Incidence of Serum Creatinine Monitoring and Outpatient Visit Follow-Up among Acute Kidney Injury Survivors after Discharge: A Population-Based Cohort Study

Acute kidney injury (AKI) affects 20% of hospitalized patients and is associated with poor health outcomes [1-3]. Even among the 80–90% of hospitalized patients who survive an AKI episode, long-term morbidity is increased. Chronic kidney disease (CKD), end-stage kidney disease, [4-7] and cardiovascular diseases [8] are more common in AKI survivors than in comparable patients without AKI. To limit the burden of long-term complications in AKI survivors, follow-up care is recommended [9, 10].

Epidemiological studies of AKI follow-up have focused on select sub-populations, including Medicare and Medicaid patients and those affiliated with a Veteran’s Affairs program; however, these findings may not be broadly generalizable [11, 12]. For example, the US Renal Data System data suggest that 85% of Medicare patients have a follow-up creatinine test in 6 months after discharge. Insurance claims data from a large US healthcare company indicate the frequency is only 54% [11]. A population-based study in a geographically defined region may provide additional insights into AKI follow-up.

Determining factors that predict follow-up for AKI survivors is also needed to identify individuals potentially vulnerable to gaps in care. High-risk conditions, such as diabetes and/or CKD may lead to differences in involvement with follow-up care [11, 12]. In contexts unrelated to AKI, social and behavioral determinants of health influence follow-up [13, 14]. It is unclear whether factors related to acuity of illness or sociodemographic factors predominately influence AKI follow-up. Once identified, these factors may be used to target patient subgroups likely to benefit from systems-based interventions to enhance kidney care follow-up [15, 16].

Thus, the purpose of this study was to use the robust population-based longitudinal data available in the Rochester Epidemiology Project (REP) to characterize the follow-up care of nearly all local residents with stage II or III AKI dismissed from the hospital, not on dialysis who survived for at least 30 days. We described the cumulative incidence of an outpatient serum creatinine (SCr) check and/or an in-person healthcare visit in the 1 year after discharge and predictors of these core elements of follow-up care.

This was a population-based cohort study of adults (≥18 years of age) hospitalized at Mayo Clinic in Rochester, MN, who experienced an episode of AKI. Mayo Clinic is the only provider of Nephrology subspecialty care (including dialysis) in Olmsted County. Patients were identified and followed up using the REP. The REP is a medical record linkage system that amasses data from virtually all patients and healthcare providers (Mayo Clinic and non-Mayo Clinic including county medical facilities, private practices, regional hospitals, and nursing homes) in Olmsted County, Minnesota. While local residents may seek care outside of data captured by the REP, primary care, specialty care, and acute care services are available in Olmsted County which decreases the likelihood. Additional detail on the REP infrastructure is available in the detailed overview from St Sauver et al. [17]. Women comprise approximately 50% of the population of Olmsted County, and approximately 17% of Olmsted County residents are non-Caucasian. The REP includes demographic data and comprehensive information about medical diagnoses, hospital admissions, and outpatient follow-up care regardless of where that care was delivered or patient insurance status [17-19].

Included individuals were Olmsted County residents admitted to Mayo Clinic Hospital between January 02 2006 and December 31 2014 who experienced stage II or stage III AKI [20]. The cohort was created as part of development and validation of an AKI electronic surveillance tool based on KDIGO AKI criteria [20, 21]. Briefly, the surveillance tool screened electronic patient health records at 15 min intervals from intensive care unit admission to hospital discharge. The tool was optimized to detect any rise in SCr or decline in urine output during that timeframe. During tool development, cases of AKI were adjudicated independently in duplicate using manual chart review. The cohort remains open for longitudinal clinical outcomes follow-up and was queried secondarily for this evaluation of the short-term processes of care. For the present study only those identified with stage II or III AKI (i.e., at least a doubling of SCr from baseline or a urine output <0.5 mL/kg/hr for ≥12 h) were included. The preadmission baseline creatinine for this tool was the median of all values in the 6-months to −7 days before the index admission. If unavailable, this value was estimated using an assumed estimated glomerular filtration rate (eGFR) of 60 mL/min/1.73 m² [22]. Time 0 for this evaluation was the date of hospital discharge. Patients were followed forward in time for up to 1 year using longitudinal follow-up data from the REP to capture post-discharge care [23]. Excluded individuals were those on dialysis at admission, dismissed on dialysis, or who required dialysis in the 30 days after discharge. Individuals with a kidney transplant before, during, or in the 30 days after the index hospital admission, who were discharged on hospice care, who died within 30 days, or who declined the use of their medical record for research were also excluded [24]. Institutional Review Board approval was obtained from both Mayo Clinic (IRB 18-007889) and Olmsted Medical Center (032-OMC-18) and the need for written informed consent was waived.

