D514 Analytics

D514 TOPIC 1 Review Reading material: Statistics for Healthcare Professionals, Chapters 2, 3, 12, 13, 15–17, and 18 ● ANOVA test: Analysis of variance (ANOVA) may be used to compare research studies with two or more groups. ● Chi-square tests determine if an association exists between two categorical variables. ● Control group: In a healthcare environment, this group of patients does not receive the treatment being studied. ● Experimental group: This group of patients receives the treatment being studied with follow-up observation to determine the effect of the treatment. ● F-test: The F-test tests whether two population variances are equal. The ratio of the two variances is compared. If they are equal, the ratio of the variances will be 1. ● Frequency: Frequencies measure how often a particular value occurs to assess the importance of a value or check the variation of the values in a study. ● Hypothesis: A proposed explanation for an observation that leads to a prediction. Through investigation and the use of statistical data, those doing the study will either confirm or reject the hypothesis. Testing the hypothesis will show if there is a link (or not) between two or more variables. ● Integrity: Research always makes some assumptions, depending on the type of method used. Research assumptions must be identified to determine possible breaches of integrity. ● Interval data: Interval data includes units of equal size, such as IQ results. There is no zero point. An example of an interval scale is time: Time is measured in 24 hours daily; the time between each hour is the same, 60 minutes. ● Mean: Mean is the arithmetic average. Divide the sum of all scores by the total number of scores. ● Median: Median is the midpoint of the distribution of values, or the point above or below which 50 percent of the values fall. ● Methods section components: When analyzing the quality of a study, a careful evaluation of the research methods can reveal critical details about the population and sample, covariables and hypothesis, data presentation, statistical analysis, and study limitations. ● Misleading statistics: Interpreting and presenting the results of data analysis affords many opportunities for accidental or deliberate misrepresentations of data. Common examples include implying causation, extrapolating beyond the reasonable, relying on a biased or incomplete sample, and using inappropriate graphical representations. ● Mode: Mode is the value that occurs most frequently in the data. ● Multivariate regression analyses: Multivariate regression analyses can be used to analyze and adjust risk. This analysis model contrasts each measured factor to the patient’s risk of a particular outcome. ● Nominal data: Nominal data can be measured as a frequency or percentage, and the mean of these data cannot be calculated. Nominal data in healthcare might include demographic information about patients. The word nominal means "about a name." ● Ordinal data: Ordinal data can be measured as a frequency, and the mean of ordinal data is often calculated. Ordinal data in healthcare might include patient satisfaction surveys using a Likert scale. The word ordinal means to "put in order." ● Parametric and nonparametric tests: Parametric tests are based on probability distributions. Nonparametric tests are used when data are not normally distributed. ● Pearson's correlation: Pearson's correlation is used with interval and ordinal scale data and determines the extent to which a change in one variable tends to be associated with another change. ● Qualitative research methods: Qualitative research aims to understand perceptions, perspectives, interpretations, and opinions. Qualitative research methods often include questionnaires, interviews, written documents, observations, and focus groups. ● Ratio data: Divide one quantity by another, and you have a value. You will have a proportion, a percentage, or a rate.

D514 ● Reliability, validity, and analysis of questionnaires: Questionnaires can be evaluated for reliability based on their consistency (stability) or repeatability over time; questionnaires are valid if they measure or record what they purport to measure. Data from questionnaires may then be grouped according to nominal, ordinal, interval, or ratio data. ● Research: Research can inform decisions regarding the development and efficacy of new processes, systems, technologies, environments, and organizational structures to support operations. ● Research platform: Research is built on a platform of previous knowledge, the scientific method. ● Risk adjustment: Risk adjustment is essential for comparing data across systems, especially among patients with varying comorbid diseases and complex treatment modalities. Multivariate regression analyses can be used to analyze and adjust risk. This analysis model looks at each measured factor to the patient’s risk of a particular outcome. ● Risk of error and harm: Studies should include an analysis of any sources of error and a thorough explanation of the consequences associated with a particular study treatment. ● Sample size: The study's design provides insight into each variable's appropriate number and volume . The calculation of statistical confidence factors informs the validity testing of the study sample size . ● Standard Deviation: Standard deviation determines the amount of variance in a data set and evaluates the degree to which each case deviates from the average or mean. ● Statistical significance: If the null hypothesis is rejected, the observation is statistically significant. In a research study, the null hypothesis states there is no association between the independent and dependent variables in a study. ● T-test: The t-test helps the researcher compare whether two groups have different average values. A paired t-test is used when each observation in one group is paired with a related observation in another. ● Variables: The independent variable is the factor that the researchers directly manipulate. The dependent variable is the measurable variable that depends on the independent variable. Topic 1 Video notes: (Principles of Research Analytics) 24% of the exam ● Quantitative data: numerical, measurable (graphs/tables) ○ Example: Sample size: mean, median, mode; minimum, maximum; range, Standard deviation; quartiles ● Qualitative data: categorical data, subjective data, ○ Example: Types of cancers, insurance carriers, blood types, MD specialty ● Standard Deviation: Measures variability of data, difference of the average. ● P-value: Probability value. Results of research happen by chance. ○ Lover P value: the less probability: that the experiment was effective. ○ An example of a P value of 0.0452 is 4.52% results, which could have happened by chance. ○ Null Hypothesis : Idea or assumption: true until proven wrong.

