Local Instrumental Variable Methods to Address Confounding and Heterogeneity when Using Electronic Health Records: An Application to Emergency Surgery.

BACKGROUND: Electronic health records (EHRs) offer opportunities for comparative effectiveness research to inform decision making. However, to provide useful evidence, these studies must address confounding and treatment effect heterogeneity according to unmeasured prognostic factors. Local instrumental variable (LIV) methods can help studies address these challenges, but have yet to be applied to EHR data. This article critically examines a LIV approach to evaluate the cost-effectiveness of emergency surgery (ES) for common acute conditions from EHRs. METHODS: This article uses hospital episodes statistics (HES) data for emergency hospital admissions with acute appendicitis, diverticular disease, and abdominal wall hernia to 175 acute hospitals in England from 2010 to 2019. For each emergency admission, the instrumental variable for ES receipt was each hospital's ES rate in the year preceding the emergency admission. The LIV approach provided individual-level estimates of the incremental quality-adjusted life-years, costs and net monetary benefit of ES, which were aggregated to the overall population and subpopulations of interest, and contrasted with those from traditional IV and risk-adjustment approaches. RESULTS: The study included 268,144 (appendicitis), 138,869 (diverticular disease), and 106,432 (hernia) patients. The instrument was found to be strong and to minimize covariate imbalance. For diverticular disease, the results differed by method; although the traditional approaches reported that, overall, ES was not cost-effective, the LIV approach reported that ES was cost-effective but with wide statistical uncertainty. For all 3 conditions, the LIV approach found heterogeneity in the cost-effectiveness estimates across population subgroups: in particular, ES was not cost-effective for patients with severe levels of frailty. CONCLUSIONS: EHRs can be combined with LIV methods to provide evidence on the cost-effectiveness of routinely provided interventions, while fully recognizing heterogeneity. HIGHLIGHTS: This article addresses the confounding and heterogeneity that arise when assessing the comparative effectiveness from electronic health records (EHR) data, by applying a local instrumental variable (LIV) approach to evaluate the cost-effectiveness of emergency surgery (ES) versus alternative strategies, for patients with common acute conditions (appendicitis, diverticular disease, and abdominal wall hernia).The instrumental variable, the hospital's tendency to operate, was found to be strongly associated with ES receipt and to minimize imbalances in baseline characteristics between the comparison groups.The LIV approach found that, for each condition, there was heterogeneity in the estimates of cost-effectiveness according to baseline characteristics.The study illustrates how an LIV approach can be applied to EHR data to provide cost-effectiveness estimates that recognize heterogeneity and can be used to inform decision making as well as to generate hypotheses for further research.

Effectiveness of emergency surgery for five common acute conditions: an instrumental variable analysis of a national routine database.

The effectiveness of emergency surgery vs. non-emergency surgery strategies for emergency admissions with acute appendicitis, gallstone disease, diverticular disease, abdominal wall hernia or intestinal obstruction is unknown. Data on emergency admissions for adult patients from 2010 to 2019 at 175 acute National Health Service hospitals in England were extracted from the Hospital Episode Statistics database. Cohort sizes were: 268,144 (appendicitis); 240,977 (gallstone disease); 138,869 (diverticular disease); 106,432 (hernia); and 133,073 (intestinal obstruction). The primary outcome was number of days alive and out of hospital at 90 days. The effectiveness of emergency surgery vs. non-emergency surgery strategies was estimated using an instrumental variable design and is reported for the cohort and pre-specified sub-groups (age, sex, number of comorbidities and frailty level). Average days alive and out of hospital at 90 days for all five cohorts were similar, with the following mean differences (95%CI) for emergency surgery minus non-emergency surgery after adjusting for confounding: -0.73 days (-2.10-0.64) for appendicitis; 0.60 (-0.10-1.30) for gallstone disease; -2.66 (-15.7-10.4) for diverticular disease; -0.07 (-2.40-2.25) for hernia; and 3.32 (-3.13-9.76) for intestinal obstruction. For patients with 'severe frailty', mean differences (95%CI) in days alive and out of hospital for emergency surgery were lower than for non-emergency surgery strategies: -21.0 (-27.4 to -14.6) for appendicitis; -5.72 (-11.3 to -0.2) for gallstone disease, -38.9 (-63.3 to -14.6) for diverticular disease; -19.5 (-26.6 to -12.3) for hernia; and - 34.5 (-46.7 to -22.4) for intestinal obstruction. For patients without frailty, the mean differences (95%CI) in days alive and out of hospital were: -0.18 (-1.56-1.20) for appendicitis; 0.93 (0.48-1.39) for gallstone disease; 5.35 (-2.56-13.28) for diverticular disease; 2.26 (0.37-4.15) for hernia; and 18.2 (14.8-22.47) for intestinal obstruction. Emergency surgery and non-emergency surgery strategies led to similar average days alive and out of hospital at 90 days for five acute conditions. The comparative effectiveness of emergency surgery and non-emergency surgery strategies for these conditions may be modified by patient factors.

