The KidneyIntelX test would affect primary care physician (PCP) decision-making, and PCPs would use the results of KidneyIntelX more than albuminuria and estimated glomerular filtration rate when making decisions about diabetic kidney disease management.
Objectives: Estimated glomerular filtration rate (eGFR) and albuminuria, the current standard-of-care tests that predict risk of kidney function decline in early-stage diabetic kidney disease (DKD), are only modestly useful. We evaluated the decision-making impact of an artificial intelligence–enabled prognostic test, KidneyIntelX, in the management of DKD by primary care physicians (PCPs).
Study Design: This was a prospective web-based survey administered among PCPs in the United States.
Methods: We used conjoint analysis with multivariable logit models to estimate PCP preferences. The survey included hypothetical patient profiles with 6 attributes: albuminuria, eGFR, age, blood pressure (BP), hemoglobin A1c (HbA1c), and KidneyIntelX result. Each PCP viewed 8 patient profiles randomly selected from 42 unique profiles having 1 level from each attribute. For each patient, PCPs were asked to indicate whether they would prescribe a sodium-glucose cotransporter-2 (SGLT2) inhibitor, increase angiotensin receptor blocker (ARB) dose, and/or refer to a nephrologist.
Results: A total of 401 PCPs completed the survey (response rate, 8.8%). The relative importance of the top 2 attributes for each decision were HbA1c (52%) and KidneyIntelX result (23%) for prescribing SGLT2 inhibitors, BP (62%) and KidneyIntelX result (13%) for increasing ARB dose, and eGFR (42%) and KidneyIntelX result (27%) for nephrologist referral. A high-risk KidneyIntelX result was associated with significantly higher odds of PCPs prescribing SGLT2 inhibitors (odds ratio [OR], 1.64; 95% CI, 1.29-2.08), increasing ARB dose (OR, 1.49; 95% CI, 1.17-1.89), and referring to a nephrologist (OR, 2.47; 95% CI, 1.99-3.08) compared with no test.
Conclusions: The KidneyIntelX test had greater relative importance than albuminuria and eGFR to PCPs in making treatment decisions and was second only to eGFR for nephrologist referrals. Because of its significant impact on decision-making, KidneyIntelX has high clinical utility in DKD management.
Am J Manag Care. 2022;28(12):654-661. https://doi.org/10.37765/ajmc.2022.89207
A prognostic test, such as KidneyIntelX, that offers better risk stratification than the Kidney Disease Improving Global Outcomes classification, which is based solely on albuminuria and estimated glomerular filtration rate, would have a significant impact on physician decision-making pertaining to treatment selection and nephrologist referrals in patients with diabetic kidney disease (DKD).
Management of diabetic kidney disease (DKD) has emerged as a major public health challenge in the United States. CDC data estimate that more than 34 million (or 1 in 10) adults in the United States have diabetes, of whom approximately 90% to 95% have type 2 diabetes (T2D); further, an estimated 40% of patients with T2D have DKD.1,2 DKD is a chronic, progressive disease that is associated with substantial health care costs and leads to end-stage kidney disease (ESKD), requiring dialysis and/or a kidney transplant. DKD accounted for 14% of all Medicare spending for patients 65 years and older in 2017.3 It is a type of chronic kidney disease (CKD) caused by diabetes. The financial burden imposed by CKD and ESKD results in an estimated cost of $120 billion annually, representing approximately 7.2% of overall Medicare fee-for-service reimbursed claims.3 The incidence of ESKD is predicted to rise by approximately 11% to 18% from 2015 to 2030, due to changes in age and race distribution and the prevalence of obesity and diabetes.4,5
