Investigating prescribing patterns for as-yet-undiagnosed cancer symptoms and their utility in managing diagnostic uncertainty in primary care

早期发现癌症可以极大地改善癌症患者的预后[1]。全科医生(GP)是对有症状的患者进行进一步调查的看门人。在初级保健中，癌症的许多早期症状是非特异性的，很难与其他疾病的症状区分开来，这些症状中的大多数不需要专家诊断评估。当患者出现低风险(但不是无风险)症状时，全科医生需要对这些患者进行分类，以确定哪些患者需要进一步调查。最佳实践安全网指南旨在确保患者得到跟踪，直到症状得到解释或解决[2]。然而，全科医生访谈显示，需要在有助于有效安全网的临床实践方面提供更明确的指导[3]。这将减少与初级保健诊断错误有关的不充分的安全网实施[4]。了解其他临床诊断前事件如何影响全科医生的安全网、患者寻求帮助的行为，以及这些因素之间的相互作用，有助于澄清有效的安全网实践。研究将全科医生与患者的对话定义为患者自我管理行为和健康结果的关键决定因素[5]。确定积极促进适当患者求助行为的做法是有必要的，因为许多基于全科医生与患者互动的定性证据不包括可能导致初级保健诊断错误的患者行为[6]。诊断前的患者医疗保健利用模式显示出更高的诊断活跃度和在诊断前长达一年的全科医生会诊数量增加。这些模式表明，一些患者存在6至4个月的早期诊断机会之窗。患者通常在初级保健咨询后收到处方，在那里决定不转诊。了解如何使用处方来管理低风险症状可能是识别早期诊断机会之窗的有用方式。来自丹麦的新研究观察到，在肺癌诊断之前，COPD药物、抗生素和阿片类药物的处方模式增加，在结直肠癌或结肠癌诊断之前，痔疮药物、泻药、口服铁和阿片类药物的处方模式增加。研究还开始区分与处方药和治疗尚未确诊的癌症症状的处方药有关的癌症风险。在英国，最近国家癌症登记数据与国家处方数据的联系开启了进行类似研究的可能性。此外，自2015年以来，CPRD是一个非常成熟的研究资源，包括来自英国诊所患者电子健康记录的匿名信息(包括处方信息)，也与癌症登记数据有关。机器学习方法已经确定了肺癌和结直肠癌诊断前不同处方模式的统计证据。需要对其他肿瘤组的处方模式进行分析，以帮助确定整个疾病组的早期诊断机会窗口。建议研究的宗旨是，通过了解全科医生处方决定和全科医生安全网实践之间的相互作用，我们可以提高癌症诊断的速度。

Early cancer detection can vastly improve the prognosis of cancer patients [1]. General practitioners (GPs) serve as gatekeeper for further investigation in symptomatic patients. In primary care, many early symptoms of cancer are non-specific and hard to differentiate from symptoms of other conditions, most of which do not require specialist diagnostic assessment. When patients present with low risk (but not no risk) symptoms, GPs need to triage these patients to identify those needing further investigation. Best practice safety-netting guidelines aim to ensure patients are followed up until symptoms are explained or resolved [2]. However GP interviews show that clearer guidance on the aspects of clinical practice that contribute towards effective safety-netting is needed [3]. This would reduce inadequate safety-netting implementation which is linked to diagnostic error in primary care [4]. Understanding how other clinical pre-diagnostic events influence how GPs safety-net, patient help-seeking behaviours, and the interplay between these factors could help clarify effective safety-netting practice. Research defines the GP-patient conversation as a key determinant of patients' self-management behaviour and health outcomes [5]. Identifying practices that actively promote appropriate patient help-seeking behaviours is warranted as much of the qualitative evidence base on GP-patient interactions doesn't include potential patient behaviours that could lead to diagnostic error in primary care [6]. Patient healthcare utilisation patterns prior to diagnosis show higher diagnostic activity and increasing numbers of GP consultations for up to a year before diagnosis. These patterns indicate a window of opportunity for earlier diagnosis exists in some patients, ranging between 6 to 4 months. Patients often receive prescriptions after consulting in primary care where a decision was taken not to refer. Understanding how prescriptions are used to manage low-risk symptoms may be a useful way to identify windows of opportunity for earlier diagnosis. Emerging research from Denmark has observed increased prescribing patterns of COPD drugs, antibiotics and opioids prior to lung cancer diagnosis, and haemorrhoid drugs, laxatives, oral iron and opioids prior to colorectal or colon cancer diagnosis. Research has also begun discriminate cancer risk associated with prescription drugs, and drugs prescribed to treat symptoms of not-yet-diagnosed cancer. In the UK, recent linkage of national cancer registration data with national prescriptions data opened the possibility of conducting similar studies. Additionally, since 2015 CPRD, a very well-established research resource including anonymised information from patient electronic health records from UK practices (including information on prescription) is also linked to cancer registration data. Machine learning approaches have identified statistical evidence of differential prescribing patterns in lung and colorectal cancers pre-diagnosis. Analysis of prescribing patterns for additional tumour groups is required to help identify windows of opportunity for earlier diagnosis across the disease group. The tenet of the proposed studies is that by understanding the interplay between GP prescribing decisions and GP safety-netting practices we can improve the speed of cancer diagnosis.