Hilde Philips

Background: Out-of-hours primary care (OOH-PC) has emerged as a promising solution to improve efficiency, accessibility, and quality of care and to reduce the strain on emergency departments. As this modality gains traction in diverse healthcare settings, it is increasingly important to fully assess its societal value-for-money and conduct thorough process evaluations. However, current economic evaluations mostly emphasise direct- and short-term effect measures, thus lacking a broader societal perspective. Aim: This study offers a comprehensive overview of current effect measures in OOH-PC evaluations and proposes additional measures from the evaluation of integrated care programmes. Approach and Development: First, we systematically identified the effect measures from published cost-effectiveness studies and classified them as process, outcome, and resource use measures. Second, we elaborate on the incorporation of ‘productivity gains’, ‘health promotion and early intervention’, and ‘continuity of care’ as additional effects into economic evaluations of OOH-PC. Seeking care affects personal and employee time, potentially resulting in decreased productivity. Challenges in taking time off work and limited access to convenient care are often cited as barriers to accessing primary care. As such, OOH-PC can potentially reduce opportunity costs for patients. Furthermore, improving access to healthcare is important in determining whether people receive promotional and preventive services. Health promotion involves empowering people to take control of their health and its determinants. Given the unscheduled nature and the fragmented or rotational care in OOH-PC, the degree to which interventions and modalities provide continuity should be monitored, assessed, and included in economic evaluations. Continuity of care in primary care improves patient satisfaction, promotes adherence to medical advice, reduces reliance on hospitals, and reduces mortality. Conclusion: Although it is essential to also address local settings and needs, the integration of broader scope measures into OOH-PC economic evaluations improves the comprehensive evaluation that aligns with welfare gains.

Background: During out-of-hours (OOH) primary care, GPs overprescribe antibiotics for respiratory tract infections (RTIs). Many interventions have been shown to improve antibiotic prescribing quality, but their implementation in practice remains difficult. Participatory action research (PAR) aims to explore, implement and evaluate change in practice with an active involvement of local stakeholders, while generating knowledge through experience. Objectives: To evaluate whether PAR improves antibiotic prescribing quality for RTIs during OOH primary care and simultaneously identify the pivotal lessons learned. Methods: A mixed-methods study with a PAR approach in three OOH GP cooperatives (GPCs). Each GPC co-created a multifaceted intervention focusing on improving antibiotic use for RTIs through plan-do-study-act cycles. We quantified antibiotic prescribing quality indicators and formulated the lessons learned from a qualitative process analysis. Results: Interventions were chosen with the GPs and adapted to be context-relevant. The willingness to work on quality and engagement of local stakeholders led to ownership of the project, but was time-consuming. In one GPC, antibiotic prescribing significantly decreased for tonsillitis, bronchitis, otitis media and acute upper RTI. In all three GPCs, use of guideline-recommended antibiotics for otitis media significantly increased. Conclusions: Implementing multifaceted interventions through PAR can lower total and increase guideline-recommended antibiotic prescribing for RTIs in OOH primary care. Co-creating interventions with GPs to suit local needs is feasible, but reaching all GPs targeted is challenging.

Objectives In the TRIAGE trial, a cluster randomised trial about diverting emergency department (ED) patients to a general practice cooperative (GPC) using a new extension to the Manchester Triage System, the difference in the proportion of patients assigned to the GPC was striking: 13.3% in the intervention group (patients were encouraged to comply to an ED or GPC assignment, real-world setting) and 24.7% in the control group (the assignment was not communicated, all remained at the ED, simulated setting). In this secondary analysis, we assess the differences in the use of the triage tool between intervention and control group and differences in costs and hospitalisations for patients assigned to the GPC. Setting ED of a general hospital and the adjacent GPC. Participants 8038 patients (6294 intervention and 1744 control). Primary and secondary outcome measures proportion of patients with triage parameters (reason for encounter, discriminator and urgency category) leading to an assignment to the ED, proportion of patients for which the computer-generated GPC assignment was overruled, motivations for choosing certain parameters, costs (invoices) and hospitalisations. Results An additional 3.1% (pConclusion Nurses used a triage tool more risk averse when it was used to divert patients to primary care as compared with a theoretical assignment to primary care. Outcomes from a simulated setting should not be extrapolated to real patients. Trial registration number NCT03793972.

This study aims to compare trends in antibiotic prescribing behaviour for lower urinary tract infections among different out-of-hours primary care services.Cross-sectional study using routine presc...

