QI, Patient Safety, and Statistics

Question 1

What are the three types of measures in a QI study

Answer 1

1. Outcome – what we are trying to change
2. Process – actions needs to achieve change
3. Balancing – potential undesired side effects of change

Question 2

What are 5 key analysis tools of QI

Answer 2

1. Histogram – detect categories having the highest frequency of event or other attribute of interest
2. Scatter diagram – detect if there is association between two variables.
3. Cause and effect diagram (fish bone) – Identify causes of outcome of interest in a range of domains: measurements, materials, personnel, environment, methods, and machines
4. Pareto chart – combination of histogram and cumulative totals to identify the subset of factors that account for 80% of the outcome.
5. Run Chart or Statistical process control chart – identify if average, upper, and lower bounds for an outcome to detect shift over time

Question 3

You are designing a QI project. You want to understand what are the most reasons providers are not ordering a HFNC holiday on their patients. You survey providers to ascertain the extent to which 5 different factors impact their failure to order a HFNC holiday. What type of QI tool will help you visualize the most common factors identified by the survey?

Answer 3

A histogram.

Question 4

Your team wants to understand all the factors related to whether or not providers order a HFNC holiday on their patients. What QI tool will help you identify these factors?

Answer 4

A cause and effect diagram.

Question 5

You want to identify the smallest set of factors related to ordering a HFNC holiday that accounts for 80% of the failures to order a HFNC holiday. What diagram will help you identify these factors?

Answer 5

A Pareto chart.

Question 6

After completing your intervention, you want to know if there has been a change in HFNC ordering. What QI tool will help you determine if your project has succeeded in changing behavior with respect to ordering HFNC holidays?

Answer 6

A process control chart or run chart.

Question 7

Describe Lewin’s change theory

Answer 7

Unfreezing by increasing driving force or decreasing restraining forces, change, refreeze.

Question 8

What are the 5 principles of Lean methodology

Answer 8

Precisely specify value by specific product
Identify the value stream for each product (value stream mapping)
Make value flow without interruptions
Let customer pull value from the producer
Pursue perfection

The key to the Lean methodology is the concept of WASTE. Identify and eliminate waste by understanding what exactly the customer wants, and then efficiently provide that value without interruption or wasted movement or effort.

Question 9

What is the difference between the Six Sigma and Lean methodologies of process improvement?

Answer 9

Lean is focused on eliminating waste (by understanding the customer’s wants and optimizing the flow of materials and effort). Six Sigma is focused on reducing variation (through statistical analysis and controls).

Question 10

What are the three ethical principles that are the foundation for the Belmont Report’s guidance on human subjects research?

Answer 10

1. Respect for persons
2. Beneficence (which has been expanded to add nonmaleficence)
3. Justice

Question 11

What are 4 possible frameworks for distributing scarce resources described by Emanuel?

Answer 11

1. Maximize benefit
2. Treat people equally.
3. Promote instrumental value
4. Sickest first / Youngest first

Question 12

What are the 4 principles for equitable allocation of healthcare resources?

Answer 12

1. Focus on improving health
2. Patients should be informed
3. Patients should give consent
4. Conflict of interest should be minimized.

Question 13

What is considered the best type of study in terms of validity and generalizability of evidence?

Answer 13

Meta-analysis (and/or systematic reviews)

Question 14

What is a p-value?

Answer 14

The probability that the observed effect is due to random sampling error or “chance”.

Question 15

What is Receiver Operator Characteristic (ROC) curve used for?

Answer 15

To quantify the power of a screening or diagnostic test by comparing its sensitivity and specificity at different cut off values. The greater the “area under the curve” the more effective the test is because high sensitivity can be achieved without sacrificing too much sensitivity.

(In the attached figure, “True Positive Rate” is equivalent to sensitivity, and “False Positive Rate” is equivalent to 1 - specificity.)

Question 16

In research, what is a Type I error and what is a Type II error?

