Methods and computer-implemented methods for screening subjects for risk of chronic kidney disease

Abstract

The present disclosure relates to a method for screening a subject for a risk of chronic kidney disease (CKD), comprising: receiving marker data indicating a plurality of marker parameters of the subject, such plurality of marker parameters indicating an age value, a creatinine sample level, and an albumin sample level of the subject during a measurement period; and determining a risk factor indicating a risk of the subject suffering from CKD from the plurality of marker parameters, wherein the determining comprises: giving a higher weight to the age value than to the albumin level, and giving a higher weight to the creatinine sample level than to the albumin sample level. In addition, a computer-implemented method for screening a subject and a method for screening a subject for a risk of chronic kidney disease (CKD) are provided.

Description

Technical Field

[0001] The present invention relates to methods and computer-implemented methods for screening subjects for risk of chronic kidney disease. Background Art

[0002] In chronic kidney disease (CKD), kidney function is lost gradually, beginning with a decrease in glomerular filtration rate and / or proteinuria, and progressing to end-stage renal disease. As a result, dialysis or kidney transplantation may be necessary (see Unger, J., Schwartz, Z., “ 糖尿病初级保健管理 ” 2nd ed., LippincottWilliams & Wilkens, Philadelphia, USA, 2013. CKD is a serious problem, with an adjusted prevalence of 7% in 2013 (Glassock, RJ et al., “ 慢性肾脏病的全球负担:估计、 变异性和陷阱 ”, Nat Rev Nephrol 13,104–114, 2017. Early identification of CKD can slow progression, prevent complications, and reduce cardiovascular-related consequences (Platinga, LC et al., “ 认知度 患者和医疗服务提供者对慢性肾脏病的认知度 ”, Adv Chronic Kidney Dis 17, 225-236, 2010. CKD may be a long-term microvascular complication of diabetes (Fioretto, P. et al., “ 糖尿病中的残余微血管风险:未满足的需求和未来方向 ”, Nat Rev Endocrinol 6, 19–25, 2010).

[0003] Algorithms for CKD risk prediction in diabetic patients have been published by, for example, Dunkler et al. (Dunkler et al. 2型糖尿病早期慢性肾脏病的风险预测 ”, Clin J Am Soc Nephrol 10,1371–1379, 2015; Vergouwe et al. (Vergouwe, Y. et al. “ 进展为 1型糖尿病微量白蛋白尿:预测规则的制定与验证 预测规则 ”, Diabetologia 53, 254–262, 2010; Keane et al. (Keane WF et al., “ 风险 预测2型糖尿病和肾病患者预后的评分:RENAAL研究 预测2型糖尿病和肾病患者预后的评分:RENAAL研究 ”, Clin J Am Soc Nephrol 1, 761–767, 2006) and Jardine et al. (Jardine MJ et al. “ 2型糖尿病患者肾脏相关预后的预测 2型糖尿病患者" Am J Kidney Dis. 60, 770-778, 2012). Such published algorithms are derived from data originating from primary clinical studies.

[0004] Further models for CKD risk prediction have been described by Adler Perotte et al. (“Risk prediction for chronic kidney disease progression using heterogeneous electronic health record data and time series analysis,” Journal of the American Medical Informatics Association, Volume 22, Issue 4, April 20, 2015 (2015-04-20), Pages 872-880); Paolo Fraccaro et al. (“An external validation of models to predict the onset of chronic kidney disease using population-based electronic health records from Salford, UK,” BMC Medicine, Volume 14, Issue 1, July 12, 2016 (2016-07-12); and Justin B. Echouffo-Tcheugui et al. (“Risk Models to Predict Chronic Kidney Disease and Its Progression: A Systematic Review,” Plos One Medicine, vol. 9, no. 11, 20 November 2012 (2012-11-20), p. e1001344).

[0005] Such a prediction model based on clinical data represents an ideal setting with a pre-selected population, cross-checked and validated clinical data entries, and usually a narrow observation time window. Therefore, when inferred from clinical studies, the results do not necessarily reveal the best approach in terms of efficacy and effectiveness in real-world populations. In addition, most of the literature focuses on the progression of diabetic nephropathy or CKD, and therefore misses the early stages of this diabetic complication. Finally, patients are usually selected based on the entire set of corresponding features. Summary of the invention

[0006] The aim is to provide improved methods for screening subjects for risk of chronic kidney disease, thereby allowing early risk assessment of CKD based on real-world data (RWD).

[0007] To solve this problem, methods for screening a subject for risk of chronic kidney disease (CKD) are provided according to independent claims 1 and 15, respectively. Further, a computer implemented method is provided according to independent claim 14. Further embodiments are disclosed in the dependent claims.

[0008] According to one aspect, a method for screening a subject for risk of chronic kidney disease (CKD) is provided. The method includes receiving marker data indicating a plurality of marker parameters of the subject, such plurality of marker parameters indicating age values, creatinine sample levels, and albumin sample levels for the subject during a measurement period; and determining a risk factor indicating a risk of the subject suffering from CKD from the plurality of marker parameters. The determination includes giving a higher weight to the age value than to the albumin level, and giving a higher weight to the creatinine sample level than to the albumin sample level.

[0009] According to another aspect, a computer-implemented method for screening a subject for risk of chronic kidney disease (CKD) in a data processing system having a processor and a non-transitory memory storing a program that causes the processor to:

[0010] - receiving marker data indicative of a plurality of marker parameters for a subject, such plurality of marker parameters indicative of age values, albumin sample levels, and creatinine sample levels for the subject over a measurement period; and

[0011] - determining a risk factor indicative of a risk of the subject suffering from CKD from the plurality of marker parameters, wherein the determining comprises

[0012] - give age values ​​a higher weight than albumin levels, and

[0013] - Give higher weight to creatinine sample levels compared to albumin sample levels.

[0014] According to a further aspect, a method for screening a subject for risk of chronic kidney disease (CKD) is provided. The method includes receiving marker data indicating a plurality of marker parameters of the subject, such plurality of marker parameters indicating an age value of the subject, a creatinine sample level within a measurement period, and an albumin sample level within the measurement period; and determining a risk factor indicating a risk of the subject suffering from CKD from the plurality of marker parameters. The determination includes giving a higher weight to the age value than to the albumin level, and giving a higher weight to the creatinine sample level than to the albumin sample level. At least one of the creatinine sample level and the albumin sample level indicates a generalized value of a sample level within a corresponding measurement period of each subject in a reference subject group that does not include the subject.

[0015] With regard to such a method, for each subject of the reference subject group, the measurement period may be limited to two years and may end with a diabetes diagnosis for the respective subject of the reference subject group.

[0016] For labeled data, a screening or determination of outliers can be performed before determining a risk value. In the case of an outlier determined (e.g., by checking whether the value exceeds a specific range allowed for the value), the value can be replaced by a value within the (expected) standard deviation or the upper or lower limit of the specific allowed range for the feature. For example, in the process of collecting the data, the person entering the data may have mistakenly provided the wrong decimal places for the value. Such obviously wrong values ​​can be corrected. For example, if the feature value is above the upper limit of the specific allowed range for the feature, the value can be replaced by the upper limit of the range before using it in the prediction formula. If the feature value is below the lower limit of the specific allowed range for the feature, the value can be replaced by the lower limit of the range before using it in the prediction formula.

[0017] For labeled data, screening or determination of missing data or values ​​can be performed before determining risk values. Missing data can be imputed using the mean of the group.

[0018] One or both of the above metrics may be applied to provide improved marker data for determining risk factors.

[0019] The generalized value of the sample level of the reference subject group not including the subject can be, for example, a maximum value, a minimum value, a mean value, a median value or a slope determined for a plurality of sample levels of the corresponding measurement period of each subject of the reference subject group. The subject of the subject reference group can be a diabetic patient. For example, all subjects of the subject reference group can be diabetic patients.

