KR102388887B1 - Predicting system and method of postoperative acute kidney injury risk after non-cardiac surgery - Google Patents

Abstract

The present invention relates to a system and method for predicting the risk of acute renal failure after non-cardiac surgery, and more specifically, the system includes: a variable selection unit; classification reference point setting unit; and an acute renal failure risk prediction unit after non-cardiac surgery, wherein the method includes: a first step of selecting a variable; a second step of setting a classification reference point; and a third step of predicting the risk of developing acute renal failure after non-cardiac surgery. Since the present invention is simple and accurate, it is an invention with very high clinical applicability.

Description

PREDICTING SYSTEM AND METHOD OF POSTOPERATIVE ACUTE KIDNEY INJURY RISK AFTER NON-CARDIAC SURGERY

The present invention relates to a system and method for predicting the risk of developing acute renal failure after non-cardiac surgery.

Acute kidney injury after surgery is an important event associated with postoperative death/dialysis/extension of hospital stay, and it occurs in 1-30% of patients who underwent surgery. In previous reports, it is known that acute kidney injury increases the risk of short-term/long-term death or dialysis due to renal failure, and also significantly increases the hospital stay and medical costs during hospitalization. Although there have been many reports of kidney injury after cardiac surgery in existing studies, studies on kidney injury after non-cardiac surgery have been relatively rare. In particular, a predictive model that evaluates patients at high risk of postoperative acute renal injury and predicts the risk of occurrence has been rarely reported. did not have the report.

Korean Patent No. 1940014

An object of the present invention is to provide a system and method for predicting the risk of acute renal failure after non-cardiac surgery using clinical data before non-cardiac surgery of patients subject to non-cardiac surgery.

1. A variable selection unit that selects, as a variable, a factor related to the occurrence of acute renal failure after non-cardiac surgery, which consists of clinical data before non-cardiac surgery of patients subject to non-cardiac surgery, and sets an index set for each variable;

a classification reference point setting unit calculating sensitivity and specificity with respect to the sum of combinations of indices for each index set, and setting cut-off values according to the sensitivity and specificity; and

For the non-cardiac surgery target patient requiring prediction, the sum of the indices determined according to the set of indices set for each variable selected by the variable selection unit is calculated, and the non-cardiac surgery requiring the prediction based on the classification reference points set in the classification reference point setting unit Includes; a risk prediction unit for acute renal failure after non-cardiac surgery that predicts the risk of acute renal failure after non-cardiac surgery of the target patient;

The factors related to the occurrence of acute renal failure after noncardiac surgery were age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, sex, and expected duration of surgery (Expected). surgical duration, emergency operation, diabetes mellitus, renin-aldosterone-angiotensin-system blockade use (RAAS blockade use), hypoalbuminemia , Anemia and hyponatremia risk prediction system after non-cardiac surgery, including at least one selected from the group consisting of.

2. In the above 1, the index set is set according to the selection categories for each variable except for the expected surgical duration, and 0, 3, 4, 6 to 9, 13, 15 and 22 for each selection category. Any one index selected from among is set, and the total sum of the indices for each selection category of the variable is 0 to 81, the system.

3. The system of 1 above, wherein the index set for the expected surgical duration among the variables is set to be an index five times the expected surgical duration (time).

4. In the above 1, age among the variables is divided into selection categories of less than 40 years old, 40 years old and less than 60 years old, 60 years old and less than 80 years old, and 80 years old or more,

The estimated glomerular filtration rate (eGFR) is 60 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 60 mL/min/1.73 m ² or less, 30 mL/min/1.73 m ² or more and less than 45 mL/min/1.73 m ² , 15 mL/min/1.73 m ² or more and less than 30 mL/min/1.73 m ² The system is divided into selection categories.

5. In the above 1, the factors related to the occurrence of acute renal failure after non-cardiac surgery are age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, and sex. , Expected surgical duration, Emergency operation, diabetes (diabetes mellitus), renin-aldosterone-angiotensin-system blockade use (RAAS blockade use) , including Hypoalbuminemia, Anemia and Hyponatremia.

6. In the above 1, the prediction of the risk of acute renal failure after non-cardiac surgery is A, The system, which is classified into a total of four grades, B, C, and D.

7. In the above 6, the A grade is when the sum of the indices is less than 20, the B grade is when the sum of the indices is 20 or more and less than 40, and the C grade is when the sum of the indices is 40 or more and less than 60, The system, wherein the D grade is classified when the sum of the indices is 60 or more.

8. In the above 7, the A grade, the probability of occurrence of acute renal failure and severe acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is less than 2%,

In the B grade, the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 2% or more, and the probability of occurrence of severe acute renal failure is less than 2%,

The grade C indicates that the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 10% or more, and the probability of occurrence of severe acute renal failure is 2% or more,

In the D grade, the probability of occurrence of acute renal failure after non-cardiac surgery of the non-cardiac surgery target patient requiring the prediction is 20% or more, and the probability of occurrence of severe acute renal failure is 10% or more, the system.

9. A first step of selecting as a variable a factor related to the occurrence of acute renal failure after non-cardiac surgery, which consists of clinical data before non-cardiac surgery of patients subject to non-cardiac surgery, and setting an index set for each variable;

a second step of calculating sensitivity and specificity for the sum of combinations of indices for each index set, and setting cut-off values according to the sensitivity and specificity; and

For the non-cardiac surgery target patient requiring prediction, the sum of the indices determined according to the set of indices set for each variable selected in the first step is calculated, and the non-cardiac surgery target requiring the prediction based on the classification reference points set in the second step A third step of predicting the risk of acute renal failure after non-cardiac surgery of the patient;

The factors related to the occurrence of acute renal failure after noncardiac surgery were age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, sex, and expected duration of surgery (Expected). surgical duration, emergency operation, diabetes mellitus, renin-aldosterone-angiotensin-system blockade use (RAAS blockade use), hypoalbuminemia , Anemia and hyponatremia (Hyponatremia), including at least one selected from the group consisting of, a method of predicting the risk of acute renal failure after non-cardiac surgery.

10. The set of indices according to 9 above, wherein the set of indices is set according to the selection categories for each variable except for the expected surgical duration, and 0, 3, 4, 6 to 9, 13, 15 and 22 for each selection category. Any one of the indices selected from among is set, and the total sum of the indices for each selection category of the variable is 0 to 81, the method.

11. The method of 9 above, wherein the index set for the expected surgical duration among the variables is set to be five times the index of the expected surgical duration (time).

12. In the above 9, age among the variables is divided into selection categories of less than 40 years old, more than 40 years old and less than 60 years old, more than 60 years old and less than 80 years old, and more than 80 years old,

The estimated glomerular filtration rate (eGFR) is 60 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 60 mL/min/1.73 m ² or less, 30 mL/min/1.73 m ² or more and 45 mL/min/1.73 m ² or more, 15 mL/min/1.73 m ² or more and 30 mL/min/1.73 m 2 or more and less than 30 mL/min/1.73 m ² .

13. The above 9 above, wherein the factors related to the occurrence of acute renal failure after non-cardiac surgery are age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, and sex. , Expected surgical duration, Emergency operation, diabetes (diabetes mellitus), renin-aldosterone-angiotensin-system blockade use (RAAS blockade use) , including Hypoalbuminemia, Anemia and Hyponatremia.

14. The method of 9 above, wherein the prediction of the risk of acute renal failure after non-cardiac surgery is A, B, C, D classified into a total of four grades, the method.

15. In the above 14, the A grade is when the sum of the indices is less than 20, the B grade is when the sum of the indices is 20 or more and less than 40, and the C grade is when the sum of the indices is 40 or more and less than 60, The method of claim 1, wherein the D grade is classified when the sum of the indices is 60 or more.

16. The grade A according to the above 15, the probability of occurrence of acute renal failure and severe acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is less than 2%,

In the B grade, the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 2% or more, and the probability of occurrence of severe acute renal failure is less than 2%,

The grade C indicates that the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 10% or more, and the probability of occurrence of severe acute renal failure is 2% or more,

The D grade is the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 20% or more, and the probability of occurrence of severe acute renal failure is 10% or more, the method.

The present invention can predict a patient's acute renal failure or related prognosis before the actual operation is performed with measurable/predictable indicators before surgery. Furthermore, through the present invention, it is possible to reduce the risk of short-term/long-term death, the risk of dialysis, and the like, and use an accurate, simple and easy scoring system, which is an invention highly applicable in clinical practice.

1 is a diagram summarizing an embodiment of the present invention;
2 is a diagram illustrating a variable selection process performed in a discovery cohort.
3 is a diagram illustrating a correction chart and a receiver operation characteristic curve.
4 is a diagram showing the sensitivity and specificity for the SPARK index.
5 is a diagram showing the SPARK risk classification and incidence rates of low-stage AKI and severe AKI.
6 is a diagram illustrating a preoperative PO-AKI risk assessment strategy based on the proposed SPARK classification.
7 is a diagram illustrating a correction chart and a receiver operation characteristic curve for each surgical department.

