Factors Associated With Potentially Inappropriate Phosphodiesterase-5 Inhibitor Use for Pulmonary Hypertension in the United States, 2006 to 2015
Use of phosphodiesterase-5 inhibitors (PDE5i) for groups 2 and 3 pulmonary hypertension (PH) is rising nationally, despite guidelines recommending against this low-value practice. Although receiving care across healthcare systems is encouraged to increase veterans’ access to specialists critical for PH management, receiving care in 2 systems may increase risk of guideline-discordant prescribing. We sought to identify factors associated with prescribing of PDE5i for group 2/3 PH, particularly, to test the hypothesis that veterans prescribed PDE5i for PH in the community (through Medicare) will have increased risk of subsequently receiving potentially inappropriate treatment in Veterans Health Administration (VA).
Methods and Results:
We constructed a retrospective cohort of 34 775 Medicare-eligible veterans with group 2/3 PH by linking national patient-level data from VA and Medicare from 2006 to 2015. We calculated adjusted odds ratios (ORs) of receiving daily PDE5i treatment for PH in VA using multivariable models with facility-specific random effects. In this cohort, 1556 veterans received VA prescriptions for PDE5i treatment for group 2/3 PH. Supporting our primary hypothesis, the variable most strongly associated with PDE5i treatment in VA for group 2/3 PH was prior treatment through Medicare (OR, 6.5 [95% CI, 4.9–8.7]). Other variables strongly associated with increased likelihood of VA treatment included more severe disease as indicated by recent right heart failure (OR, 3.3 [95% CI, 2.8–3.9]) or respiratory failure (OR, 3.7 [95% CI, 3.1–4.4]) and prior right heart catheterization (OR, 3.8 [95% CI, 3.4–4.3]).
Our data suggest a missed opportunity to reassess treatment appropriateness when pulmonary hypertension patients seek prescriptions from VA—a relevant finding given policies promoting shared care across VA and community settings. Interventions are needed to reinforce awareness that pulmonary vasodilators are unlikely to benefit group 2/3 pulmonary hypertension patients and may cause harm.
What Is Known
Despite the US and European guidelines recommending against routine use of pulmonary vasodilators such as phosphodiesterase-5 inhibitors (PDE5i) in pulmonary hypertension (PH) secondary to left heart disease (group 2 PH) and chronic lung disease (group 3 PH), use is rising over time, exposing patients to harm and increasing costs.
A growing national trend toward shared care for veterans may increase the potential for care fragmentation and risk of guideline-discordant treatment with PDE5i for groups 2 and 3 PH.
What the Study Adds
The strongest predictor of PDE5i treatment in Veterans Health Administration for group 2/3 PH was prior treatment in the community, with an odds ratio of 6.5.
Other factors associated with PDE5i treatment for group 2/3 PH included more severe PH disease, prior right heart catheterization, and higher facility PH volume.
Understanding these factors associated with use of PDE5i for group 2/3 PH may help address larger issues driving inappropriate prescribing in other contexts, such as prescribing inertia, lack of provider awareness, workplace culture, subspecialty influences, and care fragmentation.
Providing patient-centered, high-value care is a fundamental goal of clinicians and health systems. Yet, an estimated 30% of all medical spending in the United States is wasteful and does not add value.1 One significant contributor to waste—overuse or misuse of medications—can result in adverse drug effects, decreased quality of life, increased hospitalizations, and even death.2,3 The American Board of Internal Medicine’s Choosing Wisely Campaign identifies low-value practices, including inappropriate prescribing, to curb overuse and mitigate patient harm.4 Within this campaign, the American College of Chest Physicians and American Thoracic Society identified routine use of pulmonary vasodilators for groups 2 and 3 pulmonary hypertension (PH) as one such practice.5
PH is a challenging condition to manage, with high morbidity and mortality. While patients with group 1 PH (also known as pulmonary arterial hypertension) clearly benefit from treatment with pulmonary vasodilators, patients with the most common forms of PH—PH secondary to underlying left-sided heart disease (group 2 PH) or chronic hypoxic lung disease (group 3 PH)—have no established benefit from treatment.6–8 In fact, some studies suggest serious harm for patients with groups 2 and 3 PH treated with vasodilators, including worsened hypoxemia, renal failure, right-sided heart failure, shock, and potentially higher mortality.9–12 Given the lack of benefit and potential for harm, clinical practice guidelines recommend against routine use of pulmonary vasodilators for groups 2 and 3 PH and instead direct clinicians to optimize treatment for the underlying cardiac or pulmonary condition.13,14
Despite these recommendations, use of pulmonary vasodilators, particularly phosphodiesterase-5 inhibitors (PDE5i), for group 2/3 PH is increasing over time.15,16 Given the high cost of these medications, veterans who share care across the Veterans Health Administration (VA) and community settings may seek to fill PDE5i prescriptions from VA, where copays tend to be substantially lower.17 With a growing national trend toward shared care for veterans,18,19 comanagement of PH patients is also likely to increase. While shared care may increase veterans’ access to specialty care critical for PH management, it also increases the potential for care fragmentation and the risk of guideline-discordant care.20,21 To understand the influences on potentially inappropriate PDE5i prescribing for PH in VA, and in particular the impact of shared care, we performed a national retrospective cohort study of veterans diagnosed with group 2 or 3 PH over a decade. Our primary hypothesis was that veterans with group 2/3 PH initiated on PDE5i treatment via Medicare would be at increased risk for subsequently receiving potentially inappropriate prescriptions in VA.
Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to the study's principal investigator, Dr Renda Wiener, at the Center for Healthcare Organization and Implementation Research (renda.
Study Design and Data Source
We conducted a retrospective cohort analysis of all Medicare-eligible veterans with group 2/3 PH diagnosed between January 1, 2006, and December 31, 2015, linking national patient-level data from the VA and Centers for Medicare and Medicaid Services. The Edith Nourse Rogers Memorial VA Hospital Institutional Review Board approved this study.