[0020] The marker parameters may be indicative of real-world data, which is not restricted in terms of, for example, completeness or accuracy of the data (unlike clinical data).

[0021] The subject's age value for a measurement period may be the subject's age value at the end of the measurement period.

[0022] Within the meaning of the present disclosure, weighting a first value or sample level higher than a second value or sample level means using the first value or sample level and the second value or sample level in an equation (such as an equation for determining a risk factor) in such a way that a relative change in the first value or sample level (e.g., a 10% change in the first value) has a greater impact on the outcome of the equation (e.g., the risk factor) than the same relative change in the second value or sample level (in the example above, a 10% change in the second value). For example, weighting can include multiplying the first value or sample level and the second value or sample level by an appropriate corresponding constant. Depending on the expected first value or sample level and the expected second value or sample level and their corresponding units, weighting the first value or sample level higher than the second value or sample level can include multiplying the first value or sample level by a larger or smaller constant than the second value or sample level.

[0023] The method may further include a plurality of marker parameters indicating the subject's creatinine blood sample level. Thus, the requirement to take the creatinine sample level as a concentration in urine may be avoided. For the subject, the plurality of marker parameters may indicate a creatinine blood sample level selected from a plurality of creatinine blood sample levels. For example, the selected creatinine blood sample level may be a maximum value from a plurality of creatinine blood sample levels. Alternatively or additionally, the plurality of marker parameters may indicate the subject's creatinine blood sample level calculated from a plurality of creatinine blood sample levels. For example, the calculated creatinine blood sample level may be a statistical value, such as a mean, calculated from a plurality of creatinine blood sample levels.

[0024] Creatinine sample levels may be provided in units of mg / dl (such as milligrams of creatinine per deciliter of blood).

[0025] The method may further include a plurality of marker parameters indicating the albumin blood sample level of the subject. Thus, requiring the albumin sample level as a concentration in urine may be avoided. For the subject, the plurality of marker parameters may indicate an albumin blood sample level selected from a plurality of albumin blood sample levels. For example, the selected albumin blood sample level may be a minimum value among the plurality of albumin blood sample levels. Alternatively or additionally, the plurality of marker parameters may indicate an albumin blood sample level of the subject calculated from a plurality of albumin blood sample levels. For example, the calculated albumin blood sample level may be a statistical value, such as a mean, calculated from a plurality of albumin blood sample levels.

[0026] The albumin sample level may be provided in units of g / dl (such as grams of albumin per deciliter of blood).

[0027] The subject may be a diabetic patient. Thus, the risk of chronic kidney disease in diabetic patients can be screened.

[0028] Alternatively, all of the plurality of marker parameters may be for subjects for whom a diabetes diagnosis is unavailable. For example, the subject may be at risk of becoming a diabetic. Thus, the risk of chronic kidney disease in subjects who have not been diagnosed with diabetes (e.g., subjects at risk of becoming diabetic) may be screened. The receiving may include receiving marker data indicating a plurality of marker parameters for subjects for which a diabetes diagnosis is unavailable.

[0029] The measurement period may be limited to two years. Thus, values ​​and / or sample levels of a substance collected over a period of up to two years may be provided, wherein the risk factor indicates the risk of the subject suffering from CKD from the end of the measurement period.

[0030] The subject may not have been diagnosed as having diabetes at the end of the measurement period. For example, the risk of CKD can be screened in a subject who has recently been diagnosed with diabetes, and the marker data can indicate multiple marker parameters of the subject in a measurement period that is completely before the subject's diabetes diagnosis. Alternatively, the risk of CKD can be screened for a subject who has not yet been diagnosed with diabetes, and therefore, the marker data indicates multiple marker parameters of the subject in a measurement period in which the subject has not yet been diagnosed with diabetes.

[0031] The measurement period may be at least partially after the subject's diabetes diagnosis. For example, up to 20% of the measurement period (preferably up to 10% of the measurement period) may be after the time when the subject was diagnosed with diabetes. For example, the subject may be a diabetic patient who has been diagnosed with diabetes for less than two years, and the marker data may indicate a plurality of marker parameters of the patient within a measurement period (such as a two-year measurement period), the measurement period ending directly at or shortly before the risk factor is determined, so that a portion of the plurality of marker parameters is for a time period before the patient's diabetes diagnosis, and a portion of the plurality of marker parameters is for a time period after the patient's diabetes diagnosis.

[0032] The measurement period may be completely after the diabetic patient's diagnosis of diabetes. For example, the subject may be a diabetic patient who has been diagnosed with diabetes for more than two years, and the marker data may indicate a plurality of marker parameters of the patient over a measurement period (such as a two-year measurement period) that ends directly before or shortly before the risk factor is determined.

[0033] The risk factor may indicate the risk of a subject suffering from CKD within a three-year forecast period from the end of the measurement period. The risk factor may be the probability of a subject developing CKD within three years from the last value and / or sample level being determined. Alternatively, the risk factor may indicate the risk of a subject suffering from CKD within a time period less than three years (e.g., two years) from the end of the measurement period. As a further alternative, the risk factor may indicate the risk of a subject suffering from CKD within a time period more than three years from the end of the measurement period.

[0034] The determining may further include weighting age higher than the sample creatinine level.

[0035] According to the foregoing, the marker parameters include age values, creatinine sample levels and albumin sample levels, thereby providing a simple method for calculating risk factors indicative of risk of suffering from CKD. In further embodiments, as will be described in more detail below, further marker parameters are optionally included in the risk calculation, and the further marker parameters include at least one of estimated glomerular filtration rate sample levels, body mass index, glucose sample levels and HbA1c sample levels.

[0036] Receiving may include receiving marker data indicating a plurality of marker parameters of a subject having a sample level of HbA1c less than 6.5%. HbA1C is the C fraction of glycated hemoglobin A1. The sample level of HbA1c may be provided in units of %, such as a percentage in blood. Alternatively, the sample level of HbA1c may be provided in units of mmol / mol, such as millimoles of HbA1c per mole of blood.

[0037] The method may further include a plurality of marker parameters indicative of a glomerular filtration rate sample level of the subject, and in the determination, weighting each of the age value, the albumin sample level, and the creatinine sample level higher than the glomerular filtration rate sample level.

[0038] The multiple marker parameters can indicate a selected glomerular filtration rate of the subject selected from a plurality of glomerular filtration rates. For example, the selected glomerular filtration rate can be the minimum value among the plurality of glomerular filtration rates. Alternatively or additionally, the multiple marker parameters can indicate a calculated glomerular filtration rate of the subject calculated from a plurality of glomerular filtration rates. For example, the calculated glomerular filtration rate can be a statistical value calculated from a plurality of glomerular filtration rates, such as a mean.

[0039] The glomerular filtration rate is known in the art and indicates the flow rate of filtered fluid through the kidneys and is an important indicator for estimating renal function. The glomerular filtration rate may be reduced due to kidney disease. In an embodiment, the glomerular filtration rate can be estimated using a modification of the diet in renal disease (MDRD) formula known in the art such as this. For example, the MDRD formula using four variables relies on the subject's age, gender, race, and serum creatinine to estimate the glomerular filtration rate. In an alternative embodiment, the glomerular filtration rate can be estimated using a CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) formula known in the art such as this. The CKD-EPI formula relies on the subject's age, gender, race, and serum creatinine to estimate the glomerular filtration rate. In a further embodiment, the glomerular filtration rate can be estimated using other methods or can be determined directly. The sample glomerular filtration rate can be expressed in ml / min / 1.73m 2 It is given in units of (ml per minute per 1.73 m2 of body surface).

[0040] Risk factors ( ) can be determined according to the following equation:

[0041]

[0042] In this paper, P` CKD预测 It can be calculated as:

[0043] P` CKD_预测 = c` CKD1 ∙ Age + c` CKD2 ∙ Creatinine + c` CKD3 ∙Albumin + c` CKD4 ,

[0044] where age is the subject's age in years, creatinine is the subject's sample creatinine level, albumin is the subject's sample albumin level, and c` CKD1 ,c` CKD2 ,c` CKD3 and c` CKD4 is a constant.

