Combination Antihypertensive Therapy Prescribing and Blood Pressure Control in a Real-World Setting

Abstract

BACKGROUND

Specific combinations of two drug classes are recommended in a variety of clinical situations in the management of hypertension. These preferred combinations are based on complimentary blood pressure (BP) lowering mechanisms or benefit for a concomitant disease.

METHODS

Using electronic health records (EHRs) data from 27,579 ambulatory hypertensive patients, we investigated antihypertensive therapy prescribing patterns and associations of preferred two drug classes with BP control.

RESULTS

Overall, BP control, defined as BP <140/90 mm Hg, was 65% among treated patients. Preferred dual antihypertensive therapy was prescribed in 55% of patients with uncomplicated hypertension, 49% of patients with diabetes, and 47% of patients with a history of myocardial infarction (MI); these prescribing frequencies of preferred combinations were not explained by worse BP control on those combinations. In fact, we found suggestive evidence of association between prescribing of preferred two drug classes and improved BP control among post-MI (OR: 1.21, 95% CI: 0.99–1.48, P = 0.061) and uncomplicated hypertensive (OR: 1.11, 95% CI: 0.98–1.26, P = 0.089) patients.

CONCLUSIONS

Prescribing of guideline-recommended antihypertensive drug classes for concomitant diseases is suboptimal and prescribing of preferred/optimized drug class combinations was moderate. We did not find a clear association between the use of optimized drug class combinations and greater BP control. Overall, using EHR data, we identified potential opportunities for re-examining prescribing practices with implications for clinical decision support and healthcare improvement at the community and health system-wide levels.

Hypertension is a major risk factor for cardiovascular diseases.¹ In the United States, hypertension identification, treatment, and control have been increasing; yet, 32% of adults have hypertension, defined as blood pressure (BP) ≥140/90 mm Hg, 24% are taking antihypertensive medication(s); of those, 39% have uncontrolled BP (≥140/90 mm Hg) despite treatment.² Moreover, most hypertensive individuals require ≥2 antihypertensive drug classes to achieve BP control³ especially with the new hypertension definition (BP ≥130/80 mm Hg).² Compared to using higher doses of one drug class, combining multiple mechanistically synergistic drug classes is associated with greater and quicker BP-lowering and fewer side effects when given as initial therapy.^4–7

Based on evidence from numerous studies, the latest hypertension treatment guidelines suggest avoiding drug combinations that target similar BP control systems; instead, they recommend combining drugs with complementary mechanisms of action to achieve greater BP-lowering efficacy.^7–12 Both the Eight Joint National Commission (JNC8 2014) and American Heart Association/American College of Cardiology (AHA/ACC 2017) guidelines recommend avoiding the combination of the two renin–angiotensin system blockers, angiotensin-converting enzyme inhibitor (ACEI), and angiotensin-receptor blocker (ARB), based on studies such as ONTARGET that showed worse cardiovascular and renal outcomes with this combination.^8–10 Using evidence from prior studies including a large meta-analysis of patients treated with multiple drug classes, the AHA/ACC 2017 guideline specifically recommends combining drugs with complementary activity to achieve greater BP-lowering effect.^7–16 Only one major clinical trial, ACCOMPLISH, directly compared two-drug class combinations against each other, namely ACEI + calcium-channel blocker (CCB) and ACEI + diuretic, and found improved outcomes with ACEI + CCB and comparable BP-lowering (13–14 mm Hg systolic BP decrease) between the two combinations.¹²

