Long-Term Outcomes Among Patients Discharged From the Hospital With Moderate Anemia

Abstract

Anemia is common in hospitalized patients and has been independently associated with short- and long-term morbidity and mortality (17). Traditionally, treatment of moderate anemia (hemoglobin levels between 7 and 10 g/dL) in hospitalized patients has included transfusion of allogeneic red blood cells (RBCs) to increase hemoglobin level (8). However, a series of randomized clinical trials showed similar outcomes with less frequent RBC transfusion, thereby avoiding potential infectious, cardiopulmonary, and immunomodulatory complications (916). The trial results supporting a more restrictive approach have been translated into practice guidelines by multidisciplinary societies, including medical, critical care, anesthesia, and pathology specialties (1722). These recommendations also play a key role in patient blood management strategies that emphasize evidence-based transfusion practice and the tolerance of anemia in addition to optimizing RBC counts and minimizing perioperative blood loss. Within our network of community hospitals, multidisciplinary patient blood management programs for specific clinical departments (such as cardiac and orthopedic surgery) were implemented beginning in 2010 (23, 24). Strategies focused on identification and treatment of suboptimal iron stores before surgery, increased use of cell salvage and hemostatic agents (such as tranexamic acid), and blood-sparing approaches to medical and surgical procedures. These department-specific initiatives were supported by hospital- and system-level quality improvement projects to improve blood use more broadly. System-level initiatives included promotion of evidence-based transfusion practice through peer review and clinical decision support within electronic order sets for blood transfusion. We found that a substantial reduction in RBC use was not associated with changes in 30-day or hospital mortality rates, consistent with short-term safety findings of randomized clinical trials and other observational studies (2327). However, the effect of restrictive transfusion practice on anemia persistence and longer-term clinical outcomes has not been well described (28, 29). In this study, we examined the incidence and prevalence of anemia in an integrated health care delivery system at and within 6 months of hospital discharge after implementation of blood management initiatives in our network. We hypothesized that an increase in moderate anemia would not adversely affect subsequent RBC transfusion, rehospitalization, or mortality events after hospital discharge. Methods Design Overview We conducted a retrospective cohort study using electronic health record data from Kaiser Permanente Northern California (KPNC), which serves 4 million members. We included all adult nonobstetric KPNC patients who were hospitalized and discharged alive from 21 hospitals during the 5 years from 1 January 2010 through 31 December 2014. The KPNC Institutional Review Board approved this study. Sample We identified all hospitalizations of adult patients. Patients were eligible if they survived an overnight inpatient hospital stay and were discharged from a KPNC hospital. We excluded patients who were hospitalized for observation or 1-day surgery that did not result in an overnight stay. We also excluded hospitalizations for childbirth; however, women who were hospitalized for postdelivery complications were included. Definition of Anemia We determined the presence and magnitude of anemia by using hemoglobin levels measured by the clinical laboratory during inpatient and outpatient care. We used the World Health Organization criteria for anemia, defined as a hemoglobin value of less than 12 g/dL in women and less than 13 g/dL in men (30). Hemoglobin levels below these thresholds but greater than 10 g/dL were defined as mild anemia. Moderate anemia was defined as a hemoglobin level less than 10 g/dL but greater than or equal to 7 g/dL. On the basis of RBC transfusion guidelines, severe anemia was defined as a hemoglobin level less than 7 g/dL (22). We defined admission hemoglobin as the lowest hemoglobin level within 72 hours before hospital entry or the most recent hemoglobin level within 30 days before an elective admission. We defined discharge hemoglobin as the most proximate hemoglobin level before hospital discharge and postdischarge hemoglobin as the initial hemoglobin level after hospital discharge. Outcomes Rates of RBC transfusion, rehospitalization, and mortality within 6 months of hospitalization were calculated for each patient discharged alive with moderate anemia. Resolution of anemia was assessed by examining all available hemoglobin levels within 6 months of hospital discharge, referencing the highest value during that period. Transfusion events were identified from blood bank transfusion records and included both inpatient and outpatient encounters. Rehospitalization and mortality rates were linked from