[0045] In an alternative approach, the risk factor can be determined according to the following equation:

[0046]

[0047] In this paper, P CKD预测 It can be calculated as:

[0048] P CKD预测 = c CKD1 ∙ Age + c CKD2 ∙ Creatinine + c CKD3∙Albumin + c CKD4 ,

[0049] and P 死亡预测 It can be calculated as:

[0050] P 死亡预测 = c 死亡1 ∙ Age + c 死亡2 ∙ Creatinine + c 死亡3 ∙Albumin + c 死亡4 ,

[0051] where age is the subject's age in years, creatinine is the subject's sample creatinine level, albumin is the subject's sample albumin level, and c CKD1 、c CKD2 、c CKD3 、c CKD4 、c 死亡1 、c 死亡2 、c 死亡3 and c 死亡4 is a constant. In the case where there is a death prediction revealed from the RWD analysis, such a formula can be applied. Otherwise, as outlined above, the constant for the death prediction can be omitted.

[0052] The creatinine sample level may be a creatinine sample level from a plurality of creatinine sample levels. The albumin sample level may be an albumin sample level from a plurality of albumin sample levels. The creatinine sample level and / or the albumin sample level may be representative sample levels from a corresponding plurality of sample levels of creatinine and / or albumin, such as a maximum sample level, a minimum sample level, an average sample level, and / or a median of the sample levels. In an exemplary embodiment, creatinine is a maximum creatinine sample level from a plurality of creatinine sample levels of a subject, and albumin is a minimum albumin sample level from a plurality of albumin sample levels of a subject.

[0053] Constant c` CKD1 ,c` CKD2 ,c` CKD3 and c` CKD4 Can be a model specific constant. In an embodiment, the constant c' CKD1 ,c` CKD2 and c` CKD3 It can be a constant weighting factor associated with the corresponding marking parameter.

[0054] Constant c CKD1 、c CKD2 、c CKD3 and c CKD4 and c 死亡1 、c 死亡2 、c 死亡3 and c死亡4 Can be a model specific constant. In an embodiment, the constant c CKD1 、c CKD2 and c CKD3 and c 死亡1 、c 死亡2 and c 死亡3 It can be a constant weighting factor associated with the corresponding marking parameter.

[0055] For example, the constants can be:

[0056] c CKD1 :0.02739 / year;

[0057] c CKD2 : 1.387 dl / mg;

[0058] c CKD3 : -0.3356 dl / g; and

[0059] c CKD4 : -3.1925;

[0060] c 死亡1 :0.06103 / year;

[0061] c 死亡2 :0.8194 dl / mg;

[0062] c 死亡3 : -0.9336 dl / g; and

[0063] c 死亡4 :-3.3325.

[0064] In an embodiment, any or each constant may be selected from a range of + / - 30% around such respective value, preferably from a range of + / - 20%, and more preferably from a range of + / - 10%.

[0065] Risk factors ( ) can be determined according to the following equation:

[0066]

[0067] In this paper, P`' CKD预测 It can be calculated as:

[0068] P`' CKD预测 = c`' CKD1 ∙ Age + c`' CKD2 ∙ Creatinine + c`' CKD3 ∙Albumin + c`' CKD4 + c`' CKD5 ∙eGFR

[0069] where age is the subject's age in years, creatinine is the subject's sample creatinine level, albumin is the subject's sample albumin level, eGFR is the subject's sample estimated glomerular filtration rate level, and c`' CKD1 、c`' CKD2 、c`' CKD3 、c`' CKD4 and c`' CKD5 is a constant.

[0070] In another example, the risk factor may be determined according to the following equation ( ):

[0071]

[0072] In this paper, P' CKD预测 It can be calculated as:

[0073] P' CKD预测 = c' CKD1 ∙ Age + c' CKD2 ∙ Creatinine + c' CKD3 ∙Albumin + c' CKD4 + c' CKD5 ∙eGFR,

[0074] and P' 死亡预测 It can be calculated as:

[0075] P' 死亡预测 = c' 死亡1 ∙ Age + c' 死亡2 ∙ Creatinine + c' 死亡3 ∙Albumin + c' 死亡4 + c' 死亡5 ∙eGFR,

[0076] where age is the subject's age in years, creatinine is the subject's sample creatinine level, albumin is the subject's sample albumin level, eGFR is the subject's sample estimated glomerular filtration rate level, and c' CKD1 、c' CKD2 、c' CKD3 、c' CKD4 、c' CKD5 、c' 死亡1 、c' 死亡2 、c' 死亡3 、c' 死亡4 and c' 死亡5 is a constant. In the case where there is a death prediction revealed from the RWD analysis, such a formula can be applied. Otherwise, as outlined above, the constant for the death prediction can be omitted.

[0077] The creatinine sample level may be a creatinine sample level from a plurality of creatinine sample levels. The albumin sample level may be an albumin sample level from a plurality of albumin sample levels.

[0078] With respect to estimating glomerular filtration rate, glomerular filtration rate can be estimated from the multiple levels available for a subject.

[0079] The creatinine sample level, the albumin sample level, and / or the estimated glomerular filtration rate sample level may be representative sample levels from a corresponding plurality of sample levels of creatinine, albumin, and / or estimated glomerular filtration rate, such as a maximum sample level, a minimum sample level, an average sample level, and / or a median of the sample levels. In an exemplary embodiment, creatinine is the maximum creatinine sample level from a plurality of creatinine sample levels from a subject, albumin is the minimum albumin sample level from a plurality of albumin sample levels from a subject, and eGFR is the minimum estimated glomerular filtration rate sample level from a plurality of estimated glomerular filtration rate sample levels from a subject.

[0080] Constant c`' CKD1 、c`' CKD2 、c`' CKD3 、c`' CKD4 and c`' CKD5 Can be a model specific constant. In an embodiment, the constant c' CKD1 、c`' CKD2 , and c`' CKD3 and c`' CKD5 It can be a constant weighting factor associated with the corresponding marking parameter.

[0081] Constant c' CKD1 、c' CKD2 、c' CKD3 、c' CKD4 and c' CKD5 , and c' 死亡1 、c' 死亡2 、c' 死亡3 、c' 死亡4 and c' 死亡5 Can be a model specific constant. In an embodiment, the constant c' CKD1 、c' CKD2 、c' CKD3 and c' CKD5 , and c' 死亡1 、c' 死亡2 、c' 死亡3 and c' 死亡5 It can be a constant weighting factor associated with the corresponding marking parameter.

[0082] For example, in such an embodiment, the constants may be as follows:

[0083] c' CKD1 :0.02739 / year;

[0084] c' CKD2 : 1.387 dl / mg;

[0085] c' CKD3 : -0.3356 dl / g;

[0086] c' CKD4 : -1.3013; and

[0087] c' CKD5 :-0.02843 min∙1.73m 2 / ml;

[0088] c' 死亡1 :0.06103 / year;

[0089] c' 死亡2 :0.8194 dl / mg;

[0090] c' 死亡3 : -0.9336 dl / g;

[0091] c' 死亡4 : -4.4328; and

[0092] c' 死亡5 :0.01654 min∙1.73m 2 / ml.

[0093] In an embodiment, any or each of said constants may be selected from a range of + / - 30% around such respective value, preferably from a range of + / - 20%, and more preferably from a range of + / - 10%.