Preadmission data captured included age, sex, race/ethnicity, and education level. Socioeconomic status was determined using the previously validated HOUsing-based SocioEconomic Status index (HOUSES index) [25, 26]. The HOUSES index uses publically available housing data (e.g., square footage, building value, and ownership of the housing unit) to construct an individual rather than an area-based measure of socioeconomic status. Comorbid conditions were identified using ICD 9/10 codes (online suppl. Table 1; for all online suppl. material, see www.karger.com/doi/10.1159/000519375). Preadmission and dismissal eGFRs were calculated with the CKD EPI eGFR_creatinine equation (mL/min/1.73 m²) [27]. Hospital length of stay, maximum AKI severity (stage II or III), and degree of recovery by discharge were collected. Degree of recovery was classified as recovered or stage 0 (SCr <1.5x baseline, but not back to baseline level), stage 1 (1.5–1.9x baseline), stage 2 (2–2.9x baseline), and stage 3 (3x baseline or ≥4 mg/dL) [28].

The primary outcome of interest was the cumulative incidence of both outpatient SCr level and an in person healthcare visit with a physician or advanced practice provider within 30 days, 90 days, or 1 year after discharge. Secondarily, we described the cumulative incidence of either a SCr level or an in person healthcare visit at these time points. To identify completion of an outpatient visit, the date of the first outpatient diagnosis code(s) billed after discharge was identified. Manual validation of 10% of the records was performed to ensure the accuracy of this method and confirm that the codes corresponded to the visit date of interest. An additional analysis was performed to evaluate the healthcare visit type. We captured any nephrology visit within the follow-up period. Patients were censored at the time of their last clinical encounter (recorded in the REP) or 1 year (the end of the follow-up period).

Cumulative incidence of follow-up with the 95% confidence interval (CI) was estimated using the Aalen-Johansen method, where death, end-stage kidney disease, or kidney transplantation were considered competing risks. Rates were described at 30 days, 90 days, and 1 year after discharge. Cox proportional hazards models were used to assess the association of preadmission or hospitalization variables and follow-up care. Predictors of interest included age, sex, race/ethnicity, self-reported education level, socioeconomic status, body mass index, comorbidity burden described with the Charlson Comorbidity Index, preadmission kidney function, AKI stage (II or III), and discharge eGFR. The occurrence of a hospital readmission during the follow-up interval could increase the likelihood of the follow-up outcome. Thus, hospital readmission was explored as a time-dependent variable, where all patients started in the no readmission group, and at the time of readmission (assuming this occurred prior to the outpatient follow-up event) the patient was moved to the readmission group. Variables satisfied the proportional hazards assumption. Multivariable models were constructed via the least absolute shrinkage and selection operator regression. Appreciating that the degree and intensity of follow-up may differ depending on the discharge disposition, a sensitivity analysis excluded individuals discharged to a skilled nursing facility. A sensitivity analysis was also performed for individuals where the preadmission SCr was observed not estimated. All analyses were performed using SAS version 9.4 software (SAS Institute, Inc., Cary, NC, USA) and R version 3.6.2 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria).

Of the 879 individuals who experienced an episode of stage II or III AKI during their hospitalization and survived to hospital discharge, 627 were included in the study cohort (Fig. 1; stage II: N = 400 [64%]; and stage III: N = 227 [36%]). The primary reason for exclusion was the use of dialysis at hospital discharge (N = 118). Included individuals were of a mean age of 68 years; 46% were men and 89% self-identified as non-Hispanic Caucasian (Table 1). The mean ± standard deviation preadmission eGFR was 63 ± 24 mL/min/1.73 m² and 379 (61%) had CKD. The median change in eGFR from preadmission to discharge was −31% (IQR, −56–0%). Patients with stage III AKI during hospitalization were generally similar to those with stage II AKI with the exception of a lower discharge eGFR in patients with stage III AKI (37 ± 27 vs. 51 ± 27, respectively; p = < 0.001; online suppl. Table 2). No patients experienced end-stage kidney disease or kidney transplantation before the endpoints of interest; thus, these were not ultimately included as competing risks in endpoint analyses.

