INTRODUCTION: Initiation of medications for opioid use disorder (MOUD) within the hospital setting may improve outcomes for people who inject drugs (PWID) hospitalized because of an infection. Many studies used International Classification of Diseases (ICD) codes to identify PWID, although these may be misclassified and thus, inaccurate. We hypothesized that bias from misclassification of PWID using ICD codes may impact analyses of MOUD outcomes. METHODS: We analyzed a cohort of 36 868 cases of patients diagnosed with Staphylococcus aureus bacteremia at 124 US Veterans Health Administration hospitals between 2003 and 2014. To identify PWID, we implemented an ICD code-based algorithm and a natural language processing (NLP) algorithm for classification of admission notes. We analyzed outcomes of prescribing MOUD as an inpatient using both approaches. Our primary outcome was 365-day all-cause mortality. We fit mixed-effects Cox regression models with receipt or not of MOUD during the index hospitalization as the primary predictor and 365-day mortality as the outcome. RESULTS: NLP identified 2389 cases as PWID, whereas ICD codes identified 6804 cases as PWID. In the cohort identified by NLP, receipt of inpatient MOUD was associated with a protective effect on 365-day survival (adjusted hazard ratio, 0.48; 95% confidence interval, .29-.81; P < .01) compared with those not receiving MOUD. There was no significant effect of MOUD receipt in the cohort identified by ICD codes (adjusted hazard ratio, 1.00; 95% confidence interval, .77-1.30; P = .99). CONCLUSIONS: MOUD was protective of all-cause mortality when NLP was used to identify PWID, but not significant when ICD codes were used to identify the analytic subjects.

Background

Antibiotic overuse at hospital discharge is common, but there is no metric to evaluate hospital performance at this transition of care. We built a risk-adjusted metric for comparing hospitals on their overall post-discharge antibiotic use.

Methods

This was a retrospective study across all acute-care admissions within the Veterans Health Administration during 2018-2021. For patients discharged to home, we collected data on antibiotics and relevant covariates. We built a zero-inflated negative binomial mixed-model with two random intercepts for each hospital to predict post-discharge antibiotic exposure and length of therapy (LOT). Data were split into training and testing sets to evaluate model performance using absolute error. Hospital performance was determined by the predicted random intercepts.

Results

1,804,300 patient-admissions across 129 hospitals were included. Antibiotics were prescribed to 41.5% while hospitalized and 19.5% at discharge. Median LOT among those prescribed post-discharge antibiotics was 7 (IQR 4-10). The predictive model detected post-discharge antibiotic use with fidelity, including accurate identification of any exposure (area under the precision-recall curve=0.97) and reliable prediction of post-discharge LOT (mean absolute error = 1.48). Based on this model, 39 (30.2%) hospitals prescribed antibiotics less often than expected at discharge and used shorter LOT than expected. Twenty-eight (21.7%) hospitals prescribed antibiotics more often at discharge and used longer LOT.

Conclusion

A model using electronically-available data was able to predict antibiotic use prescribed at hospital discharge and showed that some hospitals were more successful in reducing antibiotic overuse at this transition of care. This metric may help hospitals identify opportunities for improved antibiotic stewardship at discharge.

