STIRS Framework Keywords

In order of their occurrence in the STIRS framework

Evidence- Information which support a conclusion. Evidence may be derived from research investigations, observations, or other methods and submitted for careful review with an opportunity to challenge the evidence. 

Information- Data is raw unorganized measurement. When data are processed, organized, structured, or presented in a given context so as to make it useful for a particular purpose, it is called information.

Reductionist approach- An approach to obtaining evidence that looks at one factor or variable at a time while controlling for or taking into account other factors or variables.

Study and control groups- A study group is made up of participants or objects which have the factor under investigation or are exposed to the factor being investigated. Control or comparison groups in contrast do not have the factor under investigation or are not exposed to the factor being investigated.

Inductive logic- Inductive logic starts with observations and develops hypotheses for testing based on these observations. That is, inductive logic goes from the specific to the general.

Deductive logic- Deductive logic starts with accepted generalizations or theories and develops a hypothesis about a specific situation. That is, deductive reasoning goes from the general to the specific.

Intervention- A general term encompassing a wide range of human action aimed at improving outcome.

Efficacy- The extent to which an intervention produces a beneficial effect when assessed under the ideal conditions of an investigation.

Integrative approach - Obtaining evidence by examining multiple factors or variables at the same time and examining their influence on outcomes as well as the interactions that occur between factors.

Multiple influence or determinants- Influences or determinants are factors, often occurring long before the outcome, which eventually affect the probability of an outcome. The terms influences and determinants are often used with an integrative approach and have been referred to as the "causes of causes".

Interactions- Interaction occurs when the probability of an outcome resulting from the presence of one variable is altered by the level of a second variable. Interaction implies that the probability of an outcome is either less than or more than expected from adding together the impacts of each factor. (Synonym: effect modification)

System- A group of parts combined to form a whole that functions as a unit.

Theories- A set of statements or principles devised to explain a group of facts which have been repeatedly tested or are widely accepted and can be used to provide explanations or make predictions about the occurrence of outcomes or phenomena.

Paradigm shift- A paradigm is a broad intellectual framework or theory used to analyze a disciplinary issue. Paradigm shifts may result in a new framework or theory that is developed when current frameworks or theories do not adequately explain phenomenon or outcomes which they are intended to explain or predict.

Net-effectiveness- A measure of outcome that reflects the benefits minus the harms.

Visual display of data- A term which refers to a large and growing number of methods for presenting data in visual formats e.g. charts and graphics, geographic information systems.

Statistical significance testing- A statistical technique for determining the probability that the data observed in a sam­ple, or more extreme data, would occur by chance if there is no true difference or association in the larger population (i.e., if the null hypothesis of no association or difference is true). (Synonym: inference, hy­pothesis testing)

Sample- A subset of a population obtained in order to draw conclusions or inference about the larger population from which the subset is obtained.

Population- A large group, not necessarily made up of individuals, which includes all the units which define the universe of interest e.g. all residents of a state, all high schools in the country.

P-value- In statistics the P-value is the probability of obtaining the observed data, or more extreme data if the null hypothesis is true.

95% Confidence intervals- In statistical terms, the in­terval of numerical values within which one can be 95% confident that the value in the larger popu­lation lies. (Synonym: interval estimate)

Type I Error- An error that occurs when data dem­onstrate a statistically significant result when no true association or difference exists in the larger population.

Type II Error - An error that occurs when the sample’s observations fail to demonstrate statistical significance even though a true association or difference actually exists in the larger popula­tion.

Differences- A measurement obtained by subtracting one measurement from another measurement

Proportion- A measurement which includes a numerator and a denominator that is a ratio, in which the numerator is a subset of the denominator. Prevalence is a proportion which measures the probability that a condition is present at one point in time.

Rates- A measurement which includes a numerator and a denominator in which the numerator is a subset of the denominator and there is also a time interval over which the numerator is measured.  Incidence is the rate at which new conditions develop.

Estimation- A measurement of the strength of an association or the magnitude of a difference between an independent and a dependent variable e.g. correlation coefficients, odds ratios, relative risks etc.

Adjustment-  Statistical methods used after the collection of data to take into account or control for the effect of known or potential confounding variables and interactions.

Confounding- A factor or variable that is different in the study group compared to the control group and the factor af­fects the outcome being assessed. Excludes factors included in the chain of causation.

Association - An association is present when the presence of a factor represented by an independent variable occur along with an outcome represented by a dependent variable more often than expected by chance alone i.e. a statistically significant relationship. Correlation may be used as a synonym for association or may imply a special type of association in which both the independent and dependent variable are measured using continuous data.

Independent variable- A factor or variable being measured to explain or predict the occurrence of a dependent variable.

Dependent variable- An outcome or endpoint which is measured for participants in the study and in the control group.

