- This topic has 13 replies, 14 voices, and was last updated 1 week, 5 days ago by
.
-
Posts
-
-
-
Ways to cope with challenges in big health data/EHRs:
1.Minimizing Data Inconsistency: Database Management Systems should include interchange and translation mechanisms to standardize data across hospitals and platforms. Furthermore, machine learning algorithms should be leveraged to extract meaningful diagnoses and investigation results from unstructured clinical text. This allows important information to become usable for analysis and decision-making.2.Improving Data Quality: EHRs should require mandatory completion of key CVD fields such as blood pressure, lipid profiles, ECG findings, and diagnosis codes before a record can be closed. This helps ensure clinicians capture essential information during routine care. In practice, smart alerts can gently remind clinicians when essential data are missing, making data quality part of daily workflow rather than an extra burden.
3.Building the culture of data ownership: a culture where clinicians and relevant persons feel ownership of the data. For instance, showing clinicians dashboards of their own patient outcomes or service performance can help them see that the data benefits their work, not just publications and research purposes.
4.Enhancing Translational Applicability: Utilization of machine learning can help standardize image interpretation, such as echocardiography, ECGs, and radiology images. This supports clinicians with consistent assessments. In addition, researchers should improve transparency by clearly describing dataset variables. When clinicians understand what the data mean, research findings become usable.
5.Reducing Selection Bias: In clinical research, stronger statistical methods should be applied, including hypothesis testing and appropriate sampling techniques. When possible, randomized controlled trials and population-based datasets should be combined with routine clinical data to reduce bias. This helps ensure that findings reflect real-world situation.
6.Investment in Training: Investment should be considered in formal training curricula in informatics, coding, data management, and advanced statistical tools. Training should not only target researchers but also clinicians who enter data daily. When staff understand how data are used, they become more motivated to collect high-quality information and apply research results in practice.
7.Solving Privacy and Ethical Issues: Updated cybersecurity and encryption systems are needed to implement ahead of cyber threats. Clear governance frameworks, consent procedures, and role-based access control help protect patient confidentiality.
-
To cope with the challenges
1. Addressing Missing Data
Improve source data entry: Agree-on standardized variable and mandatory data entry field ensuring all essential data are filled.
Capacity building: sufficient training is provided for the assigned data entry staff.
Use alternative analyses for handling high missing values: for example, amputation techniques, mixed effects regression models, generalized estimating equation.
Use appropriate statistical methods based on the level of missingness (e.g. multiple imputation, mixed-effects models).
Conduct regular review sessions: data audits are regularly conducted to identify operational and systematic gaps to smoothen workflow and improve data quality.2. Reducing Selection Bias
Use advanced data analytic methods: including propensity score analysis, instrumental variable analysis and Mendelian randomization.
Use big data mainly for hypothesis-generation: always check and validate with RCT or triangulate multiple studies to be used for clinical practice.
Ensure transparency: about inclusion/exclusion criteria and participants characteristics.3. Strengthening Data Analysis Capacity
Build a team including experts with various skills for data handling: to handle very large datasets with multiplicity requiring multiple analyses to establish the significance of a hypothesis and identify correlations.
Build a multidisciplinary team: including clinicians, researchers, health informaticians, data scientists, statisticians and others.
Capacity building programs: for researchers including data science, health informatics, statistics and machine learning.
Standardize analytical protocols: to reduce multiple testing inappropriately and false positives.
Use validated algorithms and reproducible methods: for data analyses ensuring accuracy, transparency and ability to verify independently and therefore improve reliability of findings.4. Improving Interpretation and Translational Use
Early involvement of relevant stakeholders: for example, involve clinicians from the beginning of the study (designing to interpretation of results) to ensure clinical relevance and produce actionable results.
Produce results in clinical usable/meaningful formats: focus to provide actionable insights, not complex ones.