Ethnic variations in the risk of hypoglycaemia among people with Type 2 diabetes prescribed insulins and/or sulfonylureas: a historical cohort study using general practice-recorded data.

AIM: To identify ethnic differences in hypoglycaemic risk among people with Type 2 diabetes prescribed insulins and/or sulfonylureas in community settings. METHODS: Using routine general practice-recorded data, two cohorts of adults with Type 2 diabetes from east London were studied between January 2013 and December 2015: (1) adults prescribed insulins ± other antidiabetes medications (n=7269) and (2) adults prescribed sulfonylureas ± other antidiabetes medications excluding insulins (n=12 502). Incidence rate ratios of hypoglycaemia by ethnicity, adjusting for age, sex, socio-economic status and clustering within Clinical Commissioning Groups, were estimated using random effects Poisson regression. RESULTS: Compared with white British people prescribed insulins, those of black Caribbean ethnicity were at increased hypoglycaemic risk [adjusted incidence rate ratio 1.56 (95% CI 1.21,2.01)], while Bangladeshi people had a lower risk [adjusted incidence rate ratio 0.49 (95% CI, 0.38,0.64)]. In the sulfonylurea cohort, black Caribbean, black African and Indian people all had increased risks of hypoglycaemia compared with white British people [adjusted incidence rate ratios 1.63 (95% CI 1.15,2.29), 1.90 (95% CI 1.32,2.75) and 1.93 (95% CI 1.39,2.69), respectively]. CONCLUSION: The differences in hypoglycaemic risk among people with Type 2 diabetes prescribed insulin and/or sulfonylureas warrant further investigation of any differing biological responses and/or cultural attitudes to antidiabetes therapy among ethnic groups, and should be considered by clinicians evaluating the treatment goals of people with Type 2 diabetes using insulins or sulfonylureas.

A Systematic Review of the Burden of Pancreatic Cancer in Europe: Real-World Impact on Survival, Quality of Life and Costs.

PURPOSE: The purpose of this study was to assess the overall burden of pancreatic cancer in Europe, with a focus on survival time in a real-world setting, and the overall healthy life lost to the disease. METHODS: Real-world data were retrieved from peer-reviewed, observational studies identified by an electronic search. We performed two de novo analyses: a proportional shortfall analysis to quantify the proportion of healthy life lost to pancreatic cancer and an estimation of the aggregate life-years lost annually in Europe. RESULTS: Ninety-one studies were included. The median, age-standardised incidence of pancreatic cancer per 100,000 was 7.6 in men and 4.9 in women. Overall median survival from diagnosis was 4.6 months; median survival was 2.8-5.7 months in patients with metastatic disease. The proportional shortfall analysis showed that pancreatic cancer results in a 98 % loss of healthy life, with a life expectancy at diagnosis of 4.6 months compared to 15.1 years for an age-matched healthy population. Annually, 610,000-915,000 quality-adjusted life-years (QALYs) are lost to pancreatic cancer in Europe. Patients had significantly lower scores on validated health-related quality of life instruments versus population norms. CONCLUSIONS: To the best of our knowledge, this is the first study to systematically review real-world overall survival and patient outcomes of pancreatic cancer patients in Europe outside the context of clinical trials. Our findings confirm the poor prognosis and short survival reported by national studies. Pancreatic cancer is a substantial burden in Europe, with nearly a million aggregate life-years lost annually and almost complete loss of healthy life in affected individuals.

