Suggestion for dealing with challenges:
1. Missing data – to dealing with missing data, we have to explore at upstream of data, The data sources. Try to define what is the main problem of data missing such as some variable is out-of-date, some variables are hard to entry or understanding, some variables is not to be added at first. If we try to understand these problems, we could reduce missing data at downstream. However, missing data is inevitable. We could use statistical methods to deal with them like imputation or complete-cases analysis but be caution, we are introducing information bias into the result.
2. Selection bias – selection bias usually occurs when sample size is too small and specific in some of populations. To dealing with this issue, we could identify and validate data in our data and data in general population in which representative or not. We also increase sample size and recruit more eligible samples to reduce bias.
3. Data analysis and training – In general, big data could not use old-fashioned method to manipulate and analyzing. So, we have to provide infrastructure for data storage, manipulation, analyzing, and interpretation insight from data.
4. Interpretation and applicability of result – we need to use result to build an application or optimization of process in healthcare, or visualize result, and insight of our data.
5. Privacy – data encryption is the one of standard method to protect healthcare data, need to apply in all level of operation. Another issue is authentication and authorization management by using methods like 2FA, Access control, or logging issue

I would like to suggest some solutions for the challenges.

1. Missing data can be corrected by established strict rules or policies whether data can be obtained or not and those data that cannot be collected must provide cause in the the incident report for further prevention.

2. Selection bias, because of large amount of information, the researchers are able to select the best data that suited their research purposes, this led to bias result and abuse fundamental concept of big data that more amount of data could provide more accurate results. To prevent this type of problem, the researcher should conduct a study based on real representation of data.

3. Data analysis and training, if the dataset is too large or complex to be analyze by standard set of statistic and the study required more advance statistical analysis, the researcher should consult the statistician. Moreover, many modern packaged softwares provide comprehensive list of statistic to use and intuitive graphic user interface.

4. Interpretation and translational application of result, in my opinion, the researchers should conduct big data studies on real problem in healthcare or result from the studies can benefits current health practices.

5. Privacy and ethical issue, using technologies to assist study always have privacy and cybersecurity issues but implementing right policy and punctual practice can reduce risk of these problem.

To be able to receive the benefits of big health data- studying large-scale population, using advanced techniques for data analysis, and answering medical questions in a timely way, we need to overcome many challenges as the following;

– Missing data: missing data can reduce the statistical power of a study and can produce biased estimates, leading to invalid conclusions. There are many statistical methods to deal with missing data. However, the best way to deal with it is preventing the missing data problems by well-planned data collection and creating a general protocol for widely use as a standard of data collection.

– Selection bias: we need to confirm the results of observational research that use big data analysis by doing RCTs.

– Data analysis and training: Big data analysis can be complexity and Human Resources of health data science is inadequate. To overcome this challenge, we need to promote more training about big data analysis and build more capacity for health informaticians that can analysts big data in a effective way.

– Interpretation and translational application of results: if we want to use the results from the big data for medical practice, we need to ensure that we can scope the data and can explain the description of the dataset variables and associated metadata.

– Privacy and ethical issues: Basically, we need a committee to monitor the use of big data in ethical way and ensure the well-protection of patients’ data.

According to the big health data challenges, the article give some examples of such complexity issues including missing data, selection bias, data analysis and training, interpretation and translational applicability of results, and privacy and ethical issue. I would like to add some opinion and suggestion in each topics.

1. Missing data: This challenges is the most often problems that found when deal with data analysis, but hardest to protect. It directly effect to validity and reliability of the result. The authors suggest to use statistical method to solve the problem. In addition, I think good design data collection procedure that conclusion from the brainstormed of stakeholders also can reduce some omitted data. Because incomprehensibility between data collector (clinician, health care provider, laboratorian) and designer or informatician, listening the opinions from the stakeholders can provide essential information to create suitable and user friendly data collection procedure.

2. Selection bias: In addition from the using many statistic techniques, to reduce selection bias, study design is also important. Criteria for include subjects is beneficial tool to control many factors or variables that affect treatment outcome. However, the validity of the data depend on variables that you can define and control and randomized control trial (RCT) is necessary.

3. Data analysis and training: Knowledge of researchers and clinicians and the development of algorithms are important for big data analysis. To develop both of human resource and knowledge resource need encouragement from many part such as funding and policy support from the government, cooperation from university etc.

4. Interpretation and translational applicability of results: The results from big health data analysis need to integrated and applied with daily practice. For information utilization, researchers have to explain their works in easier to understand while practitioners also have to understand the basic knowledge and applied the results with daily practice. In my opinion, it necessary soft skills for both of them need to develop to full filled the gap.

5. Privacy and ethical issue: Cyber crimes is concerned from everyone. For big health data analysis, it is inevitable to access personal data but consent of owner is also important. Consent form should carefully develop. In the other hand, researchers who have responsible to protect the private data must have preventive measures to reduce risk of breaches as much as possible.

