Methods, computer-implemented methods, systems, and computer program products for screening subjects for risk of chronic kidney disease - Patents.com

Abstract

A method for screening a subject for risk of chronic kidney disease (CKD) is provided, comprising receiving marker data indicative of a plurality of marker parameters for the subject, the plurality of marker parameters indicative of at least an age value, a time since diagnosis value indicative of a time since diabetes diagnosis for the subject, a sample level of creatinine, an estimated glomerular filtration rate, a sample level of albumin, and a sample level of blood urea nitrogen, and determining a risk factor indicative of the risk of the subject suffering from CKD from the plurality of marker parameters. Further provided is a computer-implemented method for screening a subject for risk of CKD. Also provided is a system comprising a processor and a non-transitory memory storing a program that causes the processor to execute a method for screening a subject for risk of CKD. In another aspect, a computer program product is provided, the computer program product including instructions, the instructions causing the computer to execute a method for screening a subject for risk of CKD when the program is executed by a computer.

Description

[Technical field]

[0001] The present invention relates to methods, computer-implemented methods, systems, and computer program products for screening subjects for risk of chronic kidney disease. [Background technology]

[0002] Chronic kidney disease (CKD) involves a gradual loss of kidney function, beginning with a decline in glomerular filtration rate and / or albuminuria, and progressing to end-stage renal disease. As a result, dialysis or kidney transplantation may be required (see Unger, J., Schwartz, Z., Diabetes Management in Primary Care, 2nd edition. Lippincott Williams & Wilkens, Philadelphia, USA, 2013). CKD is a serious problem, with an adjusted prevalence of 7% in 2013 (Glassock, RJ et al., The global burden of chronic kidney disease: estimates, variability and pitfalls, Nat Rev Nephrol 13, 104-114, 2017). Early recognition of CKD may slow its progression, prevent complications, and reduce cardiovascular-related outcomes (Platinga, LC, et al., Awareness of chronic kidney disease among patients and providers, Adv Chronic Kidney Dis 17, 225-236, 2010). CKD may be a long-term microvascular complication of diabetes (Fioretto, P., et al., Residual microvascular risk in diabetes: unmet needs and future directions, Nat Rev Endocrinol 6, 19-25, 2010).

[0003] Algorithms for predicting the risk of CKD in diabetic patients have been described, for example, by Dunkler et al. (Dunkler, D. et al., Risk Prediction for Early CKD in Type 2 Diabetes, Clin J Am Soc Nephrol 10, 1371-1379, 2015), Vergouwe et al. (Vergouwe, Y. et al., Progression to microalbuminuria in type 1 diabetes: development and validation of a prediction rule, Diabetologia 53, 254-262, 2010), Keane et al. (Keane, WF et al., Risk Scores for Predicting Outcomes in Patients with Type 2 Diabetes and Nephropathy: The RENAAL Study, Clin J Am Soc Nephrol 1, 761-767, 2006) and Jardine et al. (Jardine, MJ et al., Prediction of Kidney-Related Outcomes in Patients With Such published algorithms are derived from data derived from primary clinical studies.

[0004] Such predictive models based on clinical data represent ideal settings with preselected populations, cross-checked and validated clinical data inputs, and often narrow observation time windows. Thus, results, when extrapolated from clinical studies, do not necessarily reveal optimal pathways in terms of efficacy and effectiveness in real-world populations. Moreover, most literature focuses on the progression of diabetic nephropathy or CKD, thus missing the early stages of this diabetic complication. Finally, patients are usually selected based on a complete set of respective characteristics.

[0005] EP 3 543 702 A1 discloses a method for screening a subject for risk of chronic kidney disease, comprising receiving marker data indicative of a plurality of marker parameters for the subject, the plurality of marker parameters indicative of an age value, a sample level of creatinine, and a sample level of albumin for the subject for a measurement period, and determining a risk factor indicative of the risk of the subject suffering from CKD from the plurality of marker parameters. The determining comprises weighting the age value higher than the sample level of albumin, and weighting the sample level of creatinine higher than the sample level of albumin. A computer-implemented method is disclosed that applies a logistic regression (LR) model to determine the risk of chronic kidney disease for a subject, such as a patient with a diabetes diagnosis.

[0006] A method for long-term risk prediction of CKD in diabetic patients has been published by Song et al. (Song et al., Longitudinal risk prediction of chronic kidney disease in diabetic patients using a temporal-enhanced gradient boosting machine: retrospective cohort study, JMIR Med Inform 8(1), 2020, e15510). A boosted machine learning model has been proposed to predict the risk of CKD. Summary of the Invention

[0007] It is an object of the present invention to provide an improved method for screening subjects for risk of chronic kidney disease, which allows a reliable risk assessment of CKD based on real world data (RWD).

