Retrospective analysis of the use of osteoporosis medication at the presentation of non-vertebral fragility fractures in a predominantly Hispanic population.

Ethics statement

The study conformed to the Declaration of Helsinki and the U.S. Federal Policy for the Protection of Humans Subjects and was approved by the Institutional Review Board; protocol number 1542366-6. This retrospective study received approval on February 3, 2020, by the DHR Health Institute for Research and Development IRB under the expedited mechanism of review. A full waiver of consent to participate was submitted and approved by the IRB, specifying the retrospective nature of the study and the age of the participants to be included. All data was collected and analyzed de-identified.

Study design and setting

This is a three-year, retrospective cohort study of patients from a single hospital, Level 2 Trauma Center in South Texas. The retrospective period of chart review was set from January 1, 2017, to December 31, 2019.

Participants

Male and female participants 50 years and older presenting to any of the clinics or emergency rooms at our health system were included in the study. The electronic medical record at the hospital links with that of primary care providers and other clinics within the health system (including urgency room, orthopedic, endocrinology, and radiology depts., etc.), allowing for prescriptions of multiple providers to be captured in the patient’s medical record. The following types of fractures were included: upper end of the humerus or the shoulder griddle, lower end of radius or ulna and forearm, displaced or non-displaced femoral neck, intertrochanteric/ subtrochanteric fractures, the lower end of tibia, medial malleolus, lateral malleolus, or lower leg. Patients were excluded from the study if the onset of the injury or presentation was outside the study period, if the fracture was already treated at another institution (e.g., referred to a rehabilitation hospital), or if the fracture was nonunion. Additional exclusion criteria were based on the mechanism of injury: fractures produced as a result of projectiles, motor vehicle accidents, or falls from a height higher than one meter were excluded from the cohort. Fractures due to diagnosed/ documented cancer were excluded as well. Fragility fractures of the vertebrae (as defined by the International Classification of Disease-10 as M80.08) were initially considered in this study. However, the incidence of such fractures, as documented in the medical records, was lower than any of the other types of fragility fractures included herein. Therefore, fractures of the vertebrae were underdiagnosed, similar to what has been previously reported for this type of injuries^21,22. To reduce the introduction of information bias, these fractures were purposefully excluded from data extraction or analysis²³.

Variables

The demographic variables included were age, sex, ethnicity, body mass index, date of arrival, presentation site, and health insurance or self-pay. From the medical record, we obtained the onset of the injury, the primary diagnosis and secondary diagnosis, the mechanism of injury, treatment received and date, intensive care unit (ICU) usage, comorbidities, home medications use at the time of presentation, zip code of primary residency, discharge disposition, and history of prior dual-energy X-ray absorptiometry (DXA) scan. The type and number of drugs for osteoporosis treatment and prevention were classified based on their mechanism of action: biphosphonates, RANK ligand inhibitors, estrogen, selective estrogen receptor modulators, sclerostin inhibitor, parathyroid hormone analog, calcium, and vitamin D. As a surrogate for patient’s mortality at six months and 12 months, we used the last known reported activity at the hospital, regardless of services requested. In some cases, detailed notes from clinicians indicated the patient’s status.

Data source/measurements

Patients meeting inclusion criteria were identified via an electronic report from the hospital trauma registry, and the business intelligence department at the hospital system. The trauma registry is populated from information directly extracted from the patient’s medical record and is used for purposes of quality improvement and trauma level certification by the American College of Surgeons. The first author has unlimited access to generate reports from the trauma registry using the software DI Report Writer (DI Data Management System) and the second author has unlimited access to the electronic medical record used in the hospital. Home medications, diagnosis of osteoporosis and occurrence of a DXA scan were not part of the trauma registry, and were extracted via a report from the hospital business intelligence department. The files were filtered and processed using the Pandas package for Python (https://pandas.pydata.org/). Each medication was classified using string processing of its generic and the brand names as published by the National Osteoporosis Foundation (https://www.nof.org/patients/treatment/medicationadherence/), plus calcium and vitamin D. The resulting classifications were thenanalyzed for frequencies. The process was verified by the first author to ensure that no data was missing and all possible medication classes (grouped by their mechanism of action) were accounted. All presented information was part of the participant’s standard of care as documented in their medical record, and no data were collected directly from the patient. The orthopedic surgeon (second author) provided oversight of the data collection process and designed the group comparisons.

