due to the lack of records of urine volume in mildly and moderately ill patients. Thirdly, the detailed extent of lung involvement was not recorded by us which may affect the severity of kidney injury. Finally, we did not divide overall enrolled patients into training set and validation set due to the limited number of AKI event. Further study with larger sample size could be conducted in other medical centers to verify the value of our risk score.

| CON CLUS IONS
Single PCT value is a valuable predictive marker of AKI in COVID-19 patients, the risk score we constructed using serum PCT level could readily and efficiently help clinicians evaluate the possibility of developing AKI in COVID-19 patients.

CO N FLI C T O F I NTE R E S T
All authors declare no conflict of interest.

AUTH O R ' S CO NTR I B UTI O N S
Ruoran Wang was involved in the data analysis, and drafting of the manuscript. Min He was involved in the data collection and interpretation of the results. Yan Kang interpreted the results and contributed to the reviewing of the draft.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used and analyzed during this study are available from the corresponding author on reasonable request.