In this study, using electronic health records (EHRs) data from ambulatory patients diagnosed with hypertension, we examined antihypertensive drug prescribing patterns and BP control on combinations of antihypertensive drug classes that are preferred for patients according to comorbidity (e.g. postmyocardial infarction (MI) or diabetic patients). Previous studies have examined comorbidity-specific antihypertensive drug prescribing patterns,^17–21 but few have focused on the prescribing of preferred drug class combinations and tested their associations with BP control in a real-world setting.²² Consistent with the latest hypertension treatment guidelines and for this study, we considered ACEI or ARB (hereafter “ACEI/ARB”) plus a CCB or diuretic (hereafter “CCB/diuretic”) as a preferred drug class combination because of its mechanistic synergism and greater BP-lowering efficacy.^7–12 We made an exception for patients with prior MI where we considered ACEI/ARB + beta-blocker (BB) (hereafter “ACEI/ARB + BB) as a preferred combination. Although both drug classes affect the renin pathway, hypertension treatment guidelines recommend their combination for all post-MI patients for cardioprotection.^8,9,23,24 Lastly, in addition to understanding prescribing patterns and BP control on preferred drug class combinations, we identified the sociodemographic/clinical risk factors associated with uncontrolled BP on multiple drug classes so that clinicians may better identify and treat patients at risk of uncontrolled BP.

We used EHR data for our study. EHR data can be a useful resource for understanding real-world prescribing practices and related BP control with different antihypertensive regimens. In recent years, EHRs have gained increasing recognition as a rich resource for studying healthcare issues in a real-life clinical setting.^25,26 Unlike randomized controlled trial and cohort data, EHR data are collected as part of routine clinical care (thereby less cost/time-intensive), have more data elements and typically larger sample sizes.²⁶ EHR-based studies may also help validate randomized controlled trial findings in the real world.²⁶ One important limitation is EHR data quality variability; accordingly, many healthcare institutions are developing data repositories to standardize EHR data for large-scale biomedical research.²⁷

METHODS

Data source

We used EHR data from patients treated at University of Florida (UF) Health, a large health system serving North-Central Florida. UF Institutional Review Board approved the study and granted waiver of consent for data acquisition and analysis. De-identified EHR data came from UF Health Integrated Data Repository (IDR). Informatics for Integrating Biology and the Bedside (i2b2) (https://idr.ufhealth.org/i2b2) software was used to query and select de-identified patient records.²⁸

Patient inclusions, exclusions, and analysis datasets

Figure 1 shows processes of patient inclusions, exclusions, and analysis dataset creation. First, we selected patients with an outpatient clinic visit (generalist and specialist) from 1 August 2013–1 August 2016 (study period), age ≥18 years at visit, and history of primary hypertension. Next, we identified “treated” patients based on antihypertensive drug prescription records (both UFHealth-prescribed and self-reported) during the study period. We included patients treated for ≥1 month (≥28 days). We excluded untreated patients, patients treated <1 month, patients diagnosed with end-stage renal disease, stage-5 chronic kidney disease (CKD), dialysis, heart failure, and renal insufficiency/failure, ACEI/ARB allergy, angioedema, or hyperkalemia, and patients without BP measurements after ≥1 month on antihypertensive therapy. In Analysis 1, we studied patients on two antihypertensive drug classes only. In Analysis 2, among patients on ≥2 antihypertensive drug classes, we examined the association between BP control and treatment with the preferred two drug classes. In supplementary analysis, we also examined the proportion of patients with controlled BP among those on two drug classes and among those on ≥3-drug classes. In Analysis 3, which explored sociodemographic/clinical correlates of uncontrolled BP on combination therapy, we included patients that met cases–controls definitions (Figure 1). Additional details on data-processing are under Supplementary Information online.

Figure 1.

Flow diagram showing screening, inclusion, and analysis of study population. Abbreviations: BP, blood pressure; HTN, hypertension; HF, heart failure; MI, myocardial infarction; RF, renal failure; CKD, chronic kidney disease.

As shown in Figure 1, we first examined (Analysis 1) prescribing patterns of preferred two drug classes among patients on two drug classes only (n = 11,018). Second (Analysis 2), we examined associations between preferred vs. other or non-preferred two drug classes and controlled (systolic BP <140 mm Hg and diastolic BP <90 mm Hg) vs. uncontrolled BP (≥140/90 mm Hg) (n = 27,579). In supplementary analyses, we examined BP control among patients on two drug classes and ≥3-drug classes; among those on ≥3-drug classes, we also compared BP control by diuretic therapy. BP control was determined using office measurements taken closest to the time at which the prescribing of ≥2-drug classes overlapped for ≥28 days. We used ≥140/90 mm Hg as the cutoff for uncontrolled BP because this was the definition of hypertension at the time from which the data were extracted. All aforementioned analyses were stratified by history of MI, diabetes, and uncomplicated hypertension (no ischemic heart disease, diabetes, CKD, and stroke/transient ischemic attack).

