Summary
Background and objectives
The Agency for Healthcare and Research Quality patient safety indicators track adverse safety events in hospitalized patients but overlook safety incidents specific to CKD. This study considers candidate CKD-pertinent patient safety indicators and compares them with the Agency for Healthcare and Research Quality patient safety indicators.
Design, setting, participants, & measurements
Using a national Veterans Health Administration database of hospitalized veterans from fiscal year 2005, 247,160 hospitalized veterans with prehospitalization measures of renal function were retrospectively examined for proposed CKD patient safety indicators versus Agency for Healthcare and Research Quality patient safety indicators using International Classification of Diseases, Ninth Revision diagnosis codes. Candidate CKD-pertinent patient safety indicators included in-hospital acute kidney failure; in-hospital congestive heart failure (and related diagnostic codes); electrolyte disturbances; and medication errors, poisoning, and intoxication. Patients with a prehospital estimated GFR<60 ml/min per 1.73 m2 (CKD group) were compared with a non-CKD group. For CKD patient safety indicators, hospitalizations were excluded if the admitting condition was a potential cause of the secondary condition. Regression methods were used to present adjusted rates in study groups of interest.
Results
The CKD patient safety indicators were generally more common than the Agency for Healthcare and Research Quality patient safety indicators in all groups, tended to occur in different patients than those patients who experienced Agency for Healthcare and Research Quality patient safety indicators, and were more common in the CKD group than the non-CKD group, except for hypoglycemia, hypokalemia, and hyponatremia. The adjusted composite CKD patient safety indicators rate (per 1000 patient-hospitalizations) was 398.0 (95% confidence interval, 391.2 to 405.0) for patients in the CKD group and 250.0 (95% confidence interval, 247.4 to 252.7) for patients in the non-CKD group. The prevalence ratio of CKD patient safety indicators to Agency for Healthcare and Research Quality patient safety indicators was 23.4 (95% confidence interval, 21.9 to 25.0).
Conclusion
The candidate CKD patient safety indicators that occur in hospitalized patients are distinct from the Agency for Healthcare and Research Quality patient safety indicators and tend to be more common in CKD than non-CKD patients. These measures have the potential to serve as sentinel tools for identifying patients with CKD who warrant examination for disease-pertinent safety events.
Introduction
Reduction of medical errors is an important objective in health care, and various classifications have been developed to identify preventable hospital-based adverse safety events. The Agency for Healthcare and Research Quality (AHRQ) patient safety indicators (PSIs) (1) track patient safety events across hospitals, demographic groups, and disease populations (2–5). Adverse safety events detected with AHRQ PSIs are more common in patients with predialysis CKD than patients without CKD (6). Despite the more frequent occurrence of AHRQ PSIs in CKD, these indicators disproportionately identify procedural and surgical errors in the general hospital population, while potentially overlooking adverse safety events common to the medical care of CKD (7,8).
Our objective was to identify a set of candidate CKD-pertinent PSIs and contrast their incidence in hospitalized patients with impaired renal function with their incidence in those hospitalized patients with normal renal function. In the same population, we also compared the frequency of the candidate CKD PSIs with previously determined AHRQ PSIs to show the preponderance of disease-specific versus general indicators and how the latter fail to identify patients at risk for CKD-pertinent adverse safety events.
Materials and Methods
Study Design
This study was a retrospective observational study of national Veterans Health Administration (VHA) data from fiscal year (FY) 2005.
Setting and Data Sources
The study dataset was derived from the national VHA acute inpatient data file and linked to outpatient laboratory data for the veterans with at least one hospitalization. The source data files included the Patient Treatment File (PTF), the Outpatient Care File (OPC), and the Decision Support System (DSS) Laboratory Result files (9,10). The cohort was described previously (6,7) and has been deemed exempt by the University of Maryland, Baltimore Institutional Review Board and the Baltimore Veterans Affairs Research and Development Committee.
Participants
The study population included veterans with an acute care hospitalization during FY 2005, complete demographic data (age, sex, and race), and an outpatient serum creatinine measured from between 1 week and 1 year before first admission of the study year. Estimated GFR (eGFR) was calculated from the creatinine value closest to the first hospital admission using the four-component Modification of Diet in Renal Disease (MDRD) Equation, which was the standard for electronic and automated reporting at that time. Eligibility in the impaired renal function group (hereafter designated as the CKD group) was limited to those individuals and their associated first hospitalizations where the antecedent eGFR was <60 ml/min per 1.73 m2. Participants were excluded if there was a diagnosis of ESRD, solid organ transplant, or dialysis-related clinic visit. The remaining patients who had an eGFR≥60 ml/min per 1.73 m2 before hospitalization were assigned to the non-CKD group. Only the first hospitalization for each participant was included to avoid the inclusion of hospitalizations that could be the consequence of an adverse safety event occurring in the index hospitalization. We excluded those hospitalizations occurring in nonacute care facilities according to previously described methods (3).
PSIs
We included a set of candidate CKD-specific PSIs conceived as part of the preparatory phase of the Safe Kidney Care (SKC) cohort study (ClinicalTrials.gov NCT-01407367) when community and academic nephrologists were surveyed to identify and rank important disease-specific safety indicators. The survey responses were reviewed, prioritized, and endorsed by an expert consensus panel convened to establish candidate CKD PSIs. From these proceedings, 12 CKD PSIs related to in-hospital incidents were developed based on International Classification of Diseases, Ninth Revision (ICD-9) codes, their pertinence to the processes of care common in CKD, and their potential to identify individuals at risk for an adverse safety event.
The candidate CKD PSIs include in-hospital acute kidney failure (ICD-9 code 584); in-hospital congestive heart failure (ICD-9 code 428.x); fluid overload (ICD-9 code 276.6); acute edema of lung, unspecified (ICD-9 code 518.4); hypoglycemia (ICD-9 code 251.x); hyperkalemia (ICD-9 code 276.7); hypokalemia (ICD-9 code 276.8); hypernatremia (ICD-9 code 276.0); hyponatremia (ICD-9 code 276.1); hypercalcemia (ICD-9 code 275.42); hypocalcemia (ICD-9 code 275.41); hypophosphatemia (ICD-9 code 275.3); electrolyte disturbances (ICD-9 code 276.x, excluding hyperkalemia, hypokalemia, hyponatremia, and hypernatremia); and medication errors, poisoning, and intoxication (ICD-9 codes 960.xx–979.x,x E850.x–E858.x, E870.x–E876.x, and E930.x–E949.x).