Hereinafter, the present invention will be described in detail.

The present invention relates to a system for predicting the risk of acute renal failure after non-cardiac surgery.

The system of the present invention includes a variable selection unit that selects, as a variable, a factor related to the occurrence of acute renal failure after non-cardiac surgery, which consists of clinical data before non-cardiac surgery of patients subject to non-cardiac surgery, and sets an index set for each variable. .

The non-cardiac surgery means any surgery other than cardiac surgery.

The non-cardiac surgery pre-clinical data of the patients subject to non-cardiac surgery is at least one selected from an electronic medical record (EMR) of a hospital, an interview with a patient or guardian, an inpatient nursing information record, a nursing activity assessment stop, and a hospital record record. It may depend on the path, but is not limited thereto.

The factors related to the occurrence of acute renal failure after non-cardiac surgery may be collected from a discovery cohort and a validation cohort, but are not limited thereto.

The variable selection unit can select factors with large model coefficients as variables while considering statistical assumptions and multivariable analysis for a proportional odds regression technique, but is limited thereto it is not

In addition, the variable selection unit uses a proportional odds regression technique with the selected variables to construct a multivariable model targeting an ordinal variable composed of negative prognosis related to acute renal failure. At the same time, the index set for each variable may be set so that the sum of the indices set for each variable reflects the risk by converting the model coefficients into integers, but is not limited thereto.

The factors related to the occurrence of acute renal failure after noncardiac surgery were age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, sex, and expected duration of surgery (Expected). surgical duration, emergency operation, diabetes mellitus, renin-aldosterone-angiotensin-system blockade use (RAAS blockade use), hypoalbuminemia , may include at least one selected from the group consisting of anemia and hyponatremia, and preferably include all, but is not limited thereto.

In an embodiment of the present invention, an index set may be set for each index for each variable except for the expected surgical duration. The exponent set is set to (0,X), and X is any one of exponents selected from 3, 4, 6 to 9, 13, 15 and 22. For example, a patient will have an index from one set of indices according to the definition of that variable. For example, when the index set of the gender variable is (0, 8), the index may be 10 points if the patient is a man, and 0 points if the patient is a woman.

In one embodiment of the present invention, the size of the exponent set may vary by the number of selection categories for each variable. For example, if the predetermined variable has three selection categories, the set of exponents set in the predetermined variable may be (0, Y, Z). In this case, Y and Z are values corresponding to any one of X, and Y does not have the same value as Z.

Specifically, the age (age) may be divided into a selection category of less than 40 years old, more than 40 years old and less than 60 years old, more than 60 years old and less than 80 years old, and more than 80 years old,

The estimated glomerular filtration rate (eGFR) is 60 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 60 mL/min/1.73 m ² or less, 30 mL/min/1.73 m ² or more and 45 mL/min/1.73 m ² or more, 15 mL/min/1.73 m ² or more and less than 30 mL/min/1.73 m ² can be divided into selection categories.

More specifically, an index set of 0 if age is less than 40 years old, 6 if 40 or more and less than 60 years old, 9 if it is 60 or more and less than 80 years old, and 13 if it is 80 years old or more may be set; Estimated glomerular filtration rate (eGFR) is 0 for 60 mL/min/1.73 m ² or more, 8, 30 mL/min for 45 mL/min/1.73 m ² or more and less than 60 mL/min/1.73 m ² an exponent set of 15 if min/1.73 m ² or more and 45 mL/min/1.73 m ² or more and 22 if 15 mL/min/1.73 m ² or more and less than 30 mL/min/1.73 m ² can be set; An index set of 6 if dipstick albuminuria is present and 0 otherwise may be set; An exponent set of 8 if sex is male and 0 if female may be set; If it is an emergency operation, an exponent set of 7, otherwise 0 may be set; An exponent set of 4 if diabetes mellitus is present and 0 otherwise may be set; An exponent set of 6 if there is renin-aldosterone-angiotensin-system blockade use (RAAS blockade use) and 0 otherwise may be set; An exponent set of 8 if there is hypoalbuminemia and 0 if not; An exponent set of 4 if anemia is present and 0 otherwise may be set; An exponent set of 3 if hyponatremia is present and 0 if not present can be set. Therefore, the total sum of the indices for each variable excluding the expected surgery period can be calculated from 0 to 81 at the lowest.

In addition, the unit of expected surgical duration among variables may be time, and an index five times the expected surgical duration may be set. For example, if the expected duration of surgery is 3 hours, the index of that variable is 15.

In addition, the system for predicting the risk of acute renal failure after non-cardiac surgery of the present invention calculates sensitivity and specificity for the sum of the combinations of indices for each index set, and classifies them according to the sensitivity and specificity and a classification reference point setting unit for setting cut-off values.

The classification reference point setting unit may set the classification reference point by calculating the sensitivity and specificity for the sum of the combinations of indices for each index set using a receiver operating characteristic curve (ROC curve).

The index set means an index set set according to the size of the selection category of each variable, and the combination of indices for each index set is each corresponding variable except for the expected surgery period obtained from clinical data of patients subject to non-cardiac surgery. As meaning a combination of indices according to the selection category of

Sensitivity and specificity for each combination sum of each index can be calculated according to the area under the curve (AUC) value of the discovery cohort and validation cohort by performing receiver operation characteristic curve analysis. can

On the other hand, to set the classification reference point, a receiver manipulation characteristic (ROC) curve analysis can be performed to obtain the prediction probability with the estimated regression coefficient value, and in this case, it can be set as the classification reference point by obtaining the sum of the sensitivity and specificity.

More specifically, a cut-off value may be set using a youden index of a receiver operation characteristic curve (ROC curve).

In addition, the system for predicting the risk of acute renal failure after non-cardiac surgery of the present invention calculates the sum of the indices determined according to the set of indices set for each variable selected in the variable selection unit for the non-cardiac surgery target patient requiring prediction, and the classification and a risk prediction unit for acute renal failure after non-cardiac surgery that predicts the risk of acute renal failure after non-cardiac surgery of a patient subject to non-cardiac surgery requiring the prediction based on the classification reference points set in the reference point setting unit.

According to one embodiment of the present invention, the prediction of the risk of occurrence of acute renal failure after non-cardiac surgery is based on the sum of indices determined according to the set of indices set for each variable selected in the variable selection unit for the non-cardiac surgery target patient requiring prediction. It can be classified into a total of four grades: A, B, C, and D.

The A grade is when the sum of the indices is less than 20, the B grade is when the sum of the indices is 20 or more and less than 40, the C grade is when the sum of the indices is 40 or more and less than 60, the D grade is the index of the index It may be classified when the total is 60 or more.

In addition, according to an embodiment of the present invention, in the grade A, the probability of occurrence of acute renal failure and severe acute renal failure after non-cardiac surgery of a patient subject to non-cardiac surgery requiring the prediction may be less than 2%, and the grade B is the prediction The probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring this may be 2% or more, and the probability of occurrence of severe acute renal failure may be less than 2%, and the grade C is The probability of occurrence of acute renal failure after non-cardiac surgery may be 10% or more, and the probability of occurrence of severe acute renal failure may be 2% or more, and the grade D indicates that the probability of occurrence of acute renal failure after non-cardiac surgery is 20% or more, severe The probability of developing acute renal failure may be greater than 10%.

In addition, the present invention relates to a method for predicting the risk of acute renal failure after non-cardiac surgery (see FIG. 1 ).

The method includes a first step of selecting as a variable a factor related to the occurrence of acute renal failure after non-cardiac surgery, which consists of clinical data before non-cardiac surgery of patients subject to non-cardiac surgery, and setting an index set for each variable;

a second step of calculating sensitivity and specificity for the sum of combinations of indices for each index set, and setting cut-off values according to the sensitivity and specificity; and

For the non-cardiac surgery target patient requiring prediction, the sum of the indices determined according to the set of indices set for each variable selected in the first step is calculated, and the non-cardiac surgery target requiring the prediction based on the classification reference points set in the second step A third step of predicting the risk of acute renal failure after non-cardiac surgery of the patient; includes.

The factors related to the occurrence of acute renal failure after noncardiac surgery were age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, sex, and expected duration of surgery (Expected). surgical duration, emergency operation, diabetes mellitus, renin-aldosterone-angiotensin-system blockade use (RAAS blockade use), hypoalbuminemia , may include at least one selected from the group consisting of anemia and hyponatremia, and preferably include all, but is not limited thereto.

According to an embodiment of the present invention, the index set is set according to the selection category for each variable except for the expected surgical duration, and 0, 3, 4, 6 to 9, 13, 15 for each selection category. and 22 may be set, and the total sum of the indices for each selection category of the variable may be 0 to 81.

In addition, according to an embodiment of the present invention, the index set for the expected surgical duration among the variables may be set to five times the index of the expected surgical duration (time).

