Exam 3

What is big data?

analyzing large sets of data computational to reveal, patterns, trends, and associations relating to human behavior and interactions

what are the established quality domains of big data?

-safe
-timely
-effective
-efficient
-equitable
-patient centered

reasons for the chasm of big data?

-growing complexity of science and technology
-increase in chronic conditions
-poorly organized delivery system
-constraints on exploiting the revolution in information technology

how does big data have increased focus on quality across healthcare system?

-patients
-providers
-payers
-regulators
-researchers
-external rating agencies and reporting

what are two publicly available resources for research?

Health Data and NCBI

What data tools are available for use?

-Tableau
-SlicerDicer
-alteryx
-python
-SQL
-IBM SPSS
-MATLAB
-RStudio

Systematic Review Purpose

draw a conclusion from the cumulative weight of evidence related to a specific question/topic

Components of SR research design

-subjects = individual studies
- rigorous methodology guiding the search for studies to minimize bias
-majority of intervention focused SRs are of RCTs
-Interpretation of findings accounts for quality of evidence

Narrative Review

-general or focused topic
-summary of findings from individual studies
-may/may not have established search and selection methods
-limited to no critical analysis of individual studies included in review
-authors interpretation

systematic review

-focused question
-summary of findings from selected individual studies
-rigorous search and selection methods established ahead of time
-extensive critical analysis of individual studies included in review

what is the gold standard of systematic reviews

Cochrane reviews

Steps in systematic reviews

1. phrase the research question
2. search for all evidence
3. select relevant studies
4. extract characteristics
5. assess study quality
6. data/meta analysis

methodological design of an SR

-should have robust methodological design
-can range from poor to high quality
-critical appraisal of SR important for EBP

1. Specify the clinical/ research question

-clear clinical / research question (PICO)
-patient, intervention, comparison, outcome

2. Specific Methodological Steps in an SR

-clear, explicit, and reproducible search strategy
-data range clear
-addresses potential for study selection bias
-languages other than English
-unpublished works/grey literature
-explicit inclusion/exclusion criteria
-minimum of 3 databases searched (PubMed, CINAHL, Cochrane)

3. Assessing the quality / risk of bias of qualified studies

-clear description of processes and tools used to asses the quality / risk of bias of included studies
-masked, independent review of the full text article of each study
-minimum of 3 reviewers (2 independently score the quality and 1 to tie break)

Threats to validity of RCT that impact quality

-selection bias
-assignment / allocation bias
-lack of blinding
-attrition
-maturation
-statistical regression to the mean
-history
-instrumentation
-testing bias

4. Summarize the evidence

evidence tables

5. interpret the findings

interpret the results in context of evidential quality ratings / risk of bias for each study

meta-analysis

-extension of SR: mathematical pooling of data across multiple studies
-statistical analyses of data pooled from individual studies to form an aggregate statistic
-sample size (variability) of each study drives the analysis (size of square)
-not all SR's allow for meta-analysis

potential benefits of meta-analyses

-increase statistical power to find effects (1 study versus > 1)
-improve estimates of effect size, risk or odds
-reduce uncertainty in results across studies
-improve external validity

Not all SRs lead to a

meta-analysis

commonly reported statistics in meta-analysis

-effect size (ES)
-standardized effect size (SES)
-likelihood ratio (LR)
-odds ratio (OR)
-relative risk (RR)

for a meta- analysis to be performed, there must be

homogeneity of included studies

homogeneity

-degree to which the studies are similar
-determined by visual examination or statistical analysis: chi-square statistic or Cochrane chi-square
-there should not be significant difference between articles (p> 0.5)

key issues interpreting Forest plots

-number of contributing studies
-size of each square is proportional to the weight of each study contributing to overall result
-diamond at bottom represents the pooled data for that estimate
-width of the diamond represents the confidence interval (CI) of pooled result
-solid vertical midline = no effect
-if CI crosses solid vertical line, no clear effect since it includes the no effect line

exploratory research

-find relationships
-systematic investigation of relationships among 2 or more variables

examples of exploratory relationships

-cohort studies
-case control studies
-correlational / predictive
-methodological
-historical research

exploratory relationships use

-describe relationships
-predict effect of 1 variable on another
-test relationships supported by clinical theory

correlation

-measure of the degree of association font variables
-foundation of exploratory research
-correlation coefficient (r) = -1 to +1

strength of correlation

-little to no = 0 to 0.25
-fair = 0.25 to 0.50
-moderate to good = 0.5 to 0.75
-good to excellent = > 0.75

regression

-when researcher wants to predict value of outcome variable
-linear regression: know 1 variable and predict the other

correlational research

-describe the nature of existing relationships among variables
-rationale for clinical decisions
-search for new hypotheses
-determine which variables are related

predictive research studies

predict a behavior or response based on the observed relationships between that behavior and other variables

uses of predictive research studies

-develop methods used for decision making
-validation of measurement tool

theory testing

correlational data: used to test theories

advantages of exploratory research

-important role in clinical research
-how variables are related
-how they occur in nature

disadvantages of exploratory research

-no cause and effect determined
-risk of bias in data interpretation
-multiple variable interplay

cohort studies: prospective

follow group of participants over time
-calculate rate at which disease develops
-identify risk factors

largest/longest prospective cohort study

-framingham heart study
-1948
-5200 participants every 2 years
-risk factors for cardiovascular disease

value of prospective cohort studies

demonstrates association/correlation between risk factors and disease development

prospective research data

-variables measured by direct recording
-data collected in present
-follow progress through treatment or evaluation
-control quality and quantity of data