[0094] In a further embodiment, the risk factor (P'') may be determined according to the following equation: CKD ):

[0095]

[0096] In this paper, P`'' CKD预测 It can be calculated as:

[0097] P`'' CKD预测 = c`'' CKD1 ∙ Age + c`'' CKD2 ∙ Creatinine + c`'' CKD3 ∙Albumin + c`'' CKD4+ c`'' CKD5 ∙eGFR + c`'' CKD6 ∙BMI + c`'' CKD7 ∙ Glucose + c`'' CKD8 ∙HbA1c,

[0098] where age is the subject's age in years, creatinine is the subject's creatinine sample level, albumin is the subject's albumin sample level, eGFR is the subject's estimated glomerular filtration rate sample level, BMI is the subject's body mass index (BMI) value, glucose is the subject's glucose sample level, HbA1c is the subject's glycated hemoglobin A1c fraction C sample level, and c''' CKD1 、c`'' CKD2 、c`'' CKD3 、c`'' CKD4 、c`'' CKD5、 c`'' CKD6、 c`'' CKD7 and c`'' CKD8 is a constant. BMI can be expressed in kg / m 2 The minimum sample level of glucose may be provided in units of mg / dl (such as milligrams of glucose per deciliter of blood).

[0099] In another example, the risk factor (P`'') may be determined according to the following equation: CKD ):

[0100]

[0101] In this paper, P'' CKD预测 It can be calculated as:

[0102] P'' CKD预测 = c'' CKD1 ∙ Age + c'' CKD2 ∙ Creatinine + c'' CKD3 ∙Albumin + c'' CKD4 + c'' CKD5 ∙eGFR + c'' CKD6 ∙BMI + c'' CKD7 ∙ Glucose + c'' CKD8 ∙HbA1c,

[0103] and P'' 死亡预测 It can be calculated as:

[0104] P'' 死亡预测 = c'' 死亡1 ∙ Age + c''死亡2 ∙ Creatinine + c'' 死亡3 ∙Albumin + c'' 死亡4 + c'' 死亡5 ∙eGFR + c'' 死亡6 ∙BMI + c'' 死亡7 ∙ Glucose + c'' 死亡8 ∙HbA1c,

[0105] where age is the subject's age in years, creatinine is the subject's creatinine sample level, albumin is the subject's albumin sample level, eGFR is the subject's estimated glomerular filtration rate sample level, BMI is the subject's body mass index (BMI) value, glucose is the subject's glucose sample level, HbA1c is the subject's glycated hemoglobin A1c C fraction sample level, and c'' CKD1 、c'' CKD2 、c'' CKD3 、c'' CKD4 、c'' CKD5 、c'' CKD6 、c'' CKD7 、c'' CKD8 、c'' 死亡1 、c'' 死亡2 、c'' 死亡3 、c'' 死亡4 、c'' 死亡5 、c'' 死亡6 、c'' 死亡7 and c'' 死亡8 is a constant. BMI can be expressed in kg / m 2 The formula may be provided in units of kilograms per square meter and determined as known in the art. The lowest sample level of glucose may be provided in units of mg / dl (such as milligrams of glucose per deciliter of blood). In the event that there is a prediction of death revealed from the RWD analysis, such a formula may be applied. Otherwise, the constants for the prediction of death may be omitted, as outlined above.

[0106] In an embodiment, any or each constant may be selected from a range of + / - 30% around such respective value, preferably from a range of + / - 20%, and more preferably from a range of + / - 10%.

[0107] The creatinine sample level may be a creatinine sample level from a plurality of creatinine sample levels from the subject, the albumin sample level may be an albumin sample level from a plurality of albumin sample levels from the subject, the estimated glomerular filtration rate sample level may be an estimated glomerular filtration rate sample level from a plurality of estimated glomerular filtration rate sample levels from the subject, the BMI value may be a BMI value from a plurality of body mass index (BMI) values ​​from the subject, the glucose sample level may be a glucose sample level from a plurality of glucose sample levels from the subject, and / or the C fraction of glycated hemoglobin A1 sample level may be a C fraction of glycated hemoglobin A1 sample level from a plurality of C fraction of glycated hemoglobin A1 sample levels from the subject.

[0108] The creatinine sample level, albumin sample level, estimated glomerular filtration rate sample level, body mass index value, glucose sample level and / or glycated hemoglobin A1 fraction C sample level may be representative sample levels from a corresponding plurality of sample levels of creatinine, albumin, estimated glomerular filtration rate, body mass index, glucose and / or glycated hemoglobin A1 fraction C, such as a maximum sample level, a minimum sample level, an average sample level and / or a median sample level. In an exemplary embodiment, creatinine is a maximum creatinine sample level among a plurality of creatinine sample levels from the subject, albumin is a minimum albumin sample level among a plurality of albumin sample levels from the subject, eGFR is a minimum estimated glomerular filtration rate sample level among a plurality of estimated glomerular filtration rate sample levels from the subject, BMI is a minimum body mass index (BMI) value among a plurality of body mass index (BMI) values ​​from the subject, glucose is a minimum glucose sample level among a plurality of glucose sample levels from the subject, and HbA is an average sample level of the C fraction of glycated hemoglobin A1 among a plurality of C fraction of glycated hemoglobin A1 sample levels from the subject.

[0109] Constant c`'' CKD1 、c`'' CKD2 、c`'' CKD3 、c`'' CKD4 、c`'' CKD5 、c`'' CKD6 、c`'' CKD7 and c`'' CKD8 Can be a model specific constant. In an embodiment, the constant c`'' CKD1 、c`'' CKD2 、c`'' CKD3 、c`'' CKD4 、c`'' CKD5 、c`'' CKD6 、c`'' CKD7 and c`'' CKD8It can be a constant weighting factor associated with the corresponding marking parameter.

[0110] Constant c'' CKD1 、c'' CKD2 、c'' CKD3 、c'' CKD4 、c'' CKD5 、c'' CKD6 、c'' CKD7 and c'' CKD8 and c'' 死亡1 、c'' 死亡2 、c'' 死亡3 、c'' 死亡4 、c'' 死亡5 、c'' 死亡6 、c'' 死亡7 and c'' 死亡8 Can be a model specific constant. In an embodiment, the constant c'' CKD1 、c'' CKD2 、c'' CKD3 、c'' CKD4 、c'' CKD5 、c'' CKD6 、c'' CKD7 and c'' CKD8 and c'' 死亡1 、c'' 死亡2 、c'' 死亡3 、c'' 死亡4 、c'' 死亡5 、c'' 死亡6 、c'' 死亡7 and c'' 死亡8 It can be a constant weighting factor associated with the corresponding marking parameter.

[0111] For example, in such an embodiment, the constants may be as follows:

[0112] c'' CKD1 :0.02739 / year;

[0113] c'' CKD2 : 1.387 dl / mg;

[0114] c'' CKD3 : -0.3356 dl / g;

[0115] c'' CKD4 : -2.409;

[0116] c'' CKD5 : -0.02843 min∙1.73 m / ml;

[0117] c'' CKD6:0.01128 m 2 / kg;

[0118] c'' CKD7 : 0.0004946 dl / mg; and

[0119] c'' CKD8 :0.0893 / %;

[0120] c'' 死亡1 :0.06103 / year;

[0121] c'' 死亡2 :0.8194 dl / mg;

[0122] c'' 死亡3 : -0.9336 dl / g;

[0123] c'' 死亡4 : -4.557;

[0124] c'' 死亡5 :0.01654 min∙1.73m 2 / ml;

[0125] c'' 死亡6 :-0.0101 m 2 / kg;

[0126] c'' 死亡7 :0.0009107 dl / mg; and

[0127] c'' 死亡8 : 0.04368 / %.

[0128] In an embodiment, any or each of the constants may be selected from a range of + / - 30% around such respective value, preferably from a range of + / - 20%, and more preferably from a range of + / - 10%.

[0129] In an embodiment, for any or all of creatinine, albumin, eGFR, BMI, glucose, and HbA, a generalized value (creatinine 广义 ,albumin 广义 , eGFR 广义 , BMI 广义 ,glucose 广义 , HbA 广义 ) is used to replace the value of the subject. For example, the average value of the general population or the average value of the relevant sub-population can be used. As a generalized value, the average value of representative values ​​of the corresponding multiple values ​​from each population member, such as the average value of the corresponding maximum value, the corresponding minimum value, the corresponding average value and / or the corresponding median value can be used.