To reduce the clinical and economic burden of DKD, accurate staging and prognosis with respect to risk of progression are vital to enable appropriate interventions and referral to specialists early in the management of the disease.6 The international Kidney Disease: Improving Global Outcomes (KDIGO) 2012 Clinical Practice Guideline for CKD Evaluation and Management7 includes a cause–glomerular filtration rate (GFR)–albuminuria (C-G-A) CKD definition and classification system to optimize risk stratification of kidney disease progression based on the estimated GFR (eGFR) and urine albumin-creatinine ratio (uACR). However, the predictive ability of the combination of eGFR and uACR is limited by the underuse of uACR and of nuanced interpretation that considers both parameters in the CKD classification.8 Current methods of risk stratification based on albuminuria and eGFR can be enhanced to identify which patients are more likely to experience progressive decline in kidney function, especially in earlier stages of DKD (G1-G3b). A more precise risk of progression that holistically accounts for blood-based biomarkers in combination with clinical parameters could better enable clinicians to implement appropriate interventions, such as prescription of guideline-based treatments to lower blood pressure (eg, an angiotensin-converting enzyme [ACE] inhibitor or an angiotensin II receptor blocker [ARB]), efforts to lower elevated blood glucose and hemoglobin A1c (HbA1c) levels (eg, metformin and sodium-glucose cotransporter-2 [SGLT2] inhibitors), or, for high-risk patients, referral to a nephrologist.9 Furthermore, such approaches to risk stratification may also avoid expensive treatments and unnecessary interventions in low-risk patients.10
In this study, we sought to evaluate the impact of a novel prognostic test that has been reported to potentially offer improved risk stratification for physician decision-making in patients with early-stage DKD.11 The test evaluated in this study, KidneyIntelX, can identify the risk of progression in early-stage DKD (stages G1-G3b, excluding G1A1 and G2A1) using a combination of blood-based biomarkers, including soluble tumor necrosis factor receptor (sTNFR)–1 and sTNFR-2, and kidney injury molecule 1 (KIM-1), and clinical factors, including eGFR, uACR, serum calcium, HbA1c, systolic blood pressure, platelets, and liver enzyme aspartate aminotransferase, derived from electronic health records. Based on previous research among primary care physicians (PCPs) eliciting unmet needs in DKD patient management,9 PCPs may find the KidneyIntelX test to be a suitable alternative to current standard of care, particularly when the degree of albuminuria is uncertain, as described in a validation study.11 The positive predictive value for progressive decline in kidney function in the high-risk group included in the validation study was 61% for KidneyIntelX vs 40% for the highest-risk stratum by KDIGO categorization (P < .001). Only 10% of those scored as low risk by KidneyIntelX went on to experience progression, which represents a high negative predictive value of 90%.11 Based on this performance, KidneyIntelX provides an alternate method of risk-stratifying patients with DKD, which could help simplify physicians’ decision-making regarding these patients. eAppendix A (eAppendices available at ajmc.com) shows the net reclassification from KDIGO risk strata to KidneyIntelX risk strata in derivation and validation sample. eAppendix B shows the treatment and patient management recommendations provided by KidneyIntelX, stratified by risk result.
In this study, we sought to address the question: Does availability of the information provided by the KidneyIntelX test affect physicians’ treatment decisions compared with the information available from the standard tests recommended by KDIGO practice guidelines? To answer this question, we undertook a physician preference study assessing whether the results from a novel prognostic test such as KidneyIntelX would aid physician decision-making in the real-world setting more than current standard-of-care tests, such as those measuring albuminuria and eGFR.
We administered a web survey in a convenience sample of PCPs currently managing patients with DKD. The PCPs were recruited via an external recruiting agency that enlists panels of medical providers for primary research. Respondents were allowed to participate in the study if they (1) were board-certified in family medicine, primary care, or internal medicine; (2) had at least 2 years of experience managing patients; (3) managed more than 20 patients with T2D in the past 6 months; (4) tested at least 11 patients with T2D for DKD in the past 6 months; (5) spent at least 50% of their time in direct clinical care; and (6) did not reside in Maine or Vermont due to regulations in those states that limit physicians from receiving compensation for participation in market research. If respondents did not satisfy a criterion, their survey was terminated at the end of the screener. PCPs completing the survey were offered $25 as honorarium.