Codes and Algorithms

The fractures were identified using the International Classification of Diseases 10^th revision (ICD10) as follows: fractures of the upper end of the humerus (S42.2) or the shoulder griddle (S42.9); fractures of the lower end of radius (S52.5), lower end of ulna (S52.60), fractures of the forearm (S52.9), displaced or non-displaced femoral neck, intertrochanteric/ subtrochanteric fractures (S72.0, S72.1, and S72.2), fracture of the lower end of tibia, medial malleolus, lateral malleolus or lower leg (S82.3, S82.5, S.82.6, and S82.8). All modifiers to the above fractures based on the ICD10 were also included. Once all patients presenting with these fractures were identified, data was further filtered to retain only those patients who were older than 50 years of age at the time of injury. Subsequently, participants were excluded if the mechanism of injury was other than a fall from a height greater than one meter. Both exclusion rules were done using the parameters reported in the trauma registry. The exclusion criteria was validated by randomly verifying 20 patients from each ICD10 classification (S42, S52, S72, S82) using the clinical notes documented in the patient’s medical record.

Statistical methods

Descriptive statistics were used for the entire study population. Frequencies and column percentages were used to summarize categorical variables. The normal distribution of continuous variables was measured using the Shapiro-Wilk goodness-of-fit test. Non-normally distributed variables were analyzed using the Wilcoxon test, and normally distributed variables were analyzed using the Student t-test for independent samples. Chi-square or Fisher exact tests were used for categorical variables. Multinomial regression analyses were used to explore the changes in medication across injury types, taking into consideration the age and sex of the patients. The statistical analyses were two-sided and conducted using JMP 15.0 (SAS Institute, Inc, Carry, NC, USA). The statistical significance was set at p < 0.05.

An earlier version of this article can be found on bioRxiv (doi: https://doi.org/10.1101/2020.12.24.424289).

Participants

A total of 864 cases of non-vertebral fractures in patients older than 50 years of age were identified: 121 shoulder cases, 230 wrist cases, 297 hip cases, and 216 ankle cases. Most of the excluded clinical cases were due to the mechanisms of high energy injuries: pedestrian-MVA injuries, motor vehicle accidents, and falls from a height greater than one meter. A handful of cases were excluded due to inaccurate coding of fracture. The final cohort consisted of 719 patients.

Demographic characteristics and medication use

Table 1 describes the number of cases included in each cohort grouped by the site of injury. Patients were predominantly female (>60%), except for shoulder injuries, where 16% of the cases were females. 81% of the patients self-reported as Hispanic/Latino ethnicity, which is representative of the demographic distribution for the region. The average age of the cohort was 74.03 ± 11.90 years. Patients with ankle injuries were the youngest in the cohort, with an average age of 66.72 ± 11.08. The length of stay in the hospital was highest for hip injuries (3.19 ± 5.05 days) and lowest for wrist injuries (0.32 ± 1.87 days). Only 62 patients (8.6%) had a prior diagnosis of osteoporosis. The highest percentage of prior diagnosis was in patients with hip fractures and the lowest was in patients with ankle fractures (13.1% and 5.3%, respectively). Of the patients with prior diagnosis, 51 (82%) were taking osteoporosis medications, while 11 (17.7%) were not taking any osteoporosis medication. 98% of the patients diagnosed with osteoporosis were female, while only one patient was male. 10% of the cohort (72 patients) had a dual-energy X-ray absorptiometry (DXA) scan, from which 66 patients were female, and only six were male. The highest percentage of patients who received a DXA presented with shoulder fractures (12%). The following three classes of medications were not reported being used by any of the participants: sclerostin inhibitor (Romosozumab), parathyroid hormone analog (Teriparatide), and parathyroid hormone-related protein analog (Abaloparatide); hence these were not included in any of the results tables.

Patients were stratified based on the number of medications for osteoporosis they were taking (zero to four). 42% of the patients, regardless of the fracture site, were taking at least one medication, while 58% were not taking any medication for osteoporosis at the time of fracture (Table 2). Within those taking medications, the vast majority took one or two. The most common medications were a combination of Vitamin D plus calcium and bisphosphonates (Table 3). Other than Vitamin D plus calcium, only 16.7% were on medications for osteoporosis. The percentage of patients taking medications also varied by fracture site, with the highest percentage of patients in the hip fracture group (58.9%) and the lowest in the ankle fractures group (27.8%). Patients who had Medicare as the principal payer constituted 70% of the cohort, which is expected due to the age group. The type of health insurance influenced whether the patients took osteoporosis medication (X²= 66.78, d.f.= 4, p< 0.001; Figure 1). The odds of taking osteoporosis medication for patients on Medicare compared to self-pay was 6.84 (95% CI: 2.62 to 17.85). The odds were even higher for patients in Medicare than other payment types grouped together (government, charity, indigent): 11.76 (95% CI: 3.56 to 38.86). However, the odds of taking osteoporosis medications for those patients in Medicare compared to private insurance was similar at 1.49 (95% CI: 0.55 to 4.08; Figure 1).