From the population of veterans who used VA services during the study period, we identified all patients with incident PH between 2006 and 2015, defined by at least 2 International Classification of Diseases, Ninth or Tenth Revision, diagnosis codes for PH (416.xx or I27.x) occurring either in VA or Medicare. We defined the incident PH date as the date of the first PH code. Those with a PH code between October 1, 1999 (the inception of the VA Corporate Data Warehouse), and December 31, 2005, were excluded. We restricted our sample to Medicare-eligible (age ≥65 years or in the Medicare denominator file) veterans who were active pharmacy users (excluding those with no VA or Medicare prescriptions in the year before PH diagnosis). We also excluded veterans who received PH care exclusively outside VA (ie, no PH-associated outpatient visits in VA during the study period). Finally, we narrowed our sample to veterans with group 2/3 PH using a previously validated algorithm designed to exclude veterans with pulmonary arterial hypertension, other PH groups, or uncertain pathogenesis of PH, which demonstrated a high positive predictive value for identifying group 2/3 PH.15 To maximize the specificity of group 2/3 PH, the validated algorithm erred on the side of labeling patients as group 1 over group 2/3. For example, patients with diagnosis codes for both group 1 PH-associated diagnoses and group 2 or 3 diagnoses were classified as group 1 PH. The derivation of our study cohort is shown in Figure 1.
The goal of our analysis was to identify patient- and facility-level factors associated with the probability of receiving PDE5i treatment of group 2/3 PH in VA. Our primary outcome was a binary indicator of treatment with daily PDE5i, defined by ≥1 VA prescriptions for daily PDE5i (sildenafil, vardenafil, or tadalafil) treatment, supplied for at least 15 days per month. We selected this algorithm as indicative of PDE5i treatment for PH based on (1) our prior validation against gold standard medical record abstraction,15 during which we noted that prescriptions often included instructions to cut pills in half; (2) VA policy restricting PDE5i for treatment of erectile dysfunction to ≤4 doses per month.
All exposure variables were derived using all available data from VA and Medicare. Variables were measured at the index date, defined as the initiation of daily PDE5i treatment in VA (date of first VA prescription) for cases, or for a calculated phantom prescription date for patients who did not receive VA treatment, as described below and shown in Figure 2.
Following our primary hypothesis, our primary exposure variable was prior daily PDE5i treatment for group 2/3 PH through Medicare. We also evaluated other patient- and facility-level exposures we hypothesized might be associated with prescribing of PDE5i for group 2/3 PH based on clinical experience and prior literature on inappropriate prescribing in other contexts, as defined below. Exposures were identified by their presence during a defined look-back period from the index date (Figure 2); details of the look-back period for each variable are shown in Table I in the Data Supplement.
Additional Patient-Level Exposures
We included demographic characteristics (age, sex, race, marital status, and VA enrollment priority status, which is a proxy for level of disability and VA benefits22); factors related to PH etiology, severity, or management (underlying conditions associated with group 2/3 PH, recent right heart or respiratory failure, performance of a right heart catheterization, receipt of other pulmonary vasodilators); contraindications to PDE5i use (coprescription of nitrates), acute and chronic comorbidities and vital status changes indicative of general health, and markers of healthcare utilization including long-term care, emergency department visits, hospitalizations, and intensive care unit stays in the prior year (see Table I in the Data Supplement for full list and definitions of all variables).
Additional Facility-Level Exposures
Patients were assigned to VA medical centers (facilities) based on the location of the most recent PH-associated visit on or before the date of the daily PDE5i prescription or phantom prescription date. For each VA facility, we calculated rates of daily PDE5i prescribing among all patients with group 2/3 PH seen at the facility, excluding facilities with a volume of <25 group 2/3 PH patients seen over the study period (one VA facility excluded). We then considered VA facility-level characteristics that may influence PDE5i prescribing for PH. These included factors relevant to PH care (volume of PH patients seen annually, volume of right heart catheterizations performed annually, presence of specialty pulmonary or cardiology care [defined by outpatient clinic stop codes]); geographic factors (US census region, rurality [defined by VA urban or rural designation23]) and facility complexity rating, which is comprised of patient volume and risk, level of teaching and research, number of specialists, and presence of intensive care units within the facility.24
Our statistical analyses proceeded in 2 stages. In the first stage, we imputed phantom prescription dates by estimating the number of days since PH diagnosis for patients who never received PDE5i treatment in VA. To do so, we developed a prediction model for the number of days between PH diagnosis and VA PDE5i prescription among those prescribed daily PDE5i in VA based on a comprehensive list of demographic and clinical variables. The fit of this model was then applied to the cohort that never received VA PDE5i treatment to obtain the imputed phantom prediction dates. This imputation model was created by fitting a gradient boosted tree25 for the time since PH diagnosis until VA PDE5i prescription. The approach of imputing phantom prescription dates has close connections to estimating missing disease onset dates in cohort studies.26,27
In the second stage, we modeled a binary indicator of VA PDE5i prescription based on the full cohort as a function of covariate information at the time of PDE5i initiation in VA for the treated patients and the phantom prescription dates for patients who did not receive PDE5i treatment in VA. We considered 2 models. First, we modeled VA PDE5i prescription as an ordinary multivariable logistic regression depending on the patient and facility exposures described above. Second, we extended the logistic regression model to include normally distributed facility-specific random effects. To address potential issues of multicollinearity among the patient-level predictors, we computed variance inflation factors for each predictor in the model.28 No variables had a variance inflation factor >10, a conventional threshold for collinearity concerns, suggesting that collinearity was not problematic. Missing data for weight difference were imputed using median values. All analyses were performed using SAS statistical software, version 9.4 (SAS Institute, Inc, Cary, NC).
We performed 3 sensitivity analyses to check the validity of our conclusions about factors associated with PDE5i prescribing for group 2/3 PH in VA. First, to ensure our population reflected active VA pharmacy users, we repeated our analyses excluding veterans who received no VA prescriptions of any kind during the study period. Second, to increase our confidence that PDE5i prescriptions were intended for PH and not erectile dysfunction, we repeated our primary analyses excluding veterans with an International Classification of Diseases code for sexual impotence and those with a prescription for PDE5i with a pill per month ratio <15 (indicating less than daily use). Finally, we determined the level of residual confounding necessary to nullify the effect of our primary exposure (prior daily PDE5i in Medicare).29
We identified 34 775 Medicare-eligible veterans with group 2/3 PH in our cohort. The mean age was 74.0 years; 97.4% were men, and 79.4% were non-Hispanic white (Table). Veterans in our cohort had high rates of comorbid conditions, most commonly substance use disorders and underlying pulmonary, cardiac, or chronic kidney disease. Most veterans were seen at urban facilities with high complexity ratings and both pulmonary and cardiology specialty care available.