Influence of surgical rate on patients' reported clinical need and outcomes in English NHS.

BACKGROUND: To determine if higher rates of surgery are associated with lower levels of need (patients' pre-operative reports of their symptoms, functional status and quality of life) and with less benefit (patients' post-operative reports). METHODS: Patient-reported outcome measures (PROMs) collected before and after joint replacement, hernia repair or varicose vein (VV) surgery in National Health Service (NHS)-funded patients (2009/11). Regression analysis for associations between 10% increase in rates and mean PROM score for Primary Care Trust (PCT) populations. RESULTS: National rate for hip and knee replacement increased by 6%, unchanged for hernia repair and decreased by 26% for VV surgery. Changes in PCT rates varied but had little or no association with the mean level of need of patients: 10% increase in the rate was associated with only 0.3% decline in the pre-operative PROM score for knee replacement (P < 0.05) and VV surgery (P < 0.001) and no significant change for other procedures. There was no significant association between a 10% change in the rate and the amount of benefit from surgery apart from a slight reduction (0.46%; P < 0.001) in the disease-specific PROM score for VV surgery. CONCLUSION: Policies by commissioners to reduce surgical rates in the English NHS cannot be justified on the grounds of avoiding inappropriate operations or increasing cost-utility.

Review of health-related quality of life data in multiple myeloma patients treated with novel agents.

In multiple myeloma (MM), health-related quality of life (HRQoL) data is becoming increasingly important, owing to improved survival outcomes and the impact of treatment-related toxicity on HRQoL. Researchers are more frequently including HRQoL assessments in clinical trials, but analysis and reporting of this data has not been consistent. A systematic literature review assessed the effect of novel agents (thalidomide, bortezomib and lenalidomide) on HRQoL in MM patients, and evaluated the subsequent reporting of these HRQoL results. A relatively small body of literature addresses HRQoL data in MM patients treated with novel MM therapeutic agents: 9 manuscripts and 15 conference proceedings. The literature demonstrates the complementary value of HRQoL when assessing clinical response, progression, overall survival and toxicity. However, weaknesses and inconsistencies in analysis and presentation of HRQoL data were observed, often complicating interpretation of the impact of treatment on HRQoL in MM. Further evaluation of HRQoL in MM patients treated with novel agents is required in larger cohorts, and ideally in head-to-head comparative studies. Additionally, the development of standardised MM-specific best practice guidelines in HRQoL data collection and analysis is recommended. These would ensure that future data are more useful in guiding predictive models and clinical decisions.

Socioeconomic differences in patient-reported outcomes after a hip or knee replacement in the English National Health Service.

