The Impact of Multi-Institution Datasets on the Generalizability of Machine Learning Prediction Models in the ICU.

OBJECTIVES: To evaluate the transferability of deep learning (DL) models for the early detection of adverse events to previously unseen hospitals. DESIGN: Retrospective observational cohort study utilizing harmonized intensive care data from four public datasets. SETTING: ICUs across Europe and the United States. PATIENTS: Adult patients admitted to the ICU for at least 6 hours who had good data quality. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Using carefully harmonized data from a total of 334,812 ICU stays, we systematically assessed the transferability of DL models for three common adverse events: death, acute kidney injury (AKI), and sepsis. We tested whether using more than one data source and/or algorithmically optimizing for generalizability during training improves model performance at new hospitals. We found that models achieved high area under the receiver operating characteristic (AUROC) for mortality (0.838-0.869), AKI (0.823-0.866), and sepsis (0.749-0.824) at the training hospital. As expected, AUROC dropped when models were applied at other hospitals, sometimes by as much as -0.200. Using more than one dataset for training mitigated the performance drop, with multicenter models performing roughly on par with the best single-center model. Dedicated methods promoting generalizability did not noticeably improve performance in our experiments. CONCLUSIONS: Our results emphasize the importance of diverse training data for DL-based risk prediction. They suggest that as data from more hospitals become available for training, models may become increasingly generalizable. Even so, good performance at a new hospital still depended on the inclusion of compatible hospitals during training.

Assessing the impact of discordant antibiotic treatment on adverse outcomes in community-onset UTI: a retrospective cohort study.

OBJECTIVES: To investigate the risk of adverse outcomes following discordant antibiotic treatment (urinary organism resistant) for culture-confirmed community-onset lower urinary tract infection (UTI). METHODS: Cohort study using routinely collected linked primary care, secondary care and microbiology data from patients with culture-confirmed community-onset lower UTI (COLUTI). Antibiotic treatment within ±3 days was considered concordant if the urinary organism was sensitive and discordant if resistant.The primary outcome was the proportion of patients experiencing urinary infection-related hospital admission (UHA) within 30 days. Secondary outcomes were the proportion of patients experiencing reconsultation within 30 days, and the odds of UHA and reconsultation following discordant treatment, adjusting for sex, age, risk factors for complicated UTI, previous antibiotic treatment, recurrent UTI and comorbidities. RESULTS: A total of 11 963 UTI episodes in 8324 patients were included, and 1686 episodes (14.1%, 95% CI 13.5%-14.7%) were discordant. UHA occurred in 212/10 277 concordant episodes (2.1%, 95% CI 1.8%-2.4%) and 88/1686 discordant episodes (5.2%, 95% CI 4.2%-6.4%). Reconsultation occurred in 3961 concordant (38.5%, 95% CI 37.6%-39.5%) and 1472 discordant episodes (87.3%, 95% CI 85.6%-88.8%). Discordant treatment compared with concordant was associated with increased odds of UHA (adjusted OR 2.31, 95% CI 1.77-3.0, P < 0.001) and reconsultation (adjusted OR 11.25, 95% CI 9.66-13.11, P < 0.001) on multivariable analysis. Chronic kidney disease and diabetes mellitus were also independently associated with increased odds of UHA. CONCLUSIONS: One in seven COLUTI episodes in primary care were treated with discordant antibiotics. In higher risk patients requiring urine culture, empirical antibiotic choice optimization could meaningfully reduce adverse outcomes.

Effect of SGLT2 Inhibitors on Discontinuation of Renin-angiotensin System Blockade: A Joint Analysis of the CREDENCE and DAPA-CKD Trials.

