In 1899, when Sir William Osler advised the students of Albany Medical College to “care more particularly for the individual patient than for the special features of the disease,” he probably could not have imagined a scenario in which millions of patient records could be analyzed en masse to distinguish “features of a disease.”1 Over a century later, 96% of US hospitals have adopted an electronic medical record system, with an estimated 153 exabytes of health data generated in 2013 alone (1 exabyte = 1 billion gigabytes).2,3 This vast quantity of information is now beginning to inform “learning health systems” that have the potential to span disease prevention, cure, and management to improve care of the individual patient by efficiently moving data from clinical care to research and back.4 This data-driven approach to health care discovery and delivery is in its infancy, much the same as Osler was at the forefront of “modern medicine” in his time. Large-scale, population-level data have already proven to be of great value in the exploration of rare diseases.
Epidemiologic studies of rare diseases are necessary on many levels, but particularly to assess disease burden, identify outcomes, and inform preventive strategies and improvements in care. Go et al. evaluated the risks of kidney, cardiovascular, and mortality outcomes among nondiabetic adults with primary nephrotic syndrome due to minimal change disease, FSGS, or membranous nephropathy, identified within Kaiser Permanente Northern California (KPNC) between 1996 and 2012.5 Controls were nondiabetic, matched for age and indicators of KPNC resource utilization, and served as a comparison group for a unique assessment of risk among those with primary nephrotic syndrome. Although not unexpected, the burden of nephrotic syndrome far exceeded that of controls in ESKD, hospitalizations for acute cardiovascular events, and all-cause mortality. These outcomes are highly meaningful and relevant to patients, caregivers, and clinicians. Considering the high burden of cardiovascular disease in patients with CKD, there is a need to better understand its effect, and modifiable risk factors and optimal management among individuals with glomerular disease.
The use of data from electronic medical records holds great promise for the study of rare diseases, including glomerular disease. Similar to Go et al. others have made use of electronic medical record data in this patient population, including within Kaiser Permanente,6 MarketScan administrative data,7 and PEDSnet, a national network of children's hospitals.8 These studies focused on identifying renal biopsy diagnoses,6 biopsy in combination with International Classification of Diseases 9th revision (ICD-9) codes,8 or ICD-9 codes alone.7 The inclusive approach used by Go et al. to identify nephrotic syndrome associated with three specific glomerular diseases is a different and useful, albeit labor-intensive, approach. Newer, more efficient methods are needed that leverage novel computational tools that incorporate free text capture and updated diagnosis codes (i.e., ICD-10). Nonetheless, the rigor used to identify and adjudicate cases within the integrated KPNC health system, long-term follow-up, and presence of a comparator group, all lend confidence to the outcome estimates described by Go et al. In contrast, the data presented are at least a decade old, and treatment regimens have evolved. For example, a contemporary observational cohort of patients with membranous nephropathy (CureGN) described rates of renin-angiotensin-aldosterone system inhibition at 85%,9 well above the rate in this study of 32.6%. In the larger CureGN cohort, inclusive of minimal change disease and FSGS, renin-angiotensin-aldosterone system inhibition use is similar to that reported in membranous nephropathy alone (unpublished data).
With respect to the population studied, the 4.5 million members of KPNC are reported to be sociodemographically representative of the regional and state population.5,10 Go and colleagues acknowledge that inferences to other geographic regions, health systems, and uninsured patients may be limited. Compared with the overall US population, California has far fewer Black (6.5% versus 13.4%), and White non-LatinX (36.5% versus 60.1%) inhabitants, and far more LatinX (39.4 versus 18.5%) and Asian (15.5 versus 5.9%) inhabitants (US Census https://www.census.gov/quickfacts/fact/table/CA,US/PST045219). In contrast, other metrics such as completion of high school and percent living in poverty are similar to the overall US population.
Differences in health insurance coverage and access to care likely also limit the external validity of these data. The Kaiser Permanente integrated health care and insurance system provides the distinct advantage, for both patients and research, of capturing multiple levels of care and encounter types, often with long-term follow-up. These are strengths when studying rare diseases. This type of health care system stands in stark contrast to how health care is insured and delivered across much of the United States, where there is a patchwork of employment-based and Medicaid options for insurance coverage before age 65. During the time of the KPNC study (1996–2012), which preceded the Affordable Care Act, approximately 18% of adults under age 65 were without insurance and another 12% were underinsured, so 30% had little to no care,11 much less “integrated care.” As such, the outcome estimates presented likely underestimate the true burden of these outcomes in individuals with glomerular disease.