[0008] To address this, there is provided a method for screening a subject for risk of chronic kidney disease (CKD) according to independent claim 1. Furthermore, there is provided a computer implemented method according to independent claim 8. Also provided is a system and a computer program product according to claims 13 and 14, respectively. Further embodiments are disclosed in the dependent claims.

[0009] According to one embodiment, a method for screening a subject for risk of chronic kidney disease (CKD) is provided, comprising receiving marker data indicative of a plurality of marker parameters for the subject, the plurality of marker parameters indicative of at least the following: age value, time since diagnosis value indicative of time since diabetes diagnosis for the subject, sample level of creatinine, estimated glomerular filtration rate, sample level of albumin, and sample level of blood urea nitrogen.A risk factor indicative of risk of suffering from CKD is determined for the subject from the plurality of marker parameters.

[0010] According to another aspect, there is provided a computer-implemented method for screening a subject for risk of chronic kidney disease (CKD) in a data processing system, the data processing system having a processor and a non-transitory memory storing a program that causes the processor to receive marker data indicative of a plurality of marker parameters for the subject, the plurality of marker parameters indicative of an age value, a time since diagnosis value indicative of a time since diabetes diagnosis for the subject, a sample level of creatinine, a sample level of estimated glomerular filtration rate, a sample level of albumin, and a sample level of blood urea nitrogen, and determine a risk factor indicative of the risk of the subject suffering from CKD from the plurality of marker parameters.

[0011] A system is provided that includes a processor and a non-transitory memory that stores a program that causes the processor to execute a method for screening a subject for risk of chronic kidney disease (CKD).

[0012] A computer program or computer program product is provided, the computer program or computer program product comprising instructions, which when executed by a computer, cause the computer to perform steps of a method for screening a subject for risk of chronic kidney disease (CKD).

[0013] Marker parameters may, for example, represent real-world data that (unlike clinical data) is not limited with respect to data completeness or accuracy.

[0014] Time from diagnosis refers to the time from the date and time of initial diagnosis of diabetes for a subject to the date of assessment of a risk factor for the subject.

[0015] The method may further include a plurality of marker parameters indicative of a blood sample level of creatinine for the subject. As the sample level of creatinine, a serum sample level or a plasma sample may also be used. Thus, it may be avoided to require a sample level of creatinine as a concentration in urine. The plurality of marker parameters may indicate a selected blood sample level of creatinine (or a serum or plasma sample level, respectively) for the subject, selected from a plurality of blood sample levels of creatinine (or a serum or plasma sample level, respectively). For example, the selected blood sample level of creatinine may be a maximum value from a plurality of blood sample levels of creatinine. Alternatively or additionally, the plurality of marker parameters may indicate a calculated blood sample level of creatinine for the subject, calculated from a plurality of blood sample levels of creatinine. For example, the calculated blood sample level of creatinine may be a statistical value, such as a mean, a median, or a mode, calculated from a plurality of blood sample levels of creatinine. The sample level of creatinine may be provided in units of mg / dl (such as milligrams of creatinine per deciliter of blood).

[0016] The method may further include a plurality of marker parameters indicative of at least one of a blood sample level of albumin and a urine sample level of albumin for the subject. In one embodiment, the sample level of albumin is a blood sample level. As the sample level of albumin, a serum sample level or a plasma sample may also be used. The plurality of marker parameters may also indicative of a selected blood sample level of albumin (or a serum or plasma sample level, respectively) for the subject, selected from a plurality of blood sample levels of albumin (or a serum or plasma sample level, respectively). For example, the selected blood sample level of albumin may be a minimum value from a plurality of blood sample levels of albumin. Alternatively or additionally, the plurality of marker parameters may indicative of a calculated blood sample level of albumin for the subject, calculated from a plurality of blood sample levels of albumin. For example, the calculated blood sample level of albumin may be a statistical value, such as a mean value, a median value, or a mode value, calculated from a plurality of blood sample levels of albumin. Albumin sample levels may be provided in units of mg / dl (eg, milligrams of albumin per deciliter of blood).

