Complementary and alternative medicine use and absenteeism among individuals with chronic disease

The U.S. Center for Disease Control’s National Center for Health Statistics used a continuous sampling and interviewing design to obtain information on basic demographics, general health, and current health topics through the cross-sectional U.S. National Health Interview Survey (NHIS) [17]. Data were collected from the civilian non-institutionalized United States population, oversampling African American and Hispanic populations. Individuals in the armed forces and those living in institutional group quarters, long-term care institutions, correctional facilities, and countries outside the U.S. were not included. From each household sampled, one adult and one child (if there were any children) were interviewed and provided information on the rest of the household. Beginning in 2002, the NHIS included questions on 27 complementary health approaches every 5 years [18]. The conditional sample adult response rate for the 2012 NHIS was 79.7 % [19].

These analyses made use of the 2012 NHIS. NHIS data are available publically and permission was not required for access. All individuals gave verbal consent prior to participating in this study as indicated in the Field Representative Manual [17]. As this study made use of a de-identified publically available dataset, the study was exempt from IRB oversight by the Columbia University Medical Center IRB.

A total of 108,131 individuals participated in the 2012 NHIS and 34,525 participants provided data on CAM use. Participants under age 18 and/or were eligible to receive full retirement benefits in 2012 (age 66 and older) were excluded from analysis (n = 40,050) [20]. Of the 20 conditions defined as chronic disease by the Multiple Chronic Conditions working group within the Health and Human Services Office of the Assistant Secretary of Health, the 2012 NHIS collected responses on 13 of these conditions [21]. Individuals were included if they answered yes to any of the following: ever having either arthritis, cancer, chronic obstructive pulmonary disease, non-gestational diabetes, hepatitis, high cholesterol, or being informed in the past twelve months of having hypertension, hepatitis, coronary heart disease, asthma, chronic kidney disease, depression, or substance abuse. Participants who did not have one or more chronic disease at the time of the interview (n = 43,448), had not worked for pay in the past week (n = 10,500), or had reported missing 364 or more days from work in the past twelve months (n = 64) were excluded [22]. Working for pay in the past week was the best indicator of employment in the past year opposed to ever having a job. Those who did not or refused to answer any of the survey questions regarding CAM use were excluded from analysis (n = 3,873). After removing observations with missing values for the number days missed from job or business in the past twelve months and CAM use the final population size was 10,196.

CAM practices considered for inclusion were determined by the National Center for Complementary and Integrative Health’s (NCCIH) definition of “complementary health approaches” for mind and body practices and natural products. Natural products were included in our definition of CAM as these therapies are the most commonly used [9]. CAM practices included acupuncture, massage, meditation, movement therapies (does not include general exercise), relaxation techniques, mind-body practices, vitamins (excluding multi-vitamins), minerals, and herbs [13]. Participants reported not using any of the NCCIH defined practices were considered non-users. CAM types of interest were categorized as either dietary supplements or mind-body practices. Mind-body practices included biofeedback, mantra meditation/mindfulness meditation/spiritual meditation/guided imagery/progressive relaxation, and yoga/tai-chi/qi-gong. Dietary supplements included use of non-multivitamins, non-vitamin supplements, minerals, and herbs.

The primary outcome was the self-reported number of days missed from job or business in the past twelve months due to illness or injury (0–364 days), excluding maternity leave. Participants who had self-reported working in the past week were asked during the past twelve months, “about how many days were missed from a job or business due to injury or illness,” with a possible range of 0–366 days.

Demographic variables included race, age, gender, household income, and highest level of education achieved. A priori hypothesized confounders included BMI (underweight: <18.5, normal: 18.5-24.9, overweight: 25.0-29.9, obese: 30.0+ kg/m²), marital status, smoking status, level of alcohol consumption, number of chronic conditions, surgery in the past twelve months, self-perceived general health, health insurance status, number of employees at job or business, class of worker, and reasons for using the self-reported top three CAM therapies. Type of employment was categorized as working for a privately owned company, any type of government (local, state, and federal government), or for one’s self. Possible reasons for CAM use included for general wellness or general disease prevention, to improve energy, to improve immune function, and to improve memory or concentration.

Univariable analysis was performed to determine the raw frequency, weighted percentage, and significance of the association between each variable and either days missed from job/business or CAM use. Using the chi-square test, variables associated with both the main exposure and outcome at α = 0.10 were treated as potential confounders. This α level allowed more variables to be considered for modeling since the effects of these variables on days missed from job or business are not well known. Multivariable Poisson regression was used to evaluate the association between CAM use and days missed from job/business due to illness or injury. The minimally adjusted model included race, age, gender, income, and education as covariates. Potential confounders identified in univariable analyses were included in the fully adjusted models if they modified the beta coefficient of any CAM use in minimally adjusted model by 10 % or greater. Age, BMI, and number of chronic conditions were kept as continuous variables for regression analysis (categorized in Tables 1, 2 and 3). Eigenvalues and variance inflation factors were examined to determine any collinear variables; none of the significant variables were collinear. To make results generalizable to the U.S. population and to adjust for clustering, stratification, and oversampling of specific population subgroups, weighted Poisson regression was performed using final person-level weight which included design, ratio, non-response, and post-stratification adjustments [23]. All analyses were performed using SAS 9.4 (Cary, NC). The SURVEYFREQ commands were used with strata, cluster, and weight to determine weighted percentages. The final model was constructed using PROC GENMOD for weighted regression using Poisson distribution.