In addition, among the above variables, age is divided into selection categories of less than 40 years old, 40 years old and less than 60 years old, 60 years old and less than 80 years old, and 80 years old or more, and the estimated glomerular filtration rate (eGFR) is 60 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 60 mL/min/1.73 m ² or more, 30 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 15 mL /min/1.73 m ² or more and less than 30 mL/min/1.73 m ² can be divided into selection categories.

In addition, according to an embodiment of the present invention, the prediction of the risk of acute renal failure after non-cardiac surgery is the sum of the indices determined according to the set of indices set for each variable selected in the variable selection unit for the non-cardiac surgery target patient requiring prediction. Accordingly, it can be classified into a total of four grades: A, B, C, and D.

The A grade is when the sum of the indices is less than 20, the B grade is when the sum of the indices is 20 or more and less than 40, the C grade is when the sum of the indices is 40 or more and less than 60, the D grade is the index of the index It may be classified when the total is 60 or more.

In addition, according to an embodiment of the present invention, in the grade A, the probability of occurrence of acute renal failure and severe acute renal failure after non-cardiac surgery of a patient subject to non-cardiac surgery requiring the prediction may be less than 2%, and the grade B is the prediction The probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring this may be 2% or more, and the probability of occurrence of severe acute renal failure may be less than 2%, and the grade C is The probability of occurrence of acute renal failure after non-cardiac surgery may be 10% or more, and the probability of occurrence of severe acute renal failure may be 2% or more, and the grade D indicates that the probability of occurrence of acute renal failure after non-cardiac surgery is 20% or more, severe The probability of developing acute renal failure may be greater than 10%.

The details of the first, second and third steps of the method are as performed by the variable selection unit, the classification reference point setting unit, and the acute renal failure risk prediction unit after non-cardiac surgery, respectively, of the system of the present invention described above. Therefore, it is omitted below.

Hereinafter, examples will be given to describe the present invention in detail.

Example

1. Experimental method

(1) the hospital in which the experiment of the present invention was carried out and the design of the invention

The present invention relates to a retrospective observational cohort experiment conducted in a government-designated tertiary care hospital in Korea. This discovery cohort includes adult patients aged 18 years and older who underwent surgery at Seoul National University Hospital between 2004 and 2013, and the validation cohort includes adults who underwent surgery at Seoul National University Bundang Hospital between 2006 and 2015. Both hospitals have more than 1000 beds and top clinicians affiliated with Seoul National University School of Medicine. However, the two hospitals are located in different administrative divisions in Korea and do not share a patient pool or key medical staff.

We included the first surgical cases of the trial period in the following five surgical departments: general surgery, orthopedics, obstetrics and gynecology, neurosurgery and urology. Exclusion criteria were: 1) cardiac surgery, 2) surgery in a deceased patient (e.g., transplantation from a deceased donor), 3) patient with nephrectomy or kidney transplant, 4) defined as an operation time of less than 1 hour. Minor surgical procedures, 5) history of renal replacement therapy, preoperative serum creatinine (sCr) levels of 4 mg/dL or higher, and estimated glomerular filtration rate (eGFR) of 15 mL/min/1.73 m ² or less, or patients with preoperative renal dysfunction, defined as a baseline increase in sCr of ≥0.3 mg/dL or ≥1.5 fold from the minimum value 2 weeks prior to surgery, and 6) Postoperative acute kidney injury; PO-AKI) patients without baseline or follow-up sCr values to confirm events.

(2) Data collection and variables for model building

Information that can be collected or planned before surgery is included because preoperative risk classification is the purpose of the present invention. Most of the continuous variables have been categorized by ranges commonly used for practical application of the present invention. Detailed information on the collected variables is as follows.

The following demographic data were collected from the discovery and validation cohort: age, gender and baseline body mass index (BMI) at time of acquisition. Ages were classified as <40, ≥40 and <60, ≥60 and <80 or ≥80 years, and the interval was determined to limit the number of classifications for brevity. BMI was classified as underweight (<18.5 kg/m ² ), in the normal range (≥18.5 and <25 kg/m ² ), and obese (≥25 kg/m ² ). Co-morbidities of heart disease including a history of heart failure or coronary artery disease (eg, angina pectoris or myocardial infarction), hypertension and diabetes were collected. Co-morbidity histories were mainly confirmed by reviewing records of anesthesiologists, taking related medications, and diagnostic codes. Before performing the surgery to schedule anesthesia and reserve the operating room, data on the actual surgery duration (hours) and expected surgery duration (hours) entered by the attending physician were collected. Only the expected surgical duration was included in the model as it was a collectible variable before surgery. The type of anesthesia (general or non-general) and whether the surgery was performed as scheduled or as an emergency surgery was collected. Preoperative systolic and diastolic blood pressures were recorded. Among well-known PO-AKI-related drugs, preoperative use of a renin-aldosterone-angiotensin-system blocker was included in the variable of the present invention. Diuretics, non-steroidal anti-inflammatory drugs, or nephrotoxic antibiotics were not collected because they were frequently used newly after surgery to control dose overload, pain, or infection. The laboratory values collected were the final examination results within 3 months before surgery. Baseline eGFR was calculated based on sCr levels using the CKD-EPI equation and stratified into 4 classifications (baseline range = 60 mL/min/1.73 m ² or greater; CKD 3A = 45 mL/min/1.73 m ² or greater; and less than 60 mL/min/1.73 m ² ; CKD 3B = more than 30 mL/min/1.73 m ² and less than 45 mL/min/1.73 m ² ; or CKD 4 = more than 15 mL/min/1.73 m ² and 30 mL less than /min/1.73 m ² ). The presence of another renal function variable, baseline proteinuria, was confirmed by a simple dipstick test. Abnormalities in the leukocyte count were classified as leukopenia (<4,000/μL) and leukopenia (≧10,000/μL). Anemia was defined as a hemoglobin level of < 12 g/dL in women and < 13 g/dL in men. A serum albumin level of less than 3.5 g/dL is the definition of hypoalbuminemia. hyponatremia (blood sodium < 135 mEq/L), hypernatremia (blood sodium > 145 mEq/L), hypokalemia (blood potassium < 3.5 mEq/L), hyperkalemia (blood potassium > 5.5 mEq/L) A baseline electrolyte imbalance was recorded, including

(3) Experimental results

PO-AKI was defined according to the sCr-criteria of Kidney Disease: Improving Global Outcomes guidelines using peak levels of sCr within 2 weeks after surgery. The term “PO-AKI” included all AKIs regardless of AKI severity. To address the patient-oriented outcomes of severity and PO-AKI, we defined outcomes that represent the turn for predictive model construction involving three outcome classifications: “no AKI”, “low-stage AKI” and “severe AKI”. ". Among PO-AKI patients, severe AKI was defined by two or more AKI stages and consequently AKI leading to death and dialysis within 90 days after AKI. When several patients started or died of kidney transplantation outside the trial hospital, we reviewed the national mortality database from the National Statistical Office and the national dialysis records maintained by the Korean Renal Association and confirmed the results. Other patients who developed stage 1 PO-AKI but did not have severe AKI were included in the "low stage AKI" classification of sequential outcomes.

(4) Variable selection

First, the variable selection process was performed in the discovery cohort (Fig. 2). To identify variables that violated the parallel regression estimation, univariate cumulative logistic regression analysis with binary results defined as different reference values from sequential results was performed. Parallel regression estimates were checked by examining the direction and magnitude of the model coefficients rather than the usual semi-conservative statistical tests on large data sets. Next, we fit the sequential outcomes to a multivariate proportional odds model, leaving only the variables that were shown to have statistically significant relationships with independent, sequential outcomes. Finally, we reduced the number of variables included in the model according to the absolute size of the model coefficients, and excluded variables with a relatively small effect size. After the variable selection, additional procedures were performed with patients with no missing values in the selected variables.

(5) Simple Postoperative AKI risK (SPARK) Index and Classification

Further simplifications were made to construct the Simple Postoperative Aki RisK (SPARK) index. After confirming the calibration of the simple model, we multiplied the coefficients to set the maximum sum of the model coefficients in the discovery cohort to 100, and rounded each coefficient to an integer to generate the SPARK index. Finally, cutoff values defining the four classifications in the discovery cohort were identified to generate a comprehensive classification that could be easily interpreted in practice. A cutoff value for grade A/B was defined to present a threshold for screening PO-AKI with high sensitivity (90%), whereas the threshold for grade B/C was PO-AKI Values with high specificity (90%) for Finally, in patients with a SPARK index higher than the cutoff value for grade B/C, we additionally determined a cutoff value for grade C/D, high specificity for severe AKI. A threshold with (90%) was chosen to define the D grade. Considering practical problems, we round the threshold to the nearest ten.

(6) sensitivity analysis and other statistical analysis

Other statistical analysis methods, including sensitivity analyzes, were as follows.