BACKGROUND: We investigated socioeconomic differences in patient-reported outcomes after a hip or knee replacement and the contribution of health differences beforehand. METHODS: Our sample included 121 983 patients in England who had an operation in 2009-2011. Socioeconomic status was measured with quintiles of the ranking of areas by the English Index of Multiple Deprivation. Outcomes at 6 months were the Oxford hip or knee score (OHS or OKS) that measure pain and disability on a scale from 0 (worst) to 48 (best), and the percentage reporting no improvement in problems. Adjustment was made for age, sex, ethnicity, comorbidity, general health, revision surgery, primary diagnosis, preoperative OHS or OKS and having longstanding problems. RESULTS: Comparing the most- with the least-deprived group, the mean OHS was 5.0 points lower and the OKS 5.4 lower. Adjusted differences, reflecting the differences in improvement in the condition, were 2.8 [95% confidence interval (CI): 2.5-3.0] on OHS and 2.4 (95% CI: 2.2-2.7) on OKS. Adjusted odds ratios for reporting no improvement were 1.4 (1.2-1.6) for the hip and 1.4 (1.3-1.5) for the knee. CONCLUSIONS: On average, patients living in socioeconomically deprived areas had worse outcomes after surgery, partly related to preoperative differences in health and disease severity and partly to less postoperative improvement.

Sociodemographic differences in the severity and duration of disease amongst patients undergoing hip or knee replacement surgery.

BACKGROUND: Differences in the use of hip and knee replacement by sex, age, ethnicity or socioeconomic status may lead to differences in disease severity between those who have surgery. METHODS: Analyses used data collected from 117,736 patients in 2009-10 via the Patient Reported Outcome Measures (PROMs) programme in England. Adjusted differences were estimated in the Oxford Hip Score (OHS) or the Oxford Knee Score (OKS), both expressed on a scale from 0 to 48, and the proportion with longstanding problems (>5 years), expressed as odds ratios (ORs). RESULTS: Women had more severe pain and disability than men on average (difference OHS 2.3 and OKS 3.3), but less often longstanding problems. Compared with white patients, average severity was higher in South Asian patients (difference OHS 2.7 and OKS 3.0) and in black patients (difference OHS 0.9 and OKS 1.6), who also more often had longstanding problems (OR 1.40 for hip and 1.54 for knee). Patients from deprived areas had more severe disease (difference OHS 3.6 and OKS 3.3 between least and most deprived quintile). CONCLUSIONS: There is evidence that non-white and deprived patients tend to have hip and knee replacement surgery at a later stage in the course of their disease.

The association between quality of primary care, deprivation and cardiovascular outcomes: a cross-sectional study using data from the UK Quality and Outcomes Framework.

BACKGROUND: The Quality and Outcomes Framework, a financial incentive scheme for general practitioners in the UK, seems to have improved the quality of primary care and reduced inequalities in primary care delivery. It remains unclear, however, whether higher-quality primary care improves health outcomes or reduces health inequalities. METHODS: We conducted a cross-sectional study examining the association between quality of cardiovascular care and coronary heart disease (CHD) outcomes in 1531 general practices in London. We calculated CHD quality achievement scores (ranging from 0 to 100) for each practice using the 2006-2007 data from the Quality and Outcomes Framework. We used weighted linear regression models to assess the practice-level association between the CHD quality score and CHD admissions and deaths. FINDINGS: Overall, practices with higher CHD quality achievement scores had better CHD outcomes. Each one point increase in the CHD quality achievement score was associated with 4.28 (95% CI 1.19 to 7.38; p=0.007) fewer admissions per 100,000 for practices serving highly deprived populations and 2.11 (95% CI 0.68 to 3.55; p=0.004) fewer admissions per 100 000 for practices serving populations of average deprivation. There was no association between the CHD quality achievement score and the CHD admissions for practices serving affluent populations (p=0.906). We observed a similar deprivation-dependent gradient between quality achievement and CHD deaths. INTERPRETATION: High-quality primary care is associated with improved health outcomes. This association is strongest in deprived areas, suggesting that high-quality primary care may play an important role in reducing health inequalities.

Mortality due to injuries by place of occurrence in the European region: analysis of data quality in the WHO mortality database.

