Jianfang Liu

Challenges Associated With Using Large Data Sets for Quality Assessment and Research in Clinical Settings

The rapidly expanding use of electronic records in health-care settings is generating unprecedented quantities of data available for clinical, epidemiological, and cost-effectiveness research. Several challenges are associated with using these data for clinical research, including issues surrounding access and information security, poor data quality, inconsistency of data within and across institutions, and a paucity of staff with expertise to manage and manipulate large clinical data sets. In this article, we describe our experience with assembling a data-mart and conducting clinical research using electronic data from four facilities within a single hospital network in New York City. We culled data from several electronic sources, including the institution’s admission-discharge-transfer system, cost accounting system, electronic health record, clinical data warehouse, and departmental records. The final data-mart contained information for more than 760,000 discharges occurring from 2006 through 2012. Using categories identified by the National Institutes of Health Big Data to Knowledge initiative as a framework, we outlined challenges encountered during the development and use of a domain-specific data-mart and recommend approaches to overcome these challenges.

Infection in home health care: Results from national Outcome and Assessment Information Set data

BACKGROUND Patients in home health care (HHC), the fastest growing health care sector, are at risk for infection. The existing research on infection in HHC is often limited by small sample sizes, local scope of inquiry, and a lack of current data. There is no national study examining agency-level infection rates. METHODS This secondary data analysis used a 20% random sample of the 2010 national Outcome and Assessment Information Set (OASIS) data. An infection case was identified when the HHC patient was hospitalized or received emergency care for respiratory infection, urinary tract infection, intravenous catheter-related infection, wound infection, or deterioration. Proportions of infection cases out of the total number of patients were calculated for the whole sample and for each HHC agency. RESULTS The final analysis included 199,462 patients from 8,255 HHC agencies. Approximately 3.5% of patients developed infections during their HHC stay, leading to emergency care treatment or hospitalization. Seventeen percent of unplanned hospitalizations among HHC patients were caused by infections. The agency-level infection rate ranged from 0%-34%, with an average of 3.5%. CONCLUSION To our knowledge, this is the first study to examine the proportion of hospitalizations or emergency care treatment caused by infection in HHC and the agency-level infection rate at a national level by using OASIS data. These data demonstrate that infection is a serious problem in HHC, and infection rates varied between agencies. The variance in agency level rates may be caused by differences in infection control policies and practices. Better infection surveillance system in HHC is needed to benchmark quality of care.

Carbapenem-Resistant Klebsiella pneumoniae Infection in Three New York City Hospitals Trended Downwards From 2006 to 2014

Background. Carbapenem-resistant Klebsiella pneumoniae (CRKP) infection is a rising public health threat since its first outbreaks in New York City (NYC) in the early 2000s. We investigated annual trends of CRKP infection in hospital-acquired infections (HAIs) and community-onset infections (COIs) treated in 3 NYC hospitals from 2006 to 2014. Methods. We extracted K pneumoniae infection data including carbapenem susceptibility and anatomical sites, compared clinical characteristics between CRKP and carbapenem-susceptible K pneumoniae infections, and determined CRKP infection proportions in total K pneumoniae infections in HAI and COI to identify statistically significant trends from 2006 to 2014 using the Cochran-Armitage trend test. Results. Carbapenem-resistant K pneumoniae contributed 17.3% (601 of 3477) of hospital-acquired K pneumoniae infection compared with 7.7% (149 of 1926) in COI from 2006 to 2014. Carbapenem-resistant K pneumoniae proportions in HAI and COI were positively correlated over time (r = 0.83, P < .01), and there were downward annual trends of CRKP proportions from 2006 to 2014 in both HAI and COI (25.8% to 10.5% in HAI, P < .001; 13.6% to 3.1% in COI, P < .001). By anatomical site, significant downward annual trends were present only in urinary tract infection (P < .001 for both HAI and COI) from 2006 to 2014. Conclusions. Annual trends of CRKP proportions from 2006 to 2014 were downward in both HAI and COI, and HAI and COI were positively correlated. Efforts to reduce and prevent CRKP infections in both hospital and community settings were successful and warrant continuation.

