Research Theory - Easy to Understand PDF

**Research theory -- enkelt forklart** [EBP: Clinical expertise, best evidence, patient values] **1. Categories of Clinical Questions** These categories include: - **Diagnosis**: Questions about how to properly identify a disease or condition. Example: "What is the accuracy of an MRI in di...

**Research theory -- enkelt forklart** [EBP: Clinical expertise, best evidence, patient values] **1. Categories of Clinical Questions** These categories include: - **Diagnosis**: Questions about how to properly identify a disease or condition. Example: "What is the accuracy of an MRI in diagnosing ligament tears?" - **Prognosis (the likely course of a medical condition)**: Questions about the expected outcome of a disease. Example: "What is the five-year survival rate of breast cancer? - **Therapy**: Questions about the best treatment for a condition. Example: "Is medication X more effective than medication Y in lowering blood pressure?" - **Etiology/Harm (cause)**: Questions about the cause of a disease or the effects of exposures. Example: "Does smoking increase the risk of lung cancer?" - **Prevention**: Questions about how to prevent a condition or disease. Example: "Does a low-sodium diet reduce the risk of heart disease?" **2. Preferred Study Designs for Different Questions** Each type of clinical question often has a preferred study design for answering it: - **Therapy**: **Randomized Controlled Trials (RCTs)** - **Diagnosis**: **Cross-sectional studies** - **Prognosis**: **Cohort studies**, where groups are followed over time, are preferred to understand the natural progression of a disease. - **Etiology/Harm**: **Cohort studies and case-control studies** - **Prevention: RCTs or cohort studies** can be used to assess preventive measures. **3. Primary and Secondary Studies** - **Primary Studies**: These involve original research and data collection, typically from experimental or observational studies. Examples: - **Randomized Controlled Trials (RCTs)**: Participants are randomly assigned to intervention or control groups, helping to establish causality. - **Cohort Studies**: Follow groups over time to study the effect of an exposure. - **Case-Control Studies**: Compare patients with a disease (cases) to those without (controls) to find a casual factor. - **Secondary Studies**: These synthesize data from primary studies, providing a broader overview of evidence. Examples: - **Systematic Reviews**: Summarize the findings of multiple studies on the same topic. - **Meta-Analyses**: Statistically combine data from several studies to provide a more precise estimate of effect size. **4. Experimental Designs** Experimental designs are crucial in clinical research, and some are considered more rigorous than others based on their ability to minimize bias. - **Randomized Controlled Trials (RCTs)**: This is the most rigorous design, often called the \"gold standard.\" Participants are randomly assigned to different groups, which minimizes selection bias and allows for strong conclusions about causality. - **Crossover Studies**: In this design, participants receive multiple treatments in a random sequence. This allows each participant to serve as their own control. - **Quasi-experimental Studies**: These lack randomization but still involve an intervention. They are considered less rigorous than RCTs because of potential biases. - **Observational Studies**: These are non-experimental designs where researchers observe outcomes without intervening. Examples include: - **Cohort Studies**: Follow groups over time to measure outcomes. - **Case-Control Studies**: Retrospectively compare participants with and without a condition. **5. Hierarchy of Evidence** **- Causality: Causality refers to a direct cause-and-effect relationship between two events, where one event (the cause) leads to another event (the effect). For example, smoking causes lung cancer.** **The independent variable** is the *presumed **cause*** and the **dependent variable** is the ***potential effect.*** **Conditions necessary to determine casuality:** - **Emperical association**: Finding an association between dependent and independent variable. - **Time order**: The cause needs to proceed the effect. Cases need to be exposed to a variation in the independent variable before variation in the dependent variable - **Nonspuriousness**: A relationship between two variables is not due to a variation in a third variable. **Not because of confounding.** In research, there is a hierarchy of evidence, which ranks study designs based on their reliability and validity. From highest to lowest: 1. **Systematic Reviews and Meta-Analyses (Level 1A)**: These provide the most reliable evidence because they summarize findings from multiple studies, reducing bias. 2. **Randomized Controlled Trials (RCTs) (Level 1B)**: These are the best individual studies for determining cause and effect. Therapeutic. A prospective trial with one or more interventions. 3. **Cohort Studies (Level 2B)**: These are strong for studying prognosis and etiology but are less powerful than RCTs in establishing causality. A prospective look on a disease over time. 4. **Case-Control Studies (Level 3)**: Useful for rare conditions, but prone to bias since they are **retrospective.** 5. **Cross-sectional Studies (Level ?)**: These provide a snapshot in time, often used in diagnostic research, but they can\'t establish causality. **A study of a given population at one point in time.