|Treatment Status in VA|
|Treated (n=1556)||Untreated (n=33 219)|
|Age, y; mean (SD)||70.7 (9.2)||74.2 (9.6)|
|Female sex||23 (1.5%)||890 (2.7%)|
|White||1163 (74.7%)||26 455 (79.6%)|
|Black||266 (17.1%)||4483 (13.5%)|
|Hispanic||48 (3.1%)||987 (2.9%)|
|Other||79 (5.1%)||1334 (4.0%)|
|Married||923 (59.3%)||17 492 (52.7%)|
|VA benefits (priority status)|
|Highly disabled||478 (30.7%)||9288 (28.0%)|
|Low/moderately disabled||238 (15.3%)||4824 (14.5%)|
|Limited with copayments||167 (10.7%)||2864 (8.6%)|
|Poverty/no copayments||673 (43.3%)||16 243 (48.9%)|
|PH diagnosis and management|
|Other pulmonary vasodilator||35 (2.2%)||100 (0.3%)|
|Calcium channel blockers||858 (55.1%)||19 861 (59.8%)|
|Right heart catheterization||818 (52.6%)||8104 (24.4%)|
|Year of first PDE5i prescription|
|2006||69 (4.4%)||65 (0.2%)|
|2007||98 (6.3%)||1661 (5.0%)|
|2008||129 (8.3%)||3152 (9.5%)|
|2009||162 (10.4%)||3784 (11.4%)|
|2010||173 (11.1%)||4105 (12.4%)|
|2011||141 (9.1%)||4639 (14.0%)|
|2012||189 (12.1%)||4464 (13.4%)|
|2013||187 (12.0%)||3972 (12.0%)|
|2014||191 (12.3%)||3733 (11.2%)|
|2015||175 (11.2%)||3646 (11.0%)|
|General health measures|
|Underlying PH condition associated with group 2 PH|
|Ischemic heart disease||395 (25.4%)||10 641 (32.0%)|
|Heart failure||908 (58.4%)||21 139 (63.6%)|
|Arrhythmia||704 (45.2%)||17 949 (54%)|
|Underlying PH condition associated with group 3 PH|
|Chronic lung disease*||993 (63.8%)||20 239 (60.9%)|
|Interstitial lung disease||339 (21.8%)||4682 (14.1%)|
|Obstructive sleep apnea||355 (22.8%)||6274 (18.9%)|
|Other comorbid conditions|
|Stroke or transient ischemic attack||206 (13.2%)||6365 (19.2%)|
|Chronic kidney disease||876 (56.3%)||21 344 (64.3%)|
|Dialysis dependency||109 (7.0%)||2496 (7.5%)|
|Chronic liver disease||96 (6.2%)||1730 (5.2%)|
|Hepatitis C||72 (4.6%)||2603 (7.8%)|
|Malignancy||248 (15.9%)||6727 (20.3%)|
|Neurological disorder†||27 (1.7%)||2002 (6.0%)|
|Mental health disorder‡||37 (2.4%)||1357 (4.1%)|
|Substance abuse||579 (37.2%)||16 104 (48.5%)|
|Oxygen dependency||597 (38.4%)||9888 (29.8%)|
|Ventilator dependency||42 (2.7%)||837 (2.5%)|
|Medications for comorbidities|
|Opioids||864 (55.5%)||17 084 (51.4%)|
|Low-dose PDE5i||509 (32.7%)||6890 (20.7%)|
|Recent organ failures§|
|Renal failure||50 (3.2%)||785 (2.4%)|
|Liver failure||1 (0.1%)||22 (0.1%)|
|Neurological failure||18 (1.2%)||344 (1.0%)|
|Hematologic failure||21 (1.3%)||198 (0.6%)|
|Metabolic failure||4 (0.3%)||37 (0.1%)|
|Circulatory failure||36 (2.3%)||575 (1.7%)|
|Respiratory failure||230 (14.8%)||1316 (4.0%)|
|Right heart failure||310 (19.9%)||2579 (7.8%)|
|Vital sign changes|
|Change in weight in pounds, mean (SD)‖||−2.7 (11.5)||−1.1 (11.4)|
|Dual VA and community care|
|Daily PDE5i in Medicare||102 (6.6%)||232 (0.7%)|
|Contraindications to PDE5i|
|Nitrates||97 (6.2%)||2317 (7.0%)|
|Outpatient visit days, mean (SD)||54.1 (46.5)||52.0 (49.4)|
|Long-term care||13 (0.8%)||315 (0.9%)|
|Emergency room visit days, mean (SD)||1.4 (1.2)||1.4 (1.2)|
|Hospitalizations, mean (SD)||2.0 (2.8)||2.1 (2.9)|
|Intensive care unit stays, mean (SD)||0.5 (0.9)||0.4 (0.8)|
|North Atlantic||263 (16.9%)||8617 (25.9%)|
|Southeast||370 (23.8%)||5889 (17.7%)|
|Midwest||323 (20.8%)||8084 (24.3%)|
|Continental||283 (18.2%)||4755 (14.3%)|
|Pacific||317 (20.4%)||5874 (17.7%)|
|Rural||83 (5.3%)||2244 (6.8%)|
|Urban||1428 (91.8%)||30 443 (91.6%)|
|Unknown||45 (2.9%)||532 (1.6%)|
|Specialty care at the facility|
|Pulmonary||1518 (97.6%)||32 464 (97.7%)|
|Cardiology||1538 (98.8%)||32 991 (99.3%)|
|1a||767 (49.3%)||14 470 (43.6%)|
|1b||276 (17.7%)||6593 (19.8%)|
|1c||191 (12.3%)||5112 (15.4%)|
|2||159 (10.2%)||4129 (12.4%)|
|3||118 (7.6%)||2383 (7.2%)|
|Volume of PH patients by quartile|
|Q1 (lowest quartile)||685 (44.0%)||17 594 (53.0%)|
|Q2||480 (30.8%)||8601 (25.9%)|
|Q3||259 (16.6%)||5021 (15.1%)|
|Q4 (highest quartile)||132 (8.5%)||2003 (6.0%)|
|Volume of right heart catheterization by quartile|
|Q1 (lowest quartile)||665 (42.7%)||14 421 (43.4%)|
|Q2||487 (31.3%)||9963 (30.0%)|
|Q3||122 (7.8%)||2379 (7.2%)|
|Q4 (highest quartile)||282 (18.1%)||6456 (19.4%)|
Relative to the multivariable logistic regression, the model with random effects was significantly more predictive (P<0.001) based on a likelihood ratio test for the random effects variance greater than zero.30,31 We, therefore, report our findings based on the random effects logistic regressions. Odds ratios (ORs) from these models adjust for the estimated facility-level random effects.