Big health data challenges

1. Missing data

1.1 Assign must input data in HIS (don’t blank)
1.2 Mapping to correct data field in many HIS have different filed data name that is same data filed.
1.3 Data type correction, in some type of data such as date time is very difficult to use together especially in Thai date.
1.4 Statistical correction such as data imputation.

2. Selection bias

2.1 Use as much as data from multiple site to analysis.
2.2 Data pre-analysis for evaluate quality of data and distribution of data.

3. Data analysis and training
Machine learning method

– multiple time for training and compare the result.
– try different multiple data source to compare result if result very different may due to missing data, selection bias
– compare to statistical method

Statistical method

– clean data before analysis
– blind in all process as possible.

4. Interpretation and applicability of result
Post analysis study for the result from big data analysis to evaluate result.

5. Privacy
Inform consent at point of service that include data analysis.

Big data analytics face certain challenges which lead to inappropriate disease management and could not defeat the burden of diseases. This article suggests some solution which is useful to overcome the challenge of big data management. However, I would like to add some suggestions and solutions as below:

1. Missing data: the common error in many systems is missing data. This error can indeed resolve with statistical methods but if the missing data over 10% of the total data, it cannot use statistical methods to solve the problem. To deal with the missing data issue, data standards and clear definitions will be involved in the data collection process.

2. Selection bias: a high volume of data set may lead to bias observations sometimes so to reduce a selection bias, hypotheses will be confirmed with RCT.

3. Data analysis and training: due to the large datasets, a single dataset analysis will be out and not powerful enough, thus the algorithms with multiple testing will be the solution for big data analytics. However, training in informatics, coding, data analysis, or other increasingly relevant skills to users will optimize the use of large volumes datasets.

4. Interpretation and Translational Applicability of Results: designing and defining the dataset will increase the possibility to integrate the between the different systems.

5. Privacy and ethical issue: maintaining security, data encryption, and access control will help protect the people’s identity in the server.

-Missing data
Establish policies for collecting the necessary information. Ask for cooperation in recording the necessary parts of the work.

-Data analysis and training
Data Analysis can use tools for Big Data Analysis. Use outsource expert help data analysis and additional training.

-Interpretation
In big health data, a model can be designed to integrate the results of the analysis. Using a data management tool.

-Privacy and ethical issue
Access controls including the restriction of access rights to networks, systems, applications, functions, and data.

After reviewing the provided article Big Health Data and Cardiovascular Diseases: A Challenge for Research, an Opportunity for Clinical Care by Alessandro Gialluisi et al., there are many challenges which can be occurred in three groups/steps; source identification, data processing and clinical application of results. However, I will provide my suggestions on how to dealing with those five challenges as mentioned in the article as follows:
1. Missing data: We need to consider about the main sources of missing data prior. The missing data might be occurred from various reasons as mentioned in this publication e.g., about non-formulated/ systematic of database in EHR. This will lead to the unqualified data analysis when calculate by statistic power. We should initial handle this problem by formulate the set of data collection that need. This data set will be finalized by brainstorming the ideas among the creators, users, colleagues, and stakeholders what they need from these data set. Then, the standard operating procedure (SOP) need to create and distribute to let them work rely on the SOP. Parallelly, the data verification and monitoring process need to be performed for checking the completeness and quality of data.
2. Selection bias: It can be reduced by well-plan establishing of the standard protocol of recruitment of subjects into the study projects prior. The increasing of sample size in each group of RCT study also need to consider reducing the selection bias of the study and benefit when perform the data analysis.
3. Data analysis and training: We can handle this problem by establishing the standardization of Data analysis for the project. The code of factors, acceptance criteria and algorithm of data analysis need to be standardized and need to make sure that everyone in the team clearly understand in the same manner. The SOP need to develop, and training plan need to be performed to ensure about this understanding. Additionally, regular monitoring needs to be done to ensure the consistency of the understanding.
4. Interpretation and translational applicability of results: To reduce the gap between team with clinicians who need the analyze data to treat the patients. We need to discuss and finalize among the team as mentioned in topic 1 by following the SOP/ protocol. After finalization, these data set need to integrate to the data system, and It should be easy to interpret.
5. Privacy and ethical issue: Inform consent to the patient and the data security (include private information, address, pictures etc.) in term of prevent the breaching of data need to be performed in every study. Additionally , recruiting more cybersecurity professionals to protect their data sold be considered as well and other steps taken for securing data include: Data encryption, Data segregation, Identity and access control, Implementation of endpoint security, Real-time security monitoring, Use Big Data security tools.