** 6. **Case Reports and Case Series (Level 4)**: These describe individual or small groups of cases and provide insights into rare conditions but lack generalizability. A grouping of similar case reports. [Scientific Orientation: Diagnosis]  **Diagnostic accuracy: Proportion of people with accurate test outcome** - TP+TN/ all (observed agreement) = A+D/A+B+C+D **1. Prevalence** - **Definition**: Prevalence refers to the proportion of individuals in a population who have a specific disease or condition at a particular point in time. - **Formula**: TP + FN/ All results - **Interpretation**: A higher prevalence means the disease is more common in the population. For example, if the prevalence of diabetes in a population is 10%, it means 10 out of every 100 individuals have diabetes. **2. Sensitivity (True Positive Rate)** - **Definition**: Sensitivity measures how effectively a test identifies those who actually have the disease (true positives). - **Formula**: TP/(TP+FN) = A/A+C - **Interpretation**: A highly sensitive test correctly identifies most people with the disease. For example, if sensitivity is 90%, the test detects 90 out of 100 people who truly have the disease. **3. Specificity (True Negative Rate)** - **Definition**: Specificity measures how effectively a test identifies those who do not have the disease (true negatives). - **Formula**: TN/(FP+TN) = D/B+D - **Interpretation**: A highly specific test correctly rules out most people without the disease. For example, if specificity is 95%, the test correctly identifies 95 out of 100 people without the disease. **4. Positive Predictive Value (PPV)** - **Definition**: PPV is the probability that individuals with a positive test result actually have the disease. - **Formula**: TP/(TP+FP) = A/A+B - **Interpretation**: A higher PPV indicates that when the test is positive, it's more likely the person actually has the disease. For example, if the PPV is 85%, there is an 85% chance that a person with a positive test result truly has the disease. **\>10 is good.** **5. Negative Predictive Value (NPV)** - **Definition**: NPV is the probability that individuals with a negative test result truly do not have the disease. - **Formula**: TN/(TN+FN) = D/C+D - **Interpretation**: A high NPV means that a negative test result is very likely to indicate the absence of the disease. For example, if the NPV is 90%, then 90% of those with a negative test result truly do not have the disease. **SpIN: Specific test when Positive rules IN the disease. If Specificity is high, a positive test will rule the disorder IN.** - High specificity - Low rate of false positive - High PPV  **SnOUT: Sensitity test:** **If Sensitivity is high, a negative test will rule the disorder OUT** - High sensitivity - Low rate of false negative - High NPV **Roc Curve: numerical** - The ROC curve shows how well your test works. - It helps you find the sweet spot between catching the right answers (true positives) and avoiding false alarms (false positives). - AUC tells you how good your test is overall: Closer to 1 is great; closer to 0.5 is like guessing randomly. **6. Kappa Statistic (κ)** - **Definition**: The Kappa statistic measures the agreement between two raters or diagnostic tests, beyond what would be expected by chance. It assesses the reliability or consistency of the results. - **Formula**: - **Interpretation**: Kappa values range from -1 to 1: - A value of 1 indicates perfect agreement. - A value of 0 indicates agreement is no better than chance. - Negative values suggest disagreement. For example, a Kappa value of 0.7 indicates substantial agreement. **7. Correlation** - **Definition**: Correlation measures the strength and direction of a linear relationship between two variables. The most used measure is **Pearson's correlation coefficient (r)**. can be used for intra/ inter - **Formula: øverst: covariance, nederst SD** - **Interpretation**: - **Strong correlation: 0.5-1.0** - **Moderate correlation: 0.30 -- 0.49** - **Low correlation: 0.10 -- 0.29** For example, a correlation of 0.8 between two variables indicates a strong positive relationship, meaning as one increase, the other tends to increase as well. **Positive Likelihood Ratio:** Sensitivity/1-specificity 1: no info \>10: is good **Negative Likelihood Ratio:** Specificity/1-sensitivity 1: no info \ 0.05) suggests there is not enough evidence to reject the null hypothesis. - **Estimates and Confidence Intervals (CIs)**: - **Estimates**: These are the point estimates (e.g., mean difference, odds ratio) derived from the data. - **Confidence Intervals (CIs)**: The probability of a result being representative for a global population (macro) - 95% CI contains 0 - -- do not reject H0 - -- no statistical significant different - -- result is based upon coincidence - A confidence interval gives a range of values that is likely to contain the true population parameter with a certain level of confidence (usually 95%). - Can be influenced by sample size, and variation in the population. - Example: A 95% CI of (2.5, 6.0) means we are 95% confident that the true mean difference lies between 2.5 and 6.0. - **Measures of Central Tendency**: - **Mean**: The arithmetic average. It is sensitive to outliers and is used when data is normally distributed. - **Median**: The middle value when data is ordered. It is useful for skewed data as it is not influenced by outliers. - **Mode**: The most frequently occurring value in the data. It can be useful for categorical data or multimodal distributions. - **Measures of Variability**: - **Minimum and Maximum**: The lowest and highest values in the data set. They give a sense of the range of the data but are sensitive to outliers. - **Interquartile Range (IQR)**: The range between the 25th and 75th percentiles, which represents the spread of the middle 50% of the data. It is less affected by outliers. **Test:** Boxplot - **Standard Deviation (SD)**: A measure of the average spread of data around the mean. It is most appropriate for normally distributed data and indicates the dispersion of the data points. - **Confidence Intervals**: As discussed above, CIs provide a range for the estimate of a population parameter, allowing for uncertainty in the point estimate. **Central Tendency**: **Criterion validity:** Compares test results with internal criterium. **Construct validity:** Compares test results with expectations based upon theory **Content validity:** To which degree the test covers all aspects of the topic. [Scientific Orientation: Therapy] **1. Confounding: cause surprise or confusion in (someone), especially by not according with their expectations. Third variable.