Abstract
Context:
Testosterone prescribing rates have increased substantially in the past decade. However, little is known about the context within which such prescriptions occur.
Objective:
We evaluated provider- and site-level determinants of receipt of testosterone and of guideline-concordant testosterone prescribing.
Design:
This study was cross-sectional in design.
Setting:
This study was conducted at the Veterans Health Administration (VA).
Participants:
Study participants were a national cohort of male patients who had received at least one outpatient prescription within the VA during fiscal year (FY) 2008 to FY 2012. A total of 38,648 providers and 130 stations were associated with these patients.
Main Outcome Measure:
This study measured receipt of testosterone and guideline-concordant testosterone prescribing.
Results:
Providers ranging in age from 31 to 60 years, with less experience in the VA [all adjusted odds ratio (AOR), <2; P < 0.01] and credentialed as medical doctors in endocrinology (AOR, 3.88; P < 0.01) and urology (AOR, 1.48; P < 0.01) were more likely to prescribe testosterone compared with older providers, providers of longer VA tenure, and primary care providers, respectively. Sites located in the West compared with the Northeast [AOR, 1.75; 95% confidence interval (CI), 1.45–2.11] and care received at a community-based outpatient clinic compared with a medical center (AOR, 1.22; 95% CI, 1.20–1.24) also predicted testosterone use. Although they were more likely to prescribe testosterone, endocrinologists were also more likely to obtain an appropriate workup before prescribing compared with primary care providers (AOR, 2.14; 95% CI, 1.54–2.97).
Conclusions:
Our results highlight the opportunity to intervene at both the provider and the site levels to improve testosterone prescribing. This study also provides a useful example of how to examine contributions to prescribing variation at different levels of the health care system.
We examined providers and sites as predictors of testosterone use and guideline-concordant testosterone prescribing in the VA. Prescribing varied by provider age, specialty, years in VA, and region.
There has been a large increase in testosterone prescribing in the United States over the past decade (1). New testosterone prescriptions within the Veterans Health Administration (VA) also increased substantially between 2009 and 2012 (2). From fiscal year (FY) 2008 to FY 2014, among all classes of medications, testosterone was the 13th most commonly prescribed drug in the VA, ranking just below cardiovascular medications, opioids, antidepressants, and antipsychotics. Although there was an increase in testosterone use from 2000 to 2012 attributed primarily to promotion, marketing, and direct-to-consumer advertising of testosterone products (3), since 2013 there has been a drop in testosterone prescription sales (4) after some reports that have suggested an increased risk of cardiovascular events in men taking testosterone (5, 6).
The issue of prescribing of testosterone without appropriate baseline evaluation has gained considerable attention (2, 7) and has led the US Food and Drug Administration to review the appropriate labeling for testosterone products (8, 9). However, the clinical context within which testosterone prescriptions occur remains poorly understood. The decision to prescribe medications results from complex interactions between patient-, provider-, and system-level factors (10). In a recent evaluation of patient-level factors associated with testosterone use in the VA, we found that age, obesity, and opioid use were the strongest predictors of testosterone receipt (11).
Beyond the contributions of patient-level factors, prescribing testosterone, like any medication, may be affected by provider- (12, 13) and site-level factors (14, 15). No previous study has examined whether some providers are more or less likely to prescribe testosterone than others and whether prescription use varies by site in various geographic regions. Therefore, in this study, we analyzed the provider- and site-level predictors of testosterone use, taking advantage of the multilevel data available within the VA system. Understanding the relative influence of these factors at the level of the region, site, and provider will enable the design of interventions that can improve prescribing of testosterone and, by extension, other medications.
Materials and Methods
The study was approved by the Institutional Review Board of the Bedford VA Medical Center. We examined provider and site characteristics associated with an index dispensing of testosterone among patients receiving outpatient medications in the national VA system from 1 October 2007 to 30 September 2012 (FY 2008 to FY 2012).
Study population
The study included male patients who had at least one outpatient testosterone prescription filled between FY 2009 and FY 2012 (testosterone patients, n = 132,764) and a random 10% sample of male patients who received at least one outpatient fill for another medication and did not receive testosterone between FY 2009 and FY 2012 (nontestosterone patients, n = 550,151), for a total of 682,915 patients. Patients with HIV or diagnosed disorders of the testis, pituitary, or hypothalamus were excluded. In a separate study, we have shown that most patients who receive testosterone do not have a diagnosed disorder of the pituitary, hypothalamus, or testes (11). Prescribing testosterone to patients without a strong indication might be termed a “discretionary” prescription; this is the type of prescribing we were most interested in examining.
The provider and site characteristics associated with these patients were captured at the time of the index prescription fill, which was the earliest testosterone fill for the testosterone patients or a prescription fill chosen at random for the nontestosterone patients. The provider who wrote the index prescription was designated the “provider of interest” in these analyses. The VA is organized into 130 stations, which generally consist of a main hospital and several satellite clinics. Approximately 90% of VA patients receive all their care at one station. When a patient received care at more than one station, we considered the station where the patient had the most encounters. The current study included 38,648 providers and 130 stations (Fig. 1).
