We collected data from 21,572 hospitalized patients aged 18 years or older whose serum creatinine (Cr) levels were identified once or more during their first admission for a period of 1 year (2013) at a tertiary care hospital in Seongnam, Korea. All admissions through general practice, accident, and emergency were included. Among them, 21,172 patients were deemed eligible for this study after excluding those with missing BP data (n = 91), pre-existing end-stage renal disease (n = 309), and intensive care unit admission during hospitalization (n = 3,048). Finally, we enrolled only 1,612 patients diagnosed with AKI according to the KDIGO definition based on serum Cr measurements (Fig. 1). All clinical investigations were conducted with respect to the 2008 Declaration of Helsinki and good clinical practice guidelines. This study was approved by the Institutional Review Board (IRB) of Seoul National University Bundang Hospital (IRB number: H-1508-310-115), with no written consent required because patient records/information were anonymized and de-identified prior to analysis.

Demographic, physiological, and laboratory data at the index admission were collected from an electronic medical records database. Serum Cr values were measured via the rate-blanked compensated kinetic alkaline picrate Jaffe kinetic method using an automatic analyzer (Toshiba-200FR; Toshiba Corp., Tokyo, Japan). The estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease (MDRD) study equation [18] adjusted for a Korean ethnic coefficient [19]. To detect AKI, at least two Cr measurements were required: First taken during the baseline time window and a second recorded on a different day during hospital admission. We defined the baseline Cr level using the 6-month Cr data from before patients’ admission, as this directly affected the detection and staging of AKI, in the manner as follows: By selecting the lowest Cr value measured less than 90 days prior to admission. If this value was unavailable, the lowest Cr value measured between 90 and 180 days before admission was considered. If this value was also unavailable, the Cr level was estimated using the MDRD study equation [18] adjusted for a Korean ethnic coefficient [19], assuming that the baseline eGFR was 75 mL/min/1.73 m² [20]. Inpatient Cr measurements were defined as those obtained during admission. Using both a baseline and an inpatient Cr value measured on different days, AKI was defined as an increase of 0.3 mg/dL or more in or of 1.5 times or more of the serum Cr, using the serum Cr criteria proposed by the KDIGO [20]. Severe AKI was defined as stage III of the KDIGO using the serum Cr criteria (threefold increase in the serum Cr from baseline or increase in serum Cr to 4.0 mg/dL or more). We defined AKI solely based on changes in the measured serum Cr values because urine output data were not consistently available for all inpatients. Bacteremia was elucidated via positive blood culture during hospitalization. BacT/Alert FA and FN bottles (bio-Mérieux, Durham, NC, USA) were used for all blood cultures.

BP measurements were conducted using an electronic sphygmomanometer at the discretion of the nurse during the hospitalization period. For the present analysis, the average SBP was calculated for each patient within 48 hours after the development of AKI. The risk of mortality or severe AKI was then evaluated as a function of BP. BP values were categorized into 10-mmHg increments to determine the association with mortality or severe AKI. Because the average follow-up BP variables represented the effects of BP over a period of time rather than at a single point in time, we considered this to be more important in predicting the outcomes.

We assessed the comorbidities based on the code of the International Classification of Disease, 10th Revision (ICD-10) as follows: cardiovascular disease including angina pectoris (I20), myocardial infarction (I21–I23), ischemic heart disease (I24–I25), and heart failure (I50); cancer (C); cerebrovascular disease (I63–I66); hypertension (I10–I15); and diabetes mellitus (E10–14). The presence of hypertension and diabetes mellitus were also confirmed by the use of antihypertensive medications and antihyperglycemic agents, respectively. Antihypertensive drugs included alpha- and beta-blockers, calcium-channel blockers, renin–angiotensin system inhibitors, and diuretics.

The composite outcome of 90-day all-cause mortality or severe AKI was considered in this research. We combined 90-day data from the Ministry of Security and Public Administration with our dataset using each individual’s unique identifier. Mortality data were obtained until August 31, 2015.

Continuous variables were expressed as means ± standard deviations, while categorical variables were expressed as percentages. Differences in continuous variables were analyzed using one-way analysis of variance or the Kruskal–Wallis rank test, whereas chi-square tests and Fisher’s exact test were used for categorical variables. Cox’s hazard proportion analysis was used to estimate the hazard ratios (HRs) for composite outcome according to the average SBP. We conducted the tests of proportional hazards assumptions and restricted cubic spline curves, respectively. The HRs and 95% confidence intervals (CIs) for the composite outcomes were calculated after a stepwise adjustment for multiple confounders including age, gender, body mass index, Charlson’s comorbidity index, hypertension, operation, bacteremia, white blood cell count, hemoglobin, albumin, cholesterol, alanine aminotransferase, sodium, total CO₂, and maximal Cr during admission.