BACKGROUND AND OBJECTIVE: The 10th revision of the International Classification of Diseases (ICD-10) has codes for the place of occurrence of external causes of mortality. The purpose of this study was to investigate the quality of data available in the World Health Organization (WHO) mortality database on the place of occurrence of fatal injuries in the European region. METHODS: Data on external causes of mortality from countries in the European region according to the ICD-10 with four-character subdivision, between the years 1998 and 2003, were analysed. The quality of ICD-10 place of occurrence data was analysed for each country, based on the completeness, coverage and percentage of unspecified place of injury occurrence. RESULTS: Only three countries in the European region (Hungary, Iceland and Lithuania) had high quality of data on place of occurrence of injuries, and six had medium-quality data. CONCLUSIONS: Few countries in the European region have injury mortality data of adequate quality by place of occurrence.

Tension pneumothorax secondary to automatic mechanical compression decompression device.

The details are presented of the first published case of a tension pneumothorax induced by an automatic compression-decompression (ACD) device during cardiac arrest. An elderly patient collapsed with back pain and, on arrival of the crew, was in pulseless electrical activity (PEA) arrest. He was promptly intubated and correct placement of the endotracheal tube was confirmed by noting equal air entry bilaterally and the ACD device applied. On the way to the hospital he was noted to have absent breath sounds on the left without any change in the position of the endotracheal tube. Needle decompression of the left chest caused a hiss of air but the patient remained in PEA. Intercostal drain insertion in the emergency department released a large quantity of air from his left chest but without any change in his condition. Post-mortem examination revealed a ruptured abdominal aortic aneurysm as the cause of death. Multiple left rib fractures and a left lung laceration secondary to the use of the ACD device were also noted, although the pathologist felt that the tension pneumothorax had not contributed to the patient's death. It is recommended that a simple or tension pneumothorax should be considered when there is unilateral absence of breath sounds in addition to endobronchial intubation if an ACD device is being used.

Does the decline in road traffic injury in London vary by ethnicity?

OBJECTIVE: To examine ethnic variations in trends in road traffic injuries in London. DESIGN: Analysis of STATS19 data comparing trends in road traffic casualty rates by ethnic group. SETTING: London, 2001-6. SUBJECTS: Children (or=15 years). MAIN OUTCOME MEASURES: Annual casualty rates in white, black, and Asian pedestrians, cyclists, and car occupants. RESULTS: Casualty rates in London declined each year between 2001 and 2006 by an average of 8.8% (95% CI 8.5% to 9.0%). After adjustment for area-level deprivation, there was good evidence that the average annual reduction in injury rates in car occupants was significantly less in Asian than in white adults (10.9% vs 14.4%, p<0.001). There was some evidence that average annual reductions in injury rates were lower in black than in white adult pedestrians (7.4% vs 9.3%, p = 0.041) and car occupants (13.2% vs 14.4%, p = 0.03). CONCLUSIONS: Casualty rates in London have declined for pedestrians, cyclists, and car occupants in three broad ethnic groups. Asian car drivers appear to have benefited least from these reductions.

Is intimal hyperplasia a marker of neuro-ophthalmic complications of giant cell arteritis?

OBJECTIVE: The ischaemic complications of giant cell arteritis (GCA) such as blindness and stroke may result from luminal narrowing of the affected arteries. This study focuses on the association between the severity of intimal proliferation on temporal artery biopsy (TAB) histology and neuro-ophthalmic complications (NOCs) of GCA. METHOD: We identified 30 cases of biopsy-proven temporal arteritis. One histopathologist (blinded to the clinical details) evaluated the TAB specimens and categorized the degree of maximum stenosis due to intimal hyperplasia into four grades: grade 1 is <50% luminal occlusion due to intimal hyperplasia, grade 2 is 50-75%, grade 3 is >75% and grade 4 is complete luminal occlusion. A second histopathologist (also blinded to the clinical details) independently evaluated the TAB specimens using the same grading system. The NOCs in these patients were noted after a case record review. RESULTS: Of the 30 patients, 12 had NOC-10 with eye complications (complete visual loss, anterior ischaemic neuropathy, visual field defects), one patient had cerebral infarcts and one had both cerebral infarcts and vision loss. There was evidence for a statistically significant trend of NOC associated with higher intimal hyperplasia scores (P = 0.001). The scores of the histopathologists agreed for 23 (77%) patients and differed by 1 category for the remaining 7 (kappa-statistic 0.88). CONCLUSIONS: Our study suggests that the degree of intimal hyperplasia on TAB histology (routinely available to all hospital units) seems to be closely associated with NOCs of GCA. The study highlights the possible prognostic as well as diagnostic role of the biopsy. We feel that intimal hyperplasia noted in biopsy specimens may help us in the risk stratification of GCA patients and targeting of appropriate and novel therapies.