Lastly, in Analysis 3, we examined the sociodemographic/clinical risk factors associated with uncontrolled BP despite prescription of multiple drug classes (n = 25,507); here, we selected the treatment period in which patients were on the maximum number of antihypertensive drug classes (Figure 1).

Definition of preferred combination therapy

In this study, we considered ACEI/ARB + CCB/diuretic as preferred two drug classes for patients with diabetes and uncomplicated hypertension, as previously described (Tables 1 and 2). Although the latest hypertension treatment guidelines emphasize ACEI/ARB for diabetics with proteinuria,⁸ we conducted the primary analyses among all diabetics regardless of proteinuria because we considered that our definition of preferred two drug classes will not differ by proteinuria status. Among post-MI patients, we considered ACEI/ARB + BB as the preferred two drug classes regardless of time since MI, consistent with the most recent MI²³ and US hypertension treatment guidelines.⁸ However, given debate surrounding length of BB treatment post-MI,^29,30 we also conducted analysis limited to recent (≤1-year) post-MI patients.

Table 1.

Definition of preferred vs. non-preferred two drug class antihypertensive therapy by comorbidity groups

Combination therapy	Post-MI	Diabetes	CKD	Uncomplicateda
Preferred	ACEI/ARB + BB	ACEI/ARB + CCB/diuretic	ACEI/ARB + CCB/diuretic	ACEI/ARB + CCB/diuretic or BB + CCB/diuretic
Non-preferred	All other combinations	All other combinations	All other combinations	All other combinations

Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, beta-blocker; CCB, calcium-channel blocker; IHD, ischemic heart disease; MI, myocardial infarction; CKD, chronic kidney disease; TD, thiazide or thiazide-like diuretic.

^aUncomplicated hypertensive defined as lack of diabetes, IHD, CKD, and stroke/transient ischemic attack; ACEI/ARB refers to either ACEI or ARB; CCB/TD refers to either CCB or TD.

Table 2.

Clinical guideline recommendations on combination antihypertensive therapy by major comorbidity

Guideline	Population	Recommended combination antihypertensive therapy
2014 JNC8 Guideline9	General	• Combine different drug classes from ACEI, ARB, CCB, and TD • Avoid combination of ACEI + ARB
2017 ACC/AHA Guideline1	General	• Combine drug classes with complementary mechanisms of action (e.g. ACEI/ARB + diuretic (TD), ACEI/ARB + CCB) • Do not combine drugs of the same class (exceptions: diuretic combinations, dihydropyridine CCB with non-dihydropyridine CCB) • Combinations including ACEI/ARB are preferred in diabetic and/or CKD patients with proteinuriaa • Avoid combination of ACEI + ARBb
	Diabetes
	CKD
	Post-MI	• ACEI/ARB + BB

Abbreviations: BP, blood pressure; MI, myocardial infarction; CKD, chronic kidney disease; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, beta-blocker; CCB, calcium-channel blocker; TD, thiazide or thiazide-like diuretic; JNC, Joint National Committee; ACC/AHA, American College of Cardiology/American Heart Association.

^aOther acceptable but less preferred combinations for CKD patients include BB + diuretic, BB + ACEI/ARB, alpha-blocker/centrally acting alpha agonist + diuretic.

^bAlso, avoid ACEI or ARB combined with direct renin inhibitor or aldosterone antagonist.

Statistical analysis

Analyses were performed using SAS 9.4 (Cary, NC). In Analysis 1, we examined sociodemographic/clinical characteristics, prescribing patterns of two-drug class regimens, and BP control on preferred vs. non-preferred two drug classes using chi-square or Student’s t-test.

