biostatistics in obstetrics - slide 1
TRANSCRIPT
Biostatistics in Obstetrics
Mitra Ahmad Soltani
2008
In the Name of God
References:• Ahmad Soltani M. Regression Analysis of Labor Duration. The
Internet Journal of Gynecology and Obstetrics. Texas: Vol 5, No 2. 2006
• Clements JM. Synergy Medical Education Alliance Research Design Core Curriculum. Module2&3.2008
• Kramer M et al. Prepregnancy Weight and the Risk of Adverse Pregnancy Outcomes. New England Journal of Medicine.1998 Vol:338, N3:147-152
• Lyon D. Use of Vital Statistics in Obstetrics. emedicine. Dec 2007
• Pritchard JA, MacDonald PC, Gant NF. Obstetrics in broad perspective. In: Williams Obstetrics. 22nd ed. New York, NY: McGraw-Hill; 2005
Birth rate
number of births 1000 population
• It includes men in the population.
Fertility Rate
number of live births 1000 women aged 15-44 years
• While a woman with 2 second-trimester miscarriages would be considered fertile, her deliveries would not be included in the fertility rate.
Reproductive Mortality rate
contraceptive use plus direct maternal deaths
100000 women• This is perhaps the most sensitive measure of
a population's ability to provide safety for women.
Maternal Mortality Rate
number of direct or indirect maternal deaths100,000 live births
• A condition in which both mother and fetus are lost would both increase the numerator (maternal death) and decrease the denominator (live birth).
Infant Mortality Rate
infants who die prior to their first birthday 1000 live Births
• IMR is often one of the sentinel indicators used to evaluate a population's overall health and access to health care.
Neonatal Mortality Rate
losses between 0-28 d of life (inclusive) 1000 live births
• This rate is often divided into early (first 7 d) and late (8-28 d) rates, as etiologies within these 2 categories vary somewhat.
Fetal Death rate (stillbirth rate)
number of stillbirths 1000 infants (total Births)
• Infants means “live and still” born.
Perinatal Mortality Rate
Fetal deaths+neonatal deaths 1000 total Births
Still birth
• Delivery after 20 weeks' EGA (and more than 500 g birthweight) in which the infant displays no sign of life (gasping, muscular activity, cardiac activity) is considered a stillbirth.
Live Birth
• Delivery after 20 weeks' EGA in which any activity is noted is classified as a live birth. This is a difficult definition, as the lower limit of reasonable viability currently remains around 23 weeks' EGA. Thus, a spontaneous delivery at 21 weeks' EGA with reflex motion but no ability to survive with or without intervention would nonetheless be considered a live birth.
Abortion
• The most common definition of an abortion is any loss of a fetus that is less than 20 weeks' completed gestational age (since last menstrual period) or that weighs less than 500 grams.
Preterm Infants
• Preterm infant is another arbitrary definition because a subtle gradient of maturity exists. Most states define premature as a delivery before 37 completed weeks' gestational age, although the vast majority of babies born after 35 weeks‘ GA have uncomplicated perinatal courses.
Postterm Infants
• The generally accepted definition of a postterm pregnancy is one that progresses beyond 42 weeks' completed gestational age based on last menstrual period (LMP). In practice, many clinicians use a lower cutoff such as 41 weeks‘ GA when LMP is certain.
