And we can talk about​

High Quality RWE.

RWE recommendations and guidance by Health technology assessment agencies and regulators state that High-quality RWE is based on a principled scientific process:

Deepening in the process:

01 - Study design.

Description:

Savana’s study protocols and statistical analysis plans have been approved by multiple research ethics boards in different countries across Europe, North and Latin America. Results have been published in top scientific Journals.

Task involved:

  • Strong emphasis on clear and purpose driven study objectives.
  • Research methodology explanation in protocols.
  • Methodology check points across the complete research study process.
  • Advanced and methodologically  validated NLP and ML software models.

Objectives:

  • Fit-for-purpose study design.
  • Transparent protocol development.

Exemplar:

02 - Data source.

Description:

Savana participates only when EHR text is relevant to the objectives of the study.

Other structured databases can be included.

Data reliability checks performed: including accuracy, validity, conformance, plausibility, completeness, data provenance, and transparency in data processing.

Task involved:

  • Project qualification by Savana’s scientific department.
  • Determining the structured and non structured data sources needed for the study.
  • Data-driven quality assessment and quality checks across the complete research study process.
  • Implement contingency strategies.

Objectives:

  • Determining ‘data quality’ of RWD.
  • Identifying ‘fit-for-purpose’ RWD sources.

02.1 - Data-driven quality assessment:

Summary statistics for:

  • Population demographics.
  • Study key variables.

Objectives:

  • To guarantee data quality to answer objectives.
  • To adapt analysis strategy:
    • Missing values.
    • Data source heterogeneity.
    • Adjust population selection criteria.
  • To implement contingency strategies where needed.

02.1 - Fit-for-purpose NL2

Fit-for-purpose NLP models are built based on the objectives of the study.

Advanced NLP goes beyond codifying or one-time extraction of variables from text. It requires a deep clinical understanding of how EHRs are populated and what inferences need to be made to extract knowledge from each situation and for a given purpose.

Task involved

  • Data assessment by clinical experts.
  • Scientific inferences to build ad-hoc models.
  • Creating a structured database for each study
  • NLP model external validation.

Objectives

  • Developing ‘fit-for-purpose’ NLP models.
  • Evaluating ‘NLP models quality’.

This is unique to Savana's methodology.

This is unique to Savana's methodology.

02.1.1 - External Evaluation Methodology:

02.1.2 - Evaluation of the NLP system:

03 - Analytical methods.

Description:

We leverage clinical expertise to guide our analysis strategy: 

  • Ad-hoc definition of study objectives to ensure study feasibility.
  • Tailored analyses using multiple AI-based methodologies to satisfy specific clinical questions.

From single terms to complex medical concepts and patient populations:

  • Flexible time point and window definition for data collection to minimize measuring bias.
  • Complex variables definition through the aggregation of several terms and conditions.

Task involved:

  • Definition of the study objectives.
  • Tailored clinical analysis.
  • Traditional and AI-based analytics
  • Time point and window definition. 
  • Complex variables definition.

Objectives:

  • Selecting appropriate ‘analytical methods’.
  • Enduring transparency.

Exemplar:

Deep learning.

Random forest.

Neural networks.

Logistic regression.

04 - Transparency and reproducibility.

Description:

  • Analysis strategy updated according to data availability and granularity. Questions and conclusions adjusted accordingly to maximize clinical value.
  • Detailed description of methods used for descriptive and predictive analysis. Code sharing to maximize reproducibility.
  • Data visualization of intermediate and final results to ensure transparency and results interpretability.

Task involved:

  • Analysis strategy.
  • Code sharing.
  • Data visualization.

Objectives:

  • Transparency and reproducibility, especially in “study report” development. 

Exemplar:

The Impact of COVID-19 on Patients with Asthma.

05 - Final report evaluation.

Description:

There are several tools to assist decision makers in evaluating the quality of RWE studies. However, as yet, there is no consensus on a gold standard tool. 

Savana’s dRWE Scientific Methodology incorporates considerations in each point in the scientific process.

  • A priori considerations.
  • Study design.
  • Data sources.
  • Analytical methods.
  • Transparency and reproducibility.
  • Results reporting.
  • Interpretation of findings.

Objectives:

  • Understanding of how decision-makers will evaluate the quality of the RWE study.

Exemplar:

Clinical characteristics and prognostic factors for ICU admission of patients with covid-19: Retrospective Study Using Machine Learning and Natural Language Processing.

Clinical Management of COPD in a Real-World Setting. A Big Data Analysis.

Generate deeper and more agile clinical evidence, thanks to the use of AI

Fill in this form carefully so that we can check if there is a match between you and us:

We will contact you even if we are not interested in your research project.

Enrol in a project