For the primary endpoint, the cumulative incidence of follow-up with an outpatient SCr level and a healthcare visit at 30 days was 70% (95% CI: 66 and 73%) (Table 2). Incidence of follow-up increased over the 1 year study time period to 91%. Forty-eight individuals had no outpatient SCr or healthcare visit within 30 days. Within 30 days, those with both an outpatient SCr and a healthcare visit had a median of 2 (IQR 1–5) outpatient SCr assessments and 3 (IQR 2–4) visits.

Several factors significantly predicted the completion of an outpatient SCr and a healthcare visit in univariate analyses (online suppl. Table 3). In multivariable models, lower discharge eGFR, higher Charlson Comorbidity Index, longer length of hospitalization, and greater maximum severity of AKI predicted completion of both follow-up care elements (Table 3; Fig. 2). None of the sociodemographic factors in Table 1 independently predicted follow-up.

In secondary analyses, the cumulative incidence of an outpatient SCr level was 80% (95% CI 76 and, 83%) at 30 days. For provider visits, the cumulative incidence was 82% (95% CI: 79 and 85%) at 30 days. Predictive models for follow-up SCr or follow-up provider visit were generally similar to when both follow-up elements were combined, but sample sizes were more limited (Table 3; online suppl. Fig. 1, 2). An intensive care unit stay during hospitalization significantly predicted creatinine evaluation at 30 days (p = 0.022), 90 days (p = 0.015), and 1 year (p = 0.015). Within 1 year, 23% of individuals with a follow-up healthcare visit (n/N = 133/578) saw a nephrologist (online suppl. Table 4).132 (23%) individuals completed both outpatient SCr and saw a nephrologist within 1 year, of which 113 (86%) had CKD. The median time to outpatient nephrology consultation was 21 (IQR 7–54) days. Among patients with preadmission diabetes or CKD, 29% saw a nephrologist within 1 year.

A series of sensitivity analyses were performed including an evaluation of an SCr and a healthcare visit where the visit was only counted if it was with a primary care provider or nephrologist, an analysis excluding the 189 individuals discharged to a skilled nursing facility or an acute care hospital, an analysis excluding the 211 individuals with an estimated baseline creatinine value, and stratified evaluations according to AKI severity, degree of recovery by discharge, and baseline CKD. The findings were generally similar in these analyses to the overall group. The frequency of follow-up was higher in patients with more severe AKI, nonrecovery, and baseline CKD. When non-nephrology specialty care visits were not included in the follow-up visit definition, incidence of a healthcare visit (primary care/nephrology) and SCr fell from 70 to 59% at 30 days (Table 2).

In this population-based cohort study, one-third of AKI survivors lacked follow-up in the 30 days after hospital discharge. Among those who only had 1 element of follow-up care, the SCr was more often missing than the healthcare visit. By 1 year, 88% of individuals experienced an outpatient SCr check and a healthcare visit. Non-nephrologists oversaw the majority of healthcare visits (90%). Sociodemographic factors did not predict follow-up care. Instead, factors related to illness severity, including a worse severity of AKI, lower discharge eGFR, longer length of stay, and greater comorbidity burden independently predicted a higher likelihood of follow-up.

