Health Facts Database Research Articles

Hemoglobin A1c and Reoperation After Surgery for Stress Incontinence or Prolapse.

Few studies compare the link between hemogobin A1c (HbA1c) and urogynecologic surgical complications. The objective of this study was to determine the association between HbA1c and reoperation in women undergoing surgery for stress urinary incontinence (SUI) or pelvic organ prolapse (POP). We performed 2 separate retrospective cohort analyses using Cerner's HealthFacts Database (750 hospitals; 519,000,000 patient encounters from January 1, 2010, to November 30, 2018). We included women undergoing surgery for (1) SUI or (2) apical POP by International Classification of Diseases coding who had HbA1c at the initial procedure. Each analysis compared those undergoing reoperation for complications or recurrence and those who did not. Multivariable logistic regression assessed the association between reoperation and HbA1c both as a continuous variable and comparing the commonly accepted cutoff ≥8. Of 30,180 SUI surgical procedures and 26,389 POP surgical procedures, 1,625 (5.4%) and 805 (3.1%) had HbA1c. Median (interquartile range) HbA1c in grams per deciliter was similar by reoperation status (SUI: 6.0 [5.6-6.8] vs 6.1 [5.6-6.9], P = 0.35; POP: 6.2 [5.6-6.6] vs 6.1 [5.7-6.8], P = 0.60). Reoperation was also similar using the HbA1c ≥8% cutoff (SUI: 6.9% vs 7.4%, P = 0.79; POP: 6.3% vs 5.4%, P = 0.77). On multivariate analysis, HbA1c value was not a significant predictor of reoperation either as a continuous (SUI: odds ratio [OR] = 0.966, 95% CI = 0.833-1.119; POP: OR = 1.040, 95% CI = 0.801-1.350) or dichotomous variable ≥8 (SUI: OR = 0.767, 95% CI = 0.407-1.446; POP: OR = 0.988, 95% CI = 0.331-2.951). Mean follow-up was 4.28-5.13 years. Although other studies have shown a link between diabetes and complications, we were unable to show an association between HbA1c values and rates of reoperation.

Incidence of Maternal and Perinatal Morbidity in Sickle Cell Disease and Sickle Cell Trait Patients during Pregnancy

Background: Sickle cell disease (SCD) patients have a higher risk of maternal and neonatal morbidity during pregnancy than the general population. Pregnancy in SCD is associated with increased metabolic demands, a hypercoagulable state, endothelial dysfunction, severe anemia, and increased vaso-occlusive episodes. We designed a study to retrospectively review the clinical outcomes of pregnant patients with SCD and SCD trait to examine the incidence of maternal and perinatal morbidity and mortality. We hypothesized that SCD patients should have a higher morbidity during pregnancy and SCD trait patients should not have higher morbidity. Methods: We conducted a retrospective cohort study and reviewed 29,822 patient encounters. Pregnant patients with the SCD genotype - Hb SS, Hb S beta 0 thalassemia, Hb S beta + thalassemia and Hb SC disease, and sickle cell trait (SCD trait) - Hb AS were included. Maternal clinical outcome variables for the entire pregnancy duration were vaso-occlusive crisis (VOC), deep vein thrombosis (DVT), pulmonary embolism (PE), infections - sepsis, septic shock, asymptomatic bacteriuria, acute pyelonephritis, and urinary tract infection (UTI), cerebrovascular accident (CVA), myocardial infarction (MI) and transient ischemic attack (TIA). Clinical outcome variables for pregnancy complications were preeclampsia, gestational diabetes mellitus (GDM), and C-section rate. Clinical outcome variables for neonatal complications included preterm birth and low birth weight. ICD 9 and ICD 10 codes were used to search for the pregnancy, SCD, sickle cell trait, maternal and neonatal outcomes. Data was collected from the Health Facts Database, which contains de-identified information from electronic medical records (EMR) of hospitals in which Cerner© has a data use agreement. The categorical clinical outcome variables were analyzed using Chi-square test. Results: There were 29,822 patient encounters reviewed. Patients with SCD had higher percentages for C-section rates, CVA, preeclampsia, sepsis, septic shock, and VOC compared to those with SCD trait. C-section rates were higher in SCD patients with 76.43% versus 13.76% in SCD trait patients. CVA was higher in SCD patients with 73.78% versus 0% in SCD trait patients. Preeclampsia was higher in SCD patients with 75.92% in SCD patients versus 5.43% in SCD trait patients. Sepsis and septic shock were higher in SCD patients with 73.75% in SCD patients versus 0% in SCD trait patients. VOC was evident in SCD patients with 0.06% versus 0% in SCD trait patients. Patients with SCD trait had higher percentages for asymptomatic bacteriuria, GDM, low birth weight, preterm birth, and UTI compared to patients with SCD. Asymptomatic bacteriuria was higher in SCD trait patients with 0.25% versus 0.08 in SCD patients. GDM was higher in SCD trait patients with 3.09% versus 0% in SCD patients. Low birth weight was higher in SCD trait patients with 0.61% versus 0.17% in SCD patients. Preterm birth was higher in SCD trait patients with 0.78% versus 0.44% in SCD patients. UTI was noted in SCD trait patients with 0.02% versus 0% in SCD patients. There were no findings for DVT, PE, acute pyelonephritis, MI, and TIA in both SCD and SCD trait patients. Conclusion: Pregnant patients with SCD were more likely to have C-sections, CVA, preeclampsia, sepsis, septic shock, and VOC. Pregnant patients with SCD trait were more likely to have asymptomatic bacteriuria, GDM, low birth weight, preterm birth, and UTI. Our findings indicate that prevention and management of CVA, preeclampsia, sepsis, and septic shock is important for pregnant patients with SCD to prevent adverse maternal and neonatal outcomes. The higher rate of C-sections may have been to reduce the severity of maternal and perinatal morbidity in SCD patients. Our study has significant limitations - retrospective nature and use of ICD9 and ICD10 codes for clinical outcome variables. Certain clinical outcome variables, such as CVA, sepsis or septic shock, may have been overestimated as these diagnoses may have been recorded prior to the start of pregnancy in the EMR. Nevertheless, SCD patients have higher maternal and perinatal morbidity that needs to be addressed.