Reporting ethics committee approval and patient consent by study design in five general medical journals.

BACKGROUND: Authors are required to describe in their manuscripts ethical approval from an appropriate committee and how consent was obtained from participants when research involves human participants. OBJECTIVE: To assess the reporting of these protections for several study designs in general medical journals. DESIGN: A consecutive series of research papers published in the Annals of Internal Medicine, BMJ, JAMA, Lancet and The New England Journal of Medicine between February and May 2003 were reviewed for the reporting of ethical approval and patient consent. Ethical approval, name of approving committee, type of consent, data source and whether the study used data collected as part of a study reported elsewhere were recorded. Differences in failure to report approval and consent by study design, journal and vulnerable study population were evaluated using multivariable logistic regression. RESULTS: Ethical approval and consent were not mentioned in 31% and 47% of manuscripts, respectively. 88 (27%) papers failed to report both approval and consent. Failure to mention ethical approval or consent was significantly more likely in all study designs (except case-control and qualitative studies) than in randomised controlled trials (RCTs). Failure to mention approval was most common in the BMJ and was significantly more likely than in The New England Journal of Medicine. Failure to mention consent was most common in the BMJ and was significantly more likely than in all other journals. No significant differences in approval or consent were found when comparing studies of vulnerable and non-vulnerable participants. CONCLUSION: The reporting of ethical approval and consent in RCTs has improved, but journals are less good at reporting this information for other study designs. Journals should publish this information for all research on human participants.

An experimental study of determinants of the extent of disagreement within clinical guideline development groups.

OBJECTIVE: To assess the effect of design features and clinical and social cues on the extent of disagreement among participants in a formal consensus development process. METHODS: Factorial design involving 16 groups consisting of 135 general practitioners (GPs) and 42 mental health professionals from England. The groups rated the appropriateness of four mental health interventions for three conditions (chronic back pain, irritable bowel syndrome, and chronic fatigue syndrome) in the context of various clinical and social cues. The groups differed in three design features: provision of a systematic literature review (versus not provided), group composition (mixed versus GP only), and assumptions about the healthcare resources available (realistic versus idealistic). Disagreement was measured using the mean absolute deviation from a group's median rating for a scenario. RESULTS: None of the design features significantly affected the extent of disagreement within groups (all p>0.3). Disagreement did differ between treatments (closer consensus for cognitive behavioural therapy and behavioural therapy than for brief psychodynamic intervention therapy and antidepressants) and cues (closer consensus for depressed patients and patients willing to try any treatment). CONCLUSION: In terms of the extent of disagreement in the groups in this study, formal consensus development was a robust technique in that the results were not dependent on the way it was conducted.

Increased risk of tuberculosis among health care workers in Samara Oblast, Russia: analysis of notification data.