General practitioners (GPs) use safety netting advice to communicate with patients when and how to seek further help when their condition fails to improve or deteriorate. Although many respiratory ...

Background: Electronic medical records have opened opportunities to analyze clinical practice at large scale. Structured registries and coding procedures such as the International Classification of Primary Care further improved these procedures. However, a large part of the information about the state of patient and the doctors’ observations is still entered in free text fields. The main function of those fields is to report the doctor’s line of thought, to remind oneself and his or her colleagues on follow-up actions, and to be accountable for clinical decisions. These fields contain rich information that can be complementary to that in coded fields, and until now, they have been hardly used for analysis. Objective: This study aims to develop a prediction model to convert the free text information on COVID-19–related symptoms from out of hours care electronic medical records into usable symptom-based data that can be analyzed at large scale. Methods: The design was a feasibility study in which we examined the content of the raw data, steps and methods for modelling, as well as the precision and accuracy of the models. A data prediction model for 27 preidentified COVID-19–relevant symptoms was developed for a data set derived from the database of primary-care out-of-hours consultations in Flanders. A multiclass, multilabel categorization classifier was developed. We tested two approaches, which were (1) a classical machine learning–based text categorization approach, Binary Relevance, and (2) a deep neural network learning approach with BERTje, including a domain-adapted version. Ethical approval was acquired through the Institutional Review Board of the Institute of Tropical Medicine and the ethics committee of the University Hospital of Antwerpen (ref 20/50/693). Results: The sample set comprised 3957 fields. After cleaning, 2313 could be used for the experiments. Of the 2313 fields, 85% (n=1966) were used to train the model, and 15% (n=347) for testing. The normal BERTje model performed the best on the data. It reached a weighted F1 score of 0.70 and an exact match ratio or accuracy score of 0.38, indicating the instances for which the model has identified all correct codes. The other models achieved respectable results as well, ranging from 0.59 to 0.70 weighted F1. The Binary Relevance method performed the best on the data without a frequency threshold. As for the individual codes, the domain-adapted version of BERTje performs better on several of the less common objective codes, while BERTje reaches higher F1 scores for the least common labels especially, and for most other codes in general. Conclusions: The artificial intelligence model BERTje can reliably predict COVID-19–related information from medical records using text mining from the free text fields generated in primary care settings. This feasibility study invites researchers to examine further possibilities to use primary care routine data.

BACKGROUND Electronic medical records have opened opportunities to analyze clinical practice at large scale. Structured registries and coding procedures such as the International Classification of Primary Care further improved these procedures. However, a large part of the information about the state of patient and the doctors’ observations is still entered in free text fields. The main function of those fields is to report the doctor’s line of thought, to remind oneself and his or her colleagues on follow-up actions, and to be accountable for clinical decisions. These fields contain rich information that can be complementary to that in coded fields, and until now, they have been hardly used for analysis. OBJECTIVE This study aims to develop a prediction model to convert the free text information on COVID-19–related symptoms from out of hours care electronic medical records into usable symptom-based data that can be analyzed at large scale. METHODS The design was a feasibility study in which we examined the content of the raw data, steps and methods for modelling, as well as the precision and accuracy of the models. A data prediction model for 27 preidentified COVID-19–relevant symptoms was developed for a data set derived from the database of primary-care out-of-hours consultations in Flanders. A multiclass, multilabel categorization classifier was developed. We tested two approaches, which were (1) a classical machine learning–based text categorization approach, Binary Relevance, and (2) a deep neural network learning approach with BERTje, including a domain-adapted version. Ethical approval was acquired through the Institutional Review Board of the Institute of Tropical Medicine and the ethics committee of the University Hospital of Antwerpen (ref 20/50/693). RESULTS The sample set comprised 3957 fields. After cleaning, 2313 could be used for the experiments. Of the 2313 fields, 85% (n=1966) were used to train the model, and 15% (n=347) for testing. The normal BERTje model performed the best on the data. It reached a weighted F1 score of 0.70 and an exact match ratio or accuracy score of 0.38, indicating the instances for which the model has identified all correct codes. The other models achieved respectable results as well, ranging from 0.59 to 0.70 weighted F1. The Binary Relevance method performed the best on the data without a frequency threshold. As for the individual codes, the domain-adapted version of BERTje performs better on several of the less common objective codes, while BERTje reaches higher F1 scores for the least common labels especially, and for most other codes in general. CONCLUSIONS The artificial intelligence model BERTje can reliably predict COVID-19–related information from medical records using text mining from the free text fields generated in primary care settings. This feasibility study invites researchers to examine further possibilities to use primary care routine data.