Enhance documentation: standardized essential data variables and documentation to be interpreted and effectively used.5. Managing Privacy and Ethical Issues
Data governance, oversight and data protection: regular audit trials with access control, encryption. Clear laws and policies should be in place to mitigate the breach of personally identifiable information.
Anonymize data: to reduce identification risks.
Minimize data: only necessary information should be provided and used by researchers to reduce data breach.
Ethical Board consent: to request consent from board members to balance privacy with public health benefits and ensure research is conducted ethically.6. System-Level and Policy Solutions
Data sharing policy: ensure responsible data sharing across organisations and departments with clear regulatory and ethical safeguards.
Data standards: develop national standards for EHR to enhance interoperability, data quality and health information exchange.
Digital infrastructure: safeguarding infrastructure according to minimum standards to prevent cyberattacks and data breaches.In conclusion, many adjunctive and robust procedures should be planned and implemented at each steps of data processing to make the greatest possible use of big data and improve public health.
-
To cope with the challenges
1. Use standard formats
Use common rules and coding systems so data from different hospitals and systems can be easily combined and compared.
For example, public hospitals and private hospitals in Myanmar should use the same diagnosis and reporting formats so patient data can be shared nationwide.
2. Improve data quality
Check data regularly to remove errors, duplicate records and incorrect information.
In Myanmar hospitals, regular review of electronic and paper-based records can reduce mistakes caused by manual data entry.
3. Manage missing data properly
Use suitable statistical methods to deal with missing information instead of ignoring it.
This is important in Myanmar, where some patient records may be incomplete due to limited resources or emergency situations.
4. Link data from different sources
Combine data from electronic health records, medical devices and registries to get a complete picture of patients.
For example, linking hospital records with data from community clinics and screening programs in Myanmar can improve disease monitoring.
5. Train healthcare staff and researchers
Teach doctors and researchers basic data analysis and computer skills to handle big data better.
Training programs for Myanmar healthcare workers can improve confidence in using electronic health systems.
6. Use advanced analysis tools carefully
Apply machine learning and computer models to analyze large datasets but always test and validate the results.
These tools could help Myanmar researchers identify trends in common diseases such as hypertension and heart disease.
7. Protect patient privacy
Use strong security systems and follow laws to keep patient information safe and confidential.
This is especially important as Myanmar moves from paper records to digital health systems.
8. Clear rules for data sharing
Create clear legal and ethical guidelines so data can be shared safely for research.
National policies can guide how hospitals and universities in Myanmar share health data responsibly.
9. Reduce bias in data
Use proper methods to reduce bias and make sure results represent real patient populations.
Including data from both urban and rural areas in Myanmar helps ensure fair and accurate results.
10. Apply results to real healthcare
Make sure research findings are easy to understand and can be used by doctors to improve patient care.
For example, research findings can help Myanmar clinicians improve prevention and treatment strategies for cardiovascular diseases. -
The followings are my suggestions to cope with those challenges.