[0130] For example, in such an embodiment, the generalized values ​​may be as follows:

[0131] Creatinine 广义 : 1.055 mg / dl;

[0132] albumin 广义 : 3.835 g / dl;

[0133] eGFR 广义 :66.523 ml / min / 1.73m 2 ;

[0134] BMI 广义 :32.295 kg / m 2 ;

[0135] glucose 广义 : 129.691 mg / dl; and

[0136] HbA 广义 :7.607%.

[0137] In an embodiment, any or each generalized value may be selected from a range of + / - 30% around such respective value, preferably from a range of + / - 20%, and more preferably from a range of + / - 10%.

[0138] The method may further include determining a subject value recommendation and providing a recommendation output indicating the subject value recommendation. Determining the subject value recommendation may include determining a first marker parameter based on a weight of the marker parameter, receiving a generalized value for the first marker parameter, and the first marker parameter has a higher weight than a second marker parameter for which the generalized value is received, and determining the subject value recommendation as a recommendation to obtain a value for the first marker parameter for the subject. The recommendation output may indicate instructions to obtain a value for the first marker parameter for the subject and re-perform a method for screening a subject for risk of CKD, thereby providing marker data including the value for the first marker parameter for the subject.

[0139] The method may include only determining a subject value recommendation, and providing a recommendation output indicating the subject value recommendation if the accuracy value for determining the risk factor is below an accuracy threshold. The accuracy value for the risk factor may be determined based on which marker parameters the generalized values ​​are used for. In an embodiment, the accuracy value for the risk factor may be determined in comparison to a reference risk factor determined using subject values ​​for all or any marker parameters whose generalized values ​​were used when determining the risk factor.

[0140] Within the meaning of the present disclosure, screening a subject for risk of CKD means identifying a subject at risk of developing or having CKD.

[0141] In the sense of the present disclosure, a sample level is a level of a substance (such as creatinine or albumin) in a sample of a subject's bodily fluid. The sample level may be determined in the same or different samples. Alternatively or additionally, in order to determine the sample level, measurements may be performed in the same or different samples. For example, the sample level of a substance may be determined from multiple measurements of the same substance in the same sample, for example by determining an average value. In another example, at least one of the multiple sample levels of the same substance may be determined in a first sample, and at least another of the multiple sample levels of the same substance may be determined in a second sample. The sample level of a first substance and the sample level of a second substance may be determined in the same sample. Alternatively, the sample level of a first substance may be determined in a first sample, and the sample level of a second substance may be determined in a second sample.

[0142] A computer program product may be provided that includes a computer readable medium embodying program code executable by a processor of a computing device or system, which, when executed, causes the computing device or system to perform a computer-implemented method for screening a subject for risk of chronic kidney disease.

[0143] With respect to the computer implemented method, computer program product and further method for screening a subject for risk of chronic kidney disease, the alternative embodiments described above may apply mutatis mutandis.

[0144] In the computer-implemented method, the albumin sample level may be the albumin sample level in a bodily fluid sample, and the creatinine sample level may be the creatinine sample level in another bodily fluid.

[0145] In a computer-implemented method, the program may further cause the processor to perform generating output data indicative of the risk factor and outputting the output data to an output device of the data processing system. The output device may be any device suitable for outputting the output data, such as a display device of the data processing system, such as a monitor, and / or a transmitter device for transmission of wired and / or wireless data transmission. The output data may be output to a user, such as a physician. The output data may be output via a display of the data processing system.

[0146] The data processing system may include a plurality of data processing devices, each having a processor and a memory. The marker data may be provided in a first data processing device. For example, the marker data may be received in the first data processing device via an input device by a user input and / or by data transmission. The marker data may be sent from the first data processing device to a second data processing device, which may be located at a distance relative to the first data processing device. The marker data may be received in the second data processing device, and then the risk factor may be determined in the second data processing device. Result data indicating the risk factor may be sent from the second data processing device to the first data processing device, or alternatively or additionally, to a third data processing device. The result data may then be stored in the first and / or third data processing device and / or output via an output device of the first and / or third data processing device.

[0147] The first data processing device and / or the third data processing device may be a local device, such as a client computer, and the second data processing device may be a remote device, such as a remote server.

[0148] Alternatively, at least the functionality of the first data processing device and the second data processing device may be provided in the same data processing device, eg a computer, such as a computer in a doctor's office.All steps of the computer-implemented method may be performed in the same data processing device. BRIEF DESCRIPTION OF THE DRAWINGS

[0149] Further embodiments are described below by way of example. In the figures:

[0150] Figure 1 is the distribution of ages in the example teaching training set, validation set, and further validation set;

[0151] Figure 2 is the distribution of HbA1C in the example teaching training set, validation set and further validation set;

[0152] Figure 3 is a comparison of algorithms used to predict CKD;

[0153] Figure 4 is a comparison of algorithms using subgroups to predict CKD;

[0154] Figure 5 is another comparison of algorithms for predicting CKD; and

[0155] Figure 6 A further comparison of algorithms for predicting CKD was performed. DETAILED DESCRIPTION

[0156] Generally speaking, in any embodiment of the method for screening a subject for risk of CKD, creatinine max can be the maximum creatinine sample level among a plurality of creatinine sample levels from a subject, albumin min can be the minimum albumin sample level among multiple albumin sample levels from a subject, eGFR min can be the minimum estimated glomerular filtration rate sample level among a plurality of estimated glomerular filtration rate sample levels from the subject, BMI min can be the minimum BMI value among multiple BMI values ​​from the subject, glucose min may be the minimum glucose sample level among a plurality of glucose sample levels from a subject, and HbA mean It may be the average C-fraction sample level of glycated hemoglobin A1 from a plurality of C-fraction sample levels of glycated hemoglobin A1 from a subject. Such values ​​and / or sample levels may be determined from values ​​and / or sample levels for a subject that have been archived. Alternatively or additionally, values ​​and / or sample levels for a subject may be determined, particularly values ​​and / or sample levels for use with a method for screening a subject for the risk of CKD. Values ​​and / or sample levels may be real-world data, i.e., unlike clinical data, they may not be subject to restrictions regarding, for example, the integrity or accuracy of the data.

[0157] In the method for screening a subject for risk of CKD, creatinine max It can be expressed in mg / dl. min It can be expressed in g / dl, eGFR min It can be adjusted in ml / min / 1.73m 2 BMI is expressed in units of min Can be expressed in kg / m 2 Glucose min It can be expressed in mg / dl, and HbA 平均值 It can be expressed in %. The glomerular filtration rate can be estimated using the MDRD formula known in the art as such. Alternatively, the glomerular filtration rate can be estimated using the CKD-EPI formula known in the art as such.

[0158] The marker data may be received for a subject suffering from diabetes. Alternatively, the subject does not suffer from diabetes, but may be at risk of suffering from diabetes in the future. The marker data indicates the marker parameters age, creatinine, max and albumin min The parameter "age" indicates the age of the subject in years. The parameter "creatinine max” indicates the maximum creatinine sample level from multiple creatinine sample levels on file from the subject and collected from blood samples within the previous 2 years. The parameter “Albumin min ” indicates the lowest albumin sample level from multiple albumin sample levels on file from the subject and collected from blood samples within the previous 2 years.

[0159] According to this embodiment, the marker data indicates the marker parameters age, creatinine, max and albumin min , thereby providing a simplified method for calculating a risk factor indicative of a subject's risk of suffering from CKD. In a further embodiment, as will be described in more detail below, a marker parameter eGFR indicative of a subject may be included in the calculation min , BMI min ,glucose min and HbA mean to provide a more accurate calculation of the risk factor.

