Outcomes of Patients Receiving Maintenance Dialysis Admitted Over Weekends

  • Published on
    23-Dec-2016

  • View
    212

  • Download
    0

Embed Size (px)

Transcript

  • Original Investigation

    aike

    1,2 GD. N

    ave butcomributenatio

    missiopria

    200

    all-caital stl chaialysons o6 (5.

    138,517 (5.4%) for weekday admissions (P 0.001). In a multivariable model, patients admitted over

    Asiohaincintpueuneremencliityan

    faiofepwimilen

    Amweekends had higher all-cause in-hospital mortality (OR, 1.06; 95% CI, 1.01-1.10) in comparison to thoseadmitted over weekdays and higher mortality during the first 3 days of admission (OR, 1.18; 95% CI,1.10-1.26). Patients admitted over weekends were less likely to be discharged to home, had longer hospitalstays, and had shorter times to death compared with those admitted over weekdays on adjusted analysis.

    Limitations: Use of ICD-9-CM codes to identify patients, defining weekend as midnight Friday to midnightSunday.

    Conclusions: Maintenance dialysis patients admitted over weekends have increased mortality rates andlonger lengths of stay compared with those admitted over weekdays. Further studies are needed to identify thereasons for worse outcomes for weekend admissions in this patient population.Am J Kidney Dis. 62(4):763-770. 2013 by the National Kidney Foundation, Inc.

    INDEX WORDS: Dialysis; admissions; mortality.

    growing body of literature demonstrates unfa-vorable outcomes for weekend hospital admis-

    ns in comparison to weekdays. This phenomenons been highlighted in several patient populations,luding those with myocardial infarction,1 stroke,2racerebral hemorrhage,3 gastrointestinal bleeding,4lmonary embolism,5 ruptured abdominal aortic an-rysm,5 acute epiglottitis,5 and recently, acute kid-y injury.6 Although the cause of this phenomenon

    ains elusive, potential explanations include differ-tial staffing models with limited availability ofnician expertise, unmeasured differences in sever-of illness, and decreased accessibility to diagnostic

    d therapeutic procedures.End-stage renal disease (ESRD) defined as kidneylure necessitating kidney replacement therapy, mostten in the form of hemodialysis, is a growingidemic in the United States. The present population-de prevalence is estimated to be 1,700 patients perllion, with a projection of more than 700,000 preva-t patients by 2015.7 According to the US Renal

    Data System 2012 annual data report, there were 340new cases of ESRD per million population in 2010.8ESRD constitutes a unique group of patients who arehospitalized frequently, and the in-hospital mortalityis on average 6-8 times greater than that in the generalpopulation (Collins et al,8 volume 2, chapters 1 and5). Given their complex nature and propensity to

    From the Departments of 1Nephrology and Hypertension and2Quantitative Health Sciences, Cleveland Clinic, Cleveland, OH;Divisions of 3Pulmonary and Critical Care Medicine and 4Ne-phrology, Department of Medicine, Medical College of Wisconsin,Milwaukee, WI; and 5Renal-Electrolyte Division, University ofPittsburgh Medical Center, Pittsburgh, PA.

    Received October 9, 2012. Accepted in revised form March 6,2013. Originally published online May 13, 2013.

    Address correspondence to Ankit Sakhuja, MD, Department ofNephrology and Hypertension, Glickman Urological and KidneyInstitute, Cleveland Clinic, Q7, 9500 Euclid Ave, Cleveland, OH44195. E-mail: sakhuja@ccf.org

    2013 by the National Kidney Foundation, Inc.0272-6386/$36.00http://dx.doi.org/10.1053/j.ajkd.2013.03.014Outcomes of Patients Receiving MOver Wee

    Ankit Sakhuja, MD,1 Jesse D. Schold, PhD,Puneet Sood, MD,5 and Sankar

    Background: Hospital admissions over weekends hpatient populations. The cause of this difference in opatterns over weekends have been speculated to contnance dialysis therapy admitted over weekends using a

    Study Design: Retrospective cohort study.Setting & Participants: We included nonelective ad

    dialysis therapy (n 3,278,572) identified using apprRevision, Clinical Modification (ICD-9-CM) codes fordatabase.

    Predictor: Weekend versus weekday admission.Outcomes: The primary outcome measure was

    included mortality by day 3 of admission, length of hospMeasurements: We adjusted for patient and hospita

    primary discharge diagnosis common to maintenance dResults: There were an estimated 704,491 admissi

    Unadjusted all-cause in-hospital mortality was 40,66J Kidney Dis. 2013;62(4):763-770ntenance Dialysis Admittedndsagan Kumar, MD,3 Aaron Dall, MD,4avaneethan, MD, MPH1

    een associated with worse outcomes in differentes remains unclear; however, different staffing. We evaluated outcomes in patients on mainte-nal database.