Missing Data
Missing data can be minimized by strengthening data entry processes through regular training, robust standard operating procedures (SOPs), and built-in data validation mechanisms at the point of entry. When missing values occur, they can be addressed by validating and supplementing information from alternative data sources, such as laboratory machines, laboratory registers, and linked records across multiple systems. In addition, unstructured data can be transformed into usable formats through the application of natural language processing (NLP) tools, enabling the recovery or inference of otherwise missing information.Selection Bias
Selection bias can be mitigated through the use of advanced analytical techniques, including propensity score analysis, instrumental variable analysis, and Mendelian randomization. Large-scale datasets should primarily be used for hypothesis generation rather than direct clinical decision-making. Findings derived from big data analyses should be carefully validated through randomized controlled trials (RCTs) or triangulated with evidence from multiple independent studies before being translated into clinical practice.Data Analysis and Training
Effective data analysis requires the establishment of a multidisciplinary team comprising clinicians, researchers, health informaticians, data scientists, statisticians, and other relevant experts. Such teams can collaboratively develop comprehensive training materials and innovative training approaches tailored to diverse skill levels. The creation of standardized templates and analytical tools can further reduce entry barriers, shorten learning curves, and promote consistent and high-quality data analysis practices.Interpretation and Translational Applicability of Results
Early and continuous involvement of key stakeholders—particularly clinicians—throughout the research process, from study design to interpretation of results, is essential to ensure clinical relevance. Studies should be designed with a backward approach, starting from the intended clinical application to ensure the results are actionable. Emphasis should be placed on clinically meaningful outcomes rather than solely on statistical significance, and analytical models should be communicated in clinician-friendly terms, such as risk, benefit, and potential harm.Privacy and Ethical Issues
Privacy and ethical considerations should be addressed through a privacy-by-design approach, including data de-identification and strict access control mechanisms. Broad consent models can be adopted, provided they are accompanied by clear and transparent communication with participants. Researchers should explicitly articulate both the potential risks and the anticipated societal benefits associated with data use to maintain trust and ethical integrity. -
1. Missing data
The reasons are due to omitted by clinicians, refused by patients, not attending for data collection. Statistical methods have limitations for analyzing missing data and make it difficult to produce plausible results.
Suggestions: data collection protocol should be standardized for the system, not be subjective by clinicians or others. Essential variables should be collected to avoid big volume without meanings. Data Management Trainings should be conducted regularly to reduce entry error to database. Advanced statistical methods can be used to get meaningful data analysis results.
2. Selection bias
The reasons are due to variations in subjects’ geographic, medical history profiles and insurance etc. Exploited statistical analysis can have confounding implications. Large volume is not advantage for representative sample.
Suggestions: sample selection should be following statistical methods/calculations such as weighting, randomization in extracting from big data. Blinding and masking of researcher and clinician can also help to reduce this bias. Appropriate statistical analysis should be chosen for comparing data.
3. Data analysis and training
Big data analysis requires multiple analysis to establish hypothesis and significant correlations. So, the researcher’s skill to use statistical and methodological tools is important.
Suggestions: the clinicians/researchers should collaborate with informaticians, statistician and data scientists in big data analysis. They also need to attend required basic methodological trainings.
4. Interpretation and translational applicability of results
Important to integrate the analysis output to daily clinical practice.
Suggestions: Improve transparency and documentation of datasets, including standardize disease definitions to ensure interpretability and reproducibility between clinicians and researchers. Advocacy and sharing practices with research findings should be carried out as event for raising application of results.
5. Privacy and ethical issues
Data protection policy and regulations must be set up for all big dada storage in server. Data encryption, data authorization and access should be strictly maintained. Data security protocols, safeguarding practice and trainings should be updated regularly. -
Here are my suggestions on coping with big health data challenges
1. Missing data
Some methods to deal with missing data are the following:
a) complete-case analysis (CCA)
– all persons with missing values on one or more variables are excluded from the analysis. This method has a lot of drawbacks and should be avoided in general because it generates unbiased results only in some situations
b) imputation
– replacement of missing data by real values
– multiple imputation is recommended over single imputation methods (mean imputation, imputation based on linear regression, and last value/observation carried forward) because most single imputation methods lead to an artificial decreased standard deviation in the variables to be analyzed, resulting in too small standard errors
– multiple imputation consists of three phases: imputation, analysis, and pooling2. Selection bias
In a paper published by Rojas-Saunero et al. (2023), the following solutions can prevent selection bias in health research:
a) Clearly specify the target population
b) Collect primary data in a way that ensures accessibility for participants who are often marginalized. Ideally, all social groups should be recruited from the same source, rather than creating a distinct recruitment pipeline that draws from different populations to achieve diversity.
