OBJECTIVE: to evaluate the application of clinical pathway simulation in machine learning, using clinical audit data, in order to identify key drivers for improving use and speed of thrombolysis at individual hospitals. DESIGN: Computer simulation modelling and machine learning. SETTING: Seven acute stroke units. PARTICIPANTS: Anonymised clinical audit data for 7864 patients. RESULTS: Three factors were pivotal in governing thrombolysis use: (1) the proportion of patients with a known stroke onset time (range 44%-73%), (2) pathway speed (for patients arriving within 4 hours of onset: per-hospital median arrival-to-scan ranged from 11 to 56 min; median scan-to-thrombolysis ranged from 21 to 44 min) and (3) predisposition to use thrombolysis (thrombolysis use ranged from 31% to 52% for patients with stroke scanned with 30 min left to administer thrombolysis). A pathway simulation model could predict the potential benefit of improving individual stages of the clinical pathway speed, whereas a machine learning model could predict the benefit of 'exporting' clinical decision making from one hospital to another, while allowing for differences in patient population between hospitals. By applying pathway simulation and machine learning together, we found a realistic ceiling of 15%-25% use of thrombolysis across different hospitals and, in the seven hospitals studied, a realistic opportunity to double the number of patients with no significant disability that may be attributed to thrombolysis. CONCLUSIONS: National clinical audit may be enhanced by a combination of pathway simulation and machine learning, which best allows for an understanding of key levers for improvement in hyperacute stroke pathways, allowing for differences between local patient populations. These models, based on standard clinical audit data, may be applied at scale while providing results at individual hospital level. The models facilitate understanding of variation and levers for improvement in stroke pathways, and help set realistic targets tailored to local populations.

BACKGROUND: the prevalence of viral hepatitis (hepatitis B virus and hepatitis C virus) in migrants is higher than among the general population in many high-income countries. We aimed to determine whether incentivising and supporting primary-care physicians in areas with a high density of migrants increases the numbers of adult migrants screened for viral hepatitis. METHODS: HepFREE was a multicentre, open, cluster-randomised controlled trial in general practices in areas of the UK with a high density of migrants (Bradford, Yorkshire, and northeast and southeast London). Participants were adult patients (aged 18 years or older) in primary care, who had been identified as a first or second generation migrant from a high-risk country. General practices were randomly assigned (1:2:2:2:2) to an opportunistic screening (control) group or to one of four targeted screening (interventional) groups: standard (ie, hospital-based) care and a standard invitation letter; standard care and an enhanced invitation letter; community care and a standard invitation letter; or community care and an enhanced invitation letter. In control screening, general practitioners (GPs) were given a teaching session on viral hepatitis and were asked to test all registered migrants. In the intervention, GPs were paid a nominal sum for setting up searches of records, reimbursed for signed consent forms, and supported by a dedicated clinician. Patients who were eligible for testing and tested positive for viral hepatitis in the intervention groups were eligible to enrol in a second embedded trial of community versus hospital based care. The primary outcomes were the proportion of patients eligible for screening, the proportion of those eligible who were sent an invitation letter in the intervention groups, the uptake of viral hepatitis screening (in the intention-to-treat population), the proportion of patients who tested positive for viral hepatitis, the proportion who complied with treatment, and the cost-effectiveness of the intervention. This trial is registered with ISRCTN, number ISRCTN54828633. FINDINGS: Recruitment and testing ran from Oct 31, 2013, to Feb 4, 2017, and each practice recruited for 18 consecutive calendar months. We approached 70 general practices in three areas with a high density of migrants, of which 63 general practices agreed to participate. Five practices withdrew and 58 practices were randomly assigned: eight to control and 50 to an intervention. In control practices, 26 046 (38·4%) of 67 820 patients who were initially registered were eligible for testing, as were 152 321 (43·3%) of 351 710 patients in the interventional groups in London and Bradford. of 51 773 randomly selected eligible patients in the intervention groups in London and Bradford, letters were sent to 43 585 (84·2%) patients. In the eight control general practices, screening was taken up by 543 (1·7%) of 31 738 eligible participants, which included 5692 newly registered patients. However, in the 50 general practices that used the intervention, screening was taken up by 11 386 (19·5%) of 58 512 eligible participants (including 6739 newly registered patients; incidence rate ratio 3·70, 95% CI 1·30-10·51; p=0·014) and this intervention was cost-effective. 720 (4·5%) of 15 844 patients who received a standard letter versus 1032 (3·7%) of 28 095 patients who received the enhanced letter were tested (0·70, 0·38-1·31; p=0·26). In the control group, 17 patients tested positive for viral hepatitis, as did 220 patients (one with a co-infection) in the intervention groups. In the embedded study, 220 patients were randomly assigned to either hospital-based care or community care; 80 (87·9%) of 91 patients in the hospital setting complied with treatment versus 105 (81·4%) of 129 patients in the community setting. The intervention was cost-effective at willingness to pay thresholds in excess of £8540. One serious adverse event (thyroiditis) was noted. INTERPRETATION: Screening migrants for viral hepatitis in primary care is effective if doctors are incentivised and supported. Community care is expensive and there is no evidence that this offers benefits in this setting or that bespoke invitation letters add value. We suggest that bespoke invitation letters should not be used, and we suggest that outreach, community-based services for migrants should not be developed. FUNDING: National Institute for Health Research.

