Publications
scientifiques

BACKGROUND: Predictive models show promise in healthcare, but their successful deployment is challenging due to limited generalizability. Current external validation often focuses on model performance with restricted feature use from the original training data, lacking insights into their suitability at external sites. Our study introduces an innovative methodology for evaluating features during both the development phase and the validation, focusing on creating and validating predictive models for post-surgery patient outcomes with improved generalizability. METHODS: Electronic health records (EHRs) from 4 countries (United States, United Kingdom, Finland, and Korea) were mapped to the OMOP Common Data Model (CDM), 2008-2019. Machine learning (ML) models were developed to predict post-surgery prolonged opioid use (POU) risks using data collected 6 months before surgery. Both local and cross-site feature selection methods were applied in the development and external validation datasets. Models were developed using Observational Health Data Sciences and Informatics (OHDSI) tools and validated on separate patient cohorts. RESULTS: Model development included 41 929 patients, 14.6% with POU. The external validation included 31 932 (UK), 23 100 (US), 7295 (Korea), and 3934 (Finland) patients with POU of 44.2%, 22.0%, 15.8%, and 21.8%, respectively. The top-performing model, Lasso logistic regression, achieved an area under the receiver operating characteristic curve (AUROC) of 0.75 during local validation and 0.69 (SD = 0.02) (averaged) in external validation. Models trained with cross-site feature selection significantly outperformed those using only features from the development site through external validation (P < .05). CONCLUSIONS: Using EHRs across four countries mapped to the OMOP CDM, we developed generalizable predictive models for POU. Our approach demonstrates the significant impact of cross-site feature selection in improving model performance, underscoring the importance of incorporating diverse feature sets from various clinical settings to enhance the generalizability and utility of predictive healthcare models.

OBJECTIVES: The incidence of gout in the UK appears to have declined since 2013; however, whether such a trend occurred across participants born in different years (ie, birth cohort) is unknown. We aimed to examine the effects of the birth cohort on gout incidence using an age-period-cohort (APC) model. DESIGN: Cross-sectional study. SETTING: Nationwide data from the UK primary care database. PARTICIPANTS: Individuals between 30 and 89 years of age were included. We excluded individuals who had gout history when entering the database and individuals with less than 1 year of continuous follow-up between 1 January 1999 and 31 December 2019. PRIMARY AND SECONDARY OUTCOME MEASURES: Gout was identified using READ codes assigned by general practitioners. The incidence of gout between 1999-2013 and 2011-2019 was analysed with APC model. RESULTS: The incidence of gout between 1999 and 2013 increased with birth cohorts. Compared with those born in 1949-1953 (reference), the age-adjusted and period-adjusted rate ratios (RRs) of incident gout increased from 0.39 (95% CI 0.34 to 0.46) in participants born in 1910-1914 to 2.36 (95% CI 2.09 to 2.66) in participants born in 1979-1983 (p for trend <0.001). In contrast, the incidence of gout between 2011 and 2019 decreased with birth cohorts. Compared with those born in 1949-1953 (reference), the age-adjusted and period-adjusted RRs of incident gout declined from 2.75 (95% CI 2.30 to 3.28) in participants born in 1922-1926 to 0.75 (95% CI 0.65 to 0.87) in participants born in 1976-1980 but then increased slightly to 0.95 (95% CI 0.77 to 1.17) in participants born in 1985-1989. CONCLUSIONS: The gout incidence between 1999 and 2013 in the UK increased with the birth cohorts and then decreased between 2011 and 2019 except for those born after 1980. Future monitoring is needed to help identify aetiological factors and guide preventive and treatment strategies for gout.

Hepatic encephalopathy (HE) is a severe complication of cirrhosis, independently associated with a poor survival. The objectives of this study were to describe the prevalence of overt hepatic encephalopathy (OHE) requiring hospitalization, and the healthcare pathways and outcomes of patients hospitalized for OHE in France. Data from the French Hospital-Discharge Database (Programme de Medicalisation des Systemes d'information, PMSI) within the 5-year period from 2014 to 2018 were analysed. Since the disease lacks a PMSI code in the ICD-10, an identification algorithm was developed. The analysis identified 57,191 patients with OHE including 48,566 patients (85 %) who had been hospitalized twice or more during the study period. Each year, an average of over 20,000 patients were hospitalized in France for OHE as the primary or secondary reason for hospitalization. Among these patients, between 11,500 and 13,500 had been hospitalized at least twice in that year with an average of 3.4 hospitalisations per year. 25 % of admissions occurred following consultation at the emergency unit. Among hospitalisations, 15 % involved admission to the critical care resuscitation unit or intensive care. For all patients identified as suffering from OHE and hospitalized, the 5-year mortality was 46.5 % (26,621 patients). This pioneering study revealed that, in France, despite a probable underestimation of OHE episodes due to the lack of specific PMSI coding, the prevalence of OHE was very high, with frequent recurrences and readmissions, and high mortality.