c) Design retention strategies to prevent differential loss to follow up3. Data analysis and training
This can be solved by training the clinicians and researchers on informatics and tools for big health data analysis.4. Interpretation and Translational Applicability of Results
Tools for visualization of big data such as Tableau, Microsoft Power BI, Google Looker Studio, and D3.js can be used to present big data into information that is easy to understand and interpret.5. Privacy and Ethical Issue
Regulations on the use of data, for example the General Data Protection Regulation (GDPR) in Europe and Data Privacy Act in the Philippines, must be followed to ensure data privacy and confidentiality.References:
Heymans, M. W., & Twisk, J. W. (2022). Handling missing data in clinical research. Journal of Clinical Epidemiology, 151, 185–188. https://doi.org/10.1016/j.jclinepi.2022.08.016Rojas-Saunero, L. P., Glymour, M. M., & Mayeda, E. R. (2023). Selection Bias in Health Research: Quantifying, Eliminating, or Exacerbating Health Disparities? Current Epidemiology Reports, 11(1), 63–72. https://doi.org/10.1007/s40471-023-00325-z
Staff, C. (2025, April 11). Big Data Visualization tools: Types, benefits, and how to choose. Coursera. https://www.coursera.org/articles/big-data-visualization-tools-types-benefits-and-how-to-choose
-
I want to suggest the following actions to cope with the big health data challenges mentioned in “Big Health Data and Cardiovascular Diseases: A Challenge for Research, an Opportunity for Clinical Care”:
Missing data
To address this challenge, I would like to propose standardizing terminology and variables across all systems to enhance interoperability. In addition, the data quality will be improved by embedding a structured data system and key variables. The regular data audit will be performed, and a missing data report will be extracted from the system with an action plan. And appropriate advocacy and orientation will be conducted to all the relevant stakeholders with the collaboration of local authorities.Selection Bias
I would like to develop predefine data analysis plan to identify which variable is mandatory, and the application of advanced data analysis techniques will be beneficial to cope with this challenge.Data Analysis and Training
I would like to form multidisciplinary teams with clinicians, statisticians, and data scientists. Then, I will develop models and share reusable data analytical pipelines with the collaboration of this team. In addition, I will provide training to ensure the interpretation with the collaboration of academic and industry experts.Interpretation and Translational Applicability of Results
I would propose developing clinically interpretable models with explainability tools, with the involvement of clinicians whether outputs align with real-world workflows.Privacy and Ethical Issues
The data governance system will be strengthened. Societal benefits and individual rights need to be balanced. -
Coping With Big Health Data Challenges
1. Missing Data
Accept that missing data are common due to paper records, displacement, and weak follow-up.
Define minimum essential data for key programs (NCDs, TB, MCH).
Use simple methods always and apply complex, multiple imputation, and mixed models only when needed.
Improve data collection using simple digital tools (DHIS2, Kobo, ODK).2. Selection Bias
Large datasets might not be representative of the whole population due to selection bias.
Clearly state who is included and excluded in analyses.
Use triangulation (routine data + surveys + qualitative data).
Treat findings as hypothesis-generating, not practice-changing.3. Data Analysis & Skills
Limited local skills are a major challenge.
Train clinicians and public health staff in basic data analysis and interpretation.
Use standard analysis templates to reduce errors.
Avoid advanced AI or machine learning unless data quality is strong.4. Interpretation & Use of Results
Translate results into simple, actionable messages for decision-makers.
Involve program staff in interpreting findings.
Use big data to identify trends and gaps, not to dictate clinical care.5. Privacy & Ethics
Build trust through clear explanations of how data are used.
Collect only necessary data and limit access.
Use broad consent models suitable for conflict-affected settings. -
Sometime I think it is hard to do what the paper proposed because people work in different steps of collecting and using data, therefore if you only collect data you may not be aware of how poor quality data can result in poor analysis and poor conclusion. While data analysts may not be aware of challenges the data collectors face from the start of the project. Good communication is essential.
-
I would like to add some thoughts about the communication gap between the frontline collectors and the analysts regarding the main challenges in the article.
Missing Data: Data is often missing because workers are busy and sometimes don’t see the point of completing it. They should be shown how their data turns into better patient care or improves their workflow. If they see positive results, they would be less likely to skip the info. Automated error-checking can also be included in the process of checking for missing data and errors.