[0017] Glomerular filtration rate is known in the art to indicate the flow rate of filtered fluid through the kidney. It is an important index for estimating renal function. Glomerular filtration rate may decrease due to renal disease. In an embodiment, glomerular filtration rate may be estimated using the Modification of Diet in Renal Disease (MDRD) formula known in the art. For example, the MDRD formula, which uses four variables, depends on the subject's age, sex, ethnicity and serum creatinine to estimate glomerular filtration rate. In an alternative embodiment, glomerular filtration rate may be estimated using the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) formula, which is known in the art per se. The CKD-EPI formula depends on the subject's age, sex, ethnicity and serum creatinine to estimate glomerular filtration rate. In a further embodiment, glomerular filtration rate may be estimated using other methods or may be determined directly. Estimated glomerular filtration rate may be provided in units of ml / min / 1.73m2 (milliliters / min per 1.73 square meters of body surface area).

[0018] The plurality of marker parameters may represent a selected estimated glomerular filtration rate selected from a plurality of estimated glomerular filtration rates for a subject. For example, the selected estimated glomerular filtration rate may be the minimum value from the plurality of estimated glomerular filtration rates. Alternatively or additionally, the plurality of marker parameters may represent a statistical value as a rated estimated glomerular filtration calculated from a plurality of estimated glomerular filtration rates, such as the mean, median or mode, for a subject.

[0019] The blood urea nitrogen (BUN) sample level may be provided in units of mg / dl (e.g., milligrams of urea nitrogen per deciliter of blood). Thus, the blood urea nitrogen (BUN) sample level may represent the mass of nitrogen in urea / volume of blood sample, rather than the mass of total urea. The plurality of marker parameters may represent a selected blood urea nitrogen sample level selected from a plurality of blood sample levels of urea nitrogen for the subject. For example, the selected urea nitrogen blood sample level may be a minimum value from a plurality of urea nitrogen blood sample levels. Alternatively or additionally, the plurality of marker parameters may represent a calculated blood sample level of urea nitrogen calculated from a plurality of blood sample levels of urea nitrogen for the subject. For example, the calculated urea nitrogen blood sample level may be a statistical value calculated from a plurality of blood sample levels of urea nitrogen, such as a mean, median, or mode.

[0020] The sample level of creatinine, the sample level of albumin, the sample level of blood urea nitrogen and / or the estimated glomerular filtration rate may be a representative sample level and / or rate from a plurality of sample levels and / or rates, e.g., a maximum sample level and / or rate, a minimum sample level and / or rate, an average sample level and / or rate and / or a median sample level and / or rate, respectively. In an exemplary embodiment, creatinine is the maximum sample level of creatinine from a plurality of sample levels of creatinine for the subject, albumin is the minimum sample level of albumin from a plurality of sample levels of albumin for the subject, eGFR is the minimum estimated glomerular filtration rate from a plurality of estimated glomerular filtration rates for the subject, and blood urea nitrogen is the minimum sample level of blood urea nitrogen from a plurality of sample levels of blood urea nitrogen for the subject.

[0021] Marker data can be derived from a measurement period of 2 years or less than 2 years.Accordingly, the measurement period can be limited to 2 years.Therefore, the substance values ​​and / or sample levels collected within a period of up to 2 years can be provided, in which risk factors indicate the risk of the subject suffering from CKD from the end of the measurement period.

[0022] In one embodiment, at least the sample level of creatinine, the sample level of albumin, the sample level of blood urea nitrogen, and the estimated glomerular filtration rate are from a measurement period of 2 years or less. The samples for determining the sample level of creatinine, the sample level of albumin, the sample level of blood urea nitrogen, and the estimated glomerular filtration rate may have been taken and / or determined during a measurement period of 2 years or less.

[0023] The age value may correspond to the age (eg, in years) of the patient at the time the risk factor is determined.

[0024] The time since diagnosis value may indicate the time since diabetes diagnosis for the subject when the risk factor is determined. In one embodiment, the date of determining the risk factor for the subject may be defined as the end of the measurement period.

[0025] The risk factors indicative of the risk of a subject suffering from CKD are determined from a number of marker parameters including at least a subject's age value, a time since diagnosis value indicative of the time since diabetes diagnosis for the subject, a sample level of creatinine from the subject, an estimated glomerular filtration rate from the subject, a sample level of albumin from the subject, and a sample level of blood urea nitrogen from the subject.

[0026] The risk factor can indicate the risk of the subject suffering from CKD within a predicted period of 3 years from the end of the measurement period.The risk factor can be the probability of the subject developing CKD within 3 years from the time when the sample level is determined.Alternatively, the risk factor can indicate the risk of the subject suffering from CKD within a period of less than 3 years, for example within 2 years, from the end of the measurement period.As a further alternative, the risk factor can indicate the risk of the subject suffering from CKD within a period of more than 3 years from the end of the measurement period.