Introduction: Demonstration of cost-effectiveness is an established hurdle for new treatments and technologies. However, evidence synthesis, including simulation modelling, can be very difficult in the absence of good quality research that addresses pertinent questions, and people with rare conditions may have to forego the best available treatments. We illustrate this point with regard to the current choice between intravitreal ranibizumab, verteporfin photodynamic therapy (VPDT) or combinations of these (combination therapy) for polypoidal choroidal vasculopathy. Methods: We developed a Markov model to simulate equivalent cohorts of 65-year-old patients over a lifetime horizon. We obtained costs from the NHS national tariff, and utility values based on unilateral visual function deterioration. We carried out deterministic and probabilistic sensitivity analyses to investigate the sensitivity of the results to uncertainty in the model parameters. Results: Our model predicts that both VPDT and combination therapy offer cost savings but lower clinical efficacy over a lifetime horizon at δ81 165 and δ14 826 per quality-adjusted life year (QALY) respectively. VPDT monotherapy has a 99% chance of cost-effectiveness at a willingness to pay of δ30 000 per QALY gained. Combination therapy has a low (29%) probability of cost-effectiveness, however, this is heavily dependent on the modelled incidence of haemorrhagic adverse events. Conclusion: Based on the results of our model, VPDT might be commissioned according to a strict decision rule interpretation. The outcome regarding combination therapy is uncertain. These conclusions are dependent on the available evidence. There is considerable modelling and parameter uncertainty, which needs to be urgently addressed.

BACKGROUND: Keratoconus is a progressive degenerative corneal disorder of children and young adults that is traditionally managed by refractive error correction, with corneal transplantation reserved for the most severe cases. UVA collagen crosslinking is a novel procedure that aims to prevent disease progression, currently being considered for use in the UK NHS. We assess whether it might be a cost-effective alternative to standard management for patients with progressive keratoconus. METHODS: We constructed a Markov model in which we estimated disease progression from prospective follow-up studies, derived costs derived from the NHS National Tariff, and calculated utilities from linear regression models of visual acuity in the better-seeing eye. We performed deterministic and probabilistic sensitivity analyses to assess the impact of possible variations in the model parameters. RESULTS: Collagen crosslinking is cost effective compared with standard management at an incremental cost of £ 3174 per QALY in the base case. Deterministic sensitivity analysis shows that this could rise above £ 33,263 per QALY if the duration of treatment efficacy is limited to 5 years. Other model parameters are not decision significant. Collagen crosslinking is cost effective in 85% of simulations at a willingness-to-pay threshold of £ 30,000 per QALY. CONCLUSION: UVA collagen crosslinking is very likely to be cost effective, compared with standard management, for the treatment of progressive keratoconus. However, further research to explore its efficacy beyond 5 years is desirable.