** - **Definition**: A confounder is a variable that influences both the independent variable (what is being studied) and the dependent variable (the outcome). It can distort the true relationship between the variables being studied. - **Example**: In a study on smoking and lung cancer, age could be a confounder because older people may smoke more and are also more likely to get lung cancer, independently of smoking. - **Impact**: Confounding can lead to false conclusions about cause and effect if not properly controlled. **2. Bias: at du lener mer til noe. Subjektiv mer enn objektiv.** - **Definition**: Bias refers to any systematic error in the design, conduct, or analysis of a study that leads to incorrect estimates of the association between the exposure and the outcome. - **Types of Bias**: - **Selection Bias**: Occurs when participants selected for the study are not representative of the general population. - **Observer bias**: the tendency of observers to see what they expect or want to see, rather than accurate facts - **Recall bias**: hukommelsesbias, der studiedeltakere husker fenomener som undersøkes i forskjellig grad - **Response** **bias:** a term for various tendencies for participants to respond inaccurately or falsely to questions in self-report research - **Information Bias**: Happens when there are errors in measurement or classification of the variables. - **Publication Bias**: Studies with significant results are more likely to be published, skewing the literature. - **Performance Bias**: the effects of unequal treatment between study groups. As a result, study participants alter their behaviour. F.ex Hawthorne. - **Attrition Bias**: the selective dropout of some participants who systematically differ from those who remain in the study. Some groups of participants may leave because of bad experiences, unwanted side effects, or inadequate incentives for participation. - **Impact**: Bias can undermine the validity of study results, leading to incorrect conclusions. **3. Randomization** - **Definition**: Randomization is the process of assigning participants to different groups (e.g., treatment vs. control) in a random manner to reduce bias. - **Purpose**: It ensures that known and unknown confounders are evenly distributed across groups, minimizing their effect on the outcome. - **Impact**: Randomization increases the internal validity of a study by reducing the likelihood that confounders will bias the results. **4. Blinding** **- Allocation: how things are divided, and used.** - **Definition**: Blinding is the practice of keeping study participants, and sometimes researchers, unaware of which group (treatment or control) participants are in to reduce bias. - **Types of Blinding**: - **Single-Blind**: Only the participants are unaware of their group allocation. - **Double-Blind**: Both participants and researchers are unaware of group allocation. - **Triple-Blind**: Participants, researchers, and data analysts are all blinded. - **Impact**: Blinding reduces **performance bias** and **detection bias**, leading to more accurate results. **5. Follow-Up** - **Definition**: Follow-up refers to the ongoing monitoring of study participants over a period of time to collect data on outcomes. - **Impact**: A good follow-up ensures comprehensive data collection and helps measure the true effects of an intervention or exposure. **6. Drop-Out** - **Definition**: Drop-out occurs when participants leave a study before it is completed. - **Impact**: A high drop-out rate can reduce the study's power and lead to **attrition bias**, where those who remain in the study are systematically different from those who left, potentially skewing the results. **Interpreting the Outcome of a Scientific Study** **7. Internal and External Validity** - **Internal Validity**: Refers to the extent to which a study's results are due to the treatment or exposure being studied, and not to other factors (such as bias or confounding). - **High Internal Validity**: The study design effectively rules out other explanations for the outcome, meaning that the observed effect is likely due to the independent variable. - **External Validity**: Refers to the extent to which the results of a study can be generalized to other populations, settings, or times. - **High External Validity**: The study's findings are applicable beyond the specific conditions of the study. - **Example**: If a drug trial has high external validity, its results can be applied to a broader population, not just the trial participants. **8. Placebo** - **Definition**: A placebo is an inactive substance or treatment used in controlled experiments to test the efficacy of another treatment. It helps to determine whether the effect of the treatment is due to the **treatment itself or participants\' expectations.** - **Placebo Effect**: Participants may experience improvements simply because they believe they are receiving treatment, even if they are only receiving a placebo. - **Impact**: Placebo groups are crucial in randomized controlled trials to control for psychological effects and isolate the actual effects of the treatment being tested. **9. Maturation** - **Definition**: Maturation refers to natural changes in participants over time that may influence the outcome of a study, independent of the treatment or intervention. - **Example**: In a long-term study of children's cognitive development, improvements could be due to natural brain development rather than the educational intervention being tested. - **Impact**: Maturation can be a confounder in longitudinal studies, leading to incorrect attributions of the effect to the intervention. **10. Hawthorne Effect** - **Definition**: The Hawthorne Effect o ccurs when participants alter their behaviour because they know they are being observed, which can influence study results. - **Example**: In a study on physical activity, participants may exercise more frequently just because they are aware they are part of a study, not because of the intervention. - **Impact**: This effect can inflate the perceived efficacy of an intervention, leading to biased results. **11**. **Test-effect** - Learing effect caused by using the same test twice. **Chi-Square: Relative Risk** **RR = 1 no effect** **RR \< 1 protection** **RR \> 1 harm**