Incidence of Postoperative Venous Thromboembolism, Hemorrhage, and Infection in Patients with Bleeding Disorders after Hip and Knee Arthroplasty

Introduction: Patients undergoing hip or knee arthroplasty (HA or KA) developed asymptomatic postoperative (post-op) venous thromboembolism (VTE) in around 30% of cases without any pharmacologic thromboprophylaxis (ThPPx). The use of pharmacological ThPPx after HA or KA has reduced the incidence of post-op VTE within 90 days to around 0.6-1.5%. People with bleeding disorders tend to have chronic joint damage from recurrent hemarthroses and frequently require HA or KA. This was especially true in the era before chronic factor prophylaxis. The exact risk estimate of post-op VTE in this population is unclear. Furthermore, the exact risk benefit for pharmacologic ThPPx is also unknown. Methods: A retrospective cohort study was conducted by looking at patients who underwent HA or KA and stratifying them based upon whether they had bleeding disorder diagnosis. Data was collected from the Health Facts database, which contains de-identified information from the electronic medical records of all US hospitals in which Cerner© has a data use agreement. Patients with bleeding disorders included Hemophilia A, B, C and Von Willebrand disease. Incidence rates of post-op VTE, hemorrhage (Hge) and infection (Inf) were calculated and compared between patients with bleeding disorders and those without bleeding disorders regardless of pharmacologic ThPPx. The same outcomes were obtained and compared between patients with/without bleeding disorders who received ThPPx. Chi-square and Fischer exact test were used to determine statistical significance. Results: 333,584 patients underwent HA or KA. Out of these, 438 patients had bleeding disorder. The incidence rates of post-op VTE were 0.01% in patients with bleeding disorders vs 0.02% in those without (p=0.1850) regardless of ThPPx which was not statistically significant. The incidence rate of post-op Hge was higher in patients with bleeding disorders (0.01%) compared to patients without bleeding disorder (0.004%; p=0.0404). The difference in incidence rate of Inf was not statistically significant between the two groups (p=0.1768). Out of total patients, 202,936 patients who underwent KA or HA received ThPPx. The difference in incidence rates of VTE as well as Inf were not statistically significant between patients with and without bleeding disorders who received ThPPx (p=0.2357 for VTE and p=0.2519 for Inf). The incidence rate of post-op Hge in patients who received ThPPX and had bleeding disorder was 0.014% in contrast to 0.004% in patients without bleeding disorder, though, it did not reach statistical significance (p=0.0684). Conclusion: Patients with bleeding disorders have a similar risk of post-op VTE as the general population after HA or KA. Patients with bleeding disorders have higher post-op hemorrhagic complications. However, when comparing only those patients who received ThPPx, the difference in incidence of post-op Hge was not statistically significant. This implies that all patients including the ones with bleeding disorders, undergoing HA or KA may be treated with a similar ThPPx approach. Patients with bleeding disorders undergoing HA or KA do not have a higher predisposition to infections compared to the general population and may not require additional infection control measures. We did not look at the differences in factor treatment perioperatively which would have been standard for all bleeding disorder patients if they followed in a hemophilia treatment center. Our study limitation was its retrospective nature as well as all HA, KA, post-op VTE, Hge and Inf were identified by CPT and ICD codes.

Hospital survival following pediatric HSCT: changes in complications, ICU therapies and outcomes over 10 years.

Hematopoietic stem cell transplantation (HSCT) is an increasingly utilized therapy for malignant and non-malignant pediatric diseases. HSCT complications, including infection, organ dysfunction, and graft-versus-host-disease (GVHD) often require intensive care unit (ICU) therapies and are associated with mortality. Our aims were to identify the HSCT characteristics, complications and ICU therapies associated with (1) survival, and (2) survival changes over a ten-year period in a national dataset. A national sample from the Health Facts (Cerner Corporation, Kansas City, MO) database from 2009 to 2018 was utilized. Inclusion criteria were age 30 days to <22 years and HSCT procedure code. For patients with >1 HSCT, the first was analyzed. Data included demographics, hospital length of stay (LOS), hospital outcome, transplant type and indication. HSCT complications included GVHD and infections. ICU therapies were positive pressure ventilation (PPV), vasoactive infusion, and dialysis. Primary outcome was survival to discharge. Statistical methods included bivariate analyses and multivariate logistic regression. 473 patients underwent HSCT with 93% survival. 62% were allogeneic (89% survival) and 38% were autologous (98% survival). GVHD occurred in 33% of allogeneic HSCT. Infections occurred in 26% of all HSCT. ICU therapies included PPV (11% of patients), vasoactive (25%), and dialysis (3%). Decreased survival was associated with allogeneic HSCT (p < 0.01), GVHD (p = 0.02), infection (p < 0.01), and ICU therapies (p < 0.01). Survival improved from 89% (2009-2013) to 96% (2014-2018) (p < 0.01). Allogeneic survival improved (82%-94%, p < 0.01) while autologous survival was unchanged. Survival improvement over time was associated with decreasing infections (33%-21%, p < 0.01) and increasing vasoactive infusions (20%-28%, p = 0.05). On multivariate analysis, later time period was associated with improved survival (p < 0.01, adjusted OR 4.28). Hospital survival for HSCT improved from 89% to 96% from 2009 to 2018. Factors associated with mortality included allogeneic HSCT, GVHD, infections and ICU therapies. Improving survival coincided with decreasing infections and increasing vasoactive use.