To attempt to differentiate between events that were iatrogenic and events that were the consequence of the disease (e.g., congestive heart failure after a myocardial infarction), we used secondary diagnosis codes that target in-hospital events more closely associated with adverse safety events rather than the primary diagnoses, which typically code admitting diagnoses. We then developed two sets of exclusion criteria to eliminate hospitalizations from consideration for a PSI. The first set of exclusion criteria (Supplemental Appendix 1) accounted for disorders that may lead to acute kidney failure. This exclusion criterion was used for the acute kidney failure (hereafter referred to as AKI PSI) as well as the PSIs for metabolic/electrolyte disturbances cited above (e.g., sepsis preceding AKI and metabolic disturbances). The second set of exclusion criteria (Supplemental Appendix 2) was developed to account for cardiovascular disorders that may lead to congestive heart failure (CHF) and only applied to the CHF PSI (e.g., myocardial infarction preceding CHF).
For comparison with the CKD PSIs, we used previously determined incidence rates for 16 AHRQ PSIs adapted in prior analyses of this cohort (6,7). Development of these AHRQ PSIs and methods for estimating their incidence rates have been previously described (1). The publicly available software AHRQ-PSI 3.0a (1) identified eligible hospitalizations and determined the occurrence of each AHRQ PSI with necessary modifications required for the data structure and study population.
Covariates
Demographic characteristics include sex, age at hospitalization, race (black or other), and length of stay. The Charlson Comorbidity Index (CCI) was used as a measure of comorbidity, excluding diabetes and cardiovascular and kidney diseases, which were considered separate covariates. Patients were classified into CKD stages 3a, 3b, and 4/5 (eGFR values of 45–59, 30–44, and <29 ml/min per 1.73 m2, respectively).
Statistical Methods
Continuous variables were reported with mean ± SD, and comparisons across groups used t test or ANOVA. Categorical variables were expressed with N (%), and comparisons across groups were made using the chi-squared test. Rates of PSI were expressed as the number of events per 1000 person-hospitalizations at risk for each particular PSI. With the AHRQ PSIs, only hospitalizations determined at risk for a specific PSI were included in that rate determination. For example, only surgical discharges were eligible for the AHRQ PSI postoperative hip fracture. All hospitalizations were eligible for each CKD PSI unless the primary ICD-9 code presented a plausible explanation for the secondary incident and associated diagnostic code (Supplemental Appendices 1 and 2).
Adjusted rates for the CKD or AHRQ PSIs were determined using Poisson regression with CKD status set as the predictor variable and total count of CKD and AHRQ PSIs designated as the response variable in their respective models. Additional covariates in the models are outlined above, with two-way interactions between the CKD status and each of the other covariates also included. Length of stay was not included in the determination of risk factors predictive of a CKD PSI or to define risk strata for comparing odds of CKD versus AHRQ PSIs (Figure 1), because length of stay was a potential consequence of a safety event rather than a cause. To account for the fact that patients may be at risk for different CKD or AHRQ PSI incidents during a hospitalization, we defined the effective exposure for each of the combined CKD and AHRQ PSIs as the total number of PSIs that each participant was at risk for divided by the total number of possible CKD PSIs or total number of possible AHRQ PSIs, respectively. We then used the logarithm transformation of the effective exposure as an offset variable in the Poisson regression models. Where individual strata-adjusted PSI rates were reported, all other covariates were set at the population mean for that factor. All rates for CKD and AHRQ PSIs were expressed for both CKD and non-CKD groups.
Figure 1.
The odds of either an Agency for Healthcare and Research Quality (AHRQ) patient safety indicator (PSI) only or a CKD PSI only in the CKD cohort based on the number of CKD safety risk factors, including age of 75 years or older, estimated GFR of less than 30 ml/min per 1.73 m2, Charlson comorbidity index of four or greater, and cardiovascular disease as a comorbidity. Estimates of the prevalence ratio of CKD PSI relative to AHRQ PSI are depicted by plotted points, and bars represent 95% confidence intervals. Sample sizes for groups 0–4 are n=10,273, n=18,865, n=23,019, n=15,514, and n=2301, respectively. OR, odds ratio.
A logistic regression model determined key risk factors for the occurrence of one or more candidate CKD PSIs in the CKD group. The four significant predictors of CKD PSI included age≥75 years, comorbid cardiovascular disease, CCI≥4, and eGFR<30 ml/min per 1.73 m2. These predictors were used to construct risk strata to exhibit the differential odds of CKD versus AHRQ PSIs. The odds ratios of CKD and AHRQ PSIs were used to compute the prevalence ratios of the former versus the latter in each risk strata. Analyses were performed using SAS 9.2 (SAS Institute Inc., Cary, NC).
Results
Veterans with acute hospitalizations (247,160) during the study year were in the cohort. Table 1 presents the demographic characteristics of the CKD versus non-CKD groups. CKD patients were more likely to be men, be nonblack race, be older, and have a higher level of comorbidity (higher CCI score). They were also more likely to have cancer and cardiovascular disease with longer hospitalizations.
Table 1.