Figure 1.
Composition of analytic population of providers (n = 38,648).
Independent variables: provider characteristics
Data on provider characteristics were obtained from the VA Corporate Data Warehouse and supplemented by VA human resources files. Providers were characterized by age, years in the VA, sex, credentials, and specialty. Provider age was calculated from the date of birth, and years in the VA were calculated from the employment start date. We were uncertain whether older providers might prescribe more testosterone (because of less familiarity with evidence-based medicine) or less (because the phenomenon of increased testosterone prescribing is new to them). We expected providers with more years in the VA system to prescribe testosterone less often because their practice may have been more influenced by the VA’s generally parsimonious approach to prescribing.
Credentials refer to the provider’s training and included categories such as the Doctor of Medicine (MD), Doctor of Osteopathy, physician assistant, nurse practitioner, pharmacist, and other. Specialty refers to the area of medical specialization, such as primary care/general, endocrinology, urology. We combined the provider credentials and specialty to create a single provider type variable for our analyses, such as MD, primary care; MD, urology; and physician assistant, primary care. We expected that specialists such as endocrinologists would prescribe more testosterone because patients may be referred to them specifically to consider such therapy. We were nevertheless interested in quantifying this effect. Missing values for each of the provider characteristics (age, 7%; years in VA, 11%; sex, 3%; credentials, 0.4%; specialty, 0.2%) were supplemented using multiple imputation.
Independent variables: site characteristics
Data pertaining to site characteristics were obtained from the following VA entities: VA Corporate Data Warehouse, Planning Systems Support Group, Veterans Equitable Resource Allocation system, and Office of Productivity, Efficiency & Staffing. Sites were characterized by region, patient load, complexity, and apportionment of care between a hospital and its satellite clinics. Region refers to geographic location within the United States (Midwest, South, West, and Northeast). Patient load represents the number of unique patients seen at a specific VA station. Geographic region and patient load may affect rates of testosterone prescribing, although the direction is difficult to predict. In addition, we stratified facilities by complexity rating (category 1a, 1b, 1c, 2, or 3 on the VA Allocation Resource Center Complexity Scale). The VA rates each VA Medical center on this five-level scale, with high-complexity facilities (category 1a) seeing the largest volume of patients and having the highest patient risk and complexity (16). VA complexity ratings have previously been used to measure facility complexity in VA research (17). We expected high-complexity facilities to have lower testosterone prescribing rates due to greater availability of local expertise (e.g., an academically oriented endocrinologist) to guide practice.
Because patient load and site complexity were highly correlated (correlation coefficient, −0.77), we excluded patient load from the final model. One additional site characteristic was derived using patient-level data: specifically, the apportionment of each patient’s outpatient visits between the hospital and community-based outpatient clinic (CBOC). This variable was used to ascertain whether a patient primarily received care at a CBOC or at the main facility. We expected patients who received at least some of their care at any CBOC to be more likely to receive testosterone because CBOC providers may have less access to and interaction with experts at the main site. Although other potentially relevant site-level characteristics, including academic affiliation, location of a site in a rural area, and number of providers at a site, were examined, we did not include them in the final models due to lack of significant associations with the outcome (i.e., testosterone prescribing).
Dependent variables: testosterone prescribing and appropriateness of prescribing
Our primary outcome was whether the patient received testosterone from a VA pharmacy. Among patients who did receive testosterone, we also evaluated the appropriateness of prescribing as an outcome. To determine the appropriateness of the testosterone prescription, we examined the association of each of the provider and site characteristics with testing of testosterone levels before initiation of therapy in new patients in FY 2009 to FY 2012. Clinical guidelines require documentation of low morning testosterone levels on two or more occasions along with consistent signs and symptoms (18–20). Consistent with our previous study (2) and guidelines, we defined total testosterone <300 ng/dL or free testosterone <70 pg/mL as low. We examined several different levels of appropriateness in our models: at least one low testosterone level (minimal appropriateness), at least two low levels, and at least two low levels measured in the morning (blood draw time between 5:00 and 10:00 am; maximal appropriateness). For our analyses of whether patients had received an appropriate workup before receiving testosterone, we restricted the sample to patients who first received testosterone between FY 2009 to FY 2012 (n = 99,045) to allow a 1-year look-back period for checking of testosterone levels before the index testosterone prescription. The corresponding number of providers and sites for these appropriateness analyses were 12,912 and 129, respectively.
Statistical analyses
Logistic regression was used to ascertain the likelihood of receiving testosterone and the appropriateness of testosterone testing before prescription at both the provider and site level. Mixed models were generated to adjust for clustering at the provider and site level. Analyses were conducted using SAS, version 9.3 (SAS Corp., Cary, NC).