Changes in serum potassium with concomitant administration of renin-angiotensin system agents and potassium-sparing diuretics: U.S. cohort study using electronic health record data

BackgroundPotassium-sparing diuretics (KSDs) and renin-angiotensin system (RAS) agents (including angiotensin-converting enzyme inhibitors [ACEIs] and angiotensin II receptor blockers [ARBs]) are commonly prescribed for the management of hypertension, cardiovascular diseases, and diabetic nephropathy. However, there is a concern of developing hyperkalemia when these drugs are concomitantly used. This study investigates the change in serum potassium during concomitant administration of RAS agents (ACEI or ARB) and KSD. MethodsThis is a retrospective observational study using the Cerner Health Facts database. Electronic health records from unique patient hospitalization encounters were used to compare change in the serum potassium among 5 different cohorts (ACEI, ARB, KSD, ACEI-KSD, and ARB-KSD) and a non-RAS non-KSD exposed one (acetaminophen [APAP] cohort). Descriptive data of the cohorts and the risk of increase in potassium serum were estimated by generalized linear regression. ResultsA total of 5816 patients’ unique encounters were analyzed. After admission, 5.2% (N = 303) had serum potassium greater than 5.5 mmol/L. Results showed a significant increase in serum potassium during concomitant use of KSD and RAS agents and when only using KSD compared with the control APAP group (all P < 0.05). Other significant predictors of increased serum potassium levels included male (P = 0.002), black race (P < 0.001), and individuals with low estimated glomerular filtration rate (P < 0.001). ConclusionsThis study found no evidence of hyperkalemia when combining KSD and other medications that may increase potassium levels, even when controlling for other potential factors that may affect serum potassium levels, for patients within institutional settings. Warnings concerning excessive potassium levels when a KSD is prescribed with a medication that may affect potassium should be limited to patients with pre-existing higher levels of potassium.

Outcomes of Hospital-Acquired Hypernatremia.

Hospital-acquired hypernatremia is highly prevalent, overlooked, and associated with unfavorable consequences. There are limited studies examining the outcomes and discharge dispositions of various levels of hospital-acquired hypernatremia in patients with or without CKD. We conducted an observational retrospective cohort study, and we analyzed the data of 1,728,141 patients extracted from the Cerner Health Facts database (January 1, 2000, to June 30, 2018). In this report, we investigated the association between hospital-acquired hypernatremia (serum sodium [Na] levels >145 mEq/L) and in-hospital mortality or discharge dispositions with kidney function status at admission using adjusted multinomial regression models. Of all hospitalized patients, 6% developed hypernatremia after hospital admission. The incidence of in-hospital mortality was 12% and 1% in patients with hypernatremia and normonatremia, respectively. The risk of all outcomes was significantly greater for serum Na >145 mEq/L compared with the reference interval (serum Na, 135-145 mEq/L). In patients with hypernatremia, odds ratios (95% confidence interval) for in-hospital mortality, discharge to hospice, and discharge to nursing facilities were 14.04 (13.71 to 14.38), 4.35 (4.14 to 4.57), and 3.88 (3.82 to 3.94), respectively ( P < 0.001, for all). Patients with eGFR (Chronic Kidney Disease Epidemiology Collaboration) 60-89 ml/min per 1.73 m 2 and normonatremia had the lowest odds ratio for in-hospital mortality (1.60 [1.52 to 1.70]). Hospital-acquired hypernatremia is associated with in-hospital mortality and discharge to hospice or to nursing facilities in all stages of CKD.

Reliability of Admission Procalcitonin Testing for Capturing Bacteremia Across the Sepsis Spectrum: Real-World Utilization and Performance Characteristics, 65 U.S. Hospitals, 2008-2017.

Serum procalcitonin is often ordered at admission for patients with suspected sepsis and bloodstream infections (BSIs), although its performance characteristics in this setting remain contested. This study aimed to evaluate use patterns and performance characteristics of procalcitonin-on-admission in patients with suspected BSI, with or without sepsis. Retrospective cohort study. Cerner HealthFacts Database (2008-2017). Adult inpatients (≥ 18 yr) who had blood cultures and procalcitonin drawn within 24 hours of admission. None. Testing frequency of procalcitonin was determined. Sensitivity of procalcitonin-on-admission for detecting BSI due to different pathogens was calculated. Area under the receiver operating characteristic curve (AUC) was calculated to assess discrimination by procalcitonin-on-admission for BSI in patients with and without fever/hypothermia, ICU admission and sepsis defined by Centers for Disease Control and Prevention Adult Sepsis Event criteria. AUCs were compared using Wald test and p values were adjusted for multiple comparisons. At 65 procalcitonin-reporting hospitals, 74,958 of 739,130 patients (10.1%) who had admission blood cultures also had admission procalcitonin testing. Most patients (83%) who had admission day procalcitonin testing did not have a repeat procalcitonin test. Median procalcitonin varied considerably by pathogen, BSI source, and acute illness severity. At a greater than or equal to 0.5 ng/mL cutoff, sensitivity for BSI detection was 68.2% overall, ranging between 58.0% for enterococcal BSI without sepsis and 96.4% for pneumococcal sepsis. Procalcitonin-on-admission displayed moderate discrimination at best for overall BSI (AUC, 0.73; 95% CI, 0.72-0.73) and showed no additional utility in key subgroups. Empiric antibiotic use proportions were not different between blood culture sampled patients with a positive procalcitonin (39.7%) and negative procalcitonin (38.4%) at admission. At 65 study hospitals, procalcitonin-on-admission demonstrated poor sensitivity in ruling out BSI, moderate-to-poor discrimination for both bacteremic sepsis and occult BSI and did not appear to meaningfully alter empiric antibiotic usage. Diagnostic stewardship of procalcitonin-on-admission and risk assessment of admission procalcitonin-guided clinical decisions is warranted.