[0027] With respect to the computer-implemented method, determining the risk factor may include providing a machine learning model, providing input data indicative of a plurality of marker parameters to the machine learning model, and determining the risk factor by the machine learning model. Thus, the risk factor is determined by applying a machine learning model that has been previously trained and tested (validated), and such training / testing includes training a machine learning algorithm to create or determine a machine learning model that is a result of such training, including training and testing / validation.

[0028] Providing the machine learning model may include providing an XGBoost machine learning model. XGBoost provides a decision tree-based ensemble machine learning algorithm that uses a gradient boosting framework. Gradient boosting is a machine learning technique for regression and classification problems that produces a predictive model in the form of a weak predictive model, typically an ensemble of decision trees. When decision trees are the weak learners, the resulting algorithm is called a gradient boosted tree and typically outperforms random forests. It builds models incrementally as other boosting methods do and generalizes them by allowing optimization of any discriminable loss function.

[0029] Providing a machine learning model may include, in preprocessing, providing a set of training data for a population of subjects, the training data indicating a plurality of training parameters for the population of subjects, the training marker parameters including age, creatinine level, estimated glomerular filtration rate level, albumin level, blood urea nitrogen level, and an indication of whether the subject has developed CKD; providing diabetes diagnosis data indicating a time or date when a diabetes diagnosis was determined for a subject from the population of subjects; determining supplemental training data from the diabetes diagnosis data indicating a time since diagnosis parameter indicating a time since a diabetes diagnosis was determined for a subject from the population of subjects; providing an expanded set of training data including the set of training data and the supplemental training data; and training a machine learning model, such as an XGBoost machine learning model, based on the expanded set of training data. The additional parameters referring to the time since the diagnosis parameter was determined are determined in preprocessing to expand the size and number of training data applied to train the machine learning model.

[0030] Pre-processing may also include determining a set of pre-processed training data from training data, comprising one or more statistical values ​​and / or selected values ​​for one or more of the levels of creatinine, estimated glomerular filtration rate, albumin level, and blood urea nitrogen level for each one or more subjects from a population of subjects.For example, a plurality of sample levels of creatinine may be determined for the subjects of a population.Thus, in pre-processing, one or more statistical values ​​and / or selected values ​​may be determined from a plurality of sample levels of creatinine, such as the average creatinine value and / or maximum creatinine value of the subject from the population.

[0031] Also provided herein is a method for training a machine learning model for determining a risk factor indicative of a test subject's risk of suffering from CKD from a plurality of marker parameters of the test subject. The training method comprises: providing a set of training data for a population of training subjects, the training data indicative of a plurality of training parameters for the population of training subjects, the training marker parameters including at least age, a level of creatinine, a level of estimated glomerular filtration rate, a level of albumin, and a level of blood urea nitrogen, the training data further including, for each training subject, an indication of whether the training subject has developed CKD; Optionally, determining from the training data a set of pre-processed training data including one or more statistics and / or selected values ​​for one or more of creatinine level, estimated glomerular filtration rate, albumin level, and blood urea nitrogen level for each training subject from the population of training subjects; providing diabetes diagnosis data indicating the time or date when a diabetes diagnosis was determined for each training subject from the population of training subjects; determining, from the diabetes diagnosis data, supplemental training data indicative of a time since diagnosis parameter indicative of a time since a diabetes diagnosis was determined for a subject from the population of subjects; an augmented set of training data, a set of training data and / or a set of preprocessed training data; and Supplemental Training Data providing an expanded set of training data, including training a machine learning model based on the expanded set of training data to determine risk factors indicative of the test subject's risk of developing CKD; Includes.

[0032] 1. A method for screening a test subject for risk of chronic kidney disease (CKD), comprising: training a machine learning model according to the training method as described above, thereby obtaining a trained machine learning model; receiving marker data indicative of a plurality of marker parameters for the subject, the plurality of marker parameters indicative of at least an age value, a time since diagnosis value indicative of a time since diabetes diagnosis for the subject, a sample level of creatinine, an estimated glomerular filtration rate, a sample level of albumin, and a sample level of blood urea nitrogen; Using the trained machine learning model, risk factors indicative of the subject's risk of developing CKD are determined from the multiple marker parameters. Further provided herein is a method comprising:

[0033] Risk factors may be determined using a machine learning model in which no marker data is imputed, and thus the machine learning model was trained / tested by training (and testing or validation) on data without imputed marker data.