Classified variables are expressed as frequencies (percentages), and continuous variables are expressed as medians (interquartile ranges). A chi-squared test and a Mann-Whitney U test were implemented to compare the basic characteristics of the discovery and validation cohorts. Model calibration was first visually examined with calibration plots for probabilities for both estimated low-grade AKI and severe AKI. Because the present invention included a large number of patients (>25000), it is not recommended to apply the Hosmer-Lemeshow test as a default. Therefore, we calculated the P values of the Hosmer-Lemeshow test from thousands of random subsamples with a fixed subsample size (n=1000). The discriminative power of the model was confirmed with c-stat. A final test of the predictive power of the SPARK index was performed for PO-AKI and severe AKI results along with receiver operating characteristics curve (ROC curve) analysis, and an area under curve (AUC) of 0.7 or greater was acceptable. considered possible. To examine whether a significant bias existed because of our exclusion criteria, a sensitivity analysis was performed. In the analyzes, the discriminant power of the final SPARK index for sequential outcomes was determined by: 1) the substituted data set using non-linear additional strains and substitutions, 2) the input of actual surgical duration instead of the expected, and 3) after merging of the present trial and validation cohorts. , data sets separated according to three eras (2004 to 2007, 2008 to 2011, 2012 to 2015) and 4) tightly controlled for possible inclusions of patients with preoperative undefined AKI The risk was calculated with the data set after additional exclusion of patients with sCr ≥0.3 mg/dL or 1.5-fold higher than the lowest level within 3 months postoperatively after surgery, regardless of the interval between surgeries. In addition, the performance of the SPARK index and classification was tested in each surgical division after combining the discovery and validation cohorts. All analyzes were performed with complete cases with no missing values except for the sensitivity analysis with the imputed dataset. Statistical analyzes were performed with R (version 3.4.3, the R foundation), and a bilateral P value of less than 0.05 was considered statistically significant.

2. Experimental results

(1) Characteristics of the tested cohort

A total of 162,095 patients were included in the screening in this trial, which was the sum of 93,370 and 68,725 surgical cases tested at SNUH (Seoul National University Hospital) and SNUBH (Seoul National University Bundang Hospital), respectively (Fig. 2). After exclusion criteria were applied, 51,041 and 39,764 patients were screened for model building in the discovery and validation cohorts, respectively. The number of patients with low-grade AKI and severe AKI was 2,132 (4.2%) and 605 (1.2%) in the discovery cohort. Of the patients in the discovery cohort with severe AKI, 511 (1.0%), 167 (0.3%), and 88 (0.2%) PO-AKI patients, respectively, had at least stage 2 AKI, death after AKI, and dialysis within 90 days. The incidence of adverse outcomes increased progressively in the discovery cohort, with 1,774 (4.5%) and 727 (1.8%) patients having low-grade AKI and severe AKI. Stage 2 or higher AKI, death after AKI, and dialysis within 90 days were 644 (1.6%), 176 (0.4%), and 64 (0.2%), respectively, in the discovery cohort. Other characteristics of the two cohorts were significantly different as the validation cohort consisted of older, male patients with a greater proportion (Table 1). Obstetrics and gynecology surgery was relatively common in the discovery cohort, but orthopedic surgery was the most common surgery in the discovery cohort. Significant differences were also identified with respect to baseline experimental values and drug use.

Variables Discovery cohort (N=51,041) Validation cohort (N=39,764) P Demographics Age(years) 56 [44; 66] 60 [48; 70] <0.001 <40 9,206 (18.0%) 5,559 (14.0%) ≥40 and (and) <60 20,877 (40.9%) 13,492 (33.9%) ≥60 and <80 19,684 (38.6%) 18,698 (47.0%) ≥80 1,274 (2.5%) 2015 (5.1%) Gender (Sex) <0.001 female 28,306 (55.5%) 18,706 (47.0%) male 22,735 (44.5%) 21,058 (53.0%) Body mass index (kg/m ² ) 23.8 [21.7; 26.0] 24.1 [21.9; 26.4] <0.001 <18.5 1,919 (3.9%) 854 (3.8%) ≥18.5 and <30 45,098 (91.5%) 20,624 (90.9%) ≥30 2,294 (4.7%) 1,218 (5.4%) Preexisting comorbidities Heart disease 1,629 (3.2%) 1,163 (2.9%) 0.022 Hypertension 9,824 (19.2%) 9,161 (23.0%) <0.001 Diabetes mellitus 3,956 (7.8%) 3,581 (9.0%) <0.001 Surgery characteristics Departments <0.001 General surgery 22,447 (44.0%) 14,733 (37.1%) Neurosurgery 5,063 (9.9%) 3,842 (9.7%) Obstetrics and gynecology 7,894 (15.5%) 908 (2.3%) Orthopedics 11,372 (22.3%) 16,823 (42.3%) Urology surgery 4,265 (8.4%) 3,458 (8.7%) Surgery duration (hours) 2.2 [ 1.5; 3.3] 2.5 [1.7; 3.6] <0.001 Expected surgery duration (hours) 2.5 [ 2.0; 3.0] 3.0 [ 2.0; 4.0] < 0.001 Anesthesia type <0.001 General 43,921 (86.6%) 30,570 (76.9%) Non-general 6,789 (13.4%) 9,194 (23.1%) Emergency operation 732 (1.4%) 1,965 (4.9%) <0.001 BP before operation (mmHg) systolic blood pressure (SBP) 124 [113; 135] 128 [116; 142] <0.001 diastolic blood pressure (DBP) 77 [70; 85] 73 [65; 81] <0.001 Normotensive 36,976 (75.0%) 22,934 (57.7%) Hypertensive (SBP ≥140 or DBP ≥90) 10,206 (20.7%) 11,472 (28.9%) Hypotensive (SBP <90 or DBP <60) 2,134 (4.3%) 5,358 (13.5%) Medication usage RAAS blockade 2,881 (5.6%) 2,915 (7.3%) <0.001 Laboratory findings eGFR (mL/min/1.73 m ² ) 82.1 [71.4; 95.1] 87.7 [73.1; 99.7] <0.001 No CKD or CKD stage 1 or 2 (≥60) 46,971 (92.0%) 35,881 (90.2%) CKD stage 3A (≥45 and <60) 3,226 (6.3%) 2,918 (7.3%) CKD stage 3B (≥30 and <45) 641 (1.3%) 703 (1.8%) CKD stage 4 (≥15 and <30) 203 (0.4%) 262 (0.7%) Dipstick albuminuria (≥ 1+) 4,682 (9.3%) 2,169 (7.0%) <0.001 White blood cell count (/mm ² ) 6,100 [ 5,000; 7,500] 6,500 [ 5,400; 8,000] <0.001 Reference range (4000~10000) 43,262 (84.8%) 33,902 (85.3%) Leukopenia (<4000) 3,934 (7.7%) 1,806 (4.5%) Leukocytosis (≥10000) 3,823 (7.5%) 4,033 (10.1%) Hemoglobin (g/dL) 13.2 [12.1; 14.4] 13.6 [12.4; 14.8] <0.001 Anemia (<12 for female, <13 for male) 14,177 (27.8%) 9,212 (23.2%) <0.001 Platelet (x10 ³ /μL) 200 [152; 256] 239 [199; 284] < 0.001 Thrombocytopenia (<10) 4,160 (8.2%) 493 (1.2%) < 0.001 Albumin (g/dL) 4.2 [3.9; 4.5] 4.3 [4.0; 4.5] <0.001 Hypoalbuminemia (<3.5) 5,148 (10.1%) 2,679 (6.8%) <0.001 Sodium (mEq/L) 140 [139; 142] 141 [139; 142] <0.001 Normal sodium blood concentration (Normonatremia) (135-145) 48,288 (94.8%) 31,156 (92.8%) Hyponatremia (<135) 1,291 (2.5%) 986 (2.9%) Hypernatremia (>145) 1,361 (2.7%) 1,418 (4.2%) Potassium (mEq/L) 4.2 [4.0; 4.4] 4.2 [4.0; 4.5] <0.001 Normal potassium blood level (Normokaleima) (3.5-5.5) 49,711 (97.6%) 32,649 (97.3%) Hypokalemia (<3.5) 1,016 (2.0%) 711 (2.1%) Hyperkalemia (>5.5) 213 (0.4%) 200 (0.6%)

CKD = chronic kidney disease

(2) Variable selection

Table 2 below shows the patient characteristics according to the sequential results tested in the discovery cohort. In cumulative logistic regression analysis, surgical department, body mass index and blood pressure (BP) classification, type of anesthesia, and hypernatremia were excluded from further model building because they were inadequately met in parallel regression estimates (Table 3). Additionally, serum potassium level categories and leukocytosis did not show a significant relationship with sequential outcomes in our multivariate proportional Oz model (Table 4). Finally, heart disease, hypertension and leukopenia were excluded from the model because the model coefficients were relatively small compared to the others (Table 5). Other variables were included for final index construction and validation.

Finally, a total of 49,803 and 29,715 cases in the discovery and validation cohort with complete information of the finally selected variables, respectively, were used for further analysis for simplified model construction and validation (Fig. 2). The number of low-stage AKI patients included was 2,062 (4.1%) and 1,109 (3.7%), respectively, in the discovery and validation cohorts, and there were no missing values. Serious AKI events appeared in complete cases as follows: 563 (1.1%) in the discovery cohort and 445 (1.5%) in the validation cohort. Finally, the model coefficients of the variables selected in the proportional odds model are shown in Table 6 below.