Impact of elevated body mass index on burn injury–associated mortality in a representative US sample

BackgroundThe impact of obesity on burn-related mortality is inconsistent and incongruent; despite being a risk factor for numerous comorbidities that would be expected to increase complications and worsen outcomes, there is evidence of a survival advantage for patients with high body mass index—the so-called obesity paradox. We used a national data set to explore further the relationship between body mass index and burn-related mortality. MethodsDeidentified data from patients with second and third-degree burns between 2014 and 2018 were obtained from the Cerner Health Facts Database. Univariate and multivariate regression models were created to identify potential factors related to burn-related mortality. A restricted cubic spline model was built to assess the nonlinear association between body mass index and burn-related mortality. All statistical analyses were conducted using R (R Foundation for Statistical Computing). ResultsThe study included 9,405 adult burn patients. Univariate and multivariate analyses revealed that age (odds ratio = 2.189 [1.771, 2.706], P < .001), total burn surface area (odds ratio = 1.824 [1.605, 2.074], P < .001), full-thickness burns (odds ratio = 1.992 [1.322, 3.001], P < .001), and comorbidities (odds ratio = 2.03 [1.367, 3.014], P < .001) were associated with increased mortality. Sensitivity analysis showed similar results. However, a restricted cubic spline indicated a U-shaped relation between body mass index and burn-related mortality. The nadir of body mass index was 28.92 kg/m2, with the lowest mortality. This association persisted even after controlling for age, total burn surface area, full-thickness burns, and comorbidities, which all remained significant. ConclusionThis study confirms a U-shaped association between body mass index and burn-related mortality along with age, total burn surface area, full-thickness burns, and comorbidities as risk factors.

Association between piperacillin/tazobactam MIC and survival among hospitalized patients with Enterobacterales infections: retrospective cohort analysis of electronic health records from 161 US hospitals.

A recent randomized trial has suggested an increased risk of mortality for ceftriaxone-non-susceptible Enterobacterales infections treated with piperacillin/tazobactam compared with meropenem despite MICs within the susceptible range. We conducted a retrospective cohort study of clinical encounters within the Cerner Health Facts database to identify all encounters between 2001 and 2017 in which Enterobacterales infections were treated empirically with piperacillin/tazobactam and for which MICs to the drug were available. Multivariate regression analysis was performed to enable partitioning of MICs into discrete strata based on statistically significant difference in mortality risk. During the study period, 10 101 inpatient encounters were identified meeting inclusion criteria. The crude in-hospital mortality for the entire cohort was 16.5%. Partitioning analysis identified a breakpoint of ≤16/4 mg/L that dichotomized encounters into lower versus higher mortality risk strata in the primary cohort of overall infections. This finding persisted in sequentially granular subsets where specific MICs ≤8/4 mg/L were reported (in lieu of ranges) as well as in the high-reliability subset with bloodstream infections. A higher clinical breakpoint of ≥128/4 mg/L dichotomized encounters with respiratory tract infection. No breakpoint was identified when restricting to encounters with urinary tract infections, ICU admits or upon restricting analysis to encounters with ceftriaxone-resistant isolates. Clinical data suggest improved outcomes when piperacillin/tazobactam is prescribed for Enterobacterales infections with an MIC of ≤16/4 mg/L compared with ≥32/4 mg/L.

Effect of Hepatitis C Treatment on Risk of Developing Cardiovascular Disease and Cardiac Arrhythmias

Background: Chronic hepatitis C virus (HCV) infection affects more than 70 million people worldwide, is a systemic disease and has been implicated as a risk factor for cardiovascular disease (CVD). Objective: We sought to study the association between risk of CVD between 0-12 month and 12-24 months of HCV treatment initiation. Methods: 567,956 Hepatitis C patients without prior history of CVD were identified in the Cerner Health Facts database. Of these, 1446 patients received HCV treatment. We included both demographic and as covariates in the multivariate logistic regression model. To control for baseline differences among treatment and control groups, propensity score matching (PSM) comparative analysis was used to adjust for potential confounding baseline characteristics between these two groups, and finally 2892 patients (1:1 match) (1446 in each group) were enrolled in the final analysis. Results: There was no statistically significant association between CVD and treatment of HCV (at both 0-12 and 12-24 months). The odds of developing CVD for females were 1.52 times greater than males during the first 12 months of treatment. Advanced age was also associated with a high risk of CVD during the initial 12 months (OR 1.05 95% CI 1.03-1.07). HCV patients with HIV were 45% less likely to experience CVD Expanded =0.55, 95% CI: 0.33- 0.91, p&lt;0.05) compared to those without HIV for 12-24 months. Treatment’s association with the development of cardiac arrhythmias was not statistically significant between 0-12 months or 12-24 months. Older age was associated with a higher risk of cardiac arrhythmias following 0-12 months after treatment with OR 1.08 95% CI 1.04-1.12. There was also a significantly high risk of arrhythmias with hypothyroidism for 0-12 months (OR 4.23 95% CI 1.48-12.08) Conclusion: Female sex and advanced age were associated with increased odds of CVD, and advanced age and hypothyroidism were associated with increased odds of arrhythmias development during the first 12 months of treatment. Further studies are needed to explore this in a prospective fashion.

Timely Albumin Improves Survival in Patients With Cirrhosis on Diuretic Therapy Who Develop Acute Kidney Injury: Real-World Evidence in the United States.