In Analysis 2, we tested associations between preferred vs. non-preferred two drug classes (explanatory variable) and controlled vs. uncontrolled BP (dependent variable) using multivariable logistic regression analyses stratified by comorbidity. Covariates were selected using univariate analysis (P < 0.20), followed by stepwise logistic regression modeling, with P < 0.20 as the criterion for a variable entering the model and P < 0.05 as the criterion for a variable staying in the model. The list of tested covariates is provided in Table 3. All regression models adjusted for age, sex, ethnicity (African American vs. other), baseline systolic BP, number of concomitant antihypertensive drugs prescribed, and treatment duration (time between prescribing of two drug classes and closest subsequent BP measurement date). Among post-MI patients, the model was further adjusted for history of diabetes and CKD. Among diabetics, the model was also adjusted for history of ischemic heart disease and CKD and body mass index (BMI), and among uncomplicated patients, the model was further adjusted for BMI and uninsured status. Covariates were selected through prior knowledge of association with uncontrolled BP^31–34 as well as the stepwise regression covariate selection process. In supplementary analysis, we examined the percent of patients with controlled BP among those on two drug classes and ≥3-drug classes; chi-square test was used to compare BP control by diuretic therapy among those on ≥3-drug classes.

Table 3.

Characteristics of patients on two drug class antihypertensive therapy by comorbidity

Patient characteristics	Comorbidity
	Post-MI	Diabetes	Uncomplicateda
	n = 1,582	n = 4,746	n = 4,690
Age	65.3 ± 12.1	62.1 ± 12.8	57.9 ± 14.1
Male	930 (59)	2,104 (44)	1,803 (38)
BMI, kg/m2	29.9 ± 6.6	32.8 ± 7.6	31.5 ± 7.4
Ethnicity
White	1,253 (79)	3,156 (66)	3,472 (74)
African American	254 (16)	1,297 (27)	930 (20)
Other	70 (4)	264 (6)	245 (5)
Unknown	5 (1)	29 (1)	43 (1)
Hispanic	42 (3)	171 (4)	136 (3)
Disease history
IHD	1,582 (100)	2,099 (44)	b
MI	1,582 (100)	757 (16)	b
Diabetes	757 (48)	4,746 (100)	b
CKD	701 (44)	1,713 (36)	b
Stroke/TIA	556 (35)	1,023 (21)	b
Insurance
Private	320 (20)	1,352 (28)	2,137 (46)
Medicaid	146 (9)	503 (11)	387 (8)
Medicare	966 (61)	2,461 (52)	1,528 (33)
Other	15 (1)	64 (1)	71 (1)
Uninsured	135 (9)	366 (8)	567 (12)
Baseline SBP	131.3 ± 18.5	134.7 ± 17.5	136.9 ± 17.3
Number of drugs	2.2 ± 1.0	2.1 ± 1.0	1.7 ± 0.8
Treatment duration, days	105.8 ± 118.9	106.5 ± 119	126.4 ± 140.3

Number of patients with percentage (%) in parentheses and mean ± SD shown.

Abbreviations: BMI, body mass index; IHD, ischemic heart disease; MI, myocardial infarction; CKD, chronic kidney disease; TIA, transient ischemic attack; BP, blood pressure.

^aUncomplicated hypertension defined as lack of IHD, diabetes, CKD, stroke, or TIA.

^bNot applicable.

Lastly, in Analysis 3, we identified the most important sociodemographic/clinical characteristics (explanatory variables) associated with uncontrolled BP (vs. controlled BP) on two drug classes (dependent variable) using multivariable logistic regression analysis. Cases were patients with uncontrolled BP on ≥2 drugs, and controls were patients with controlled BP on 1–2 drugs; patients with controlled BP on monotherapy were included among controls assuming those controlled on one drug are also controlled on two. Covariates were selected through univariate analysis followed by stepwise logistic regression modeling using the same aforementioned model entry/retention criteria. Tested covariates are summarized in Supplementary Table S1 online. Age and BMI were entered as categorical variables (age, 10-year and BMI, 5-unit (kg/m²) increments).

Power calculation for association analyses with BP (Analyses 2 and 3)

In Analysis 2 for the associations between preferred two drug classes and BP control, assuming alpha = 0.05, power = 80%, and given the sample sizes, minimum detectable odds ratios were OR = 1.06, OR = 1.03, and OR = 1.03 among post-MI (n = 4,074), diabetes (n = 12,848), and uncomplicated hypertensive (n = 12,697) patients, respectively. In Analysis 3 examining risk factors for uncontrolled BP on ≥2-drug combination therapy, minimum detectable odds ratio was OR = 1.03 given n = 25,507, alpha = 0.05 and power = 80%.