SIGNIFICANCE STATEMENT: Angiotensin-converting enzyme (ACE) inhibitors and angiotensin receptor blockers (ARBs) are foundational therapy for CKD but are underused, in part because they are frequently withheld and not restarted due to hyperkalemia, AKI, or hospitalization. Consequently, ensuring persistent use of ACE inhibitors and ARBs in CKD has long been a major clinical priority. In this joint analysis of the CREDENCE and DAPA-CKD trials, the relative risk of discontinuation of ACE inhibitors and ARBs was reduced by 15% in patients randomized to sodium-glucose cotransporter 2 (SGLT2) inhibitors. This effect was more pronounced in patients with urine albumin:creatinine ratio ≥1000 mg/g, for whom the absolute benefits of these medications are the greatest. These findings indicate that SGLT2 inhibitors may enable better use of ACE inhibitors and ARBs in patients with CKD. BACKGROUND: Strategies to enable persistent use of renin-angiotensin system (RAS) blockade to improve outcomes in CKD have long been sought. The effect of SGLT2 inhibitors on discontinuation of RAS blockade has yet to be evaluated. METHODS: We conducted a joint analysis of canagliflozin and renal events in diabetes with established nephropathy clinical evaluation (CREDENCE) and dapagliflozin and prevention of adverse outcomes in CKD (DAPA-CKD), two randomized, double-blind, placebo-controlled, event-driven trials of SGLT2 inhibitors in patients with albuminuric CKD. The main outcome was time to incident temporary or permanent discontinuation of RAS blockade, defined as interruption of an ACE inhibitor or ARB for at least 4 weeks or complete cessation during the double-blind on-treatment period. Cox regression analyses were used to estimate the treatment effects from each trial. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were pooled with fixed effects meta-analysis to obtain summary treatment effects, overall and across key subgroups. RESULTS: During median follow-up of 2.2 years across both trials, 740 of 8483 (8.7%) patients discontinued RAS blockade. The relative risk for discontinuation of RAS blockade was 15% lower in patients randomized to receiving SGLT2 inhibitors (HR, 0.85; 95% CI, 0.74 to 0.99), with consistent effects across trials ( P -heterogeneity = 0.92). The relative effect on RAS blockade discontinuation was more pronounced among patients with baseline urinary albumin:creatinine ratio ≥1000 mg/g (pooled HR, 0.77; 95% CI, 0.63 to 0.94; P -heterogeneity = 0.009). CONCLUSIONS: In patients with albuminuric CKD with and without type 2 diabetes, SGLT2 inhibitors facilitate the use of RAS blockade. CLINICAL TRIAL REGISTRY NAME AND REGISTRATION NUMBER: ClinicalTrials.gov, NCT02065791 and NCT03036150 . PODCAST: This article contains a podcast at https://dts.podtrac.com/redirect.mp3/www.asn-online.org/media/podcast/JASN/2023_11_21_JASN0000000000000248.mp3.

Predicting success of intrauterine insemination using a clinically based scoring system.

PURPOSE: To develop a predictive score for the success of intrauterine insemination (IUI) based on clinical parameters. METHODS: We performed a retrospective cohort study evaluating the homologous IUI cycles performed at a single university-based reproductive medical center between 2009 and 2017. The primary outcome measure was pregnancy, defined as positive serum human chorionic gonadotropin (hCG) 12-14 days after IUI. Predictive factors for pregnancy after IUI were identified, and a predictive score was developed using a multivariable continuation ratio model. RESULTS: Overall, 1437 IUI cycles in 758 couples were evaluated. We found a per cycle pregnancy rate of 10.9% and a cumulative pregnancy rate of 19.4%. In a multivariable analysis, the probability of pregnancy was negatively associated with female age ≥ 35 years (OR 0.63, 95% CI 0.41-0.97, p = 0.034), endometriosis, unilateral tubal factor, or anatomical alteration (OR 0.54, 95% CI 0.33-0.89, p = 0.016), anti-Mullerian hormone (AMH) < 1 ng/ml (OR 0.50, 95% CI 0.29-0.87, p = 0.014), and total progressive motile sperm count (TPMSC) < 5 mil (OR 0.47, 95% CI 0.19-0.72, p = 0.004). We developed a predictive clinical score ranging from 0 to 5. Following 3 cycles, couples in our cohort with a score of 5 had a cumulative probability of achieving pregnancy of nearly 45%. In contrast, couples with a score of 0 had a cumulative probability of only 5%. CONCLUSION: IUI success rates vary widely depending on couples' characteristics. A simple to use score could be used to estimate a couple's chance of achieving pregnancy via IUI, facilitating individualized counseling and decision-making.

Antibiotic Prescribing Before and After the Diagnosis of Comorbidity: A Cohort Study Using Primary Care Electronic Health Records.