Many US hospitals lack Infectious Disease (ID) specialists, which may hinder antibiotic stewardship efforts. We sought to compare patient-level antibiotic exposure at Veterans Health Administration (VHA) hospitals with and without an on-site ID specialist, defined as an ID physician and/or ID pharmacist. This retrospective VHA cohort included all acute-care patient-admissions during 2016. A mandatory survey was used to identify hospitals' antibiotic stewardship processes and their access to an on-site ID specialist. Antibiotic use was quantified as days of therapy (DOT) per days-present and categorized based on National Healthcare Safety Network definitions. A negative binomial regression model with risk adjustment was used to determine the association between presence of an on-site ID specialist and antibiotic use at the level of patient-admissions. Eighteen of 122 (14.8%) hospitals lacked an on-site ID specialist; there were 525,451 (95.8%) admissions at ID hospitals and 23,007 (4.2%) at non-ID sites. In the adjusted analysis, presence of an ID specialist was associated with lower total inpatient antibacterial use [OR 0.92, (95% CI, 0.85-0.99)]. Presence of an ID specialist was also associated with lower use of broad-spectrum antibacterials [OR 0.61 (95% CI, 0.54-0.70)] and higher narrow-spectrum beta-lactam use [OR 1.43 (95% CI, 1.22-1.67)]. Total antibacterial exposure (inpatient plus post-discharge) was lower among patients at ID versus non-ID sites [OR 0.92 (95% CI, 0.86-0.99)]. Patients at hospitals with an ID specialist received antibiotics in a way more consistent with stewardship principles. The presence of an ID specialist may be important to effective antibiotic stewardship.

Background

Improving the identification of people who inject drugs (PWID) in electronic medical records can improve clinical decision making, risk assessment and mitigation, and health service research. Identification of PWID currently consists of heterogeneous, nonspecific International Classification of Diseases (ICD) codes as proxies. Natural language processing (NLP) and machine learning (ML) methods may have better diagnostic metrics than nonspecific ICD codes for identifying PWID.

Methods

We manually reviewed 1000 records of patients diagnosed with Staphylococcus aureus bacteremia admitted to Veterans Health Administration hospitals from 2003 through 2014. The manual review was the reference standard. We developed and trained NLP/ML algorithms with and without regular expression filters for negation (NegEx) and compared these with 11 proxy combinations of ICD codes to identify PWID. Data were split 70% for training and 30% for testing. We calculated diagnostic metrics and estimated 95% confidence intervals (CIs) by bootstrapping the hold-out test set. Best models were determined by best F-score, a summary of sensitivity and positive predictive value.

Results

Random forest with and without NegEx were the best-performing NLP/ML algorithms in the training set. Random forest with NegEx outperformed all ICD-based algorithms. F-score for the best NLP/ML algorithm was 0.905 (95% CI, .786-.967) and 0.592 (95% CI, .550-.632) for the best ICD-based algorithm. The NLP/ML algorithm had a sensitivity of 92.6% and specificity of 95.4%.

Conclusions

NLP/ML outperformed ICD-based coding algorithms at identifying PWID in electronic health records. NLP/ML models should be considered in identifying cohorts of PWID to improve clinical decision making, health services research, and administrative surveillance.

Background

Many US hospitals lack infectious disease (ID) specialists, which may hinder antibiotic stewardship efforts. We sought to compare patient-level antibiotic exposure at Veterans Health Administration (VHA) hospitals with and without an on-site ID specialist, defined as an ID physician and/or ID pharmacist.

Methods

This retrospective VHA cohort included all acute-care patient admissions during 2016. A mandatory survey was used to identify hospitals' antibiotic stewardship processes and their access to an on-site ID specialist. Antibiotic use was quantified as days of therapy per days present and categorized based on National Healthcare Safety Network definitions. A negative binomial regression model with risk adjustment was used to determine the association between presence of an on-site ID specialist and antibiotic use at the level of patient admissions.

Results

Eighteen of 122 (14.8%) hospitals lacked an on-site ID specialist; there were 525 451 (95.8%) admissions at ID hospitals and 23 007 (4.2%) at non-ID sites. In the adjusted analysis, presence of an ID specialist was associated with lower total inpatient antibacterial use (odds ratio, 0.92; 95% confidence interval, .85-.99). Presence of an ID specialist was also associated with lower use of broad-spectrum antibacterials (0.61; .54-.70) and higher narrow-spectrum β-lactam use (1.43; 1.22-1.67). Total antibacterial exposure (inpatient plus postdischarge) was lower among patients at ID versus non-ID sites (0.92; .86-.99).

