Publications
scientifiques

INTRODUCTION: In the UK, most people with lung cancer are diagnosed at a late stage when curative treatment is not possible. To aid earlier detection, the socio-demographic and early clinical features predictive of lung cancer need to be identified. METHODS: We studied 12,074 cases of lung cancer and 120,731 controls in a large general practice database. Logistic regression analyses were used to identify the socio-demographic and clinical features associated with cancer up to 2 years before diagnosis. A risk prediction model was developed using variables that were independently associated with lung cancer up to 4 months before diagnosis. The model performance was assessed in an independent dataset of 1,826,293 patients from the same database. Discrimination was assessed by means of a receiver operating characteristic (ROC) curve. RESULTS: Clinical and socio-demographic features that were independently associated with lung cancer were patients' age, sex, socioeconomic status and smoking history. From 4 to 12 months before diagnosis, the frequency of consultations and symptom records of cough, haemoptysis, dyspnoea, weight loss, lower respiratory tract infections, non-specific chest infections, chest pain, hoarseness, upper respiratory tract infections and chronic obstructive pulmonary disease were also independently predictive of lung cancer. On validation, the model performed well with an area under the ROC curve of 0.88. CONCLUSIONS: This new model performed substantially better than the current National Institute for Health and Clinical Excellence referral guidelines and all comparable models. It has the potential to predict lung cancer cases sufficiently early to make detection at a curable stage more likely by allowing general practitioners to better risk stratify their patients. A clinical trial is needed to quantify the absolute benefits to patients and the cost effectiveness of this model in practice.

To examine time trends and the influence of socio-demographic and geographic factors on incidence of Parkinson's disease (PD) diagnosis in a large UK population-based cohort from The Health Improvement Network (THIN), a UK primary care database. All patients aged over 50 years and contributing data to THIN between January 1999 and December 2009 were extracted and the incidence rates of PD diagnoses were determined by age, gender, time period, social deprivation score and urban/rural status. The overall incidence of PD diagnosis for people over 50 years was 84 per 100,000 person years (95 % CI 82-85). After accounting for socio-demographic factors, the adjusted incidence rates in men were 46 % higher (95 % CI 43-48 %) than in women. Adjusted incidence rates were also 12 % higher in urban than rural areas (95 % CI 4-20 %) and slightly lower in less socially deprived areas. Over time there was a downward trend in PD diagnosis with the adjusted incidence rates declining by around 6 % every calendar year (95 % CI 5-6 %) between 1999 and 2009. Broadening the definition of PD from diagnostic codes to include symptoms and antiparkinsonian drug prescriptions increased the overall incidence to 118 per 100,000 person years (95 % CI 116-120) and showed a much weaker downward trend over time of around 1 % per calendar year (95 % CI 1-2 %). With the broader definition, the adjusted incidence rates remained significantly higher in men compared to women and in urban areas compared with rural areas but not in socially deprived areas. The PD diagnosis rates in the primary care setting were, as expected, higher in men, and slightly higher in urban areas, but not different between socio-economic groups. There was a decline in PD diagnosis in the primary care setting, which may largely represent changes in diagnosis and/or coding rather than a true decline in incidence.

AIMS: To determine if a diagnostic record of poor treatment compliance (medication non-compliance and/or non-attendance at medical appointments) was associated with all-cause mortality in people with type 1 diabetes. METHODS: This is an observational cohort study of data extracted from The Health Improvement Network (THIN) database, comprising data on patients served by over 350 primary care practices in the U.K. Participants were included in the study if they had diagnostic codes indicative of type 1 diabetes. Treatment non-compliance was defined as missing one or more scheduled appointment, or one or more codes indicating medication non-compliance. RESULTS: Of 2946 patients with type 1 diabetes, 867 (29.4%) had a record of either appointment non-attendance or medication non-compliance in the 30 month compliance assessment period. The crude, unadjusted mortality rate for those patients who were treatment non-compliant was 1.462 (95% CI 0.954-2.205). Following adjustment for confounding factors, treatment non-compliance was associated with increased all-cause mortality (HR=1.642; 95% CI 1.055-2.554). CONCLUSIONS: Treatment non-compliance was associated with increased all-cause mortality in patients with type 1 diabetes. Understanding and addressing factors that contribute to patient treatment non-compliance will be important in improving the life expectancy of patients with type 1 diabetes.

