Unit III ArtRev CommHea

HCA 3306, Community Health 1

Course Learning Outcomes for Unit III Upon completion of this unit, students should be able to:

2. Recognize effective principles of health programming for community health on a global scale. 2.1 Identify sources of data in community and public health biostatistics. 2.2 Identify the types of statistical analysis utilized in epidemiological research and community and

public health studies. Required Unit Resources Chapter 4: Descriptive Biostatistics in Community and Public Health Chapter 5: Inferential Biostatistics in Community and Public Health Unit Lesson

Biostatistics Statistics—the very word often brings about fear and anxiety. That feeling applies to many students and to many health care professionals alike, but it does not have to be that way. In this lesson, we will explain what you really need to know about descriptive and inferential statistics in order to be effective as a health care leader. Remember, there are experts known as professional statisticians who are available to help us when we get stuck. This lecturer/author has requested that kind of help often over the course of a career. Our professional statisticians are ready and, in fact, anxious to help. A statistic is a value that describes a sample taken from a particular population. That could be a population of just about anything from spotted owls to sea turtles to redwood trees. But in our case, it is usually a human population such as the entire population of a nation, the entire population of a state, or the entire population within a certain age group or a certain ethnic group. The reason that we study a sample from a population to develop a statistic is that typically the entire population is just too large to work with. For example, as of 2020, the population of the United States is over 329 million people (The United States Census Bureau, n.d.). It would be awesome to survey, test, or screen every U.S. citizen in our research, but the time and cost involved with doing so would simply be prohibitive. So we work with samples most of the time in community health. In the unusual case that we do have a measure for the entire population of anything studied, we call that statistic a parameter.

Descriptive Statistics Let us consider the two primary types of biostatistics: descriptive and inferential. Descriptive statistics are all about describing data. We can organize data, summarize it for users, simplify it, and then present it in meaningful ways. Descriptive statistics are generally categorized into three primary types: frequency distributions, graphical representations, and summary statistics. Frequency distributions tell us the number of people who fall into a particular category. Let’s take a simple example of a frequency distribution. Perhaps, we want to know what percentage of patients who use our facility are Medicare beneficiaries. We take the number of Medicare patients, divide by the total number of patients, and multiply by 100. Just that simply, we have the frequency distribution for Medicare patients who utilize our facility. For many U.S. hospitals today, the Medicare percentage is in the range of 40–50%, making hospitals very dependent on the Medicare program for survival!

UNIT III STUDY GUIDE Biostatistics in Community Health

HCA 3306, Community Health 2

UNIT x STUDY GUIDE Title

With frequency distributions, sometimes, we want to categorize on the basis of more than one variable simultaneously. For example, we might want to look at Medicare patients but also look at patients with a particular diagnosis. Chronic obstructive pulmonary disease (COPD) is a hot topic for hospitals today because of the 30-day readmission penalty, which has been instituted for patients with that diagnosis. So a reasonable thing to cross-tabulate today would be the percentage of patients who are on Medicare and also have a COPD diagnosis. View Image 1 below to see a frequency distribution table demonstrating data on blood pressure.

Graphical representations of data are so helpful, and they have really benefitted us in community health. Important factors and trends can sometimes seem buried in long columns and rows of numbers. But once we display the same data graphically, important patterns become obvious. The most common graphical representations in community health are the bar graph, histogram, polygon line graph, and frequency distribution. The bar graph in Image 2 indicates the relationship between HCV infection and type 2 diabetes.

Image 1: Frequency Distribution Table: Blood Pressure Data

HCA 3306, Community Health 3

UNIT x STUDY GUIDE Title

Summary statistics describe data in just two numbers. The measure of central tendency, such as the typical average score for a measure, and a measure of variability, such as the typical average variation for a measure. A practical example includes how the typical average adult fasting blood sugar might be 100 mg/dl, with typical average variation of 20 mg/dl above and below. Many of our normal ranges in health care are created on this basis. Among measures of central tendency, for quantitative data we can consider these items:

• mean, the arithmetic average of the data; • mode, the most frequently occurring observation in a set of data; and • median, the middle value in the data.