There is increased recognition of the importance of AKI as a contributor to long-term kidney and cardiovascular outcomes [29]. Kidney follow-up care is recommended during the period of vulnerability after an episode of AKI [10, 28]. Delivery of this follow-up care for AKI survivors remains suboptimal. The US Renal Data System data suggest that follow-up laboratory testing in AKI survivors, including a SCr or urine albumin evaluation within 6-months of discharge, only occurs between 54 and 14% of patients, respectively [11]. Primary care visits occur in 75% of Medicare and Medicaid patients within 3 months, but in other cohorts, the frequency is much lower. In Medicare and Medicaid patients and those affiliated with a Veteran’s Affairs program, estimates indicate that only 4–17% of AKI survivors visit nephrologists by 1 year [11, 12]. In the present study, we comprehensively characterized the adequacy of follow-up using a population-based cohort from Olmsted County, MN. Our cohort is slightly younger, with better baseline kidney function and fewer comorbidities than previous patient subsets [11, 12]. AKI in this study was defined by biochemical markers, SCr, and urine output, rather than diagnosis codes, which facilitates the detection and inclusion of survivors with less severe AKI during hospitalization [30]. We demonstrated that one-third of AKI survivors lacked an outpatient SCr assessment and a healthcare visit in the 1 month after follow-up. This improved over time, but even out to 1 year after the episode of moderate to severe AKI, 12% of patients lacked both outpatient SCr evaluation and a follow-up healthcare visit.

An important finding in the present study was the infrequency with which nephrology follow-up occurred for AKI survivors. We demonstrated that only 23% of individuals with a follow-up healthcare visit within 1 year saw a nephrologist. If a patient was to see the nephrologist, it was relatively close to the time of discharge (median 21 days [IQR 7–54]). Among patients with diabetes or CKD, only one-third of our cohort saw a nephrologist within 1 year. These data are similar to the 8.5% cumulative incidence of nephrology referral in 1 year observed in AKI survivors in a Department of Veterans Affairs dataset [12]. Even in AKI survivors who required dialysis during hospitalization, population-level data from Taiwan indicated follow-up by a nephrologist occurred in only 37% of patients [31]. This gap in follow-up care is despite evidence that nephrologists would like to see these high-risk patients in the clinic [32] and when they do, there is an association with better outcomes [31, 33]. Several factors likely contribute to the low frequency of nephrology follow-up even among very high-risk AKI survivors. AKI is often underdiagnosed in hospitalized patients and underreported on dismissal summaries [34]. Interviews with AKI survivors indicate that many factors limit engagement with post-AKI care, including prioritizing other health concerns, lack of education about the significance of AKI, anxiety from competing health demands, and complex discharge plans [35]. A proportion of patients may be referred for specialty follow-up care, but cancel for these or other reasons. Further, nephrology follow-up is a sparse resource that may not be widely accessible to all patients, especially those in rural communities [36, 37]. Although the observational data suggest a health benefit for AKI survivors when seen by nephrology, it remains to be determined whether structured kidney health follow-up delivered by trained individuals in primary care could reach patients not seen by nephrologists with a similar health benefit [9].

To identify individuals at high risk for gaps in care, we conducted a multivariable analysis of factors that might predict patient follow-up. These data reveal that a greater comorbidity burden was associated with a higher incidence of a SCr assessment and a healthcare visit within 1 year. This reinforces the hypothesis that greater medical complexity portends clinical follow-up. Among the other factors evaluated, increased severity of AKI during the hospitalization, longer length of stay, and lower discharge eGFR also predicted follow-up. Multivariable risk factors were generally consistent at 30 days, 90 days, and 1 year. We did not observe any association between age, sex, race/ethnicity, education level, or socioeconomic status with post-hospital follow-up among AKI survivors. Our sample comprised 89% non-Hispanic Caucasian individuals, with a relatively high baseline education level and socioeconomic status, which likely affected the observed follow-up patterns. In previous prognostic studies, non-Caucasian race/ethnicity and worse socioeconomic status were associated with a higher AKI incidence and worse long-term outcomes [38-40].

Nephrology consensus groups recommend follow-up for all AKI survivors but indicate that it should be graded in urgency and intensity based on the severity of AKI, duration of injury, degree of comorbidities, and magnitude of renal recovery by discharge [10]. Our data suggest existing approaches to care delivery moderately achieve these follow-up goals. Based on these findings, individuals at high risk for experiencing gaps in kidney care follow-up are those with less severe AKI (stage II), fewer comorbidities, short hospital stays, and improved GFR by discharge. While this lower risk group likely would benefit from post-discharge monitoring of kidney health (i.e., SCr, medication reconciliation and counseling, and blood pressure evaluation), these care elements may be able to be delivered by a primary care provider rather than a specialist.