Variables No AKI (N=48,304) Low stage AKI (N=2,132) Critical AKI (N=605) P Demographics Age 55 [44; 66] 63 [53; 71] 65 [55; 73] <0.001 <40 9,016 (18.7%) 152 (7.1%) 38 (6.3%) ≥40 and <60 19,990 (41.4%) 698 (32.7%) 189 (31.2%) ≥60 and <80 18,194 (37.7%) 1,158 (54.3%) 332 (54.9%) ≥80 1,104 (2.3%) 124 (5.8%) 46 (7.6%) Sex <0.001 female 27,402 (56.7%) 706 (33.1%) 198 (32.7%) male 20,902 (43.3%) 1,426 (66.9%) 407 (67.3%) Body mass index (kg/m ² ) 23.8 [21.6; 26.0] 24.0 [21.9; 26.3] 23.7 [21.3; 26.1] 0.005 <18.5 (underweight) 1,802 (3.9%) 77 (3.8%) 40 (7.1%) ≥18.5 and <30 (normal range) 42,743 (91.5%) 1,854 (90.4%) 501 (88.5%) >30 (obesity) 2,150 (4.6%) 119 (5.8%) 25 (4.4%) Preexisting comorbidities heart disease 1,445 (3.0%) 132 (6.2%) 52 (8.6%) <0.001 Hypertension 8,989 (18.6%) 653 (30.6%) 182 (30.1%) <0.001 Diabetes mellitus 3,494 (7.2%) 357 (16.7%) 105 (17.4%) <0.001 Surgery characteristics Departments <0.001 General surgery 20,962 (43.4%) 1,167 (54.7%) 318 (52.6%) Neurosurgery 4,920 (10.2%) 106 (5.0%) 37 (6.1%) Obstetrics and gynecology 7,802 (16.2%) 61 (2.9%) 31 (5.1%) Orthopedics 10,834 (22.4%) 476 (22.3%) 62 (10.2%) Urologic surgery 3,786 (7.8%) 322 (15.1%) 157 (26.0%) Surgery duration (hours) 2.2 [1.4; 3.2] 3.8 [2.2; 5.8] 3.9 [2.3; 6.2] <0.001 Expected surgery duration (hours) 2.5 [2.0; 3.0] 3.0 [2.5; 5.0] 3.0 [2.5; 5.0] <0.001 Anesthesia type <0.001 General 41,480 (86.4%) 1,878 (88.6%) 563 (93.5%) Non-general 6,508 (13.6%) 242 (11.4%) 39 (6.5%) emergency operation 613 (1.3%) 74 (3.5%) 45 (7.8%) <0.001 BP before operation SBP 123 [113; 135] 126 [114; 138] 125 [113; 137] <0.001 DBP 77 [70; 85] 76 [69; 84] 75 [68; 83] <0.001 Normotensive 35,199 (75.2%) 1,389 (69.4%) 388 (72.4%) Hypertensive (SBP ≥ 140 or DBP ≥ 90) 9,601 (20.5%) 493 (24.6%) 112 (20.9%) Hypotensive (SBP < 90 or DBP < 60) 1,978 (4.2%) 120 (6.0%) 36 (6.7%) Preoperative RAAS blockade 2,482 (5.1%) 294 (13.8%) 105 (17.4%) <0.001 laboratory findings eGFR (mL/min/1.73 m ² ) 82.3 [71.8; 95.1] 76.9 [61.9; 93.6] 74.0 [57.6; 92.9] <0.001 No CKD or CKD stage 1 or 2 (≥60) 44,900 (93.0%) 1,633 (76.6%) 438 (72.4%) CKD stage 3A (≥45 and <60) 2,836 (5.9%) 311 (14.6%) 79 (13.1%) CKD stage 3B (≥30 and <45) 459 (1.0%) 132 (6.2%) 50 (8.3%) CKD stage 4 (≥15 and <30) 109 (0.2%) 56 (2.6%) 38 (6.3%) Presence of albuminuria (dipstick ≥1+) 4,060 (8.6%) 427 (20.5%) 195 (33.2%) <0.001 White blood cell count (1000/mm ² ) 6.1 [5.0; 7.5] 6.3 [4.9; 7.9] 6.6 [4.9; 8.9] <0.001 Reference range (4000~10000) 41,211 (85.3%) 1,627 (76.3%) 424 (70.3%) Leukopenia (<4000) 3,581 (7.4%) 278 (13.0%) 75 (12.4%) Leukocytosis (≥10000) 3,493 (7.2%) 226 (10.6%) 104 (17.2%) Hemoglobin (g/dL) 13.3 [12.1; 14.4] 12.6 [10.9; 14.0] 12.1 [10.4; 13.8] <0.001 Anemia (<12 for female, <13 for male) 12,819 (26.6%) 1,010 (47.4%) 348 (57.6%) <0.001 Platelet (10 ³ /μL) 200 [152; 255] 206 [140; 294] 207 [135; 324] <0.001 Thrombocytopenia (<10) 3,807 (7.9%) 263 (12.6%) 90 (15.2%) <0.001 Albumin (g/dL) 4.2 [3.9; 4.5] 3.9 [3.3; 4.3] 3.7 [3.0; 4.2] <0.001 Hypoalbuminemia (<3.5) 4,322 (9.0%) 591 (27.7%) 235 (38.8%) <0.001 Sodium (mEq/L) 140 [139; 142] 140 [138; 142] 140 [137; 142] <0.001 Normonatremia (135-145) 45,900 (95.2%) 1,881 (88.2%) 507 (83.8%) Hyponatremia (<135) 1,028 (2.1%) 190 (8.9%) 73 (12.1%) Hypernatremia (>145) 1,275 (2.6%) 61 (2.9%) 25 (4.1%) Potassium (mEq/L) 4.2 [4.0; 4.4] 4.2 [4.0; 4.5] 4.2 [3.9; 4.5] <0.001 Normokaleima (3.5-5.5) 47,108 (97.7%) 2,041 (95.7%) 562 (92.9%) Hypokalemia (<3.5) 920 (1.9%) 64 (3.0%) 32 (5.3%) Hyperkalemia (>5.5) 175 (0.4%) 27 (1.3%) 11 (1.8%)

(Low stage AKI + Critical AKI) vs. (No AKI) (Critical AKI) vs. (Low stage AKI + No AKI) Coefficient 95% CI P Coefficient 95% CI P Age <40 Reference Reference ≥40 and <60 0.745 0.588, 0.906 <0.001 0.790 0.453, 1.154 <0.001 ≥60 and <80 1.357 1.207, 1.513 <0.001 1.420 1.098, 1.773 <0.001 ≥80 1.989 1.772, 2.205 <0.001 2.201 1.769, 2.640 <0.001 Male sex (vs. female) 0.978 0.896, 1.060 <0.001 0.951 0.782, 1.123 <0.001 Departments General surgery Reference Reference Neurosurgery -0.891 -1.069, -0.720 <0.001 -0.669 -1.027, -0.341 <0.001 Obstetrics and gynecology -1.793 -2.012, -1.587 <0.001 -1.293 -1.683, -0.941 <0.001 Orthopedics -0.355 -0.457, -0.255 <0.001 -0.964 -1.246, -0.698 <0.001 Urologic surgery 0.580 0.470, 0.688 <0.001 0.978 0.782, 1.170 <0.001 BMI category Underweight (vs. reference range) 0.164 -0.032, 0.351 0.093 0.639 0.299, 0.951 <0.001 Obesity (vs. reference range) 0.195 0.017, 0.365 0.028 -0.019 -0.449, 0.362 0.925 Heart disease (vs. none) 0.849 0.688, 1.005 <0.001 1.069 0.769, 1.348 <0.001 Hypertension (vs. none) 0.652 0.567, 0.737 <0.001 0.599 0.422, 0.772 <0.001 Diabetes mellitus (vs. none) 0.957 0.850, 1.062 <0.001 0.932 0.714, 1.141 <0.001 Expected surgery duration (hours) 0.539 0.515, 0.563 <0.001 0.458 0.417, 0.500 <0.001 Non-general anesthesia (vs. general) -0.310 -0.438, -0.185 <0.001 -0.810 -1.151, -0.499 <0.001 Emergency operation (vs. non-emergency) 1.277 1.073, 1.474 <0.001 1.803 1.477, 2.105 <0.001 BP category Hypotensive before surgery (vs. normotensive) 0.222 0.126, 0.316 <0.001 0.045 -0.170, 0.253 0.675 Hypertensive before surgery (vs. normotensive) 0.446 0.273, 0.613 <0.001 0.481 0.120, 0.811 0.006 Preoperative RAAS blockade use (vs. no use) 1.148 1.033, 1.260 <0.001 1.282 1.064, 1.492 <0.001 eGFR ≥60 Reference Reference ≥45 and <60 1.092 0.977, 1.206 <0.001 0.981 0.732, 1.217 <0.001 ≥30 and <45 2.151 1.972, 2.327 <0.001 2.196 1.881, 2.490 <0.001 ≥15 and <30 2.928 2.648, 3.207 <0.001 3.197 2.819, 3.551 <0.001 Leukopenia (vs. reference range) 0.683 0.564, 0.800 <0.001 0.675 0.420, 0.916 <0.001 Leukocytosis (vs. reference range) 0.641 0.518, 0.761 <0.001 1.039 0.817, 1.251 <0.001 Anemia (vs. none) 1.004 0.926, 1.081 <0.001 1.280 1.118, 1.443 <0.001 Hypoalbuminemia (vs. none) 1.481 1.393, 1.568 <0.001 1.772 1.605, 1.937 <0.001 Hyponatremia (vs. reference range) 1.593 1.450, 1.733 <0.001 1.731 1.472, 1.976 <0.001 Hypernatremia (vs. reference range) 0.260 0.030, 0.475 0.022 0.567 0.136, 0.950 0.006 Hypokalemia (vs. reference range) 0.636 0.416, 0.844 <0.001 1.045 0.664, 1.390 <0.001 Hyperkalemia (vs. reference range) 1.369 1.002, 1.710 <0.001 1.561 0.888, 2.125 <0.001 Urine albuminuria (vs. none) 1.175 1.079, 1.270 <0.001 1.605 1.429, 1.778 <0.001 Thrombocytopenia (vs. none) 0.564 0.446, 0.680 <0.001 0.707 0.475, 0.928 <0.001