BACKGROUND: Comorbidities such as diabetes and chronic obstructive pulmonary disease (COPD) increase patients' susceptibility to infections, but it is unclear how the onset of comorbidity impacts antibiotic use. We estimated rates of antibiotic use before and after diagnosis of comorbidity in primary care to identify opportunities for antibiotic stewardship. METHODS: We analyzed UK primary care records from the Clinical Practice Research Datalink. Adults registered between 2008-2015 without prior comorbidity diagnoses were eligible for inclusion. Monthly adjusted rates of antibiotic prescribing were estimated for patients with new-onset stroke, coronary heart disease, heart failure, peripheral arterial disease, asthma, chronic kidney disease, diabetes, or COPD in the 12 months before and after diagnosis and for controls without comorbidity. RESULTS: 106 540/1 071 943 (9.9%) eligible patients were diagnosed with comorbidity. Antibiotic prescribing rates increased 1.9- to 2.3-fold in the 4-9 months preceding diagnosis of asthma, heart failure, and COPD before declining to stable levels within 2 months after diagnosis. A less marked trend was seen for diabetes (rate ratio, 1.55; 95% confidence interval, 1.48-1.61). Prescribing rates for patients with vascular conditions increased immediately before diagnosis and remained 30%-39% higher than baseline afterwards. Rates of prescribing to controls increased by 17%-28% in the months just before and after consultation. CONCLUSIONS: Antibiotic prescribing increased rapidly before diagnosis of conditions that present with respiratory symptoms (COPD, heart failure, asthma) and declined afterward. Onset of respiratory symptoms may be misdiagnosed as infection. Earlier diagnosis of these comorbidities could reduce avoidable antibiotic prescribing.

Antibiotic prescribing for lower UTI in elderly patients in primary care and risk of bloodstream infection: A cohort study using electronic health records in England.

BACKGROUND: Research has questioned the safety of delaying or withholding antibiotics for suspected urinary tract infection (UTI) in older patients. We evaluated the association between antibiotic treatment for lower UTI and risk of bloodstream infection (BSI) in adults aged ≥65 years in primary care. METHODS AND FINDINGS: We analyzed primary care records from patients aged ≥65 years in England with community-onset UTI using the Clinical Practice Research Datalink (2007-2015) linked to Hospital Episode Statistics and census data. The primary outcome was BSI within 60 days, comparing patients treated immediately with antibiotics and those not treated immediately. Crude and adjusted associations between exposure and outcome were estimated using generalized estimating equations. A total of 147,334 patients were included representing 280,462 episodes of lower UTI. BSI occurred in 0.4% (1,025/244,963) of UTI episodes with immediate antibiotics versus 0.6% (228/35,499) of episodes without immediate antibiotics. After adjusting for patient demographics, year of consultation, comorbidities, smoking status, recent hospitalizations, recent accident and emergency (A&E) attendances, recent antibiotic prescribing, and home visits, the odds of BSI were equivalent in patients who were not treated with antibiotics immediately and those who were treated on the date of their UTI consultation (adjusted odds ratio [aOR] 1.13, 95% CI 0.97-1.32, p-value = 0.105). Delaying or withholding antibiotics was associated with increased odds of death in the subsequent 60 days (aOR 1.17, 95% CI 1.09-1.26, p-value < 0.001), but there was limited evidence that increased deaths were attributable to urinary-source BSI. Limitations include overlap between the categories of immediate and delayed antibiotic prescribing, residual confounding underlying differences between patients who were/were not treated with antibiotics, and lack of microbiological diagnosis for BSI. CONCLUSIONS: In this study, we observed that delaying or withholding antibiotics in older adults with suspected UTI did not increase patients' risk of BSI, in contrast with a previous study that analyzed the same dataset, but mortality was increased. Our findings highlight uncertainty around the risks of delaying or withholding antibiotic treatment, which is exacerbated by systematic differences between patients who were and were not treated immediately with antibiotics. Overall, our findings emphasize the need for improved diagnostic/risk prediction strategies to guide antibiotic prescribing for suspected UTI in older adults.

Opportunities to reduce antibiotic prescribing for patients with COPD in primary care: a cohort study using electronic health records from the Clinical Practice Research Datalink (CPRD).