OBJECTIVE: The current guidelines recommend weight loss for patients with overweight or obesity and knee or hip osteoarthritis (OA); however, there is a paucity of data on the relation of weight loss to death among patients with OA. We aimed to examine the relation of the rate of weight loss induced by antiobesity medications over one year to all-cause mortality among patients with overweight or obesity and knee or hip OA. METHODS: Using the IQVIA Medical Research Database, we identified people with overweight or obesity and knee or hip OA. We emulated analyses of a hypothetical target trial to assess the effect of slow-to-moderate (2%-10%) or fast (≥10%) weight loss induced by the initiation of antiobesity medications within one year on all-cause mortality and secondary outcomes over five years' follow-up. RESULTS: Among 6,524 participants, the five-year all-cause mortality rates were 5.3%, 4.0%, and 5.4% for weight gain or stable, slow-to-moderate weight loss, and fast weight loss arms, respectively. Compared with the weight gain or stable arm, hazard ratios of all-cause mortality were 0.72 (95% confidence interval [CI] 0.56-0.92) for the slow-to-moderate weight loss arm and 0.99 (95% CI 0.67-1.44) for the fast weight loss arm. We found dose-response protective effects of weight loss on incident hypertension, type 2 diabetes, and venous thromboembolism but a slightly higher risk of cardiovascular disease, albeit not statistically significant, in the fast rate of weight loss arm than in the weight gain or stable arm and no significant relations of weight loss to the risk of cancer. CONCLUSION: In this population-based study, a slow-to-moderate, but not fast, rate of weight loss induced by antiobesity medications is associated with a lower risk of all-cause mortality in people with overweight or obesity and knee or hip OA.

BACKGROUND AND OBJECTIVES: Previous studies have reported a possible prodrome in multiple sclerosis (MS) defined by nonspecific symptoms including mood disorder or genitourinary symptoms and increased health care use detected several years before diagnosis. This study aimed to evaluate agnostically the associations between diseases and symptoms diagnosed in primary care and the risk of MS relative to controls and 2 other autoimmune inflammatory diseases with similar population characteristics, namely lupus and Crohn disease (CD). METHODS: A case-control study was conducted using electronic health records from the Health Improvement Network database in the United Kingdom and France. We agnostically assessed the associations between 113 diseases and symptoms in the 5 years before and after diagnosis in patients with subsequent diagnosis of MS. Individuals with a diagnosis of MS were compared with individuals without MS and individuals with 2 other autoimmune diseases, CD and lupus. RESULTS: The study population consisted of patients with MS (n = 20,174), patients without MS (n = 54,790), patients with CD (n = 30,477), and patients with lupus (n = 7,337). Twelve codes were significantly positively associated with the risk of MS compared with controls without MS. After considering codes suggestive of neurologic symptoms as the first diagnosis of MS, 5 codes remained significantly associated with MS: depression (UK: odds ratio 1.22, 95% CI 1.11-1.34), sexual dysfunction (1.47, 1.11-1.95), constipation (1.5, 1.27-1.78), cystitis (1.21, 1.05-1.39), and urinary tract infections of unspecified site (1.38, 1.18-1.61). However, none of these conditions was selectively associated with MS in comparisons with both lupus and CD. All 5 codes identified were still associated with MS during the 5 years after diagnosis. DISCUSSION: We identified 5 health conditions associated with subsequent MS diagnosis, which may be considered not only prodromal but also early-stage symptoms. However, these health conditions overlap with prodrome of 2 other autoimmune diseases; hence, they lack specificity to MS.

The QRISK cardiovascular disease (CVD) risk assessment model is not currently optimized for patients with type 2 diabetes mellitus (T2DM). We aim to identify if the abundantly available repeatedly measured data for patients with T2D improves the predictive capability of QRISK to support the decision-making process regarding CVD prevention in patients with T2DM. We identified patients with T2DM aged 25 to 85, not on statin treatment and without pre-existing CVD from the IQVIA Medical Research Data United Kingdom primary care database and then followed them up until the first diagnosis of CVD, ischemic heart disease, or stroke/transient ischemic attack. We included traditional, nontraditional risk factors and relevant treatments for our analysis. We then undertook a Cox's hazards model accounting for time-dependent covariates to estimate the hazard rates for each risk factor and calculated a 10-year risk score. Models were developed for males and females separately. We tested the performance of our models using validation data and calculated discrimination and calibration statistics. The study included 198,835 (180,143 male with 11,976 outcomes and 90,466 female with 8,258 outcomes) patients. The 10-year predicted survival probabilities for females was 0.87 (0.87 to 0.87), whereas the observed survival estimates from the Kaplan-Meier curve for all female models was 0.87 (0.86 to 0.87). The predicted and observed survival estimates for males were 0.84 (0.84 to 0.84) and 0.84 (0.83 to 0.84) respectively. The Harrell's C-index of all female models and all male models were 0.71 and 0.69 respectively. We found that including time-varying repeated measures, only mildly improved CVD risk prediction for T2DM patients in comparison to the current practice standard. We advocate for further research using time-varying data to identify if the involvement of further covariates may improve the accuracy of currently accepted prediction models.