AKI = acute kidney injury (acute renal failure), CI = confidence interval, BMI = body mass index, BP = blood pressure, RAAS = renin-angiotensin-aldosterone system, eGFR = estimated glomerular filtration rate. General surgery, Neurosurgery, Obstetrics and gynecology, Orthopedics, Urologic surgery), body mass index (Underweight (vs. reference range), Obesity (vs. reference range)), Non-general anesthesia (vs. general)), Blood pressure (Hypotensive before surgery (vs. normotensive), Hypertensive before surgery (vs. normotensive)), and Hypernatremia (vs. reference range) were not included in the additional model construction due to significant differences in coefficients at each critical point, Parallel estimation can hardly be estimated with variables.

Model coefficient 95% CI P Age <40 Reference ≥40 and <60 0.473 0.302, 0.644 <0.001 ≥60 and <80 0.787 0.617, 0.956 <0.001 80≥ 1.120 0.875, 1.366 <0.001 Male sex (vs. female) 0.714 0.624, 0.804 <0.001 Heart disease (vs. none) 0.190 0.006, 0.374 0.042 Hypertension (vs. none) 0.176 0.069, 0.282 0.001 Diabetes mellitus (vs. none) 0.263 0.135, 0.391 <0.001 Expected surgery duration (hours) 0.451 0.424, 0.478 <0.001 Emergency operation (vs. non-emergency) 0.690 0.448, 0.931 <0.001 Preoperative RAAS blockade use (vs. none) 0.473 0.338, 0.608 <0.001 eGFR ≥60 ≥45 and <60 0.663 0.532, 0.794 <0.001 ≥30 and <45 1.321 1.113, 1.528 <0.001 ≥15 and <30 1.921 1.607, 2.235 <0.001 Leukopenia (vs. reference range) 0.395 0.254, 0.536 <0.001 Leukocytosis (vs. reference range) 0.027 -0.117, 0.172 0.712 Anemia (vs. none) 0.292 0.192, 0.392 <0.001 Hypoalbuminemia (vs. none) 0.683 0.563, 0.802 <0.001 Hyponatremia (vs. reference range) 0.307 0.162, 0.452 <0.001 Hypokalemia (vs. reference range) 0.169 -0.080, 0.418 0.184 Hyperkalemia (vs. reference range) 0.270 -0.155, 0.696 0.212 Urine albuminuria (vs. none) 0.533 0.421, 0.646 <0.001 Thrombocytopenia (vs. none) 0.041 -0.090, 0.172 0.543

CI = confidence interval, RAAS = renin angiotensin aldosterone system, eGFR = estimated glomerular filtration rate Leukocytosis (Leukocytosis ( vs. reference range), Hypokalemia (vs. reference range), Hyperkalemia (vs. reference range), and Thrombocytopenia (vs. none) were significant for each threshold coefficient. Due to differences, they were not included in further model building.

model coefficient 95% CI P Age (vs. <40) ≥40 and <60 0.502 0.332, 0.672 <0.001 ≥60 and <80 0.807 0.638, 0.976 <0.001 ≥80 1.136 0.892, 1.379 <0.001 Male sex (vs. female) 0.705 0.616, 0.794 <0.001 Heart disease (vs. none) 0.187 0.004, 0.370 0.040 Hypertension (vs. none) 0.171 0.065, 0.276 < 0.001 Diabetes mellitus (vs. none) 0.275 0.148, 0.401 <0.001 Expected surgical duration (continuous, hours) 0.458 0.432, 0.484 <0.001 emergency operation 0.669 0.432, 0.906 <0.001 Preoperative RAAS blockade use (vs. none) 0.453 0.319, 0.588 <0.001 eGFR (vs. ≥60 mL/min/1.73 m ² ) ≥45 and <60 0.680 0.551, 0.809 <0.001 ≥30 and <45 1.331 1.127, 1.536 <0.001 ≥15 and <30 2.021 1.714, 2.328 <0.001 Leukopenia (vs. none) 0.204 0.097, 0.310 <0.001 Anemia (vs. none) 0.305 0.206, 0.404 <0.001 Hypoalbuminemia (vs. none) 0.667 0.550, 0.784 <0.001 Hyponatremia (vs. none) 0.296 0.153, 0.438 <0.001 Albuminuria (vs. none) 0.505 0.394, 0.617 <0.001

CI = confidence interval, RAAS = renin angiotensin aldosterone system, eGFR = estimated glomerular filtration rate The additional simple model consists of age and additional ten variables according to the size of the model coefficients, and among the variables in Table 5, heart disease (Heart disease ( vs. none), Hypertension (vs. none)), and Leukopenia (vs. none)) were not included in the further model construction because of the relatively small model coefficients.

model coefficients 95% CI P Age (vs. <40) ≥40 and <60 0.522 0.353, 0.691 <0.001 ≥60 and <80 0.852 0.686, 1.019 <0.001 ≥80 1.203 0.962, 1.443 <0.001 Male sex (vs. female) 0.705 0.616, 0.794 <0.001 Diabetes mellitus (vs. none) 0.347 0.227, 0.467 <0.001 Expected surgical duration (continuous, hours) 0.459 0.433, 0.484 <0.001 emergency operation 0.678 0.441, 0.915 <0.001 RAAS blockade use (vs. none) 0.506 0.375, 0.638 <0.001 eGFR (vs. ≥60 mL/min/1.73 m ² ) ≥45 and <60 0.690 0.561, 0.818 <0.001 ≥30 and <45 1.345 1.141, 1.549 <0.001 ≥15 and <30 2.012 1.705, 2.319 <0.001 Anemia (vs. none) 0.319 0.221, 0.418 <0.001 Hypoalbuminemia (vs. none) 0.705 0.590, 0.820 <0.001 Hyponatremia (vs. none) 0.298 0.156, 0.441 <0.001 Albuminuria (vs. none) 0.510 0.399, 0.621 <0.001

CI = confidence interval, RAAS = renin angiotensin aldosterone system, eGFR = estimated glomerular filtration rate Model coefficients are multiplied by 11.0306 and rounded to an integer to construct a SPARK index.

(3) SPARK index and classification

The selected variables were fitted to sequential results with a proportional odds model. Although some underestimations were found in the high probability range of the discovery cohort, the calibration plot showed an acceptable distribution of estimated and expected probabilities (Figure 3). In 1000 random subsamples of fixed size (n = 1,000), the Hosmer-Lemeshow test showed 0.372 [interquartile range (IQR) 0.171 to 0.592, P 90.9% of samples with ≧0.05] and a median P value of 0.485 (IQR 0.171 to 0.739, 88.3% of samples with P≧0.05). Median P values within the discovery cohort from the same test also indicated that the model fit was significantly acceptable (P > 0.05). However, the calibration results were relatively good within the validation cohort, as a small proportion of subsamples indicated good model fit. The median P value for low-stage AKI was 0.119 (67.8% of samples with IQR 0.032 to 0.294, P ≥ 0.05), and the median P value for severe AKI was 0.130 (samples with IQR 0.016 to 0.437, P ≥ 0.05). of 65.0%). c-stats were 0.798 and 0.715, respectively, in the discovery and validation cohorts, which were acceptable ranges. After converting the model coefficients to integer scores, the final SPARK index is PO-AKI [area under the discovery cohort curve (AUC) 0.800 (95% CI 0.791-0.809), validation cohort AUC = 0.717 (95% CI 0.705-0.730)] and severe AKI [discovery cohort AUC = 0.826 (95% CI 0.810-0.843), validation cohort AUC = 0.765 (95% CI 0.743-0.786)] outcomes.