[0160] In an example, a risk factor indicative of a risk of the subject suffering from CKD is determined from the plurality of marker parameters according to the following equation:

[0161]

[0162] P CKD预测 = 0.02739∙age / year + 1.387∙creatinine max ∙dl / mg - 0.3356∙albumin min ∙dl / g- 3.1925

[0163] P 死亡预测 = 0.06103 ∙ age / year + 0.8194 ∙ creatinine max ∙dl / mg – 0.9336∙albumin min ∙dl / g- 3.3325.

[0164] Thus, the age value is weighted higher than the albumin sample level, and the creatinine sample level is weighted higher than the albumin sample level.

[0165] The marker data may be received for a subject suffering from diabetes. Alternatively, the subject does not suffer from diabetes but may be at risk of suffering from diabetes in the future. The marker data indicates the marker parameters age, creatinine, max ,albumin min and eGFR min The parameter "age" indicates the age of the subject in years. The parameter "creatinine max” indicates the maximum creatinine sample level from multiple creatinine sample levels on file from the subject and collected from blood samples within the previous 2 years. The parameter “Albumin min ” indicates the lowest albumin sample level from multiple albumin sample levels on file from the subject and collected from blood samples within the previous 2 years. The parameter “glomerular filtration rate min ” indicates the lowest estimated glomerular filtration rate sample level from multiple estimated glomerular filtration rate sample levels on file for the subject and collected within the previous 2 years.

[0166] In an example, a risk factor indicative of a subject's risk of suffering from CKD is determined from a plurality of marker parameters according to the following equation:

[0167]

[0168] P CKD预测 = 0.02739∙age / year + 1.387∙creatinine max ∙dl / mg - 0.3356∙albumin min ∙dl / g-0.02843∙eGFR min ∙Minimum value∙1.73m 2 / ml - 1.3013

[0169] P 死亡预测 = 0.06103 ∙ age / year + 0.8194 ∙ creatinine max ∙dl / mg – 0.9336∙albumin min ∙dl / g+ 0.01654∙eGFR min ∙Minimum value∙1.73m 2 / ml – 4.4328

[0170] Thus, the age value is weighted higher than the albumin sample level, and the creatinine sample level is weighted higher than the albumin sample level, and each of the age value, albumin sample level, and creatinine sample level is weighted higher than the glomerular filtration rate sample level.

[0171] The marker data may be received for a subject suffering from diabetes. Alternatively, the subject does not suffer from diabetes but may be at risk of suffering from diabetes in the future. The marker data indicates the marker parameters age, creatinine, max ,albumin min , eGFR min , BMI min ,glucose min and HbA mean The parameter "age" indicates the age of the subject in years. The parameter "creatinine max” indicates the maximum creatinine sample level from multiple creatinine sample levels on file from the subject and collected from blood samples within the previous 2 years. The parameter “Albumin min ” indicates the lowest albumin sample level from multiple albumin sample levels on file from the subject and collected from blood samples within the previous 2 years. The parameter “glomerular filtration rate min " indicates the lowest estimated glomerular filtration rate sample level from multiple estimated glomerular filtration rate sample levels on file from the subject and collected in the previous 2 years. The parameter "BMI min " indicates the minimum BMI value from multiple BMI values ​​on file for the subject and collected within the previous 2 years. The parameter "Glucose min " indicates the minimum blood glucose sample level from multiple blood glucose sample levels archived by the subject and collected within the previous 2 years. The parameter "HbA mean " indicates the average glycated hemoglobin A1 fraction C sample level from a plurality of glycated hemoglobin A1 fraction C sample levels that are archived from the subject and collected within the previous 2 years.

[0172] A risk factor indicative of a subject's risk of suffering from CKD is determined from the plurality of marker parameters according to the following equation:

[0173]

[0174] P CKD预测 = 0.02739∙age / year + 1.387∙creatinine max ∙dl / mg - 0.3356∙albumin min ∙dl / g-0.02843∙eGFR min ∙Minimum value∙1.73m 2 / ml + 0.01128∙BMI min + 0.0004946∙Glucose min ∙dl / mg +0.0893∙HbA mean / % - 2.409

[0175] P 死亡预测 = 0.06103 ∙ age / year + 0.8194 ∙ creatinine max ∙dl / mg – 0.9336∙albumin min ∙dl / g+ 0.01654∙eGFR min ∙Minimum value∙1.73m 2 / ml - 0.0101 BMI min + 0.0009107∙Glucose min ∙dl / mg +0.04368∙HbA mean / % - 4.557

[0176] Thus, the age value is weighted higher than the albumin sample level, the age is weighted higher than the creatinine sample level, the creatinine sample level is weighted higher than the albumin sample level, and each of the age value, albumin sample level, and creatinine sample level is weighted higher than the glomerular filtration rate sample level. Further, each of the age value, albumin sample level, creatinine sample level, and glomerular filtration rate sample level is weighted higher than each of the body mass index, blood glucose sample level, and C fraction sample level of glycated hemoglobin A1.

[0177] In the method for screening a subject for risk of CKD, in determining P CKD预测 and / or P 死亡预测 When , all or any of the values ​​multiplied by the subject's value and / or sample level can be determined as follows.

[0178] The algorithm is taught using electronic health record (EHR) data, for example, from 417,912 people with diabetes (type 1 and type 2) among more than 55 million people represented in the database. Data within a time window starting 2 years before the initial diagnosis of diabetes and continuing until 3 years after the diagnosis is retrieved. The data can be considered as real-world data (RWD), without general restrictions on, for example, the completeness or accuracy of the data. Before feature selection and teaching the algorithm, the missing data is imputed using the mean of the group. Logistic regression is selected for teaching, rather than black-box methods such as deep learning. This can allow medical interpretation of data-driven analysis. After teaching, an independent sample data set, for example, from 104,504 further individuals in the same database, is used for independent validation. In addition, the algorithm is applied to data from 82,912 people with type 2 diabetes included in a further database, for example.

[0179] In the analysis of the validation results, the ICD code can be used as the target variable for training as well as the CKD reference diagnosis. The definition of the target feature "CKD" can be based solely on the occurrence of the corresponding ICD code in the database. In order to maintain the RWD nature of the dataset, no additions or changes can be made to the database. Such ICD codes may include ICD-9 codes and ICD-10 codes, such as the following ICD codes: 250.40, 250.41, 250.42, 250.43, 585.1, 585.2, 585.3, 585.4, 585.5, 585.6, 585.9, 403.00, 403.01, 403.11, 403.90, 403.91, 404.0, 404.00, 404.01, 404.02, 404.03, 404.1, 404.10, 404.11, 404.12, 404.13, 404.9, 404.90, 404.91, 404.92, 4.9, N05.8, N08 and / or N25.9.

[0180] ICD-9 codes 250.40, 403.90, 585.3, 585.9 were likely the most abundant diagnoses in the corresponding time windows of the datasets, and they appeared in >5% of cases within each dataset.

[0181] In a further method for screening a subject for risk of CKD, after determining P CKD预测 and / or P 死亡预测 When , all or any of the values ​​multiplied by the subject's value and / or sample level can be determined as follows.

[0182] To allow early risk assessment of CKD, EHR data were extracted from a database that includes longitudinal data from more than 55 million patients with thousands of individual-specific characteristics. The survey data extracted from the database originated from 522,416 people newly diagnosed with diabetes. Data were retrieved within a time window starting 2 years before the initial diagnosis of diabetes and continuing until 3 years after that diagnosis. In order to perform an unbiased risk assessment for the later development of CKD, people with previous renal dysfunction were excluded. According to the guidelines for the diagnosis of diabetes, it is required to determine the concentration of the β-N-1-deoxyfructose component of hemoglobin (HbA1C), an important clinical laboratory parameter in the diagnosis and treatment of diabetes, at least once before (or within 7 days after) the initial diagnosis of diabetes. The data selected from the database can be considered RWD because no further restrictions are imposed on the completeness or accuracy of the data. In order to address these challenges caused by the use of RWD, the following methods can be implemented:

[0183] 1. The data selected from the database were randomly divided into a teaching set (417,912 people) and a validation set (104,504 people);

[0184] 2. Select features based on data-driven correlation analysis within the teaching set and cross-check the conceptual (especially medical) relevance;

[0185] 3. Impute missing values ​​using the mean of the data set. Optionally, screening or elimination of outliers has been performed prior to teaching. Where an outlier is identified, the value has been replaced with an appropriate value (if the feature value is above the upper limit of a specific allowed range for the feature, the value may be replaced by the upper limit of the range before the value is used in the prediction formula. If the feature value is below the lower limit of the specific allowed range for the feature, the value may be replaced by the lower limit of the range before the value is used in the prediction formula);

[0186] 4. Exclusively teach the risk predictor in the RWD teaching set;

[0187] 5. After the teaching is complete, apply the algorithm to the validation set in order to assess the quality of the algorithm. No further retraining of the algorithm is performed;

[0188] 6. In addition, RWD from 82,912 individuals represented in a further database was used as a further independent validation set.

