chapter 44: statistics and patient safety Flashcards
rejects null hypothesis incorrectly -> falsely assumed there was a difference when no difference exists
type 1 error
rejects null hypothesis incorrectly -> falsely assumed there was a difference when no difference exists
type 1 error
accepts null hypothesis incorrectly because of small sample size -> the treatments are interpreted as equal when there is actually a difference
type 2 error
hypothesis that no difference exists between groups
null hypothesis
p value that rejects the null hypothesis
p
p value: > 95% likelihood that the difference between the populations is true
p
likelihood that the difference is not true and occurred by chance alone with p
spread of data around a mean
variance
population
parameter
most frequently occurring value
mode
average
mean
middle value of a set of data
median
prospective study with random assignment to treatment and non treatment groups
randomized controlled trial (avoids treatment biases)
prospective study in which patient and doctor are blind to the treatment
double-blind controlled trial
- avoids observational bias
prospective study -> compares disease rate between exposed and unexposed groups (nonrandom assignment)
cohort study
retrospective study in which those who have the disease are compared with a similar population who do not have the disease; the frequency of the suspected risk factor is then compared between the 2 groups
case-control study
combining data from different studies
meta-analysis
2 independent groups and variable is quantitative -> compares means (mean weight between 2 groups)
student’s t test
variable is quantitative; before and after studies (e.g. weight before and after, drug versus placebo)
paired t tests
compares quantitative variables (means) for more than 2 groups
ANOVA
compare categorical (qualitative) variables (race, sex, medical problems and diseases, medications)
nonparametric statistics
compares 2 groups with categorical (qualitative) variables (number of obese patients with and without diabetes versus number of non obese patients with and without diabetes)
chi-squared test
small groups -> estimates survival
Kaplan-Meyer
incidence in exposed / incidence in unexposed
relative risk
probability of making the correct conclusion = 1 - probability of type 2 error
- likelihood that the conclusion of the test is true
- larger sample size increases power of a test
power of test
number of people with disease in a population (Eg number of patents in US with colon CA)
- long-standing disease increases prevalence
prevalence
number of new cases diagnosed over a certain time frame in a population (e.g. number of patients in the US newly diagnosed with colon CA in 2003)
incidence
ability to detect disease = true-positives/(true-positives+false-negatives)
- indicates the number of people who have the disease who test positive
sensitivity
with high sensitivity, a negative test result means patient is very unlikely to have disease
ability to state no disease is present = true-negatives/(true-negatives + false-positives)
- indicates the number of people who do not have the disease who test negative
specificity
with high specificity, a positive test result means patient is very likely to have disease
true-positives / (true-positive + false-positive)
- likelihood that with a positive result, the patient actually has the disease
positive predictive value
true-negatives / (true-negatives + false-negatives)
- likelihood that with a negative result, the patient does not have the disease
negative predictive value
depends on disease prevalence
predictive value
depends on disease prevalence
predictive value
accepts null hypothesis incorrectly because of small sample size -> the treatments are interpreted as equal when there is actually a difference
type 2 error
hypothesis that no difference exists between groups
null hypothesis
p value that rejects the null hypothesis
p
p value: > 95% likelihood that the difference between the populations is true
p
likelihood that the difference is not true and occurred by chance alone with p
less than 5%
spread of data around a mean
variance
population
parameter
most frequently occurring value
mode
average
mean
middle value of a set of data
median
prospective study with random assignment to treatment and non treatment groups
randomized controlled trial (avoids treatment biases)
prospective study in which patient and doctor are blind to the treatment
double-blind controlled trial
- avoids observational bias
prospective study -> compares disease rate between exposed and unexposed groups (nonrandom assignment)
cohort study
retrospective study in which those who have the disease are compared with a similar population who do not have the disease; the frequency of the suspected risk factor is then compared between the 2 groups
case-control study
combining data from different studies
meta-analysis
2 independent groups and variable is quantitative -> compares means (mean weight between 2 groups)
student’s t test
variable is quantitative; before and after studies (e.g. weight before and after, drug versus placebo)
paired t tests
compares quantitative variables (means) for more than 2 groups
ANOVA
compare categorical (qualitative) variables (race, sex, medical problems and diseases, medications)
nonparametric statistics
compares 2 groups with categorical (qualitative) variables (number of obese patients with and without diabetes versus number of non obese patients with and without diabetes)
chi-squared test
small groups -> estimates survival
Kaplan-Meyer
incidence in exposed / incidence in unexposed
relative risk
probability of making the correct conclusion = 1 - probability of type 2 error
- likelihood that the conclusion of the test is true
- larger sample size increases power of a test
power of test
number of people with disease in a population (Eg number of patents in US with colon CA)
- long-standing disease increases prevalence
prevalence
number of new cases diagnosed over a certain time frame in a population (e.g. number of patients in the US newly diagnosed with colon CA in 2003)
incidence
ability to detect disease = true-positives/(true-positives+false-negatives)
- indicates the number of people who have the disease who test positive
sensitivity
with high sensitivity, a negative test result means patient is very unlikely to have disease
ability to state no disease is present = true-negatives/(true-negatives + false-positives)
- indicates the number of people who do not have the disease who test negative
specificity
with high specificity, a positive test result means patient is very likely to have disease
true-positives / (true-positive + false-positive)
- likelihood that with a positive result, the patient actually has the disease
positive predictive value
true-negatives / (true-negatives + false-negatives)
- likelihood that with a negative result, the patient does not have the disease
negative predictive value
true-positives + true-negatives / true-positives+true-neg+false-pos+false-neg
accuracy
depends on disease prevalence
predictive value
are independent of prevalence
sensitivity and specificity
seeks to collect outcome date to measure and improve surgical quality in the US. outcomes are reported as observed vs expected ratios
National Surgery Quality Improvement Program (NQSIP)
JCAHO prevention of wrong site/procedure/patient protocol
- preop verification of patient and procedure
- operative site and side
- time out before incision is made (verifying patient, procedure, position site + side, and availability of implants or special requirements)
promoting culture of safety
- confidential system of reporting errors
- emphasis on learning over accountability
- flexibility in adapting to new situations or problems
risk factors for retained object after surgery (MC sponge)
emergency procedure, unplanned change in procedure, obesity, towel used for closure
unexpected occurrence involving death or serious injury, or the risk thereof; hospital undergoes root cause analysis to prevent and minimize future occurrences (Eg wrong site surgery)
sentinel event (JCAHO)
GAP protection technique
gaps in care (shift change, etc) can lead to loss of information and error. prevention - structured handoffs and checklists (face to face if possible; standardizing orders; reading back orders if verbal