Conjoint Analysis Experimental Design
Conjoint analysis is a preference elicitation technique in which the implicit values for the outcome are derived from some overall score for a profile consisting (conjointly) of 2 or more attributes.12,13 Conjoint analysis allows the investigator to measure how sensitive a respondent is to changes in the levels of 1 or more attributes relative to others. Sawtooth Software’s Conjoint Value Analysis module version 3.0 was used for the experimental setup of this study.14
First, a set of attributes and levels that may affect physician decision-making in DKD treatment and management was developed based on a literature review and discussion with 2 clinical experts.15-20 A total of 6 attributes were evaluated in this study: albuminuria, eGFR, age, blood pressure, HbA1c, and KidneyIntelX test result. The levels for each attribute presented in the patient profiles are displayed in Table 1.
Hypothetical profiles were then constructed using a combination of 1 level from each attribute. Given the number of attributes (6) and levels (5 attributes with 3 levels and 1 attribute with 4 levels), asking respondents to review 972 profiles was not feasible. To improve response efficiency and maintain statistical precision, a balanced orthogonal design was generated with a fewer number of profiles using the Sawtooth Software. A balanced orthogonal design is one in which all attribute levels vary independently (and thus are not correlated) and each level of each attribute is represented an approximately equal number of times in the subset of profiles.12 This yielded a total of 42 profiles. To reduce the burden on respondents and improve survey response rates, a random selection of 8 of the 42 profiles was shown to each respondent. The profiles were picked at random using the least fill methodology, such that each of the 42 profiles was shown an approximately equal number of times.
The outcome in the conjoint experiment was PCP decision-making, and it was operationalized using 3 binary variables. For each profile seen, PCPs were asked to select “yes” or “no” for the following 3 questions: (1) Would you prescribe an SGLT2 inhibitor with a DKD indication to this patient?; (2) Would you increase the dose of losartan from 50 mg per day to 100 mg per day in this patient?; and (3) Would you refer this patient to a nephrologist? These outcomes (decisions) are in accordance with both the 2020 KDIGO guidelines and 2021 American Diabetes Association guidelines and were validated for accuracy by clinical experts.
Survey Instrument and Data Collection
The self-administered, anonymous, web-based survey was hosted by the recruiting agency. The survey was designed to be completed in 15 minutes. For the purposes of the survey, the KidneyIntelX test was blinded as Test X to minimize any potential biases. The full survey and the product profile shown to the respondents are presented in eAppendices C and D, respectively.
The survey was pretested with 2 PCPs prior to full launch to detect ambiguities in the patient profiles or the treatment choice questions. In addition to the 8 patient profiles, the pretest survey included 2 additional profiles (hold-out profiles) that were identical in terms of the attributes and levels (Table 1) to assess for test-retest reliability (consistency). Feedback received in the pretest phase was implemented in the final survey before full launch. Full launch data from the final survey were collected between November 16, 2020, and December 3, 2020. The study was reviewed by an independent institutional review board (IRB) and deemed exempt from full review, as no sensitive or identifying information was being collected from the respondents (Sterling IRB approval number APP819).
A sample size of 400 PCPs was deemed sufficient after a preliminary multivariable logit analysis using dummy data produced standard errors that were approximately 0.10 or less.21
Analysis of impact of the KidneyIntelX test result and other attributes on PCP decision-making was conducted using a multivariable logit model for each binary outcome (prescription of SGLT2 inhibitors, dose change for losartan, and referral to a nephrologist). The output of the logit model includes preference weights (beta coefficients) for each level of each attribute. Relative attribute importance was calculated as the difference in the minimum and maximum values of relative preference weights for levels in each attribute. These values were normalized to sum to 100%. Raw relative preference weights for each level were also converted to odds ratios and 95% CIs to assess the impact of the KidneyIntelX test result on decision outcomes compared with no test.