Suggestion for coping with challenges related to big health data:
1. Missing data
– Missing values of less than 10% can be managed with the application of statistical methods.
– However, it is best to avoid missing data starting from the data collection process. At the operational level, strict rules must be set on which values to be collected.
– Trainings are given regularly to the users of the EHR system to reduce entry errors.
– Moreover, the system itself must provides features for required fields, type of data entered, etc.
– Data accessibility polices must be well defined for availability of data from different sources.
– Data integrity is another component to be considered to reduce the number of missing values by intentional or unintentional modification of the data.
2. Selection bias
– For diminishing selection bias in big data analysis, which is often observational study, the participants selected for the study must represent the population. One of the steps that can be used is well defining inclusion criteria.
3. Data analysis and training
– Researchers should be trained on the knowledge of big data analysis tools. To make it possible, adequate funding is necessary.
– If the required skills are out of the knowledge of the researcher, expert opinion from other sources should be sought.
4. Interpretation and translational applicability of results
– In my opinion, new research SOP should be made available for clear and adequate description of the dataset variables and results without misunderstandings between researchers and the clinicians who would apply the results to the daily practice.
5. Privacy and ethical issues
– Appropriate data security measures must be integrated to prevent from cybercriminals.
– All patients must be informed of types of data collected and shared, the risks and benefits of data sharing, and only after that, consent must be taken.

In my opinion , the challenge for big health data for suggesssion as follows;

1.Missing data; In order to resolve the missing data problem, we may need to start by finding the cause of the missing data and what stage it is, perhaps it may be due to too many duplicate data. In order to brainstorm, solve problems, and develop methods that will help complete the information gathered.

2. Selection bias; The use of statistical methods is used to analyze a wide variety of data to help reduce data selection bias. It requires randomized trials (RCT) to help analyze the data.

3. Data analysis and training; Big data analysis requires knowledge, analytical skills and knowledge of personnel to be updated. The application of technology in particular to the job is very challenging.

4. Interpretation and Translational Applicability of Results; Utilizing the results from the analyzed data to be useful is an essential skill. Due to the dissemination of information or its use, it must be understood by both the sender and the receiver so as not to cause problems.

5. Privacy and ethical issue; Data privacy and security are important to maintain. Today’s technology is able to help protect and protect information from cybercriminals.

Missing Data: This issue is usually found in my experiences. And as for the big data, this would, even more, be a greater challenge for the data collection process at a high velocity with a high volume for a high number of variables. Anyway, there are some ways to address this issue. Firstly, there should be mandatory data fields to fill for key variables. Then the data collection policy should also be adopted and advocated to the data collectors, such as doctors, nurses, and pharmacists. Anyway, in spite of implementing these measures, there can also be missing data. Then these missing data will be treated with statistical methods such as single and multiple imputations, likelihood-based methods, and some advanced statistical models. And sensitivity analysis could also be used in some cases as well.

Selection Bias: As most of the big data analyses fall under observational studies, it bears some cons. Of these, selection bias would be an important issue to deal with. The selection bias could contribute to confounding effects that might interfere with the hypothesis. Though randomization would cure this problem, the big data analyses themselves undermine the randomization process. Anyway, it would be wise to check the indications from baseline imbalance before analysis. And advanced regression models can also be used for confounding variables that arise from selection bias.

Data analysis and training: As described in the paper, the clinicians are fine for the data analysis of the small datasets. Anyway, big data analysis requires a significant amount of analytical skills to get the job done. For that matter, professional data scientists should do this job. Or the clinicians and the medical personnel can also be trained for such matters.

Interpretation and translational applicability of results: There are always steps between research literature and clinical practices. One should not be discouraged by this problem. Anyway sometimes, there are much more challenges than anticipated as to translate the results of the big data analysis to action. Issuing standard operating procedure guidelines to guide the whole big data analysis steps. The other possible solution would be to encourage the big data culture and context in the clinical setting so that it gives a large community to hold seminars and workshops on a regular basis to discuss the interpretation problems and applicability of the results.

Privacy and ethical issue: Privacy is very fundamental, and always should be regarded as a high priority. In big data analyses, it is, even more, a broad challenge. To face this challenge, there should be policies and guidelines to ensure the privacy of the patients. And the SOPs should be adopted to guide the steps at the operation level. The security of both software and hardware should also be considered as important. And the rules and regulations should also be adapted to support the privacy and ethical issue from a legal aspect.

Please find suggestions for coping with challenges related to big health data as follow.
Missing Data – This is quite an uncommon issue either in traditional recording or electronic recording system. In term of big data, it may have negative impact for data analysis as it is high volume. We should start finding loophole for missing data. If the data collecting tool has loophole, it should be fixed immediately. Health care professionals who are filling data should have access to feedback system and should encourage by reward system to promote awareness on importance of valid data. However, as mentioned in the article, when missing data cannot be filled because of various factors, it is wise to use alternative analysis methods such as imputation techniques, mixed method regression model etc.
Selection bias
To reduce selection bias, it is better to include data from various sources as much as possible.
Data Analysis and Training
It is advisable that frequent trainings to researchers should be done to empower knowledge on analysis. To avoid complexity and errors, data cleaning should be done first followed by advanced analysis tools like logistic regression, latent class analysis, square-root transformations etc.
Interpretation and Translational applicability of results
As big data analysis itself has its own limitation, it might affect data interpretation for applicable results. Researchers from both clinical and public health background should involve in interpretation process by considering culture and context in multi-disciplinary approach.
Privacy and ethical issues
Data security must be set the first priority of all. All possible measures for data security preventing from cybercriminals must be taken in advance. Informed consent from beneficiaries must be taken.