BACKGROUND: In primary care there is uncertainty about which patients with acute exacerbations of COPD (AECOPD) benefit from antibiotics. OBJECTIVES: To identify which types of COPD patients get the most antibiotics in primary care to support targeted antibiotic stewardship. METHODS: Observational study of COPD patients using a large English primary care database with 12 month follow-up. We estimated the incidence of and risk factors for antibiotic prescribing relative to the number of AECOPD during follow-up, considering COPD severity, smoking, obesity and comorbidity. RESULTS: From 157 practices, 19594 patients were diagnosed with COPD, representing 2.6% of patients and 11.5% of all prescribed antibiotics. Eight hundred and thirty-three (4.5%) patients with severe COPD and frequent AECOPD were prescribed six to nine prescriptions per year and accounted for 13.0% of antibiotics. Individuals with mild to moderate COPD and zero or one AECOPD received one to three prescriptions per year but accounted for 42.5% of all prescriptions. In addition to COPD severity, asthma, chronic heart disease, diabetes, heart failure and influenza vaccination were independently associated with increased antibiotic use. CONCLUSIONS: Patients with severe COPD have the highest rates of antibiotic prescribing but most antibiotics are prescribed for patients with mild to moderate COPD. Antibiotic stewardship should focus on the dual goals of safely reducing the volume of prescribing in patients with mild to moderate COPD, and optimizing prescribing in patients with severe disease who are at significant risk of drug resistance.

Diagnostic uncertainty and urinary tract infection in the emergency department: a cohort study from a UK hospital.

BACKGROUND: Suspected urinary tract infection (UTI) syndromes are a common reason for empirical antibiotics to be prescribed in the Emergency Department (ED), but differentiating UTI from other conditions with a similar presentation is challenging. We investigated how often an ED diagnosis of UTI is confirmed clinically/microbiologically, and described conditions which present as UTI syndromes. METHODS: Observational study using electronic health records from patients who attended the ED with suspected UTI and had a urine sample submitted for culture. We compared the ED diagnosis to diagnosis at discharge from hospital (ICD-10 codes), and estimated the proportion of cases with clinical/microbiological evidence of UTI. RESULTS: Two hundred eighty nine patients had an ED diagnosis of UTI syndrome comprising: lower UTI (191), pyelonephritis (56) and urosepsis (42). In patients admitted to hospital with an ED diagnosis of lower UTI, pyelonephritis or urosepsis, clinical/microbiological evidence of UTI was lacking in 61/103, 33/54 and 31/42 cases respectively. The ED diagnosis was concordant with the main reason for admission in less than 40% of patients with UTI syndromes, and antibiotics were stopped within 72 h in 37/161 patients. CONCLUSIONS: Clinical/microbiological evidence of UTI was lacking in 60-70% of patients, suggesting scope to revise empirical prescribing decisions for UTI syndromes in light of microbial culture and clinical progression.

Linking individual-level data on diagnoses and dispensing for research on antibiotic use: Evaluation of a novel data source from English secondary care.

PURPOSE: There has been a focus on stewardship programmes to curb inappropriate antibiotic prescribing and reduce antimicrobial resistance. In-hospital, patient-level prescribing linked to indication is needed to support surveillance, evaluation of stewardship initiatives, as well as other antibiotic research. We evaluated whether a novel dataset linking hospital pharmacy records to Hospital Episode Statistics data can be used for antibiotic research. METHODS: Using the Hospital Treatment Insights (HTI) database, which links Hospital Episode Statistics to pharmacy records from 43 out of 153 hospital trusts in England, we estimated the proportion of missed linkage and identified characteristics associated with missing data. RESULTS: Linkage of antibiotics to patients was inconsistent and dependent on drug type and clinical setting, so that linkage for some specific antibiotics was high (80-100%), but overall, only 27.6% (CI: 27.4%-27.8%) for all antibiotics dispensed. Linkage was best for quinolones (62.6%; CI: 61.8%-63.8%), but only 21.1% (CI: 21.1%-21.2%) for penicillins. Linkage was lower for common antibiotics and in emergency departments; however, 80% linkage was achieved for individual drugs like clindamycin, especially on wards with reduced ward stock use. CONCLUSIONS: For those antibiotics with high linkage, HTI might be used to study associations between indication, dispensing, and outcomes. However, the majority of common antibiotics had insufficient linkage, likely due to extensive use of ward stocks. Therefore, HTI in its current form is not suitable for general antibiotic surveillance or evaluation of stewardship initiatives. For drugs in HTI other than antibiotics, linkage should be similarly evaluated before a study is conducted.

