Short-Term Mortality Associated with Outpatient Vascular Access Restoration Procedures

Abstract

Key Points 

• This study confirms the safety of endovascular interventions for thrombosis of hemodialysis access in outpatient and office-based settings.
• Risk of death in the week after vascular access procedure was not associated with hemodialysis access type (fistula versus graft).

In the ESKD population, the overall high risk of death may obscure short-term increases in mortality related to endovascular procedures. Single institution studies have reported little to no increased risk of short-term mortality after outpatient endovascular intervention (¹,²); however, no studies have evaluated procedure-related mortality risks in the US ESKD population nationally. This study aimed to estimate the risk of all-cause mortality within the first 7 days after an endovascular access (VA) restoration procedure due to thrombosis and to identify patient and procedural risk factors associated with 7-day mortality among a national cohort of US ESKD patients.

We conducted a case-cohort study of US ESKD patients identified from the United States Renal Data System (USRDS) from January 1, 2011, to September 30, 2015 (³). This end date was chosen because it marks the conversion of Medicare claims data to ICD-10. Baseline inclusion criteria were hemodialysis (HD) treatment for at least 1 year, use of an arteriovenous fistula or graft, age ≥18 years, and Medicare as the primary payer. Within this base population, we identified the first VA procedure with thrombosis diagnosis for each patient (index procedure). We then excluded patients for whom the index procedure was performed in an inpatient hospital or emergency department setting and all patients who had additional VA procedures within 90 days of the index procedure. Patients for whom the index procedure was not endovascular were also excluded.

We identified 356 deaths within 90 days among 6381 unique patients meeting the inclusion criteria. Among patients experiencing mortality, the cause of death was missing for 50% of patients, and it was not possible to associate this information with VA procedures among people for whom these data were available. There was no increase in mortality 0–7 days after VA procedures (n=31, mean 3.9 deaths per day) compared with 8–90 days (n=325, mean 4 deaths per day). A modified self-controlled case series analysis (⁴) yielded a relative incidence of 1.0 (95% confidence interval, 0.7 to 1.5).

We used logistic regression to examine risk factors for mortality within 7 days (cases) versus 8–90 days post procedure (controls), adjusting for age and Liu Comorbidity Index (⁵). We found an increased risk associated with a duration of 3 days since last dialysis relative to 0–2 days (Table 1). This finding and the increased risk associated with inadequate dialysis dose measured by single-pool Kt/V ≤1.2 at the time of the procedure may have been related to the indication for the procedure rather than the procedure itself. Although we did not detect a difference in risk associated with dialysis ≥4 days relative to 0–2 days since last dialysis, our estimates were highly imprecise (odds ratio=0.6, 95% confidence interval, 0.1 to 4.7) due to a small number of cases and controls whose last dialysis was ≥4 days before their procedure.

Table 1. - Adjusted risk ratios comparing death within 7 days of an outpatient endovascular procedure for thrombosis relative to death between 8 and 90 days among ESKD patients, United States Renal Data System, January 1, 2011, to September 30, 2015

Characteristic	Adjusted Risk Ratio a	95% Confidence Interval
Age at time of VAP (continuous, yr)	1.0	1.0	1.1
Liu Comorbidity Index	1.1	1.0	1.4
Days since last dialysis (3 versus ≤2)	3.7	1.3	10.6
Days since last dialysis (≥4 versus ≤2)	0.6	0.1	4.7
Single-pool Kt/V <1.2 b	3.8	1.1	13.3
Heart failure	4.2	1.3	13.9
Procedure setting (outpatient/ambulatory surgery versus office)	1.1	0.5	2.5
Hemodialysis access type (arteriovenous fistula versus graft)	0.8	0.3	2.2

^aAll risk ratios adjusted for age at time of procedure, comorbidity index, congestive heart failure, and days since last dialysis. Complete case analysis: all models except single-pool Kt/V included 30 cases and 323 controls due to missing data for days since last dialysis.

^bSingle-pool Kt/V was missing for 35 patients.

The only risk factor associated with increased risk that was not potentially due to the procedure indication was a diagnosis of heart failure. Risk was not associated with procedure setting, type of VA, proceduralist specialty, sex, race or ethnicity, or other comorbidities available on the USRDS dataset (data not shown). The lack of an association between death within 7 days and age or Liu Comorbidity Index indicates that cases and controls represent a similar pool of patients. The median age was 77 years for cases and 75 years for controls.

To our knowledge, this is the first US study evaluating the risk of mortality related to HD restoration endovascular procedures nationally that was controlled in a comprehensive manner. Because we restricted the study population to deaths within 90 days of a VA procedure, we limited bias related to mortality risk associated with the primary indication for the VA procedure (i.e., thrombosis) (⁶). Restriction to patients on HD for at least 1 year limited the increased mortality risks associated with early loss of AV patency (⁷). Restriction to the outpatient setting also reduced bias due to competing causes and confounding by comorbidities not captured in the USRDS or the comorbidity index. Further, because we only included patients for whom the indication for the endovascular procedure was thrombosis, we limited confounding by indication in our analyses from other causes of VA failure, such as infection, where the short-term risk of mortality is likely to be higher.

This study confirms the safety of endovascular interventions performed in the outpatient and office-based settings, at least within the patient population we studied. Given that procedure techniques and greater patient access to VA centers have likely improved survival in the time period after 2015, our results may be viewed as conservative. These findings should be reassuring to providers and patients. Our finding of increased risk associated with time since last dialysis supports early referral for intervention. Our design did not enable us to examine the time between VA procedure and subsequent dialysis. As a result, we could not determine if VA procedure followed immediately by dialysis was protective for individuals who had not been adequately dialyzed. Further, our design did not allow us to examine the risk of mortality among patients experiencing multiple VA procedures within 90 days, who may be at higher risk of mortality by indication. Additional studies are also needed to confirm the role of heart failure in short-term mortality and to investigate potential preventive measures for this population.

Disclosures

K.E. Cardone reports that her spouse is employed by Fresenius Medical Care; consultancy for Otsuka Pharmaceuticals, Vifor, and Wolters Kluwer Kelly; research funding from Merck; honoraria from the American College of Clinical Pharmacy, American Society of Health-System Pharmacy, and Pharmacy Times; and other interests or relationships with the Community Foundation for the Greater Capital District (grant funding) and New York State Department of Health (contract funding). M.K. Doll reports research funding from the St. Luke’s Wood River Foundation (subaward) and the US National Institute of Health (subaward). All remaining authors have nothing to disclose.

Funding

None.