This study has several limitations that must be accounted for when interpreting these findings. Consensus recommendations suggest AKI follow-up should generally include kidney function assessment, patient education, medication review and reconciliation, fluid status and blood pressure evaluation, and self-management counseling during acute illness (i.e., sick-day counseling) [10]. The focus of this evaluation was on the SCr assessment and healthcare visit components. Other laboratory tests beyond SCr may be warranted for AKI follow-up, but data suggest these are performed even less frequently than the SCr [11]. Moreover, we cannot assess the interpretation of and reaction to the results of kidney monitoring tests. Access to the SCr data does not necessarily result in a change to the care provided, especially if clinicians were unaware of the patient’s AKI episode. A creatinine evaluation may be included as part of a broader panel of laboratory data, but evaluation of its results may not be the care team’s focus. Not all healthcare visits likely provide the degree of kidney care evaluation (nephrology services, non-nephrology specialty referral, and primary care). Still, we feel that at least 1 encounter with a healthcare provider lends an opportunity for a kidney health review. In a sensitivity analysis where visits were defined only as with a primary care provider or a nephrologist, the cumulative incidence of follow-up with a SCr and a visit within 30 days fell from 70% in the main analysis to 59%. In the 6 years that have passed since the end of the study follow-up period, an increased emphasis has been placed on the importance of kidney follow-up care [10]. Further, the advent of telehealth has increased access to follow-up care [37]. It is possible that a more contemporary cohort might reveal an improved incidence of follow-up. We recognize that 30% of our patients were dismissed to skilled nursing facilities or acute care hospitals, where nursing care and clinical evaluations that constitute follow-up may have gone undocumented in the electronic health record. A subgroup analysis which excluded these individuals the results were similar. While the included individuals represent the demographics of southeast Minnesota, the sample was mostly non-Hispanic Caucasians with a higher baseline education level which could affect generalizability of these findings. One challenge in epidemiological studies of this nature is how to handle patients who die or are rehospitalized during the follow-up period. Those that die within 30 days often have a care plan designed to focus on comfort and quality of life. Laboratory monitoring or healthcare visits may not be consistent with these goals of care. These individuals were thus excluded. Death in the included cohort was otherwise modeled as a competing risk. At a rehospitalization, the potential for a future healthcare visit or creatinine assessment could be affected. To address this limitation, we modeled rehospitalization during the follow-up window as a time-dependent variable and restricted the endpoint to include only outpatient creatinine values and outpatient visits. Finally, further studies are needed to determine the impact of post-AKI follow-up on clinical outcomes such as CKD.

In this population-based cohort study, we demonstrated that 30% of patients with moderate to severe AKI failed to receive follow-up in the 30-day post-discharge interval. Medical risk factors, rather than social/demographic characteristics, were the primary determinants of follow-up. More work is needed to enhance care delivery in AKI survivors.

This study protocol was reviewed and approved by Institutional Review Boards from both Mayo Clinic (#18-007889) and Olmsted Medical Center (#032-OMC-18). All included individuals consented to the use of their medical records for research.

Dr. Barreto reports consultancy for FAST Biomedical, unrelated.

This project was supported in part by the Mayo Clinic Department of Pharmacy, the National Institutes of Health National Center for Advancing Translational Sciences (NCATS) under Award No. UL1 TR002377, the National Institute on Aging under Award No. R01AG034676, the National Institute of Allergy and Infectious Diseases of the under Award No. K23AI143882 (PI; EFB), and the Agency for Healthcare Research and Quality HS028060-01 (PI; EFB). The funding sources had no role in study design; data collection, analysis, or interpretation; writing the report; or the decision to submit the report for publication. Its contents are solely the authors’ responsibility and do not necessarily represent the official views of the NIH.

E.B., D.S., K.M., and A.R. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. E.B., D.S., H.M., K.M., A.C., K.K., S.K.G. and A.R. designed and conducted the study. E.B., D.S., K.M., S.P., C.W., V.S. gathered data. K.M. performed the statistical analysis which was reviewed by E.B., A.C., and A.R. All authors reviewed the data and participated in discussions related to interpretation. E.B. drafted the manuscript. All authors reviewed and edited the manuscript and have seen and approved the final draft.

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants but are available from the corresponding author E.F.B. upon reasonable request.