Impact of Physical Activity on Disease and Mortality: Evidence from National Health and Nutrition Examination Survey

2.1. Data

The National Health and Nutrition Examination Survey (NHANES) is the primary data used in this study [12-14]. The NHANES is specifically designed to measure the health and nutritional status of adults and children in the United States. The survey is comprehensive in terms of methodology as it combines physical examinations and interviews for the same subject (i.e., individual participants). Due to the scientific importance and confidentiality clause of the data, all the survey process and protocol, from the design of the questionnaires to data collection, is approved by CDC as well as the National Center for Health Statistics (NCHS) Ethics Review Board (ERB). Furthermore, the informed consent for the survey is publicly available on the NHANES website.

More specifically, it is a national-level data set that was started in 1999, with over 10,000 individuals randomly selected in each two-year assessment cycle. The random section is conducted through a statistical process using the U.S. Census information. Briefly describing the sample selection procedure, all the counties in the U.S. are divided into 15 groups based on their characteristics. From each of the 15 selected groups, between 20 and 24 neighborhoods are also selected based on the neighborhood characteristics. Then, a sample of about 30 households is chosen within each neighborhood. Finally, the randomly selected individuals are invited to an interview. After the interview, individuals go through a health examination. As specified in the Helsinki Declaration, confidentiality of personal information is strictly protected, and therefore personal information, such as names and addresses, are not disclosed in the data. In 2005-2006, the NHANES also released individual-level accelerometer data in addition to their health status information. This study particularly uses the information collected until December 31, 2015, for an average follow-up time of 10 years.

2.2. Physical Activity

NHANES collects objective physical activity data using hip-worn accelerometers, and it is publicly available from the CDC’s webpage. The self-reported data covers two time periods, between 2003-2004 and 2005-2006, and is organized in the R-package “rnhanesdata” [15]. Following household interviews, participants are invited to have a health examination in an NHANES mobile examination center. The participants wore the ActiGraph AM-7164 device for seven consecutive days from the day of the NHANES examination and removed it during sleep and water-related activities. The NHANES data reports both the wear and non-wear time. Specifically, the non-wear time was identified as an interval of at least 60 consecutive minutes of 0 activity counts. The physical activity was summarized as an activity count at the minute level weekly for each participant. Therefore, individuals with high PA time are considered to move actively throughout the day.

2.3. Analysis

The NHANES 2003-2006 survey participants were followed up for mortality status, and their mortality status was available from NHANES via a link to the National Mortality Index [16]. This study specifically uses the data collected until December 31, 2015, for an average follow-up time of a 10-year period.

To distinguish individuals’ physical activity levels, the total per week physical activity amount for each individual was calculated, and then indicator variables were created for the analysis, based on the 25th, 50th, or 75th percentile of total physical activity to identify low, medium, and high PA individuals to examine the PA levels. After deleting the missing information, the sample size became 11,170 participants with accelerometry data between the 2003-2004 and 2005-2006 cycles. This study first analyzes the dependent variables of high and low physical activity levels with different diseases and covariates such as age, gender, and education level to study the active and less active individuals’ characteristics. Then, by running nine logit models, the following section of the study analyzes the impact of different physical activity levels on mortality.

3.1. Descriptive Statistics

Table 1 summarizes all variables in the final dataset, separating the data between alive (i.e., assigned as 0) and dead (deceased, assigned as 1). The first column is for the entire sample. Participants with a critical underlying disease are associated with a high 10-year mortality rate in the sample. Those underlying diseases include diabetes, heart disease, and cancer. For example, 410 individuals out of 1057 with diabetes, 252 out of 464 with coronary heart disease (CHD), 250 out of 377 with congestive heart failure (CHF), 246 out of 404 with stroke, and 398 out of 992 with cancer died within ten years. Moreover, smokers are low-mobility individuals who are also at high risk. The data confirms that 731 individuals out of 2620 former smokers and 1036 out of 2184 with mobility difficulty also did not survive within ten years.