Utility of Electronic Medical Records to Assess the Relationship Between Parenteral Nutrition and Central Line–Associated Bloodstream Infections in Adult Hospitalized Patients

BACKGROUND Parenteral nutrition is associated with increased central line-associated bloodstream infections (CLABSIs). Electronic databases are important for identifying independent risk factors for prevention strategies. Our aims were to evaluate the utility of using electronic data sources to identify risk factors for CLABSIs, including parenteral nutrition (PN), and to assess the association between CLABSI and PN administration. METHODS Data were obtained for all discharges of adult patients in whom a central line was inserted between September 1, 2007, and December 31, 2008, in a large, academically affiliated hospital in New York City. CLABSI was defined electronically using a modified definition from the Centers for Disease Control and Prevention. A manual chart review was also undertaken to assess validity/reliability of the electronic database and gather additional information. Risk factors for CLABSI were examined using logistic regression. RESULTS Among 4840 patients, there were 220 CLABSIs, an incidence of 5.4 CLABSIs per 1000 central line days. Risk factors included PN (odds ratio [OR], 4.33; 95% confidence interval [CI], 2.50-7.48), intensive care unit stay (OR, 2.26; 95% CI, 1.58-3.23), renal disease (OR, 2.79; 95% CI, 2.00-3.88), and immunodeficiency (OR, 2.26; 95% CI, 1.70-3.00). Diabetes mellitus was associated with reduced CLABSI rates (OR, 0.63; 95% CI, 0.45-0.88). CONCLUSIONS The utility of electronic medical records for determining risk factors is limited by such things as free-text data entry. Using a hybrid between fully electronic and manual chart review, reliable data were obtained. PN is associated with a high risk for CLABSI in a population highly selected for indications for PN.

Risk Factors for Bacteremia in Patients With Urinary Catheter–Associated Bacteriuria

Background Catheter‐associated bacteriuria is complicated by secondary bacteremia in 0.4% to 4.0% of cases. The directly attributable mortality rate is 12.7%. Objective To identify risk factors for bacteremia associated with catheter‐associated bacteriuria. Methods Data were acquired from a large electronic clinical and administrative database of consecutive adult inpatient admissions to 2 acute care hospitals during a 7‐year period. Data on patients with catheter‐associated bacteriuria and bacteremia were compared with data on control patients with catheter‐associated bacteriuria and no bacteremia, matched for date of admission plus or minus 30 days. Urine and blood cultures positive for the same pathogen within 7 days were used to define catheter‐associated bacteriuria and bacteremia. Multivariable conditional logistic regression was used to determine independent risk factors for bacteremia. Results The sample consisted of 158 cases and 474 controls. Independent predictors of bacteremia were male sex (odds ratio, 2.76), treatment with immunosuppressants (odds ratio, 1.68), urinary tract procedure (odds ratio, 2.70), and catheter that remained in place after bacteriuria developed (odds ratio, 2.75). Patients with enterococcal bacteriuria were half as likely to become bacteremic as were patients with other urinary pathogens (odds ratio, 0.46). Odds of secondary bacteremia increased 2% per additional day of hospital stay (95% CI, 1.01‐1.04) and decreased 1% with each additional year of age (95% CI, 0.97‐0.99). Conclusions The results add new information about increased risk for bacteremia among patients with catheters remaining in place after catheter‐associated bacteriuria and confirm evidence for previously identified risk factors.

