Prognosis

Question 1

Prognosis

Answer 1

Prognosis speaks to outcomes in patients
with disease

Question 2

Prognostic factors versus risk factors

Answer 2

• Prognosis speaks to outcomes in patients
with disease • Risk factors predict disease development

• Some factors are both risk and prognostic
factors

ex// • Hypertension in acute coronary
syndromes

Question 3

aspects of prognosis in patient care

Answer 3

1. telling families and patients what’s in store
2. identifying the high-risk/poor prognosis patients
3. matching resources to need
4. implicit prognostcation is constant and ongoing– must be evidenced based

Question 4

three layers of prognosis (LVM)

Answer 4

1. prognostic significance of specific variables
   - history, physical exam, labs
2. likelihood of a given outcome in a given disease
3. multi-variable models to predict outcome.

Question 6

what studies inform prognosis

Answer 6

cohort studies

RCTs

administrative data

systematic reviews

Question 7

factors that determine result validity

Answer 7

1. representative sample
2. patients should be homogenous with respect to pronostic risk
3. was follow up complete?
4. were objective and unbiased outcome criteria used?

Question 8

factors to think about when asking the question “what are the results”?

Answer 8

1. how likely are the outcomes over time?
2. how precise are the estimates of likelihood?

Question 9

Simple Odds Ratios represent the results of
a __ analysis.

Multivariate analysis yields adjusted ___
Ratios and ___ predictors

Answer 9

Simple Odds Ratios represent the results of
a univariate analysis.

Multivariate analysis yields adjusted Odds
Ratios and independent predictors

Question 10

questions to ask yourself when considering “How can I apply the results to patient care? “

Answer 10

1. Were the study patients and their management similar to my own?
2. Was the follow-up sufficiently long?
3. Can I use the results in managing patients in my practice?

• Simple Odds Ratios represent the results of a univariate analysis • Multivariate analysis yields adjusted Odds Ratios and independent predictors

Question 11

Odds Ratio

Answer 11

measure of association between an exposure and an outcome. OR represents odds that an outcome will occur given a particular exposure compared to the odds of the outcome occurring in the absence of that exposure

Question 12

The exposure-odds ratio for case control data is

Answer 12

s the ratio of the odds in favor of exposure among cases to the odds in favor of exposure among noncases

Question 13

the disease-odds ratio for a cohort or cross section is:

Answer 13

is the ratio of the odds in favor of disease among the exposed to the odds in favor of disease among the unexposed

Question 14

prevalence-odds ratio

Answer 14

refers to an odds ratio derived cross-sectionally from studies of prevalent cases.

Question 15

formula for odds ratio

Answer 15

Question 16

when does an odds ratio kind of estimate relative risk

Answer 16

when the disease is infrequent. odds ratio DOES NOT equal relaitve risk when the disease is more common.

Question 17

Prognostic models use multiple prognostic factors in combination to predict the risk of future clinical outcomes in individual patients. A useful model provides accurate predictions that inform patients and their care givers, supports clinical research, and allows for informed decisions to improve patient outcomes.

Prognostic model research has three main phases:

Answer 17

model development (including internal validation), external validation, and investigations of impact in clinical practice. Although many prognostic models are proposed, relatively few are currently used in clinical practice.

Question 18

Development of a prognostic model needs to consider various steps, such as the __ and coding of predictors for the model, and how to estimate the model ___.

__ modeling is the most common approach, while ___ learning techniques are gaining interest.

It is important to evaluate the ___ of the predictions for the derivation cohort (___ validation) as well as for new settings that may differ from the derivation cohort (____ validation)

Answer 18

Development of a prognostic model needs to consider various steps, such as the specification and coding of predictors for the model, and how to estimate the model parameters.

Regression modeling is the most common approach, while machine learning techniques are gaining interest.

It is important to evaluate the quality of the predictions for the derivation cohort (internal validation) as well as for new settings that may differ from the derivation cohort (external validation)

Question 19

Apparent aka ___ validation implies assessment of model performance directly in the derivation cohort. This approach yields an optimistic estimate of model performance, because the regression coefficients are optimized for the derivation cohort

Answer 19

Internal Validation

Question 20

____ validation relates to the generalizability and transportability of the prognostic model to another population

Answer 20

external

Question 21

The classic measures to express model performance are __ and __.

___ refers to the ability of the prognostic model to distinguish between high and low risk patients, and is commonly quantified with the ___ statistic

. ___ indicates the agreement between observed outcomes and predicted probabilities.

Answer 21

The classic measures to express model performance are discrimination and calibration. Discrimination refers to the ability of the prognostic model to distinguish between high and low risk patients, and is commonly quantified with the concordance statistic.

Calibration indicates the agreement between observed outcomes and predicted probabilities.