Patients admitted with decompensated cirrhosis who develop acute kidney injury (AKI) tend to experience poor outcomes, even if provided with increased organ support such as renal replacement therapies. We assessed the association of albumin administered ≤24 hours of admission with hospital length of stay (LOS) and in-hospital mortality. The Cerner Health Facts Database was queried for hospitalized patients with cirrhosis who had >0.3 mg/dL increase in serum creatinine within 48 hours and received diuretics following admission between January 2009 and April 2018. This study received institutional review board exemption through federal regulation 45CFR46. Albumin infusion was "timely" if administered ≤24 hours after admission and "nontimely" if administered >24 hours after admission or not at all. Two subgroups were assessed: the AKILOS subgroup (patients who survived to discharge) and the AKIMORTALITY RISK subgroup (patients with the highest risk of mortality, ie, AKI stage 3). A total of 4135 hospitalizations with cirrhosis and AKI were grouped into AKILOS (n= 3321) and AKIMORTALITY RISK (n= 609) subgroups. Albumin administration occurred in 59.7% of the AKILOS subgroup and 77.8% of the AKIMORTALITY RISK subgroup, but timely treatment only occurred in 25.9% and 35.8% of encounters within these subgroups, respectively. Risk-adjusted analysis showed timely albumin administration to be associated with a 15.5% reduction (P < .01) in LOS in the AKILOS subgroup and a 49% reduction in the odds of death (adjusted odds ratio: 0.51; P < .01) in the AKIMORTALITY RISK subgroup, when compared to the nontimely group. Among patients with cirrhosis and AKI, treatment with albumin ≤24 hours after admission was associated with a shorter LOS and lower risk of death in patients with stage 3 AKI.

364: PREDICTORS OF HOSPITAL MORTALITY FOLLOWING PEDIATRIC HEMATOPOIETIC STEM CELL TRANSPLANTATION (HSCT)

Introduction: HSCT is an increasingly utilized therapy for malignant and non-malignant pediatric diseases. HSCT complications, including infection, organ failure, and graft-versus-host disease (GVHD), often necessitate critical care interventions and are associated with mortality. Our aim was to examine the clinical variables and HSCT complications associated with survival. Methods: A multicenter, 2009-2018 cohort from the Health Facts (Cerner Corporation, Kansas City, MO) database was used. Inclusion criteria were age 30 days to < 22 years and HSCT procedure code. For patients with >1 HSCT, only the first was analyzed. Data included demographics, hospital length of stay (LOS) and outcome, diagnostic and procedure codes, medications and laboratory results including microbiology. Critical care interventions included positive pressure ventilation (PPV), vasoactive infusion, and dialysis. Primary outcome was survival to transplant admission discharge. Results: 477 patients underwent HSCT with 92.7% survival. Survival from 2009-2013 was 88%, increasing to 95% from 2014-2018 (p< 0.01). 62% were allogeneic and 38% were autologous, with 90% and 98% survival respectively (p< 0.01, OR 0.2 [0.1-0.6]). Common indications for transplant were leukemia (34%), solid malignancy (33%), bone marrow failure (9%) and lymphoma (8%). After controlling for transplant type, indication was not associated with survival. GVHD occurred in 32% of allogeneic HSCT and was associated with decreased survival (p=0.02, OR 0.4 [0.2-0.9]). Hospital LOS was 31 days [23-43] for survivors and 85 days [63-116] for non-survivors (p< 0.01). PPV, vasoactive infusion, and dialysis were utilized in 11%, 26%, and 3% of patients respectively. Positive cultures occurred in 27% of patients. Variables associated with decreased survival included PPV (p< 0.01 [OR 0.1 [0.05-0.2]), vasoactive infusion (p< 0.01 OR 0.03 [0.01-0.06]), dialysis (p< 0.01 OR 0.03 [0.01-0.09]), and positive blood (p< 0.01, OR 0.2 [0.1-0.5]) or respiratory (p< 0.01 OR 0.1 [0.05-0.3]) culture. Conclusions: Mortality following HSCT improved during the study period. One third of patients required at least one of the assessed critical care interventions. Factors associated with poor outcome were GVHD, infectious complications, and need for critical care interventions.

Characteristics Associated With Failure to Rescue After Open Abdominal Aortic Aneurysm Repair

IntroductionFailure to Rescue (FTR), defined as mortality following a complication of care, is an important indicator of hospital care quality. Understanding risk factors associated with FTR in the elective Abdominal Aortic Aneurysm (AAA) population may help surgeons prevent operative mortality. MethodsElective open AAA repairs (2008-2018) were identified from Cerner's HealthFacts database using ICD-9 and ICD-10 diagnosis and procedure codes. Patient, hospital, and encounter characteristics were analyzed. Multivariate logistic regression models determined the relative contribution of patient and encounter characteristics leading to FTR. ResultsFor 1761 patients who underwent open repair for nonruptured AAA, overall mortality was 6.1%. Of patients with one or more complications (40%), mortality was 9.6%, increasing to 21.5% for patients with ≥4 major complications. Complications of care most associated with death were myocardial infarction (MI), gastrointestinal (GI) bleeding, and pulmonary failure. After multivariable adjustment, FTR was associated with advanced age (odds ratio [OR] 1.19 for 5 y, 95% confidence interval [CI] 1.06-1.34); female sex (OR 1.74, 95% CI 1.12-2.70); congestive heart failure (OR 1.65, 95% CI 1.00-2.73); peptic ulcer disease (OR 3.99, 95% CI 1.18-13.5); diabetes (OR 4.90, 95% CI 1.90-12.6), and the number of complications of care. ConclusionsComplications of care were common following open elective AAA repair. The complications with the highest mortality included MI, GI bleeding, and respiratory failure. FTR was associated with female sex, comorbidities, and increasing numbers of complications of care. Often, the lowest occurring complications had the highest FTR. Adopting gender-specific assessment tools, a protocol-driven approach for perioperative GI prophylaxis, and preoperative MI risk mitigation may lead to reduced FTR.