A total of 401 respondents completed the survey. The sample attrition leading to the final number of respondents included in this study can be found in eAppendix E.
A total of 6104 invitations to the web survey were sent out via email: 670 physicians opened the email, 540 physicians clicked the link and entered the survey, and 401 physicians qualified based on the inclusion/exclusion criteria and completed the full survey. A total of 139 physicians were excluded because they did not meet inclusion criteria for the survey. The median time taken to complete the survey was 15.2 minutes.
The characteristics of the 401 qualified respondents who completed the survey can be found in a table in eAppendix F. The geographic distribution of respondents was relatively even; the highest proportion of respondents was from the South (31%), followed by the Midwest (25%), which was representative of the geographic distribution in the United States.22
PCPs in our study acknowledged the benefits of accurately assessing the risk of progressive decline in kidney function in both high-risk and low-risk patients with T2D (Figure 1).
PCPs reported several clinical benefits of assessing the risk of progressive kidney function decline. Approximately 80% (322/401) PCPs noted that risk assessment would support the decision to take more aggressive clinical action early in high-risk patients; 77% (307/401) suggested it would be useful to help avoid prescribing nephrotoxic drugs to high-risk patients; and 72% (290/401) indicated that with proper risk assessment, high-risk patients could be monitored more frequently. A relatively smaller yet substantial proportion (55%; 220/401) reported avoidance of unnecessary nephrologist visits in low-risk patients as a benefit. Almost 98% of PCPs in our sample indicated that they were somewhat, very, or extremely likely to order KidneyIntelX for their patients with DKD.
Conjoint Analysis Findings
The relative preference weights for each level of each patient attribute for the decisions to prescribe SGLT2 inhibitors, increase the dose of losartan, and refer patients to a nephrologist are presented in Figure 2 (part A and parts B and C). Severely increased albuminuria and a high-risk KidneyIntelX test result were the 2 levels that had significantly higher relative preference weights compared with their respective attribute level reference categories across all outcomes.
Figure 3 presents the overall relative importance of the attributes for the ranges of levels presented in the 3 outcomes tested in the conjoint analysis. For the decision to prescribe SGLT2 inhibitors, HbA1c was the most important attribute (relative importance, 52%), followed by the KidneyIntelX test result (23%). For the decision to increase losartan dose, blood pressure was the most important attribute (62%), followed by the KidneyIntelX test result (13%). For referral to a nephrologist, eGFR was the most important attribute (42%), followed by the KidneyIntelX test result (27%).
Table 2 displays odds ratios calculated from relative preference weights of the multivariable logit models.
A KidneyIntelX high-risk test result was associated with significantly higher odds (1.64 times) of prescribing SGLT2 inhibitors, significantly higher odds (1.49 times) of increasing losartan dose, and significantly higher odds (2.47 times) of referring patients to a nephrologist compared with not having the test available, while controlling for other attributes. Interestingly, odds ratios for a low-risk test result were less than 1 across all outcomes when compared with no test, although these did not reach statistical significance.
This study sought to assess the impact of a prognostic test on physician decision-making in the management of patients with DKD. The KidneyIntelX test result had a greater impact than albuminuria and eGFR in PCPs’ decision to prescribe SGLT2 inhibitors and increase the dose of ARBs (losartan) and was second only to eGFR in the decision to refer patients to a nephrologist.
