Data from the publicly available, deidentified National Health and Nutrition Examination Survey (NHANES) dataset for the years 2005–2006, 2007–2008, 2009–2010, 2013–2014, and 2017–2018 were used for these analyses. NHANES is a cross-sectional survey designed to monitor the health and nutritional status of the non-institutionalized civilian U.S. population [6]. The NHANES sample is selected through a complex, multistage probability design. The sample design includes oversampling to obtain reliable estimates of health and nutritional estimates for various sub-populations. The survey consists of interviews conducted in participants’ homes and standardized physical examinations conducted in mobile examination centers. Femur neck bone mineral density (BMD) scores were measured using a dual energy X-ray absorptiometry (DXA) scan on Hologic Discovery Model A Densitometers [7]. The sample was limited to individuals 50 years of age and older. This study was reviewed by an institutional review board and determined to be not research and therefore exempt, effective April 8, 2024.

Variables for gender (male and female); ethnicity/race (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, Other Race — including multi-racial); age group (reclassified as 50–59 years of age; 60–69 years of age; 70–70 years of age; and 80 years of age and older); and gender and ethnicity/race (male Mexican American, male Other Hispanic, male non-Hispanic White, male non-Hispanic Black, male Other Race — including multi-racial, female Mexican American, female Other Hispanic, female non-Hispanic White, female non-Hispanic Black, and female Other Race — including multi-racial) were analyzed across NHANES years 2005–2006, 2007–2008, 2009–2010, 2013–2014, and 2017–2018 to determine if there was a statistically significant change across the time intervals.

Prevalence was defined as the number of people in a given survey cycle with the presence of osteoporosis divided by the total number of eligible people (age 50 years old and above) in our sample. NHANES sampling weights were used to generate nationally representative prevalence estimates. For the purpose of this study, the World Health Organization definition of osteoporosis as having a femoral neck BMD of 2.5 standard deviations or more below the young female adult mean (using normative data from the NHANES reference database on Caucasian women aged 20–29 years) was used to classify individuals with osteoporosis [8]. The narrow neck (NN) BMD (g/cm2) variable in the NHANES III dataset is the recommended metric for analysis [9] and is constructed from the Hip Structural Analysis (HSA) Program. The average BMD value for the NN site among Caucasian women aged 20–29 years was used to set the reference value against which standard deviation values were calculated for all individuals included in this study. If an individual’s BMD value was 2.5 or more standard deviations below this reference value, that individual was classified as having osteoporosis; otherwise, they were classified as not having osteoporosis.

The presence of an osteoporosis diagnosis being made (yes; no) served as the secondary outcome variable and was assessed using descriptive statistics. This secondary outcome was used to determine and report the proportion of undiagnosed osteoporosis. In order to assess whether an individual was diagnosed with osteoporosis, self-reported NHANES questionnaire data was cross-referenced with NHANES examination data. If an individual responded to the NHANES survey question “Ever told had osteoporosis/brittle bones?” with “Yes” and DXA examination data showed the presence of osteoporosis, an individual was classified as being diagnosed with osteoporosis. Conversely, if an individual responded to that NHANES survey question “Ever told had osteoporosis/brittle bones?” with “No” and DXA examination data showed the presence of osteoporosis, an individual was classified as having undiagnosed osteoporosis.

The NHANES survey creates primary sampling units (PSUs), which are typically single counties or may be groups of contiguous counties using probability proportional to a measure of size. Stratification parameters, which are defined by geography, metropolitan statistical area status, and various other population demographics, are applied to PSU creation to improve sample precision. The corresponding NHANES survey mobile examination center (MEC) sample strata and weights were applied for all descriptive and statistical calculations to ensure representativeness and generalizability to the U.S. civilian non-institutionalized population.

In the present study, we applied sampling weights to assess the temporal changes in the osteoporosis prevalence between 2005 and 2018. Although NHANES response rates have been steadily falling since 2011, with more accelerated declines reported in recent years, efforts have been made to adjust biases and ensure comparability across survey years [9, 10]. Specifically, weights were adjusted for non-response to the in-home interview when creating the interview weights and further adjusted for non-response to the MEC exam when creating the exam weights. Also, the NHANES sample weights are post-stratified to match estimates of the U.S. civilian non-institutionalized population available from the U.S. Census Bureau [11]. Moreover, as declines in the response biased toward income and education in 2017–2018, a series of alternative weighting adjustments to reduce bias were explored. After the initial weighting process was applied, the weights were adjusted to education and then further adjusted to income [12]. The intensive adjustments suggest that bias in the outcome statistics was mostly reduced through the enhanced weighting adjustments, and estimates in various NHANES cycles are comparable.

Descriptive analyses were conducted to capture osteoporosis prevalence estimates for gender, ethnicity/race, and age group (Table 1), and gender and ethnicity/race (Table 2) across NHANES survey years 2005 to 2018. The prevalence for the sub-group analyses was defined as the number of people in a sub-group with presence of osteoporosis divided by the total number of population at risk in that specific group. Sampling weights were used to generate nationally representative estimates of the prevalence. Descriptive analyses were also conducted using the NHANES weighting to capture osteoporosis prevalence estimates for diagnosed and undiagnosed osteoporosis among gender, ethnicity/race, and age group (Table 3) for NHANES years 2017 to 2018, in addition to a descriptive assessment of the proportion of undiagnosed osteoporosis.

Temporal trend analyses were conducted utilizing linear probability models (LPMs) to assess for changes in prevalence across all included NHANES cycles. LPMs are preferred over logistic regression in such cases because of the intuitive and easily interpretable estimates [13]. Appropriate sampling weights were applied in the trend analysis. These results are listed in Tables 1 and 2, and a two-sided p-value less than 0.05 was considered statistically significant. Analyses were conducted using the STATA statistical software package [14].