Factors Associated With PDE5i Prescribing for Group 2/3 PH
One thousand five hundred fifty-six (4.5%) veterans received VA prescriptions for PDE5i treatment for group 2/3 PH. Of 334 veterans who had received PDE5i treatment in the community, 31% had them continued in the VA. Supporting our primary hypothesis, by far the strongest predictor of PDE5i prescribing for group 2/3 PH in VA was prior PDE5i treatment in the community (OR, 6.5 [95% CI, 4.9–8.7]; Figure 3).
PH severity and active management of PH were also associated with VA PDE5i prescribing for group 2/3 PH (Figure 3). Veterans with more severe PH disease as indicated by recent episodes of right heart failure (OR, 3.3 [95% CI, 2.8–3.9]) or acute respiratory failure (OR, 3.7 [95% CI, 3.0–4.4]) were more likely to be treated. Veterans who had previously received PH-specific interventions including right heart catheterization (OR, 3.8 [95% CI, 3.4–4.3]) or treatment with other pulmonary vasodilators (OR, 3.3 [95% CI, 2.1–5.2]) were also more likely to receive PDE5i for group 2/3 PH. Among our cohort, veterans with conditions associated with group 2 PH (underlying cardiac disease) were less likely to receive PDE5i treatment, whereas veterans with conditions associated with group 3 PH (underlying lung disease, particularly interstitial lung disease) were more likely to receive treatment. Veterans with age ≥75 years and those with indications of poor general health were less likely to receive PDE5i treatment.
We found substantial variation across facilities in rates of PDE5i prescribing for group 2/3 PH in VA, ranging from 0% to 35% (Figure 4). Relative to facilities in the Northeast, facilities in the Southeast (OR, 2.0 [95% CI, 1.4–2.9]), Continental (OR, 1.5 [95% CI, 1.1–2.2]), and Pacific regions (OR, 1.6 [95% CI, 1.1–2.3]) were all associated with increased odds of VA PDE5i treatment (Figure 3). Facilities with lower PH patient volume had lower odds of VA prescriptions for PDE5i treatment of group 2/3 PH. The intraclass correlation for the random effects model is 0.076, suggesting that very little of the total variation is attributable to facility effects.32,33
There was little change in the direction of our main findings in our sensitivity analyses (Tables II and III in the Data Supplement). Of note, in our model excluding veterans potentially treated with PDE5i for erectile dysfunction, the effect of prior PDE5i treatment through Medicare was even stronger (OR, 12.1 [95% CI, 8.5–17.1]). For our primary exposure of prior daily PDE5i treatment for PH in Medicare, to nullify our observed OR of 6.5, an unmeasured confounder would need be associated with our exposure variable with a relative risk of 20 and with our primary outcome with a relative risk of at least 10.
Two recent international conferences of PH experts issued consensus statements raising concern about increasing indiscriminate use of pulmonary vasodilators for groups 2 and 3 PH.34–36 To develop interventions to mitigate this problem, it is critical to understand factors driving potentially inappropriate treatment. While we found overall rates of PDE5i prescribing for group 2/3 PH were relatively low in VA (≈5%), some VA facilities were dispensing PDE5i prescriptions to ≈35% of their patients with group 2/3 PH, exposing these patients to potential harm and wasting resources. At a cost of $10 000 to $13 000 to treat 1 patient for 1 year,37 the burden to the healthcare system of unnecessary PDE5i treatment for group 2/3 PH could be substantial. Furthermore, rates of potentially inappropriate treatment may actually be higher for patients treated in non-VA settings as VA has already taken steps to restrict PDE5i prescribing, such as requiring pharmacy review and approval to fill daily PDE5i prescriptions. Importantly, the factors we identified as associated with PDE5i treatment for group 2/3 PH in VA speak to larger issues that may drive inappropriate prescribing in other contexts, such as prescribing inertia, lack of provider awareness, workplace culture, subspecialty influences, and care fragmentation (Figure 5).
The strongest predictor of treatment with PDE5i in VA was prior treatment initiated in the community, representing a missed opportunity to reevaluate appropriateness of treatment and discontinue guideline-discordant prescriptions. The tendency to continue prescriptions initiated by another provider, even when potentially inappropriate, likely reflects multiple factors at the patient, provider, and system levels. Providers who inherit the prescribing of a medication may not always apply the same level of scrutiny as when initiating the medication themselves,38,39 tending instead to respect their colleagues’ prescribing autonomy even when there is little evidence to support using the medication.40 Providers may be reluctant to discontinue treatment for a complex condition like PH, particularly if they have limited experience managing the condition, concerned that doing so may worsen symptoms or worsen clinical status.41 Similarly, patients may be resistant to change, pressuring a new provider into continuing a medication that was started by another clinician the patient trusts.42 Finally, inadequate care coordination across VA and non-VA systems may contribute to reluctance to discontinue medications.43 For example, VA providers with incomplete documentation of care received in the community may give their colleagues the benefit of the doubt in continuing a potentially inappropriate medication. With recent policy changes such as the Veterans Choice and Maintaining Internal Systems and Strengthening Outside Networks (MISSION) acts, more veterans can be expected to share care in the future,18,19 raising the importance of educating VA providers on the need to carefully consider the appropriateness of prescriptions initiated in the community rather than defaulting to continuing them without much scrutiny.
