Emergency Department Use Among Adults Receiving Dialysis

This cohort study assesses factors associated with potentially preventable emergency department use by adults receiving maintenance dialysis in Canada.


Introduction
The considerable increase in emergency department (ED) use across North America over the past decade has strained the capacity of health systems to respond to urgent health concerns in a timely and effective way. 1,2Many efforts to improve the appropriateness of ED use and prevent the need for emergency care focus on frequent ED users, who reflect a minority of all ED users but a disproportionately high number of ED presentations. 3 Observational reports suggest people with kidney failure receiving dialysis are frequent ED users, with a mean of 3 ED encounters per year and reported rates of ED use up to 8 times greater than the general population. 4,5In addition to patients with kidney failure being older, more frail, and having more comorbid conditions than the general public, their reliance on dialysis as life-sustaining therapy both contributes to and complicates the high ED use observed in this population. 6ople with kidney failure receive emergency care for a variety of health issues that occur both with and as a result of their underlying kidney disease. 7][10][11] However, the medical reasons prompting ED visits are broadly distributed across many diagnoses that can be but are not necessarily related to kidney failure or the need for dialysis. 12Previous research by our group has found that 6% of all ED encounters among people with chronic kidney disease were for conditions commonly attributed to advanced kidney disease that are potentially preventable through early identification and intervention (eg, hyperkalemia, volume overload). 4The higher rates of potentially preventable ED use and hospitalizations observed among people receiving dialysis compared with those with earlier stages of kidney disease 4,13 underscore a need to understand the individual and contextual drivers in this complex population.
5][16][17] However, the factors associated with potentially preventable ED encounters among people receiving dialysis are poorly understood yet necessary to identify and enable the design of interventions that improve appropriateness of acute care use, system efficiencies, and patient outcomes and experiences.In this population-based study, we used linked administrative health data to determine rates and characteristics of potentially preventable ED encounters among a cohort of patients with kidney failure receiving maintenance dialysis.

Study Design and Population
We conducted a retrospective cohort study using linked laboratory and administrative health data within the Alberta Kidney Disease Network database. 18This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline and the Reporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. 19e study received ethics approval and a waiver of patient informed consent from the University of

JAMA Network Open | Nephrology
Emergency Department Use Among Adults Receiving Dialysis The study population included all Alberta residents aged 18 years and older who were receiving maintenance dialysis (ie, hemodialysis or peritoneal dialysis) between April 1, 2010, and March 31, 2019.Receipt of dialysis was identified within the provincial dialysis registries (Alberta Kidney Care-North and -South) for more than 90 days to establish maintenance dialysis status and to prevent misclassification of people receiving short-term dialysis for acute kidney injury. 20All patients were followed up from cohort entry (defined as dialysis start date plus 90 days) until death, outmigration from the province, receipt of a kidney transplant, or end of study follow-up.We excluded patients with a prior transplant and those who died or received a kidney transplant between the start of dialysis and cohort entry date.

Outcomes
We linked the cohort to provincial administrative data sources via each individual's unique provincial health care number.The National Ambulatory Care Reporting System was used to count the number of all-cause ED encounters from cohort entry until death, outmigration from the province, receipt of a kidney transplant, or end of study follow-up.Encounter-level information was extracted, including the top 10 most frequent ED diagnoses and the proportion of ED visits resulting in hospital admission or death.Multiple encounters on the same day were reported as 1 event, with information being reported from the first event.
We defined potentially preventable ED use as encounters for kidney disease-specific ambulatory care-sensitive conditions (ACSCs).We defined ACSCs as health conditions for which adequate outpatient care can prevent the need for acute care presentation and are recognized internationally as a measure of the adequacy of ambulatory and primary health care performance. 21,22 previously defined kidney disease-specific ACSCs (hyperkalemia, malignant hypertension, heart failure, and volume overload) listed as the ED diagnosis code responsible for the ED encounter (eTable 1 in Supplement 1). 23,24Heart failure and volume overload are distinguished in these coding algorithms by the latter comprising extracellular fluid volume expansion that is not attributed directly to poor cardiac function.