Question 22

Decision Support

Some prognostic models explicitly aim to support clinical decision making. For these models, an additional __-___ evaluation is required, beyond the normal discrimination and calibration for other prognostic models.

A __-__ measure that can be used to express this balance is __ __ (NB).

NB is calculated as a weighted sum of true and false positive classification: (true positives – weight × false positives) / total number of patients

Answer 22

Decision Support

Some prognostic models explicitly aim to support clinical decision making. For these models, an additional decision-analytic evaluation is required, beyond discrimination and calibration.

A decision-analytic measure that can be used to express this balance is net benefit (NB).

NB is calculated as a weighted sum of true and false positive classification: (true positives – weight × false positives) / total number of patients

Question 23

how is net benefit calculated?

Answer 23

NB is calculated as a weighted sum of true and false positive classification: (true positives – weight × false positives) / total number of patients

Question 24

Q1. Research that addresses a prognosis question can inform which of the following scenarios?

a. The likelihood that a former smoker will develop lung cancer
b. The clinical findings, in a comatose survivor of cardiac arrest which can help physicians and loved ones decide when to withdraw life support.
c. Which elements of the history and physical exam can be combined as a rule to eliminate the need for radiography in patients with ankle sprains.
d. Which genetic variants of aspirin metabolism determine response to treatment in acute myocardial infarction.

Answer 24

The answeris B

a. The likelihood that a former smoker will develop lung cancer. (no this is a study about harm, where smoking is a risk factor)
b. The clinical findings, in a comatose survivor of cardiac arrest which can help physicians and loved ones decide when to withdraw life support. (yes, the clinical findings may be used to predict outcome, which can help guide decisions about care)

c. Which elements of the history and physical exam can be combined as a rule to eliminate the need for radiography in patients with ankle sprains. (This is a clinical decision rule designed for diagnostic purposes)
d. Which genetic variants of aspirin metabolism determine response to treatment in acute myocardial infarction. (This is about guiding choice of therapy)

Question 25

Q2. A 2006 study used retrospective chart data to determine the 30 and 90-day risk of stroke in patients of a US Health Maintenance Organization presenting to the ED with a TIA. Which factor is most likely to lead to an over-estimation of stroke risk for a general population of patients with TIA?

a. ED referral patterns in Health Maintenance Organizations (HMOs).
b. Variable rates of disagreement for primary outcome assessment of stroke among the neurologist reviewers
c. Retrospective data collection.
d. Significant loss to follow-up of study participant

Answer 25

A

a. ED referral patterns in Health Maintenance Organizations (HMOs). (yes may have selected for more severe symptom profiles)
b. Variable rates of disagreement for primary outcome assessment of stroke among the neurologist reviewers (possible in theory but could have biased the other way too)
c. Retrospective data collection. (a risk of bias from a data integrity perspective but not necessarily associated with making outcomes look worse)

d. Significant loss to follow-up of study participants (depending on whether those lost to follow- up were more or less healthy than those not lost, this could bias the result either way)

Question 26

Q3. Which of the following is the BEST example of an independent prognostic factor?

a. A clinical or laboratory variable that demonstrates a highly significant statistical association with the outcome of interest
b. A clinical characteristic with a statistically significant association with the outcome after multivariable logistic regression analyses
c. A clinical variable that is predictive of the outcome of interest after consideration of similarly predictive but less discriminating covariates
d. A clinical or laboratory variable that is statistically distinct from covariates

Answer 26

C IS THE CORRECT ANSWER

a. A clinical or laboratory variable that demonstrates a highly significant statistical association with the outcome of interest (no it’s not just about statistical significance, and it may be confounded)
b. A clinical characteristic with a statistically significant association with the outcome after multivariable logistic regression analyses (helps remove issue of confounding but still not the best answer)
c. A clinical variable that is predictive of the outcome of interest after consideration of similarly predictive but less discriminating covariates (comparison with other possible predictors is key)
d. A clinical or laboratory variable that is statistically distinct from covariates (distracting choice)

Question 27

Q5. Which of the following MOST influences the application of a prognosis study in practice?

a. Clinical variable information was derived from administrative data
b. Clinical outcome data was derived from clinical records
c. Recruiting of study subjects from a tertiary care hospital
d. Lack of blinding for outcome adjudication

Answer 27

C IS THE CORRECT ASNWER

a. Clinical variable information was derived from administrative data (not necessarily a weakness)
b. Clinical outcome data was derived from clinical records (same)
c. Recruiting of study subjects from a tertiary care hospital (yes, patients may represent a sicker spectrum)
d. Lack of blinding for outcome adjudication (this threatens internal validity – i.e., are the results true but is not specific to the application of the study results in practice)

Question 28

T/F: A study with a larger cohort or longer duration of follow-up (and higher numbers of events)
would have narrower CIs.

Answer 28

true. would be more representative of the population, allowing you to have a more precise answer.