The Kansas City Bleeding Index: A Clinical Assessment Tool to Predict Risk of Major Bleeding Events on Doac Therapy

BACKGROUND Direct oral anticoagulants (DOACs) are widely used medications for treatment and prevention of venous thromboembolism (VTE) and arterial thromboembolism (ATE). DOACs are favorable to the more traditional oral vitamin K antagonist (VKA)-as they do not require routine laboratory monitoring, generally have fewer drug interactions, and can be safely taken with all foods. However, DOACs can also increase the risk of bleeding events which can be severe and life-threatening. Moreover, many patients requiring DOAC therapy have chronic medical conditions, are of older age, and have history of cancers-putting them at increased risk of bleeding events. There have been multiple attempts to predict bleeding risk in patients that need DOACs, but no standard validated tool is preferred in clinical practice. Our study aims to quantify the risk of major bleeding while on DOAC therapy based upon patient characteristics. We hypothesized based upon prior evidence that a group of risk factors (Table 1) - liver disease, cerebral disease, heart disease, history of surgery, kidney disease, blood disorders, cancers, history of fall and type 2 diabetes mellitus - each increase the risk of bleeding during DOAC therapy. METHODS We retrospectively analyzed patient charts of 210,682 patients prescribed DOAC therapy (apixaban, rivaroxaban, edoxaban, betrixaban and dabigatran) who had a clinically significant bleeding event at 6 months and 5 years that required a hospital visit. ICD 9 and ICD 10 codes were used to search for bleeding events at different body sites and for risk factor groups based upon diagnosis. Kansas City Bleeding Index (KCBI) risk score from 0 to 9 increments was assigned depending upon the number of grouped risk factors present. Data was collected from the Health Facts database, which contains de-identified information from the electronic medical records of hospitals in which Cerner© has a data use agreement. Multivariate logistic regression modeling was used to calculate bleeding risk. RESULTS A total of 210,682 patients were reviewed, with an average age of 68 years old (18 to 90 years). KCBI risk scores in patients ranged from 0-7 (maximum score 9); however, none of the patients had risk of 8 or 9. Majority of patients had KCBI risk scores between 1 and 2. Of the 210, 682 patients, 7.07% had bleeding events by 6 months and 10.86% had bleeding events by 5years. KCBI risk score, Caucasian race, female gender and age were all statistically significant risk factors to predict bleeding risk at 6 months (Table 2; p value <0.05). With a one-point increase in KCBI risk score, the chance of bleeding at 6 months increased by 46.9%. Caucasian race was associated with protection from bleeding with a 21.2% less likely chance to have an event at 6 months. Female gender had a 5.2% less likely chance to have a bleeding event at 6 months. With one-year increase in age, the chance of a bleeding event at 6 months increased by 0.8%. At 5 years, KCBI risk score, Caucasian race and age were significant risk factors to predict bleeding (Table 2; p value < 0.05). With a one-point increase in KCBI score, the chance of bleeding at 5 years on DOACs increased by 40.6%. Again, Caucasian race was 18.5% less likely to have a bleeding event at 5 yrs. With a one-year increase in age, the chance of bleeding at 5 years increased by 0.8%. CONCLUSION Several risk factors exist that increase the risk of bleeding in patients on DOACs and calculating the actual risk in clinical setting is complicated. Younger age, female gender and Caucasian race were all associated with lower bleeding risk. KCBI risk score (Table 1), can be used to predict the likelihood of clinically significant bleeding requiring hospital visit based upon our study. A higher KCBI risk score correlates with a medically complex patient with higher bleeding risk requiring greater caution during anticoagulation. However, actual validation of the KBCI risk score as a bleeding predictor tool might be needed in a prospective study in an ambulatory setting. Nevertheless, it can be still used as a tool for predicting risk both at 6 months and 5 years, especially if the benefit of anticoagulation is low. Our study was limited by retrospective analysis, in-patient demographic, sicker patient condition, and clinically significant bleeding requiring hospital visit. Additionally, our patient demographic was predominantly Caucasian, limiting generalizability. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal

An integrated LSTM-HeteroRGNN model for interpretable opioid overdose risk prediction

Opioid overdose (OD) has become a leading cause of accidental death in the United States, and overdose deaths reached a record high during the COVID-19 pandemic. Combating the opioid crisis requires targeting high-need populations by identifying individuals at risk of OD. While deep learning emerges as a powerful method for building predictive models using large scale electronic health records (EHR), it is challenged by the complex intrinsic relationships among EHR data. Further, its utility is limited by the lack of clinically meaningful explainability, which is necessary for making informed clinical or policy decisions using such models. In this paper, we present LIGHTED, an integrated deep learning model combining long short term memory (LSTM) and graph neural networks (GNN) to predict patients' OD risk. The LIGHTED model can incorporate the temporal effects of disease progression and the knowledge learned from interactions among clinical features. We evaluated the model using Cerner's Health Facts database with over 5 million patients. Our experiments demonstrated that the model outperforms traditional machine learning methods and other deep learning models. We also proposed a novel interpretability method by exploiting embeddings provided by GNNs to cluster patients and EHR features respectively, and conducted qualitative feature cluster analysis for clinical interpretations. Our study shows that LIGHTED can take advantage of longitudinal EHR data and the intrinsic graph structure of EHRs among patients to provide effective and interpretable OD risk predictions that may potentially improve clinical decision support.