SETTING: Samara Oblast, Russia. OBJECTIVE: To compare the rates of tuberculosis (TB) in health care workers (HCWs) working in TB services, general health services (GHS) and the general population in a region of the Russian Federation. DESIGN: Analysis of notification rates of TB among HCWs, GHS workers and the general population during the 9-year period from 1994 to 2002. RESULTS: During 1994-2002, TB incidence among staff employed at the TB services in Samara Oblast was ten times higher than among the general population, reaching 741.6/100 000 person years at risk. Staff working at in-patient TB facilities were found to be at highest risk, with an incidence rate ratio of 17.7 (95% CI 11.6-27.0) compared to HCWs at the GHS. CONCLUSIONS: HCWs at TB services in the Russian Federation are at substantially increased risk for TB, suggesting significant risks from nosocomial transmission. Control of institutional spread of TB in the Russian Federation is an area that requires urgent attention, especially given the epidemic of human immunodeficiency virus that Russia is currently witnessing.

The effects of acute ethanol intake on isoniazid pharmacokinetics.

AIM: To assess effects of acute ethanol intake on the pharmacokinetics of isoniazid in healthy male volunteers. METHODS: Sixteen healthy male, drug-free subjects were studied. Each received in the fasting state, on two occasions separated by at least 1 week, isoniazid (200 mg orally). On one occasion (assigned randomly), subjects received ethanol 0.73 g/kg, 1 h before isoniazid, followed by 0.11 g/kg ethanol orally every hour thereafter for 7 h. Plasma isoniazid and acetylisoniazid concentrations were measured by means of high-performance liquid chromatography. Blood ethanol concentrations were measured hourly by breath analysis. Plasma concentrations of isoniazid and acetylisoniazid were analysed using TOPFIT software. RESULTS: Peak concentrations of isoniazid were reached within 90 min, in both the ethanol-treated and control groups. The ethanol dosage regimen used resulted in peak blood ethanol concentrations between 78 mg/l and 103 mg/l. There was no significant difference in area under the curve, half-life of elimination or the ratio of acetylisoniazid to isoniazid (AcINH/INH) in the sample withdrawn 3 h after isoniazid dose. Acetylator phenotype for patients was the same in both phases, whether assessed by half-life of isoniazid or the AcINH/INH ratio at 3 h. CONCLUSIONS: Acute ethanol intake at this dose is unlikely to affect results of acetylation studies in which isoniazid is used as a substrate, whether the half-life of isoniazid or the AcINH /INH ratio at 3 h is used to phenotype patients.

Inappropriate requests for serum anti-epileptic drug levels in hospital practice.

BACKGROUND: Serum anti-epileptic drug (AED) levels are indicated to assess AED adherence or toxicity, and are applicable to only a few AEDs. Expert consensus views on the clinical role of serum AED levels are summarized in the evidence-based guidelines published by the Scottish Intercollegiate Guidelines Network. AIM: To examine local compliance with these guidelines. DESIGN: Retrospective case-note audit. METHODS: We included all serum AED level measurements requested from our hospital over two months. Our audit standards were first, that serum AED levels should be requested only for suspicion of poor AED adherence or toxicity ('indication-compliant'), and secondly, for 'full compliance', that 'indication-compliant' requests should be made only for AEDs with established dose-response and dose-toxicity relationships (phenytoin, carbamazepine, phenobarbitone). RESULTS: There were 114 measurements in 102 patients. Serum AED level requests were for phenytoin (n = 50), valproate (n = 27), carbamazepine (n = 22), lamotrigine (n = 8), phenobarbitone (n = 7), and were made by physicians (n = 46), paediatricians (n = 30), neurologists (n = 15), neurosurgeons (n = 14), psychiatrists (n = 7), and intensivists (n = 2). AED toxicity was queried in 29 requests (25%), and adherence in 10 (9%); thus 34% of requests were 'indication-compliant'. However, 16 of these were for valproate or lamotrigine; thus only 23 requests (20%) were 'fully compliant'. Clinical management changed in only 17 of the 47 patients whose levels fell outside target ranges, and only two of these followed indication-compliant AED measurement. DISCUSSION: The audit identified a failure locally to comply with standard evidence-based guidelines. If, as is likely, this reflects practice elsewhere in the UK, there are potentially major clinical management and resource implications.