The adjusted HR for each category of SBP was calculated based on the BP group in which the event rate was lowest, for which the HR was considered as 1. P values of less than 0.05 were considered to be statistically significant. All analyses and calculations were conducted using the IBM SPSS Statistics version 24.0 (IBM Corp., Armonk, NY, USA) and STATA version 14.0 (StataCorp LP, College Station, TX, USA) software programs.

Among the 1,612 patients, the mean age, Charlson’s comorbidity index value, and serum Cr levels at baseline were 65.8 ± 16.0 years, 2.8 ± 2.2, and 1.00 ± 1.09 mg/dL, respectively. The average SBP/diastolic BP (DBP) within 48 hours after admission and within 48 hours after AKI were 125.8/69.9 mmHg and 123.8/68.3 mmHg, respectively. Of the 1,612 patients, 55.4% were males, 30.1% had hypertension, and 35.6% had diabetes mellitus. The median duration of AKI development was 1.0 (interquartile range: 0–4.0) days after hospital admission. Additionally, the median duration of hospital stay and mean duration of total follow-up were 9.0 (interquartile range: 5–16) and 76.7 days, respectively. In total, 467 patients (29.0%) underwent an operation and 133 patients (8.3%) suffered from bacteremia episodes during admission (Table 1).

Table 2 shows the demographic and baseline characteristics according to SBP categories within 48 hours after AKI. Patients with low SBP were younger and leaner; less likely to have previous hypertension; had high levels of total bilirubin but low levels of baseline Cr, hemoglobin, albumin, cholesterol, and total CO₂; and had a high incidence of bacteremia.

Among the 1,612 patients, 90-day mortality rate, severe AKI rate, and composite outcome rate for AKI inpatients were 11.5% (n = 185), 10.1% (n = 163), and 18.7% (n = 302), respectively.

In the composite outcome group, a higher proportion of patients had diabetes mellitus and cancer as compared with in the group without the composite outcome. Additionally, the composite outcome group was older, leaner; had a higher white blood cell count and maximal Cr level during admission but low levels of albumin, sodium, total CO₂, and hemoglobin; and had a high incidence of bacteremia and low incidence of operation. The relationship between SBP and composite outcome followed a U-shaped curve, with an increased event rate at both low and high SBP values (Fig. 2, 3). The average SBP after AKI predicted the composite outcome after adjusting for baseline variables (reference SBP: 120–129 mmHg; < 100 mmHg: HR 1.84, P = 0.015; 100–109 mmHg: HR 1.56, P = 0.038; 110–119 mmHg: HR 1.15, P = 0.483; 130–139 mmHg: HR 1.51, P = 0.045; ≥ 140 mmHg: HR 1.73, P = 0.005). The curve was relatively flat for SBPs of 110 to 130 mmHg after AKI (Table 3). Separately, bias-corrected confidence intervals using bootstrap analysis (n = 10,000) to validate the SBP groups after AKI in predicting composite outcomes were calculated (reference SBP: 120–129 mmHg; < 100 mmHg, 1.047–3.246; 100–109 mmHg, 1.014–2.391; 110–119 mmHg, 0.770–1.721; 130–139 mmHg, 0.972–2.360; and ≥ 140 mmHg, 1.063–2.829).

We performed a subgroup analysis according to the underlying disease to evaluate the effects of BP categories on the composite outcome in patients with AKI. Supplementary Tables 1 and 2 (available online) display the relation between BP and composite outcomes across the patient comorbidities. In 485 hypertensive patients, the proportion taking antihypertensive medication is shown in Table 1. In these patients, the risk of composite outcome increased 2.5-fold in the group with SBPs of less than 100 mmHg as compared with the reference group (SBP of 120–129 mmHg). In contrast, the relationship between SBP and composite outcome significantly followed a J-shaped curve, with an increased event rate observed at high SBP values in the nonhypertensive patients (reference SBP: 120–129 mmHg; 130–139 mmHg: HR 2.125; 95% CI, 1.301–3.469; ≥ 140 mmHg: HR 2.485; 95% CI, 1.489–4.147).

Diabetic patients showed a similar pattern of BP and composite outcome as that of nonhypertensive patients. In nondiabetic patients, the risk of composite outcome increased 2.1-fold in the group with SBPs of less than 100 mmHg in comparison with the reference group (SBP 120–130 mmHg). A difference in the relationship between SBP and the composite outcome was observed, depending on the underlying disease.