Multiple regression - A statistical method that allows simultaneous adjustment for a large number of potential confounding variable and interactions.

Analytical reasoning - The ability to understand the structure of relationships and to draw logical conclusions about that structure.  Requires the ability to reason deductively from a set of statements and rules or principles that describe relationships among persons, things, or events.

Logic models- A graphical or descriptive representation of cause and effect relationships between components of an intervention or program.

Intuitive thinking- Intuition is receiving input and ideas without knowing exactly how and where they came from. It utilizes multiple senses and thought processes to produce conventional as well as unconventional results.

Qualitative evidence-Investigations which use methods such as listening or observing to gather and organize data into patterns or themes.

Quantitative evidence-Investigations in which numbers are used to measure variables such as characteristics, concepts, or objects.

Experimental design- An interventional investigation using study and control groups which aims to ensure, often using randomization, that the study and control groups are similar or identical except for the factor being investigated.

Observational design- An investigation using study and control groups comparing participants with and without a factor by observing the occurrence of events rather than intervening. Observational studies may begin with study and control groups with and without a possible risk factor in a cohort study. They may also begin with study and control groups with and without a condition in a case-control study. Cross-sectional studies, in which the factor and the outcome are measured at the same point in time, measure the prevalence rather than the incidence limiting their contribution to studying etiology.

Inclusion and exclusion criteria- Defining a study population for an investigation requires delineating characteristics which permit acceptance of participants and prohibit acceptance of participants into an investigation.

Statistical power- The probability of demonstrating statistical significance in a subsequent investigation based on a proposed sample size assuming a difference or association of a defined size is present in the larger population. Statistical power equals (1-Type 2 error).

Assignment- The process by which participants become part of a study or a control group. Assignment may occur by observation or by randomization.

Randomization- A method of assignment in which in­dividuals or groups have a known, but not necessarily equal, probabil­ity of being assigned using a chance process to a particular study group or control group.  (Synonym: random assignment)

Pairing - A method for assignment in which similar participants in a study group and in a control group are treated as a unit and their outcomes are compared to each other and analyzed as a unit.

Assessment of outcome- The process by which the outcome or endpoint of an investigation is measured for participants in the study and also in the control groups.

Surrogate outcome- An outcome measurement which substitutes for the measurement of interest such as substitution of a test result for measurement of a clinically or socially important outcome.

Primary outcome- An outcome which is used as the basis for calculating the required sample size for an investigation and therefore is expected to be statistically significant if a difference or association is present in the larger population.

Masked- A participant in an investigation and the investigator(s) ideally are not aware of their group assignment when assessing the outcome. (Synonym: blinded)

Accurate- On average the results approximate those of the phenomenon under study. Without systematic error or bias in either direction

Precise- Without variability from measurement to measurement of the same phenomenon. Without random or chance error. (Synonyms: reproducibility, reliability)

Complete- An investigation is complete if all participants included in study and control groups are included in the measurement of outcome and the measurements of outcome are made with equal intensity.

Continuous data- Data which is measured on a scale with an unlimited number of equally spaced potential values or categories e.g. blood pressure or IQ.

Ordinal data- Data which is measured on a scale with a limited number of potential values or categories which have an inherent ordering from lowest to highest but the distance between the categories is not known or is not equal e.g. stages or cancer, stages of intellectual development

Nominal data- Data with a limited number of named categories which cannot be ordered e.g. race, hair color or with only two categories i.e. dichotomous e.g. dead/ alive, pass test/ fail test.

Multiple comparisons- When more than one outcome reflecting more than one hypothesis is examined, the investigator needs to take into account these multiple comparisons when determining whether the results are statistically significant. Lowering the P- value required to establish statistical significance is one method for taking into account multiple comparisons.

Interpretation - Drawing conclusions for participants or situations included in an investigation.

Contributory cause- A definition of cause and effect which implies that three definitive criteria for contribute cause have been established i.e. an association exists between an independent and a dependent variable, the "cause" precedes the "effect", and "altering the "cause" alters the "effect".  ". Communicable conditions use association, isolation, and transmission as a parallel set of criteria to establish contributory cause.

Supportive criteria- Criteria used to provide additional support for the existence of a contributory cause when the definitive criteria cannot be established. Supportive criteria commonly used to make evidence-based judgments include strength of the relationship, dose-response relationship, consistency, and the biological plausibility of findings.  (Synonym: ancillary criteria)

Subgroup analysis- An examination of smaller groups within a population defined by characteristics such as age, gender, race, socioeconomic status etc.  Subgroups should be defined prior to collecting the data rather than after the overall results are known i.e. pre-hoc rather than post hoc.

Extrapolation - Drawing conclusions for participants or situations not included in an investigation (Synonyms: generalizability).

Prediction- A special form of extrapolation in which the investigator extrapolates to a future point in time.  May also refer to efforts to develop a prognosis or predict outcome for one particular individual.