Then, we tested the sensitivity/specificity values (Fig. 4), and cutoff values of 20 and 40 were selected as the threshold values between SPARK grades A/B and B/C, respectively (Table 7). . Additionally, a threshold of 60 was identified as the threshold between SPARK C and D grades. After classification was performed with the designated cutoff values, the incidence rates of both AKI and severe AKI showed a grade-dependent increase within the discovery and validation cohorts ( FIG. 5 ). With the above-mentioned results, we constructed a SPARK index and classification, and proposed a preoperative AKI monitoring strategy (Fig. 6).

Sensitivity Specificity Negative predictive value Positive predictive value PO-AKI Discovery cohort Cutoff = 20 96.0% 27.2% 99.2% 6.8% Cutoff = 40 51.0% 88.6% 97.0% 20.0% Validation cohort Cutoff = 20 95.9% 16.0% 98.6% 5.9% Cutoff = 40 38.5% 85.8% 96.2% 13.0% Critical AKI (among SPARK index ≥40) Discovery cohort Cutoff = 60 25.8% 91.7% 96.0% 13.9% Validation cohort Cutoff = 60 18.2% 94.9% 96.0% 14.8%

PO-AKI = postoperative acute kidney injury, AKI = acute kidney injury, SPARK = simple postoperative acute kidney injury risk

(4) sensitivity analysis

A sensitivity analysis was performed to investigate whether there was any significant bias due to our exclusion criteria. The discriminant power of the SPARK index in the imputed dataset, including cases where missing values were present, was acceptable in the discovery cohort [c-stat = 0.802 (N = 51,041)]. However, within the validation cohort with a high proportion of missing values, discriminant power, which was largely missing the dipstick proteinuria variable, was relatively decreased [c-stat = 0.698 (N = 39,764)]. There was no significant decrease in discriminant power when the actual surgical duration was included instead of the expected duration [c-stat = 0.810 in the discovery cohort (N = 49,803) and c-stat = 0.723 in the validation cohort (N = 29,715)]. Then, after combining the discovery and validation cohorts, we tested for the presence of obvious differences over time, and no significant differences or reductions in c-stat were observed in the three time periods of this experiment [2004 to c-stat = 0.754 for 2007 (N = 18,560), c-stat = 0.779 for 2008 to 2011 (N = 34,016) and c-stat = 0.768 for 2012 to 2015 (N = 26,942)]. Finally, to control for a potential bias from the set of preoperative subacute or chronic progressive renal injury, we found a distinct preoperative creatinine ≥0.3 mg/dL or ≥1.5 fold from a minimum value within 3 months preoperatively, regardless of duration. Patients with this elevation were excluded. Also in the analysis, the discriminant power of the SPARK index remained within an acceptable range [c-stat = 0.792 in the discovery cohort (N = 48,124) and c-stat = 0.711 in the validation cohort (N = 29,315)].

(5) SPARK index performance and classification in each surgical department

When we merged cases without missing values in the discovery and validation cohort, certain differences related to clinical characteristics existed between departments (Table 8). The performance results of the simplified Biret-type odds model and the SPARK index of each department are shown in FIG. 7 . The correction results indicated that all corrections were acceptable in general surgery and orthopedic surgery and good in obstetrics and gynecology. However, it was found that our model underestimates the risk of adverse outcomes in urological surgery, while marked overestimation in neurosurgery. This showed a similar trend in discriminant power, and the AUC value in neurosurgery and urology departments was 0.7 or less, indicating relatively low discriminant power. Nevertheless, when the SPARK classification was applied, an increase in the incidence according to the significant grade of adverse outcomes was again observed (Table 9).

General surgery
(N=31,810) Orthopedic surgery (N=24,873) Obstetrics and gynecology (N=8,351) Neurosurgery (N=7,280) Urologic surgery (N=7,204) P Age (years) 58 [49; 68] 59 [44; 69] 45 [36; 52] 56 [45; 65] 66 [57; 72] <0.001 <40 3,023 (9.5%) 5,171 (20.8%) 2,827 (33.9%) 1,186 (16.3%) 692 (9.6%) ≥40 and <60 13,949 (43.9%) 7,559 (30.4%) 4,332 (51.9%) 3,185 (43.8%) 1,473 (20.4%) ≥60 and <80 13,869 (43.6%) 11,149 (44.8%) 1,133 (13.6%) 2,813 (38.6%) 4,780 (66.4%) ≥80 969 (3.0%) 994 (4.0%) 59 (0.7%) 96 (1.3%) 259 (3.6%) eGFR (mL/min/1.73 m ² ) 82.7 [71.7; 95.2] 85.7 [72.6; 98.0] 86.5 [75.4; 100.8] 85.5 [72.8; 99.4] 76.6 [64.7; 88.2] <0.001 ≥60 29,432 (92.5%) 22,516 (90.5%) 8,042 (96.3%) 6,740 (92.6%) 5,936 (82.4%) ≥45 and <60 1,895 (6.0%) 1,830 (7.4%) 250 (3.0%) 440 (6.0%) 963 (13.4%) ≥30 and <45 366 (1.2%) 401 (1.6%) 43 (0.5%) 74 (1.0%) 226 (3.1%) ≥15 and <30 117 (0.4%) 126 (0.5%) 16 (0.2%) 26 (0.4%) 79 (1.1%) Dipstick albuminuria 2,806 (8.8%) 1,459 (5.9%) 699 (8.4%) 402 (5.5%) 1,316 (18.3%) <0.001 male sex 16,900 (53.1%) 11,025 (44.3%) 0 (0.0%) 3,366 (46.2%) 6,455 (89.6%) <0.001 Expected surgery duration (hours) 3.0 [2.0; 3.5] 2.5 [2.0; 3.0] 2.0 [2.0; 3.0] 4.0 [3.0; 5.0] 3.0 [2.0; 4.0] <0.001 emergency department 444 (1.4%) 296 (1.2%) 164 (2.0%) 233 (3.2%) 58 (0.8%) <0.001 Diabetes mellitus 2,705 (8.5%) 2,522 (10.1%) 255 (3.1%) 595 (8.2%) 597 (8.3%) <0.001 RAAS blockade use 1,647 (5.2%) 1,756 (7.1%) 235 (2.8%) 712 (9.8%) 472 (6.6%) <0.001 Albumin (g/dL) 4.2 [3.9; 4.5] 4.4 [4.1; 4.6] 4.3 [4.1; 4.5] 4.2 [3.9; 4.5] 4.4 [4.2; 4.6] <0.001 Hypoalbuminemia (<3.5) 3,126 (9.8%) 1,478 (5.9%) 554 (6.6%) 683 (9.4%) 197 (2.7%) Hemoglobin (g/dL) 13.3 [12.0; 14.4] 13.6 [12.5; 14.9] 12.7 [11.6; 13.5] 13.4 [12.4; 14.5] 14.3 [13.2; 15.2] <0.001 Anemia (<12 for female, <13 for male) 9,425 (29.6%) 4,444 (17.9%) 2,555 (30.6%) 1,631 (22.4%) 1,280 (17.8%) Sodium (mEq/L) 141 [139; 142] 141 [139; 142] 140 [139; 141] 141 [139; 142] 141 [139; 142] <0.001 Hyponatremia (<135) 933 (2.9%) 545 (2.2%) 102 (1.2%) 169 (2.3%) 111 (1.5%)

eGFR = estimated glomerular filtration rate, RAAS = renin-angiotensin-aldosterone system

Class A Class B Class C Class D P General surgery (N=6,228) (N=19,988) (N=4,988) (N=606) Any AKI 47 (0.8%) 785 (3.9%) 827 (16.6%) 347 (57.3%) <0.001 Low-stage AKI 41 (0.7%) 606 (3.0%) 610 (12.2%) 259 (42.7%) <0.001 Critical AKI 6 (0.1%) 179 (0.9%) 217 (4.4%) 88 (14.5%) <0.001 Orthopedic surgery (N=6,119) (N=16,776) (N=1,908) (N=70) Any AKI 78 (1.3%) 604 (3.6%) 219 (11.5%) 14 (20.0%) <0.001 Low-stage AKI 71 (1.2%) 539 (3.2%) 169 (8.9%) 8 (11.4%) <0.001 Critical AKI 7 (0.1%) 65 (0.4%) 50 (2.6%) 6 (8.6%) <0.001 Obstetrics and gynecology (N=4,300) (N=3,900) (N=145) (N=6) Any AKI 26 (0.6%) 68 (1.7%) 20 (13.8%) 1 (16.7%) <0.001 Low-stage AKI 17 (0.4%) 44 (1.1%) 13 (9.0%) 0 (0.0%) <0.001 Critical AKI 9 (0.2%) 24 (0.6%) 7 (4.8%) 1 (16.7%) <0.001 Neurosurgery (N=287) (N=5,133) (N=1,781) (N=79) Any AKI 7 (2.4%) 91 (1.8%) 76 (4.3%) 15 (19.0%) <0.001 Low-stage AKI 6 (2.1%) 71 (1.4%) 52 (2.9%) 6 (7.6%) <0.001 Critical AKI 1 (0.3%) 20 (0.4%) 24 (1.3%) 9 (11.4%) <0.001 Urologic surgery (N=552) (N=4,935) (N=1,603) (N=114) Any AKI 10 (1.8%) 526 (10.7%) 361 (22.5%) 57 (50.0%) <0.001 Low-stage AKI 6 (1.1%) 376 (7.6%) 240 (15.0%) 37 (32.5%) <0.001 Critical AKI 4 (0.7%) 150 (3.0%) 121 (7.5%) 20 (17.5%) <0.001