In addition to our primary exposure of prior treatment in the community, we identified several other clinically relevant patient factors associated with higher odds of PDE5i treatment for group 2/3 PH. Veterans with more severe PH, as indicated by recent right heart failure or acute respiratory failure, were more likely to be treated. Patients with groups 2 and 3 PH experience high mortality with significant symptom burden and poor prognosis.44 Though a provider may understand guideline recommendations at the population level, those beliefs may not always translate to how they manage individuals, particularly when faced with a patient with severe or end-stage disease.45 In particular, when providers perceive they have nothing else to offer, they are more likely to turn to low-value practices as a last resort to try to alleviate symptoms of a very sick patient.46 Indeed, in our study, we found that among veterans with group 2/3 PH, patients with interstitial lung disease were more likely to be treated with daily PDE5i while patients with underlying cardiac disease (group 2 PH) were less likely to receive daily PDE5i treatment. This may reflect the limited options for treatment of interstitial lung disease as compared with a wider range of effective options for treatment of cardiac disease. This willingness to use pulmonary vasodilators for patients with interstitial lung disease despite established benefit and potential for harm was previously identified in a survey of pulmonary physicians.47
Prior management for PH, including right heart catheterization or treatment with other pulmonary vasodilators, also increased the odds of treatment with PDE5i. It is possible that these patients represent the subset of group 2/3 PH for whom guidelines allow trials of pulmonary vasodilators on a case-by-case basis—those whose symptoms and hemodynamics are out of proportion to their underlying heart or lung disease.13,14 Alternatively, once providers have committed to seeking the diagnosis of PH with a right heart catheterization, they may be more prone to treat PH if the diagnosis is confirmed, regardless of whether the hemodynamics support the use of pulmonary vasodilators.
We also identified several facility factors associated with PDE5i treatment. Veterans receiving care in facilities with higher volume of PH patients were more likely to be treated, likely reflecting providers’ familiarity with the disease and the medications, as well as increased access to PH specialists. While some of these prescriptions may reflect appropriate trials of daily PDE5i for patients with group 2/3 PH out-of-proportion to the underlying disease, other cases may represent a spillover effect in which the threshold for treatment even for group 2/3 PH is lower in PH specialty clinics where providers routinely use pulmonary vasodilators for pulmonary arterial hypertension and are more comfortable using these medications. Indeed, prior survey data show that a majority of PH expert referral centers use pulmonary vasodilators for groups 2 and 3 PH, despite limited evidence to support this practice.48
Our findings may inform the design of multilevel interventions to deimplement guideline-discordant prescribing of PDE5i for group 2/3 PH, preventing patient harm and waste of resources. At the system level, open sharing of patient records across systems, clearly defined provider roles, and bidirectional communication between community and VA providers are all necessary to improve comanagement and appropriate prescribing.49,50 Additional resources such as dedicated care trackers, patient navigators, and record administrators may be necessary to help bridge divides across systems.50 At the provider level, effective strategies to limit inappropriate prescribing in other contexts include focused educational sessions, audit and feedback, and clinical decision support integrated in the electronic medical record.51,52 Pharmacist-led provider education and integration of pharmacists into clinical teams can also mitigate inappropriate prescribing, particularly when pharmacists are empowered to challenge potentially inappropriate prescriptions.53–55 Finally, as suggested previously, establishing appropriate use criteria may also help to reform this current practice.56
Our study has limitations. First, our retrospective analysis of administrative claims data cannot capture the nuanced thought processes of prescribers and there are likely additional influences on prescribing that we could not capture with this study design. Second, administrative data cannot assess physiological data, without which we cannot definitively determine the appropriateness of prescriptions. Additionally, while more recent consensus statements definitively recommend against treatment of group 2/3 PH, earlier guidelines released during the study period were less clear and allowed for treatment in group 2/3 PH with a precapillary component. Because our validated algorithm15 was conservatively designed to increase specificity for identifying patients with group 2/3 PH, our results likely represent an underestimate of the rate of potentially inappropriate PDE5i prescribing. Third, while we sought to comprehensively capture both VA and community care by using VA and Medicare data and restricting our cohort to Medicare-eligible veterans, we may have missed information about prescriptions or care received and reimbursed through other insurers. Finally, because our analysis is focused on Medicare-eligible veterans, our results may not be generalizable to factors affecting treatment among non-veterans or younger patients.
Given the growing rates of use of pulmonary vasodilators for groups 2 and 3 PH and the recent calls by PH expert conferences to reduce this practice, the time is now to intervene to deimplement this low-value, high-cost, potentially harmful practice. Our study identifies previously unknown influences on potentially inappropriate PDE5i prescribing for PH and suggests concrete solutions to address them.
Dr Gillmeyer, Dr Rinne, Dr Glickman, Dr Klings, Dr Maron, J.T. Hanlon, D.R. Miller, and Dr Wiener contributed to the study concept and design. Dr Lee, Dr Shao, S.X. Qian, and D.R. Miller contributed to the acquisition of data. Analysis and interpretation of data was done by all authors. Drs Gillmeyer and Wiener contributed to the drafting of the manuscript. All authors contributed to the critical revision of the manuscript for important intellectual content.
Sources of Funding
This work was supported by Veterans Health Administration (VA) Health Services Research and Development IIR 15-115 and by resources from the Edith Nourse Rogers Memorial Veterans Hospital and the VA Boston Healthcare System. In addition, Dr Gillmeyer receives support through a National Institutes of Health (NIH) Institutional Training Grant, T32 HL007035; Dr Maron receives support though NIH grants R56HL131787, 1R01HL139613-01, R21HL145420, 1U01HL125215-01, the Cardiovascular Medical Research and Education Foundation, and National Scleroderma Foundation; Dr Rinne was funded through a Parker B. Francis Fellowship and a Veterans Integrated Service Networks 1 Career Development Award. The views expressed in this article do not necessarily represent the views of the Department of Veterans Affairs or the US Government.