Ethical research principles- Conduct of research under internationally accepted principles designed for the protection of participants. These principles include respect for persons, beneficence, and justice.

Informed consent - Method used in human research which aims to ensure that participants in an investigation choose to participate or not participate based on adequate information on the potential benefits and harms of participation as well an understanding of the process that will occur.

Evidence-based problem solving- An approach to presenting problems and identifying and evaluating their etiology as well as potential interventions to improve outcome which relies to the extent possible on data from research investigations.  

Benefits - The desirable outcomes which result from an intervention.

Harms - The undesirable or adverse outcomes which result from an intervention. Harms are often referred to as risks. Technically, however, risk is a measure of the probability of an outcome either a benefit or a harm. Use of the term risk as a synonym for harm can cause confusion.

Systematic reviews - A systematic review is a critical assessment and evaluation of all available evidence including qualitative evidence that address a broadly defined issue. The investigators use a defined method of locating, presenting, and evaluating all available published and unpublished literature.

Meta-analysis -A series of quantitative method for potentially combining data from multiple research investigations which may allow investigators to draw conclusions which could not be obtained based on the data from any one of the investigations. 

Translational research- A framework for connecting multiple types of research from basic or laboratory research to their use in improving individual and population outcomes.

Heuristics or rules of thumb - Methods used in non-quan­titative or subjective decision-making that use only a portion of the potentially available data and thus simplifies the decision-making process.

Decision rules- A set of criteria often expressed as a formulae which is used to make a choice between different alternatives.

Maximizing expected utility- A decision rule in which the option with the greatest expect utility is selected. Expected utility incorporates measures of probability, utility, and timing.

Satisficing- A decision rule often used when complete data on the options is not available at the time of the decision. Satisficing implies selecting a good enough solution that satisfies the most important desired outcomes.

Utilities- A measurement of the value or importance placed on a particular outcome using a scale of 0 to 1 with 0 indicating immediate death and 1 equaling full health or wellbeing. Using this scale allows utilities to be multiplied by probabilities which are measured on same 0 to 1 scale.

Timing of the outcome(s)- The timing of occurrence of potential benefits and potential harms relative to the timing of the intervention. Outcomes which occur in the future need to be discounted or adjusted to take into account the timing of their occurrence.

Risk- A risk or chance is an option in which there is more than one possible outcome. When the probability of an outcome is 1 i.e. 100% or 0 i.e. 0% there is only one possible outcome. That is, the outcome is certain. When the outcome is certain a decision maker is not taking a risk or chance.

Risk-taking- Risk-taking can be defined as selecting an option which incorporates risk and has a lower expected utility rather than an option with a certain outcome and a higher expected utility.

Risk avoiding - Risk-avoiding is defined as selecting an option which includes a certain outcome and a lower expected utility rather than an option which incorporates risk but has a higher expected utility.

Reference point- Benefits and harms are often judged compared to a reference point. Reference points may be the current status quo but they often utilize past states of wellbeing. The choice of reference point can determine whether an outcome is viewed as a loss or a gain.

Decision analysis-A description of the decision options, the probability of specific outcomes, as well as the benefits and harms of each outcome. May be displayed and summarized as a decision tree.

Cost-effectiveness analysis- A type of decision analysis in which financial costs are included as well as the non-financial benefits and harms.

False negative - An individual whose results on a test are negative but who actually has the condition as determined by the definitive or gold standard.

False positives- An individual whose results on a test are positive but who actually does not have the condition as determined by the definitive or gold standard.

Gold standard test- The criteria used to unequivocally define the presence and absence of a condition. (Synonym: reference standard)

Sensitivity- The proportion of those with the condition, as measured by the gold standards, who have positive results on a test being investigated.

Specificity- The proportion of those without the condition, as measured by the gold standards, who have negative results on a test being investigated.

Predictive value of a positive test- The proportion of individuals with a positive test who actually have the condition as measured by the gold standard.

Predictive value of a negative test-The proportion of individuals with a negative test who actually do not have the condition as measured by the gold standard.

Systems analysis - A structured effort to understand the influences of multiple factors and their interactions on outcomes and how one intervention may influence other interventions or outcomes.

Bottlenecks - A bottleneck occurs in a system when there are barriers to implementation of an intervention which slow down or reduce the benefit or increase the harm of the intervention.

Leverage points- A leverage point occurs in a system when location(s) for intervention(s) are identified which produce large improvements in outcome.

Maximizing overall group benefit- Selecting an option which maximizes the benefits minus the harms for the group or population as a whole.  Doing the greatest good for the greatest number.

Distributional justice - Giving consideration or special importance to subgroup(s) which disproportionately experience the harms of an intervention. Unique vulnerabilities or social inequities may be justification for giving special consideration or importance to a particular subgroup.