Review 1: "Antibiotic Prescribing in Remote Versus Face-to-Face Consultations for Acute Respiratory Infections in English Primary Care: An Observational Study Using TMLE"

Guanbo Wang

doi:doi:10.1162/2e3983f5.12751734

Antibiotic prescribing in remote versus face-to-face consultations for acute respiratory infections in English primary care: An observational study using TMLE

by Emma Vestesson, Kaat De Corte, Paul Chappell, Elizabeth Crellin, and Geraldine M. Clarke

Show Description
Published on Mar 20, 2023
www.medrxiv.org

Description

Abstract Background The COVID-19 pandemic has led to an ongoing increase in the use of remote consultations in general practice in England. Though the evidence is limited, there are concerns that the increase in remote consultations could lead to more antibiotic prescribing.Methods We used patient-level primary care data from the Clinical Practice Research Datalink to estimate the association between consultation mode (remote vs face-to-face) and antibiotic prescribing in England for acute respiratory infections (ARI) between April 2021 – March 2022. We used targeted maximum likelihood estimation, a causal machine learning method with adjustment for patient-, clinician- and practice-level factors.Findings There were 45,997 ARI consultations (34,555 unique patients), of which 28,127 were remote and 17,870 face-to-face. For children, 48% of consultations were remote whereas for adults 66% were remote. For children, 42% of remote and 43% face-to-face consultations led to an antibiotic prescription; the equivalent in adults was 52% of remote and 42% face-to-face. Adults with a remote consultation had 23% (Odds Ratio (OR) 1.23 95% Confidence Interval (CI): 1.18-1.29) higher chance of being prescribed antibiotics compared to if they had been seen face-to-face. We found no significant association between consultation mode and antibiotic prescribing in children (OR 1·04 95% CI 0·98-1·11).Interpretation This study uses rich patient-level data and robust statistical methods and represents an important contribution to the evidence base on antibiotic prescribing in post-COVID primary care. The higher rates of antibiotic prescribing in remote consultations for adults are cause for concern. We see no significant difference in antibiotic prescribing between consultation mode for children. These findings should inform antimicrobial stewardship activities for health care professionals and policy makers. Future research should examine differences in guideline-compliance between remote and face-to-face consultations to understand the factors driving antibiotic prescribing in different consultation modes.Funding No external funding.Research in context Evidence before this study Use of remote consultations in general practice has increased rapidly since the onset of the COVID-19 pandemic. Concerns have been raised that antibiotic prescribing rates may be higher in remote compared with face-to-face consultations. Acute respiratory infection (ARI) is the most common reason for an antibiotic prescription in adults making it one of the most important areas of prescription practice for antibiotic use. Empirical studies investigating the differences in antibiotic prescribing rates between online and remote consultations have produced mixed findings, in general and for ARIs specifically. Recent review-type articles on the topic - including a 2020 qualitative systematic review and a 2021 meta-analytic systematic review – have reported mixed results when comparing online and face-to-face consultations with some showing higher and others lower antibiotic prescribing in remote consultations. Furthermore, many of the studies that were included in the reviews were at risk of bias due to a failure to control for demographic and clinical differences between patients in remote versus face-to-face consultations.Added value of this study This is the first England wide study estimating the difference in antibiotic prescribing between consultations modes in the post-covid setting where remote consultations are as common as face-to-face consultations. It is also the first study in this setting to apply TMLE – doubly robust causal machine learning method. We found that an adult was 23% more likely to be prescribed an antibiotic for an ARI in a remote compared with a face-to-face consultation with a general practitioner in England. There was no evidence for a difference in children. Our findings are based on an analysis of a representative sample of almost 46,000 GP consultations for ARIs in general practice in England and controls for patient-, clinician- and practice-level factors that are associated with both consultation mode and with antibiotic prescribing. As such, our findings are at a smaller risk of bias from unobserved confounding than the previous research examining this issue and therefore represent an important contribution to the evidence base.Implications of the available evidence Taken together with the existing body of evidence on this topic, our results showing higher prescribing in remote consultations are cause for concern. The factors affecting antibiotic prescribing and the interaction with consultation mode are complex and will require further research to unpick. The existing evidence including this study have largely focused on prescribing rates, and do not investigate the appropriateness of antibiotics prescribing in remote compared to face-to-face consultations. Further investigation is required to explain the discrepancy between consultation modes. The growing body of evidence in this area has relevance for future antimicrobial stewardship activities and should be used to inform the ongoing development of antibiotic prescribing guidelines for remote consultations.

RR:C19 Evidence Scale rating by reviewer:

Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.