RESULTS

Based on UFHealth IDR query using i2b2, 324,046 adult outpatients were treated at UFHealth during August 2013–August 2016; of those, 22% (n = 72,693) had a history of hypertension. Among hypertensive patients, 58,965 (81%) were prescribed antihypertensives (treated) during this period. Of treated patients, using historical BP-value cutoff 140/90 mm Hg, 65% had controlled BP.

Table 3 shows the demographic/clinical characteristics of patients prescribed two antihypertensive drug classes. Most prior-MI patients were male and most diabetic patients were of African American ethnicity. Information on the proportion of patients prescribed each major antihypertensive drug class that comprised treatment with two drug classes is in Supplementary Table S2 online. Across all comorbidities, most patients were prescribed ACEI/ARB (68–72%)—except for BB (72%) among post-MI patients—followed by diuretics, BB, and CCB. Among diabetics, 72% received ACEI/ARB, and among post-MI patients, 72% received BB, as part of antihypertensive therapy with two drug classes. Additional information on ACEI/ARB prescribing as part of treatment with two drug classes in CKD patients is in Supplementary Table S3 online.

Figure 2 compares by comorbidity proportions of patients who were prescribed preferred two drug classes. Additional information on specific combinations of common two drug classes is in Table 4. In Figure 2, panel a, across all comorbidity groups, <60% of hypertensive patients on two drug classes were prescribed a comorbidity-specific preferred combination. Uncomplicated/usual care patients were significantly more likely to receive preferred therapy (55%) than diabetic (49%) or post-MI (47%) patients. Among post-MI patients, clinicians were less likely to prescribe ACEI/ARB + BB for those with concomitant diabetes/CKD (42%) than for those without diabetes/CKD (56%) (Figure 2, panel b). In additional analysis, among recent post-MI patients (≤1-year), 51% were prescribed ACEI/ARB + BB.

Figure 2.

Proportion of patients on preferred two drug class antihypertensive therapy by comorbidity. (a) pairwise comparisons by comorbidity, (b) comparison between post-MI patients with and without diabetes and/or CKD. Sample sizes: post-MI, n = 1,582; diabetes without MI, n = 3,989; uncomplicated hypertension, n = 4,690; post-MI with diabetes and/or CKD, n = 1,053, post-MI without diabetes and/or CKD, n = 529. Abbreviations: CKD, chronic kidney disease; MI, myocardial infarction; DIAB, diabetes; Uncomp, uncomplicated.

Table 4.

Proportion of patients treated with preferred vs. non-preferred two drug class antihypertensive therapy by comorbidity

Two drug class antihypertensive therapy	Comorbidityb
	Post-MI	Diabetes	Uncomplicated
	n = 1,582	n = 3,989	n = 4,690
Preferreda
ACEI/ARB + DIUR	188 (12)	1,283 (32)	1,839 (39)
ACEI/ARB + CCB	124 (8)	685 (17)	733 (16)
ACEI/ARB + BB (post-MI)	737 (47)	781 (20)	539 (11)
Non-preferred
BB + DIUR	237 (15)	429 (11)	592 (13)
BB + dCCB	104 (6)	226 (6)	229 (5)
CCB + DIUR	65 (4)	225 (5)	446 (9)
Other	105 (6)	286 (7)	224 (5)
Harmful
ACEI + ARB	2 (0.1)	36 (1)	45 (1)
BB + ndCCB	20 (1)	38 (1)	43 (1)

Number of patients with percentage (%) shown in parentheses.

Abbreviations: MI, myocardial infarction; CKD, chronic kidney disease; UC, uncomplicated hypertension; IHD, ischemic heart disease; TIA, transient ischemic attack; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, beta-blocker; DIUR, diuretic; CCB, calcium-channel blocker; dCCB, dihydropyridine calcium-channel blocker; ndCCB, non-dihydropyridine calcium-channel blocker.