Conclusions

Patients at hospitals with an ID specialist received antibiotics in a way more consistent with stewardship principles. The presence of an ID specialist may be important to effective antibiotic stewardship.

Objective

In the absence of pyuria, positive urine cultures are unlikely to represent infection. Conditional urine reflex culture policies have the potential to limit unnecessary urine culturing. We evaluated the impact of this diagnostic stewardship intervention.

Design

We conducted a retrospective, quasi-experimental (nonrandomized) study, with interrupted time series, from August 2013 to January 2018 to examine rates of urine cultures before versus after the policy intervention. We compared 3 intervention sites to 3 control sites in an aggregated series using segmented negative binomial regression.

Setting

The study included 6 acute-care hospitals within the Veterans' Health Administration across the United States.

Participants

Adult patients with at least 1 urinalysis ordered during acute-care admission, excluding pregnant patients or those undergoing urological procedures, were included.

Methods

At the intervention sites, urine cultures were performed if a preceding urinalysis met prespecified criteria. No such restrictions occurred at the control sites. The primary outcome was the rate of urine cultures performed per 1,000 patient days. The safety outcome was the rate of gram-negative bloodstream infection per 1,000 patient days.

Results

The study included 224,573 urine cultures from 50,901 admissions in 24,759 unique patients. Among the intervention sites, the overall average number of urine cultures performed did not significantly decrease relative to the preintervention period (5.9% decrease; P = 0.8) but did decrease by 21% relative to control sites (P < .01). We detected no significant difference in the rates of gram-negative bloodstream infection among intervention or control sites (P = .49).

Conclusions

Conditional urine reflex culture policies were associated with a decrease in urine culturing without a change in the incidence of gram-negative bloodstream infection.

Antimicrobial stewardship is vital to reducing the spread of antimicrobial resistance. A group of investigators and clinicians within the Veterans Health Administration set forth a research agenda for antimicrobial stewardship, including research targets for inpatient and outpatient stewardship activities, metrics, and antimicrobial dosing and duration.

To investigate factors that influence antibiotic prescribing decisions, we interviewed 49 antibiotic stewardship champions and stakeholders across 15 hospitals. We conducted thematic analysis and subcoding of decisional factors. We identified 31 factors that influence antibiotic prescribing decisions. These factors may help stewardship programs identify educational targets and design more effective interventions.

Introduction

The optimal method for implementing hospital-level restrictions for antibiotics that carry a high risk of Clostridioides difficile infection has not been identified. We aimed to explore barriers and facilitators to implementing restrictions for fluoroquinolones and third/fourth-generation cephalosporins.

Methods

This mixed-methods study across a purposeful sample of 15 acute-care, geographically dispersed Veterans Health Administration hospitals included electronic surveys and semi-structured interviews (September 2018 to May 2019). Surveys on stewardship strategies were administered at each hospital and summarized with descriptive statistics. Interviews were performed with 30 antibiotic stewardship programme (ASP) champions across all 15 sites and 19 additional stakeholders at a subset of 5 sites; transcripts were analysed using thematic content analysis.

Results

The most restricted agent was moxifloxacin, which was restricted at 12 (80%) sites. None of the 15 hospitals restricted ceftriaxone. Interviews identified differing opinions on the feasibility of restricting third/fourth-generation cephalosporins and fluoroquinolones. Some participants felt that restrictions could be implemented in a way that was not burdensome to clinicians and did not interfere with timely antibiotic administration. Others expressed concerns about restricting these agents, particularly through prior approval, given their frequent use, the difficulty of enforcing restrictions and potential unintended consequences of steering clinicians towards non-restricted antibiotics. A variety of stewardship strategies were perceived to be effective at reducing the use of these agents.

Conclusions

Across 15 hospitals, there were differing opinions on the feasibility of implementing antibiotic restrictions for third/fourth-generation cephalosporins and fluoroquinolones. While the perceived barrier to implementing restrictions was frequently high, many hospitals were effectively using restrictions and reported few barriers to their use.