Covariates
Included patients were linked to additional administrative data sources obtained from the provincial health ministry to capture sociodemographic data, clinical characteristics, and measures of prior health care use.We used the Andersen behavioral model of health services use as a conceptual framework to identify variables related to health care need, predisposing factors, and enabling factors, where available, from our data sources. 25 We reported on comorbidities including previous acute myocardial infarction, atrial fibrillation, cancer, chronic pain, chronic pulmonary disease, severe constipation, depression, dementia, diabetes, heart failure, hypertension, hypothyroidism, peripheral vascular disease, and stroke.These were defined using validated algorithms of ICD-9-CM and ICD-10-CA codes (eTable 1 in Supplement 1). 27Dialysis-related characteristics, such as initial modality, duration (in years), and modality switch (ie, switch from hemodialysis to peritoneal dialysis or vice versa at any time during follow-up), were identified within the provincial dialysis registries.Outpatient laboratory tests were also reported using the most recent measurement within the 6 months before the date of the first ED encounter.These included hemoglobin A 1c , serum potassium, and serum sodium levels.Cutoff values for laboratory parameters were defined according to the Medical Council of Canada. 28Health care resource use in the year prior to cohort entry was also measured and included the number of primary care visits, primary care attachment (ie, usual provider of care index), 29 number of cardiologist visits, number of prior ED visits, number of hospitalizations, and characteristics of the hospitalizations (ie, cumulative length of stay, intensive care unit admission, or long-term care placement at discharge).

Statistical Analysis
Patient characteristics were summarized using counts for categorical and dichotomous variables and means and SDs for continuous variables.All variables of interest were stratified by rate of ED use, which was operationalized as 0 visits per person-year (ie, no ED encounters during follow-up), less than 1 visit per person-year, 1 to less than 3 visits per person-year, and 3 or more ED visits per personyear.Crude rates of all-cause ED encounters by year of cohort entry were also calculated to explore trends in ED use over time.In a separate analysis, patient characteristics were stratified by the presence or absence of at least 1 ACSC ED presentation during the follow-up period.All statistical comparisons for potential differences in patient clinical and sociodemographic characteristics were performed using χ 2 tests, t tests, and analysis of variance, as appropriate.When covariate data were missing, a category was included for missing data.
Crude rates of all-cause and kidney-specific ACSC ED use (per 1000 person-years) were calculated using generalized linear models with a negative binomial distribution and log link.To identify patient and clinical factors associated with rates of potentially preventable ED encounters, multivariable negative binomial regression models were created initially among those with at least 1 ED encounter.In sensitivity analyses, models were then rerun among the entire cohort (including those with no ED encounters) to assess the robustness of study findings.Factors that were statistically significant in univariable models were included in a multivariable model.A parsimonious model was then developed using backward elimination, with the Akaike information criterion and likelihood ratio test used to determine model goodness of fit. 30All variables in the final adjusted model were reported as incident rate ratios (IRRs).For all statistical tests, P < .05 was considered statistically significant.Data were analyzed in March 2024 using SAS, version 9.4 (SAS Institute Inc).
Nearly one-third of patients (28.3% and 29.5%) scored in the highest quintiles (ie, most deprived) on the Pampalon social and material deprivation indices, respectively (Table 1).More than half (64.6%) of the cohort was receiving maintenance hemodialysis at study entry, and the mean (SD) duration of dialysis was 2.4 (2.0) years.The most common comorbidities among the cohort were hypertension (90.7%), diabetes (61.0%), chronic pain (53.4%), and heart failure (39.0%).
Approximately one-third (37.8%) died during the study.
There were 1048 patients (21.3%) who had 0 ED encounters during the study period and thus a rate of 0 all-cause ED visits per year.There were 773 patients (15.7%) with less than 1 ED visit per year, 1498 (30.4%) with 1 to less than 3 ED visits per year, and 1606 (32.6%) with 3 or more ED visits per year.Throughout the study period, the rate of ED visits remained stable (approximately 3 visits per person-year) (Table 1).The proportion of patients who were female and those living in rural locations within the province increased as the rate of ED use increased.Furthermore, the proportions of patients in the highest material and social deprivation quintiles and lowest income quintile also increased as the rate of ED use increased (Table 1).The prevalence of all measured comorbidities increased with increasing rates of ED use; for example, chronic pain was present in 44.0% of patients with 0 ED visits per person-year and 64.3% of those with 3 or more ED visits per person-year (P < .001).Patients receiving maintenance dialysis who had the highest rates of ED use (Ն3 ED encounters per year) were more likely to die during the study than those with low rates of ED use (<1 (defined as 3 or more ED encounters in the prior year) had higher rates of ACSC presentation (Figure 2; eTable 4 in Supplement 1).In sensitivity analyses, results were similar for regression models that included the subset of patients with no ED encounters (eFigure and eTable 5 in Supplement 1).