BACKGROUND: Subclinical hypothyroidism is diagnosed when serum thyroid stimulating hormone levels are higher whilst free thyroxine levels remain within their respective reference ranges. These reference ranges are uniformly applied in all adults, despite serum thyroid stimulating hormone levels naturally increasing with age. Research has found that mildly elevated thyroid stimulating hormone levels may be associated with some benefits in ageing patients, including reduced mortality and better cardiorespiratory fitness. Levothyroxine is typically prescribed to patients with hypothyroidism, but no conclusive evidence exists on whether levothyroxine therapy is beneficial or detrimental in older subclinical hypothyroid patients. Despite this, prescriptions for levothyroxine are increasing year-on-year. This study aims to determine if receiving levothyroxine affects the cardiovascular and bone health outcomes of subclinical patients in primary care aged 50 years and over. METHODS: This project includes a retrospective cohort analysis and a target trial emulation study using electronic patient records collected between 2006 and 2021 and recorded in The Health Improvement Network database. The primary outcome of this study is to compare the cardiovascular outcomes of subclinical hypothyroid patients aged over 50 years treated with levothyroxine compared to those untreated. Secondary outcomes are bone health and all-cause mortality outcomes. Descriptive and inferential statistics will both be employed to analyse the data. Secondary analysis will explore confounding factors, including age, sex, smoking status, body mass index, co-morbidities, and levothyroxine dosage. DISCUSSION: There needs to be a greater understanding of the potential risks of the current treatment for older patients with subclinical hypothyroidism in a primary care setting. We will investigate the clinical importance of this issue and whether older subclinical hypothyroid patients have poorer outcomes when treated. Clarifying this concern may help address the healthcare resource implications of ageing patients being misclassified as having mild hypothyroidism, as these patients are more likely to repeat their blood tests. This could reduce prescription wastage and improve patient outcomes and quality of life in the ageing population. TRIAL REGISTRATION: Not applicable.

BACKGROUND: Previous research has shown that people who have been diagnosed autistic are more likely to die prematurely than the general population. However, statistics on premature mortality in autistic people have often been misinterpreted. In this study we aimed to estimate the life expectancy and years of life lost experienced by autistic people living in the UK. METHODS: We studied people in the IQVIA Medical Research Database with an autism diagnosis between January 1, 1989 and January 16, 2019. For each participant diagnosed autistic, we included ten comparison participants without an autism diagnosis, matched by age, sex, and primary care practice. We calculated age- and sex-standardised mortality ratios comparing people diagnosed autistic to the reference group. We used Poisson regression to estimate age-specific mortality rates, and life tables to estimate life expectancy at age 18 and years of life lost. We analysed the data separately by sex, and for people with and without a record of intellectual disability. We discuss the findings in the light of the prevalence of recorded diagnosis of autism in primary care compared to community estimates. FINDINGS: From a cohort of nearly 10 million people, we identified 17,130 participants diagnosed autistic without an intellectual disability (matched with 171,300 comparison participants), and 6450 participants diagnosed autistic with an intellectual disability (matched with 64,500 comparison participants). The apparent estimates indicated that people diagnosed with autism but not intellectual disability had 1.71 (95% CI: 1.39-2.11) times the mortality rate of people without these diagnoses. People diagnosed with autism and intellectual disability had 2.83 (95% CI: 2.33-3.43) times the mortality rate of people without these diagnoses. Likewise, the apparent reduction in life expectancy for people diagnosed with autism but not intellectual disability was 6.14 years (95% CI: 2.84-9.07) for men and 6.45 years (95% CI: 1.37-11.58 years) for women. The apparent reduction in life expectancy for people diagnosed with autism and intellectual disability was 7.28 years (95% CI: 3.78-10.27) for men and 14.59 years (95% CI: 9.45-19.02 years) for women. However, these findings are likely to be subject to exposure misclassification biases: very few autistic adults and older-adults have been diagnosed, meaning that we could only study a fraction of the total autistic population. Those who have been diagnosed may well be those with greater support needs and more co-occurring health conditions than autistic people on average. INTERPRETATION: The findings indicate that there is a group of autistic people who experience premature mortality, which is of significant concern. There is an urgent need for investigation into the reasons behind this. However, our estimates suggest that the widely reported statistic that autistic people live 16-years less on average is likely incorrect. Nine out of 10 autistic people may have been undiagnosed across the time-period studied. Hence, the results of our study do not generalise to all autistic people. Diagnosed autistic adults, and particularly older adults, are likely those with greater-than-average support needs. Therefore, we may have over-estimated the reduction in life expectancy experienced by autistic people on average. The larger reduction in life expectancy for women diagnosed with autism and intellectual disability vs. men may in part reflect disproportionate underdiagnosis of autism and/or intellectual disability in women. FUNDING: Dunhill Medical Trust, Medical Research Council, National Institute for Health and Care Research, and the Royal College of Psychiatrists.