AKI = acute kidney injury

Claims (18)

a variable selection unit that selects, as a variable, a factor related to the occurrence of acute renal failure after non-cardiac surgery, which consists of clinical data before non-cardiac surgery of patients subject to non-cardiac surgery, and sets an index set for each variable;
a classification reference point setting unit calculating sensitivity and specificity with respect to the sum of combinations of indices for each index set, and setting cut-off values according to the sensitivity and specificity; and
For the non-cardiac surgery target patient requiring prediction, the sum of the indices determined according to the set of indices set for each variable selected by the variable selection unit is calculated, and the non-cardiac surgery requiring the prediction based on the classification reference points set in the classification reference point setting unit Includes; a risk prediction unit for acute renal failure after non-cardiac surgery that predicts the risk of acute renal failure after non-cardiac surgery of the target patient;
The factors related to the occurrence of acute renal failure after noncardiac surgery were age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, sex, and expected duration of surgery (Expected). surgical duration, emergency operation, diabetes mellitus, renin-aldosterone-angiotensin-system blockade use (RAAS blockade use), hypoalbuminemia , A system for predicting the risk of acute renal failure after non-cardiac surgery, including anemia and hyponatremia.
The method according to claim 1, wherein the set of indices is set according to the selection criteria for each variable except for the expected surgical duration, and each selection category is selected from 0, 3, 4, 6 to 9, 13, 15, and 22. A system for predicting the risk of acute renal failure after non-cardiac surgery, in which any one index is set, and the sum of the indices for each selection category of the variable is 0 to 81.
The system for predicting the risk of acute renal failure after non-cardiac surgery according to claim 1, wherein the index set for the expected surgical duration among the variables is set to be five times the expected surgical duration (time).
The method according to claim 1, wherein the age (age) of the variables is divided into selection categories of less than 40 years old, 40 years old and less than 60 years old, 60 years old or more and less than 80 years old, and 80 years old or more,
The estimated glomerular filtration rate (eGFR) is 60 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 60 mL/min/1.73 m ² or less, 30 mL/min/1.73 A system for predicting the risk of acute renal failure after non-cardiac surgery, divided into selection categories of m ² or more and 45 mL/min/1.73 m ² , 15 mL/min/1.73 m ² or more and 30 mL/min/1.73 m ² .
The method according to claim 1, wherein the non-cardiac surgery target patients after surgery to confirm the occurrence of acute renal injury and further comprising a severity determination unit for classifying 'no AKI', 'low-stage AKI' and 'severe AKI' according to the severity, A system for predicting the risk of acute renal failure after non-cardiac surgery.
The method according to claim 1, wherein the prediction of the risk of acute renal failure after non-cardiac surgery is A, B, depending on the sum of the indices determined according to the set of indices set for each variable selected in the variable selection unit for the non-cardiac surgery target patient requiring prediction. A system for predicting the risk of acute renal failure after non-cardiac surgery, classified into a total of four grades, C and D.
The method according to claim 6, wherein the A grade is when the sum of the indices is less than 20, the B grade is when the sum of the indices is 20 or more and less than 40, and the C grade is when the sum of the indices is 40 or more and less than 60, the D The grade will be classified when the sum of the indices is 60 or more, the acute renal failure risk prediction system after non-cardiac surgery.
The method according to claim 7, wherein the A grade, the probability of occurrence of acute renal failure and severe acute renal failure after non-cardiac surgery of the non-cardiac surgery target patient requiring the prediction is less than 2%,
In the B grade, the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 2% or more, and the probability of occurrence of severe acute renal failure is less than 2%,
The grade C indicates that the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 10% or more, and the probability of occurrence of severe acute renal failure is 2% or more,
The D grade is the probability of occurrence of acute renal failure after non-cardiac surgery of patients subject to non-cardiac surgery requiring the prediction is 20% or more, and the probability of occurrence of severe acute renal failure is 10% or more, A system for predicting the risk of acute renal failure after non-cardiac surgery.
A first step of selecting as a variable a factor related to the occurrence of acute renal failure after non-cardiac surgery, which consists of clinical data before non-cardiac surgery of patients subject to non-cardiac surgery, and setting an index set for each variable;
a second step of calculating sensitivity and specificity for the sum of combinations of indices for each index set, and setting cut-off values according to the sensitivity and specificity; and
For the non-cardiac surgery target patient requiring prediction, the sum of the indices determined according to the set of indices set for each variable selected in the first step is calculated, and the non-cardiac surgery target requiring the prediction based on the classification reference points set in the second step A third step of predicting the risk of acute renal failure after non-cardiac surgery of the patient;
The factors related to the occurrence of acute renal failure after noncardiac surgery were age, estimated glomerular filtration rate (eGFR), dipstick albuminuria, sex, and expected duration of surgery (Expected). surgical duration, emergency operation, diabetes mellitus, renin-aldosterone-angiotensin-system blockade use (RAAS blockade use), hypoalbuminemia , A method for predicting the risk of acute renal failure after non-cardiac surgery, including anemia and hyponatremia.
The method according to claim 9, wherein the set of indices is set according to a selection category for each variable except for the expected surgical duration, and for each selection category, a selection of 0, 3, 4, 6 to 9, 13, 15, and 22 A method for predicting the risk of acute renal failure after non-cardiac surgery, in which any one index is set, and the sum of the indices for each selection category of the variable is 0 to 81.
The method for predicting the risk of acute renal failure after non-cardiac surgery according to claim 9, wherein the index set for the expected surgical duration among the variables is set to be five times the expected surgical duration (time).
The method according to claim 9, wherein age among the variables is divided into selection categories of less than 40 years old, more than 40 years old and less than 60 years old, more than 60 years old and less than 80 years old, and more than 80 years old,
The estimated glomerular filtration rate (eGFR) is 60 mL/min/1.73 m ² or more, 45 mL/min/1.73 m ² or more, 60 mL/min/1.73 m ² or less, 30 mL/min/1.73 A method for predicting the risk of acute renal failure after non-cardiac surgery, divided into selection categories of m ² or more and 45 mL/min/1.73 m ² , 15 mL/min/1.73 m ² or more and 30 mL/min/1.73 m ² .
The method of claim 9, further comprising a severity determination step of confirming the occurrence of acute renal injury after surgery in the non-cardiac surgery target patients and classifying 'No AKI', 'Low AKI' and 'Severe AKI' according to the severity , A method for predicting the risk of acute renal failure after non-cardiac surgery.
The method according to claim 9, wherein the prediction of the risk of occurrence of acute renal failure after non-cardiac surgery is A, B, according to the sum of indices determined according to the set of indices set for each variable selected in the first step for the non-cardiac surgery target patient requiring prediction. A method for predicting the risk of acute renal failure after non-cardiac surgery, classified into four grades, C and D.
The method of claim 14, wherein the A grade is when the sum of the indices is less than 20, the B grade is when the sum of the indices is 20 or more and less than 40, and the C grade is when the sum of the indices is 40 or more and less than 60, the D The grade will be classified when the sum of the indices is 60 or more, a method of predicting the risk of acute renal failure after non-cardiac surgery.
The method according to claim 15, wherein the A grade is that the probability of occurrence of acute renal failure and severe acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is less than 2%,
In the B grade, the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 2% or more, and the probability of occurrence of severe acute renal failure is less than 2%,
The grade C indicates that the probability of occurrence of acute renal failure after non-cardiac surgery in patients subject to non-cardiac surgery requiring the prediction is 10% or more, and the probability of occurrence of severe acute renal failure is 2% or more,
The D grade is a method for predicting the risk of acute renal failure after non-cardiac surgery, in which the probability of occurrence of acute renal failure after non-cardiac surgery is 20% or more, and the probability of occurrence of severe acute renal failure is 10% or more in patients subject to non-cardiac surgery requiring the prediction.
The method according to claim 1, wherein the variable selection unit selects the factors related to the occurrence of acute renal failure after non-cardiac surgery as variables through proportional odds regression analysis in consideration of the severity of acute renal injury. A system for predicting the risk of developing acute renal failure.
The method according to claim 9, wherein in the second step, the factors related to the occurrence of acute renal failure after non-cardiac surgery are selected as variables through proportional odds regression analysis in consideration of the severity of acute renal injury. A method for predicting the risk of developing acute renal failure.

Images