Comparison of demographic characteristics between CKD and non-CKD patients
| Characteristic | CKD | Non-CKDa | Total |
|---|---|---|---|
| Patients (row %) | 71,156 (28.8) | 176,004 (71.2) | 247,160 |
| Sex (column %) | |||
| Men | 68,844 (96.8) | 168,171 (95.6) | 237,015 (95.9) |
| Women | 2312 (3.2) | 7833 (4.4) | 10,145 (4.1) |
| Race | |||
| Black | 11,366 (15.9) | 34,177 (19.4) | 45,543 (18.4) |
| Other | 59,790 (84.1) | 141,827 (80.6) | 201,617 (81.6) |
| Age (yr)b | 71.6±10.8 | 62.6±12.5 | 65.2±12.7 |
| <45 | 624 (0.9) | 10,748 (6.1) | 11,372 (4.6) |
| 45–60 | 12,289 (17.3) | 74,970 (42.6) | 87,259 (35.3) |
| 61–75 | 28,202 (39.6) | 57,800 (32.8) | 86,002 (34.8) |
| 76+ | 30,041 (42.2) | 32,486 (18.5) | 62,527 (25.3) |
| Charlson Comorbidity Index | |||
| 0–1 | 12,543 (17.6) | 65,030 (36.9) | 77,573 (31.4) |
| 2–3 | 22,262 (31.3) | 58,011 (33.0) | 80,273 (32.5) |
| ≥4 | 36,351 (51.1) | 52,963 (30.1) | 89,314 (36.1) |
| Cancer | |||
| No | 49,570 (69.7) | 130,037 (73.9) | 179,607 (72.7) |
| Yes | 21,586 (30.3) | 45,967 (26.1) | 67,553 (27.3) |
| Diabetes | |||
| No | 34,294 (48.2) | 115,377 (65.6) | 149,671 (60.6) |
| Yes | 36,862 (51.8) | 60,627 (34.4) | 97,489 (39.4) |
| Cardiovascular disease | |||
| No | 27,834 (39.1) | 113,875 (64.7) | 141,709 (57.3) |
| Yes | 43,322 (60.9) | 62,129 (35.3) | 105,451 (42.7) |
| CKD stage | |||
| Controls | — | 176,004 (100.0) | 176,004 (71.2) |
| Stage 3ac | 37,768 (53.1) | — | 37,768 (15.3) |
| Stage 3bd | 20,268 (28.5) | — | 20,268 (8.2) |
| Stage 4 or 5e | 13,120 (18.4) | — | 13,120 (5.3) |
| Length of hospital stay (d) | |||
| 0–2 | 24,147 (33.9) | 68,126 (38.7) | 92,273 (37.3) |
| 3–6 | 26,517 (37.3) | 64,822 (36.8) | 91,339 (37.0) |
| 7+ | 20,492 (28.8) | 43,056 (24.5) | 63,548 (25.7) |
All comparisons were significant at P<0.01.
Control: GFR≥60 ml/min per 1.73 m2.
Mean ± SD.
Stage 3a: GFR=45–59.99 ml/min per 1.73 m2.
Stage 3b: GFR=30–44.99 ml/min per 1.73 m2.
Stage 4 or 5: GFR<30 ml/min per 1.73 m2.
Table 2 lists the candidate CKD and AHRQ PSIs, the counts of hospitalizations, the crude rates (per 1000 patient-hospitalizations) of each, and the associated rate ratio of each PSI in the CKD versus non-CKD group. Candidate CKD PSIs tended to be more common in the CKD than the non-CKD group, except for hypoglycemia, which was not significantly different between the two groups, and hypokalemia and hyponatremia, which were more common in the non-CKD group. CKD PSIs were generally more frequent than AHRQ PSIs in both the CKD and non-CKD groups. The most common CKD PSI in the CKD group was CHF, and the least common CKD PSI was hypoglycemia. Among patients in the CKD group, failure to rescue was the most common AHRQ PSI, and it was of comparable frequency with the three most common CKD PSIs: CHF, AKI, and electrolyte disturbances. Among non-CKD patients, the AHRQ PSI failure to rescue was more common than all the CKD PSIs in that group, but AKI, CHF, hypokalemia, hyponatremia, electrolyte disturbances, and medication errors occurred more commonly than all other AHRQ PSIs listed.
Table 2.
Count (hospitalizations) and rates (per 1000 patient-hospitalizations) of both Agency for Healthcare and Research Quality (AHRQ) and candidate CKD patient safety indicators (PSIs)
| Patient Safety Indicators | CKDa | Non-CKDb | Risk Ratio | Risk Ratio P Value | ||||
|---|---|---|---|---|---|---|---|---|
| Count | At-Risk Hospitalizations | Rate | Count | At-Risk Hospitalizations | Rate | |||
| Candidate CKD PSIs | ||||||||
| Acute kidney failure | 5758 | 66,367 | 86.8 (84.5–89.0) | 4300 | 170,428 | 25.2 (24.5–26.0) | 3.4 (3.3–3.6) | <0.001 |
| Congestive heart failure | 8709 | 53,497 | 162.8 (159.4–166.3) | 8247 | 144,835 | 56.9 (55.7–58.2) | 2.9 (2.8–2.9) | <0.001 |
| Hypoglycemia | 160 | 66,316 | 2.4 (2.1–2.8) | 404 | 170,335 | 2.4 (2.2–2.6) | 1.0 (0.8–1.2) | 0.85 |
| Hyperkalemia | 2230 | 65,993 | 33.8 (32.4–35.2) | 1446 | 170,353 | 8.5 (8.1–8.9) | 4.0 (3.7–4.3) | <0.001 |
| Hypokalemia | 1475 | 66,291 | 22.3 (21.1–23.4) | 4742 | 170,267 | 27.9 (27.1–28.7) | 0.8 (0.8–0.8) | <0.001 |
| Hypernatremia | 335 | 66,332 | 5.1 (4.5–5.6) | 551 | 170,378 | 3.2 (3.0–3.5) | 1.6 (1.4–1.8) | <0.001 |