Can the application of machine learning to electronic health records guide antibiotic prescribing decisions for suspected urinary tract infection in the Emergency Department?

Urinary tract infections (UTIs) are a major cause of emergency hospital admissions, but it remains challenging to diagnose them reliably. Application of machine learning (ML) to routine patient data could support clinical decision-making. We developed a ML model predicting bacteriuria in the ED and evaluated its performance in key patient groups to determine scope for its future use to improve UTI diagnosis and thus guide antibiotic prescribing decisions in clinical practice. We used retrospective electronic health records from a large UK hospital (2011-2019). Non-pregnant adults who attended the ED and had a urine sample cultured were eligible for inclusion. The primary outcome was predominant bacterial growth ≥104 cfu/mL in urine. Predictors included demography, medical history, ED diagnoses, blood tests, and urine flow cytometry. Linear and tree-based models were trained via repeated cross-validation, re-calibrated, and validated on data from 2018/19. Changes in performance were investigated by age, sex, ethnicity, and suspected ED diagnosis, and compared to clinical judgement. Among 12,680 included samples, 4,677 (36.9%) showed bacterial growth. Relying primarily on flow cytometry parameters, our best model achieved an area under the ROC curve (AUC) of 0.813 (95% CI 0.792-0.834) in the test data, and achieved both higher sensitivity and specificity compared to proxies of clinician's judgement. Performance remained stable for white and non-white patients but was lower during a period of laboratory procedure change in 2015, in patients ≥65 years (AUC 0.783, 95% CI 0.752-0.815), and in men (AUC 0.758, 95% CI 0.717-0.798). Performance was also slightly reduced in patients with recorded suspicion of UTI (AUC 0.797, 95% CI 0.765-0.828). Our results suggest scope for use of ML to inform antibiotic prescribing decisions by improving diagnosis of suspected UTI in the ED, but performance varied with patient characteristics. Clinical utility of predictive models for UTI is therefore likely to differ for important patient subgroups including women <65 years, women ≥65 years, and men. Tailored models and decision thresholds may be required that account for differences in achievable performance, background incidence, and risks of infectious complications in these groups.

Hospital admission after primary care consultation for community-onset lower urinary tract infection: a cohort study of risks and predictors using linked data.

BACKGROUND: Urinary tract infections (UTIs) are a common indication for antibiotic prescriptions, reductions in which would reduce antimicrobial resistance (AMR). Risk stratification of patients allows reductions to be made safely. AIM: To identify risk factors for hospital admission following UTI, to inform targeted antibiotic stewardship. DESIGN AND SETTING: Retrospective cohort study of East London primary care patients. METHOD: Hospital admission outcomes following primary care consultation for UTI were analysed using linked data from primary care, secondary care, and microbiology, from 1 April 2012 to 31 March 2017. The outcomes analysed were urinary infection-related hospital admission (UHA) and all-cause hospital admission (AHA) within 30 days of UTI in primary care. Odds ratios between specific variables (demographic characteristics, prior antibiotic exposure, and comorbidities) and the outcomes were predicted using generalised estimating equations, and fitted to a final multivariable model including all variables with a P-value <0.1 on univariable analysis. RESULTS: Of the 169 524 episodes of UTI, UHA occurred in 1336 cases (0.8%, 95% confidence interval [CI] = 0.7 to 0.8) and AHA in 6516 cases (3.8%, 95% CI = 3.8 to 3.9). On multivariable analysis, increased odds of UHA were seen in patients aged 55-74 years (adjusted odds ratio [AOR] 1.49) and ≥75 years (AOR 3.24), relative to adults aged 16-34 years. Increased odds of UHA were also associated with chronic kidney disease (CKD; AOR 1.55), urinary catheters (AOR 2.01), prior antibiotics (AOR 1.38 for ≥3 courses), recurrent UTI (AOR 1.33), faecal incontinence (FI; AOR 1.47), and diabetes mellitus (DM; AOR 1.37). CONCLUSION: Urinary infection-related hospital admission after primary care consultation for community-onset lower UTI was rare; however, increased odds for UHA were observed for some patient groups. Efforts to reduce antibiotic prescribing for suspected UTI should focus on patients aged <55 years without risk factors for complicated UTI, recurrent UTI, DM, or FI.