D514 TOPIC 2 Review: Reading material: Population Health: Creating a Culture of Wellness , Chapter 1 ● Categorical variables: The value is in the name or label. Types of cancers (breast, skin, lung, etc.) are categorical variables. ● Confounding Variable: It obscures the effect of another variable. The researcher may initially believe the variable will influence the research, but findings show it does not. As a researcher, you may not be able to control its influence on the result of the research, but you should be aware that it impacts your results and not allow it to skew your data. ● Continuous variable: This is also known as an interval variable. There is a meaningful difference between values. An example is body temperature. ● Databases: A well-designed healthcare database captures data to support the organization’s analysis and comparison of safety, quality, effectiveness, efficiency, timeliness, and efficacy of actual care and services delivered to the patient over time. ● Data warehouses: These assimilate data from multiple transaction systems. Data warehouses can be used to distinguish more significant trends in data from multiple sources . ● Dichotomous variable: This is also called a binary variable. It occurs in one of two possible states, for example, male or female. The patient has cancer or does not. ● Disease registries: These are a hybrid between transaction systems and data warehouses. They are designed for tracking explicitly defined data at a case-specific level. Some examples are trauma

D514 registries to track emergency department data, cancer registries, immunization registries, and numerous others. ● Errors: Incorrect application of a statistical test can result in a type I error, which occurs when a null hypothesis is rejected when it should have been accepted. A type II error is experienced when the alternative hypothesis is rejected when it should have been accepted. ● Measurement and decision support: Measurement is used within an organization to monitor improvements in systems and processes through analysis of current performance trends, identify key opportunities, and consider leading practices informed by new research-based knowledge. Decision support provides an information platform to evaluate leading, lagging, and real-time performance measures. ● Outcome evaluation: An outcome evaluation focuses on the result of a specific program or initiative, generally clinically measured by improvements in morbidity, mortality, or vital measures of symptoms, signs, or physiologic indicators. ● Quality improvement: Quality improvement is measured internally and externally using various benchmarks and indicators. These indicators are quantified by proportions, percentages, ratios, means, medians, and counts to measure processes, perspectives, and outcomes aligned with a specific initiative or decision. ● Quality measures: Administrators receive numerous quality reports regularly in today's healthcare environment. These reports are generated for internal quality improvement projects, for mandated external reports to government agencies, and for compliance with accrediting body requirements. As healthcare's value-based purchasing model evolves, quality measures become pivotal operational "pulse checks" to administrators. ● Statistical testing and treatment: The basis of statistical testing is whether or not the study results have a proven relationship to a change in processes or care modalities. Statistically significant results do not automatically indicate clinical significance. ● Transaction systems: Transaction systems divide data according to individual operations. The data stored by transaction systems is granular and based on specialized systems.