A key concept of community health is that the mean for any data set is greatly affected by just a few outliers, meaning values that are greatly different from most of the data presented. Back to our blood sugar example, a few patients with blood sugars of 600 mg/dl would shift the mean significantly, but the median would not be impacted. The middle value is, by definition, more stable than the mean, and for that reason, it is often chosen for reporting central tendency in health care. For qualitative data, the mode is always our choice, and the mean should not be utilized. Considering measures of variability, we can utilize these:

• variance, the average distance that each score is away from the mean (variance is noted as s2); • standard deviation, the square root of s2 ; and • standard error of the mean, SD/square root of the number of measurements.

You will see the standard deviation utilized to describe data variability in most community health research and applications. It gives us the best understanding of how wide the differences are among clinical data points.

Image 2: Prevalence of Type 2 Diabetes in HCV Infected and Healthy Subjects

HCA 3306, Community Health 4

UNIT x STUDY GUIDE Title

Inferential Statistics Next, we consider inferential statistics, which focus on generalizing from samples to populations. Keys here are hypothesis testing and understanding the relationships among variables. Through inferential statistics, we set about making predictions of clinical importance to our communities. We take out the crystal ball here, predicting the impact of specific factors and proposed changes on health outcomes. Health care policy is often based upon these predictions, making inferential statistics very important today! For example, we would love to study the impact on lung cancer rates by studying every client who stops smoking, following them for 10 years or 20 years downstream to track lung cancer diagnosis rates, mortality rates, and other considerations. The reality is that we cannot track them all, but with interoperable electronic health records (EHR) now becoming reality across America, we can track a lot of them! The larger our sample, the better our data, and the more meaningful our results. Inferential analysis lets us apply our sample’s reduction of lung cancer to the larger population of all smokers who quit. As an example, the NHLBI ARDS Network web page, About the NHLBI ARDS Network, is dedicated to sharing results from sample studies of acute respiratory distress syndrome (ARDS) patients at teaching hospitals across the world. Researchers all share their data and findings on this website, and the process leads to better protocols for treating ARDS patients (NHLBI ARDS Network, n.d.). This is an exciting time for researchers who employ inferential statistics. We are finally achieving truly interoperable medical records in America, and that means we will be able to share data much more easily and to mine data from around the nation, and even around the world, for the betterment of our clients.

Sampling Error Whenever we utilize inferential statistics, we need to consider sampling error. This refers to variability among populations, which occurs due to random chance rather than a true difference in the populations. The absolute best way to reduce the impact of sampling error is to conduct multiple studies from multiple samples and to make sure that each sample is large enough to minimize the impact of chance variability. That is where probability comes in. Statistical probability refers to the odds that what we observed in our sample did not occur because of random sampling error. It asks, what is the probability that my results are not just due to chance. That is where the concept of level of confidence comes in. We ask ourselves a couple of questions. What is the level of confidence that we have that our sample accurately reflects the total population we serve? Are the inferences that we are making valid? We can never know with 100% confidence, but for many health care purposes, we insist that we know with 95% confidence before we consider our results to be valid.

Hypothesis Testing In inferential statistics, we are generally interested in testing a hypothesis. The null hypothesis is that the two groups studied will not differ. The alternative hypothesis is that the two groups studied will not perform the same. There will be a difference (Sharma & Branscum, 2020). Back to our smoking cessation example–the null hypothesis might be that smoking cessation causes no difference in lung cancer rates after 10 years of cessation. The alternative hypothesis is that smoking cessation does reduce the incidence of lung cancer after 10 years of cessation. Based upon our data, we will either find that the null hypothesis is true, meaning that we do not reject the null hypothesis, or we will find that the null hypothesis is false, and we will reject it. Very commonly in health care, we tolerate a 5% probability that our decision about accepting or rejecting the null hypothesis is wrong, which is another way of saying that we have 95% confidence that our decision is correct.

Conclusion You will learn much about descriptive and inferential statistics in this unit. Whether you actually conduct research yourself or not, your understanding will make you a better health care leader because of your ability to interpret and use research findings in your work.

HCA 3306, Community Health 5

UNIT x STUDY GUIDE Title

References NHLBI ARDS Network. (n.d.). About the NHLBI ARDS network. http://www.ardsnet.org/ Sharma, M., & Branscum, P. W. (2020). Introduction to community and public health (2nd ed.). Jossey-Bass.

https://online.vitalsource.com/#/books/9781119633716 United States Census Bureau. (n.d.). U.S. and world population clock. U.S. Department of Commerce.

https://www.census.gov/popclock/