Berwick DM, Hackbarth AD. Eliminating waste in US health care.JAMA. 2012; 307:1513–1516. doi: 10.1001/jama.2012.362CrossrefMedlineGoogle Scholar
Cahir C, Bennett K, Teljeur C, Fahey T. Potentially inappropriate prescribing and adverse health outcomes in community dwelling older patients.Br J Clin Pharmacol. 2014; 77:201–210. doi: 10.1111/bcp.12161CrossrefMedlineGoogle Scholar
Muhlack DC, Hoppe LK, Weberpals J, Brenner H, Schöttker B. The association of potentially inappropriate medication at older age with cardiovascular events and overall mortality: a systematic review and meta-analysis of cohort studies.J Am Med Dir Assoc. 2017; 18:211–220. doi: 10.1016/j.jamda.2016.11.025CrossrefMedlineGoogle Scholar
Cassel CK, Guest JA. Choosing wisely: helping physicians and patients make smart decisions about their care.JAMA. 2012; 307:1801–1802. doi: 10.1001/jama.2012.476CrossrefMedlineGoogle Scholar
Wiener RS, Ouellette DR, Diamond E, Fan VS, Maurer JR, Mularski RA, Peters JI, Halpern SD. An official American Thoracic Society/American College of Chest Physicians policy statement: the choosing wisely top five list in adult pulmonary medicine.Chest. 2014; 145:1383–1391. doi: 10.1378/chest.14-0670CrossrefMedlineGoogle Scholar
Zheng YG, Ma H, Hu EC, Liu G, Chen G, Xiong CM. Oral targeted therapies in the treatment of pulmonary arterial hypertension: a meta-analysis of clinical trials.Pulm Pharmacol Ther. 2014; 29:241–249. doi: 10.1016/j.pupt.2014.08.005CrossrefMedlineGoogle Scholar
Cao JY, Wales KM, Cordina R, Lau EMT, Celermajer DS. Pulmonary vasodilator therapies are of no benefit in pulmonary hypertension due to left heart disease: a meta-analysis.Int J Cardiol. 2018; 273:213–220. doi: 10.1016/j.ijcard.2018.09.043CrossrefMedlineGoogle Scholar
Barnes H, Brown Z, Burns A, Williams T. Phosphodiesterase 5 inhibitors for pulmonary hypertension.Cochrane Database Syst Rev. 2019; 1:CD012621. doi: 10.1002/14651858.CD012621.pub2MedlineGoogle Scholar
Lederer DJ, Bartels MN, Schluger NW, Brogan F, Jellen P, Thomashow BM, Kawut SM. Sildenafil for chronic obstructive pulmonary disease: a randomized crossover trial.COPD. 2012; 9:268–275. doi: 10.3109/15412555.2011.651180CrossrefMedlineGoogle Scholar
Bermejo J, Yotti R, García-Orta R, Sánchez-Fernández PL, Castaño M, Segovia-Cubero J, Escribano-Subías P, San Román JA, Borrás X, Alonso-Gómez A, Botas J, Crespo-Leiro MG, Velasco S, Bayés-Genís A, López A, Muñoz-Aguilera R, de Teresa E, González-Juanatey JR, Evangelista A, Mombiela T, González-Mansilla A, Elízaga J, Martín-Moreiras J, González-Santos JM, Moreno-Escobar E, Fernández-Avilés F; Sildenafil for Improving Outcomes After Valvular Correction (SIOVAC) Investigators. Sildenafil for improving outcomes in patients with corrected valvular heart disease and persistent pulmonary hypertension: a multicenter, double-blind, randomized clinical trial.Eur Heart J. 2018; 39:1255–1264. doi: 10.1093/eurheartj/ehx700CrossrefMedlineGoogle Scholar
Redfield MM, Chen HH, Borlaug BA, Semigran MJ, Lee KL, Lewis G, LeWinter MM, Rouleau JL, Bull DA, Mann DL, Deswal A, Stevenson LW, Givertz MM, Ofili EO, O’Connor CM, Felker GM, Goldsmith SR, Bart BA, McNulty SE, Ibarra JC, Lin G, Oh JK, Patel MR, Kim RJ, Tracy RP, Velazquez EJ, Anstrom KJ, Hernandez AF, Mascette AM, Braunwald E; RELAX Trial. Effect of phosphodiesterase-5 inhibition on exercise capacity and clinical status in heart failure with preserved ejection fraction: a randomized clinical trial.JAMA. 2013; 309:1268–1277. doi: 10.1001/jama.2013.2024CrossrefMedlineGoogle Scholar
Blanco I, Gimeno E, Munoz PA, Pizarro S, Gistau C, Rodriguez-Roisin R, Roca J, Barberà JA. Hemodynamic and gas exchange effects of sildenafil in patients with chronic obstructive pulmonary disease and pulmonary hypertension.Am J Respir Crit Care Med. 2010; 181:270–278. doi: 10.1164/rccm.200907-0988OCCrossrefMedlineGoogle Scholar
Galiè N, Humbert M, Vachiery JL, Gibbs S, Lang I, Torbicki A, Simonneau G, Peacock A, Vonk Noordegraaf A, Beghetti M, Ghofrani A, Gomez Sanchez MA, Hansmann G, Klepetko W, Lancellotti P, Matucci M, McDonagh T, Pierard LA, Trindade PT, Zompatori M, Hoeper M. 2015 ESC/ERS guidelines for the diagnosis and treatment of pulmonary hypertension: The Joint Task Force for the Diagnosis and Treatment of Pulmonary Hypertension of the European Society of Cardiology (ESC) and the European Respiratory Society (ERS): Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC), International Society for Heart and Lung Transplantation (ISHLT).Eur Respir J. 2015; 46:903–975. doi: 10.1183/13993003.01032-2015CrossrefMedlineGoogle Scholar
McLaughlin VV, Archer SL, Badesch DB, Barst RJ, Farber HW, Lindner JR, Mathier MA, McGoon MD, Park MH, Rosenson RS, Rubin LJ, Tapson VF, Varga J, Harrington RA, Anderson JL, Bates ER, Bridges CR, Eisenberg MJ, Ferrari VA, Grines CL, Hlatky MA, Jacobs AK, Kaul S, Lichtenberg RC, Lindner JR, Moliterno DJ, Mukherjee D, Pohost GM, Rosenson RS, Schofield RS, Shubrooks SJ, Stein JH, Tracy CM, Weitz HH, Wesley DJ; ACCF/AHA. ACCF/AHA 2009 expert consensus document on pulmonary hypertension: a report of the American College of Cardiology Foundation Task Force on Expert Consensus Documents and the American Heart Association: developed in collaboration with the American College of Chest Physicians, American Thoracic Society Inc., and the Pulmonary Hypertension Association.Circulation. 2009; 119:2250–2294. doi: 10.1161/CIRCULATIONAHA.109.192230LinkGoogle Scholar