BACKGROUND: Despite the proven efficacy of warfarin, its use in patients with Atrial Fibrillation (AF) is reportedly low. We investigated the underuse and overuse of warfarin in the management of AF in general practices in the United Kingdom (UK) against the National Institute of Clinical Excellence (NICE, UK) guidelines whilst seeking to identify subgroups of AF patients to inform efforts to optimise warfarin use. METHODOLOGY: A retrospective database analysis to determine warfarin prescribing using tree models based on 50,361 patients with AF (classified as low, moderate and high risk of stroke using CHADS2) from 430 general practices in the UK. RESULTS: Over one-third (37.0%, 4573/12,351) of low risk AF patients were on warfarin, compared with 47.1% (8349/17,709) moderate risk AF patients and 54.9% (11,142/20,301) high risk AF patients. Clinical subgroups (n = 15 low risk subgroups, n = 15 medium risk subgroups, n = 22 high risk subgroups) were identified. Several factors not supported by current guidelines (age, BMI, dementia, gender) were associated with the use of warfarin. Gender and BMI were associated with warfarin use in low and medium risk AF patients but not in high risk AF patients. CONCLUSION: Whilst NICE guidelines suggest that all high risk AF patients should be on warfarin, half of those at moderate risk should be on warfarin and none of those at low risk should be on warfarin, we found evidence of over and under use of warfarin. Interventions to optimise warfarin therapy tailored to and targeting specific subgroups of AF patients identified by the tree models are required.

BACKGROUND: It has been estimated that, at any point in time, about 1% of the general adult population of the United Kingdom is receiving long-term (ie, ≥ 3 months) oral glucocorticoid therapy. These patients may develop neuropsychiatric disorders when the drug is discontinued. METHOD: Data were obtained for all adult patients registered from January 1, 1990, through December 31, 2008, at UK general practices that were contributors to The Health Improvement Network database. Data from 21,995 adult patients who had been exposed to long-term oral glucocorticoids and who had discontinued the drugs after an exposure ranging from 1 to 3 years were analyzed. The within-person incidence rate ratios (IRRs) for Read codes and/or prescriptions for depression, delirium/confusion, mania, panic disorders, and suicide or suicide attempt during the withdrawal period were estimated using a self-controlled case series methodology. The predictors of the outcomes were ascertained using Cox proportional hazards models. RESULTS: The risk of depression (IRR = 1.13; 95% CI, 1.00-1.28; P = .04) and of delirium/confusion (IRR = 2.67; 95% CI, 1.96-3.63; P < .001) was significantly higher during the discontinuation period compared to a reference period defined as ranging from 5 to 3 months before the drug cessation. Older people were at higher risk of delirium/confusion. The use of long-acting glucocorticoids was associated with a higher risk of both depression (adjusted hazard ratio [HR] = 1.92; 95% CI, 1.07-3.46) and delirium/confusion (adjusted HR = 4.96; 95% CI, 2.60-9.49) during the withdrawal period. CONCLUSIONS: Discontinuation of long-term glucocorticoid therapy is associated with an increased risk of both depression and delirium/confusion. People treated with long-acting glucocorticoids are particularly at risk.

