Faculty Bibliography

BACKGROUND:NHS hospitals collect a wealth of administrative data covering accident and emergency (A&E) department attendances, inpatient and day case activity, and outpatient appointments. Such data are increasingly being used to compare units and services, but adjusting for risk is difficult. OBJECTIVES:To derive robust risk-adjustment models for various patient groups, including those admitted for heart failure (HF), acute myocardial infarction, colorectal and orthopaedic surgery, and outcomes adjusting for available patient factors such as comorbidity, using England’s Hospital Episode Statistics (HES) data. To assess if more sophisticated statistical methods based on machine learning such as artificial neural networks (ANNs) outperform traditional logistic regression (LR) for risk prediction. To update and assess for the NHS the Charlson index for comorbidity. To assess the usefulness of outpatient data for these models. MAIN OUTCOME MEASURES:Mortality, readmission, return to theatre, outpatient non-attendance. For HF patients we considered various readmission measures such as diagnosis-specific and total within a year. METHODS:We systematically reviewed studies comparing two or more comorbidity indices. Logistic regression, ANNs, support vector machines and random forests were compared for mortality and readmission. Models were assessed using discrimination and calibration statistics. Competing risks proportional hazards regression and various count models were used for future admissions and bed-days. RESULTS: = 0.73 for death following stroke. Calibration was often good for procedure groups but poorer for diagnosis groups, with overprediction of low risk a common cause. The machine learning methods we investigated offered little beyond LR for their greater complexity and implementation difficulties. For HF, some patient-level predictors differed by primary diagnosis of readmission but not by length of follow-up. Prior non-attendance at outpatient appointments was a useful, strong predictor of readmission. Hospital-level readmission rates for HF did not correlate with readmission rates for non-HF; hospital performance on national audit process measures largely correlated only with HF readmission rates. CONCLUSIONS:Many practical risk-prediction or casemix adjustment models can be generated from HES data using LR, though an extra step is often required for accurate calibration. Including outpatient data in readmission models is useful. The three machine learning methods we assessed added little with these data. Readmission rates for HF patients should be divided by diagnosis on readmission when used for quality improvement. FUTURE WORK:As HES data continue to develop and improve in scope and accuracy, they can be used more, for instance A&E records. The return to theatre metric appears promising and could be extended to other index procedures and specialties. While our data did not warrant the testing of a larger number of machine learning methods, databases augmented with physiological and pathology information, for example, might benefit from methods such as boosted trees. Finally, one could apply the HF readmissions analysis to other chronic conditions. FUNDING:The National Institute for Health Research Health Services and Delivery Research programme.

Sheng-Chia Chung and colleagues report in The Lancet (23 January 2014) an international comparison of cardiovascular patient mortality between the UK and Sweden. They suggest that "more than 10000 deaths at 30 days would have been prevented or delayed had UK patients experienced the care of their Swedish counterparts." Further, they estimate that 1741 deaths would have been prevented in the UK had the Swedish pattern of primary percutaneous coronary intervention (PCI) and beta-blocker use been replicated in the NHS from 2004 to 2010. However, their study does not provide convincing evidence that faster uptake of primary PCI or beta-blockers on discharge would have had an effect on cardiovascular patient mortality in the UK.

In breast cancer patients undergoing sentinel lymph node biopsy (SLNB) analysis using one-step nucleic acid amplification (OSNA), clarity is required as to the determinants of further metastasis risk upon completion axillary lymph node dissection (ALND). This study aims to identify whether the proportion of sentinel nodes containing metastases predicts risk of further axillary disease.

OBJECTIVE: To identify risk factors for falls and generate two screening tools: an opportunistic tool for use in consultation to flag at risk patients and a systematic database screening tool for comprehensive falls assessment of the practice population. STUDY DESIGN AND SETTING: This multicenter cohort study was part of the quality improvement in chronic kidney disease trial. Routine data for participants aged 65 years and above were collected from 127 general practice (GP) databases across the UK, including sociodemographic, physical, diagnostic, pharmaceutical, lifestyle factors, and records of falls or fractures over 5 years. Multilevel logistic regression analyses were performed to identify predictors. The strongest predictors were used to generate a decision tree and risk score. RESULTS: Of the 135,433 individuals included, 10,766 (8%) experienced a fall or fracture during follow-up. Age, female sex, previous fall, nocturia, anti-depressant use, and urinary incontinence were the strongest predictors from our risk profile (area under the receiver operating characteristics curve = 0.72). Medication for hypertension did not increase the falls risk. Females aged over 75 years and subjects with a previous fall were the highest risk groups from the decision tree. The risk profile was converted into a risk score (range -7 to 56). Using a cut-off of >/=9, sensitivity was 68%, and specificity was 60%. CONCLUSION: Our study developed opportunistic and systematic tools to predict falls without additional mobility assessments.

BACKGROUND: The National Institute for Health and Care Excellence (NICE) recommends postpartum and annual monitoring for diabetes for females who have had a diagnosis of gestational diabetes mellitus (GDM). AIM: To describe the current state of follow-up after GDM in primary care, in England. DESIGN AND SETTING: A retrospective cohort study in 127 primary care practices. The total population analysed comprised 473 772 females, of whom 2016 had a diagnosis of GDM. METHOD: Two subgroups of females were analysed using electronic general practice records. In the first group of females (n = 788) the quality of postpartum follow-up was assessed during a 6-month period. The quality of long-term annual follow-up was assessed in a second group of females (n = 718), over a 5-year period. The two outcome measures were blood glucose testing performed within 6 months postpartum (first group) and blood glucose testing performed annually (second group). RESULTS: Postpartum follow-up was performed in 146 (18.5%) females within 6 months of delivery. Annual rates of long-term follow-up stayed consistently around 20% a year. Publication of the Diabetes in Pregnancy NICE guidelines, in 2008, had no effect on long-term screening rates. Substantial regional differences were identified among rates of follow-up. CONCLUSION: Monitoring of females after GDM is markedly suboptimal despite current recommendations.