Results
Provider-level predictors of testosterone prescribing
The association of provider characteristics with testosterone use is reported in Table 1. After adjusting for clustering, providers ranging in age from 31 to 60 years were more likely to prescribe testosterone compared with providers 61 and older in the fully adjusted model. Providers with ≤1 year of service in the VA were more likely to prescribe testosterone than providers with 16 or more years in the VA, even after controlling for provider age and other factors [adjusted odds ratio (AOR), 1.29; 95% confidence interval (CI), 1.24 to 1.33]. Compared with MDs in primary care, MDs in endocrinology (AOR, 3.88; 95% CI, 3.31 to 4.49) and MDs in urology (AOR, 1.48; 95% CI, 1.23 to 1.84) were more likely to prescribe testosterone. Other provider types with varying specialty and credential were somewhat less likely to prescribe testosterone than MDs in primary care.
Table 1.
Provider-Level Predictors of Testosterone Prescribing (n = 682,915)
| Variable | Number of Patients (Weighted %)a,b | Models Adjusted for Clustering by Provider Odds Ratio (95% CI) |
|
|---|---|---|---|
| Unadjusted (Separate Model for Each Variable) | Fully Adjustedc | ||
| Provider age, y | |||
| 22–30 | 15,612 (2.3) | 0.94 (0.88–1.00) | 0.80d (0.75–0.86) |
| 31–40 | 108,156 (15.8) | 1.31d (1.25–1.36) | 1.20d (1.15–1.26) |
| 41–50 | 200,000 (29.3) | 1.27d (1.23–1.32) | 1.22d (1.18–1.27) |
| 51–60 | 256,430 (37.5) | 1.18d (1.14–1.22) | 1.17d (1.13–1.20) |
| 61+ | 102,717 (15.0) | ref. | ref. |
| Years in VA | |||
| 0–1 | 105,087 (15.4) | 1.32d (1.28–1.37) | 1.29d (1.24–1.33) |
| 2–5 | 175,831 (25.7) | 1.17d (1.13–1.20) | 1.13d (1.09–1.17) |
| 6–10 | 172,425 (25.2) | 1.14d (1.10–1.17) | 1.09d (1.06–1.13) |
| 11–15 | 95,589 (14.0) | 1.04 (1.00–1.07) | 1.01 (0.97–1.04) |
| 16 or more | 133,983 (19.6) | ref. | ref. |
| Male sex | |||
| Yes | 357,451 (52.3) | 1.05d (1.02–1.08) | 1.04d (1.01–1.08) |
| Provider type | |||
| MD, primary care | 454,042 (66.5) | ref. | ref. |
| MD, endocrinology | 4,048 (0.6) | 3.70d (3.18–4.31) | 3.88d (3.33–4.51) |
| MD, urology | 1,029 (0.2) | 1.43d (1.17–1.75) | 1.48d (1.21–1.81) |
| MD, other specialty | 63,969 (9.4) | 0.61d (0.58–0.63) | 0.60d (0.58–0.63) |
| PA, primary care | 28,501 (4.2) | 1.03 (0.95–1.12) | 1.04 (0.96–1.13) |
| PA, non–primary care | 15,136 (2.2) | 0.46d (0.42–0.51) | 0.47d (0.42–0.51) |
| Nurse practitioner | 111,720 (16.4) | 0.77d (0.74–0.80) | 0.79d (0.76–0.82) |
| Pharmacist | 2,498 (0.4) | 0.54d (0.46–0.63) | 0.56d (0.48–0.65) |
| Other credentials | 1,972 (0.3) | 0.28d (0.23–0.35) | 0.27d (0.22–0.33) |
Abbreviation: PA, physician assistant.
Testosterone patients, n = 132,764; nontestosterone patients, n = 550,151.
Percentage calculation considers that nontestosterone patients were sampled with a rate of 10%.
The fully adjusted model uses all provider characteristics specified in the table.
Odds ratio differs from the reference group; that is, patients without the specified factor at the level of P < 0.01. All other odds ratios do not differ from the reference group at the 0.01 level.
Site-level predictors of testosterone prescribing
Table 2 presents the association of site characteristics with testosterone use. In the fully adjusted model, sites located in the West (AOR, 1.75; 95% CI, 1.45 to 2.11), South (AOR, 1.63; 95% CI, 1.36 to 1.95), and Midwest (AOR, 1.37; 95% CI, 1.13 to 1.67) were more likely to prescribe testosterone than sites in the Northeast. Patients receiving at least some of their care at a CBOC were more likely to receive testosterone (AOR, 1.22; 95% CI, 1.20 to 1.24) compared with patients receiving care only at the main facility. Site complexity level was not significantly associated with testosterone prescribing after adjusting for clustering and the other site characteristics.
Table 2.