STUDY QUESTION: What are the age-specific incident rates of clinically recorded fertility problems in women aged 15-49 years and how do they vary by socioeconomic group and geographic area. SUMMARY ANSWER: The incident rate of recorded fertility problems was highest in women age 30-34 years: about 1% of women per annum. Overall rates did not vary by socioeconomic group; however, age-specific rates varied substantially by socioeconomic deprivation quintile; among younger women, deprivation was associated with higher infertility rates. WHAT IS KNOWN ALREADY AND WHAT THIS PAPER ADDS: The rates of infertility in the UK range from 2 to 26%. Infertility definitions and denominators vary widely, and most current evidence is based on questionnaire studies that are subject to recall, reporting and selection bias. The current paper presents population-based estimates of clinically recorded fertility problems in women of reproductive age and the variation by age and socioeconomic deprivation quintile across different regions of the UK, using a nationally representative cohort of women that is larger than any previous study. Although infertility overall does not vary by socioeconomic status, consultation for fertility problems is closely related to socioeconomic patterns of women's age at first conception, demonstrating that many couples have pre-existing, rather than specifically age-related, infertility. STUDY DESIGN, SIZE, DURATION: This cohort study used data from The Health Improvement Network, a computerized primary care database of anonymized patient records from general practices across the UK, with prospective health records on over 1.7 million women between 1990 and 2010. PARTICIPANTS/MATERIALS, SETTING AND METHODS: Our cohort included 1,776,746 women of reproductive age (age 15-49 years) who contributed one or more years of active general practice registration. We estimated rates of new clinically recorded fertility problems in these women using medical records and medications exclusively used to treat infertility. We assessed variation in age-specific incidence by socioeconomic deprivation quintile and geographic area using Poisson regression. MAIN RESULTS AND THE ROLE OF CHANCE: The rate of incident recorded fertility problems was highest in women in the 30-34 year age group (10.9 per 1000 person-years), which equates to approximately 1% of women per annum in this age group. Lowest rates were in women in the 15-19 and 45-49 year age groups (0.7 and 0.4 per 1000 person-years, respectively). Overall rates did not vary by socioeconomic group, measured using quintiles of the Townsend index. Age-specific rates, however, varied substantially with socioeconomic deprivation quintile (P-value for interaction < 0.0001) such that up to age 25, women with more deprivation had more recorded fertility problems [rate ratio (RR) comparing most to least deprived 5.6, 95% confidence interval (CI) 4.4-7.2 at 15-20 years of age]. This reversed from age 25 to 39, when women with more deprivation had fewer recorded fertility problems (RR 0.6 95% CI 0.5-0.6 at age 30-34). After age 40, there was no socioeconomic gradient in absolute rates. LIMITATIONS, REASONS FOR CAUTION: This is by far the largest population-based study to estimate clinically recorded fertility problems in women and the first in the UK to assess variation across such a broad age group from 15 to 49 years. Our data, however, did not capture women who experience difficulty in conceiving, but do not consult their general practitioner (GP) regarding fertility problems. WIDER IMPLICATIONS OF THE FINDINGS: Compared with existing estimates, our measures of the extent and distribution of recorded fertility problems in primary care are more useful for GPs, primary care trusts and policy makers for the planning and delivery of fertility services. We have shown a high burden of infertility with little geographic variation; however, the significant burden in young, more deprived women needs recognition in light of age restrictions for treatment availability for infertility in the UK. Not only does treatment access need to be universal and more equitably allocated across socioeconomic groups, but also more resources are required to reduce fertility problems by targeting modifiable risk factors. STUDY FUNDING/COMPETING INTEREST(S): There was no direct source of funding for this research work. N.N.D. completed the work as part of an M.Sc., which was funded by Developing Solutions Scholarship provided by the International Office, University of Nottingham. J.W. is supported by a University of Nottingham/National Institute for Health Research (NIHR) Senior Clinical Research Fellowship. TRIAL REGISTRATION NUMBER: Not applicable.

INTRODUCTION: To evaluate the performance of QCancer® (Renal) for predicting the absolute risk of renal tract cancer in a large independent UK cohort of patients from general practice records. MATERIALS AND METHODS: Open cohort study to validate QCancer® (Renal) prediction model. Record from 365 practices from United Kingdom contributing to The Health Improvement Network (THIN) database. 2.1 million patients registered with a general practice surgery between 01 January 2000 and 30 June 2008, aged 30-84 years (3.7 million person years) with 2283 renal tract cancer cases. Renal tract cancer was defined as incident diagnosis of renal tract cancer during the 2 years after study entry. Model discrimination was measured using the receiver operating characteristics derived area under the curve. Calibration plots examined the relationship between predicted and observed probabilities of undetected renal tract cancer. RESULTS: The results from this independent and external validation of QCancer® (Renal) demonstrated good performance data on a large cohort of general practice patients. QCancer® (Renal) had very good discrimination with areas under the ROC curve of 0.92 and 0.95 for women and men respectively. QCancer® (Renal) was well calibrated across all tenths of risk and over all age ranges with predicted risks closely matching observed risks. QCancer® (Renal) explained 74.4% and 74.2% of the variation in men and women respectively. A limitation of our study is the recording of symptoms might be less complete, as patients with mild symptoms may not visit their general practitioner or not report mild symptoms. CONCLUSIONS: QCancer® (Renal) are useful tools to help in identifying undetected cases of undiagnosed renal tract cancer in primary care in the UK.