admission records and death indices, respectively. Covariates We classified patients as having emergency or elective admission according to whether they were admitted through the emergency department. We classified hospitalizations as medical or surgical admission by the presence or absence of surgical procedural codes during hospitalization. We defined hospital entry as the time of initial admission to a nonemergency department inpatient location, including the general medicalsurgical wards, transitional care unit, intensive care unit, and operating room. To categorize patient diagnoses and comorbid conditions, we grouped International Classification of Diseases, Ninth Revision, diagnosis codes by using Health Care Utilization Project (www.ahrq.gov/data/hcup) single-level and multilevel Clinical Classifications Software categories. We quantified comorbid disease burden using the Charlson Comorbidity Index and Comorbidity Points Score, version 2, which is based on patients medical diagnoses within the 12 months before hospitalization (3133). We quantified illness severity at admission with the Laboratory-based Acute Physiology Score, version 2, which is based on laboratory test results, vital signs, and neurologic status within 72 hours before hospital entry. Statistical Analysis Descriptive analyses are presented as counts and percentages, medians and interquartile ranges, or proportions and 95% CIs. We used 2 tests for equal proportion or Wilcoxon rank-sum tests to test differences. For uncorrelated outcomes, we compared trends in dichotomous and continuous variables by using the 2 test for linear trend (CochranArmitage) and linear regression, respectively. Anemia prevalence and incidence were calculated at the hospitalization and patient levels, respectively. For anemia prevalence calculations, all patients who survived hospitalization and had moderate anemia at hospital discharge were included in the numerator, whereas the denominator was all inpatients discharged alive from the hospital. For anemia incidence calculations, each patient was counted once annuallywith the number of patients with moderate anemia in the numerator and all patients who survived hospitalization in the denominator. Therefore, separate admissions of individual patients were included in anemia prevalence calculations and in different years for incidence calculations. For all patients discharged alive with moderate anemia, we examined annual trends in RBC transfusion characteristics (receipt of inpatient RBC transfusion, pretransfusion hemoglobin level, number of RBC units received) as well as resolution of anemia within 6 months of discharge. We used a generalized estimating equation (GEE) to account for correlated observations and included continuous year as a covariate to test for time trends. For dichotomous (receipt of inpatient RBC transfusion) and continuous (pretransfusion hemoglobin level and number of RBC units received) outcomes, we used the modified log-Poisson and identity-Normal GEE link functions, respectively (34, 35). For the multicategory outcome (resolution of anemia), we used GEEs for nominal multinomial responses assuming a baseline category logit model (36). We selected the correlation structure of the GEE models by using a goodness-of-fit measure, QIC (quasi-likelihood under the independence model criterion). We reported results for dichotomous outcomes as risk ratios (RRs), continuous outcomes as mean change in outcome, and multicategory outcomes as odds ratios, all with corresponding 95% CIs. Among patients discharged alive with moderate anemia, we examined the adjusted annual prevalence of RBC transfusion, rehospitalization, and mortality within 1 and 6 months of hospitalization. We used the aforementioned modified log-Poisson GEE approach and included continuous year, number of hospitalizations, age, sex, emergency admission, surgical status, Charlson Comorbidity Index score, and severity of illness (Laboratory Acute Physiology Score, version 2) as covariates. In parallel to the model of 6-month mortality in patients with moderate anemia, we fit a modified log-Poisson GEE model for this outcome by using the same covariates for all other patients (those with mild or no anemia who survived hospital discharge). Last, we examined patient-level trends in unadjusted 6-month mortality after the first hospital discharge with moderate anemia in subgroups based on age, sex, emergency or elective surgical or medical status, presence of anemia on admission, prior RBC transfusion or anticoagulation, principal discharge diagnosis, and individual hospital site. For these patient-level observations, we used the 2 test for linear trend (CochranArmitage). Two-sided P values less than 0.050 were considered statistically significant. Statistical analyses were performed by using Stata, version 14.1 (StataCorp); SAS, version 9.4 (SAS Institute); and R, version 3.4.1 (The R Foundation). Role of the Funding Source The funders had no role in the design of the study; collection, ma