Testing for statistical significance of the difference for nominal data
• Small unmatched sample: Fisher’s exact test• Small matched sample: Sign test• Large unmatched sample: Chi-square, with
Yates correction• Large matched sample:McNemar’s test
Testing for statistical significance of the difference for ordinal data
• One comparison(2 groups),unmatched sample: Mann-Whitney U
• One comparison(2 groups)Matched sample:Wilcoxon matched pairs
• More than 2 groups unmatched sample: Kruskal Wallis one-way ANOVA
• More than 2 groups matched sample: Friedman 2-way ANOVA
Testing for statistical significance of the difference for continuous data
• One comparison(2 groups)unmatched sample: t-test
• One comparison(2 groups)matched sample: matched t-test
• More than two groups unmatched sample:F test for analysis of variance followed by pairwise comparisons
• More than two groups matched sample: F test for analysis with blocking or analysis of covariance
Measure the size of difference
• Nominal/ordinal data: differences in proportions or percentages in
each category• Continuous data:Differences in mean values between the
groups+ SD for each group
Tests to Determine Association Between Groups Measure the degree of Association
• Nominal data: odds Ratio/Relative Risk• Ordinal Data(nonlinear):Spearman’s
rho/Kendall’s tau• Continuous Data: Pearson’s Correlation
Coefficient ( r )
Tests to Determine Association Between Groups testing for statistical significance of association
• Nominal Data: Statistical Significance of odd’s Ratio
• Ordinal data(nonlinear): Statistical Significance of rho or tau
• Continuous Data(linear):Statistical significance of Pearson’s r
Tests to Determine Association Between Groups- Extent Association Explains Variation Between Groups
• Nominal data: Attributable Risk• Ordinal data(nonlinear):Spearman’s rho or
Kendall’s tau• Continuous data(linear):Pearson’s coefficient
of determination
For describing one group
• Mean, SD for measurement of Parametric Distributions
• Median, interquartile range for rank,score or measurement of non-parametric distributions
• Proportion for Binominal (2 possible outcomes)
Compare one group to a hypothetical value
• One sample t-test for measurement of Parametric Distributions
• Wilcoxon test for rank,score or measurement of non-parametric distributions
• Chi-square for Binominal (2 possible outcomes)
Compare two unpaired groups
• Unpaired t-test for measurement of Parametric Distributions
• Mann-Whitney test for rank, score or measurement of non-parametric distributions
• Fischer test(or chi-square for large samples) for Binominal (2 possible outcomes)
Compare two paired groups
• Paired t-test for measurement of Parametric Distributions
• Wilcoxon test for rank, score or measurement of non-parametric distributions
• McNemar’s test for Binominal (2 possible outcomes)
Compare three or more unmatched groups
• One-way ANOVA for measurement of Parametric Distributions
• Kruskal Wallis test for rank, score or measurement of non-parametric distributions
• Chi-square for Binominal (2 possible outcomes)
Compare three or more matched groups
• Repeated measures ANOVA for measurement of Parametric Distributions
• Friedman test for rank, score or measurement of non-parametric distributions
• Cochrane Q for Binominal (2 possible outcomes)
Quantify association between two variables
• Pearson Correlation for measurement of Parametric Distributions
• Spearman Correlation for rank, score or measurement of non-parametric distributions
• Contingency coefficients for Binominal (2 possible outcomes)
Predict value from another measured variable
• Simple linear regression or non-linear regression for measurement of Parametric Distributions
• Non-parametric regression for rank, score or measurement of non-parametric distributions
• Simple logistic regression for Binominal (2 possible outcomes)
Predict value from several measured or binominal variables
• Multiple linear or nonlinear regression for measurement of Parametric Distributions
• Multiple logistic regression for Binominal (2 possible outcomes)
summary
P=parametric/N=nonparametric/B=binominalM=matched/ U=unmatched/G=group/~=versus/ H=Hypothetical value
P N B
Describing Mean &SD Median&range Proportion
Compare 1G~H One sample T-test Wicoxon Chi-Square
Compare 2GU Unpaired T-test Mann-Whitney Fisher’s
Compare 2GM Paired T-test wilcoxon McNemar’s
Compare 3GU 1 way ANOVA Kruskal Wallis Chi-square
Compare 3GM Repeated ANOVA Friedman Cochrane Q
association Pearson Spearman Contingency coef.
predict Linear regression Non-para. Regr. Logistic regres.
Statistics Related to Diagnostic Tests
• Sensitivity = True Positives/(True Positives + False Negatives)
• Specificity = True Negatives/(False Positives + True Negatives)
• Positive Predictive Value = True Positive/(True Positive + False Positive)
• Negative Predictive Value = True Negative/(True Negative+False Negative)
• Likelihood Ratio = compares the likelihood of a result in a patients with the disease to the likelihood of a result in patients without disease.