advantages of prospective cohort studies

-greater control of data and collection methods
-determine incidence of condition
-establish temporal relationship: risk factor; outcome
-effective for studying multiple disorders

disadvantages prospective cohort studies

-not useful for uncommon disorders
-time
-expense

retrospective research

collected from studies done in the past

advantages retrospective research

-important source of information
-databases
-medical records
-medical surveys

disadvantages retrospective research

-no control: operational definitions; reliability of data
-incomplete/missing data
-accuracy and credibility: need consideration

value of retrospective cohort study

-determines association between identified risk factors and disease
-less expensive/time consuming than prospective study

components of retrospective cohort study

-observational human study
-starts in present with exposure of interest
-using medical records and interviews, investigators look for presence of specific in patients' histories

retrospective vs. prospective

-prospective: exposed but before disease
-retrospective: exposed and developed disease

selection of subjects

-general population
-geographical area
-employees of specific companies
-unique exposure histories: army

one group cohort studies

-prospective: follow over time
-retrospective: look backwards

case control studies

-case: with disorder
-control: without disorder

methodology

-look backwards:
-direct interviews, questionnaires, chart review
-examine differences in exposure histories:
-differences suggest outcome
-identify potential exposures that may have resulted in disease

case-control study

-limitations:
-bias in subject selection
-controls may have other health conditions

selection of cases for case control studies

-general population
-benefit: good generalizability
-limitation: expensive

hospital based study for case control studies

-most common approach
-easy to obtain participants

selection of controls of case control studies

-same criteria as those chosen for case
-just no disease present
-matched: age, race, gender, occupations
-random digit dialing, medicaid enrollees

hospital recruits case control studies

-advantages: convenience
-limitations: may be sick, more likely to smoke, use other medications

analysis issues of case control studies

-bias: selection, observation, interview, recall bias
-misclassification: non-differential, differential misclassification
-assessment of exposure status: confounding effects of extraneous factors

what is epidemiology?

study of health and disease

how did epidemiology begin?

-study of epidemics
-morbidity and mortality

what does epidemiology include today

-study of chronic disease
-disabilities
-health status
-AIDS, cardiovascular disease, arthritis, cancer, back pain, traumatic injuries, birth defects, covid

epidemiology includes

-determining frequency of diseases, health states, and their trends
-determining factors that effect the development of particular disease or health state and why
-predicting the occurrence and distribution of various diseases and health states
-determining factors that prevent disease, prolong life with disease, or improve health status

classification of epidemiological studies

-observational
-quasi experimental
-experimental

observational epidemiological studies

no artificial manipulation of any of the study factors

what are the two observational epidemiological studies

-descriptive
-analytical

when to use descriptive epidemiological studies

-when little is known about the occurrence or determinants of health conditions and disease states
-set priorities for health care planning
-generate hypotheses

purpose of descriptive epidemiological studies

-who: experiences the disorder?
-where: is frequency of disease highest and lowest?
-when: does occur most? least?

classifications of descriptive epidemiological studies

-case reports/ case series
-correlational studies
-cross sectional studies

case reports/ case series

description of unique occurrence or medical condtion

purpose of case reports/ case series

-complete description: characteristics/exposures
-hypothesis about causal factors

limitations of case reports/case series

not enough control (generalizability/causality)

correlational studies

-relationship: disease/disorder and specific exposure
-analyze patterns in population

correlational studies use

evidence to formulates hypotheses for testing

advantages for correlational studies

use information from large databases
-centers for Disease control

limitations of correlational studies

no cause and effect relationship

cross sectional studies

snapshot of population at one point in time

purpose of cross sectional studies

-describe frequency of a disorder
-within defined group or population
-efficient method

limitations of cross sectional studies

-cannot detainee causality (exposure and disease)
-insensitive to duration of disorder (long or short disease)
-under representative of disease frequency due to deaths

how measures of disease frequency are measured

-population size
-period in time
-whole numbers

measures of disease frequency

-reflects risk for disease in a given group
-35 cases/ 3200 people/ 1 year = 10.94/1000
-report as 1094 cases / 100,000 in 1 year

most common measures for disease frequency

-prevalence
-incidence

prevalence

estimates probability: individual will have a particular disease in a particular period of time
-impact of disease on population
-planning health services

what is prevalence

-existing number of cases/total population at specific time
-point prevalence: single point in time
-period prevalence: over a range of time

limitations of prevalence

-does not explain cause
-long disease process = large prevalence
-short disease process = low prevalence, even if number of people with disease was high

incidence

-new cases in a specific time
-estimates risk of developing disease during that time
-discounts effects of duration of illness

two expressions of incidence

-cumulative incidence
-incidence rate

cumulative incidence (CI)

number of new cases (give time period/ total population at risk)

concerns of cumulative incidence

-specify time period of observation
-number of people at risk may vary over time
-competing risk factors may contribute to disease
-assume all subjects available for entire study period

limitations of cumulative incidence

follow up period not uniform for all participants

incidence rate (IR)

number of cases/total person-time
-total person time: sum of time periods of observation
-can account for death/attrition

usefulness of incidence rate

more efficient than CI

birth rate

number of live births (year) / total population at mid year

mortality rate

number of deaths (year) / population at mid year

cause-specific mortality rate

number of death from specific disease (year)/ average mid year population

case fatality rate

number of death from disease/ number of individuals with disease
-within specific time frame

age specific rate

mortality rate for specific age group

age adjusted mortality rate

mortality rate weighted for differences in population distribution

use of observational analytic designs

when enough is known about a condition to allow for testing of hypotheses about specific risk factors and disease