| Hyponatremia | 1239 | 66,240 | 18.7 (17.7–19.8) | 4394 | 169,923 | 25.9 (25.1–26.6) | 0.7 (0.7–0.8) | <0.001 |
| Hypercalcemia | 228 | 66,290 | 3.4 (3.0–3.9) | 491 | 170,337 | 2.9 (2.6–3.2) | 1.2 (1.0–1.4) | 0.03 |
| Hypocalcemia | 177 | 66,353 | 2.7 (2.3–3.1) | 300 | 170,417 | 1.8 (1.6–2.0) | 1.5 (1.3–1.8) | <0.001 |
| Hypophosphatemia | 191 | 66,364 | 2.9 (2.5–3.3) | 331 | 170,425 | 1.9 (1.7–2.2) | 1.5 (1.2–1.8) | <0.001 |
| Electrolyte disturbancesc | 5547 | 65,408 | 84.8 (82.6–87.1) | 9012 | 169,298 | 53.2 (52.1–54.3) | 1.6 (1.5–1.6) | <0.001 |
| Medication errors, poisoning, and intoxication | 2960 | 70,956 | 41.7 (40.2–43.3) | 6169 | 175,016 | 35.3 (34.4–36.1) | 1.2 (1.1–1.2) | <0.001 |
| AHRQ PSI | ||||||||
| Complications of anesthesia | 17 | 14,852 | 1.2 (0.7–1.8) | 41 | 48,180 | 0.9 (0.6–1.2) | 1.3 (0.8–2.4) | 0.30 |
| Death in a low-mortality DRG | 46 | 9662 | 4.8 (3.5–6.4) | 61 | 34,201 | 1.8 (1.4–2.3) | 2.7 (1.8–3.9) | <0.001 |
| Decubitus ulcer | 440 | 26,795 | 16.4 (14.9–18.0) | 736 | 57,840 | 12.7 (11.8–13.7) | 1.3 (1.1–1.5) | <0.001 |
| Failure to rescue | 521 | 4747 | 109.8 (100.5–119.6) | 970 | 8657 | 112.1 (105.1–119.3) | 1.0 (0.9–1.1) | 0.69 |
| Foreign body left in during a procedure | 7 | 71,159 | 0.1 (0.0–0.2) | 26 | 176,001 | 0.2 (0.1–0.2) | 0.7 (0.3–1.5) | 0.34 |
| Iatrogenic pneumothorax | 35 | 67,364 | 0.5 (0.4–0.7) | 119 | 165,980 | 0.7 (0.6–0.9) | 0.7 (0.5–1.1) | 0.09 |
| Infection caused by medical care | 174 | 49,475 | 3.5 (3.0–4.1) | 185 | 114,560 | 1.6 (1.4–1.9) | 2.2 (1.8–2.7) | <0.001 |
| Postoperative hip fracture | 12 | 11,212 | 1.1 (0.6–1.9) | 5 | 32,451 | 0.2 (0.1–0.4) | 6.9 (2.4–19.7) | <0.001 |
| Postoperative hemorrhage or hematoma | 44 | 14,774 | 3.0 (2.2–4.0) | 126 | 47,998 | 2.6 (2.2–3.1) | 1.1 (0.8–1.6) | 0.47 |
| Postoperative physiologic or metabolic derangement | 140 | 14,597 | 9.6 (8.1–11.3) | 99 | 48,026 | 2.1 (1.7–2.5) | 4.7 (3.6–6.0) | <0.001 |
| Postoperative respiratory failure | 384 | 9574 | 40.1 (36.2–44.3) | 702 | 35,473 | 19.8 (18.4–21.3) | 2.0 (1.8–2.3) | <0.001 |
| Postoperative deep vein thrombosis or pulmonary embolus | 181 | 14,733 | 12.3 (10.6–14.2) | 464 | 47,933 | 9.7 (8.8–10.6) | 1.3 (1.1–1.5) | 0.01 |
| Postoperative sepsis | 158 | 6482 | 24.4 (20.7–28.5) | 249 | 18,525 | 13.4 (11.8–15.2) | 1.8 (1.5–2.2) | <0.001 |
| Postoperative wound dehiscence | 21 | 2418 | 8.7 (5.4–13.3) | 59 | 9784 | 6.0 (4.6–7.8) | 1.4 (0.9–2.4) | 0.15 |
| Accidental puncture or laceration wound | 222 | 71,128 | 3.1 (2.7–3.6) | 651 | 175,940 | 3.7 (3.4–4.0) | 0.8 (0.7–1.0) | 0.03 |
| Transfusion reaction | 3 | 71,156 | 0.0 (0.0–0.1) | 1 | 176,004 | 0.0 (0.0–0.0) | 7.4 (0.8–71.3) | 0.75 |
DRG, diagnosis-related group.
CKD: GFR≥60 ml/min per 1.73 m2.
Controls: GFR<60 ml/min per 1.73 m2.
Excludes hyperkalemia, hypokalemia, hypernatremia, and hyponatremia.
Table 3 presents the count and demographic characteristics of participants in each PSI category. A significantly greater proportion of participants with or without CKD experienced a CKD PSI alone versus an AHRQ PSI only. The proportion of patients with both a CKD PSI and an AHRQ PSI was also low and comparable with the number with an AHRQ PSI alone. Veterans experiencing only a CKD PSI had a higher mean age, were less likely to be black than those veterans with only an AHRQ PSI, and were also more likely to have a higher CCI with diabetes and cardiovascular disease as comorbidities. Those patients with a CKD PSI were less likely to have cancer, advanced stage CKD before hospitalization, or prolonged hospitalization than patients with an AHRQ PSI.
Table 3.
Characteristics of CKD and non-CKD patients who experience Agency for Healthcare and Research Quality (AHRQ) and candidate CKD patient safety indicators (PSIs; n=247,160)
| Characteristic | CKD (n=71,156) | Non-CKDa (n=176,004) | ||||||
|---|---|---|---|---|---|---|---|---|
| +CKD PSI, −AHRQ PSI | −CKD PSI, +AHRQ PSI | +CKD PSI, +AHRQ PSI | −CKD PSI, −AHRQ PSI | +CKD PSI, −AHRQ PSI | −CKD PSI, +AHRQ PSI | +CKD PSI, +AHRQ PSI | −CKD PSI, −AHRQ PSI | |
| Patients (row %) | 20,528 (28.8) | 877 (1.2) | 1184 (1.7) | 48,567 (68.3) | 30,040 (17.1) | 2082 (1.2) | 1830 (1.0) | 142,052 (80.7) |
| Sex (column %) | ||||||||
