Managing comorbidities in chronic kidney disease reduces utilization and costs

This retrospective cohort study used the Humana Research Database to identify patients 19–89 years of age who were enrolled in a Medicare Advantage Prescription Drug (MAPD) plan and who had claims’ evidence of CKD Stage 3a or greater during the identification period (calendar year 2017). The Humana Research Database includes patients’ enrollment records, full medical and pharmacy claims, basic demographic information such as age, sex, and geographic region, detailed information on diagnosed medical conditions and procedures performed, and linked outpatient pharmacy records and lab results. Eligible patients had at least two outpatient medical claims with a diagnosis of CKD (ICD-10 diagnosis codes N18.3, N18.4 and N18.5), at least one inpatient medical claim with a diagnosis of CKD, or at least two serum creatinine laboratory values 90 days apart that translated to an eGFR value less than 60 ml/min/1.73m². CKD stage was determined based on the eGFR value, and the diagnoses code in the absence of an eGFR. Eligible patients were enrolled in an MAPD plan for at least 12 months prior to the index date (for measuring baseline characteristics, including the CKD diagnosis) and at least 12 months following the index date (for measuring outcomes). The index date was the claims date providing the first evidence of CKD, and the observation period for receipt of guideline-recommended care was within 1 year of the index date. We operationalized the five indicators based on internal clinician input, study team experience and guidelines from the Practical Approach to Detection and Management of Chronic Kidney Disease for the Primary Care Clinician [15]. For HbA1c monitoring, the ADA guidelines [16] recommend that HbA1c should be measured at least twice a year when meeting treatment goals. Since we could not measure actual glycemic targets/HbA1c values for all patients in the study we used 2 measures of monitoring to consider someone as regularly following up and monitoring their glycemic targets. For adherence, typically a threshold of 80% proportion of days covered is considered adherent (i.e., 292/365 days). The study team selected no less than 60 days (or 2 months) per year which would translate to a PDC of 83%. For clinical care, we selected at least 3 nephrologist visits for high renal risk and at least 2 PCP visits among low renal risk based on the practical approach to management of CKD suggested by Vassalotti et al. [15].

Patients with end-stage renal disease diagnosis codes on inpatient or outpatient medical claims or procedure codes indicating receipt of dialysis, recipients of kidney transplant, and patients who elected hospice care were excluded. Figure 1 displays the study time frame.

HCRU was measured by inpatient (IP) hospital admissions and emergency department (ED) visits during the 12 months postindex using facility claims. IP admissions were identified based on bill type code (11) and revenue codes (0190–0219). IP episodes were delineated by admit and discharge cates recorded on the claims. ED visits were identified using revenue codes (045x), place of service codes (23), and Current Procedural Terminology (CPT)/Healthcare Common Procedure Coding System (HCPCS) (CPT/HCPCS) codes indicating procedures performed in the ED setting. For both IP and ED, we created dichotomized outcomes indicating any use during the 12-month follow-up period. For cost outcomes, we calculated total medical costs by summing up allowed costs for all paid medical claims, and we calculated total pharmacy costs by summing up prescription costs for all paid pharmacy claims. Total healthcare costs were calculated by summing up medical and pharamacy costs.

Baseline demographic and clinical characteristics were summarized for the study sample. Means and standard deviations were reported for continuous variables and counts and percentages for categorical variables. Regression analyses were conducted to examine the association between disease management and outcomes, while controlling for patient demographic characteristics, comorbidities, and CKD stage. Logistic models were used for HCRU outcomes, and generalized linear models with a gamma distribution and log link function were used for cost outcomes. All analyses were conducted separately for the four sub-cohorts: patients with both diabetes and hypertension, diabetes only, hypertension only, or neither. Statistical significance was assessed at the conventional level of 0.05.

Since health plan administrative databases are built primarily for billing and reimbursement purposes, clinical information extracted from these databases may be incomplete or inaccurate. Certain diagnostic information documented in medical records may not be fully captured in submitted claims. As a result, CKD cases, comorbid conditions, and/or outcomes could be undercounted or misclassified. Without access to medical records, it is difficult to gauge the extent of such occurrences. However, since we were comparing outcomes between patient groups and it is unlikely these billing errors or omissions would differ systematically between groups, the impact on our results should not result in systematic bias. Potential misclassification of patients into the low renal risk category due to limited lab data on albuminuria and glomerular filtration rate can occur. Additionally, disease management was operationally defined using process measures (e.g., HbA1c monitoring) rather than actual clinical outcomes (e.g., HbA1c control).​ We utilized the older CKD-EPI equation [17]. given the study period preceded 2021, however, future analysis of post-2021 data should utilize the 2021 CKD-EPI equation [27]. No causal inference can be ascertained from this study, as it was an observational design using retrospective claims data. The results may reflect healthcare patterns unique to Humana beneficiaries. This study utilized data for the Humana MAPD population only, so the results may not be generalizable to individuals with Traditional Medicare or to a general adult population. However, Humana is a large national health plan with members residing in a broad array of geographic regions.