Timely treatment to help patients reach an HbA1c target of less than 7.0%, as well as the use of ACE inhibitors or ARBs to control their blood pressure, are beneficial in patients who are at high risk of progressive kidney function decline.23,24 Despite this, data indicate that fewer than 50% of patients are treated appropriately with these agents.8 Because 80% of PCPs in our study view early clinical action as a clinical benefit of assessing the risk of progressive decline in kidney function in patients with DKD, KidneyIntelX could be a very useful tool in this patient population. A major challenge to the early diagnosis and treatment of DKD is that patients with diabetes are neither routinely screened for deteriorating kidney function nor referred to nephrologists when needed.25-27 This finding was corroborated in our study, with 72% of PCPs identifying the importance of frequent monitoring and 55% of PCPs emphasizing intervention by a nephrologist as a clinical need among patients with DKD. Given the shortcomings of current standard-of-care tests and high physician preference reported in our study, a prognostic test such as KidneyIntelX could address all the aforementioned needs and challenges by accurately assessing the risk of progressive decline in kidney function in patients with T2D.
Our study has 3 important findings regarding PCP decision-making for DKD patients. First, we found that the factors most associated with treatment decisions are, as expected, HbA1c for SGLT2 inhibitor prescriptions and blood pressure for losartan dose increase. This finding is not surprising, given that several preclinical studies and clinical trials of SGLT2 inhibitors, especially canagliflozin, have consistently provided evidence for reduction of albuminuria and preservation of kidney function.28,29 Likewise, controlling blood pressure through the use of ACE inhibitors or ARBs is a common treatment strategy in patients with DKD because of the kidney-protective effects associated with these medications in the presence of albuminuria.30-32 Clinical practice guidelines recommend that when eGFR falls below 30 mL/min/1.73 m2, patients with T2D should be referred to nephrologists.7 Our study showed results that were in line with this, as eGFR was the most important attribute affecting nephrologist referrals. These findings are consistent with logical and clinical expectations, thereby providing evidence in support of the robustness of our results.
Second, we found that routine clinical factors are not sufficient for decision-making pertaining to management of these patients. The finding that KidneyIntelX test results would affect decision-making more than albuminuria and eGFR for 2 of the 3 outcomes (ie, SGLT2 inhibitor prescription and losartan dose increase) provides strong evidence for the clinical utility of this test in the DKD management pathway. The addition of the KidneyIntelX test would significantly affect treatment and management decisions for high-risk patients based on the significantly higher odds ratios obtained for all outcomes compared with no test. Given that the model for KidneyIntelX includes uACR, eGFR, blood pressure, and HbA1c, along with blood-based protein biomarkers (sTNFR-1, sTNFR-2, and KIM-1), KidneyIntelX offers management recommendations based on more comprehensive information for physician decision-making compared with individual attributes that are currently used by PCPs to make these decisions. Moreover, clinical parameters in the KidneyIntelX test are multidimensional and additive to eGFR, which has limited predictive capability. Thus, even patients who actively manifest hyperfiltration or have a treatment-induced decline in eGFR can have a reliable and robust risk score for progression with minimal confounding by these external factors when using the KidneyIntelX test.
Third, we saw several interesting patterns in PCP decision-making based on the KidneyIntelX test results. A low-risk result was associated with lower odds of prescribing SGLT2 inhibitors, prescribing losartan, and referring patients to a nephrologist compared with no test. Although this odds ratio did not reach statistical significance, the directionality indicated that a low-risk result gave the PCPs more confidence to refrain from prescribing SGLT2 inhibitors. A similar pattern was observed for the other 2 outcomes, increasing the dose of losartan and nephrologist referral. This further demonstrates the utility of a prognostic test such as KidneyIntelX in avoiding unnecessary treatments and specialist visits in patients with low risk of progressive kidney function decline.