Kim D, Lee KM, Freiman MR, Powell WR, Klings ES, Rinne ST, Miller DR, Rose AJ, Wiener RS. Phosphodiesterase-5 inhibitor therapy for pulmonary hypertension in the united states. actual versus recommended use.Ann Am Thorac Soc. 2018; 15:693–701. doi: 10.1513/AnnalsATS.201710-762OCCrossrefMedlineGoogle Scholar
Wijeratne DT, Lajkosz K, Brogly SB, Lougheed MD, Jiang L, Housin A, Barber D, Johnson A, Doliszny KM, Archer SL. Increasing incidence and prevalence of world health organization groups 1 to 4 pulmonary hypertension: a population-based cohort study in Ontario, Canada.Circ Cardiovasc Qual Outcomes. 2018; 11:e003973. doi: 10.1161/CIRCOUTCOMES.117.003973LinkGoogle Scholar
- 17. Department of Veterans Affair.Tiered pharmacy copayments for medications. Final rule.Fed Regist. 2016; 81:89383–89391.MedlineGoogle Scholar
Hynes DM, Koelling K, Stroupe K, Arnold N, Mallin K, Sohn MW, Weaver FM, Manheim L, Kok L. Veterans’ access to and use of Medicare and Veterans Affairs health care.Med Care. 2007; 45:214–223. doi: 10.1097/01.mlr.0000244657.90074.b7CrossrefMedlineGoogle Scholar
Gellad WF. The veterans choice act and dual health system use.J Gen Intern Med. 2016; 31:153–154. doi: 10.1007/s11606-015-3492-2CrossrefMedlineGoogle Scholar
Gellad WF, Thorpe JM, Zhao X, Thorpe CT, Sileanu FE, Cashy JP, Hale JA, Mor MK, Radomski TR, Hausmann LRM, Donohue JM, Gordon AJ, Suda KJ, Stroupe KT, Hanlon JT, Cunningham FE, Good CB, Fine MJ. Impact of dual use of department of veterans affairs and medicare part d drug benefits on potentially unsafe opioid use.Am J Public Health. 2018; 108:248–255. doi: 10.2105/AJPH.2017.304174CrossrefMedlineGoogle Scholar
Thorpe JM, Thorpe CT, Gellad WF, Good CB, Hanlon JT, Mor MK, Pleis JR, Schleiden LJ, Van Houtven CH. Dual health care system use and high-risk prescribing in patients with dementia: a national cohort study.Ann Intern Med. 2017; 166:157–163. doi: 10.7326/M16-0551CrossrefMedlineGoogle Scholar
- 22. Veterans Health Administration. VHA Enrollment Determination. VHA Handbook 1601A.03.2015. https://www.va.gov/VHAPUBLICATIONS/publications.cfm?pub=2Google Scholar
Lund BC, Charlton ME, Steinman MA, Kaboli PJ. Regional differences in prescribing quality among elder veterans and the impact of rural residence.J Rural Health. 2013; 29:172–179. doi: 10.1111/j.1748-0361.2012.00428.xCrossrefMedlineGoogle Scholar
- 24. Office of Quality and Safety. 2010 VHA facility quality and safety report.2010. https://www.va.gov/health/docs/HospitalReportCard2010.pdf. Accessed July 1, 2019.Google Scholar
Friedman JH. Greedy function approximation: a gradient boosting machine.Ann Stat. 2001; 29:1189–1232.CrossrefGoogle Scholar
White LF, Wallinga J, Finelli L, Reed C, Riley S, Lipsitch M, Pagano M. Estimation of the reproductive number and the serial interval in early phase of the 2009 influenza A/H1N1 pandemic in the USA.Influenza Other Respir Viruses. 2009; 3:267–276. doi: 10.1111/j.1750-2659.2009.00106.xCrossrefMedlineGoogle Scholar
Heikkinen T, Young J, van Beek E, Franke H, Verstraeten T, Weil JG, Della Cioppa G. Safety of MF59-adjuvanted A/H1N1 influenza vaccine in pregnancy: a comparative cohort study.Am J Obstet Gynecol. 2012; 207:177.e1–177.e178. doi: 10.1016/j.ajog.2012.07.007CrossrefGoogle Scholar
Kutner M, Nachtsheim C, Neter J. Applied Linear Regression Models. . Burr Ridge, IL: McGraw Hill/Irwin; 2004.Google Scholar
Ding P, VanderWeele TJ. Sensitivity analysis without assumptions.Epidemiology. 2016; 27:368–377. doi: 10.1097/EDE.0000000000000457CrossrefMedlineGoogle Scholar
Stram DO, Lee JW. Variance components testing in the longitudinal mixed effects model.Biometrics. 1994; 50:1171–1177.CrossrefMedlineGoogle Scholar
Self SG, Liang K-Y. Asymptomatic properties of maximum likelihood estimators and likelihood ratio tests under nonstandard conditions.J Am Stat Assoc. 1987; 82:605–610.CrossrefGoogle Scholar
Wu S, Crespi CM, Wong WK. Comparison of methods for estimating the intraclass correlation coefficient for binary responses in cancer prevention cluster randomized trials.Contemp Clin Trials. 2012; 33:869–880. doi: 10.1016/j.cct.2012.05.004CrossrefMedlineGoogle Scholar
Moineddin R, Matheson FI, Glazier RH. A simulation study of sample size for multilevel logistic regression models.BMC Med Res Methodol. 2007; 7:34. doi: 10.1186/1471-2288-7-34CrossrefMedlineGoogle Scholar
Rosenkranz S, Ghofrani HA, Grünig E, Klose H, Olschewski H, Hoeper MM. Cologne consensus conference on pulmonary hypertension - update 2018.Int J Cardiol. 2018; 272S:1–3. doi: 10.1016/j.ijcard.2018.09.064CrossrefMedlineGoogle Scholar
Vachiéry JL, Tedford RJ, Rosenkranz S, Palazzini M, Lang I, Guazzi M, Coghlan G, Chazova I, De Marco T. Pulmonary hypertension due to left heart disease.Eur Respir J. 2019; 53:1801897.doi: 10.1183/13993003.01897-2018CrossrefMedlineGoogle Scholar
Nathan SD, Barbera JA, Gaine SP, Harari S, Martinez FJ, Olschewski H, Olsson KM, Peacock AJ, Pepke-Zaba J, Provencher S, Weissmann N, Seeger W. Pulmonary hypertension in chronic lung disease and hypoxia.Eur Respir J. 2019; 53:1801914. doi: 10.1183/13993003.01914-2018CrossrefMedlineGoogle Scholar