Site-level Predictors of Testosterone Prescribing (n = 682,915)
| Variable | Number of Patients (Weighted %)a,b | Models Controlling for Clustering by Site Odds Ratio (95% CI) |
|
|---|---|---|---|
| Unadjusted (Separate Model for Each Variable) | Fully Adjustedc | ||
| Region | |||
| Midwest | 142,405 (21.2) | 1.38 (1.12–1.70) | 1.37d (1.13–1.67) |
| South | 285,269 (41.2) | 1.69d (1.40–2.03) | 1.63d (1.36–1.95) |
| West | 161,797 (23.1) | 1.75d (1.43–2.13) | 1.75d (1.45–2.11) |
| Northeast | 93,444 (14.5) | ref. | ref. |
| Whether patient received any care in VA at outpatient clinic outside of hospital | |||
| Yes | 491,001 (71.2) | 1.22d (1.20–1.24) | 1.22d (1.20–1.24) |
| No | 191,914 (28.8) | ref. | ref. |
| Complexity group | |||
| Category 1a (most complex) | 288,642 (41.7) | 1.32d (1.07–1.62) | 1.21 (1.01–1.45) |
| Category 1b | 86,416 (12.7) | 1.10 (0.86–1.40) | 1.01 (0.81–1.25) |
| Category 1c | 136,298 (19.9) | 1.21 (0.96–1.53) | 1.15 (0.94–1.43) |
| Category 2 | 91,598 (13.9) | 0.96 (0.78–1.18) | 0.94 (0.78–1.12) |
| Category 3 (least complex) | 79,961 (11.8) | ref. | ref. |
Testosterone patients, n = 132,764; nontestosterone patients, n = 550,151.
Percentage calculation considers that nontestosterone patients were sampled with a rate of 10%.
The fully adjusted model uses all site characteristics specified in the table.
Odds ratio differs from the reference group; that is, patients without the specified factor at the level of P < 0.01. All other odds ratios do not differ from the reference group at the 0.01 level.
Association of provider and site characteristics with appropriate testosterone testing
The association of site and provider characteristics with testosterone testing (at least one low testosterone level, at least two low testosterone levels, and at least two low morning testosterone levels) is shown in Table 3. Providers in the age range of 31 to 40 years (AOR, 1.30; 95% CI, 1.12 to 1.50) were more likely to measure two low morning testosterone levels compared with the older providers. Conversely, providers with less than 1 year in the VA were less likely to check for two low testosterone levels in the morning compared with the providers with 16 or more years in the VA (AOR, 0.82; 95% CI, 0.72 to 0.94). Male providers were less likely to check two morning low testosterone levels before prescribing testosterone, compared with female providers (AOR, 0.88; 95% CI, 0.81 to 0.96). MDs in endocrinology were more likely to record two low testosterone levels in the morning than MDs in primary care (AOR, 2.14; 95% CI, 1.54 to 2.97).
Table 3.
Provider- and Site-Level Predictors of Appropriate Testing in New Testosterone Patients (n = 99,045)
| Variable | Number of Patients (%) | Fully Adjusted Modelsa Adjusted for Clustering by Provider and Site Odds Ratio (95% CI) |
||
|---|---|---|---|---|
| At Least 1 Low T (n = 78,525) | At Least 2 Low T (n = 18,375) | At Least 2 Low T in Morning (n = 5,379) | ||
| Provider age, y | ||||
| 22–30 | 1,553 (1.6) | 0.93 (0.79–1.09) | 1.46b (1.23–1.73) | 1.29 (0.98–1.68) |
| 31–40 | 15,194 (15.3) | 0.98 (0.90–1.06) | 1.23b (1.12–1.36) | 1.30b (1.12–1.50) |
| 41–50 | 29,399 (29.7) | 0.99 (0.92–1.07) | 1.12b (1.03–1.22) | 1.14 (1.00–1.30) |
| 51–60 | 38,206 (38.6) | 1.03 (0.96–1.10) | 1.11b (1.03–1.20) | 1.12 (1.00–1.26) |
| 61+ | 14,693 (14.8) | ref. | ref. | ref. |
| Years in VA | ||||
| 0, 1 | 16,150 (16.3) | 0.96 (0.89–1.04) | 0.79b (0.73–0.86) | 0.82b (0.72–0.94) |
| 2–5 | 27,090 (27.4) | 0.99 (0.92–1.06) | 0.85b (0.79–0.92) | 0.95 (0.85–1.07) |
| 6–10 | 25,127 (25.4) | 1.03 (0.96–1.10) | 0.86b (0.79–0.92) | 0.91 (0.81–1.03) |
| 11–15 | 13,295 (13.4) | 0.96 (0.89–1.04) | 0.91 (0.84–0.99) | 1.06 (0.93–1.20) |
| 16 or more | 17,383 (17.6) | ref. | ref. | ref. |
| Male sex | ||||
| Yes | 53,590 (54.1) | 0.88b (0.84–0.93) | 0.87b (0.82–0.93) | 0.88b (0.81–0.96) |
| Provider type | ||||
| MD, primary care | 70,351 (71.0) | ref. | ref. | ref. |
| MD, endocrinology | 912 (0.9) | 1.29 (1.01–1.65) | 2.09b (1.63–2.67) | 2.14b (1.54–2.97) |
| MD, urology | 205 (0.2) | 0.98 (0.65–1.47) | 1.37 (0.88–2.14) | 0.79 (0.36–1.73) |
| MD, other specialty | 7,300 (7.4) | 1.32b (1.21–1.43) | 1.32b (1.21–1.44) | 1.12 (0.98–1.27) |
