251125P - FROM DATA TO ACTION: TRANSLATING GLOBAL RESEARCH DATA INTO EFFECTIVE STRATEGIES: HIGHLIGHTS HOW GLOBAL RESEARCH DATA CAN BE TRANSLATED INTO ACTIONABLE HEALTH CARE STRATEGIES BASED ON DATA AND WISDOM-DRIVEN APPROACHES

Presentation at the ‘Hail International Conference on Scientific Research and Innovation on the Future of Health Care’, 25th November 2025, by Professor Omar Hasan Kasule MB ChB (MUK), MPH (Harvard), DrPH (Harvard), Chairman of the Institutional Review Board at King Abdullah bin Abdulaziz University Hospital - Princess Noura University, Riyadh, Saudi Arabia.

BASICS OF CAUSAL RESEARCH: Independent determinant and outcome variables

• Research is basically uncovering the causal relation (السببية) between causes (determinant variables) and outcomes (dependent variables). 
• Statistically, the causal relation is assessed in two stages: as association using the chi-square or the t-test, and as strength of association (effect measure) using the odds ratio.
• Causal relations manifest the laws of nature sunan (سنن الكون) and are constant and fixed (ثبات السنن).
• The causal relation is considered very strong if a biological mechanism is known which reflects Hill’s causal criterion of biological plausibility.

BRADFORD HILL’S 9 CRITERIA OF CAUSALITY 

1.  Strength of association
2.  Consistency across studies
3.  Specificity of the link, 
4.  Temporal sequence (cause before outcome)
5.  Biological gradient (dose-response), 
6.  Biological plausibility, 
7.  Coherence with other knowledge, 
8.  Experimental evidence in which an intervention produces the outcome
9.  Analogy to other known causal links.

PROMOTIVE AND PREVENTIVE CAUSES

• The causal relation may be promotive (promotes the outcome) or preventive (prevents the outcome). 
• Both causative and preventive relations are important, and the general philosophical debate about which one is more important or more cost-effective is unnecessary because the correct answer is very situation-specific.
• Once the causal relation is established, action must follow as health strategies to promote health or prevent disease.
• Health strategies not based on correct causal relations can be harmful or waste resources.

ILLUSTRATION OF CONFOUNDING

 

CAUSAL RELATIONS MUST BE FREED OF CONFOUNDING EFFECTS 

• A confounding variable is a third variable in a study that is related to both the independent and dependent variables, distorting the relationship between them.
• Age, gender, and socio-economic variables are common confounding variables easy to identify and control for. Unknown or unimaginable confounders are the cause of flawed causal research 
• Failing to control for confounding variables can lead to biased conclusions, making it appear that there is a cause-and-effect relationship when there isn't, or masking an actual relationship. 
• Strategic intervention based on a confounded relationship will be flawed and even disastrous in many situations.

RESEARCH HYPOTHESES

• Research and data analysis must have underlying hypotheses to lead to efficient intervention strategies. The hypotheses are products of logic/reasoning.
• Scientific research involves both deductive reasoning and inductive reasoning. The philosophical debate about which of the two is more important is unnecessary because both are involved in the formulation of research hypotheses (deductive logic) and testing them (inductive logic). 
• The concurrence of several inductive causal relations leads to a reliable conclusion that can be the basis for health strategies if the condition of biological plausibility is fulfilled.
• A scientist must have a hypothesis before approaching data, as we learn from the famous French inventor Louis Pasteur (1822-1995) 
  
• We cannot do research as we fish using a fishing bait or a net that we throw and hope it will bring up what we want. It may bring anything.

THE PREPARED MIND OF A SCIENTIST ACCORDING TO LOUIS PASTEUR

• Louis Pasteur, on December 7, 1854, talked about what appeared as an accidental discovery of electromagnetism by Hans Christian Ørsted.

• Dans les champs de l'observation, le hasard ne favorise que les esprits préparés.
• English Translation: "In the field of observation, chance favors only the prepared mind".
• Pasteur discovered pasteurization, microbial fermentation, and vaccination because he had hypotheses from a well-prepared mind.
• Hans Christian Ørsted accidentally discovered the link between electricity and magnetism in 1820 when he observed a compass needle deflect as an electric current was switched on and off during a lecture demonstration. 
• In 1928, Scottish bacteriologist Alexander Fleming accidentally discovered penicillin when he noticed that a mold, Penicillium notatum, had contaminated a petri dish and created a bacteria-free zone around it.