Answer 16

A Type I error is saying there is a difference or effect where there isn’t one. The p-value is the probability that a Type I error has occurred.

A Type II error is saying there is not difference or effect when in fact there is one. A study’s power, or 1 - beta, is the probability that a Type II error has occurred.

For example, you conduct a study to see if HFNC holidays decrease LOS. You calculate a sample size based on an alpha of 0.05 and a beta of 0.8. You find that those patients having a HFNC holiday had an LOS or 3.2 days, and those without a holiday had an LOW of 3.3 days, and the p-value was 0.15. You conclude that HFNC holidays do not signficantly decrease LOS. Based on how the study was powered, there is a 20% (1 - beta) chance HFNC actually does decrease length of stay but you failed to detect it. That is, there is 20% chance of Type II error in this study. In this study, you cannot have committed a Type I error because you did not say there was an effect of holidays on LOS (you concluded there was no effect, or at least that there was not a statistically significant effect).

Question 17

What is a continuous variable, and what is a categorical variable?

Answer 17

A continuous variable can take on essentially any value in a given range. An example would be Systolic Blood Pressure.

A categorical variable can only take on a fixed number of values (usually a small number of possible values). Examples include “History of Premature Birth (yes/no)”, or likert scales.

Question 18

What test is used to determine if two groups differ with respect to a dichotomous variable. For example, what test would you use to determine if lung cancer (yes / no) was related to ever smoking (yes / no)?

Answer 18

Chi-square test.

Question 19

What test is used to determine if a continuous variable is different between two groups. For example, what test would you use to determine if Length of Stay was different for those patients who had a HFNC holiday vs. those who did not?

Answer 19

A t-test.

Question 20

What test is used to determine how a continuous variable relates to more than one other variables. For example, what test would you use to determine how length of stay relates to age, birth weight, white blood count, HFNC holiday status, and premature birth (yes/no)?

Answer 20

Multivariate linear regression, because the dependent variable (length of stay) is continuous. The independent variables (age, birth weight, etc.) are a combination of continuous and categorical variables, but this is no problem for multivariate linear regression.

Question 21

What test is used to determine how a dichotomous variable relates to more than one other variables. For example, what test would you use to determine if readmission to the hospital (yes / no) is predicted by age, birth weight, white blood count, HFNC holiday status, and premature birth (yes/no)?

Answer 21

Multivariate logistic regression, because the dependent variable (readmission) is dichotomous. The independent variables (age, birth weight, etc.) are a combination of continuous and categorical variables, but this is no problem for multivariate linear regression.

Question 22

What is sensitivity with respect to a clinical test?

Answer 22

True negative rate, or 1 - false negative rate. In other words, a highly sensitive test is very reliable when it is negative. If a COVID test has high sensitivity, then if it is negative you can be assured you really don’t have COVID.

Question 23

What is specificity with respect to a clinical test?

Answer 23

True positive rate, or 1 - false positive rate. In other words, a highly specific test is very reliable when it is positive. If a COVID test has high specificity, then if it is positive you can be assured you really do have COVID.

Question 24

What test is used to determine if a continuous variable is different between more than two groups. For example, what test would you use to determine if Length of Stay was different for bronchiolitis patients who received (1) albuterol alone, (2) hypertonic saline alone, (3) albuterol and hypertonic saline, or (4) neither albuterol nor hypertonic saline?

Answer 24

Analysis of variance (ANOVA). ANOVA will tell you if there is a difference between 1 or more pairs of groups, but will not tell you which groups those are. If the p-value of your ANOVA test was significant (typically <0.05), you would then perform pair-wise t-tests between all the possible pairs of groups to determine which pairs of groups were different from each other.

Question 25

What is meant by the internal validity of a study and what threatens internal validity?

Answer 25

Internal validity is the extent to which a study accurately describes the population sample it studied. Threats to internal validity include bias, technical or measurement error, confounding, and random error. Internal validity has nothing to do with whether the results can be generalized to other contexts or populations, but a study cannot be generalized at all unless it is first internally valid.