Quantitative Measures in Health Care

Abstract

For students to contin ually gain new knowledge in medicine and to maintain life-long learning beyond medical school and residency, students must understand basic statistics and research methodology (e.g., in the evaluation of scientific presentations and publications). This chapter provides a concise summary of basic statistical and methodological concepts, and applies such knowledge within the problem-based learning approach.

Appendices

Appendix A: Tables with Possible Answers to the Vignettes

Case Vignette 15.1

What are the facts?	What are your hypotheses?	What do you want to know next?	What specific information would you like to learn about?
There appears to be a link between long-term use of antipsychotic drugs and movement disorders commonly found in Parkinsonism	There is a relationship between long-term use of this particular atypical antipsychotic drug and movement disorders commonly found in Parkinsonism	Have observational or experimental studies been conducted on this topic?	How to select and conduct a study?
			Study designs
			Scales of measurement for data collection
			Types of sampling techniques
	A significant portion of the variation in occurrence of this movement disorder can be explained by the long-term use of this drug		Sensitivity and specificity tests
			How to analyze data (e.g., basic statistical tests used to analyze data)?

Case Vignette 15.2

What are the facts?	What are your hypotheses?	What do you want to know next?	What specific information would you like to learn about?
12-year-old girl	Possible anxiety disorder	Physical appearance (e.g., any bruises and residual marks)	What are the normal developmental milestones at this age?
			Other ways to approach and communicate with this girl
Mother worried about mental development	Normal separation anxiety of temperamental variation		Available diagnostic tools/instruments
	Depression	Cognitive awareness of environment
	Child abuse	Home environment
	Post-traumatic stress disorder	School performance
		Parents’ disposition (temperament)
		Possible substance use	Reliability
	Nutritional deficiency
	Possible endocrine disorder (hypothyroidism)
Difficulty interacting with peers	Mother is overly concerned	Possible parental substance use	Validity
Shy
Refuses to separate from mother

Case Vignette 15.3

What are the facts?	What are your hypotheses?	What do you want to know next?	What specific information would you like to learn about?
There may be a relationship between different types of anxiety disorders and impairment	Different anxiety disorders will have different associations with impairment	What does the scientific literature reveal?	Relationship between anxiety disorders and impairment
More advanced research designs and statistical methods may be used	Structural equation modeling might be involved	What else needs to be known before making a conclusion?	Structural equation modeling

Case Vignette 15.4

What are the facts?	What are your hypotheses?	What do you want to know next?	What specific information would you like to learn about?
The journal article purports that through the use of meta-analysis, Drug X is associated with a significant increase in the risk of a cardiovascular event	The study has many limitations and should be treated with cautious reception	Of the 64 relevant studies not selected, how many claimed no significant relationship between Drug X and cardiac events?	Meta-analysis
			Relative risk
			Odds ratio

Appendix B: Answers to Review Questions

Answers

• 1. Study designs:

• Answer: (b).

• 2. Sensitivity, specificity, positive predicative value, negative predictive value:

• Answer: (b).

Disease
	Present	Absent
Test Positive	TP	FP
Test Negative	FN	TN

1. TP  true positive, FN  false negative, FP  false positive, TN  true negative

Sensitivity = TP/(TP + FN)	Specificity = TN/(FP + TN)
Positive Predictive Value = TP/(TP + FP)	Negative Predictive Value = TN/(FN + TN)

Explanation: The prevalence of CHD is 25 %. Assume a sample of 100 and put 25 people in the CHD-present column and 75 people in the CHD-absent column. The sensitivity of the test is 60 %; therefore, 60 % of the 25 people with CHD (or 15 people) will test positive and the other 10 will test negative. The specificity is 80 %; therefore, 80 % of the 75 CHD-absent people (or 60 people) will test negative and the other 15 will test positive. Thus, 15 + 15 = 30 people will test positive, and 15 of them will really have CHD; therefore, the positive predictive value is 15/30 or 50 %. A total of 10 + 60 = 70 people will test negative, and 60 of them really will not have CHD; therefore, the negative predictive value is 60/70 or 86 %.

• 3. Correlations:

• Answer: (c).

Although the Pearson r or correlation coefficient of  +0.85 is very high, if the sample size was very small, the correlation may not be statistically significant. Although the other correlation of −0.20 is much lower, the sample size could be much larger and the correlation could be statistically significant. Therefore, because the statistical significance is dependent on both the size of the correlation and the number of participants, we do not know which of the two correlations is statistically significant. We would need to know the N size for each to determine whether each is statistically significant or not.