OBJECTIVES: Adjustment for morbidity is important to ensure fair comparison of outcomes between patient groups and health care providers. The Quality and Outcomes Framework (QOF) in UK primary care offers potential for developing a standardized morbidity score for low-risk populations. STUDY DESIGN AND SETTING: Retrospective cohort study of 653,780 patients aged 60 years or older registered with 375 practices in 2008 in a large primary care database (The Health Improvement Network). Half the practices were randomly selected to derive a morbidity score predicting 1-year mortality; the others assessed predictive performance. RESULTS: Nine chronic conditions were robust copredictors (hazard ratio = ≥1.2) of mortality independent of age and sex, producing high predictive discrimination (c-statistic = 0.82). An individual's QOF score explained more between practice variation in mortality than the Charlson index (46% vs. 32%). At practice level, mean QOF score was strongly correlated with practice standardized mortality ratios (r = 0.64), explaining more variation in practice death rates than the Charlson index. CONCLUSION: A simple nine-item score derived from routine primary care recording provides a morbidity index highly predictive of mortality and between practice variation in older UK primary care populations. This has utility in research and health care outcome monitoring and can be easily implemented in other primary and ambulatory care settings.

PURPOSE: Antibiotics are frequently prescribed for acute nonspecific respiratory infections (ARIs), presumably to avoid small risks of progression to serious bacterial illness. However, even low risks of associated adverse drug events could result in many such events at the population level. Our objective was to assess the risks and benefits of antibiotic use in a cohort of patients with ARIs, comparing outcomes of patients who were prescribed antibiotics with outcomes of patients not receiving antibiotics. METHODS: We used a June 1986 to August 2006 cohort of adult patients with ARI visits from a UK primary care database. Exposure was an antibiotic prescribed with the visit. Primary outcomes were hospitalization within 15 days for (1) severe adverse drug events (hypersensitivity, diarrhea, seizure, arrhythmia, hepatic or renal failure), and (2) community-acquired pneumonia. RESULTS: The cohort included 1,531,019 visits with an ARI diagnosis; prescriptions for antibiotics were given in 65% of cases. The adjusted risk difference for treated vs untreated patients per 100,000 visits was 1.07 fewer adverse events (95% CI, -4.52 to 2.38; P = .54) and 8.16 fewer pneumonia hospitalizations (95% CI, -13.24 to -3.08; P = .002). The number needed to treat to prevent 1 hospitalization for pneumonia was 12,255. CONCLUSIONS: Compared with patients with ARI who were not treated with antibiotics, patients who were treated with antibiotics were not at increased risk of severe adverse drug events and had a small decreased risk of pneumonia hospitalization. This small benefit from antibiotics for a common ambulatory diagnosis creates persistent tension; at the societal level, physicians are compelled to reduce antibiotic prescribing, thus minimizing future resistance, whereas at the encounter level, they are compelled to optimize the benefit-risk balance for that patient.

OBJECTIVES: The objectives of this study were to determine the prevalence of PsA in The Health Improvement Network (THIN), a large population-based medical records database in the UK, to examine factors associated with prevalent PsA among patients with psoriasis and to describe the use of DMARDs in patients with PsA. METHODS: Two cohorts were derived from THIN to examine the prevalence of PsA in a cross-sectional study among all patients aged 18-90 years and among a subcohort of 4900 psoriasis patients aged 45-65 years. Prescription codes were used to describe therapies after the diagnosis of PsA. Associations for prevalent PsA among psoriasis patients were assessed using logistic regression analysis. RESULTS: Among 4.8 million patients in THIN between the ages of 18 and 90 years, 9045 patients had at least one medical code for PsA, giving an overall prevalence of 0.19% (95% CI 0.19%, 0.19%). Of those patients, 45.9% with PsA have been prescribed DMARDs. Among the 4064 confirmed psoriasis patients, the prevalence of PsA was 8.6% (95% CI 7.7%, 9.5%). PsA was more prevalent among patients with severe psoriasis [odds ratio (OR) 3.34; 95% CI 2.40, 4.65], obesity (OR 1.77; 95% CI 1.30, 2.41) and duration of psoriasis for ≥10 years (OR 7.42; 95% CI 3.86, 14.25) in the fully adjusted model. CONCLUSION: The prevalence of PsA in THIN is consistent with previous population-based estimates. Limitations include a definition of PsA based on a diagnostic code rather than Classification Criteria for Psoriatic Arthritis (CASPAR) criteria. Given the large population of PsA patients, THIN is an important resource for the study of PsA.