[0189] A training set (from the IBM Explory database; see Kaelber, DC et al., “ 与静脉血栓栓塞事件相关的患者特征:一项 使用汇总电子健康记录数据的队列研究 ” ，J Am Med Inform Assoc 19, 965–972, 2012), validation set (from IBM Explorys database) and further validation set (from 印第安纳州患者护理网络 (INPC); see McDonald, CJ et al., “ 印第安纳州 患者护理网络:一个有效的本地健康信息基础设施 ”, Health Affairs 24, 1214-1220, 2005) 。 Logistic regression was applied in the teaching.

[0190] In the teaching and validation sets, females accounted for 50.7%, 50.9%, and 51.7% of the population, respectively. The median age of each population was 60, 60, and 59 years, respectively. The median HbA1C concentrations were 6.8%, 6.8%, and 6.6%, respectively. The distribution of age and HbA1C was Figure 1 and 2 Shown in.

[0191] In certain embodiments, for feature selection, almost 300 features are initially selected based on medical and data-driven criteria. The feature set is then picked in multiple steps. Observed features defined for less than half of the patients in the group are removed, and outliers for continuous features are also removed. Class features with ≥ 99% and continuous features with standard deviation ≤ 0.001% in a single category are not considered. Finally, only those features that have been shown to be relevant to CKD diagnosis in univariate analysis are retained, as determined by Pearson's chi-square coefficient. For predictive analysis, a logistic regression model based on forward selection was trained on the training set (see Bursac, Z. et al., “ 逻辑回归中变量的有目的选择 逻辑回归 ”, Source code for biology and medicine, March 17, 2018; and Hosmer Jr, DW et al. 应用逻辑回归 ”, Volume 398, John Wiley & Sons, 2013), and considers the person’s age, body mass index, glomerular filtration rate, and concentrations of glucose, albumin, and creatinine as the most important parameters. In contrast to “black-box” approaches based on, for example, deep learning, the medical relevance of these features can be assessed to ensure clinical applicability. HbA1 C can be added to the top 7 feature list to reflect the current state-of-the-art approach. The algorithm can be taught based on correlations, but no causal relationships may be inferred. After teaching, the algorithm is applied to two independent datasets, the validation sets.

[0192] ICD codes can be used as target variables for training and as reference CKD diagnoses in analysis of validation results. The definition of the target feature "CKD" can be based solely on the occurrence of the corresponding ICD code in the database. In order to maintain the RWD nature of the dataset, no additions or changes can be made to the database. Such ICD codes may include ICD-9 codes and ICD-10 codes, such as the following ICD codes: 250.40, 250.41, 250.42, 250.43, 585.1, 585.2, 585.3, 585.4, 585.5, 585.6, 585.9, 403.00, 403.01, 403.11, 403.90, 403.91, 404.0, 404.00, 404.01, 404.02, 404.03, 404.1, 404.10, 404.11, 404.12, 404.13, 404.9, 404.90, 404.91, 404.92, 4.9, N05.8, N08 and / or N25.9.

[0193] In an embodiment, ICD-9 codes 250.40, 403.90, 585.3, 585.9 are the most abundant diagnoses in the corresponding time windows of the datasets, and they appear in >5% of cases within each dataset.

[0194] The experimental data are discussed below.

[0195] Receiver Operating Characteristics (Compare to Swets, JA's " 测量诊断系统的准确性 系统 The area under the curve (AUC) is often used to measure the quality of clinical markers as well as machine learning algorithms (see Bradley, AP, “ 受试者工作特征曲线下面积的应用 机器学习算法评估中的曲线”, Pattern Recognition 30, 1145-1159, 1997). A perfect marker would achieve AUC = 1.0, while a coin toss would result in AUC = 0.5. After teaching a model (based on Explorys) according to the present disclosure using the seven most promising features, the AUC of the prediction algorithm totaled 0.7937 (0.790 … 0.797) when applied to the overall independent validation data (Explorys: 0.761, INPC: 0.831).

[0196] If the first 10 and first 12 features are used for evaluation, the AUC increases to 0.7939 and 0.7967, respectively. 糖尿病控制与并发症试验研究 小组。强化治疗糖尿病对胰岛素依赖型糖尿病长期并发症的发生和 进展的影响 糖尿病 ”, N Engl J Med 329, 977-986, 1993) for the same data set obtained 0.483 (0.477…0.489). Therefore, for persons newly diagnosed with diabetes, the algorithm according to the present disclosure outperforms the risk predictor of HbA1C alone.

[0197] In further analysis, the algorithm according to the present disclosure was compared with published algorithms derived from data from major clinical studies, such as the ONTARGET, ORIGIN, RENAAL and ADVANCE studies (cf. Dunkler, D. et al., “ 2型糖尿病早期慢性肾脏病的风险预测 ”, Clin J AmSoc Nephrol 10, 1371-1379, 2015; Vergouwe, Y. et al. 进展为 1型糖尿病微量白蛋白尿:预测规则的制定与验证 预测规则 ”, Diabetologia 53, 254-262, 2010; Keane, WF et al. 风险 2型糖尿病和 肾病患者预后预测评分:RENAAL研究 ”, Clin J Am Soc Nephrol 1, 761-767, 2006; and Jardine, MJ et al. 2型糖尿病患者肾脏相关结局的预测 2型糖尿病 ”, Am J Kidney Dis. 60, 770-778, 2012). Figure 3As shown in , for all RWD cohorts, the algorithm according to the present disclosure outperforms each of these algorithms. While this finding is important in terms of applicability and relevance in everyday settings, it can be argued that the validity of the published algorithm is limited to the inclusion and exclusion criteria of the corresponding clinical studies. Therefore, subcohorts of the IBM Explorys and INPC databases were formed according to the selection criteria of these studies, and the algorithm according to the present disclosure (without any retraining) was benchmarked against the literature algorithms for these subcohorts only. Although, as expected, the AUC of the disclosed algorithm increased for all specific subcohorts, the superiority of the RWD trained model according to the present disclosure prevailed ( Figure 4 ). However, for the current RWD set, the inclusion and exclusion criteria of the subgroups cannot be accurately met in all cases, because clinical studies require some information that is not available in the database (such as waist-to-hip ratio). In addition, there are differences in the selection of the time window for complication occurrence. Nevertheless, the features prioritized for classification according to the algorithm of the present disclosure are similar to those reported in the literature, thus further enhancing the effectiveness of the algorithm.

[0198] Compared to algorithms based on clinical studies, the use of RWD and in particular the inclusion of incomplete or erroneous data in the training set of the algorithm according to the present disclosure constitutes a major difference. Imputation of missing data provides a typical example for predictive analysis in RWD cohorts, while imputation is unthinkable in a clinical research setting. To further clarify the role of imputation, the algorithm according to the present disclosure was applied to RWD that only represents individuals for which a complete information set is provided (i.e., imputation is not necessary). In this case, the AUC of the overall RWD set remained comparable to previous values, i.e., 0.792 (0.787…0.797), 0.791 (0.780…0.801), and 0.809 (0.769…0.846) for the Explorys teaching training set, Explorys validation set, and INPC validation set, respectively. Further analysis showed that when testing the early algorithms, the classification accuracy was rapidly lost as the fraction of imputed data increased, while the algorithm according to the present disclosure achieved much higher stability even for a higher proportion of imputed data ( Figure 5 ). It is concluded that, at least in this example, using RWD to teach predictive analytics algorithms can achieve equivalent or even improved accuracy compared to clinical trial data, but further testing on additional datasets is necessary before these conclusions can be generalized.