| Men | 19,935 (97.1) | 853 (97.3) | 1153 (97.4) | 46,903 (96.6) | 28,958 (96.4) | 2021 (97.1) | 1780 (97.3) | 135,412 (95.3) |
| Women | 593 (2.9) | 24 (2.7) | 31 (2.6) | 1664 (3.4) | 1082 (3.6) | 61 (2.9) | 50 (2.7) | 6640 (4.7) |
| Race | ||||||||
| Black | 3399 (16.6) | 190 (21.7) | 222 (18.8) | 7555 (15.6) | 5924 (19.7) | 368 (17.7) | 353 (19.3) | 27,532 (19.4) |
| Other | 17,129 (83.4) | 687 (78.3) | 962 (81.2) | 41,012 (84.4) | 24,116 (80.3) | 1714 (82.3) | 1477 (80.7) | 114,520 (80.6) |
| Age (yr)b | 72.3±10.9 | 71.7±10.4 | 70.2±10.2 | 71.3±10.8 | 65.1±12.7 | 64.7±11.4 | 65.1±11.6 | 62.0±12.4 |
| <45 | 147 (0.7) | 6 (0.7) | 9 (0.8) | 462 (1.0) | 1296 (4.3) | 72 (3.5) | 59 (3.2) | 9321 (6.6) |
| 45–60 | 3468 (16.9) | 130 (14.8) | 226 (19.0) | 8465 (17.4) | 11,090 (36.9) | 765 (36.7) | 665 (36.3) | 62,450 (44.0) |
| 61–75 | 7523 (36.7) | 391 (44.6) | 568 (48.0) | 19,720 (40.6) | 10,006 (33.3) | 832 (40.0) | 741 (40.5) | 46,221 (32.5) |
| 76+ | 9390 (45.7) | 350 (39.9) | 381 (32.2) | 19,920 (41.0) | 7648 (25.5) | 413 (19.8) | 365 (20.0) | 24,060 (16.9) |
| CCI | ||||||||
| 0–1 | 2548 (12.4) | 128 (14.6) | 135 (11.4) | 9732 (20.0) | 8014 (26.7) | 569 (27.3) | 416 (22.7) | 56,031 (39.4) |
| 2–3 | 5897 (28.7) | 244 (27.8) | 302 (25.5) | 15,819 (32.6) | 9836 (32.7) | 729 (35.0) | 576 (31.5) | 46,870 (33.0) |
| ≥4 | 12,083 (58.9) | 505 (57.6) | 747 (63.1) | 23,016 (47.4) | 12,190 (40.6) | 784 (37.7) | 838 (45.8) | 39,151 (27.6) |
| Cancer | ||||||||
| No | 14,410 (70.2) | 578 (65.9) | 753 (63.6) | 33,829 (69.7) | 21,492 (71.5) | 1302 (62.5) | 1149 (62.8) | 106,094 (74.7) |
| Yes | 6118 (29.8) | 299 (34.1) | 431 (36.4) | 14,738 (30.3) | 8548 (28.5) | 780 (37.5) | 681 (37.2) | 35,958 (25.3) |
| Diabetes | ||||||||
| No | 9149 (44.6) | 418 (47.7) | 525 (44.3) | 24,202 (49.8) | 18,744 (62.4) | 1349 (64.8) | 1188 (64.9) | 94,096 (66.2) |
| Yes | 11,379 (55.4) | 459 (52.3) | 659 (55.7) | 24,365 (50.2) | 11,296 (37.6) | 733 (35.2) | 642 (35.1) | 47,956 (33.8) |
| CVD | ||||||||
| No | 5557 (27.1) | 355 (40.5) | 415 (35.0) | 21,507 (44.3) | 15,335 (51.0) | 1394 (67.0) | 1049 (57.3) | 96,097 (67.7) |
| Yes | 14,971 (72.9) | 522 (59.5) | 769 (65.0) | 27,060 (55.7) | 14,705 (49.0) | 688 (33.0) | 781 (42.7) | 45,955 (32.3) |
| CKD Stage | ||||||||
| No CKD | — | — | — | — | 30,040 (100.0) | 2082 (100.0) | 1830 (100.0) | 142052 (100.0) |
| Stage 3ac | 9288 (45.3) | 399 (45.5) | 554 (46.8) | 27,527 (56.7) | — | — | — | — |
| Stage 3bd | 6478 (31.6) | 249 (28.4) | 342 (28.9) | 13,199 (27.2) | — | — | — | — |
| Stage 4 or 5d | 4762 (23.1) | 229 (26.1) | 288 (24.3) | 7841 (16.1) | — | — | — | — |
| LOS (d) | ||||||||
| 0–2 | 5262 (25.6) | 69 (7.9) | 73 (6.2) | 18,743 (38.6) | 8192 (27.3) | 204 (9.8) | 169 (9.3) | 59,561 (41.9) |
| 3–6 | 7817 (38.1) | 157 (17.9) | 161 (13.6) | 18,382 (37.9) | 11,352 (37.8) | 366 (17.6) | 319 (17.4) | 52,785 (37.2) |
| 7+ | 7449 (36.3) | 651 (74.2) | 950 (80.2) | 11,442 (23.5) | 10,496 (34.9) | 1512 (72.6) | 1342 (73.3) | 29,706 (20.9) |
All comparisons (except for race among the controls) were significant at P<0.01. CCI, Charlson Comorbidity Index; CVD, cardiovascular disease; LOS, length of hospital stay.
Controls: GFR≥60 ml/min per 1.73 m2.
Mean ± SD.
Stage 3a: GFR=45–59.99 ml/min per 1.73 m2.
Stage 3b: GFR=30–44.99 ml/min per 1.73 m2.
Stage 4 or 5: GFR<30 ml/min per 1.73 m2.
Table 4 reports adjusted rates of ever having any CKD PSIs or AHRQ PSIs for both the CKD group by stage and the non-CKD group in each demographic and case mix strata. The rate of CKD PSI increased with age in the non-CKD group but did not do so consistently within stages of the CKD population. The tendency to increasing frequency of CKD PSIs was observed across CCI and diabetes status for all groups. Cancer affected the incidence rate CKD PSIs in all groups, except those groups with stage 4 or 5 CKD. Patients with cardiovascular disease and longer length of stay for their hospitalization had higher rates of CKD PSIs relative to their counterparts without cardiovascular disease or with shorter length of stay. The adjusted AHRQ PSI rates also increased across increasing comorbidity strata in both the non-CKD and CKD stage groups along with length of stay in the hospital.
Table 4.