AIM: In the last decade, new glucose lowering drugs (GLDs) have been launched, and also several warnings regarding their safety. The cardiovascular safety of thiazolidinediones (TZD) has been questioned. We analyzed the prescription pattern of GLDs from 2000 to November 2009 in the United Kingdom (UK) using the THIN database with special focus on the effects of the safety warnings about rosiglitazone issued in May 2007 and January 2008. METHODS: Annual prevalence and incidence of GLD prescriptions were measured. For TZD, the monthly prevalence and incidence of prescription were calculated from May 2006 to January 2009. The switching pattern around the FDA alert and the characteristics of subjects starting treatment with TZD before and after the alerts were observed. RESULTS: The prevalence of prescriptions of GLDs increased during the 10 year period, metformin increasing more than three times. Rosiglitazone prevalence showed an increased trend until May 2007, (2.3/1000 person-years) and decreased thereafter (January 2009: 1.1/1000 person-years). The use of pioglitazone increased surpassing rosiglitazone from April 2008 onwards. The incidence of rosiglitazone use decreased sharply after May 2007 (0.8/1000 person-years). The prevalence of use of other therapies remained rather stable from 2000 to 2007 but increased afterwards. After May 2007, rosiglitazone users were increasingly switched to pioglitazone. There was an increased proportion of new users of pioglitazone with cardiovascular risk after the alerts. CONCLUSIONS: The prescription of GLDs in the UK has increased in the last decade. For TZDs, it changed after May 2007 as well as the characteristics of the subjects treated with them.

BACKGROUND: mortality in UK care homes is not well described. OBJECTIVE: to describe 1-year mortality and predictors in older care home residents compared with community residents. METHOD: cohort study using the THIN primary care database with 9,772 care home and 354,306 community residents aged 65-104 years in 293 English and Welsh general practices in 2009. RESULTS: a total of 2,558 (26.2%) care home and 11,602 (3.3%) community residents died within 1 year. The age and sex standardised mortality ratio for nursing homes was 419 (95% CI: 396-442) and for residential homes was 284 (266-302). Age-related increases in mortality were less marked in care homes than community. Comorbidities and identification as inappropriate for chronic disease management targets predicted mortality in both settings, but associations were weaker in care homes. The number of drug classes prescribed and primary care contact were the strongest clinical predictors of mortality in care homes. CONCLUSIONS: older care home residents experience high mortality. Age and diagnostic characteristics are weaker predictors of risk of death within care homes than the community. Measures of primary care utilisation may be useful proxies for frailty and improve difficult end of life care decisions in care homes.

BACKGROUND: There has been limited examination of the contribution of socio-economic factors to the development of leg ulcers, despite the social patterning of many underlying risk factors. No previous studies were found that examined social patterns in the quality of treatment received by patients with leg ulcers. METHODS: Using The Health Improvement Network (THIN) database we identified a cohort of over 14000 patients with a diagnosis of venous leg ulceration, prospectively recorded between the years 2001 and 2006, with linked area-level socio-economic information (Townsend deprivation quintile). We assessed socio-economic differences in the incidence and prevalence of leg ulcers using negative binomial regression. Socio-economic differences in two key areas of guideline recommended leg ulcer management, arterial Doppler assessment and compression bandaging, were assessed using multilevel regression. RESULTS: The risk of incident venous leg ulceration increased for patients living in areas of higher deprivation, even after adjustment for known risk factors age and gender. Overall reported rates of Doppler assessment and provision of compression therapy were low, with less than sixteen per cent of patients having a database record of receiving these recommended diagnostic and treatment options. Patients diagnosed with incident venous leg ulcers living in the most deprived areas were less likely to receive the recommended Doppler-aided assessment for peripheral vascular disease than patients living in the least deprived areas (odds ratio 0.43, 95% confidence interval 0.24-0.78). Documented provision of compression therapy did not vary with deprivation. CONCLUSIONS: A socio-economic gradient in venous leg ulcer disease was observed. The overall rates of people with venous leg ulcers who were documented as receiving guideline recommended care (2001-2006) were low. Reported use of Doppler ultrasound assessment was negatively associated with socio-economic status. These findings suggest that the inequalities experienced by leg ulcer patients may be exacerbated by reduced access to guideline-based management.