[0034] Within the meaning of this disclosure, screening a subject for risk of CKD means identifying a subject at risk of developing or having CKD.

[0035] A sample level in the sense of the present disclosure 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, measurements may be performed in the same or different samples to determine the sample level. 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 one of the multiple sample levels of the same substance may be determined in a second sample. The sample level of the first substance and the sample level of the second substance may be determined in the same sample. Alternatively, the sample level of the first substance may be determined in a first sample, and the sample level of the second substance may be determined in a second sample.

[0036] A computer program product may be provided that includes a computer readable medium embodied with program code executable by a process of a computing device or system, the program code, when executed, causing the computing device or system to perform a computer-implemented method for screening a subject for risk of chronic kidney disease.

[0037] With regard to the computer-implemented method, the alternative embodiments described above may be applied mutatis mutandis.

[0038] In the computer-implemented method, the program may further cause the processor to generate output data indicative of the risk factors and output the output data to an output device of the data processing system. The output device may be any device suitable for outputting output data, for example, a display device of the data processing system such as a monitor, and / or a transmitting device for transmitting for wired and / or wireless data transmission. The output data may be output to a user, for example, a physician, via a display of the data processing system. Based on the output data indicative of the risk factors, further marker data may be requested from the subject, and / or a future date for further screening of the subject for CKD may be set (for example, based on at least one or more newly collected sample levels of creatinine, albumin, blood urea nitrogen, and / or a newly determined estimated glomerular filtration rate, a new age value, a new time from diagnosis value indicative of the time from diabetes diagnosis for the subject taking into account the future date).

[0039] The data processing system may comprise a plurality of data processing devices, each of which has a processor and a memory. The marker data may be provided to a first data processing device. For example, the marker data may be received at the first data processing device by user input via an input device and / or by data transfer. The marker data may be transmitted from the first data processing device to a second data processing device, which may be located remotely with respect to the first data processing device. The marker data may be received at the second data processing device, and then the risk factors may be determined at the second data processing device. The result data indicative of the risk factors may be transmitted 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.

[0040] 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.

[0041] Alternatively, the functionality of at least the first and second data processing devices may be provided in the same data processing device, e.g. a computer such as a doctor's office computer. All steps of the computer-implemented method may be performed in the same data processing device.

[0042] Description of Further Embodiments An embodiment will now be described by way of example with reference to the drawings. [Brief description of the drawings]

[0043] [Figure 1] FIG. 1 is a schematic diagram for determining an XGBoost machine learning model. [Diagram 2]FIG. 1 is a schematic diagram of a computer-implemented method for determining risk factors indicative of risk of CKD for a subject. [Diagram 3] ROC curves for the “Full XG Boosted Model”, the “Top 20 XG Boosted Model”, and the “LR Top 20 Model” using all parameters and only a limited number of parameters. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0044] 1 shows a schematic diagram for determining or creating a machine learning model by training and testing / validating a machine learning algorithm, which in one example is implemented by an XGBoost machine learning model. A dataset for a population of subjects is provided in step 10.

[0045] The machine learning model is created using, for example, Electronic Health Record (EHR) data from hundreds of thousands of people with diabetes (type 1 or type 2) represented in a database. The data is obtained for a time window after the initial diagnosis of diabetes. The data can be considered real-world data (RWD), e.g., typical restrictions on the completeness or accuracy of the data are not applied.

[0046] For teaching or learning the model (training and testing / validation), missing data are not imputed. The XGBoost machine learning process is applied.

[0047] In one example, the dataset was provided by the so-called IBM Explorys database (see Kaelber, DC et al., Patient characteristics associated with venous thromboembolic events: a cohort study using pooled electronic health record data, J Am Med Inform Assoc 19, 965-972, 2012). An alternative example of a population dataset of interest is the Indiana Network for Patient Care (INPC) database (see McDonald, CJ et al., The Indiana Network for Patient Care: a working local health information infrastructure, Health Affairs 24, 1214-1220, 2005).

[0048] The database provides an indication of the date of diabetes diagnosis for the subjects. Starting from such information, new parameters are established, which provide an indication of the period since diabetes diagnosis for each subject. A pre-processing step is applied to determine such additional parameters from the diabetes diagnosis data provided in the database. It provides supplemental training data indicative of the time since diagnosis parameter indicative of the time since the diabetes diagnosis was determined for the subjects from the population of subjects. Thus, there is an expanded set of training data further including the supplemental training data.

