Health Data Assessment

1. Create data management plan.

See instructions by the Academy of Finland for research projects. In business context, create a data strategy plan which involves business goals. See introduction for Data Strategy Plan here.

a. Plan what kind of data is needed and how it can be collected or retrieved.

Plan what kind of variables are needed to answer your questions and check if you can use existing datasets. Consult a domain expert and get familiar with previous studies.

b. Estimate how much data is needed to complete your project.

Consult a statistician or a data analyst to estimate the needed dataset size. If you want to use a machine learning approach, you generally need much more data than for a statistical approach. Template for assessing the amount of data coming soon.

c. Check if you can exploit already existing data, such as open access datasets.

See more about accessibility to health datasets in the Health data guide.

d. Figure out what is your role in data management (e.g. data owner or data processor).

European GDPR legislation involves these roles for recognizing your responsibilities, see more here.

2. Create a schedule for data collection and analysis.

This is needed especially if you need to collect data by yourself or identify and recruit commercial data analyst services or other third parties.

3. Define the required competences and availability of resources needed for data collection and analysis.

Data analysts are in high demand so take care to start resourcing as early as possible.

4. Identify needed software tools and analyze the targeted infrastructure, e.g. server, cloud.

For a compilation of software for data analytics, see here. For the needed IT infrastructure, consult the IT support of your organization

1. Check that you have needed consents from participants.

More information on consents here.

2. Check that the ethical aspects of your project are in place for gathering and processing.

More about ethical aspects on AI ethics guidelines by the European Commission.

3. Be sure that your data management follows GDPR rules, especially if person profiling and/or automatic decision making is involved.

More information about GDPR compliant processing of personal data can be found on tietosuoja.fi.

4. Identify legal roadblocks for data access and use, especially in business context.

In Finland e.g. asiakastietolaki, laki potilastietojärjestelmistä, laki lääketieteellisestä tutkimuksesta. Consult your organization's legal advisor for more information. See more Health Data Guide's Legislation chapter.

1. Check if your dataset contains enough variability (with respect to gender, age, ethnicity etc.) for generalization of the results.

Results obtained with a biased diaset may not be applicable to all groups of people.

2. Remove unnecessary variables from your dataset.

This will generally result in better (and faster) results in your data analysis, especially in predictive models.

3. Decide what to do with missing data values.

There are several methods for handling the missing data values. Check for example this article on how to compensate for missing values in a dataset.

4. Establish a plan on how to easily detect and (auto)correct erroneous values.

See guidelines here.

5. Describe metadata for your dataset and set it available for others (if you are producing new data).

Metadata describes your data. Some examples of basic metadata are author, date created, date modified, and file size.

6. Describe metadata for the variables in your dataset and set it available for others.

Metadata about variables should describe e.g. the units, labels, standards etc. See instructions here.

7. Use data labels that make sense to ‘anyone’ outside the project organization.

Use common language and abbreviations to make labels understandable to others.

8. Monitor and keep the quality and integrity of your data in good shape throughout the project — and especially in cases where the device changes.

It is a good practice to check the quality of the ongoing data gathering regularly.

9. Document your dataset gathering process and update it if changes occur.

Include at least the original purpose of the data, creators of the dataset and their contact information. Also describe how the data were collected and potential changes in the data collection routines.

1. Check if your gathered data should be accessible in an established (file) format for later use.

It is a good idea to store your gathered data in a format that is not too rare in your research or business domain- in order to ensure data reusability in future projects.

2. Establish user access protocol for your dataset and decide what you want to set available to others (whole or partial dataset, metadata or only results of analysis).

Contact your IT support for implementation of the user access control.

3. Establish a backup plan for your data.

Regular backups ensure that valuable data is not destroyed in case of hardware breakage.

4. Find a data officer for managing data issues (when GDPR requires it).

See more about GDPR requirement on tietosuoja.fi.

5. Decide how to handle the dataset after the current project - will the dataset be destroyed, stored permanently, or for a predefined time?

Keep personal data no longer than necessary for the original specified purpose, as GDPR requires. More information on European Data Protection Supervisor's Guidelines.

6. Decide whether your data should be open access after the project.

Check available open access license types here.