Is Patients’ Insurance Coverage Associated With Prescribing After Hospitalization for Severe, Poorly Controlled Hypertension?

Abstract

High medication costs may be a significant cause of nonadherence and threaten recent gains in hypertension treatment. It is unclear whether prescribing patterns differ with patients’ insurance coverage. The objective of this study was to determine whether insurance coverage, reported difficulty affording medications, or nonadherence were associated with antihypertensive prescribing in a high‐risk population. The authors conducted a cross‐sectional survey of 189 patients admitted to an inner‐city academic hospital with severe, poorly controlled hypertension. Patients’ poor medication access (one‐third lacked insurance and half reported difficulty affording medications) was not associated with admission or discharge regimen costs. Substituting the least expensive drug within each class would have reduced costs by 42%, and reducing calcium channel blocker use would have significantly reduced costs. In conclusion, markers of poor medication access were not associated with prescribing patterns. Further research is needed to explore these patterns and their impact on vulnerable populations’ financial burden and adherence.

An estimated 28.7% of the US population was hypertensive in 1999 and 2000. ¹ Despite a very high level of awareness, approaching 70%, hypertension was controlled in only 31.0% of patients. ¹ Lack of medication insurance coverage and difficulty affording antihypertensive medications may account for a portion of this poor control. Higher costs can lead to less medication use and more adverse outcomes. ² , ³ , ⁴ Although drug benefits may alleviate costs for some, even patients with medication coverage often have difficulty affording medications. ⁵ In the United States, patients are increasingly having difficulty affording their medications. More prescription drug use, rising costs for new drugs, and decreasing prescription coinsurance are all contributing to escalating costs. Higher costs are especially troublesome for vulnerable populations: lower income, minority ethnicity, less education, and poorer health are all risk factors for medication self‐restriction. ⁶

Physicians can substantially impact patients’ medication costs. This is particularly true with hypertension: although research supports the use of less expensive diuretics, ⁷ , ⁸ physicians commonly prescribe more costly medications. In one study, 40% of antihypertensive prescriptions were inconsistent with evidence‐based guidelines; changing to more appropriate regimens would have reduced costs by one‐quarter. ⁹ Although physicians want to reduce costs when patients have financial difficulties, ¹⁰ studies to date have not evaluated whether physicians actually alter their prescribing when patients are not taking their medications due to cost.

We therefore evaluated whether insurance coverage was associated with physicians’ prescribing after hospitalization for severe, poorly controlled hypertension in patients requiring antihypertensives. This adverse outcome occurs mainly in vulnerable populations and is due largely to medication nonadherence. ¹¹ We asked the following research questions: (1) was insurance coverage, including patients’ reports of difficulty affording medications, associated with how housestaff prescribed discharge antihypertensive medications for this population and (2) could alternative antihypertensive choices have substantially reduced medication costs?

Methods

Study Design, Setting, Participants, and Sampling

We conducted a cross‐sectional structured interview study of African Americans admitted to Johns Hopkins for severe, poorly controlled hypertension. This large, inner‐city teaching hospital serves an African American community with very high rates of hypertension and poor access to care. Many patients are served by the same housestaff in both the outpatient and inpatient settings, and many have multiple hospitalizations for uncontrolled blood pressure. Although pharmaceutical detailing does not occur in the inpatient setting, drug company samples are available for outpatients. Patients with financial need could receive a 2‐week voucher for discharge medications, but more comprehensive discharge planning services were not provided. We screened all patients admitted to medicine wards between August 1999 and June 2001 and between February 2002 and December 2004. The Johns Hopkins Medicine Institutional Review Board approved the protocol, and all patients provided informed consent.

We defined severe, poorly controlled hypertension as at least 2 admission blood pressure readings ≥180/110 mm Hg. A study physician reviewed all admission notes and excluded patients who appeared to have elevated blood pressure due to secondary causes (eg, acute pain). We invited patients to participate if they were older than 18 years, residents of Baltimore, and able to provide informed consent.

We identified 485 patients with eligible blood pressure readings. Of these, 193 (40%) were ineligible for clinical reasons, primarily because the elevated blood pressure was likely secondary to another condition. Of 292 eligible patients, 9% refused or withdrew, 20% were discharged or left against medical advice before they could be contacted, 10% never completed the questionnaire, and 7% died. Therefore, 193 patients completed the study, for a response rate of 70%. Since 4 patients had no prescribed discharge regimen because they died or their blood pressure normalized during hospitalization, we included 189 patients in these analyses.