D514 ○ Trauma Registry: Information system used to collect, analyze, and manage data on trauma patients and used to improve emergency care, track trends, and consider a disease registry. ○ Center of Medicare and Medicaid Services . (Division on DHHS): Develop healthcare policy and administer Medicare and the federal portion of Medicaid. ○ The National Institute of Health (part of DHHS), the nation’s medical research agency, finds clinical trials and a list of registries. ○ NPDB : National Practitioner Data Bank: Web-based repository of reports containing information on medical malpractice payments and certain adverse actions related to healthcare practitioners, providers, and suppliers. Used when onboarding physicians or making inquiries. ○ NCHS : National Center of Health Statistics: Operates through the CDC vital statistic data it collects from birth and death records, medical records, interview surveys, and through direct physical examinations and lab testing. ( https://www.cdc.gov/nchs/index.htm ) ○ CAUTI: Catheter-Associated Urinary Tract Infection ○ MPI: Master Patient Index: assigns a unique identifier to each patient admitted to the hospital. ○ AHRQ: Agency for Healthcare Research and Quality: To produce evidence to make health care safer, higher quality, more accessible, equitable, and affordable. (USDHS) ○ HCAHPS: Hospital Consumer Assessment of Healthcare Providers and Systems: Patient satisfaction survey required by CMS for all U.S. hospitals. ○ HRSA: Health Resources and Services Administration: Agency of the DHHS primary federal agency for improving healthcare for people geographically isolated, economically or medically vulnerable. ○ MEPS: Medical Expenditure Panel Survey: Large-scale survey of families, individuals, medical providers, and employers. Most complete data source on the cost and use of healthcare and insurance coverage. ○ CLABSI: Central Line-Associated Bloodstream Infection. ● Research Studies: ○ Prevalence Study: Cross-Sectional: Analyzes data collected from a subset of patients at a specific point in time or one geographic area. ○ Predictive Study: Forecasts outcomes, consequences, effects, and cost. Analyzes existing data to make predictions . ○ Case Study: Investigate one person, organization, and group in depth. A non- participant observational researcher does not interfere with the subject. (Natural setting) ○ Longitudinal Study: Data gathered on the same subjects over a lengthy period . Can detect changes in subjects( look for cause and effect.) ● Variable: Factors or features that may vary or change and have an influence on the research ○ Categorical: Value is in the name or labe l ■ Types of cancer ○ Confounding: It obscures the effect of another variable; the researcher may initially believe the variable will influence the research, but the findings show it does not. You may not be able to control the influence on the result. ○ Continuous: Known as an interval variable. There is a meaningful difference between values. ■ Body temp ○ Dichotomous: called Binary: occurs in one of two possible states ■ Male or female ■ Cancer or does not

D514 ○ Ordinal: Order matters, but not the difference between values ■ Likert scale, subjective ■ Birth order. Children are ranked according to the order in which they are born, but the years between each child may vary. ● Descriptive Statistics: ○ describes and analyzes a drive group without concluding and does not make inferences about a larger group. ○ Example: Data is enumerated, measured, organized, graphed, showing %, mean, median, mode, quartiles, and range ● Inferential Statistics: ○ It concludes the large population and draws a conclusion based on data from a sample. makes educated guess ■ Anova. T-test, regression, etc. ● Research Design: The type of study that is being done. How the research is being constructed. (longitudinal study, case study, exper study.. study) to plan to achieve the researcher’s purpose. ● Study Population: Who is being studied ? This could be persons, organizations, groups ● Data Collections: Where is the data coming from to support the research? Sources of data(patient, MR, Financial reports, databases, surveys, focus groups, etc.) ● Analysis plan: A statistical test is performed after the data is gathered to evaluate and analyze the data. It may also be called Study VALIDITY because the analysis proves if the research is valid.

D514 provides an information platform to evaluate leading, lagging, and real-time performance measures. ● Measuring the effectiveness of treatments: Statistics are necessary to measure and compare treatment outcomes. Statistically analyzing the effectiveness of treatments is the optimal method to determine validity for adoption. ● Multiple regression: Multiple regression aims to determine the relationship between several independent or predictor variables and a dependent variable. ● Needs assessment: A needs assessment collects and analyzes information about a specific population, enterprise, or cohort to gain stakeholder insight into cultural engagement. It may also identify coalition strengths, weaknesses, opportunities, issues, available resources, and constraints or barriers. The needs assessment supports clear direction for decisions involving the development of a specific health initiative or program. ● Outcome evaluation: An outcome evaluation focuses on the result of a specific program or initiative, generally clinically measured by improvements in morbidity, mortality, or vital measures of symptoms, signs, or physiologic indicators. ● Plagiarism Avoidance: Plagiarism is the uncredited use of someone else’s words or ideas. A charge of plagiarism to a researcher may have serious consequences, including loss of a job or expulsion from a university. It will result in a loss of standing in the professional community. ● Research: Research can inform decisions regarding the development and efficacy of new processes, systems, technologies, environments, and organizational structures to support operations. ● Quality improvement: Quality improvement is measured internally and externally using various benchmarks and indicators. These indicators are quantified by proportions, percentages, ratios, means, medians, and counts to measure processes, perspectives, and outcomes aligned with a specific initiative or decision. ● Transaction systems: Divide data according to individual operations. The data stored by transaction systems is granular and based on specialized systems. Topic 3 Video Notes:( Healthcare Management Decision-Making) 28% of the exam ● Risk Stratification: ○ Stratifying is separating patients into groups to identify patients/populations needing more care. ○ By creating groups with similar risks, healthcare leaders can manage the utilization of care and resource utilization. ○ After stratification, the population can be narrowed down by diagnosis, provider, care management program, level of care, utilization of services ○ Track over time: is care management successful? Calculating risk scores fall under