Incidence and risk factors for community-associated methicillin-resistant Staphylococcus aureus in New York City, 2006–2012

This study aims to describe changes in incidence and risk factors for community-associated methicillin resistant Staphylococcus aureus (CA-MRSA) infections upon admission to two New York City hospitals from 2006 to 2012. We examined the first hospitalization for adult patients using electronic health record and administrative data and determined the annual incidence/1000 admissions of total S. aureus, total MRSA, and CA-MRSA (within 48 h of admission) in clinical specimens over the study period. Logistic regression was used to identify factors associated with CA-MRSA in 2006 and 2012. In 137 350 admissions, the incidence of S. aureus, MRSA, and CA-MRSA/1000 admissions were 15·6, 7·0, and 3·5, respectively. The total S. aureus and MRSA isolations decreased significantly over the study period (27% and 25%, respectively) while CA-MRSA incidence was unchanged. CA-MRSA increased as a proportion of all MRSA between 2006 (46%) and 2012 (62%), and was most frequently isolated from respiratory (1·5/1000) and blood (0·7/1000) cultures. Logistic regression analysis of factors associated with isolation of CA-MRSA showed that age ⩾65 years [odds ratio (OR) 2·3, 95% confidence interval (CI) 1·2-4·5], male gender (OR 1·8, 95% CI 1·2-2·8) and history of renal failure (OR 2·6, 95% CI 1·6-4·2) were significant predictors of infection in 2006. No predictors were identified in 2012.

Trends in mortality, length of stay, and hospital charges associated with health care–associated infections, 2006-2012

BACKGROUND Many factors associated with hospital-acquired infections (HAIs), including reimbursement policies, drug prices, practice patterns, and the distribution of organisms causing infections, change over time. We examined whether outcomes, including mortality, length of stay (LOS), daily charges, and total charges associated with HAIs, changed during 2006-2012. METHODS Electronic data on adults discharged from 2 tertiary-quaternary hospitals and 1 community hospital during 2006-2012 were collected retrospectively. Computerized algorithms identified infections using laboratory and administrative codes. Propensity scores were used to match cases with uninfected controls. Differences in mortality, LOS, daily charges, and total charges were modeled against infection status and time period (2006-2008 vs 2009-2012), including interaction for infection status by time period. RESULTS Among 352,077 discharges, 24,466 HAIs were detected. There was no significant change in mortality. LOS declined only for bloodstream infections (3-day reduction; P < 0.01). Daily charges rose 4% for urinary tract infections but did not change significantly for other HAIs. Total charges declined by 11% for bloodstream infections and 13% for pneumonia. CONCLUSIONS We found no appreciable or consistent improvement in HAI mortality or LOS during 2006-2012. Costs of bloodstream infections and pneumonia have declined, with most of the change occurring before 2008.

Temporal trends and factors associated with pediatric hospitalizations with respiratory infection

Abstract Objectives: Asthma is the leading cause of emergency room visits in children, and those with asthma tend to suffer from more severe consequences of respiratory tract infections. The aims of this study were among children who required hospitalization, compare differences between those admitted with or without respiratory infection, describe changes in yearly proportion of asthma among the children admitted with a respiratory infection in New York City from 2006 to 2012 and ascertain how asthma affected the average length of hospital stay for those with respiratory infection. Methods: We conducted an historical prospective study using data obtained from multiple electronic sources of patients discharged from a 200-bed pediatric hospital in a major metropolitan healthcare system between 2006 and 2012. Patient characteristics were obtained from electronic health records and billing codes. Results: Among 83 348 patient discharges, there were 5057 (6.1%) with a respiratory infection. In the multivariable model, children with the following characteristics were significantly more likely to be admitted with a respiratory infection as compared with other diagnoses: asthma [odds ratio (OR) 4.68 (95% confidence limits (CL): 4.31, 5.08], male sex [OR:1.11 (1.05, 1.18)], prior hospitalization [OR: 3.65 (3.43, 3.89)], renal failure [(OR: 2.14 (1.70, 2.70)] and Medicaid coverage [OR: 1.93 (1.81, 2.05)]. Children aged ≥5 years (OR: 0.44 (0.41, 0.47) and those with diabetes (OR: 0.32 (0.21, 0.51) were less likely to be admitted with a respiratory infection. Hospitalized patients with respiratory infection and asthma, identified by ICD-9 codes, had significantly shorter lengths of stay than patients without asthma, 5.3 and 9.0 days, respectively, p < 0.001. Asthma in patients admitted with a respiratory infection increased from 19.2% in 2006 to 28.2% in 2012, peaking at 34.1% in 2010 (p < 0.001). Conclusion: This study suggests that asthma is a significant risk factor for hospitalization of children with respiratory infection, and suggests that the prevalence of asthma is increasing over time among children hospitalized with respiratory infection in this urban setting.