• Positive LR = (a/a+c)/(b/b+d)• Negative LR = (c/a+c)/(d/b+d)
True + =(a)
False + =(b)
+ sum
False – =(C)
True – =(d)
sicks sum
How much do LRs change disease likelihood?
• – LRs>10 or <0.1 cause large changes in likelihood
• – LRs 5-10 or 0.1-0.2 cause moderate changes• – LRs 2-5 or 0.2-0.5 cause small changes• – LRs between <2 and 0.5 cause little or no
changes
True + = (a)
False + =(b)
+ sum
False – =(C)
True – =(d)
sicks sum
Statistics to Interpret Importance &Precision of Therapeutic Results
• Control Event Rate (CER) = c/(c+d)• Experimental Event Rate (EER) = a/(a+b)• Relative Risk (RR) = EER/CER =
(a/a+b)/(c/c+d)• Relative Risk Reduction (RRR) = CER-EER/CER• Absolute Risk Reduction (ARR) = CER-EER• Number Needed to Treat (NNT) = 1/ARR
True + =(a)
False + =(b)
+ sum
False – =(C)
True – =(d)
sicks sum
Relative Risk and Odds ration
There is strong association of RR or OR>1There is strong association if RR>3 or OR>4
use formulaRelative risk RCT & cohorts (a/a+b)/(c/c+d)Odds Ratio Case-control (a/c)/(b/d) or
ad/bc
Sample size
If dependent variable is nominal or ordinal:
• n= p (1-p) /d²If dependent variable is
continuous:• n=s²/d²
Regression analysis of labor duration
(A sample research)
IntroductionDetermining labor duration has been the focus of different researches . The main aim is to lower the rate of cesarean section and undue hospitalization. Friedman’s, Hendrick’s , and Philpott’s Partographs and Nesheim’s regression equation are the results of such efforts. The advantage of an equation over a partograph is its predictive value in determining obstructed labor in advance and on an individualized basis.
WHO partograph
Linear Regression Estimates the coefficients of the linear equation, involving one or more independent variables, that best predict the value of the dependent variable.
The two-variable modelY = A + B X
Materials and methods
230 Laboring women were interviewed and examined according to a checklist from April – August 2004 .
The inclusion criteria were:
1-Singleton pregnancy 2- Vertex presentation 3- Gestational age 36-42 weeks 4- no medical or obstetric disease 5- Bishop score of 10-12 6- normal FHR 7- spontaneous initiation of labor 8- non elective cesarean section 9- no diagnosis of CPD
1-mother’s height 2- maternal age 3- prepregnancy weight 4- maternal BMI 5- drugs used except oxytocin 6- 9interventions ( amniotomy – c/s/vacuum/ enema or any other way of bowels preparation) 10- 12- Duration , intensity and frequency of labor pain (in the initial stages before oxcytocin administration)
The independent variables were:
13-abnormal events like cord prolapse or fetal heart abnormality or occiput posterior delivery 14- occupation 15-31- lifestyle in terms of alcohol consumption, smoking, exercise, meals, Consumption of grain , vegetables, fruit, dairy and type of dairy, meat ad meat products, fat and dressings , water, snacks , and score as the sum of items
The independent variables were:
Framingham Health Assessment Questionnaire is presented (though in small font size) as a reference for a print. It can help provide an account of life style risk for health. Four items were changed based on Iranian pregnant women characteristics:Alcohol consumptionsmokingExercise and type of dairy products
13.0 Consumption of alcoholHow often do you consume alcohol?_____1) Never drink_____2) 2 days or less per week_____3) 3 days per week_____4) 4 or more days per week