| PA, primary care | 4,016 (4.1) | 0.83b (0.73–0.93) | 0.87 (0.75–1.00) | 1.01 (0.82–1.23) |
| PA, non–primary care | 1,411 (1.4) | 1.20 (0.99–1.45) | 1.19 (0.97–1.46) | 1.30 (0.97–1.73) |
| Nurse practitioner | 14,526 (14.7) | 0.92 (0.85–0.98) | 0.97 (0.89–1.05) | 1.02 (0.90–1.15) |
| Pharmacist | 243 (0.2) | 1.04 (0.72–1.49) | 0.89 (0.60–1.31) | 0.73 (0.38–1.42) |
| Other credentials | 81 (0.1) | 0.72 (0.42–1.24) | 0.87 (0.46–1.66) | 0.60 (0.18–2.01) |
| Region | ||||
| Northeast (n = 24)c | 9,414 (9.5) | ref. | ref. | ref. |
| Midwest (n = 27)c | 19,588 (19.8) | 1.36b (1.26–1.47) | 0.90 (0.83–0.99) | 0.88 (0.78–1.01) |
| South (n = 46)c | 44,303 (44.7) | 1.28b (1.19–1.37) | 0.86b (0.79–0.93) | 0.80b (0.72–0.90) |
| West (n = 32)c | 25,740 (26.0) | 1.38b (1.28–1.49) | 0.97 (0.90–1.06) | 0.85 (0.75–0.97) |
| Whether patient received any care in VA at outpatient clinic outside of hospital | ||||
| Yes | 75,103 (75.8) | 0.97b (0.93–1.02) | 0.91b (0.87–0.96) | 0.83b (0.77–0.90) |
| Complexity group | ||||
| Category 1a (most complex) | 44,184 (44.6) | 1.10b (1.03–1.17) | 1.20b (1.12–1.29) | 1.33b (1.18–1.49) |
| Category 1b | 12,007 (12.1) | 0.95 (0.88–1.03) | 1.07 (0.98–1.17) | 1.00 (0.86–1.16) |
| Category 1c | 20,112 (20.3) | 1.00 (0.93-1.08) | 1.10 (1.02–1.20) | 1.23b (1.08–1.41) |
| Category 2 | 11,650 (11.8) | 0.82b (0.76–0.89) | 0.98 (0.89–1.07) | 1.17 (1.01–1.35) |
| Category 3 (least complex) | 11,092 (11.2) | ref. | ref. | ref. |
Abbreviation: PA, physician assistant.
Each fully adjusted model uses all provider and site characteristics specified on the table.
Odds ratio differs from the reference group; that is, patients without the specified factor, at the level of P < 0.01. All other odds ratios do not differ from the reference group at the 0.01 level.
Total number of sites per region.
Among the site characteristics, regional location of sites had a significant association with testosterone-level testing. Sites in the Northeast were more likely to check for two low testosterone levels as well as two low morning testosterone levels. Patients receiving any care at a CBOC were less likely to be checked for two low morning testosterone levels before initiating therapy (AOR, 0.83; 95% CI, 0.77 to 0.90). In terms of site complexity, overall, category 1a sites (the largest and most complex sites; AOR, 1.33; 95% CI, 1.18 to 1.49) and 1c sites were more likely to adhere to appropriate testing of two low testosterone levels in the morning compared with category 3 (AOR,1.23; 95% CI, 1.08 to 1.41).
Discussion
This study examined provider and site predictors of testosterone prescribing in the VA. Findings suggest that providers ranging in age from 31 to 60 years, with less experience in the VA, and credentialed as MDs in endocrinology and urology were more likely to prescribe testosterone to their patients. Geographic region and care received at a CBOC also significantly predicted testosterone use. Perhaps the most important finding is that physicians specializing in endocrinology and urology, those with more experience in the VA, and those practicing at a major VA medical facility were more likely to perform appropriate diagnostic evaluation before prescribing testosterone. Assuming that the physician specialists in endocrinology and urology have received a higher level of training in managing male reproductive disorders than nonspecialists, these findings raise the possibility that testosterone prescribing practices could be improved by targeted provider training or computer-based decision support.
Providers’ decision-making processes are influenced by their personal characteristics, such as sex, age, and ethnicity (21). Therefore, it is not surprising that we found that younger provider age and fewer years practicing in the VA were associated with more testosterone prescribing. Although this study does not provide direct insight into why these patterns emerged, there are some likely explanations. For example, providers with fewer years in practice in the VA could be prescribing testosterone more often because they may be less fully acculturated to the generally parsimonious prescribing norms in the VA system, including the use of drug detailing, prior authorization requests, etc., which focus on evidence for benefit rather than cost per se.