In summary, this study highlights the burden of potentially preventable adverse outcomes in nephrotic syndrome. This is done with a high level of rigor and an impressive breadth of data capture. Future efforts that leverage population-level data and incorporate other important modifiers and outcomes, such as immunosuppression exposure, disease activity, and infection,12 will add to a more comprehensive understanding of morbidity in these diseases. Although artificial intelligence methods, natural language processing of free text, and other innovative techniques are not yet well developed,13 they hold promise for increased efficiency, faster learning, and improved care.4 Multidisciplinary collaboration with health informaticians, biostatisticians, and computer scientists is needed to propel these novel methods into the mainstream. As we move toward learning health systems that efficiently translate vast amounts of clinical data into improved patient outcomes, we are, in many ways, fulfilling Osler’s recommendation to “care particularly for the individual patient.”1 Our patients with rare kidney diseases should be at the forefront to benefit from these powerful new tools.
S.L. Hogan reports receiving honoraria from the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases (grant reviewer), and as a Veterans Affairs grant reviewer. The remaining author has nothing to disclose.
1. Osler W: Address to the students of Albany Medical College. Albany Medical Annals 20: 307–309, 1899
2. Office of the National Coordinator for Health Information Technology: Non-federal Acute Care Hospital Electronic Health Record Adoption. Available at: dashboard.healthit.gov/quickstats/pages/FIG-Hospital-EHR-Adoption.php
3. Banks MA: Sizing up big data. Nat Med 26: 5–6, 2020
4. McGinnis JM, Fineberg HV, Dzau VJ: Advancing the learning health system. N Engl J Med 385: 1–5, 2021
5. Go AS, Tan TC, Chertow GM, Ordonez JD, Fan D, Law D, et al.: Primary nephrotic syndrome and risks of end-stage kidney disease, cardiovascular events and death: The Kaiser Permamente Nephrotic Syndrome Study. J Am Soc Nephrol 32: 2303–2314, 2021
6. Sun AZ, Shu YH, Harrison TN, Hever A, Jacobsen SJ, O’Shaughnessy MM, et al.: Identifying patients with rare disease using electronic health record data: The Kaiser Permanente Southern California Membranous Nephropathy Cohort. Perm J 24:19.126, 2020
7. Nazareth TA, Kariburyo F, Kirkemo A, Xie L, Pavlova-Wolf A, Bartels-Peculis L, et al.: Patients with idiopathic membranous nephropathy: A real-world clinical and economic analysis of U.S. claims data. J Manag Care Spec Pharm 25: 1011–1020, 2019
8. Denburg MR, Razzaghi H, Bailey LC, Soranno DE, Pollack AH, Dharnidharka VR, et al.: Using electronic health record data to rapidly identify children with glomerular disease for clinical research. J Am Soc Nephrol 30: 2427–2435, 2019
9. O’Shaughnessy MM, Troost JP, Bomback AS, Hladunewich MA, Ashoor IF, Gibson KL, et al.: Treatment patterns among adults and children with membranous nephropathy in the Cure Glomerulonephropathy Network (CureGN). Kidney Int Rep 4: 1725–1734, 2019
10. Gordon N, Lin T: The Kaiser Permanente Northern California adult member health survey. Perm J 20: 15–225, 2016
11. Schoen C, Hayes SL, Radely DC, Collins SR: Access to primary and preventive health care across states prior to the coverage expansions of the Affordable Care Act. Issue Brief (Commonw Fund) 17: 1–14, 2014
12. Glenn DA, Henderson CD, O’Shaughnessy M, Hu Y, Bomback A, Gibson K, et al.; CureGN Consortium: Infection-related acute care events among patients with glomerular disease. Clin J Am Soc Nephrol 15: 1749–1761, 2020
13. Garcelon N, Burgun A, Salomon R, Neuraz A: Electronic health records for the diagnosis of rare diseases. Kidney Int 97: 676–686, 2020