^aPreferred dual therapy for post-MI patients was defined as ACEI/ARB + BB. For patients with diabetes without MI, or uncomplicated hypertension, preferred dual therapy was defined as ACEI/ARB + CCB/diuretic. Uncomplicated hypertension was defined as lack of history of IHD, diabetes, CKD, stroke, or TIA.

^bComorbidity groups are not mutually exclusive; comorbidity groups defined as all post-MI patients, diabetes patients without MI, uncomplicated hypertensive patients without history of IHD, diabetes, CKD, stroke, or TIA.

The low frequency of prescribing comorbidity-specific preferred two drug classes was not explained by worse BP control on the preferred therapy. Instead, we found a trend toward improved BP control with preferred therapy, particularly among post-MI and uncomplicated patients. In descriptive analysis, post-MI patients were more likely to have controlled BP on preferred two drug classes compared to post-MI patients given other two drug classes (see Supplementary Figure S1 online). In multivariable logistic regression analysis, post-MI and uncomplicated patients prescribed preferred therapy were nominally more likely (P < 0.10) to have controlled BP than those prescribed non-preferred therapy (Figure 3, post-MI: OR: 1.21, 95% CI: 0.99–1.48, P = 0.061; Uncomplicated: OR: 1.11, 95% CI: 0.98–1.26, P = 0.089); trends for a difference were not apparent among those within 1-year post-MI (OR: 1.14, 95% CI: 0.85–1.52, P = 0.391, n = 2,088). There were no associations between preferred therapy and BP control among patients with a diabetes history (Figure 3).

Figure 3.

Association between BP control and preferred vs. non-preferred two drug class antihypertensive therapy by comorbidity: results of multivariable logistic regression analysis. Sample sizes: post-MI, n = 4,074; diabetes, n = 12,848; uncomplicated hypertension, n = 12,697. Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin-receptor blocker; BB, beta-blocker; BP, blood pressure; CCB, calcium-channel blocker; DIUR, diuretic; MI, myocardial infarction.

In BP control supplementary analysis, 63% of patients on two drug classes had controlled BP while the corresponding proportion was 60% among those on ≥3-drug classes (see Supplementary Table S4 online). Among patients on ≥3-drug classes, 83% of those with controlled BP were on a diuretic compared with 80% in those with uncontrolled BP (chi-square = 13.4, P = 0.0003).

The results on the sociodemographic/clinical correlates of uncontrolled BP on multiple drug classes are in Supplementary Table S1 and Supplementary Figure S2 online. Factors that were significantly associated with uncontrolled BP on multiple drug classes included African American ethnicity, age, BMI, uninsured status, and history of left ventricular hypertrophy, CKD, diabetes, ischemic heart disease, ischemic stroke, dyslipidemia, and obstructive sleep apnea.

DISCUSSION

We found that substantial proportions of real-world hypertensive patients were not prescribed a preferred combination of two antihypertensive drug classes by comorbidity. A variety of factors may contribute to this finding, including clinicians’ lack of familiarity with hypertension treatment guidelines.^35–38 Another explanation is that the recommended drug classes lead to inadequate BP control. However, our data do not consistently support this; in fact the prescribing of preferred vs. non-preferred two drug classes showed suggestive evidence of better BP-lowering with the preferred two drug classes among post-MI and uncomplicated patients and no differences in BP control for diabetic patients on preferred vs. non-preferred two drug classes.

Across all comorbidity groups, including uncomplicated hypertension, <60% of the patients on two drug classes were prescribed optimized/preferred two drug classes by disease history. Among diabetic patients on two drug classes, ACEI/ARB-prescribing as part of any combination was relatively high (72%) but the preferred combination, ACEI/ARB + CCB/diuretic, was prescribed in only 49%. Despite the basis of this preferred combination being pharmacological, it is likely that prescribing this combination is subject to clinician preference. This combination was not associated with improved BP control among diabetics in our study; thus, it is possible that clinicians did not prescribe this combination with greater frequency because of lack of superior BP-lowering.