    ons of adult patients (18 years) on maintenancete International Classification of Diseases, Ninth5-2009 using the Nationwide Inpatient Sample

    use in-hospital mortality. Secondary outcomesay, time to death, and discharge disposition.racteristics, payer, year, comorbid conditions, andis patients.ver weekends versus 2,574,081 over weekdays.8%) for weekend admissions in comparison to763

  • detiediareqtio

    detiopanorecallSadifwema

    Stu

    Cocar

    admResamas

    pridurma

    insregurbfacdisto

    TablinslocFacAmme

    fulhigfinone

    sta(smcor

    urbrur

    for

    Stu

    frodiatienReor

    (54

    (58negsioexc

    Stu

    admmitiooftio0-3denkidalrcodtosuppat5 gDiwhinc

    Ou

    ity.vo

    mo

    rependcheass

    admdifwiAf

    T

    FEskidnafica

    76

    Sakhuja et alvelop volume and metabolic derangements, pa-nts with ESRD, especially those on maintenancelysis therapy who need hospitalization, likely woulduire timely access to expert nephrology consulta-n and renal replacement therapy.Worse outcomes for weekend admissions, althoughmonstrated for many time-sensitive medical condi-ns, have not been studied in maintenance dialysistients. Thus, we aimed to study whether this phe-menon existed for hospital admissions of patientseiving maintenance dialysis. We used a large nation-y representative database (the Nationwide Inpatientmple [NIS] database) to describe the mortalityferences and other associated outcomes betweenekday and weekend admissions of patients onintenance dialysis therapy.

    METHODS

    dyDesign

    We designed a retrospective cohort study using the Healthcarest and Utilization Project-NIS, the largest all-payer inpatiente database publicly available in the United States. This is an

    inistrative data set created by the Agency for Healthcaresearch and Quality that contains data from the 20%-stratifiedple of US community hospitals.9 Each hospitalization is treated

    an individual entry in the database and is coded with onencipal diagnosis, up to 14 secondary diagnoses, and 15 proce-al diagnoses associated with that stay. NIS encompasses infor-tion for all patients regardless of payer, including privateurance and the uninsured. It also includes patient informationardless of type of hospital: teaching or nonteaching, rural oran, large or small volume, and private or publicly owned. Toilitate the projection of national estimates, both hospital andcharge weights are provided, along with information necessarycalculate the variance of estimates.

    he NIS collects information for common demographic vari-es, such as age, race, and sex, along with primary and secondaryurance and hospital-level characteristics such as teaching status,ation (rural vs urban), size of hospital, and hospital region.ilities are considered to be teaching hospitals if they have anerican Medical Associationapproved residency program, are a

    mber of the Council of Teaching Hospitals, or have a ratio ofltime-equivalent interns and residents to patients of 0.25 orher. Hospital location (rural/urban) and bed size also are de-ed. The bed size cutoff values are chosen so that approximately-third of the hospitals in a given region, location, and teaching

    tus combination would fall within each bed size categoryall, medium, and large). From 2004 in the NIS, hospitals with ae-based statistical area type of metropolitan were categorized asan, whereas hospitals with a core-based statistical area type ofal were categorized as rural. Data from 2005-2009 were usedthis study.

    dyPopulation

    We included all admissions of adult patients (aged 18 years)m the NIS database for 2005-2009 who were on maintenancelysis therapy. Maintenance dialysis status was defined as pa-ts with an International Classification of Diseases, Ninth

    vision, Clinical Modification (ICD-9-CM) code for ESRD (585.6)procedure code for hemodialysis (39.95) or peritoneal dialysis

    .98), but absence of ICD-9-CM code for acute kidney injury oth

    44.X). ICD-9-CM codes 584.X have 90% sensitivity andative predictive value for acute kidney injury.10 Elective admis-

    ns and patients who had undergone kidney transplantation wereluded from the study (Fig 1).

    dyVariables

    In accordance with previous studies, we defined weekendissions as those occurring after midnight on Friday through

    dnight on Sunday.5,11-13 We used the Deyo et al14 modifica-n of the Charlson Comorbidity Index to identify the burdencomorbid conditions. This index uses 17 comorbid condi-ns with differential weighting and total scores range from3, with higher scores representing greater comorbidity bur-. In accordance with the previous literature,15 we excludedney disease when calculating the index because that waseady included as patient characteristic. Using ICD-9-CMes, we also examined primary discharge diagnoses common

    maintenance dialysis patients (Table S1, available as onlineplementary material). We used NIS variables to identifyient age, sex, race, and primary payer. Age was divided intoroups: 18-34, 35-49, 50-64, 65-79, and 80 or greater years.

    scharges with missing data were excluded except for race,ich was missing in 20% of discharges. Missing race wasluded in primary analyses as a separate subgroup of race.