Discussion
In this large, population-based cohort, we found that patients receiving maintenance dialysis experienced approximately 3 all-cause ED encounters per year.While rates of ED use for potentially preventable kidney disease-related conditions were much lower, ACSC-related ED encounters were associated with important clinical and sociodemographic factors, including younger age, social disadvantage, comorbidities such as chronic pain, and high historical ED use, with half resulting in hospital admission.These findings highlight the burden of acute care use for people receiving dialysis, their families, and the health system and opportunities to address contributors to potentially preventable ED use in community-based settings.Defined as a set of conditions for which hospitalizations could be avoided through effective outpatient management, ACSCs have been used by researchers and decision-makers as factors in ambulatory care quality. 21,22,31,32Previous studies using the kidney disease-specific ACSC classification have also shown high rates of all-cause and potentially preventable acute care use among people with chronic kidney disease, with rates increasing in a graded fashion as disease severity increases. 4,13,23,24Our study complements and extends earlier findings by focusing on a subset of individuals with end-stage kidney disease receiving dialysis who have frequent ED use and addressing a limitation of using ACSC diagnoses to define preventable acute care use: the underestimation of other individual-and system-level influences. 339][40][41][42][43] Despite the relatively low rates of ACSC-related ED use, our findings point to important upstream targets to both identify and focus attention on individuals for whom community-based interventions could bring about measurable changes on acute care use and potentially costs, patient care experiences, and clinical outcomes.
In one systematic review assessing factors underlying ED use among people receiving maintenance hemodialysis, included studies reported primarily on dialysis parameters and infrequently on psychosocial and system factors (eg, psychiatric conditions, social determinants of health, or financial models). 44Only 4 included studies had evaluated the implications of an upstream intervention, such as home telemonitoring, for ED use. 44  The results are based on a multivariable regression model that was adjusted for the covariates listed.Pampalon deprivation index ranked from 1 (lowest) to 5 (highest).Potassium levels categorized as low (<3.5 mEq/L [to convert to millimoles per liter, multiply by 1]), normal (3.5 to <5.0 mEq/L), or high (Ն5.0mEq/L).IRR indicates incident rate ratio.

Figure 2 .
Figure 2. Factors Associated With Potentially Preventable Emergency Department (ED) Encounters

Table 1 .
encounter per year) (55.2% vs 23.4%, respectively).Demographic and Clinical Characteristics of Adults Receiving Dialysis Between April 1, 2010, and March 31, 2019, Overall and Stratified by Rate of All-Cause ED Encounters a receiving dialysis with at least 1 kidney disease-specific ACSC were more likely to be materially and socially disadvantaged, have a higher comorbidity burden (most notably for the prevalence of heart failure, chronic pain, and depression), and have a longer duration of dialysis compared with those with no ACSC ED encounters (Table2).They were also more likely to die during the study timeframe (41.9% for those without an ACSC vs 50.3% for those with an ACSC).Fifty-one percent (689 of 1351) of ED encounters for an ACSC resulted in a hospital admission compared with 28.8% (9396 of 32 678) for a non-ACSC-related ED encounter (P < .001).Hospital length of stay was slightly shorter for ACSC-related ED encounters resulting in admission than for non-ACSC-related admissions (median [IQR], (5 [3-9] days vs 7 [4-15] days).Differences in the top 10 diagnoses most responsible for the ED encounter (patient-and encounter-level analyses) for the first ED encounter and any ED encounter stratified by ACSC status are provided in eTables 2 and 3 in Supplement 1.Multivariable regression modeling showed that rates of ACSC-related ED encounters were significantly higher for younger patients (IRR, 1.69 [95% CI, 1.33-2.15]forthose18-44 years vs Ն65 years) and those with chronic pain (IRR, 1.35 [95% CI, 1.14-1.61]vsthosewithout chronic pain) and heart failure (IRR, 1.50 [95% CI, 1.26-1.79]vsthosewithout heart failure).Patients with a history of cancer within 5 years prior to cohort entry had lower rates of ACSC-related ED use (IRR, 0.59 [95% CI, 0.42-0.85]comparedwithpatients without a cancer diagnosis).Patients with the highest level of material deprivation also had higher rates of ACSCs (IRR, 1.57 [95% CI, 1.16-2.12]vsthosewith the a Time on dialysis was calculated during the study period only.JAMA NetworkOpen.2024;7(5):e2413754.doi:10.1001/jamanetworkopen.2024.13754(Reprinted) May 29, 2024 6/14 Downloaded from jamanetwork.comby guest on 06/06/2024 lowest level of material deprivation).Finally, patients with a history of hyperkalemia (at the measurement closest to the ED encounter within the previous 6 months) and historically high ED use