Canagliflozin, Blood Pressure Variability, and Risk of Cardiovascular, Kidney, and Mortality Outcomes: Pooled Individual Participant Data From the CANVAS and CREDENCE Trials.

Background Sodium glucose cotransporter-2 inhibitors reduce systolic blood pressure (SBP), but whether they affect SBP variability is unknown. There also remains uncertainty regarding the prognostic value of SBP variability for different clinical outcomes. Methods and Results Using individual participant data from the CANVAS (Canagliflozin Cardiovascular Assessment Study) Program and CREDENCE (Canagliflozin and Renal Events in Diabetes with Established Nephropathy Clinical Evaluation) trial, we assessed the effect of canagliflozin on SBP variability in people with type 2 diabetes across 4 study visits over 1.5 years as measured by standard deviation, coefficient of variation, and variability independent of the mean. We used multivariable Cox regression models to estimate associations of SBP variability with cardiovascular, kidney, and mortality outcomes. In 11 551 trial participants, canagliflozin modestly lowered the standard deviation of SBP variability (-0.25 mm Hg [95% CI, -0.44 to -0.06]), but there was no effect on coefficient of variation (0.02% [95% CI, -0.12 to 0.16]) or variability independent of the mean (0.08 U [95% CI, -0.11 to 0.26]) when adjusting for correlation with mean SBP. Each 1 standard deviation increase in standard deviation of SBP variability was independently associated with higher risk of hospitalization for heart failure (hazard ratio [HR], 1.19 [95% CI, 1.02-1.38]) and all-cause mortality (HR, 1.12 [95% CI, 1.01-1.25]), with consistent results observed for coefficient of variation and variability independent of the mean. Increases in SBP variability were not associated with kidney outcomes. Conclusions In people with type 2 diabetes at high cardiovascular risk or with chronic kidney disease, higher visit-to-visit SBP variability is independently associated with risks of hospitalization for heart failure and all-cause mortality. Canagliflozin has little to no effect on SBP variability, independent of its established SBP-lowering effect. Registration URL: https://www.clinicaltrials.gov; Unique identifiers: NCT01032629, NCT01989754, NCT02065791.

Reduced oxygen saturation entropy is associated with poor prognosis in critically ill patients with sepsis.

Recent studies have found that oxygen saturation (SpO2 ) variability analysis has potential for noninvasive assessment of the functional connectivity of cardiorespiratory control systems during hypoxia. Patients with sepsis have suboptimal tissue oxygenation and impaired organ system connectivity. Our objective with this report was to highlight the potential use for SpO2 variability analysis in predicting intensive care survival in patients with sepsis. MIMIC-III clinical data of 164 adults meeting Sepsis-3 criteria and with 30 min of SpO2 and respiratory rate data were analyzed. The complexity of SpO2 signals was measured through various entropy calculations such as sample entropy and multiscale entropy analysis. The sequential organ failure assessment (SOFA) score was calculated to assess the severity of sepsis and multiorgan failure. While the standard deviation of SpO2 signals was comparable in the non-survivor and survivor groups, non-survivors had significantly lower SpO2 entropy than those who survived their ICU stay (0.107 ± 0.084 vs. 0.070 ± 0.083, p < 0.05). According to Cox regression analysis, higher SpO2 entropy was a predictor of survival in patients with sepsis. Multivariate analysis also showed that the prognostic value of SpO2 entropy was independent of other covariates such as age, SOFA score, mean SpO2 , and ventilation status. When SpO2 entropy was combined with mean SpO2 , the composite had a significantly higher performance in prediction of survival. Analysis of SpO2 entropy can predict patient outcome, and when combined with SpO2 mean, can provide improved prognostic information. The prognostic power is on par with the SOFA score. This analysis can easily be incorporated into current ICU practice and has potential for noninvasive assessment of critically ill patients.

Longitudinal trends of peripheral blood counts in polycythaemia vera and essential thrombocythemia patients in the UK.

