An Informatics Consult approach for generating clinical evidence for treatment decisions

Alvina G. Lai, Wai Hoong Chang, Constantinos A. Parisinos, Michail Katsoulis, Ruth Blackburn, Anoop D Shah, Vincent Nguyen, Spiros Denaxas, George Davey Smith, Tom R. Gaunt, Krishnarajah Nirantharakumar, Murray P. Cox, Dónall Forde, Folkert W. Asselbergs, Steve Harris, Sylvia Richardson, Reecha Sofat, Richard Dobson, Aroon D. Hingorani, Riyaz Patel, Jonathan A C Sterne, Amitava Banerjee, Alastair K. Denniston, Simon Ball, Neil J. Sebire, Nigam H. Shah, Graham R. Foster, Bryan Williams, Harry HemingwaymedRxiv 2021 · DOI

Abstract

Clinicians frequently encounter complex patients for whom standard clinical guidelines offer little direction. An "Informatics Consult" has been proposed to address this by requesting novel evidence from large-scale health data resources, tailored to specific patient profiles and returned within clinical timescales. This study demonstrates the feasibility of such a consult in a scenario where a treatment indication (atrial fibrillation) and a contraindication (liver cirrhosis) coexist. By triangulating evidence from randomized controlled trial (RCT) registries, meta-analyses of observational studies, target trial emulations using population-scale electronic health records (EHR), and Mendelian randomization, we generated a robust evidence base where traditional trials were absent. Our findings show broad concordance across these sources, suggesting that warfarin use in this population is associated with lower stroke risk and all-cause mortality. A pilot survey indicates strong clinician support for this service, with 79% of respondents advocating for its integration into healthcare systems.

Introduction

Modern medicine increasingly relies on Randomized Controlled Trials (RCTs) as the gold standard for evidence. However, a significant "evidence gap" exists: approximately 80-85% of clinical recommendations lack high-level trial support. This gap is most pronounced for patients with multimorbidity, who are frequently excluded from RCTs to maintain study homogeneity. When clinicians face "treatment clashes"—where a drug indicated for one condition is potentially contraindicated for another—they are often forced to rely on intuition rather than empirical data.

This paper presents a framework for an Informatics Consult, a service designed to provide on-demand, patient-specific evidence by leveraging massive EHR databases and advanced causal inference methods. We test this framework using the clinical dilemma of anticoagulation (warfarin) for stroke prevention in patients with both atrial fibrillation (AF) and liver cirrhosis.

The Informatics Consult Framework

The Informatics Consult utilizes a four-stream approach to triangulate evidence:

  1. RCT Registry Systematic Search: A search of ClinicalTrials.gov to identify existing or ongoing trials relevant to the specific patient population.
  2. Meta-analysis of Observational Studies: Aggregating existing peer-reviewed research to synthesize current knowledge on the treatment-condition triad.
  3. Target Trial Emulation: Applying RCT design principles (e.g., defining "time zero," propensity score matching) to large-scale observational EHR data to estimate causal effects.
  4. Mendelian Randomization (MR): Using genetic variants as proxies for drug effects to provide a biological validation of the drug target, largely independent of the confounding factors found in observational data.

Methodology and Data Sources

The study utilized a linked dataset of 3,854,710 adults in England, combining primary care records (CPRD), hospital statistics (HES), and death registration data (ONS). For the genetic component, summary data from the MEGASTROKE study and genome-wide association studies (GWAS) for vitamin K1 levels were employed. The case study focused on patients aged 30 and older with a dual diagnosis of AF and cirrhosis, evaluating the efficacy (stroke risk) and safety (all-cause mortality) of warfarin.

Results

The Evidence Void

The systematic search for RCTs returned zero results. Every identified trial for anticoagulation in AF specifically excluded patients with liver cirrhosis, confirming the lack of traditional evidence for this population.