DATA

• Empirical scientific research is based on data in its widest sense. 
• Data yields information if summarized. 
• Information yields knowledge when analyzed. 
• Wisdom is needed to understand knowledge before it is used, especially in interventions that can harm human and environmental health.
• Data-driven results without wisdom (hikmat) lead to disasters. When we discover a causal relation, we need to ask fundamental questions: what is the future vision and grand objective? What are the possible harms? Not asking these questions that are easy to answer leads to regrets.

DATA WITHOUT WISDOM CAN BE DISASTROUS

• Oppenheimer regretted his work on the atomic bomb after seeing the horror it caused and is reported to have opposed the creation of the H bomb.

• We are very excited about artificial intelligence: who knows what our future vision is? Will the machine control the human? Who will be liable for mistakes made by the machine?

CONTROL OF CONFOUNDING TO REACH ACCURATE CAUSAL RELATIONS

• The term ‘confounding’ comes from the Latin word ‘confundere’, which means to mix together. It came to us through Middle English from the French.
• The effect of confounding must be removed from the causal analysis before drawing conclusions that can drive strategic interventions.
• At the study design stage, confounding is controlled by randomisation, matching, or restriction.
• At the study analysis stage, confounding can be removed by stratification and by statistical models such as regression.
• Confounding can be controlled but cannot be removed completely, and is one of the reasons that force us to state our conclusions as confidence intervals.

LARGE DATA DILUTES CONFOUNDING

• Research based on large data sets reduces the effect of confounders. We aim at big study samples as well as multi-center or multi-national studies to get more accurate results because confounding relations are diluted by the large samples. 
• There are many mathematical explanations of why large data sets are more accurate, but I prefer to make the economics analogy of Adam Smith (1723-1790), who taught the metaphor of an invisible hand that guides contradictory and individual decisions of consumers for overall economic benefit.

• A statistical invisible hand may operate in research by removing effects of confounding from large data sets.

PREFERENCE FOR LARGE DATA SETS

• Journal editors and funding agencies prefer large studies
• Researchers are encouraged to submit their data to a pool for analysis by others 
• Artificial intelligence has a spin-off. Its ability to handle large data sets has encouraged the setting up of large data portals into which data is deposited for analysis by others to give more accurate causal relations.

GLOBAL RESEARCH BASED ON OPEN DATA SOURCES ENDING THE ERA OF DATA SECRECY AND DATA HOARDING

• NASA has a portal data.nasa.gov that is a publicly available metadata repository involving aeronautics and space exploration.

• A Saudi open data initiative encourages ministries and other organizations to make their data available publicly at the portal https://od.data.gov.sa 
• Open data repositories are maintained by the US Department of Agriculture and the Saudi Ministry of Environment, Water and Agriculture (وزارة البيئة والمياه والزراعة ) https://www.mewa.gov.sa › OpenData.

GLOBAL RESEARCH BASED ON META-ANALYSIS

• Meta-analysis has, over the past 30 years, developed into a well-established branch of statistics. 
• Meta-analysis aims at solving the problem of low power small size studies with weak causal relations and turning them into stronger, more reliable studies.
• Combined effect measures are based on combining the effect measures of many small studies, each contributing according to its variance. 
• The combined effect measure that results reflects a large sample with thousands of subjects.
• The odds ratio is the most popular effect measure used in meta-analysis because it is invariant across different study designs, making it possible to combine them.

GLOBAL RESEARCH BASED ON COLLABORATIVE INTERNATIONAL RESEARCH

• Universities and funders of research encourage collaborative research to have access to larger data sets.
• Collaboration at local, national, and international levels. With researchers working from apace, we shall also have interplanetary collaborations.
• Collaboration is most effective if a common study design and methodology are fixed from the beginning.

ARE RESULTS FROM GLOBAL DATA MORE ACCURATE?

• Useful research must be global because correct inductive research must concur irrespective of where it is done and by whom. 
• Health action should follow research results that are validated on a global level.
• Creation, whether biological or physical, is the same in the whole cosmos; otherwise, the laws of physics on Earth would not apply on the moon. This is because of tauhid al rubuubiyyat. 
• Research at a global level with large samples helps mitigate the problem of confounding.

ADVANTAGES OF LARGE SAMPLES IN MITIGATING CONFOUNDING

• The bigger the sample size, the less the confounding effect.
• A large sample size assures more statistical power to detect true causal relations.
• A larger sample size is needed to controlfor confounding by using more sophisticated methods. 
• A small sample size cannot fully mitigate confounding, leading to bias, and does not have enough data for sophisticated analysis of confounding.