D514 “predictive analytics.” ● Research design: How should we construct our study? ● Study population: Who are we studying? ● Data collection: What sources do we use to gather our data? Where will we find the data? ● Analysis: What statistical test do we perform to evaluate our data and obtain results? ● Study validity: Providing the research results is what was anticipated. ● Ethical Considerations: ○ Fidelity: Loyalty, truthfulness, advocacy to pt. Faithful to your professional promises. ○ Veracity: Habitual truthfulness ○ Beneficence: the welfare of helping the patient is the goal. (I will help you) ○ Nonmaleficence: Not harming or inflicting the least possible harm to reach beneficial outcomes. (I will not harm you) ○ Justice: Fair selection of research participants ○ Respect for Autonomy: patient have the right to make their own decisions. Unless deemed legally unable ○ Plagiarism avoidance: honesty in using another’s research . ● The most frequently used data collection methods would include-> Observation: watching, taking notes, subjective -> Surveys: email, in-person, phone, structured questions -> Interview: Meeting face-to-face, structured questions -> Focus groups: guided discussion -> Experimental study: ANOVA, T-TEST,ETC -> Secondary Data Analysis/archival study -> Mixed Methods(combo of some of the above) ● Types of Research Studies: ○ Case Study: Non-participant observation of one person, group, and organization. In- depth. ○ Cohort Study: The same group of participants followed over time, which could be a prospective or retrospective—longitudinal study. (Common characteristics such as same conditions, exposure to event resulting in medical condition, genetic predisposition) ○ Meta-Analysis: Reading and analyzing a group of research studies, combining the research findings to decide for your organization. It is the combination of studies that will facilitate an evidence-based decision. ○ Literature Review: Reading various articles to build and inform a research report or further your education. Retrospective Prospective This means Looking backward (into the past) This means Looking forward (Into the future) A study is a type of cohort study that analyzes two groups of people: those with the disease under study and very similar groups of people who do not have the disease. A study is a type of cohort study where the researcher enrolls participants in the study before they develop the disease or outcome in question. Involves a group of people who already have Involves a group of people who do not have

D514 processes that pertain to large numbers of individuals. ● Histogram: A histogram is a tool to plot the frequency and number of cases and is helpful to identify areas of high variance. ● Distribution and deviations: Normal distribution visually reflects a well-formed bell curve. Variation may appear skewed outside the normal distribution and require additional testing for statistical significance. ● Standard error: The standard error reflects how close a sample mean is to the population mean. This calculation is essential for identifying variation that is a result of chance. ● Z -score: The z -score measures the distance along the x -axis of a normal distribution measured in standard deviation units. The z -score is a unit that can simplify the measurement of variation. Topic 4 Video Notes: (Predictive& Statistical Modeling) 14% of the exam ● Statistical Modeling: Examines current data to look for results, make observations on that data, examine the “now,” and collect facts. ● Predictive Modeling: Identify patterns to forecast an outcome based on current data. Identifies risks and opportunities for the future. Helps to anticipate the level of care needed. COHORT NOTES: Exam Prep session ● Type of analysis question: Time Series analysis? ● Study contents: Research design, study population, data collection, study validity? ● Type of data: Ordinal, Categorical, ratio, interval data? ● Database question: A relational database, trauma registry, object-oriented database, data warehouse ● Type of Study: Prevalence study, time-series, regression, randomized control study, ● Prospective, retrospective ● Confounding variable

D514 ● Statistical model, predictive modeling ● Nchs, cms, nih, DHHS, meps ● Predictive modeling, ANOVA, cross-sectional research. ● Indirect observation ● T-test, ANOVA, chi-square, cohort study ● CLABSI data tracked ● Chart review, literature review, indirect observation, focus group ● Population study, sample study ● factor , time series, time sensitivity(made up), dimensional analysis ● Range, ratio, correlation, inference ● Standard deviation ● ROI: RETURN ON INVESTMENT ● Prevalence rate? ● Pearson’s correlation coefficient ● Fidelity, beneficence, plagiarism, justice ● Roll up, de-identified, research ident, multidimensional ●