[0049] From the data set including supplemental training data indicative of the time from the diagnostic parameter, a set of training data and a set of test / validation data are determined (steps 11, 12). The set of training data is indicative of a plurality of parameters for the subject population (step 11). For a data set provided for the subject population, the set of training data may include training data indicative of (almost) all parameters for which data is provided in the data set of the subject population. Alternatively, a subset of parameters may be selected for training the machine learning model.

[0050] Then, in step 13, there is a training process for a machine learning model based on the set of training data. In one example, in the training process, XBoost training is applied to determine or create an XBoost machine learning model. The machine learning model is finally determined in step 14 by applying the set of test / validation data for final model evaluation.

[0051] FIG. 2 shows a schematic diagram of a computer-implemented method for determining risk factors indicative of a subject's risk of chronic kidney disease (CKD). In step 20, marker data indicative of a plurality of marker parameters for a subject for which risk factors are to be determined is provided. In one example, the plurality of marker parameters indicative of an age value, a time since diagnosis value indicative of a time since diabetes diagnosis for the subject, a sample level of creatinine, an estimated glomerular filtration rate (eGFR), a sample level of albumin, and a sample level of blood urea nitrogen (BUN). The marker data is provided as input to a machine learning model (step 21). By applying the machine learning model, a risk factor of chronic kidney disease risk for the subject is determined (step 22). The machine learning model is implemented by a software application on a data processing device having a processor and a memory.

[0052] Generally, in any of the embodiments of the method for screening a subject for risk of CKD, creatinine maxcan be a maximum sample level of creatinine from multiple sample levels of creatinine for the subject, and albumin min can be a minimum sample level of albumin from multiple sample levels of albumin for the subject, and eGFR min may be the minimum estimated glomerular filtration rate from multiple estimated glomerular filtration rates for the subject, and BUN min can be the minimum blood sample level of urea nitrogen. Such value and / or sample level can be determined from the value and / or sample level already registered for the subject. Alternatively or additionally, the value and / or sample level can be determined for the subject, specifically for use by the method for screening the subject for the risk of CKD. The value and / or sample level can be real data, i.e., unlike clinical data, it can be not limited, for example, in terms of completeness or accuracy of data.

[0053] The ICD codes may be used as target variables for training as well as CKD reference diagnoses in the analysis of validation results. The definition of the target feature "CKD" may be based solely on the occurrence of the respective ICD codes in the database. No additions or changes may be made to the database to maintain the RWD nature of the dataset. Such ICD codes may include ICD-9 codes and ICD-10 codes, for example, 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, 404.93, 581.81, 581.9, 583.89, 588.9, E10.2, E10.21, E10.22, E10.29, E11.2, E11.21, E11.22, E11.29, N17.0, N17.1, N17.2, N17.8, N17.9, N18.1, N18.2, N18.3, N18.4, N18.5, N18.6, N18.9, N19, 112.0, 112.9, 113, 113.0, 113.1, 113.10, 113.11, 113.2, N04.9, N05.8, N08 and / or N25.9.

[0054] In an embodiment, ICD-9 codes 250.40, 403.90, 585.3, 585.9 are the most abundant diagnoses in the respective time windows of data.

[0055] ICD codes may also be used to determine a diabetes diagnosis. For example, a type 1 diabetes diagnosis may be based on ICD-9 codes 250._1 and / or 250._3, and / or ICD-10 code E10.%. For example, a type 2 diabetes diagnosis may be based on ICD-9 codes 250._0 and / or 250._2, and / or ICD-10 code E11.%. "_" and "%" are placeholders, and "_" may not be empty; however, the placeholder "%" may be empty.

[0056] Experimental Data The area under the receiver operating characteristic (ROC) (Compare Swets, JA, Measuring accuracy of diagnostic systems, Science 240, 1285-1293, 1988) curve (AUC) is frequently used to measure the quality of clinical markers as well as machine learning algorithms / models (Bradley, AP, The use of the area under the ROC curve in the evaluation of machine learning algorithms, Pattern Recognition 30, 1145-1159, 1997). A perfect marker will achieve AUC=1.0, whereas flipping a coin will result in AUC=0.5.