There is sparse evidence of how well haematological targets are met in practice for essential thrombocythemia (ET) and polycythaemia vera (PV) patients. Patient data was collected between 2008 and 2020 from two UK NHS Trusts for ET and PV patients. Longitudinal changes in peripheral blood counts, including the proportion of patients meeting peripheral blood count remission, was modelled. Relative risk of cardiovascular-related events for patients achieving remission within 3-months was estimated. A total of 620 ET and 429 PV patients were analysed. For high-risk patients, haematological parameters decreased in the first months of observation then stabilised within normal reference ranges until year 5. Total time spent in peripheral blood count remission was 39.2% for ET and 29.1% for PV. A lower proportion of ET patients reached target platelet counts (48.3%) compared to WBC (79.1%), whilst PV patients were less likely to reach target haematocrit levels (56.9%) compared to platelets (77.3%) or WBC (74.6%). There was no statistically significant association between reaching target blood counts within 3-months and cardiovascular risk. Complete haematological remission remains a challenging target in managing PV and ET, however this study was unable to show statistically-significant evidence that this was associated with increased risk of cardiovascular events.

Feasibility study of hospital antimicrobial stewardship analytics using electronic health records.

BACKGROUND: Hospital antimicrobial stewardship (AMS) programmes are multidisciplinary initiatives to optimize antimicrobial use. Most hospitals depend on time-consuming manual audits to monitor clinicians' prescribing. But much of the information needed could be sourced from electronic health records (EHRs). OBJECTIVES: To develop an informatics methodology to analyse characteristics of hospital AMS practice using routine electronic prescribing and laboratory records. METHODS: Feasibility study using electronic prescribing, laboratory and clinical coding records from adult patients admitted to six specialities at Queen Elizabeth Hospital, Birmingham, UK (September 2017-August 2018). The study involved: (i) a review of AMS standards of care; (ii) their translation into concepts measurable from commonly available EHRs; and (iii) a pilot application in an EHR cohort study (n = 61679 admissions). RESULTS: We developed data modelling methods to characterize antimicrobial use (antimicrobial therapy episode linkage methods, therapy table, therapy changes). Prescriptions were linked into antimicrobial therapy episodes (mean 2.4 prescriptions/episode; mean length of therapy 5.8 days), enabling several actionable findings. For example, 22% of therapy episodes for low-severity community-acquired pneumonia were congruent with prescribing guidelines, with a tendency to use broader-spectrum antibiotics. Analysis of therapy changes revealed IV to oral therapy switching was delayed by an average 3.6 days (95% CI: 3.4-3.7). Microbial cultures were performed prior to treatment initiation in just 22% of antibacterial prescriptions. The proposed methods enabled fine-grained monitoring of AMS practice down to specialities, wards and individual clinical teams by case mix, enabling more meaningful peer comparison. CONCLUSIONS: It is feasible to use hospital EHRs to construct rapid, meaningful measures of prescribing quality with potential to support quality improvement interventions (audit/feedback to prescribers), engagement with front-line clinicians on optimizing prescribing, and AMS impact evaluation studies.

Development of risk prediction models to predict urine culture growth for adults with suspected urinary tract infection in the emergency department: protocol for an electronic health record study from a single UK university hospital.

BACKGROUND: Urinary tract infection (UTI) is a leading cause of hospital admissions and is diagnosed based on urinary symptoms and microbiological cultures. Due to lags in the availability of culture results of up to 72 h, and the limitations of routine diagnostics, many patients with suspected UTI are started on antibiotic treatment unnecessarily. Predictive models based on routinely collected clinical information may help clinicians to rule out a diagnosis of bacterial UTI in low-risk patients shortly after hospital admission, providing additional evidence to guide antibiotic treatment decisions. METHODS: Using electronic hospital records from Queen Elizabeth Hospital Birmingham (QEHB) collected between 2011 and 2017, we aim to develop a series of models that estimate the probability of bacterial UTI at presentation in the emergency department (ED) among individuals with suspected UTI syndromes. Predictions will be made during ED attendance and at different time points after hospital admission to assess whether predictive performance may be improved over time as more information becomes available about patient status. All models will be externally validated for expected future performance using QEHB data from 2018/2019. DISCUSSION: Risk prediction models using electronic health records offer a new approach to improve antibiotic prescribing decisions, integrating clinical and demographic data with test results to stratify patients according to their probability of bacterial infection. Used in conjunction with expert opinion, they may help clinicians to identify patients that benefit the most from early antibiotic cessation.

