Prediction Models for CKD Following AKI Developed and Validated

Hospitalized patients who experience acute kidney injury (AKI) are at increased risk for poor long-term outcomes. Results of studies conducted between 2008 and 2017 demonstrated that while usually reversible, in some patients AKI results in incomplete recovery of kidney function and in others accelerated loss of kidney function and an increased risk of chronic kidney disease (CKD). Current practice guidelines call for follow-up of patients with AKI within 3 months to determine the presence of CKD and plans for subsequent treatment.

However, according to Matthew T. James, MD, PhD, and colleagues, many patients do not receive a follow-up assessment or appropriate care for those whose kidney function has not recovered. Using population-based laboratory and administrative data from two Canadian provinces, the researchers developed and validated risk prediction models for advanced CKD following hospital-related AKI. The researchers aimed to develop a practical risk stratification approach for identification of patients at high risk of CKD following hospital discharge. Study results were reported in JAMA [2017;318(18):1787-1797].

Primary outcomes of interest were advanced CKD defined by a sustained reduction in estimated glomerular filtration rate (eGFR) <30 mL/min/1.73 m2 for at least 3 months during the year following discharge or initiation of chronic dialysis up to 1 year following initial discharge. Follow-up for all patients was 1 year.

The risk models were developed using data from 9973 patients hospitalized in Alberta, Canada from April 2004 to March 2014, with follow-up to March 2015. A total of 4985 patients were included in an internal validation cohort. Data from a cohort of 2761 patients hospitalized in Ontario, Canada, from June 2004 to March 2012, with follow-up to March 2013, were used to externally validate the risk models. Eligible patients had prehospitalization eGFR >45 mL/min/1.73 m2 and survived hospitalization with AKI (defined as increase in serum creatinine during hospitalization of >0.3 mg/dL or >50% of prehospitalization baseline level).

Among the 14,958 patients in the total cohort, mean age ranged from 66 to 69 years. Mean baseline serum creatinine value was 1.0 mg/dL, obtained a median of 66 days prior to admission in the derivation and internal validation cohorts and 44 days in the external validation cohort. More than 20% had stage 2 or 3 AKI, and one-third of patients had serum creatinine value of ≥1.3 mg/dL at hospital discharge.

Patients in the external validation cohort were slightly older and more likely to have received cardiovascular procedures or mechanical ventilation than those in the derivation and internal validation cohorts. Twenty-nine patients in the derivation and internal validation cohorts had only one eGFR <30 mL/min/1.73 m2 and did not have subsequent measurements during the follow-up period; those patients were not considered to have developed advanced CKD.

In the derivation and internal validation cohorts, 2.7% (n=408) developed the outcome of sustained eGFR of <30 mL/min/1.73 m2 following hospital discharge. Five point two percent (n=775) had an eGFR <30 mL/min/1.73 m2; of those, 22.2% (n=172) later developed advanced CKD (42.2% of the total number of advanced CKD events). In the external validation cohort, a similar proportion of patients experienced the outcome of sustained eGFR of <30 mL/min/1.73 m2 following hospital discharge.

In bootstrapped samples of the derivation cohort, there were six variables that were association with a higher risk of progression to advanced CKD: older age, female sex, higher baseline serum creatinine values, albuminuria, greater severity of AKI, and higher serum creatinine values at hospital discharge. The six-variable model had the highest C statistic and lowest Akaike Information Criterion.

Based on the predicted risk using model 1, 52.4% of patients in the derivation cohort (n=5228) had <1% risk of developing CKD; 35.2% (n=3512) had 1% to <5%; 6.2% (n=626) had 5% to <10%; 3.6% (n=360) had 10% to <20%; and 2.5% (n=247) had ≥20% risk of developing CKD. The lowest C statistic was seen with model 5 that included age, sex, and AKI stage only (C statistic, -.71; 95% confidence interval [CI], 0.67-0.74).

The 6-variable model remained well calibrated in the internal validation cohort; the model had a higher C statistic compared with models with fewer variables (0.87; 95% CI, 0.84-0.90). The 6-variable model also remained well calibrated in the external validation cohort; the C statistic was 0.81 (95% CI, 0.75-0.86), and the model had improved discrimination and reclassification compared with models that included age, sex, and serum creatinine value at discharge alone.

The authors cited some limitations to the study, including identification of candidate variables from secondary analysis of data, which did not include all potential risk factors for CKD such as blood pressure, urine sediment, cause of AKI, and details of dialysis treatment. Further, the possibility exists that patients at low risk for AKC were systematically excluded from the study due to lack of follow-up creatinine testing. In addition, the models were derived and validated in cohorts from Canada; there was no examination of generalizability of the findings to patients from other areas.

In conclusion, the researchers said, “A multivariable model using routine laboratory data was able to predict advanced chronic kidney disease following hospitalization with acute kidney injury. The utility of this model in clinical care requires further research.”

Takeaway Points

  1. Researchers in Canada sought to derive and validate a model to predict the possibility of progression from acute kidney injury (AKI) to advanced chronic kidney disease.
  2. Using data from 9973 patients hospitalized in Alberta, Canada, the researchers developed a model based on six variables associated with higher risk of progression to advanced CKD: older age, female sex, higher baseline serum creatinine values, albuminuria, greater severity of AKI, and higher serum creatinine values at discharge.
  3. The model was validated in an external cohort of 2761 patients hospitalized in Ontario, Canada.