[0057] Machine learning models that apply XGBoost in the learning procedure are trained and tested on different sets of training data that refer to all parameters or a subset of parameters available in the database. A machine learning model called a "Full XGBoost model" is trained using all parameters from a multiplicity of parameters, such as about 100 (about 948 features), available in the IBM Explorys database. Parameters in this context refer, for example, to creatinine, albumin, age, etc. Features in this context refer, for example, to creatinine, albumin, age, etc. max or creatinine median"Full XG boost model" refers to a selection or statistical value such as "full XG boost model" was created using all available parameters (all features). For data from a database, not all parameters are available for all patients (subjects) in a population. When working with a complete set of parameters, certain parameters have been found to be particularly important (e.g., top 5 or top 20 or top 30).

[0058] Additionally, a machine learning model called the "Top 20 XG Boosted Model" has been created (trained and tested) using only a subset of the data from the IBM Explorys database. In an embodiment, the data in the subset of data relates to (only) 20 parameters from the plurality of parameters, the 20 parameters being the parameters found to be most important in the machine learning process of the "Full XG Boosted Model." Below, 20 such parameters are listed (in no particular order of importance): age; albumin (serum and / or plasma); albumin (urine), systolic blood pressure, blood urea nitrogen (BUN), antihypertensive medication, insulin medication; number of pre-existing conditions: diabetic retinopathy, ischemic heart disease, peripheral arterial occlusive disease, cerebrovascular disease; creatinine (serum / plasma); time (days) since diabetes diagnosis; mean time span between two doctor's visits at which the parameter was measured or the diagnosis was made; diagnosis of type 2 DM with hyperglycemia; diagnosis of heart failure; estimated glomerular filtration rate (eGFR); red blood cells (serum and / or plasma); glucose (serum and / or plasma); hematocrit; hemoglobin; urinary albumin / creatinine ratio (UACR); and weight.

[0059] In training the "Top 20 XGBoost models" created as separate models, only the top 20 parameters determined from training the "Full XGBoost model" were used. Therefore, when the "Top 20 XGBoost models" were determined, other parameters (which may have been available) were ignored.

[0060] To evaluate the machine learning models, both for the "Full XG Boosted Model" and the "Top 20 XG Boosted Model", the AUC was determined for the subject population for which the database provides real-world data. Such calculations were performed for the subject population, taking into account all available parameters (features). Additionally, calculations were performed for the subject population, taking into account only data related to the following (six) parameters ("using a limited number of parameters"): age, time since diabetes, creatinine, estimated glomerular filtration rate (eGFR), albumin, and blood urea nitrogen (BUN).

[0061] For comparison, AUC was calculated for a logistic regression (LR) model similarly trained on the data of a subset of data for 20 parameters from the multiple parameters. Such machine learning models are called "LR top 20 models". For the "LR top 20 models (only limited number of parameters)", AUC calculations and specificity @ 90% sensitivity were evaluated by considering only subject-specific data for the following (six) parameters ("using limited number of parameters"): age, time since diabetes, creatinine, estimated glomerular filtration rate (eGFR), albumin, and blood urea nitrogen (BUN). For the 14 other parameters, no subject-specific data was used, but for those other parameters, data was imputed from selected cohorts or statistics, respectively.

[0062] The performance of a method for screening subjects for risk of CKD or for identifying people at high risk of developing CKD can be determined according to sensitivity (the proportion of high-risk patients correctly predicted) and specificity (the proportion of low-risk patients correctly assigned). However, either of these figures can be improved at the expense of the other by simply changing the threshold between high and low risk. Thus, the data pairs of sensitivity and specificity can be shown in the form of a so-called receiver operating characteristic (ROC) curve (see Swets, JA, Measuring accuracy of diagnostic systems, Science 240, 1285-1293, 1988), in which sensitivity is plotted as a function of 1-specificity (corresponding to the proportion of high-risk people incorrectly assigned).

[0063] The results of calculations performed on data from the Indiana Network for Patient Care (INPC) database for all parameters or only the six parameters identified above are shown in Table 1 . [Table 1]

[0064] As can be seen from Table 1, the AUC of the machine learning models is high for all the models shown, but by applying XGBoost we were able to achieve even better results than the LR models. Using only a limited number of parameters to calculate the AUC still provides reliable results.

[0065] Figure 3 shows the ROC curves for the "Full XG Boosted Model", "Top 20 XG Boosted Model" and "LR Top 20 Model" both when using all parameters and when using only a limited number of parameters. For the perfect classifier, the ROC curve reaches the upper left corner. In practice, the threshold corresponding to the data pair closest to this corner is called the "optimal threshold". If high sensitivity is aimed for, another threshold may be chosen to guarantee, for example, 90% sensitivity.