There are some limitations to this study. First, our findings are limited to the range of levels of key attributes included in this analysis and may not completely capture the scope of influence of other attributes on decision-making. However, the probability of this is very low, given that the attributes and levels included in this analysis were based on extensive secondary research and input from practicing clinical experts in the field of DKD. Second, to reduce burden, respondents were shown only a subset of the full factorial design of profiles that represented all the possible attribute and level combinations. However, the potential bias introduced from using a subset of profiles is minimal, because the 42 profiles shown to respondents (fractional factorial design) had a high D-efficiency (0.98), which is a measure of the design efficiency of the conjoint analysis (maximum possible D-efficiency = 1). Third, this is not a real-world study measuring actual physician behavior in practice. However, whereas traditional surveys may suffer from social desirability bias, wherein survey respondents tend to answer questions in a manner that will be viewed favorably by others, our study instead uses a novel preference elicitation technique with an experimental design that assesses implicit preferences, thereby minimizing this bias to a significant degree.12 Our survey also had test-retest reliability, demonstrated by the fact that both pretest participants chose the same response for the 2 hold-out cases. Finally, our study compared clinical utility of KidneyIntelX with the standard-of-care risk stratification currently being employed in practice (KDIGO risk stratification), as our goal was to derive practice-based utility. We did not test the utility of KidneyIntelX against alternative theoretical risk scores (eg, Kidney Failure Risk Equations) that predict kidney disease progression. The results of this study should be interpreted with this limitation in mind. Future studies could compare the performance and clinical utility of KidneyIntelX with other algorithms.
A prognostic test such as KidneyIntelX that uses biomarkers and clinical parameters to better stratify patients by risk of kidney function decline would have a greater impact on physician decision-making than current standard-of-care tests. It addresses unmet needs identified in current practice and allows PCPs to act in the early stages of DKD progression before significant symptoms emerge. Our study shows that if tests such as KidneyIntelX were made readily accessible, PCPs would incorporate the test results into their treatment plan, ultimately making more informed decisions in the management of their patients with DKD, potentially resulting in improved outcomes.
Author Affiliations: Boston Healthcare Associates (MD, SR, JC, TFG), Boston, MA; National Kidney Foundation (EM), New York, NY; Mount Sinai Medical Center (SGC), New York, NY; Icahn School of Medicine at Mount Sinai (JAV), New York, NY.
Source of Funding: This study was funded by RenalytixAI.
Prior Presentation: The material in this manuscript was presented at the National Kidney Foundation’s Virtual Spring Clinical Meeting, April 6-10, 2021.
Author Disclosures: Dr Datar, Mr Ramakrishnan, Ms Chong, and Dr Goss are employees of Boston Healthcare Associates, which received funding from RenalytixAI to conduct this study and prepare this manuscript. Dr Coca reports consultancies from Renalytix, Bayer, and CHF Solutions; grants pending and received from National Institutes of Health; attendance at ASN Kidney Week; patents pending from KidneyIntelX; royalties from KidneyIntelX; stock ownership in Renalytix; and employment by Mount Sinai Medical Center, which owns part of Renalytix. Dr Vassalotti is a paid advisory board member on the RenalytixAI CKD Advisory Board on biomarkers and artificial intelligence for type 2 diabetes and on the Boehringer Ingelheim/Lilly CKD Implementation Advisory Board, and has received grants from the AARP Quality Measures Innovation Project. Ms Montgomery reports no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (MD, TFG, EM, JAV); acquisition of data (MD, SR, JC, TFG); analysis and interpretation of data (MD, SR, JC, TFG, SGC, EM, JAV); drafting of the manuscript (MD, SR, TFG); critical revision of the manuscript for important intellectual content (MD, SR, JC, EM, TFG, SGC, JAV); statistical analysis (MD, SR, JC); provision of patients or study materials (MD); obtaining funding (MD, TFG); administrative, technical, or logistic support (MD, SR, JC, SGC); and supervision (MD, TFG).
Address Correspondence to: Manasi Datar, PhD, Boston Healthcare Associates, 33 Arch St, 17th Floor, Boston, MA 02110. Email: firstname.lastname@example.org.
1. Type 2 diabetes. CDC. Accessed January 15, 2021. https://www.cdc.gov/diabetes/basics/type2.html
2. Wheeler DC, James J, Patel D, et al; Improving Diabetes Steering Committee. SGLT2 inhibitors: slowing of chronic kidney disease progression in type 2 diabetes. Diabetes Ther. 2020;11(12):2757-2774. doi:10.1007/s13300-020-00930-x
3. Saran R, Robinson B, Abbott KC, et al. U.S. Renal Data System 2019 annual data report: epidemiology of kidney disease in the United States. Am J Kidney Dis. 2020;75(1 suppl 1):S1-S64.