Figure 1. Percent of patients taking osteoporosis medications by the type of health insurance or payer (vitamin D plus calcium excluded).

Patients who took medications for osteoporosis (“YES” column) had predominantly Medicare or Medicaid. However, at ages younger than 66, many patients were not taking any medications for osteoporosis (“NO” column) irrespective of the high incidence of private insurance. Data from 24.

Besides a documented diagnosis for osteoporosis, 10 other comorbidities were collected from the patient charts: hypertension, obesity, impaired gait or mobility, diabetes, hyperlipidemia, osteoarthritis, other cardiovascular conditions besides hypertension, thyroid-related disease, renal disease, and any cancer. The three most frequent comorbidities in the cohort were: hypertension (59.1%), obesity (47.7%), and impaired gait or mobility (45.2%). Neither hypertension nor obesity influenced whether the patient was taking any type of osteoporosis medication. However, patients with impaired gait or mobility and patients with thyroid-related disease had an increased odds ratio of receiving osteoporosis medications at 2.56 (95% CI, 1.80 to 3.65) and 5.33 (95% CI, 3.04 to 9.36), respectively. The remaining comorbidities did not influence osteoporosis medication patterns.

Outcomes-based on sex and age group

To further understand osteoporosis medication use patterns across sex, data were stratified based on three age groups: 50–65, 66–83, and 83 years and older (Figure 2). The age groups were constructed using the 25^th quartile (65 years) and 75^th quartile (83 years) for age. Regardless of fracture site and age, females had a significantly higher odds of using osteoporosis medications than males: 1.77 (95% CI, 1.21 to 2.58). Patients presenting with hip fractures at any age had the highest odds of taking osteoporosis medication compared to patients with fractures in any other body area: 2.66 (95% CI, 1.94 to 3.68). The odds for patients presenting with shoulder fractures to be on osteoporosis medications was very similar to other fractures at 1.13 (95% CI, 0.73 to 1.73). The older the patient, the higher the probability of patients taking osteoporosis medications. The odds for osteoporosis medication use in patients in the 84 years and older group compared to patients in the 50–65 years group were 4.92 (95% CI, 3.12 to 7.82). Similarly, patients in the 66–83 years group had odds of taking medication of 3.75 (95% CI, 2.29 to 5.19) than the younger group.

Figure 2. Percent of patients using osteoporosis medications sub-divided by sex and age groups.

Zero (0) value in the figure represents patients that were not taking any medications for osteoporosis. It is evident that males were using fewer medications (if any) than females and this was obvious for hip fractures. Females with shoulder fractures had the largest increase across age groups in the percentage of patients taking medications. Please refer to the Results sections for odds ratios. Data from 24.

Study limitations

Retrospective cohort studies provide a quick estimate when no previous data on the topic exist, especially for specific populations, but it also carries a series of limitations. Information regarding the length that the patients have been taking the medication was not collected, nor the effectiveness of such medications. While the number of patients taking medications in the current study is not suggestive of underreporting, it is always possible that patients are taking calcium plus vitamin D as part of their regular multi-vitamins but do not consider vitamins as “medicine” or “treatment.” Hence the possibility of underreporting is impossible to rule out. An attempt to verify previous fracture history was initiated, but we ruled out acquiring this information given that our region has three large hospitals. The possibility of patients seeking care for a prior fracture at a different facility is high. Thus, future studies should consider a multi-institutional design, allowing the recording of this type of information. We acknowledge an intrinsic bias in the study design by selecting patients that already present with a fracture. However, the presence of a fragility fracture is one of the critical factors for the diagnosis of osteoporosis⁷. Based on the new practice guidelines for endocrinologists, the current patient population would fall on the very high-risk group based on the presence of a fracture within the previous 12 months³⁵. Despite the limitations, the current study sets the stage for designing prospective interventions in high-risk groups for osteoporotic fractures.

Underlying data

Zenodo: Retrospective analysis of the use of osteoporosis medication at the presentation of non-vertebral fragility fractures in a predominantly Hispanic population. https://doi.org/10.5281/zenodo.4526306²⁴.

The project contains the following underlying data:

Reporting guidelines

Zenodo: STROBE checklist for “Retrospective analysis of the use of osteoporosis medication at the presentation of non-vertebral fragility fractures in a predominantly Hispanic population”. http://doi.org/10.5281/zenodo.4557854³⁷.

Data are available under the terms of the Creative Commons Attribution 4.0 International (CC BY 4.0).