Research and Quality Improvement Flashcards
To reinforce core concepts in research and quality improvement suitable for individuals studying for the PHM boards.
Which of the following is NOT one of the three principles of the Belmont Report:
A. Respect for Persons
B. Beneficence
C. Non-maleficence
D. Justice
C. Non-maleficence.
Non maleficence IS, however, one of the four principles of ethical conduct of research outlined by Tom Beauchamp & Childress in Principles of Biomedical Ethics (along with Autonomy, Beneficence, and Justice).
What are Ezekiel Emmanuel’s 4 principles for the equitable allocation of health care resources?
- Focus in improving health
- Patients should be informed
- Patients should have the opportunity to consent
- Conflicts of interest should be minimized
Which one of the following types of bias threatens external validity but not internal validity?
A. Recall bias
B. Selection bias
C. Confounding
D. Confirmation bias
B. Selection bias.
A study can have internal validity even in the face of selection bias, but such a study would have poor generalizability due to the biased selection of participants.
For a complete catalog of types of bias see: https://catalogofbias.org/biases
You are conducting a study that examines whether or not alcohol use is associated with lung cancer risk. Unfortunately, you forgot to ask about smoking history. What type of threat to internal validity is your study at risk for?
Confounding.
A confounder is an unmeasured variable that is causally related to both your exposure of interest and your outcome of interest, producing an apparent causal relationship between your exposure and outcome that does not actually exist.
How do you calculate the odds ratio from the following 2 x 2 table (see handout for details).
[ a ] [ b ]
[ c ] [ d ]
ad / bc
(or (a/b) / (c/d) )
True or false: you can calculate an odds ratio from a case control study.
True.
But you can’t calculate a relative risk from a case control study. Relative risk relates to the incidence of your outcome. In a case control study you are manipulating the apparent incidence because you are predetermining the number of cases and the number of controls. In a cohort study, you preserve the incidence because you simply select a population to study without predetermining how many “cases” it will contain.
You are abstracting the D-dimer lab values from patients in your study. What type of variable is D-dimer?
Continuous.
You have collected CRP values for 20 patients with Kawasaki Disease, 20 patients with macrophage activtation syndrome, and 20 patients with Stevens Johnson syndrom. You want to know whether CRP differs between these three groups. What statistical test would you use?
ANOVA. If the ANOVA is statistically significant, you can proceed with pairwise t-tests between each pair of conditions.
You are interested if a history of RSV infection is associated with asthma. You enroll 100 asthmatic patients and 100 healthy controls and inquire about whether they were ever diagnosed with RSV. What statistical test is appropriate to analyze this data?
Chi-square.
Bonus Round: since the number of event (HFNC exposure) may be quite small, a Fisher’s Exact Test may be required. This test is indicated if any value in your 2 x 2 table is less than about 5.
You want to know if albuterol use among patients admitted with bronchiolitis is associated with a lower risk of needing high flow nasal cannula support. You are aware that there are many potential confounders. In addition to dose and frequency of albuterol administration, you also collect data about age, adjusted gestational age, virus detected, and history of ectopy. What type of multivariate statistical test would be appropriate to analyze this data?
A. Linear regression
B. Cox proportional hazard regression
C. Friedrich Thomlinson Reverse Double Salchow test
D. Logistic regression
D. Logistic regression.
Your dependent variable (or outcome variable) is dichotomous (categorical): PICU transfer: Yes/No. Since you cannot use linear regression in this scenario, logistic regression is indicated.
You are conducting a quality improvement project to improve screening for vaccination needs prior to discharge. You have interviewed 20 nurses who failed to screen for vaccination needs at discharge. Based on their responses, you want to identify the most important factors to address in your QI project. What type of QI chart will help you address this question?
A. Run chart
B. Cause and effect diagram
C. James Moses terminal flow assessment
D. Paretto chart
D. Paretto chart
A run chart is used to analyze events over time to see if there was a shift or trend resulting from your intervention. A cause and effect diagram promotes assessment of all the factors that influence an outcome, but does not prioritize them. A James Moses terminal flow assessment is a fictional entity.
You have noticed that problem lists are not being updated at patient discharge. You convene a group to identify the reasons problem lists are not updated. What kind of QI activity is this?
A. Failure mode analysis
B. Root cause analysis.
C. Balancing measure assessment
D. Value stream mapping
A. Failure mode analysis.
A root cause analysis is an assessment of a specific event where a significant failure occurred. Balancing measures are part of the data collected during the “DO” phase of the PDSA cycle. Value stream mapping is part of the LEAN methodology and is focused on customer expectations and waste in production.
As part of a case control study, you collect data on whether a history of preterm birth (yes / no) is associated with increased risk of needing high flow nasal cannula during admission for bronchiolitis. What type of statistical test would be used for this data?
A. T-test
B. Chi-square test
C. Linear regression
D. Furtherwold’s inverse Q test.
B. Chi-square test.
You are comparing two categorical variables. T-tests are for one dichotomous (categorical) variable and one continuous variable. Linear regression is for a continuous dependent (outcome) variable, and any number of independent variables. Furtherwold’s inverse Q test does not exist.
