Biostatistics at Hackensack Meridian Health

The PICO(T) mnemonic is helpful for developing your research question:

• P: POPULATION - who do you want to recruit/study?
• I: INTERVENTION - what is the treatment in your study (if none, then you are likely conducting an observational study which still requires a good research question)?
• C: COMPARISON - who is your reference population/group to which you will compare your treatment group (AKA control group; if none, then you are likely reporting descriptive statistics on your population)?
• O: OUTCOME - what is your primary variable of interest that you are measuring? It is best to use validated and reliable tools to measure this, if available.
• (T): TIME - how long will your study last with respect to data collection?
• At a minimum, PICO should be encompassed within your research question(s). For example, a PICO(T)-based research question might look like the following: “Does one probiotic drink per day (kombucha, 20 oz) reduce the risk of heart disease in adults diagnosed with major depressive disorder (MDD) compared to adults with MDD who do not drink kombucha?” In this case, P = adults with MDD, I = kombucha (20 oz, once per day), C = adults without depression, and O = heart disease.
• see this article for more information.

• Variables are essential components of research studies.
• Variables help operationalize concepts for data collection, e.g. to measure “depression” one may use a depression scale such as the Beck Depression Inventory as the primary outcome variable.
• There are different types of variables:
• Quantitative variables represent a (typically numeric) measurable quantity.
• These can be continuous (take on any value, e.g. height) or discrete (cannot take on any value, usually whole numbers, e.g. number of skin lesions).
• Qualitative variables represent (typically descriptive, categorical, or word-based) names, labels, categories, etc.
• These can be nominal/categorical (names or categories, e.g. gender, diabetes type) or ordinal (similar to categorical but they have an inherent order, e.g. college year).
• It is important to ensure that your variable type(s) are adequate to address your research question.
• It is also important to understand what type of variable(s) you are using since this will determine, in part, what type of statistical analyses are appropriate.
• See this article for more information.

• Conduct a thorough literature review in your field to see if a measurement tool exists for the variable(s) you would like to measure.
• You can also look at the links below for existing validated measurement tools:
• https://psychology-tools.com/
• http://lib.adai.washington.edu/instruments/
• https://consult.ucsf.edu/guidance/special-populations-measures
• If you cannot find a validated measurement tool, you may have to create your own survey.
• You can see our presentation for more information on creating surveys at this link.

• It is important to choose the correct research design based on your research question.
• Is your research question descriptive?
• descriptive studies:
• case reports
• case series
• cross-sectional studies to determine frequencies and/or distributions of disease
• Is your research question analytical?
• observational studies:
• cohort
• case-control
• cross-sectional studies to compare different groups or exposures
• experimental studies
• randomized controlled trials
• See this link for more information on study designs.
• Your study design will then inform what type of statistical analysis is appropriate.
• See this link for more information on types of statistical analyses.

• Data organization is important for analysis as well as interpretation of your results.
• Spreadsheets can help keep your data organized.
• some best practices for inputting data into spreadsheets include:
• be consistent (e.g. always using “female” and not using “female” as well as “F”, “Female”, or “f”)
• indicate missing data (e.g. with “999” or “NA”)
• choose meaningful and careful variable names (e.g. “glucose_30min_fasting” versus “30 min glu fst”, the latter of which contains spaces and unclear abbreviations)
• make sure each cell contains one piece of information (e.g. a cell containing the value “1” which corresponds to the Likert scale value of “very likely” versus putting “1 very likely” in the cell together. The latter data point cannot be analyzed as is
• create a data dictionary which would be a separate tab or document containing information about variable names, ranges, detailed information, measurement units, etc.
• keep a clean “raw” version of your data that is not touched; when doing minor calculations or creating new variables always create a working copy so that you don’t risk data loss in the original file
• for more detailed information on data organization see this link.

• See this link for recorded biostatistics videos.