Although younger providers were more likely to prescribe testosterone, they were more likely to document low testosterone levels before prescribing, as were providers with more experience in the VA system. This could be because providers trained more recently may be more aware of clinical practice guidelines and the principles of evidence-based medicine. Our finding of younger provider age as a determinant for guideline-concordant prescribing echoes those of other studies that have evaluated the relationship between clinical experience (defined as physician age and time in practice) and performance (22, 23). These previous studies suggested that older physicians are less likely to adhere to appropriate standards of care and may also have poorer patient outcomes.
We also found substantial contributions to prescribing of testosterone at the site level. Sites in the Northeast were not only less likely to prescribe this therapy but also more likely to document appropriate testosterone testing before initiating therapy; indeed, the two may be related. It is possible that these geographic differences reflect cultural factors, which may in turn affect the propensity of patients in the West and Southeast to request testosterone or provider comfort with discussing it. Our results are consistent with published findings on outpatient antibiotic prescribing in the VA, which have substantial variation in the levels of antibiotic dispensing throughout regions (24, 25). Similarly, we found that patients treated at a CBOC were more likely to receive testosterone and less likely to have received appropriate testing. Providers practicing at CBOCs may be less current with medical knowledge due to their remoteness from academic medical centers, which may contribute to a lower likelihood of appropriate prescribing. However, this medical center-CBOC divide has been mitigated to some extent over the years with the availability of online educational resources and with the provision of electronic consultations of academic and specialty expertise to providers throughout the VA, including CBOCs. These findings collectively suggest that local factors may play an important role in influencing prescribing and offer the opportunity to focus quality improvement efforts in specific regions and settings. Beyond testosterone, this study provides a useful example of how to examine contributions to prescribing variation at different levels of the health care system.
To the extent that individual providers and sites vary in their decision-making regarding testosterone prescribing, our results highlight the opportunity to intervene at both the provider and the site levels. Various interventions in the literature have been used to improve and standardize provider prescribing practice, such as mailed educational materials, educational programs, audit and feedback, and academic detailing (26). Through such efforts, providers’ knowledge of how to manage male hypogonadism could be increased, which could help promote adherence to existing clinical practice guidelines for testosterone prescribing (18–20).
Patient demand for testosterone is likely to play an important role. In other areas of prescribing, research has shown that providing tools for providers to address patient demands without resorting to a prescription, such as a “non-prescription pad,” can help curb inappropriate prescribing (27–31). Using a similar tool here could help empower providers to offer an alternative to testosterone therapy for patients who want to feel “peppier,” such as advice about improving diet, exercise, or sleep. Such an approach is likely to simultaneously address patients’ symptoms and expectations. However, even with the benefit of such approaches, changing provider prescribing practices and behavior is a challenging task.
Other provider-oriented interventions could operate at the systems level, such as decision support implemented through the electronic medical record (32). Another possible approach to creating system-level changes would be to draw on the role of pharmacists to ensure appropriate documentation of a clinical need before dispensing the testosterone fill. Pharmacist-led interventions, such as point-of-care medication review and pharmacist consultations, have effectively reduced inappropriate prescribing (33, 34). System-oriented approaches may be the most powerful approaches to change practice and could be even more effective when combined with interventions targeted at the provider level (30).
Our study has important strengths and some limitations. We used a large and highly detailed dataset from the nation’s largest integrated system of care and adjusted for a comprehensive set of provider and site variables. However, this dataset does not permit an evaluation of provider attitudes and whether or how these values influence testosterone prescribing practices. We are exploring provider and patient attitudes toward testosterone therapy in an ongoing qualitative study. Also, it is possible that the provider and site determinants of testosterone prescribing outside of the VA may differ from those within this system, although we suspect that these differences would likely be minor. Another limitation is that our study did not consider dual-care veterans (i.e., patients who had been started on testosterone in the private sector and then came to VA to have the medication renewed or who had laboratory work in the private sector and then came to the VA to receive testosterone).
In summary, we characterized the provider- and site-level determinants of testosterone prescribing as well as inappropriate testosterone prescribing. The VA and other health care systems can use these insights to design more effective interventions that can help decrease inappropriate prescribing of testosterone while ensuring that those patients who can benefit the most can still receive it.
Acknowledgments
Acknowledgments
This work was supported by the Department of Veterans Affairs, Veterans Health Administration, Health Services Research and Development Service. Dr. Jasuja is a VA HSR&D Career Development awardee at the Bedford VA (CDA 13-265). The views expressed in this article are those of the author(s) and do not necessarily represent the views of the Department of Veterans Affairs.