Coyle K, Coyle D, Blouin J, Lee K, Jabr MF, Tran K, Mielniczuk L, Swiston J, Innes M. Cost effectiveness of first-line oral therapies for pulmonary arterial hypertension: a modelling study.Pharmacoeconomics. 2016; 34:509–520. doi: 10.1007/s40273-015-0366-8CrossrefMedlineGoogle Scholar
Linsky A, Simon SR, Marcello TB, Bokhour B. Clinical provider perceptions of proactive medication discontinuation.Am J Manag Care. 2015; 21:277–283.MedlineGoogle Scholar
Britten N, Brant S, Cairns A, Hall WW, Jones I, Salisbury C, Virji A, Herxheimer A. Continued prescribing of inappropriate drugs in general practice.J Clin Pharm Ther. 1995; 20:199–205. doi: 10.1111/j.1365-2710.1995.tb00649.xCrossrefMedlineGoogle Scholar
Prosser H, Walley T. New drug prescribing by hospital doctors: the nature and meaning of knowledge.Soc Sci Med. 2006; 62:1565–1578. doi: 10.1016/j.socscimed.2005.08.035CrossrefMedlineGoogle Scholar
Cullinan S, O’Mahony D, Fleming A, Byrne S. A meta-synthesis of potentially inappropriate prescribing in older patients.Drugs Aging. 2014; 31:631–638. doi: 10.1007/s40266-014-0190-4CrossrefMedlineGoogle Scholar
Anderson K, Stowasser D, Freeman C, Scott I. Prescriber barriers and enablers to minimising potentially inappropriate medications in adults: a systematic review and thematic synthesis.BMJ Open. 2014; 4:e006544. doi: 10.1136/bmjopen-2014-006544CrossrefMedlineGoogle Scholar
Byrne MM, Kuebeler M, Pietz K, Petersen LA. Effect of using information from only one system for dually eligible health care users.Med Care. 2006; 44:768–773. doi: 10.1097/01.mlr.0000218786.44722.14CrossrefMedlineGoogle Scholar
Trammell AW, Shah AJ, Phillips LS, Michael Hart C. Mortality in US veterans with pulmonary hypertension: a retrospective analysis of survival by subtype and baseline factors.Pulm Circ. 2019; 9:2045894019825763. doi: 10.1177/2045894019825763CrossrefGoogle Scholar
Cabana MD, Rand CS, Powe NR, Wu AW, Wilson MH, Abboud PA, Rubin HR. Why don’t physicians follow clinical practice guidelines? A framework for improvement.JAMA. 1999; 282:1458–1465. doi: 10.1001/jama.282.15.1458CrossrefMedlineGoogle Scholar
Hahn EE, Munoz-Plaza C, Wang J, Garcia Delgadillo J, Schottinger JE, Mittman BS, Gould MK. Anxiety, culture, and expectations: oncologist-perceived factors associated with use of nonrecommended serum tumor marker tests for surveillance of early-stage breast cancer.J Oncol Pract. 2017; 13:e77–e90. doi: 10.1200/JOP.2016.014076CrossrefMedlineGoogle Scholar
Minai OA, Nathan SD, Hill NS, Badesch DB, Stoller JK. Pulmonary hypertension in lung diseases: survey of beliefs and practice patterns.Respir Med. 2010; 104:741–748. doi: 10.1016/j.rmed.2009.12.015CrossrefMedlineGoogle Scholar
Trammell AW, Pugh ME, Newman JH, Hemnes AR, Robbins IM. Use of pulmonary arterial hypertension-approved therapy in the treatment of non-group 1 pulmonary hypertension at US referral centers.Pulm Circ. 2015; 5:356–363. doi: 10.1086/681264CrossrefMedlineGoogle Scholar
Banfield M, Gardner K, McRae I, Gillespie J, Wells R, Yen L. Unlocking information for coordination of care in Australia: a qualitative study of information continuity in four primary health care models.BMC Fam Pract. 2013; 14:34. doi: 10.1186/1471-2296-14-34CrossrefMedlineGoogle Scholar
Zuchowski JL, Chrystal JG, Hamilton AB, Patton EW, Zephyrin LC, Yano EM, Cordasco KM. Coordinating care across health care systems for veterans with gynecologic malignancies: a qualitative analysis.Med Care. 2017; 55(suppl 7 suppl 1):S53–S60. doi: 10.1097/MLR.0000000000000737CrossrefMedlineGoogle Scholar
Gerber JS, Prasad PA, Fiks AG, Localio AR, Grundmeier RW, Bell LM, Wasserman RC, Keren R, Zaoutis TE. Effect of an outpatient antimicrobial stewardship intervention on broad-spectrum antibiotic prescribing by primary care pediatricians: a randomized trial.JAMA. 2013; 309:2345–2352. doi: 10.1001/jama.2013.6287CrossrefMedlineGoogle Scholar
Meeker D, Linder JA, Fox CR, Friedberg MW, Persell SD, Goldstein NJ, Knight TK, Hay JW, Doctor JN. Effect of behavioral interventions on inappropriate antibiotic prescribing among primary care practices: a randomized clinical trial.JAMA. 2016; 315:562–570. doi: 10.1001/jama.2016.0275CrossrefMedlineGoogle Scholar
Agee C, Coulter L, Hudson J. Effects of pharmacy resident led education on resident physician prescribing habits associated with stress ulcer prophylaxis in non-intensive care unit patients.Am J Health Syst Pharm. 2015; 72(11 suppl 1):S48–S52. doi: 10.2146/sp150013CrossrefMedlineGoogle Scholar
Hughes GJ, Belgeri MT, Perry HM. The impact of pharmacist interventions on the inappropriate use of acid-suppression therapy.Consult Pharm. 2011; 26:485–490. doi: 10.4140/TCP.n.2011.485CrossrefMedlineGoogle Scholar
Smith ZR, Rangarajan K, Barrow J, Carter D, Coons JC, Dzierba AL, Falvey J, Fester KA, Guido MR, Hao D, Ou NN, Pogue KT, MacDonald NC. Development of best practice recommendations for the safe use of pulmonary hypertension pharmacotherapies using a modified Delphi method.Am J Health Syst Pharm. 2019; 76:153–165. doi: 10.1093/ajhp/zxy020CrossrefMedlineGoogle Scholar
Maron BA, Ryan JJ. A concerning trend for patients with pulmonary hypertension in the era of evidence-based medicine.Circulation. 2019; 139:1861–1864. doi: 10.1161/CIRCULATIONAHA.118.037613LinkGoogle Scholar