High Body Mass Index (BMI) is another potential cause of mortality. Among the deceased group within the BMI group in Table 1, the obese and overweight subjects have a high mortality rate. Another observation from Table 1 is that 22% of the deceased sub- jects have diabetes, 14% have coronary heart disease (CHD), 14% have congestive heart failure (CHF), 22% have cancer, 36% are overweight, 56% have mobility problems, 39% are moderate drinkers, 40% are former smokers, and 19% are current smokers. Amongst all the deceased subjects, college graduates have the lowest mortality rate compared to other education categories. Although the data used for this study does not include information on nutrition and eating habits based on education level, the analysis will later proxy the intensity of physical activity for college graduates to see the impact on the mortality rate. The mortality data merged with CDS’s data also have the cause of death. However, the cause of death could be due to complications of other preexistent diseases. Therefore, this paper focuses on physical activity and its association with disease and mortality.

The basic descriptive statistics show that the individuals in the high-PA group are associated with a higher percentage of mortality compared to the low-PA group, even though each group has an almost identical number of observations. This implies that other variables should be factored in to examine the relationship between PA and mortality status.

3.2. Physical Activity Intensity

In Table 2, a series of logistic regression models were analyzed to understand the survey data beyond descriptive statistics and study the association between objectively measured physical activity intensity and several prevalent diseases. In particular, the models reported in this table identify who is more/less engaged in physical activity. Specifically, three models (cancer, CFH, and CHD) are examined for high physical activity. Then similar models are discussed for low physical activity while keeping the diabetes dummy variable in all models. The results confirm that physically active individuals significantly decrease diabetes, chronic heart failure, and obesity. Individuals currently smoking are found to be associated with fewer PA times.

Regarding educational achievement, the highly educated group showed a significantly high level of physical activity. The odds of physical activity being high with a college education were 1.87 (i.e., exp(0.63)=1.87) times higher than with someone without.

3.3. Mortality

This study next identifies the association between objectively measured physical activity intensity and several prevalent diseases and mortality separately. As shown in Table 3, a set of logistic regressions are estimated for the 10-year mortality rate. The specific research question that is addressed in this section is, “who is more likely to be deceased after ten years, based on the physical activity level?” To answer this question, nine different models are identified to individually isolate cancer, chronic heart failure, and chronic heart disease, controlling the PA levels separately in each cohort. Although cancer, CHF, and CHD are all significantly and positively associated with mortality, the mortality rate decreases with high PA levels.

The results document that participants who reported a high level of physical activity are significantly less likely to die within the ten-year follow-up period for all types of models controlling cancer and heart diseases (both for CHD and CHF). For example, after controlling the cancer patients in models 1 through 3, the low physically active group has a 55% greater relative mortality risk within ten years (i.e., the odds ratio is exp (0.44) =1.55. In comparison, the mortality risk of high physical active individuals is decreased by almost 20% (i.e., the odds ratio is exp (−0.23) = 0.7945). Additionally, similar magnitudes for heart disease are discovered.

Comparing samples who smoke currently and formerly, the models show a higher and more significant mortality rate for currently smoking individuals. In model 1, the odds value of mortality for current smokers is 2.07 (i.e., exp (0.73) times higher than nonsmokers, while the odds value of mortality for the former smoker is 1.22 (i.e., exp (0.20) times higher than nonsmokers. Although it is good to quit smoking, it is better not to smoke at all to reduce smoke-related mortality. Although the difference is insignificant for cancer and CHF groups, for the CHF group, the result showed that those categorized under the obese groups are more likely to be alive during the follow-up period (10 years) if they have high PA. Obesity and the risk of heart failure are documented in the literature [17]. Therefore, it is unsurprising that obese individuals with high physical activity levels reduce their risk of congestive heart failure (i.e., CHF).

The results also indicate the association between gender and mortality. Although it is documented that the mortality risk of males is higher than women in each disease category, males with diabetes and cancer with a high level of physical activity are around 3.7% less likely to die compared to women (i.e., exp (0.62) −exp (0.6)=0.0369) with similar health problems [18], confirming the benefit of the high physical activities.