Adjusted rates for CKD and Agency for Healthcare and Research Quality (AHRQ) PSIs among patients by CKD stage (per 1000 patient-hospitalizations)
| Characteristica | Rates of CKD PSI Eventsb (95% Confidence Interval) | Rates of AHRQ PSI Eventsb (95% Confidence Interval) | ||||||
|---|---|---|---|---|---|---|---|---|
| Non-CKD | Stage 3a | Stage 3b | Stage 4 or 5 | Non-CKD | Stage 3a | Stage 3b | Stage 4 or 5 | |
| Sex | ||||||||
| Men | 250.2 (247.5 to 252.9) | 335.9 (327.6 to 344.4) | 448.4 (433.7 to 463.6) | 537.6 (518.5 to 557.3) | 34.3 (32.7 to 36.0) | 33.1 (29.3 to 37.4) | 41.8 (35.5 to 49.1) | 53.1 (45.0 to 62.6) |
| Women | 245.8 (232.9 to 259.5) | 328.5 (297.7 to 362.5) | 450.5 (398.7 to 509.1) | 629.2 (543.3 to 728.6) | 30.5 (25.5 to 36.5) | 33.7 (24.1 to 47.3) | 28.1 (16.1 to 49.1) | 55.8 (32.4 to 96.0) |
| Race | ||||||||
| Black | 259.5 (253.7 to 265.4) | 363.6 (346.5 to 381.5) | 445.4 (419.6 to 472.8) | 484.1 (458.7 to 510.9) | 32.9 (30.4 to 35.6) | 37.5 (31.2 to 45.0) | 45.5 (36.1 to 57.2) | 62.5 (51.2 to 76.3) |
| Other | 247.9 (245.1 to 250.8) | 329.6 (321.1 to 338.2) | 449.2 (433.9 to 465.0) | 554.8 (533.7 to 576.7) | 34.5 (32.8 to 36.2) | 32.2 (28.5 to 36.5) | 40.2 (34.0 to 47.4) | 51.3 (43.2 to 60.9) |
| Age (yr) | ||||||||
| <45 | 218.3 (207.7 to 229.5) | 347.3 (276.6 to 436.0) | 575.1 (433.1 to 763.7) | 530.7 (429.8 to 655.4) | 27.2 (23.0 to 32.2) | 24.3 (9.1 to 65.0) | 67.8 (27.9 to 164.7) | 47.1 (23.1 to 96.1) |
| 45–60 | 228.4 (224.7 to 232.3) | 345.5 (330.1 to 361.7) | 457.7 (432.1 to 484.9) | 548.8 (520.0 to 579.2) | 32.6 (30.7 to 34.6) | 32.0 (26.7 to 38.4) | 39.8 (31.4 to 50.5) | 53.1 (42.9 to 65.6) |
| 61–75 | 243.2 (238.9 to 247.5) | 315.4 (306.2 to 324.9) | 429.1 (411.3 to 447.7) | 512.9 (488.1 to 539.0) | 38.8 (36.6 to 41.2) | 40.2 (35.6 to 45.5) | 49.7 (41.6 to 59.2) | 61.1 (50.4 to 74.0) |
| 76+ | 302.0 (295.8 to 308.3) | 348.7 (338.2 to 359.6) | 442.7 (427.5 to 458.5) | 572.8 (546.6 to 600.3) | 32.0 (29.5 to 34.6) | 28.1 (24.4 to 32.3) | 30.2 (25.3 to 36.1) | 45.1 (36.7 to 55.5) |
| CCI | ||||||||
| 0–1 | 198.3 (193.7 to 203.0) | 284.3 (270.2 to 299.2) | 442.0 (412.9 to 473.2) | 529.5 (487.5 to 575.1) | 26.8 (24.8 to 29.0) | 25.8 (21.3 to 31.3) | 37.9 (28.7 to 50.0) | 38.4 (27.9 to 52.7) |
| 2–3 | 240.7 (236.5 to 245.0) | 327.6 (316.3 to 339.4) | 433.8 (414.4 to 454.1) | 528.0 (500.3 to 557.1) | 33.6 (31.5 to 35.7) | 34.0 (29.4 to 39.3) | 38.7 (31.7 to 47.3) | 48.3 (38.8 to 60.1) |
| ≥4 | 316.3 (310.4 to 322.4) | 396.0 (381.3 to 411.2) | 468.0 (446.9 to 490.2) | 563.5 (535.1 to 593.6) | 42.9 (40.1 to 45.9) | 40.1 (34.3 to 46.8) | 46.5 (38.2 to 56.6) | 77.1 (63.6 to 93.5) |
| Cancer | ||||||||
| No | 259.9 (256.6 to 263.2) | 346.4 (337.1 to 355.8) | 447.8 (432.4 to 463.8) | 544.9 (524.8 to 565.8) | 33.1 (31.4 to 34.9) | 33.1 (29.2 to 37.6) | 37.5 (31.7 to 44.4) | 56.6 (47.9 to 66.9) |
| Yes | 225.6 (220.7 to 230.6) | 308.6 (296.1 to 321.5) | 450.4 (428.2 to 473.7) | 530.9 (500.7 to 563.0) | 37.1 (34.5 to 39.8) | 33.1 (28.2 to 38.9) | 52.3 (42.6 to 64.3) | 45.1 (35.9 to 56.8) |
| Diabetes | ||||||||
| No | 258.5 (255.1 to 261.9) | 337.7 (327.7 to 348.0) | 442.5 (425.3 to 460.5) | 551.0 (526.6 to 576.5) | 36.2 (34.3 to 38.2) | 36.6 (32.1 to 41.8) | 40.8 (34.1 to 48.8) | 61.4 (51.1 to 73.8) |
| Yes | 237.5 (233.3 to 241.9) | 332.4 (321.4 to 343.8) | 457.8 (439.1 to 477.4) | 526.1 (501.2 to 552.2) | 31.3 (29.3 to 33.4) | 28.4 (24.5 to 32.8) | 41.6 (34.5 to 50.0) | 42.7 (34.9 to 52.2) |
| CVD | ||||||||
| No | 215.1 (211.9 to 218.3) | 284.4 (275.0 to 294.1) | 376.6 (360.2 to 393.8) | 461.3 (438.8 to 485.0) | 34.7 (32.8 to 36.6) | 33.9 (29.6 to 38.8) | 41.6 (34.7 to 50.0) | 52.9 (43.7 to 64.1) |
| Yes | 306.1 (301.2 to 311.1) | 419.3 (406.5 to 432.4) | 567.1 (545.4 to 589.7) | 670.2 (642.2 to 699.5) | 33.5 (31.5 to 35.7) | 32.1 (27.9 to 37.0) | 40.4 (33.5 to 48.6) | 53.6 (44.4 to 64.7) |
| LOS (d) | ||||||||
| 0–2 | 173.1 (169.7 to 176.6) | 242.1 (233.0 to 251.6) | 331.5 (315.6 to 348.1) | 415.3 (392.8 to 439.1) | 16.3 (14.8 to 18.0) | 13.9 (10.9 to 17.6) | 17.0 (12.2 to 23.5) | 26.4 (19.2 to 36.4) |
| 3–6 | 261.5 (257.2 to 265.8) | 349.3 (337.9 to 361.0) | 469.7 (450.3 to 489.9) | 558.6 (532.8 to 585.6) | 25.9 (24.1 to 27.9) | 25.7 (21.8 to 30.4) | 34.7 (28.1 to 42.9) | 42.2 (33.7 to 53.0) |
| 7+ | 399.9 (393.6 to 406.4) | 509.1 (492.5 to 526.2) | 651.1 (624.9 to 678.5) | 758.8 (725.4 to 793.8) | 148.3 (143.2 to 153.7) | 168.8 (154.2 to 184.7) | 189.0 (168.1 to 212.6) | 204.3 (179.7 to 232.1) |
Rates within strata are set at population mean for all other covariates. CCI, Charlson Comorbidity Index; CVD, cardiovascular disease; LOS, length of hospital stay.