4. Lin J, Thompson TJ, Cheng YJ, et al. Projection of the future diabetes burden in the United States through 2060. Popul Health Metr. 2018;16(1):9. doi:10.1186/s12963-018-0166-4
5. McCullough KP, Morgenstern H, Saran R, Herman WH, Robinson BM. Projecting ESRD incidence and prevalence in the United States through 2030. J Am Soc Nephrol. 2019;30(1):127-135. doi:10.1681/ASN.2018050531
6. Chen TK, Knicely DH, Grams ME. Chronic kidney disease diagnosis and management: a review. JAMA. 2019;322(13):1294-1304. doi:10.1001/jama.2019.14745
7. Kidney Disease: Improving Global Outcomes (KDIGO) Diabetes Work Group. KDIGO 2020 Clinical Practice Guideline for Diabetes Management in Chronic Kidney Disease. Kidney Int. 2020;98(4 suppl):S1-S115. doi:10.1016/j.kint.2020.06.019
8. Tuttle KR, Alicic RZ, Duru OK, et al. Clinical characteristics of and risk factors for chronic kidney disease among adults and children: an analysis of the CURE-CKD registry. JAMA Netw Open. 2019;2(12):e1918169. doi:10.1001/jamanetworkopen.2019.18169
9. Datar M, Ramakrishnan S, Montgomery E, Coca SG, Vassalotti JA, Goss T. A qualitative study documenting unmet needs in the management of diabetic kidney disease (DKD) in the primary care setting. BMC Public Health. 2021;21(1):930. doi:10.1186/s12889-021-10959-7
10. de Boer IH, Caramori ML, Chan JCN, et al. Executive summary of the 2020 KDIGO Diabetes Management in CKD Guideline: evidence-based advances in monitoring and treatment. Kidney Int. 2020;98(4):839-848. doi:10.1016/j.kint.2020.06.024
11. Chan L, Nadkarni GN, Fleming F, et al. Derivation and validation of a machine learning risk score using biomarker and electronic patient data to predict progression of diabetic kidney disease. Diabetologia. 2021;64(7):1504-1515. doi:10.1101/2020.06.01.20119552
12. Bridges JFP, Hauber AB, Marshall D, et al. Conjoint analysis applications in health—a checklist: a report of the ISPOR Good Research Practices for Conjoint Analysis Task Force. Value Health. 2011;14(4):403-413. doi:10.1016/j.jval.2010.11.013
13. Janssen EM, Hauber AB, Bridges JFP. Conducting a discrete-choice experiment study following recommendations for good research practices: an application for eliciting patient preferences for diabetes treatments. Value Health. 2018;21(1):59-68. doi:10.1016/j.jval.2017.07.001
14. Conjoint Value Analysis (CVA) version 3.0. Sawtooth Software. Accessed July 13, 2022. https://content.sawtoothsoftware.com/assets/6bb3ffd9-5209-422b-8345-35b4f9750d26