Data-driven discovery of changes in clinical code usage over time: a case-study on changes in cardiovascular disease recording in two English electronic health records databases (2001-2015).

OBJECTIVES: To demonstrate how data-driven variability methods can be used to identify changes in disease recording in two English electronic health records databases between 2001 and 2015. DESIGN: Repeated cross-sectional analysis that applied data-driven temporal variability methods to assess month-by-month changes in routinely collected medical data. A measure of difference between months was calculated based on joint distributions of age, gender, socioeconomic status and recorded cardiovascular diseases. Distances between months were used to identify temporal trends in data recording. SETTING: 400 English primary care practices from the Clinical Practice Research Datalink (CPRD GOLD) and 451 hospital providers from the Hospital Episode Statistics (HES). MAIN OUTCOMES: The proportion of patients (CPRD GOLD) and hospital admissions (HES) with a recorded cardiovascular disease (CPRD GOLD: coronary heart disease, heart failure, peripheral arterial disease, stroke; HES: International Classification of Disease codes I20-I69/G45). RESULTS: Both databases showed gradual changes in cardiovascular disease recording between 2001 and 2008. The recorded prevalence of included cardiovascular diseases in CPRD GOLD increased by 47%-62%, which partially reversed after 2008. For hospital records in HES, there was a relative decrease in angina pectoris (-34.4%) and unspecified stroke (-42.3%) over the same time period, with a concomitant increase in chronic coronary heart disease (+14.3%). Multiple abrupt changes in the use of myocardial infarction codes in hospital were found in March/April 2010, 2012 and 2014, possibly linked to updates of clinical coding guidelines. CONCLUSIONS: Identified temporal variability could be related to potentially non-medical causes such as updated coding guidelines. These artificial changes may introduce temporal correlation among diagnoses inferred from routine data, violating the assumptions of frequently used statistical methods. Temporal variability measures provide an objective and robust technique to identify, and subsequently account for, those changes in electronic health records studies without any prior knowledge of the data collection process.

An interdisciplinary mixed-methods approach to developing antimicrobial stewardship interventions: Protocol for the Preserving Antibiotics through Safe Stewardship (PASS) Research Programme.

Behaviour change is key to combating antimicrobial resistance. Antimicrobial stewardship (AMS) programmes promote and monitor judicious antibiotic use, but there is little consideration of behavioural and social influences when designing interventions.  We outline a programme of research which aims to co-design AMS interventions across healthcare settings, by integrating data-science, evidence- synthesis, behavioural-science and user-centred design. The project includes three work-packages (WP): WP1 (Identifying patterns of prescribing):  analysis of electronic health-records to identify prescribing patterns in care-homes, primary-care, and secondary-care. An online survey will investigate consulting/antibiotic-seeking behaviours in members of the public. WP2 (Barriers and enablers to prescribing in practice): Semi-structured interviews and observations of practice to identify barriers/enablers to prescribing, influences on antibiotic-seeking behaviour and the social/contextual factors underpinning prescribing. Systematic reviews of AMS interventions to identify the components of existing interventions associated with effectiveness. Design workshops to identify constraints influencing the form of the intervention. Interviews conducted with healthcare-professionals in community pharmacies, care-homes, primary-, and secondary-care and with members of the public. Topic guides and analysis based on the Theoretical Domains Framework.  Observations conducted in care-homes, primary and secondary-care with analysis drawing on grounded theory.  Systematic reviews of interventions in each setting will be conducted, and interventions described using the Behaviour Change Technique taxonomy v1. Design workshops in care-homes, primary-, and secondary care. WP3 (Co-production of interventions and dissemination). Findings will be integrated to identify opportunities for interventions, and assess whether existing interventions target influences on antibiotic use. Stakeholder panels will be assembled to co-design and refine interventions in each setting, applying the Affordability, Practicability, Effectiveness, Acceptability, Side-effects and Equity (APEASE) criteria to prioritise candidate interventions.  Outputs will inform development of new AMS interventions and/or optimisation of existing interventions.  We will also develop web-resources for stakeholders providing analyses of antibiotic prescribing patterns, prescribing behaviours, and evidence reviews.