PURPOSE: To analyze interventions implemented at the time of colorectal cancer (CRC) screening, or among individuals who have previously undergone investigation for CRC, focused on reducing CRC risk through promotion of lifestyle behavior change. Additionally, this review evaluated to what extent such interventions apply behavior change techniques (BCTs) to achieve their objectives. METHODS: Five databases were systematically searched to identify randomized control trials seeking to reduce CRC risk through behavior change. Outcomes were changes in health-related lifestyle behaviors associated with CRC risk, including changes in dietary habits, body mass index, smoking behaviors, alcohol consumption, and physical activity. Standardized mean differences (SMDs) with 95% confidence intervals (CIs) were pooled using random effects models. BCT's were coded from a published taxonomy of 93 techniques. RESULTS: Ten RCT's met the inclusion criteria. Greater increase in fruit/vegetable consumption in the intervention group were observed with respect to the control (SMD 0.13, 95% CI 0.08 to 0.18; p < 0.001). Across fiber, alcohol, fat, red meat, and multivitamin consumption, and smoking behaviors, similar positive outcomes were observed (SMD 0.09-0.57 for all, p < 0.01). However, among physical activity and body mass index, no difference between the intervention groups compared with controls were observed. A median of 7.5 BCTs were applied across included interventions. CONCLUSION: While magnitude of the observed effect sizes varied, they correspond to potentially important changes in lifestyle behaviors when considered on a population scale. Future interventions should identify avenues to maximize long-term engagement to promote sustained lifestyle behavior change.

BACKGROUND: Epidemiological studies suggest chronic and recurrent pain affects around a quarter of children, while 8% report intense and frequent pain. The long-term implications of chronic pain in childhood are uncertain. Using electronic health records (EHRs) we used both disease codes and medicines prescription records to investigate the scale of chronic pain and long-term analgesic use in children and young people (CYP), and if chronic pain and/or use of analgesic medicines at an early age is associated with substance misuse, use of prescription opioids, and poor mental health in adulthood. METHODS: We conducted a cohort study using data from IQVIA Medical Research Data UK. We identified individuals aged 2-24 with exposure to either a diagnostic code indicating chronic pain (diagnosis-exposed), repeat prescription for medicines commonly used to treat pain (prescription-exposed), or both. Follow-up began at 25, and the unexposed population acted as comparators. We calculated hazard ratios (HR) for mental health and substance misuse outcomes, and rate ratios (RR) for opioid prescriptions in adulthood. Additionally, we investigated which diagnoses, if any, were over-represented in the prescription-exposed subgroup. FINDINGS: The cohort constituted 853,625 individuals; 146,431 had one or more of the exposures of interest (diagnosis-exposed = 115,101, prescription-exposed = 20,298, both-exposed = 11,032), leaving 707,194 as comparators. Median age at index exposure was 18.7 years (IQR 14.7-22.3). On average during follow-up, the pooled exposed group had, respectively, a 31% and 17% higher risk of adverse mental health and substance misuse outcomes (adjusted HR [95% CI] of 1.31 [1.29-1.32] and 1.17 [1.11-1.24]). Exposed individuals also received prescription opioids at double the rate of unexposed individuals on average during follow-up (adjusted RR 2.01 [95% CI 1.95-2.10]). Outcomes varied between exposure subgroups, with prescription- and both-exposure tending to have worse outcomes. Unlike these two subgroups, in the diagnosis-exposed subgroup we did not detect a greater risk of substance misuse. INTERPRETATION: Chronic pain in CYP is associated with increased prescription opioid use and adverse mental health outcomes in adulthood, as is repeat prescription for analgesic medicines, but only the latter is also associated with substance misuse in adulthood. It is essential to avoid the harms of under-treating pain in CYP while giving due consideration to the risks posed by analgesic medicines. Early recognition of chronic pain in CYP and utilising non-pharmacological management options may help minimise overprescribing, and long-term reliance on dependence-forming-drugs. FUNDING: AL is an NIHR funded academic clinical fellow, and was supported by funding from UCLH Charities while carrying out this work. RS and DS are part of the Advanced Pain Discovery Platform and were supported by a UKRI and Versus Arthritis grant (MR/W002566/1) as part of the Consortium Against Pain Inequality. AW was supported by the Wellcome Trust (220558/Z/20/Z).