15. Baumgarten M, Gehr T. Chronic kidney disease: detection and evaluation. Am Fam Physician. 2011;84(10):1138-1148.
16. Chronic kidney disease tests & diagnosis. National Institute of Diabetes and Digestive and Kidney Diseases. Accessed January 11, 2021. https://www.niddk.nih.gov/health-information/kidney-disease/chronic-kidney-disease-ckd/tests-diagnosis
17. Lin YC, Chang YH, Yang SY, Wu KD, Chu TS. Update of pathophysiology and management of diabetic kidney disease. J Formos Med Assoc. 2018;117(8):662-675. doi:10.1016/j.jfma.2018.02.007
18. Chronic kidney disease in the United States, 2019. CDC. March 5, 2019. Accessed November 12, 2020. https://www.cdc.gov/kidneydisease/pdf/2019_National-Chronic-Kidney-Disease-Fact-Sheet.pdf
19. Blood pressure chart: what your reading means. Mayo Clinic. Accessed January 11, 2021. https://www.mayoclinic.org/diseases-conditions/high-blood-pressure/in-depth/blood-pressure/art-20050982
20. Nakanishi S, Hirukawa H, Shimoda M, et al. Comparison of HbA1c levels and body mass index for prevention of diabetic kidney disease: a retrospective longitudinal study using outpatient clinical data in Japanese patients with type 2 diabetes mellitus. Diabetes Res Clin Pract. 2019;155:107807. doi:10.1016/j.diabres.2019.107807
21. Orme B, ed. MBC v1.1: Software for Menu-Based Choice Analysis Manual v1.1. Sawtooth Software; 2016. Accessed December 13, 2020. https://content.sawtoothsoftware.com/assets/6f8f31ce-d0fa-434b-936d-d41598b592bd
22. 2019 national and state population estimates. United States Census Bureau. December 30, 2019. Accessed January 13, 2021. https://www.census.gov/newsroom/press-kits/2019/national-state-estimates.html
23. Snider JT, Sullivan J, van Eijndhoven E, et al. Lifetime benefits of early detection and treatment of diabetic kidney disease. PLoS One. 2019;14(5):e0217487. doi:10.1371/journal.pone.0217487
24. Kitada M, Kanasaki K, Koya D. Clinical therapeutic strategies for early stage of diabetic kidney disease. World J Diabetes. 2014;5(3):342-356. doi:10.4239/wjd.v5.i3.342
25. Stanton RC. Clinical challenges in diagnosis and management of diabetic kidney disease. Am J Kidney Dis. 2014;63(2 suppl 2):S3-S21. doi:10.1053/j.ajkd.2013.10.050
26. Szczech LA, Stewart RC, Su HL, et al. Primary care detection of chronic kidney disease in adults with type-2 diabetes: the ADD-CKD Study (awareness, detection and drug therapy in type 2 diabetes and chronic kidney disease). PLoS One. 2014;9(11):e110535. doi:10.1371/journal.pone.0110535
27. Vassalotti JA, DeVinney R, Lukasik S, et al. CKD quality improvement intervention with PCMH integration: health plan results. Am J Manag Care. 2019;25(11):e326-e333.
28. Heerspink HJL, Desai M, Jardine M, Balis D, Meininger G, Perkovic V. Canagliflozin slows progression of renal function decline independently of glycemic effects. J Am Soc Nephrol. 2017;28(1):368-375. doi:10.1681/ASN.2016030278
29. Perkovic V, Jardine MJ, Neal B, et al; CREDENCE Trial Investigators. Canagliflozin and renal outcomes in type 2 diabetes and nephropathy. N Engl J Med. 2019;380(24):2295-2306. doi:10.1056/nejmoa1811744
30. Lewis EJ, Hunsicker, Lawrence G, Bain RP, Rohde RD. The effect of angiotensin-converting-enzyme inhibition on diabetic nephropathy. N Engl J Med. 1993;329(20):1456-1462. doi:10.1056/NEJM199311113292004
31. Lewis EJ, Hunsicker LG, Clarke WR, et al; Collaborative Study Group. Renoprotective effect of the angiotensin-receptor antagonist irbesartan in patients with nephropathy due to type 2 diabetes. N Engl J Med. 2001;345(12):851-860. doi:10.1056/NEJMoa011303
32. Brenner BM, Cooper ME, de Zeeuw D, et al; RENAAL Study Investigators. Effects of losartan on renal and cardiovascular outcomes in patients with type 2 diabetes and nephropathy. N Engl J Med. 2001;345(12):861-869. doi:10.1056/NEJMoa011161