Outcomes of Patients With Primary Cardiac Diagnoses Admitted to Cardiac vs Noncardiac Intensive Care Units

BackgroundDemographics in cardiac intensive care units (CICUs) have evolved, with increased prevalence of noncardiac critical illnesses. ObjectivesThis study compares outcomes of patients with primary cardiac diagnoses admitted to CICUs vs those of patients with primary cardiac diagnoses admitted to noncardiac ICUs. MethodsThe Cerner Health Facts Database was queried to identify adults with primary cardiac diagnoses admitted to ICUs within 48 hours of presentation between 2009 and 2014. Only hospitals with multiple ICUs including a CICU were studied. Information on ICU staffing was not available. A univariate analysis of ICU type (model 1) and multivariate analyses incorporating patient- and hospital-level variables (model 2) and concurrent, noncardiac, ICU-level diagnoses (model 3) were utilized to assess the impact of ICU type on inpatient mortality. ResultsOf 16,163 encounters across 14 hospitals, 8,499 (52.6%) were admitted to CICUs and 7,664 (47.4%) to noncardiac ICUs. Univariate analysis (model 1) demonstrated increased mortality in noncardiac ICUs compared to CICUs (odds ratio [OR]: 1.47, 95% CI: 1.32-1.64; P < 0.0001). This risk dissipated (OR: 1.04, 95% CI: 0.91-1.18; P = 0.56) after controlling for patient- and hospital-level variables (model 2). Inclusion of concurrent, noncardiac, ICU-level diagnoses (model 3) lead to a reversal with decreased mortality in noncardiac ICUs (OR: 0.86, 95% CI: 0.76-0.98; P = 0.03). ConclusionsIn this historical cohort study evaluating CICU outcomes prior to the evolution of proposed staffing and care model modernization, survival of cardiac patients with concurrent, noncardiac critical illnesses may have been better with the expertise available in general system ICUs. These results may support contemporary efforts to increase the capacity to manage noncardiac critical illnesses in CICUs.

S1209 Longer Time to Recovery From Acute Kidney Injury Is Associated with Increased Major Adverse Kidney Events in Patients With Cirrhosis

Introduction: In patients with cirrhosis, non-recovery from acute kidney injury (AKI) is associated with major adverse kidney events (MAKE). However, in patients with AKI recovery, little is known about how the timing of recovery affects the risk of MAKE. Thus, we aimed to examine the association between timing of recovery and risk of MAKE in patients with AKI recovery. Methods: Hospitalized patients with cirrhosis and AKI in the Cerner Health Facts database from 1/2009-09/2017 were assessed for AKI recovery and were followed for 180 days for outcomes. The timing of AKI recovery [return of serum creatinine (sCr) < 0.3mg/dL of baseline] from AKI onset was grouped into 0-2, 3-7, 8-14, and >14 days. The primary outcome was MAKE at 90-180 days. Per consensus definition, MAKE was defined as the composite outcome of >25% decline in estimated glomerular filtration rate (eGFR) compared with baseline with CKD stage >3 or progression of CKD (defined as >50% reduction in eGFR compared with baseline) or new hemodialysis. Competing risk multivariable modeling (death/transplant as competing risk) was performed to determine the independent association between timing of recovery and risk of MAKE. Results: Out of 6,250 eligible patients, 4,655 (75%) achieved AKI recovery. The median age was 60 years [interquartile range (IQR) 25, 70], 71% White and 60% male. The most common etiologies of cirrhosis were non-alcoholic steatohepatitis (38%), alcohol (27%), and hepatitis C (17%), and the median (IQR) MELD-Na score was 23 (16, 28). 60% had ascites and the median baseline sCr was 1.00 (0.70, 1.44) mg/dL. The characteristics of patients who recovered 0-2 (n=2,791, 60%), 3-7 (n=1,455, 31%), , 8-14 (n=255, 5%), and >14 days (n=184, 4%) after AKI are shown in Table. The incidence of MAKE was 12%, 16%, 22%, and 25% for 0-2, 3-7, 8-14, and >14 days recovery groups, respectively. On adjusted multivariable competing risk analysis, compared to 0-2 days, recovery at 3-7, 8-14, and >14 days was independently associated with an increased risk for MAKE: sHR 1.48 (95% CI 1.03-2.14, p=0.036), sHR 2.92 (95% CI 1.11-4.31, p=0.023), and sHR 2.60 (95% CI 1.37-4.96, p=0.004), respectively. Conclusion: In patients with cirrhosis who recover from AKI, longer time to recovery is associated with an increased risk of major adverse kidney events. Interventions to hasten recovery from AKI should be considered in patients with cirrhosis who develop AKI. Table 1. - Comparison of Patient and Clinical Demographics Between AKI Recovery Groups Variable 0-2 Days N=2,791 3-8 Days N=1,455 8-14 days N=225 >14 days N=184 P-value Age 60 (52, 69) 62 (53, 71) 62 (55, 71) 60 (50, 68) < 0.001 Race, white 71 71 68 68 0.069 Sex, male 61 59 60 59 0.925 Etiology of cirrhosis Hepatitis C Alcohol NASH Other Unknown etiology 17 29 37 5 12 18 24 40 6 12 17 19 47 6 11 21 27 34 6 12 0.635 0.001 0.006 0.518 0.966 Ascites 56 66 68 73 < 0.001 Hepatic encephalopathy 25 27 31 32 0.054 Diabetes 50 55 64 54 < 0.001 Hypertension 57 60 63 9 0.044 Baseline creatinine, mg/dL 0.94 (0.70, 1.44) 1.00 (0.74, 1.42) 1.10 (0.80, 1.74) 1.07 (0.76, 1.82) < 0.001 MELD-Na at time of AKI 20 (14, 26) 24 (19, 29) 26 (21, 30) 26 (23, 31) < 0.001 Stage of AKI at diagnosis 1/2/3 87/10/3 69/20/11 61/21/18 63/14/13 < 0.001 Peak AKI stage 1/2/3 75/13/12 48/27/25 20/31/49 16/21/63 < 0.001 Infection 25 30 41 35 < 0.001 Required ICU care 24 28 37 33 < 0.001 Mechanical ventilation use 12 14 24 12 < 0.001 Vasopressor use 14 17 24 23 < 0.001 Continuous variables shown as median interquartile range (IQR); categorical variables presented as %.