Table 1 .
Demographic and Clinical Characteristics of Adults Receiving Dialysis Between April 1, 2010, and March 31, 2019, Overall and Stratified by Rate of All-Cause ED Encounters a (continued) a Study characteristics except laboratory values were measured at cohort entry (90 days after initial dialysis date).Laboratory values were measured at first ED encounter.Cells with n <5 are masked for privacy and indicated by NR. b P values result from χ 2 tests and analysis of variance for categorical and continuous variables, respectively.Patients with 0 ED encounters per person-year were included in the statistical testing.c Residence is based on the Statistics Canada definition of rural residence (population <1000 or a population density <400/km 2 ).d Based on the most recent laboratory measurement preceding the date of the first ED encounter within a 6-month timeframe.

Table 2 .
Demographic Characteristics of Adults Receiving Maintenance Dialysis Between April 1, 2010, and March 31, 2019, Overall and Stratified by Presence of ACSC ED Encounter a (continued) P values result from χ 2 tests and pooled t tests for categorical and continuous variables, respectively.Patients who did not have an ED encounter were not included in the comparison. b A separate scoping review further highlights a gap in available acute care avoidance interventions targeted to people experiencing complications of advanced kidney disease, such as hyperkalemia or volume overload.45Asthesecomplicationsareamong potentially preventable kidney disease-related ACSCs, their prompt detection and treatment outside of the acute care setting present opportunities for mitigating ED use in this population.The unique requirement for dialysis both as life-sustaining therapy and to treat disease-related complications means that any proposed preventative care strategies should necessarily consider the imminence of dialysis need and supports (eg, resources, expertise) for safe, effective management.While understanding the risk factors of potentially preventable ED use is grounded in a need for targeted identification and intervention, high overall ED use in patients with kidney failure continues to burden health systems internationally.With half of ACSC-related ED encounters resulting in hospital admission, the potential cost implications of our findings are significant, as patients receiving dialysis account for disproportionately high spending on both all-cause and kidney disease-specific ACSC hospitalizations compared with people with lower risk or no kidney disease.46Moreover,one-third of patients in our cohort had more than 3 ED encounters per year, reflecting a high-use subset of patients with more comorbidities, greater socioeconomic disadvantage, and increased

SUPPLEMENT 1. eTable 1. ICD
-9-CM and ICD-10-CA Codes Used to Define the Study Cohort and Covariates eTable 2. Diagnosis Most Responsible for the First ED Encounter During the Study Period, Stratified by ACSC Presentation (n=3877 ED Encounters Among 3877 Patients) eTable 3. Diagnosis Most Responsible for All ED Encounters During the Study Period, Stratified by ACSC Presentation (n=34 029 ED Encounters Among 3877 Patients) eTable 4. Association Between Characteristics and the Rate of ACSC ED Encounters Among Adults Receiving Maintenance Dialysis Between April 1, 2010, and March 31, 2019, Who Had at Least 1 ED Encounter (n=3877) eTable 5. Association Between Characteristics and the Rate of ACSC ED Encounters Among Adults Receiving Maintenance Dialysis Between April 1, 2010, and March 31, 2019 (n=4925) eFigure.Sensitivity Analysis of Factors Associated With Potentially Preventable ED Encounters Among All Patients Receiving Maintenance Dialysis (n=4925)