Disclosure Summary: S.B. has the following potential conflicts of interest. Research grants: NIA, NINR, FNIH, AbbVie, MIB, LLC; equity interest: FPT, LLC; consultant: AbbVie. The remaining authors have nothing to disclose.
Footnotes
- AOR
- adjusted odds ratio
- CBOC
- community-based outpatient clinic
- CI
- confidence interval
- VA
- Veterans Health Administration
- FY
- fiscal year.
References
- 1.Handelsman DJ. Global trends in testosterone prescribing, 2000-2011: expanding the spectrum of prescription drug misuse. Med J Aust. 2013;199(8):548–551. [DOI] [PubMed] [Google Scholar]
- 2.Jasuja GK, Bhasin S, Reisman JI, Berlowitz DR, Rose AJ. Ascertainment of testosterone prescribing practices in the VA. Med Care. 2015;53(9):746–752. [DOI] [PubMed] [Google Scholar]
- 3.Layton JB, Kim Y, Alexander GC, Emery SL. Association between direct-to-consumer advertising and testosterone testing and initiation in the United States, 2009-2013. JAMA. 2017;317(11):1159–1166. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Jasuja GK, Bhasin S, Rose AJ. Patterns of testosterone prescription overuse. Curr Opin Endocrinol Diabetes Obes. 2017;24(3):240–245. [DOI] [PubMed] [Google Scholar]
- 5.Vigen R, O’Donnell CI, Barón AE, Grunwald GK, Maddox TM, Bradley SM, Barqawi A, Woning G, Wierman ME, Plomondon ME, Rumsfeld JS, Ho PM. Association of testosterone therapy with mortality, myocardial infarction, and stroke in men with low testosterone levels. JAMA. 2013;310(17):1829–1836. [DOI] [PubMed] [Google Scholar]
- 6.Finkle WD, Greenland S, Ridgeway GK, Adams JL, Frasco MA, Cook MB, Fraumeni JF Jr, Hoover RN. Increased risk of non-fatal myocardial infarction following testosterone therapy prescription in men. PLoS One. 2014;9(1):e85805. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Layton JB, Li D, Meier CR, Sharpless JL, Stürmer T, Jick SS, Brookhart MA. Testosterone lab testing and initiation in the United Kingdom and the United States, 2000 to 2011. J Clin Endocrinol Metab. 2014;99(3):835–842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.US Food and Drug Administration. FDA drug safety communication: FDA evaluating risk of stroke, heart attack and death with FDA-approved testosterone products. Available at: http://www.fda.gov/Drugs/DrugSafety/ucm383904.htm. Accessed 6 June 2016.
- 9.US Food and Drug Administration. FDA adding general warning to testosterone products about potential for venous blood clots. Available at: http://www.fda.gov/drugs/drugsafety/ucm401746.htm. Accessed 6 June 2016.
- 10.Aagaard EM, Gonzales R, Camargo CA Jr, Auten R, Levin SK, Maselli J, Metlay J. Physician champions are key to improving antibiotic prescribing quality. Jt Comm J Qual Patient Saf. 2010;36(3):109–116. [DOI] [PubMed] [Google Scholar]
- 11.Jasuja GK, Bhasin S, Reisman JI, et al. Who gets testosterone? Patient characteristics associated with testosterone prescribing: a cross-sectional study. J Gen Intern Med. 2017;32(3):304–311. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Deepmala D, Franz L, Aponte C, Agrawal M, Jiang W. Identification of provider characteristics influencing prescription of analgesics: a systematic literature review. Pain Pract. 2013;13(6):504–513. [DOI] [PubMed] [Google Scholar]
- 13.Kanwal F, Hoang T, Spiegel BMR, Eisen S, Dominitz JA, Gifford A, Goetz M, Asch SM. Predictors of treatment in patients with chronic hepatitis C infection: role of patient versus nonpatient factors. Hepatology. 2007;46(6):1741–1749. [DOI] [PubMed] [Google Scholar]
- 14.Cox ER, Motheral BR, Henderson RR, Mager D. Geographic variation in the prevalence of stimulant medication use among children 5 to 14 years old: results from a commercially insured US sample. Pediatrics. 2003;111(2):237–243. [DOI] [PubMed] [Google Scholar]
- 15.Gellad WF, Good CB, Amuan ME, Marcum ZA, Hanlon JT, Pugh MJ. Facility-level variation in potentially inappropriate prescribing for older veterans. J Am Geriatr Soc. 2012;60(7):1222–1229. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Stefos T, LaVallee N, Holden F. Fairness in prospective payment: a clustering approach. Health Serv Res. 1992;27(2):239–261. [PMC free article] [PubMed] [Google Scholar]
- 17.Huang PY, Yano EM, Lee ML, Chang BL, Rubenstein LV. Variations in nurse practitioner use in Veterans Affairs primary care practices. Health Serv Res. 2004;39(4 Pt 1):887–904. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Bhasin S, Cunningham GR, Hayes FJ, Matsumoto AM, Snyder PJ, Swerdloff RS, Montori VM; Task Force, Endocrine Society . Testosterone therapy in men with androgen deficiency syndromes: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2010;95(6):2536–2559. [DOI] [PubMed] [Google Scholar]