All variables are held at the population’s mean value.
Per 1000 patient-hospitalizations.
Figure 1 presents the odds of experiencing CKD or AHRQ PSIs only versus no PSIs across strata identified by the cumulative number of risk factors for CKD PSI occurrence versus no such risk factors. Although both sets of odds ratios increase with the increment in risk factors, the magnitude of increased odds was greater in the CKD PSIs than the AHRQ PSIs. In the presence of all four risk factors, the odds of individuals with a CKD PSI were 4.3 times greater than the odds of individuals with none of those factors, whereas the odds of individuals with an AHRQ PSI were 2.3 times greater than the odds of individuals without risk factors. The prevalence ratio in each of the risk strata shows significantly higher frequency of CKD PSIs relative to AHRQ PSIs, with the prevalence ratio of CKD to AHRQ PSIs of 23.4 (95% confidence interval, 21.9 to 25.0) across the entire cohort.
Discussion
Using administrative data from hospitalized patients, we show a significant incidence of several candidate CKD PSIs in patients with prehospitalization impaired renal function and reveal that hospitalized patients classified with a CKD PSI were distinct from those patients identified with an AHRQ PSI. The frequency of the CKD-specific PSIs was higher than the frequency of AHRQ PSIs across the entire population, in CKD and non-CKD subgroups, and within key case mix subgroups. All CKD PSIs were more common in patients with impaired renal function, except for hypoglycemia, hypokalemia, and hypernatremia, suggesting that the latter indicators may not be specific to the care of patients with CKD. Although CKD PSIs along with AHRQ PSIs are, at most, crude proxies for processes of care that lead to adverse safety events, they can serve as useful tools to localize high-risk patient groups warranting closer examination for the incidence of verifiable adverse safety events. We anticipate that patients identified by the CKD PSIs are at high risk for adverse safety events pertinent to CKD care (except for the subset of indicators that was not specific to CKD) as opposed to those patients experiencing an AHRQ PSI, who are at risk for one of the more general processes of care (e.g., procedures). This result needs to be confirmed with more detailed evaluations of patient hospital records.
Since the release of the Institute of Medicine report entitled “To Err is Human: Building a Safer Health System” (11), various entities have established standards to evaluate patient safety events in the hospital setting (12,13). The approach formulated by the AHRQ was based on consensus proceedings intended to develop safety indicators from ICD-9 discharge codes and enable examination of safety events across hospitals or health networks (1). The AHRQ PSIs and ICD-9–derived indicators to monitor safety in hospitalized patients have been evaluated with mixed effectiveness in detecting target adverse safety events (14–17). An alternative approach to measuring adverse safety events uses the Institute for Healthcare Improvement Trigger Tool to scan hospitalization records and identify triggers that might lead to an adverse safety event. This method is an effective means to tag a set of charts for review for adverse safety events from a larger population of hospitalizations or patients (18–20). The value of the candidate CKD PSIs evaluated in this study is their use as triggers for screening large samples of hospital records for CKD-specific safety events. Although showing that predesignated renal-relevant events are more common in patients with impaired kidney function seems self-ordained, it points to the underdetection of such disease-pertinent events using current consensus-based ICD-9–derived safety indicators and their failure to identify an enriched patient group that is likely to have the highest incidence of CKD-specific safety events. It is worth noting that only a minority of hospitalizations with a safety event had both a CKD PSI and an AHRQ PSI, showing how use of a distinct set of PSIs is likely to broaden the set of individuals who are identified to experience adverse safety events.
As a retrospective analysis, the study results should be interpreted with limitations in mind. Miscoding of discharge diagnoses was possible in these hospitalizations; however, recording errors or bias are unlikely to be greater for the CKD versus AHRQ PSIs. Studies of the veterans and processes of care in the VHA health system may not generalize to the larger population. Our designation of CKD based on a single (rather than two or more) prehospitalization GFR findings of less than 60 ml/min per 1.73 m2 was necessary given the variable frequency of renal function testing in an administrative dataset and the lack of serial renal function measures in a substantial portion of the study cohort. The modified MDRD equation to estimate GFR was used in this study, because it was the prevailing means of reporting renal function at the time of this study. We recognize that it is difficult to verify whether disease-specific adverse events identified by the CKD PSIs resulted from delivery of care or were a consequence of disease progression. Excluding hospitalizations with the PSI ICD-9 codes in the primary position along with hospitalizations with predefined primary diagnoses preceding episodes of AKI and CHF strengthened the supposition that the CKD PSIs were related to an adverse safety event in the hospital, but it is dependent on the proper coding and sequencing of discharge diagnoses. We were unable to determine the accuracy of ICD-9 code-derived CKD PSIs without a gold standard for comparison. Indeed, clinical events included in the CKD PSIs relate to common clinical complications of CKD and have a likely association with adverse outcomes, which was previously shown for hypoglycemia and hyperkalemia (21,22); however, the diverse array of CKD PSIs requires additional validation for their linkage to processes of care and adverse outcome. Finally, a revision of the AHRQ PSI coding protocol has been developed since our original determination of AHRQ PSI rates in the VHA; however, we used previously determined rates of AHRQ PSIs in this population as a point of reference to the candidate CKD PSIs (6). Neither the version used for our analysis (version 3.0a, February of 2006) nor the most current version (version 4.4, March of 2012) provides PSIs that are specific to CKD patients.