[0199] In summary, it was demonstrated that the prediction algorithm for CKD performs significantly better for individuals newly diagnosed with diabetes if trained on RWD rather than clinical study data. This statement holds true when the algorithm according to the present disclosure is applied to the overall RWD cohort as well as to specific sub-cohorts defined by the corresponding clinical studies. The results support the path to high-quality prediction models that can be applied in clinical settings, thereby enabling a shift toward personalized and outcome-based healthcare.

[0200] The performance of a method for screening subjects for risk of CKD or for identifying those at high risk of developing CKD can be judged in terms of sensitivity (the fraction of high-risk patients that are correctly predicted) and specificity (the fraction of low-risk patients that are correctly assigned). However, simply by changing the threshold between high risk and low risk, one of these numbers can be improved at the expense of the other. Thus, the data pair of sensitivity and specificity can be graphically plotted in the form of a so-called receiver operating characteristic (ROC) curve (see Swets, JA, “ 测量 诊断系统的准确性 ”, Science 240, 1285-1293, 1988), where sensitivity is plotted as a function of 1-specificity (which corresponds to the fraction of people who are misclassified as high risk). Figure 6 The Explorys training set, Explorys validation set, and INPC validation set in FIG. 3 show the ROC curves of the risk model according to the present disclosure, as well as the corresponding ROC curves of the model based only on HbA1C.

[0201] For a perfect classifier, the ROC curve reaches the upper left corner. In fact, the threshold corresponding to the data pair closest to this corner is called the "optimal threshold". When high sensitivity is the goal, alternative thresholds can be chosen to guarantee, for example, 90% sensitivity. The corresponding results are summarized in the table below together with the positive predictive value (PPV) and the negative predictive value (NPV). Similar metrics from the field of bioinformatics - namely accuracy and F score (Van Rijsbergen, CJ " 信息 检索,巴特沃思-海涅曼牛顿 ”, MA, USA, 1979) – supplements the list of examples in Table 2.

[0202] Table 2

[0203] .

[0204] The full algorithm according to the present disclosure (seven values ​​per sample level for the subject, missing values ​​per imputed sample level) was compared with the reduced algorithm according to the present disclosure (age, creatinine, max and albumin min, residual value / population mean at sample level) applied to the INPC data, respectively, has resulted in an AUC of 0.831 (CI 0.827 to 0.836) for the full algorithm and 0.823 (CI 0.818 to 0.827) for the reduced algorithm. Thus, even with the reduced algorithm, useful predictions can be achieved.

Claims

1. A method for screening a subject for risk of chronic kidney disease (CKD), comprising: - receiving marker data indicative of a plurality of marker parameters of a subject, such plurality of marker parameters indicative of a value of age, a sample creatinine level, and a sample albumin level of the subject during a measurement period; as well as - determining a risk factor indicative of a risk of the subject suffering from CKD from the plurality of marker parameters, wherein the determining comprises - give age values ​​a higher weight than albumin levels, and - giving higher weight to creatinine sample levels compared to albumin sample levels; The risk factor is determined according to the following equation And among them -P CKD is a risk factor; -P CKD预测 =c CKD1 Age + c CKD2 Creatinine + c CKD3 Albumin + c CKD4 ; -Age is the age of the subject; -creatinine is the subject's sample creatinine level; - Albumin is the subject's sample albumin level; and -c CKD1 、c CKD2 、c CKD3 and c CKD4 is a constant. 2 . The method of claim 1 , further comprising a plurality of marker parameters indicative of a subject's creatinine blood sample level.

3. The method according to claim 1 or 2, further comprising a plurality of marker parameters indicating the albumin blood sample level of the subject.

4. The method according to any one of the preceding claims 1-2, wherein the subject is a diabetic patient.

5. The method according to any of the preceding claims 1-2, wherein the measurement period is limited to two years.

6. The method according to any one of the preceding claims 1-2, wherein the subject has not been diagnosed with diabetes at the end of the measurement period. The method of claim 4 , wherein the measurement period is at least partially subsequent to a diagnosis of diabetes in the subject.

8. The method according to any one of the preceding claims 1-2, wherein the risk factor indicates the risk of the subject suffering from CKD within a predicted time period of three years from the end of the measurement period.

9. The method according to any of the preceding claims 1-2, wherein the determining further comprises giving a higher weight to age than to the sample level of creatinine.

10. The method according to any of the preceding claims 1-2, wherein the receiving comprises receiving marker data indicating a plurality of marker parameters for a subject having a sample level of HbA1c below 6.5%.

11. The method according to any one of the preceding claims 1-2, further comprising: - a plurality of marker parameters indicative of the subject's sample glomerular filtration rate level; as well as - in the determining, each of the age value, the albumin sample level and the creatinine sample level is weighted higher than the glomerular filtration rate sample level.

12. The method of claim 1, wherein the risk factor is further determined according to the following equation And among them -P' CKD is a risk factor; -P' CKD预测 =c' CKD1 Age + c' CKD2 Creatinine + c' CKD3 ·Albumin+c' CKD4 +c' CKD5 eGFR -Age is the age of the subject; -creatinine is the subject's sample creatinine level; -Albumin is the subject's sample albumin level; -eGFR is the subject's estimated glomerular filtration rate sample level; and -c' CKD1 、c' CKD2 、c' CKD3 、c' CKD4 and c' CKD5 is a constant.

13. A computer-implemented method for screening a subject for risk of chronic kidney disease (CKD) in a data processing system having a processor and a non-transitory memory storing a program, the program causing the processor to: - receiving marker data indicative of a plurality of marker parameters of a subject, such plurality of marker parameters indicative of age values, albumin sample levels, and creatinine sample levels of the subject over a measurement period; as well as - determining a risk factor indicative of a risk of the subject suffering from CKD from the plurality of marker parameters, wherein the determining comprises - give age values ​​a higher weight than albumin levels, and - giving higher weight to creatinine sample levels compared to albumin sample levels; The risk factor is determined according to the following equation And among them -P CKD is a risk factor; -P CKD预测 =c CKD1 Age + c CKD2 Creatinine + c CKD3 Albumin + c CKD4 ; -Age is the age of the subject; -creatinine is the subject's sample creatinine level; - Albumin is the subject's sample albumin level; and -c CKD1 、c CKD2 、c CKD3 and c CKD4 is a constant.

14. A method for screening a subject for risk of chronic kidney disease (CKD), comprising: - receiving tag data indicating a plurality of tag parameters, such plurality of tag parameters indicating - age value of the subject, - the sample creatinine level during the measurement period, and - Albumin sample levels during the measurement period; as well as - determining a risk factor indicative of a risk of the subject suffering from CKD from the plurality of marker parameters, wherein the determining comprises - give age values ​​a higher weight than albumin levels, and - Give higher weight to creatinine sample levels compared to albumin sample levels wherein at least one of the creatinine sample level and the albumin sample level indicates a generalized value of the sample level of a reference subject group excluding the subject during a corresponding measurement period for each subject of the reference subject group; The risk factor is determined according to the following equation And among them -P CKD is a risk factor; -P CKD预测 =c CKD1 Age + c CKD2 Creatinine + c CKD3 Albumin + c CKD4 ; -Age is the age of the subject; -creatinine is the subject's sample creatinine level; - Albumin is the subject's sample albumin level; and -c CKD1 、c CKD2 、c CKD3 and c CKD4 is a constant.

Images