Selection Bias: Big data, for example, the national registries, should be checked for quality and coverage so they do not miss the full picture of representatives. Also, analysts can later use statistical techniques to adjust for bias in this observational data.
Training and applications: We can start by training the data lead in each department and then branch out. The key is making it practical by the collectors need to know how to keep data clean, while analysts need to master the stats/AI/tech tools to handle these big datasets.
Privacy and Ethics: The current system feels like every institution has its own rules and ethical committees, making data sharing difficult. We need more centralized regulation to streamline the process.
-
Method of coping with those challenges (Big Health Data or EHR)
Solving the missing data
To reduce missing data, I think standardizing the definitions and variables across data systems so data are collected consistently and shared more easily. Improve data quality by relying on structured electronic health records with clearly defined key variables and conducting regular data audits and generating missing-data reports, followed by clear and practical action plans. In addition, provide continuous orientation and engage clinicians and data collectors to reduce avoidable data gaps.
Reducing the selection Bias
Addressing selection bias by developing a clear and predefined data analysis plan that specifies inclusion criteria and essential variables in advance is vital. Apply appropriate statistical methods to adjust for confounding factors. Because most big health data are observational, it is essential to interpret findings cautiously and use them mainly to generate hypotheses rather than to directly inform clinical decision-making.
Improve the skills of Data Analysis and provide training
Analyzing complex big data such as the example from cardiovascular research requires strong multidisciplinary collaboration. Bringing clinicians, statisticians, and data scientists together to improve both the quality and relevance of analyses is a must. Also develop shared analytical models supported by targeted training to ensure accurate analysis and meaningful interpretation. Continuous capacity building for those who are responsible is essential.
Interpretation and Translational Applicability of Results
Prioritize clinically interpretable models so results remain useful in real-world cardiovascular care. By involving clinicians in model development and validation, I ensure that findings align with actual clinical workflows and can be responsibly translated into practice.
Privacy and Ethical Issues
Develop a data-sharing policy and strengthen data governance to protect patient privacy. To ensure the use of big health data balances public benefit with individual rights, apply secure access controls, ethical oversight mechanisms, and transparent data-use policies.
-
Big health data offers major opportunities to improve healthcare and research, but it also presents several challenges, especially in low-resource settings. To effectively cope with these challenges, practical and context-appropriate strategies are required.
First, start with simple data standardization such as using basic, nationally agreed disease definitions and reporting formats to reduce inconsistency across hospitals and NGOs. Secondly, improving data quality and managing missing data through routine checks and simple validation processes can increase the reliability of analyses. Thirdly, strengthening data linkage and coordination between hospitals, NGOs, and national systems helps reduce fragmentation and duplication of data.
In addition, capacity building of health professionals is crucial. Training clinicians and public health workers in basic data management and interpretation ensures that data is correctly collected and used. Furthermore, ethical and privacy safeguards must be implemented through clear guidelines to protect patient confidentiality and maintain public trust. -
Challenges & Simple Solutions for EMR,
1) Missing Data: Use consistent data entry rules and automatic checks to reduce gaps. Structured data set and control variables not to skip function in the application
2) Data Quality: Apply validation tools and regular audits to keep records accurate, EMR also needed to conduct quality check and feedback to improve the quality
3) Ethical & Confidential: Protect patient information with encryption and strict access controls like user level by level with password, admin, user, etc..
4) Volume: Store large datasets in scalable cloud systems that can expand easily and can delete or omit unnecessary information to reduce the size of dataset.
5) Variety: Standardize formats and use interoperability standards (like HL7 or FHIR) to combine different data types.
6) Velocity: Use real-time processing systems to handle fast data flow from devices and labs.
7) Training and supervision; Required skillset needed to train the all users and regular supportive supervision is recommended to improve the user capacity
-
You must be logged in to reply to this topic. Login here