Improving the wellbeing of patients with CKD is incomplete without understanding how harm from medical care contributes to adverse outcomes and whether such lapses in patient safety are preventable. CKD is unique in its collection of patient safety hazards, and tools should be developed to reduce these threats to patients’ outcomes. This analysis is a necessary first step in the development of disease-specific PSIs with applicability to CKD. Future efforts will be needed to modify, refine, and validate the candidate PSIs proposed here, such as with the ongoing Safe Kidney Care study. Consensus initiatives are necessary to define appropriate safety metrics in CKD but will be facilitated by empirical analyses such as this analysis. The candidate CKD PSIs examined here have potential to serve as useful measures to gauge the extent of adverse safety events as they continue to occur in patients with CKD who require hospitalization.
Disclosures
None.
Supplementary Material
Acknowledgments
The work included in this paper was supported serially under National Institute of Diabetes and Digestive and Kidney Diseases Grants R21 DK075675 (M.Z., L.W., and J.C.F.) and R01DK084017 (J.S.G., M.Z., and J.C.F.) and by the University of Maryland Summer Research Training Program (I.R.H.).
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
This article contains supplemental material online at http://cjasn.asnjournals.org/lookup/suppl/doi:10.2215/CJN.12671212/-/DCSupplemental.
References
- 1.Agency for Healthcare Resarch and Quality : AHRQ Quality Indicators—Guide to Patient Safety Indicators, Rockville, MD, Agency for Healthcare Quality and Research, 2003 [Google Scholar]
- 2.Rosen AK, Rivard P, Zhao S, Loveland S, Tsilimingras D, Christiansen CL, Elixhauser A, Romano PS: Evaluating the patient safety indicators: How well do they perform on Veterans Health Administration data? Med Care 43: 873–884, 2005 [DOI] [PubMed] [Google Scholar]
- 3.Rosen AK, Zhao S, Rivard P, Loveland S, Montez-Rath ME, Elixhauser A, Romano PS: Tracking rates of Patient Safety Indicators over time: Lessons from the Veterans Administration. Med Care 44: 850–861, 2006 [DOI] [PubMed] [Google Scholar]
- 4.Kaafarani HM, Rosen AK: Using administrative data to identify surgical adverse events: An introduction to the Patient Safety Indicators. Am J Surg 198[Suppl]: S63–S68, 2009 [DOI] [PubMed] [Google Scholar]
- 5.Sedman A, Harris JM, 2nd, Schulz K, Schwalenstocker E, Remus D, Scanlon M, Bahl V: Relevance of the Agency for Healthcare Research and Quality Patient Safety Indicators for children’s hospitals. Pediatrics 115: 135–145, 2005 [DOI] [PubMed] [Google Scholar]
- 6.Seliger SL, Zhan M, Hsu VD, Walker LD, Fink JC: Chronic kidney disease adversely influences patient safety. J Am Soc Nephrol 19: 2414–2419, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Chapin E, Zhan M, Hsu VD, Seliger SL, Walker LD, Fink JC: Adverse safety events in chronic kidney disease: The frequency of “multiple hits.” Clin J Am Soc Nephrol 5: 95–101, 2010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Fink JC, Brown J, Hsu VD, Seliger SL, Walker L, Zhan M: CKD as an underrecogized threat to patient safety. Am J Kidney Dis 53: 681–688, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Cowper DC, Hynes DM, Kubal JD, Murphy PA: Using administrative databases for outcomes research: Select examples from VA Health Services Research and Development. J Med Syst 23: 249–259, 1999 [DOI] [PubMed] [Google Scholar]
- 10.Maynard C, Chapko MK: Data resources in the Department of Veterans Affairs. Diabetes Care 27[Suppl 2]: B22–B26, 2004 [DOI] [PubMed] [Google Scholar]
- 11.Kohn KT, Corrigan JM, Donaldson MS: To Err Is Human: Building a Safer Health System, Washington, DC, National Academy Press, 1999 [PubMed] [Google Scholar]
- 12.Levinson DR: Adverse Events in Hospitals: National Incidence among Medicare Beneficiaries, Department of Health and Human Services, Office of the Inspector General, Washington, DC, 2010 [Google Scholar]
- 13.National Quality Forum : Safe Practices for Better Healthcare—2010 Update: A Consensus Report, Washington, DC, NQF, 2010 [Google Scholar]
- 14.Kern EF, Maney M, Miller DR, Tseng CL, Tiwari A, Rajan M, Aron D, Pogach L: Failure of ICD-9-CM codes to identify patients with comorbid chronic kidney disease in diabetes. Health Serv Res 41: 564–580, 2006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.van Walraven C, Austin PC, Manuel D, Knoll G, Jennings A, Forster AJ: The usefulness of administrative databases for identifying disease cohorts is increased with a multivariate model. J Clin Epidemiol 63: 1332–1341, 2010 [DOI] [PubMed] [Google Scholar]
- 16.Borzecki AM, Cevasco M, Chen Q, Shin M, Itani KM, Rosen AK: How valid is the AHRQ Patient Safety Indicator “postoperative physiologic and metabolic derangement?” J Am Coll Surg 212: 968–976, 2011 [DOI] [PubMed] [Google Scholar]
- 17.Sickbert-Bennett EE, Weber DJ, Poole C, MacDonald PD, Maillard JM: Utility of International Classification of Diseases, Ninth Revision, Clinical Modification codes for communicable disease surveillance. Am J Epidemiol 172: 1299–1305, 2010 [DOI] [PubMed] [Google Scholar]
- 18.Resar RK, Rozich JD, Classen D: Methodology and rationale for the measurement of harm with trigger tools. Qual Saf Health Care 12[Suppl 2]: ii39–ii45, 2003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Rozich JD, Haraden CR, Resar RK: Adverse drug event trigger tool: A practical methodology for measuring medication related harm. Qual Saf Health Care 12: 194–200, 2003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Griffin FA, Resar FA: IHI Global Trigger Tool for Measuring Adverse Events. IHI Innovation Series White Paper, Cambridge, MA, Institute for Healthcare Improvement, 2007 [Google Scholar]
- 21.Einhorn LM, Zhan M, Hsu VD, Walker LD, Moen MF, Seliger SL, Weir MR, Fink JC: The frequency of hyperkalemia and its significance in chronic kidney disease. Arch Intern Med 169: 1156–1162, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Moen MF, Zhan M, Hsu VD, Walker LD, Einhorn LM, Seliger SL, Fink JC: Frequency of hypoglycemia and its significance in chronic kidney disease. Clin J Am Soc Nephrol 4: 1121–1127, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.