Antibiotic overprescribing is one of the main drivers of the global and growing problem of antibiotic resistance, especially in primary care and for respiratory tract infections (RTIs). RTIs are the most common reason for patients to consult out-of-hours (OOH) primary care. The COVID-19 pandemic has changed the way general practitioners (GPs) work, both during office hours and OOH. In Belgian OOH primary care, remote consultations with the possibility of issuing prescriptions and telephone triage were implemented. We aimed to describe the impact of COVID-19 on GPs’ antibiotic prescribing during OOH primary care. In an observational study, using routinely collected health data from GP cooperatives (GPCs) in Flanders, we analyzed GPs’ antibiotic prescriptions in 2019 (10 GPCs) and 2020 (20 GPCs) during OOH consultations (telephone and face-to-face). We used autoregressive integrated moving average (ARIMA) modeling to identify any changes after lockdowns were implemented. In total, 388,293 contacts and 268,430 prescriptions were analyzed in detail. The number of antibiotic prescriptions per weekend, per 100,000 population was 11.47 (95% CI: 9.08–13.87) or 42.9% lower after compared to before the implementation of lockdown among all contacts. For antibiotic prescribing per contact, we found a decrease of 12.2 percentage points (95% CI: 10.6–13.7) or 56.5% among all contacts and of 5.3 percentage points (95% CI: 3.7–6.9) or 23.2% for face-to-face contacts only. The decrease in the number of prescriptions was more pronounced for cases with respiratory symptoms that corresponded with symptoms of COVID-19 and for antibiotics that are frequently prescribed for RTIs, such as amoxicillin (a decrease of 64.9%) and amoxicillin/clavulanate (a decrease of 38.1%) but did not appear for others such as nitrofurantoin. The implementation of COVID-19 lockdown measures coincided with an unprecedented drop in the number of antibiotic prescriptions, which can be explained by a decrease in face-to-face patient contacts, as well as a lower number of antibiotics prescriptions per face-to-face patient contact. The decrease was seen for antibiotics used for RTIs but not for nitrofurantoin, the first-choice antibiotic for urinary tract infections.

Objectives To determine the effectiveness and safety of a tool diverting low urgency patients eligible for primary care from an emergency department (ED) to the adjacent general practitioner cooperative (GPC). Methods Unblinded, randomised controlled trial with weekends serving as clusters (three intervention clusters for each control). The intervention was nurse-led triage using a new tool assigning patients to either ED or GPC. During intervention weekends, patients were encouraged to follow this assignment while it was not communicated to the patients during control weekends (they remained at the ED). The primary outcome was the proportion of patients assigned to and handled by the GPC during intervention weekends. The trial was randomised for the secondary outcome: the proportion of patients assigned to the GPC during intervention and control weekends. Additional outcomes were association of these outcomes with possible confounders (study tool parameters, nurse, and patient characteristics), proportion of patients referred back to the ED by the GPC, hospitalisations, and performance of the study tool to detect primary care eligible patients (with the opinion of the treating physician as the gold standard). Results In the intervention group, 838/6374 patients (13.3%, 95% CI 12.5 to 14.2) were assigned to the GPC (secondary outcome), in the control group 431/1744 (24.7%, 95% CI 22.7 to 26.8). In the intervention group, 599/6374 patients (9.5%, 95% CI 8.8 to 10.3) experienced the primary outcome which was influenced by the chosen MTS presentational flowchart, patient’s age, and the nurse. 24/599 patients (4.0%, 95% CI 2.7 to 5.9) patients were referred back to the ED of which three were hospitalised. Positive and negative predictive values of the studied tool during intervention weekends were 0.96 (95%CI 0.94 to 0.97) and 0.60 (95% CI 0.58 to 0.62). Out of the patients assigned to the GPC, 2.4% (95% CI 1.7 to 3.4) were hospitalised. Conclusions ED nurses using a new tool safely diverted 9.5% of the included patients to primary care. ClinicalTrials.gov Identifier NCT03793972 Funding Research Foundation – Flanders (FWO)

Patients in Belgium needing out-of-hours medical care have two options: the emergency department (ED) or a general practitioner (GP) on call. Currently, there is no triage system in Belgium, so pat...