Evidence Triangulation

Despite the lack of RCTs, the other three streams provided concordant evidence:

  • Meta-analysis: Prior observational studies indicated that warfarin use was associated with a lower stroke risk (Hazard Ratio [HR] = 0.71).
  • Target Trial Emulation: In the EHR-based emulation, warfarin was associated with a 39% reduction in all-cause mortality (HR = 0.61) and a 73% reduction in ischaemic stroke risk (HR = 0.27).
  • Genetic Validation: Mendelian randomization confirmed that lower levels of vitamin K1 (warfarin acts as a vitamin K1 antagonist) are associated with lower stroke risk, providing a biological "sanity check" for the observational findings.

Clinician Reception

A pilot survey of 34 independent clinicians revealed a high demand for this service. 85% of clinicians found the prognostic information (e.g., 1-year mortality rates) useful for patient management, and 79% believed the Informatics Consult should be a standard service within healthcare systems.

Discussion

The Informatics Consult demonstrates that while certain condition pairs may be rare for an individual practitioner (affecting roughly 10 in 10,000 individuals), they affect tens of thousands of patients at a national level. By pooling this data, we can generate statistically significant insights.

Limitations and Future Work

Several hurdles remain before full implementation. This study did not assess bleeding risks due to data limitations, which is a critical factor in the AF-cirrhosis dilemma. Furthermore, while the current study took approximately one month to complete manually, the authors identified steps for automation—such as automated phenotype creation and batch data extraction—that could reduce the turnaround time to 48 hours.

Conclusion

The Informatics Consult represents a shift toward a "learning health system," where the data of every patient informs the care of the next. By providing a structured framework for evidence generation in the absence of RCTs, this approach offers a scalable solution to the evidence gap in modern clinical practice.

Summary updated

Abstract

Abstract An Informatics Consult has been proposed in which clinicians request novel evidence from large scale health data resources, tailored to the treatment of a specific patient, with return of results in clinical timescales. However, the availability of such consultations is lacking. We seek to provide an Informatics Consult for a situation where a treatment indication and contraindication coexist in the same patient, i.e., anti-coagulation use for stroke prevention in a patient with both atrial fibrillation (AF) and liver cirrhosis. We examined four sources of evidence for the effect of warfarin on stroke risk (efficacy) or all-cause mortality (safety) from: (i) randomised controlled trials (RCTs), (ii) meta- analysis of prior observational studies, (iii) trial emulation (using population electronic health records (N = 3,854,710) and (iv) genetic evidence (Mendelian randomisation). We developed prototype forms to request an Informatics Consult and return of results in electronic health record systems. We found 0 RCT reports and 0 trials recruiting for patients with AF and cirrhosis. We found broad concordance across the three new sources of evidence we generated. Meta-analysis of prior observational studies showed that warfarin use was associated with lower stroke risk (hazard ratio [HR] = 0.71). In a target trial emulation, warfarin was associated with lower all-cause mortality (HR = 0.61) and ischaemic stroke (HR = 0.27). Mendelian randomisation served as a drug target validation where we found that lower levels of vitamin K1 (warfarin is a vitamin K1 antagonist) are associated with lower stroke risk. A pilot survey with an independent sample of 34 clinicians revealed that 85% of clinicians found information on prognosis useful and that 79% thought that they should have access to the Informatics Consult as a service within their healthcare systems. We identified candidate steps for automation to scale evidence generation and to accelerate the return of results within clinical timescales.
BibTeX
@misc{lai2021informatics,
  author = {Alvina G. Lai and Wai Hoong Chang and Constantinos A. Parisinos and Michail Katsoulis and Ruth Blackburn and Anoop D Shah and Vincent Nguyen and Spiros Denaxas and George Davey Smith and Tom R. Gaunt and Krishnarajah Nirantharakumar and Murray P. Cox and Dónall Forde and Folkert W. Asselbergs and Steve Harris and Sylvia Richardson and Reecha Sofat and Richard Dobson and Aroon D. Hingorani and Riyaz Patel and Jonathan A C Sterne and Amitava Banerjee and Alastair K. Denniston and Simon Ball and Neil J. Sebire and Nigam H. Shah and Graham R. Foster and Bryan Williams and Harry Hemingway},
  title = {{An Informatics Consult approach for generating clinical evidence for treatment decisions}},
  year = {2021},
  doi = {10.1101/2021.01.10.21249331},
  howpublished = {medRxiv},
}
Figure