Test Accuracy & Study Types

Defining Test Accuracy

• Sensitivity measures a test's ability to correctly identify positive results when the condition is truly present.
• Specificity measures a test's ability to correctly identify negative results when the condition is truly absent.

Types of Studies

• Primary studies collect new data directly from individuals.
  • Key questions include identifying possible source populations and determining if participant recruitment is feasible.
• Secondary studies analyze existing datasets or health records.
  • Important considerations involve identifying usable data files and potential research questions that can be explored.
• Tertiary studies review and synthesize existing literature on a specific topic.
  • Researchers need to assess the availability of adequate library resources to access necessary articles.

Correlational (Aggregate) Studies

• These studies explore the relationship between two variables and the strength of that relationship.
• They are often cross-sectional, examining data at a single point in time, but existing data can also be used.
• Example: Examining lung cancer rates in regions with high versus low smoking rates using population data.

Case-Control Studies

• Case-control studies are backward-looking, starting with individuals who have a disease (cases) and comparing them to those who do not (controls).
• Groups are divided based on outcome status, and past exposures are examined.
• The outcome involves comparing exposure histories between the two groups to find associations.
• Strengths include being faster and cheaper, making them suitable for studying rare diseases or conditions with long latency periods.
• A weakness is the inability to measure incidence or prove causation, and a susceptibility to bias, such as recall bias.
• Example: Identifying 100 people with lung cancer and 100 without, then asking both groups about their smoking history to determine if smoking is more common in the lung cancer group.
• Contingency tables are used to evaluate relationships, followed by Fisher's exact test to interpret the p-value and odds ratio (OR).

Cohort Studies

• Cohort studies examine the relationship between exposure and the development of disease over time, and they can be retrospective or prospective.
• Participants are followed over time to calculate the rate at which new diseases occur and to identify risk factors.
• The direction can be forward in time, starting with people without the outcome and following them to see if they develop it.
• Groups are divided by exposure status and then observed to see who develops the disease.
• The outcome measures the incidence rate of new cases in each group.
• Example: Starting with smokers and non-smokers, following them for 10 years, and monitoring who develops what.
• Cohort studies measure incidence and establish a sequence of events.
• Strengths include providing strong evidence of potential causal relationships by tracking exposure over time.
• Weaknesses include being time-consuming, expensive, and not ideal for rare diseases.

Cross-Sectional/Prevalence Studies

• Cross-sectional studies measure the proportion of members in a population who have a specific exposure or disease at one point in time.

Case Series Studies

• Typically describe a group of individuals exhibiting the same illness

Best Use

• Case-control studies are often best for identifying risk factors for uncommon disease

Experimental Study Parameters

• Experimental studies are the gold standard for assessing causality and involve determining if there is a causal and effective relationship between variables.
• They are commonly used for clinical trials.
• The design uses statistics to quantify the impact of an assigned exposure on the likelihood of a positive or negative outcome.

Statistical Considerations

• Must know contingency tables and how to complete them
• Must know assumptions to complete table and calculations
• Need to know the differences between the tests
• The interval containing the population mean 19 out of 20 times is the 95% Confidence Interval (CI).
• Meta-analysis can increase statistical power by calculating a pooled statistic.

Sample Populations

• The study population consists of eligible members of a sample population who consent to participate in the study.

Study Bias

• Sampling bias occurs when study participants are not representative of the larger source population.
• Cohort studies may suffer from loss of follow-up.
• Case-control studies may suffer from recall bias.

Observers

• When calculating Cohen's Kappa coefficient, both expected and observed agreement types are needed

Case Control Statistical Measures

• Odds ratio

Cross Sectional Statistical Measures

• Prevalence

Cohort Statistical Measures

• Incidence rate ratio (relative risk)

Primary Studies

• Require recruitment of study participants

Study Goals

• Observational designs, seek to understand populations and describe patterns rather than focusing on research causality

Case Control

• Can suffer from a lack of generalizability

The Ecological Fallacy

• The ecological fallacy should be avoided in coronal studies
• The term indicates the false assumption the a person follow's a trends derived from population averages

Coefficient of Determination

• R^2 is the coefficient of determination, detailing how strong a correlation is, not indication direction of association- measured between (0-1 range)

The PPT definitions

• The PPT's - Person, place and time is an important definition to extract when establishing a case series

Co-Authorship

• Authors must make substantial contributions to the study to get co-authorship, as well as intellectual contribution

Aims of Study

• A realist synthesis systematically studies why some programs succeed with others fail

EDP

• EDP (exposure, desease, population) helps generate topic question with an EDP formula

ROC curve

• ROC (receiver operating characteristics) curve graphically plots false positive rates against true positive for diagnostics

Blinding helps minimizes

• Infromation bias

Quasi-Experimental Design

• Quasiexperimental assigns participants to intervention or control using non random methods

A natural experiment

• Natural experiments are when independent variables are not manipulated, but changed due to external forces

Equipoise

• Experimental research should only be conducted with genuine uncertainty on which treatment will work best

Efficiency

• Efficiency - is a cost effectiveness analysis based on wealth

Pre-proposals

• A brief research plan before the organization wants the full submission

Authorship

• To be listed as an author, you must make substantial contributions as well as provide intellectual analysis

Types Of Validity

• Types of vaidity are listed as: Content validity, Face validity , Construct validity. Criterion (concrete) validity

Meta Analysis Aims

• Meta analysis aims to enhance the understand of the desired/study phenomenon

Observational Study

• In an observational study, participants are not exposed directly to conditions, and behaviors are not directly changed during process

Key Informants

• Key Infromants in qualitative studies are there for there relevant expertise to the study question

Stratified Sampling

• Helps with equal chance of different clusters in population in study

Qualitative Studies

• Qualitative studies, test questions that dont require the measurement of central tendency and spead

95% Relevancy

• A study reporting that Cl to be 95 also cannot show chance, must be exact and define all conditions

Primary Data Studies

• Primary data studies, involves the need to recruit direct study participants

Gold Standard

• Efficacy, success, scalability, or the key concepts that are important to be studies

Logistical Regression

• Allows to be used for data that can be discreet or continuous at times

Spurious

• If the data is extraneous and the variable is not tested or correlated, it is spurious information

Research Designs

• What the best ways to categorize if there was noise during the study This is a logistics guide

Co-efficient

Coefficient of determination helps study and test correlation, but no specific type