Among post-MI patients on two drug classes, 72% were prescribed BB, however, only 47% were on the preferred ACEI/ARB + BB; after limiting the analysis to those <1-year post-MI, this proportion increased slightly to 51%. Also, clinicians were less likely to prescribe ACEI/ARB + BB for post-MI patients with diabetes and/or CKD than for post-MI patients without diabetes and/or CKD. The low frequency of ACEI/ARB + BB prescribing among post-MI patients, regardless of other comorbidities, even at 1-year is concerning. It is also concerning that diabetics were less likely to receive the combination. Likewise, the lower ACEI/ARB-prescribing in post-MI patients with CKD may imply that clinicians are overly concerned about the renal effects of ACEI/ARB, compared to their post-MI benefits. However, we cannot rule out that ACEI/ARB was later added by a non-UF Health provider after patient stabilization post-acute MI.

In BP control analyses, post-MI and uncomplicated patients prescribed preferred two drug classes (ACEI/ARB + BB and ACEI/ARB + CCB/diuretic, respectively) showed a trend toward better BP control compared to those prescribed other combinations. ACEI/ARB + BB is not mechanistically complementary and is recommended in post-MI patients primarily for its cardioprotective benefits. The finding that this cardioprotective combination is associated with greater BP-lowering lends additional support for its use among post-MI patients. As aforementioned, among diabetic patients, we found no association between preferred two drug classes and BP control. It is possible that post-MI patients taking BB were more treatment-adherent because a “heart attack” may be perceived as more “serious” and more acutely impactful than a diabetes diagnosis. Also, compared with diabetic patients, post-MI patients may be more likely to see a second clinician, such as a cardiologist, resulting in more patient monitoring/follow-up.

The most prominent sociodemographic/clinical factors associated with uncontrolled BP on ≥2-drug classes vs. controlled BP on 1–2-drug classes among patients in a real-world clinical setting included African American ethnicity, uninsured status and left ventricular hypertrophy, CKD, and diabetes history. These findings largely corroborate factors associated with difficult to control BP and resistant hypertension reported in randomized controlled trial and observational studies.^{31–34,39–41}

This study has several strengths, including use of large datasets from a real-world clinical setting, diverse sample, ambulatory patient data from generalist and specialist clinics, cost/time-efficient study-conduction, and unbiased patient/records selection. Findings may be generalizable to similar institutions.

Limitations of the study include lack of diagnoses and prescriptions data from outside UFHealth; thus, numbers of diagnoses and percent treated may be underestimations. However, we attenuated these factors by incorporating both UFHealth-prescribed and historical/patient-reported prescriptions and using available sources of diagnoses (encounters, problem lists, and billing) since UFHealth IDR/i2b2 launch (June 2011). We do not know the extent to which the prescribed regimens were the initial vs. adjusted regimens per drug tolerance. Treatments and diagnoses were not manually verified, although the study’s goals were primarily to gain clinical understanding through large biomedical/EHR data analysis. We also did not have information on drug adherence, dosages, or refills, and BP values may not have been measured per protocol; thus, BP response analyses should be interpreted cautiously. However, these may not be limitations in that the findings may reflect patient care in a routine clinical setting. Lastly, we did not examine associations between clinicians’ characteristics and prescribing of various combinations of drug classes as we did not collect information on clinician characteristics.

In summary, among a large cohort of hypertensive patients in a real-world clinical environment, we found that BP control was 65%. Prescribing of guideline-recommended antihypertensive drug classes and drug class combinations based on concomitant diseases is suboptimal. Although we found a trend toward association of guideline-recommended therapy with BP control, our analysis does not clearly support the recommendations of preferred combinations for optimal BP control. Overall, our study demonstrates an EHR-based approach to investigating healthcare issues among patients in a real-world clinical setting where there may be implications to implement clinical decision support to improve healthcare at the community and health system-wide level.

FUNDING

University of Florida Clinical and Translational Science Institute, funded partly by the National Center for Advancing Translational Sciences (UL1TR001427), National Institutes of Health training grants: T32 DK104721 (PI: Mark S. Segal; Magvanjav), KL2 TR001429 (McDonough).

DISCLOSURE

The authors declared no conflict of interest.