Video-based Education to Reduce Distress and Improve Understanding among Pediatric MRI Patients: A Randomized Controlled Study

Purpose: Evaluate the effectiveness of an educational video vs. standard of care in improving relaxation and procedural understanding among pediatric patients undergoing a magnetic resonance imaging (MRI) procedure. Design and Methods: This pilot randomized controlled trial was conducted in a large, urban academic childrens hospital. Pediatric patients were randomized to receive either a 7‐minute educational video or standard of care. Standardized surveys, which consisted of a 1–10 visual analog scale and open‐ended questions were administered to patients to measure their level of relaxation, understanding of the procedure, and perceptions of the MRI education received. Bivariate statistics were used to compare changes in relaxation score and baseline understanding scores between study groups. Open‐ended questions were analyzed using content analysis. Results: A total of 50 pediatric patients completed the study. Improvements in relaxation scores and baseline procedural understanding scores were significantly higher among children 13–17 years of age who received the intervention compared to those that did not (P < 0.05). No statistically significant differences were noted in relaxation scores and procedural understanding scores among children < 13 years of age between study groups. A total of 26 patients, half from the control group and half from the intervention group responded to open‐ended survey questions. Content analysis revealed that nearly all respondents perceived the educational video to increase their understanding of the MRI procedure. Conclusions: Video‐based education effectively improved the relaxation and procedural understanding of children 13–17 years of age undergoing a MRI. Practice Implications: Nurses may use video‐based education to supplement existing MRI education among older children. HighlightsVideo‐based education improved the MRI procedural understanding of older children.Video‐based education decreased distress among older children undergoing MRI.Nurses may use video‐based education to improve MRI experiences among children.

Prevalence and risk factors for bloodstream infection present on hospital admission

Background: Bloodstream infection present on hospital admission (BSI-POA) is a major cause of morbidity and mortality. The purpose of this study was to measure prevalence and describe the risk factors of patients with BSI-POA and to determine the prevalence of resistance in isolates by admission source. Methods: We conducted a retrospective cohort study of patients discharged from three hospitals in New York City between 2006 and 2014. BSI-POA was defined as BSI diagnosed within 48 h of hospitalisation. Results: The prevalence for BSI-POA was 5307/315,010 discharges (1.7%). The odds of being admitted with BSI-POA were greatest among patients admitted with renal failure, chronic dermatitis, malignancies and prior hospitalisation. Odds ratios and 95% confidence intervals (CI) were 2.72 (95% CI = 2.56–2.88), 2.15 (95% CI = 1.97–2.34), 1.76 (95% CI = 1.64–1.88) and 1.59 (95% CI = 1.50–1.69), respectively. The largest proportion of BSI-POA presented with Staphylococcus aureus (48.4%), followed by Enterococcus faecalis/faecium (20.3%), Klebsiella pneumoniae (16.2%), Streptococcus pneumoniae (8.7%), Pseudomonas aeruginosa (4.2%) and Acinetobacter baumannii (2.2%). Overall, 44% of those admitted from nursing homes presented with antibiotic resistant strains versus 34% from other hospitals and 31% from private homes (P = 0.002). Conclusion: Understanding the risk factors of patients who present to the hospital with BSI could enable timely interventions and better patient outcomes.