Machine Learning in Infectious Disease for Risk Factor Identification and Hypothesis Generation: Proof of Concept Using Invasive Candidiasis.

Machine learning (ML) models can handle large data sets without assuming underlying relationships and can be useful for evaluating disease characteristics, yet they are more commonly used for predicting individual disease risk than for identifying factors at the population level. We offer a proof of concept applying random forest (RF) algorithms to Candida-positive hospital encounters in an electronic health record database of patients in the United States. Candida-positive encounters were extracted from the Cerner HealthFacts database; invasive infections were laboratory-positive sterile site Candida infections. Features included demographics, admission source, care setting, physician specialty, diagnostic and procedure codes, and medications received before the first positive Candida culture. We used RF to assess risk factors for 3 outcomes: any invasive candidiasis (IC) vs non-IC, within-species IC vs non-IC (eg, invasive C. glabrata vs noninvasive C. glabrata), and between-species IC (eg, invasive C. glabrata vs all other IC). Fourteen of 169 (8%) variables were consistently identified as important features in the ML models. When evaluating within-species IC, for example, invasive C. glabrata vs non-invasive C. glabrata, we identified known features like central venous catheters, intensive care unit stay, and gastrointestinal operations. In contrast, important variables for invasive C. glabrata vs all other IC included renal disease and medications like diabetes therapeutics, cholesterol medications, and antiarrhythmics. Known and novel risk factors for IC were identified using ML, demonstrating the hypothesis-generating utility of this approach for infectious disease conditions about which less is known, specifically at the species level or for rarer diseases.

Risk factors for thedevelopment of acute pyelonephritis in women with a positive urine culture.

In treating lower urinary tract symptoms (LUTS), the risk of overtreatment with antibiotics must be reconciled with the risk of an untreated urinary tract infection (UTI) progressing to acute pyelonephritis (APN). Using Cerner HealthFacts, a longitudinal clinical informatics database, we aimed to determine risk factors associated with the development of APN from UTI in an effort to guide the initiation of empiric antibiotics. We queried the Cerner HealthFacts database for women over age 18 with a positive urine culture. Any patient with an International Classification of Disease (ICD) code indicating chronic pyelonephritis was excluded. Development of APN within 30 days of the positive culture, specified by ICD coding, was our primary outcome. Patient and facility factors were assessed as potential risk factors for the development of APN using multivariable regression. Out of 58 344 women with a positive urine culture, 3.9% (2296) developed APN. Mean patient age was 54.4 ± 25.3 years. Overall, 12 variables were predictive for APN and 11 variables were protective against APN. Presence of obstructive and reflux uropathies (OR 4.58), presentation to an acute care facility (OR 3.19), urinary retention (OR 2.30), history of UTI (OR 2.19), and renal comorbidities (OR 2.07) conferred the highest odds of APN development. The most protective variable against APN development was cognitive impairment (OR 0.49). Identified risk factors associated with APN development may aid decisions regarding empiric antibiotic initiation for patients presenting with LUTS while awaiting urine culture results. The relationship between cognitive impairment and progression to APN deserves further study.

Estimating the prevalence of neurosurgical interventions in adults with spina bifida using the Health Facts data set: implications for transition planning and the development of adult clinics.

As the care of patients with spina bifida continues to evolve, life expectancy is increasing, leading to a critical need for transition planning from pediatric-based to adult-based care. The burden of neurosurgical care for adults with spina bifida remains unknown. In this study, the authors sought to use a large national data set to estimate the prevalence of neurosurgical interventions in adults with spina bifida. This study utilized Health Facts, which is a de-identified proprietary data set abstracted from all Cerner electronic health records. It includes 69 million unique patients with > 500 million encounters in 580 centers. Validation, technical exclusions, and data filters were applied to obtain an appropriate cohort of patients. The ICD-9 and ICD-10 codes for all types of spinal dysraphism, as well as the Current Procedural Terminology (CPT) codes for hydrocephalus procedures, spinal cord untethering, and Chiari decompression, were queried and records were retrieved. Demographic variables along with differences in age groups and temporal trends were analyzed. Overall, 24,764 unique patients with ≥ 1 encounter with a spinal dysraphism diagnosis between 2000 and 2017 were identified. The pediatric cohort included 11,123 patients with 60,027 separate encounters, and the adult cohort included 13,641 patients with 41,618 separate encounters. The proportion of females was higher in the adult (62.9%) than in the pediatric (51.4%) cohort. Annual encounters were stable from 2 to 18 years of age, but then decreased by approximately half with a precipitous drop after age 21 years. The sex distribution of adults and children who underwent procedures was similar (54.6% female adults vs 52.4% female children). Surgical interventions in adults were common. Between 2013 and 2017, there were 4913 procedures for hydrocephalus, with 2435 (49.6%) adult patients. Similarly, 273 (33.3%) of the 819 tethered cord procedures were performed in adults, as were 307 (32.9%) of 933 Chiari decompressions. The Health Facts database offered another option for studying care delivery and utilization in patients aging with spina bifida. The median age of this population has now reached early adulthood, and a significant number of neurosurgical procedures were performed in adults. An abrupt drop in the rate of encounters occurred at 21 years of age, possibly reflecting transition issues such as access-to-care problems and lack of coordinated care.