***************************************

Review:

The pre-print entitled “Antibiotic prescribing in remote versus face-to-face consultations for acute respiratory infections in English primary care: An observational study using TMLE ” by Vestesson et al. raised concern for the remote consultations for acute respiratory infections (ARI), in the sense that clinicians may tend to prescribe unnecessary antibiotics to adults, as compared to face-to-face consultations. If it is true, then policymakers ought to take that into account when promoting and regulating “telemedicine” because antibiotic resistance can threaten disease/epidemic development and management. However, certain flaws may exist in the study.

First, the conclusions made in the article are about the associations between remote consultations and the prescription rate. However, the TMLE is a method for estimating causal effects. The imprecise language may be misleading.

Second, different regions should be considered as different clusters [1]. Taking that information into consideration may remove part of the selection bias. This can be done, for example, by adjusting for the probability of being in a specific region given selected covariates [2].

Third, the current study is a subgroup analysis, stratifying the population by adults and children and estimating the effects separately. Considering the age group as an effect modifier and estimating the subgroup effects from the whole population may increase the power and reduce the variability of the estimation [3]. It would also be beneficial to investigate infection type and antibiotic type as effect modifiers [4].

Fourth, using each consultation rather than each individual as observations may violate the stable unit treatment value assumption. This is because the consultations from the same individual may be correlated. For example, a patient may exaggerate their symptoms to get the prescription in each consultation.

Fifth, the study may miss some critical confounders, which makes the no unmeasured confounding assumption unlikely to hold, though the reviewer acknowledges this may be due to the data collection mechanism.

Sixth, to reach the optimal efficiency and keep the estimator unbiased, the covariates in the outcome should include all the covariates that are associated with the outcome, and the covariates in the propensity score model should include not include the covariates that are only associated with the treatment (in this case, having a remote consultation) [5]. It seems that a same set of covariates were used in both models. In addition, the authors stated, “We also adjusted for variables that were identified by experts to be associated with either antibiotic prescribing or having a remote consultation, instrumental variable”, which is unclear if the instrumental variables were included for adjustment.

Seventh, since the covariates used for adjusting for confounding were not entirely sure to be confounders, it may be beneficial to conduct variable selection in the context of causal inference [6]. It is also possible to incorporate the structures among covariates into such variable selection [7, 8].

Overall, this work utilized a comprehensive dataset to investigate the potential relationship between remote consultation and antibiotic prescription rates. Concerns about escalating antimicrobial resistance should be considered in the decision-making about the remote consultation. However, a more appropriate study design and methods implementation should be employed.

References:

[1] Wang, G., Schnitzer, M.E., Menzies, D., Viiklepp, P., Holtz, T.H., Benedetti, A.: Estimating treatment importance in multidrug-resistant tuberculosis using targeted learning: An observational individual patient data network meta-analysis. Biometrics 76(3), 1007–1016 (2020)

[2] Dahabreh, I.J., Robertson, S.E., Petito, L.C., HernÅLan, M.A., Steingrimsson, J.A.: Efficient and robust methods for causally interpretable meta-analysis: Transporting inferences from multiple randomized trials to a target population. Biometrics (2019)

[3] Liu, Y., Schnitzer, M.E., Wang, G., Kennedy, E., Viiklepp, P., Vargas, M.H., Sotgiu, G., Menzies, D., Benedetti, A.: Modeling treatment effect modification in multidrug-resistant tuberculosis in an individual patient data meta-analysis. Statistical methods in medical research 31(4), 689–705 (2022)

[4] Siddique, A.A., Schnitzer, M.E., Bahamyirou, A., Wang, G., Holtz, T.H., Migliori, G.B., Sotgiu, G., Gandhi, N.R., Vargas, M.H., Menzies, D., et al.: Causal inference with multiple concurrent medications: A comparison of methods and an application in multidrug-resistant tuberculosis. Statistical methods in medical research 28(12), 3534–3549 (2019)

[5] Efficient adjustment sets for population average causal treatment effect estimation in graphical models. Journal of Machine Learning Research 21(188), 1–86 (2020)

[6] Shortreed, S.M., Ertefaie, A.: Outcome-adaptive lasso: variable selection for causal inference. Biometrics 73(4), 1111–1122 (2017)

[7] Wang, G., Schnitzer, M.E., Chen, T., Wang, R., Platt, R.W.: A general framework for identification of permissible variable subsets and development of structured variable selection methods. arXiv preprint arXiv:2110.01031 (2021)

[8] Wang, G., Perreault, S., Platt, R.W., Wang, R., Dorais, M., Schnitzer, M.E.: Structured variable selection: an application in identifying predictors of major bleeding among hospitalized hypertensive patients using oral anticoagulants for atrial fibrillation. arXiv preprint arXiv:2206.05337 (2022)

Review 1: "Antibiotic Prescribing in Remote Versus Face-to-Face Consultations for Acute Respiratory Infections in English Primary Care: An Observational Study Using TMLE"

Connections