14.0 Number of alcoholic beveragesOn the days you drink, on the average how many drinks do you have?
_____1) Never drink_____2) 1 to 2 drinks_____3) 3 to 4 drinks_____4) 5 or more drinks
15.0 CaffeineHow often do you consume caffeine in your diet including coffee, tea, cola or chocolate?
_____1) Never_____2) Occasionally but not every day_____3) 1 to 3 servings daily_____4) 3 to 5 servings daily_____5) More than 5 servings daily
16.0 Smoking statusIndicate which of the following best represents your current statusNOTE: Check all that apply.
_____1) Have never smoked_____2) Quit smoking less than 5 years ago_____3) Quit smoking more than 5 years ago_____4) Smoke pipe or cigar_____5) Smoke less than 1 pack of cigarettes per day_____6) Smoke more than 1 pack of cigarettes per day
LIFESTYLE ITEMS
Exercise Program
18.0 Exercise FrequencyOn the average, how many days per week do you exercise?
_____1) 3 or more days per week_____2) Less than 3 days per week_____3) No regular exercise program
19.0 Proper stretching Do you perform stretching prior to exercise?
_____1) Always_____2) Sometimes_____3) Never_____4) Currently not exercising
20.0 Warm-up and cool downDo you warm-up and cool-down after exercising?
_____1) Always_____2) Sometimes_____3) Never
_____4) Currently not exercising
Section E Nutrition Habits
21.0 Daily MealsOn the average how many meals do you consume per day?_____1) 3 meals with "healthy" snacks_____2) 3 meals_____3) 2 meals or less_____4) No regular eating pattern
22.0 Consumption of grain/bread productsOn the average, indicate the type and amount of grain products you normally consume per day.NOTE: A serving is 1 sl. bread, 1/3 cup beans / peas, 1/3 cup oatmeal, rice or other grain products.
_____1) Whole grains at least 6 to 11 servings per day_____2) Whole grains 6 servings or fewer servings per day
_____3) Refined grains such as white bread/rolls/processed flour at least 6 to 11 servings per day
_____4) Refined grains such as white bread/rolls/processed flour 6 or less servings per day _____5) Rarely consume grain products
23.0 Consumption of vegetablesOn the average, how many servings of vegetables do you consume per day? Note: A serving is approximately 1 cup of raw or 1/2 cup of cooked.
_____1) At least 3 to 5 servings per day_____2) Less than 3 servings per day_____3) Rarely consume vegetables
24.0 Consumption of fruitsOn the average, how many servings of fruit do you consume per day? Note: A serving is approximately 1 piece of fruit.
_____1) At least 2 to 4 servings per day_____2) Less than 2 servings_____3) Hardly ever consume fruit
25.0 Daily consumption of dairy productsOn the average, how many servings of dairy products do you consume per day? Note: A serving is approximately 1 cup of milk or 1 oz. of cheese.
_____1) At least 2 servings per day_____2) Less than 2 servings_____3) Hardly ever consume dairy products
26.0 Type of Dairy products
Indicate the type of dairy products you consume._____1) Nonfat selections only_____2) Both low fat and nonfat about the same_____3) Low fat only_____4) Usually high fat selections_____5) Do not consume dairy products
27.0 Daily consumption of meats and meat products
Indicate the type of meat you normally consume.
_____1) Do not consume meat or meat products_____2) Consume less than 6 oz. of poultry or fish per day_____3) Consume more than 6 oz. of poultry or fish per day_____4) Consume less than 6 oz. of red meat per day_____5) Consume more than 6 oz. of red meat per day
28.0 Consumption of fats, dressings and spreadsIndicate the type and number of servings of fat, dressings and spreads you consume each day.
High fat examples: Butter, lard, and margarineLow fat examples: Non-fat or Low-fat salad dressing-mayonnaise-cheese
_____1) Use low fat selections sparingly (less than 3 per day)_____2) Use low fat selections frequently (3 or more per day) _____3) Use both low fat and high fat about the same sparingly (3 or less)_____4) Use high fat selections sparingly (less than 3 per day)_____5) Use high fat selections (more than 3 per day)
On the average, how many glasses of water do you consume per day? Note: A serving is one 8-oz. glass of water only; do not include coffee, soda or other beverages.