- 19.Petak SM, Nankin HR, Spark RF, Swerdloff RS, Rodriguez-Rigau LJ; American Association of Clinical Endocrinologists . American Association of Clinical Endocrinologists Medical Guidelines for clinical practice for the evaluation and treatment of hypogonadism in adult male patients–2002 update. Endocr Pract. 2002;8(6):440–456. [PubMed] [Google Scholar]
- 20.Wang C, Nieschlag E, Swerdloff R, Behre HM, Hellstrom WJ, Gooren LJ, Kaufman JM, Legros JJ, Lunenfeld B, Morales A, Morley JE, Schulman C, Thompson IM, Weidner W, Wu FC. ISA, ISSAM, EAU, EAA and ASA recommendations: investigation, treatment and monitoring of late-onset hypogonadism in males. Int J Impot Res. 2009;21(1):1–8. [DOI] [PubMed] [Google Scholar]
- 21.Beers MH, Fingold SF, Ouslander JG, Reuben DB, Morgenstern H, Beck JC. Characteristics and quality of prescribing by doctors practicing in nursing homes. J Am Geriatr Soc. 1993;41(8):802–807. [DOI] [PubMed] [Google Scholar]
- 22.Choudhry NK, Fletcher RH, Soumerai SB. Systematic review: the relationship between clinical experience and quality of health care. Ann Intern Med. 2005;142(4):260–273. [DOI] [PubMed] [Google Scholar]
- 23.Cadieux G, Tamblyn R, Dauphinee D, Libman M. Predictors of inappropriate antibiotic prescribing among primary care physicians. CMAJ. 2007;177(8):877–883. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hicks LA, Bartoces MG, Roberts RM, Suda KJ, Hunkler RJ, Taylor TH Jr, Schrag SJ. US outpatient antibiotic prescribing variation according to geography, patient population, and provider specialty in 2011. Clin Infect Dis. 2015;60(9):1308–1316. [DOI] [PubMed] [Google Scholar]
- 25.Aspinall SL, Berlin JA, Zhang Y, Metlay JP. Facility-level variation in antibiotic prescriptions for veterans with upper respiratory infections. Clin Ther. 2005;27(2):258–262. [DOI] [PubMed] [Google Scholar]
- 26.Anderson GM, Lexchin J. Strategies for improving prescribing practice. CMAJ. 1996;154(7):1013–1017. [PMC free article] [PubMed] [Google Scholar]
- 27.Coenen S, Francis N, Kelly M, Hood K, Nuttall J, Little P, Verheij TJ, Melbye H, Goossens H, Butler CC; GRACE Project Group . Are patient views about antibiotics related to clinician perceptions, management and outcome? A multi-country study in outpatients with acute cough. PLoS One. 2013;8(10):e76691. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Butler CC, Rollnick S, Pill R, Maggs-Rapport F, Stott N. Understanding the culture of prescribing: qualitative study of general practitioners’ and patients’ perceptions of antibiotics for sore throats. BMJ. 1998;317(7159):637–642. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Centers for Disease Control and Prevention, National Center for Immunization and Respiratory Diseases. Get smart: know when antibiotics work. Available at: http://www.cdc.gov/getsmart/community/materials-references/print-materials/hcp/viral-rx-pad.html. Accessed 6 June 2016.
- 30.Belongia EA, Knobloch MJ, Kieke BA, Davis JP, Janette C, Besser RE. Impact of statewide program to promote appropriate antimicrobial drug use. Emerg Infect Dis. 2005;11(6):912–920. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Alweis R, Greco M, Wasser T, Wenderoth S. An initiative to improve adherence to evidence-based guidelines in the treatment of URIs, sinusitis, and pharyngitis. J Community Hosp Intern Med Perspect. 2014;4:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Shelton JB, Ochotorena L, Bennett C, Shekelle P, Kwan L, Skolarus T, Goldzweig C. Reducing PSA-based prostate cancer screening in men aged 75 years and older with the use of highly specific computerized clinical decision support. J Gen Intern Med. 2015;30(8):1133–1139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Hanlon JT, Weinberger M, Samsa GP, Schmader KE, Uttech KM, Lewis IK, Cowper PA, Landsman PB, Cohen HJ, Feussner JR. A randomized, controlled trial of a clinical pharmacist intervention to improve inappropriate prescribing in elderly outpatients with polypharmacy. Am J Med. 1996;100(4):428–437. [DOI] [PubMed] [Google Scholar]
- 34.Kaur S, Mitchell G, Vitetta L, Roberts MS. Interventions that can reduce inappropriate prescribing in the elderly: a systematic review. Drugs Aging. 2009;26(12):1013–1028. [DOI] [PubMed] [Google Scholar]