    tcomes

    Our main outcome of interest was all-cause in-hospital mortal-Additionally, we investigated length of hospital stay for survi-

    rs, time to death, and discharge disposition. We also examinedrtality by day 3 of admission because previous studies haveorted much larger differences in short-term mortality for week-

    admissions in comparison to weekday admissions.6 We alsocked for interaction between year and weekend admissions toess whether there were variations in mortality for weekend

    issions by year. Subgroup analyses were performed to look atferences in adjusted odds of mortality for weekend admissionsthin subgroups of age (65 or 65 years), sex, race (white vsrican American), primary payer (Medicare or Medicaid vs

    Total ESRD population(ESRD ICD-9-CM Code: 585.6 +

    Hemodialysis: 39.95 +Peritoneal dialysis: 54.98

    AKI ICD-9-CM Codes 584.X)N= 3,974,521

    otal Maintenance Dialysis Patientsn=3,779,880

    Study populationN=3,278,572

    Excluded if Kidney transplant: V42.0, 996.81N=194,641

    Excluded ifElective Admissionsn= 501,308

    igure 1. Selection of study population (age 18 years).timates from weighted survey data. Abbreviations: AKI, acuteney injury; ESRD, end-stage renal disease; ICD-9-CM, Inter-tional Classification of Diseases, Ninth Revision, Clinical Modi-tion.ers), hospital bed size, and hospital teaching status. The regres-

    Am J Kidney Dis. 2013;62(4):763-770

  • siotes

    Sta

    we

    nu

    ma

    betwe

    we

    traW

    mu

    wa

    hoswe

    var

    mo

    andbotunisur

    regW

    andpitliteappmo

    admdishomo

    intho

    robsam

    simandfirswitdatunithethialomo

    sen

    ingmiity

    Pa

    frama3,1alyCI

    BasehotiebecaPrnaacinatrblge

    Mo

    5.4dasioria6%we[OUnlar0.0foovbyted

    juwira(Msizadho(Fovyeni

    Di

    beferthoTaanen(Tovad

    Am

    Weekend Maintenance Dialysis Admissionsn coefficients were compared using a generalized Hausmant.16,17

    tistical Analysis

    Stata IC 11.0 (StataCorp LP) was used for all analyses. We usedights provided with the NIS to generate national estimates of thember of admissions in each age group using the survey com-nds. The 2 test was used to compare categorical variablesween patients admitted on the weekend and those admitted onekdays. Because length of stay of survivors and time to deathre not normally distributed, t test was performed on log-nsformed values.

    e examined factors associated with in-hospital mortality byltivariable logistic regression. Univariable logistic regressions used to identify risk factor variables associated with in-pital mortality. All variables that were significant at P 0.10re included in the final multivariable model. To control foriability by year, we used year as a predictor in regressiondels. We checked variables for multicollinearity using tolerancevariance inflation factor. For variables used in the final model,

    h tolerance and variation inflation factor were very close toty. Similarly, for factors associated with length of stay ofvivors and time to death, univariable and multivariable linearressions were performed.

    e examined for interactions between weekend admissionscardiovascular diseases,2,18-21 gastrointestinal bleed,11 hos-

    al bed size,6 or teaching status,22 as has been suggested in therature. Significant interaction terms were included in theropriate final multivariable models. For all-cause in-hospitalrtality, we found significant interactions between weekendission and primary discharge diagnosis of cardiovascular

    ease and weekend admission and teaching status of thespital, which were both included in the final regressiondel. Similarly for length of stay of survivors, significanteraction with weekend admissions was found only withspital bed size.We performed 5 sensitivity analyses to further evaluate theustness of our results. To account for multiple admissions ofe patients (readmissions), we identified hospital records withilar age, sex, race, primary payer, hospital identification code,year of admission, as done previously in literature.23 For the

    t sensitivity analysis, we included all unique observations alongh one of each duplicate observation randomly selected from theabase. For the second sensitivity analysis, we included allque observations along with one of each duplicate admitted infirst month for that duplicate observation. Similarly for the

    rd sensitivity analysis, we included all unique observationsng with one of each duplicate observation admitted in the lastnth for that duplicate observation. Finally, we did anothersitivity analysis including only unique observations and exclud-any duplicate observations (Table S2). Because race was

    ssing for20% of observations, we performed another sensitiv-analysis excluding race.

    RESULTS

    tient Characteristics

    Using the NIS database, during the 5-year time-me of our study (2005-2009), there were an esti-ted 3,278,572 (95% confidence interval [CI],04,322-3,452,822) discharges on maintenance di-sis therapy. Of those, an estimated 704,491 (95%

    , 654,945-754,038) were admitted over weekends. 95

    J Kidney Dis. 2013;62(4):763-770seline characteristics of patients in terms of age,x, other demographics, comorbid conditions, andspital characteristics are listed in Table 1. Pa-nts admitted over weekends were more likely toaged 18-49 or 80 years or older, African Ameri-

    ns, and admitted less often to teaching hospitals.imary discharge diagnoses common to mainte-nce dialysis patients included any infection orcess complication, cardiovascular disease (includ-g heart failure, acute coronary syndrome, andial fibrillation/flutter and stroke)...

Recommended

View more >