_____1) At least 8 glasses per day_____2) About 4 to 8 glasses per day_____3) Less than 4 glasses per day_____4) Seldom consume water
30.0 Convenience and snack food consumptionOn the average how many times per day do you eat convenience foods or forms of fast food?
_____1) Never_____2) Less than 1 time per day
_____3) More than 1 time per day
The independent variables were:
32- Gravida **33- Para** 34-Last delivery**35-education,36- residency district37- maternal BG and Rh
38- newborn weight ( not known before delivery)39- newborn sex ( not known before delivery) 40- time of delivery:( 8 am – 8 pm is considered as day time) ( not known before delivery)41- parity( based on crosstabs testing)42-last delivery ( based on crosstabs testing)43- gravida ( based on crosstabs testing)
The confounding variables were:
The dependent variables were:44 - labor duration
active phase
45 - labor duration
46 -
rate
4 cm- delivery
Rate
• Rate of cervical dilation means cm dilation per hour. Compared to other dependent variable , RATE was more reliable for analysis.
Assumptions1- First do no harm is the basic assumption of any medical intervention!2- Obstructed labor should be defined by individual characteristics of women.
Codes for nominal variables They are arranged according to less high risk to more high risk states in terms of labor duration (based on review on related literature )
• Disease:• no=1/yes=2• Hospital district :• Affluent districts=1• Non affluent districts=2• Time of delivery:• night=1/day=2• Sex of the newborn:• girl=1/boy=2
Codes for nominal variables They are arranged according to less high risk to more high risk states in terms of labor duration (based on review on related literature )
• Interventions:• done=1/not done=2• Pain intensity:• good=1/not good=2• Blood group:• A=1/Non A=2• RH:• pos=1/neg=2
Codes for nominal variables They are arranged according to low risk to high risk states in terms of labor duration (based on review on related literature )
• Occupation:• Non sedentary=1/Sedentary=2• Education:Education:• educated=1/ illiterate=2
Results
Gravida and rateSymmetric Measures
.138 .081 1.744 .083c
.183 .081 2.328 .021c
159
Pearson's RInterval by Interval
Spearman CorrelationOrdinal by Ordinal
N of Valid Cases
ValueAsymp.
Std. Errora Approx. Tb Approx. Sig.
Not assuming the null hypothesis.a.
Using the asymptotic standard error assuming the null hypothesis.b.
Based on normal approximation.c.
Young Child and rateSymmetric Measures
.278 .155 2.241 .029c
.163 .127 1.281 .205c
62
Pearson's RInterval by Interval
Spearman CorrelationOrdinal by Ordinal
N of Valid Cases
ValueAsymp.
Std. Errora Approx. Tb Approx. Sig.
Not assuming the null hypothesis.a.
Using the asymptotic standard error assuming the null hypothesis.b.
Based on normal approximation.c.
Para and rateSymmetric Measures
.129 .080 1.629 .105c
.196 .082 2.495 .014c
158
Pearson's RInterval by Interval
Spearman CorrelationOrdinal by Ordinal
N of Valid Cases
ValueAsymp.
Std. Errora Approx. Tb Approx. Sig.
Not assuming the null hypothesis.a.
Using the asymptotic standard error assuming the null hypothesis.b.
Based on normal approximation.c.
• To reduce the effect of Confounding Variables(number of pregnancies (G), number of previous
deliveries (P), and the years passed since last delivery( YC)), the stepwise regression computation was done for all independent variables and dependent variable (rate) only for those cases with previous deliveries not equal to zero (p#0).
correlation coefficients of rate and rate predictors
Independent : BMIDependent : rateMTH:LINRsquare =043Df=146F=6.60Sig f =0.011B0 =.5744B1 = .0965
Which means:
Rate=0.57 +0.09 BMIor
Conclusion• In this study, women of lower BMI had a longer labor
course. • According to Kramer, in a developed country lean
women are likely to have adequate nutritional stores to meet the basic requirements of pregnancy. So a low BMI is accompanied by a lower pregnancy risk than a higher BMI. This must not be generalized to developing countries, particularly those in which maternal undernutrition is highly prevalent.
THE END