[0066] To further compare the machine learning XGBoost model presented here, calculations were also performed on a model (algorithm) for predicting risk factors for CKD known from EP 3 543 702 A1. This model is referred to below as the "algorithmic model". The "algorithmic model" also applies a logistic regression. No data imputation was applied. The results of the calculations performed on data from the IBM Explorys database are shown in Table 2. [Table 2]

[0067] From Table 2, it can be concluded that the machine learning model with the application of XGBoost provides improved results in terms of risk factor determination than the “algorithmic model”.

[0068] In summary, we demonstrate that various machine learning models for predicting risk factors of CKD perform robustly even when the number of available marker parameters is limited (specific selection of marker parameters): age, time since diabetes, creatinine, estimated glomerular filtration rate (eGFR), albumin, and blood urea nitrogen (BUN). The results support a pathway to high-quality predictive models that can be applied in clinical practice, enabling the transition to personalized outcomes-based healthcare.

Claims

1. 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 for a subject, the plurality of marker parameters comprising at least: Age value, a time since diagnosis value indicating time since diabetes diagnosis for the subject; Creatinine sample levels, Estimated glomerular filtration rate, Sample levels of albumin, and Blood urea nitrogen sample levels receiving marker data indicative of determining a risk factor indicative of the subject's risk of developing CKD from the plurality of marker parameters; A method comprising:

2. The method of claim 1, further comprising the plurality of marker parameters indicative of blood sample levels of creatinine for the subject.

3. 3. The method of claim 1 or 2, further comprising the plurality of marker parameters indicative of at least one of a blood sample level of albumin and a urine sample level of albumin for the subject.

4. 4. The method of claim 1, wherein the receiving comprises receiving marker data indicative of a plurality of marker parameters for the subject over a measurement period of two years or less.

5. The method of claim 1 , wherein the age value corresponds to the age of the subject at the time the risk factor is determined.

6. The method of claim 1 , wherein the time since diagnostic value indicates the time since diabetes diagnosis for the subject when the risk factor is determined.

7. 7. The method of any one of claims 1 to 6, wherein the risk factor indicates the risk of the subject suffering from CKD within a prediction period of 3 years from the end of the measurement period.

8. 1. A computer-implemented method for screening a subject for risk of chronic kidney disease (CKD) in a data processing system, the data processing system comprising: a processor; and configuring the processor to: a) receiving marker data indicative of a plurality of marker parameters for a subject, the plurality of marker parameters comprising at least: Age value, a value indicative of the time since diagnosis of diabetes for the subject; Creatinine sample levels, Estimated glomerular filtration rate, Sample levels of albumin, and Blood urea nitrogen sample levels receiving marker data indicative of b) determining a risk factor indicative of the subject's risk of developing CKD from the plurality of marker parameters; a non-transitory memory for storing a program for executing the 13. A computer-implemented method comprising:

9. Determining the risk factors in step b) Providing machine learning models, providing input data indicative of the plurality of marker parameters to the machine learning model; determining the risk factors by the machine learning model; and The computer-implemented method of claim 8 , comprising:

10. 10. The computer-implemented method of claim 9, wherein the machine learning model comprises providing an XGBoost machine learning model.

11. Providing the machine learning model, providing a set of training data for a population of subjects, the training data indicating a plurality of training parameters for the population of subjects, the training parameters including age, a level of creatinine, an estimated glomerular filtration rate, a level of albumin, a level of blood urea nitrogen, and an indication of whether the subject has developed CKD; providing diabetes diagnosis data indicating a time or date when a diabetes diagnosis was determined for a subject from said population of subjects; determining from said diabetes diagnosis data supplemental training data indicative of a time since diagnosis parameter indicative of a time since a diabetes diagnosis was determined for a subject from said population of subjects; providing an expanded set of training data comprising said set of training data and said supplemental training data; training the machine learning model based on the expanded set of training data; 11. A computer-implemented method according to claim 9 or 10, comprising:

12. 12. The computer-implemented method of claim 9, wherein the risk factors are determined using the machine learning model without imputing marker data.

13. 13. A system comprising a processor and a non-transitory memory storing a program that causes the processor to perform steps a) and b) of the method of any one of claims 8 to 12 for screening a subject for risk of chronic kidney disease (CKD).

14. 13. A computer program comprising instructions which, when executed by a computer, cause the computer to perform steps a) and b) of the method of any one of claims 8 to 12 for screening a subject for risk of chronic kidney disease (CKD).

Images