Data

Measurements included insurance coverage (patient interview and medical record), reported difficulty affording medications and adherence (patient interview), and specific antihypertensives and drug classes prescribed on admission (patient interview and medical record) and discharge (patient interview and medical record). Trained interviewers administered a structured questionnaire and reviewed the admission history and discharge summary. We adapted questions from inner‐city trials to improve the control of hypertension ¹² and diabetes ¹³ and refined the questionnaire through a pilot with 10 patients. The questionnaire included history of hypertension, sociodemographic factors, and access to care. Our goal was to assess the same information that clinicians would have access to in their patient interactions. Therefore, we asked about difficulty affording medications in several ways, including “Do you have difficulty paying for your medications?” and “Do you ever miss your pills because you can’t afford them?” We assessed adherence on admission by asking “Had you missed taking your blood pressure pills before you came into the hospital?” If a patient answered yes, we asked, “Why did you miss taking them?” We asked about insurance and medication co‐pays and abstracted insurance coverage from the hospital billing system.

We asked patients to list medications prescribed before admission and reviewed the admission history for documentation about medications. We defined antihypertensive medications according to The Medical Letter. We chose this resource because physicians use it more than any other for general cost information when deciding which drug to prescribe. ¹⁴ In addition, drug prices often do not vary with the dose. We used total costs because our primary analyses were for patients without insurance or medication coverage and because our goal was to compare prescribing patterns to those for patients with coverage. We estimated costs as the average monthly patient cost for the lowest dose, based on The Medical Letter nationwide pharmacy audit data for 1999 and 2000. ¹⁵ Costs were essentially the same in a 2001 and 2002 audit. ¹⁶ If a physician or patient listed a medication by its brand name, we used the brand name cost, but we performed a sensitivity analysis using generic costs for all drugs.

Analysis

Reliability and Validity

We described key variables using more than one question or source and evaluated them for percentage of agreement. Whenever possible, we evaluated questionnaire items for reliability with information from the medical record and/or billing data. We evaluated the construct validity of the difficulty affording medications questions by measuring associations with insurance coverage. The validity of patients’ knowledge of their insurance was very high: only 1 patient with Medicare did not report this, and only 1 who reported he had insurance was uninsured in the billing database. Ninety percent of persons with Medicaid were aware that it covered medications. The questions about difficulty affording medications demonstrated good reliability and construct validity. There was good agreement between the questions about difficulty affording medications: 88% of those who reported that they sometimes missed medications because they could not afford them also reported difficulty paying for their medications. Insurance coverage and difficulty affording medications were strongly associated: in patients without insurance, 78% reported difficulty affording medications.

Univariate and Multivariable Analyses

Our conceptual model is shown in the Figure. The unit of analysis was the patient. The dependent variable was the estimated total cost for the discharge antihypertensive regimen from the current hospitalization. For our primary research question, on insurance status and patterns of prescribing, we performed separate regressions using as the independent variable whether patients had health insurance or insurance coverage of medications for the primary analyses. In secondary analyses, we performed regressions with the independent variable of whether the patient reported co‐pays for medications or difficulty affording medications. For our second research question, to determine the potential impact of drug substitution, we calculated the difference in cost in the discharge regimen with substituting the lowest‐cost, clinically equivalent medication in each class for the medication that was actually prescribed on discharge.

Figure.

 Conceptual model of physician decision making and the cost of medications.

We performed two multivariate analyses for the first research question: in the first, we adjusted for the number of discharge antihypertensive agents and the number of other discharge medications, and in the second, for those patients for whom this information was available, we adjusted for the prescribed admission drug classes and cost of the prescribed admission regimen. To evaluate the impact of comorbidities, we conducted sensitivity analyses excluding patients with end‐stage renal disease or congestive heart failure. We also performed sensitivity analyses using different definitions of clinical equivalence (only substituting within dihydropyridine calcium channel blockers and thiazide diuretics and not substituting captopril, since the cost in The Medical Letter was obtained differently and it is a twice‐daily medication).

Results

Population Characteristics and Costs

The mean age of the study population was 50 years (range, 27–88 years), 47% were male, 54% were high school graduates, and 25% were married. Only 32% were employed; 36% were on disability, and the median household income was $550 a month. The median duration of hypertension was 13 years, and patients reported a median of 2 admissions in the last 5 years related to hypertension (range, 0–19). Eleven percent had end‐stage renal disease and 22% had heart failure. One‐third of patients had no insurance, and one‐third of those with insurance reported difficulty affording medications. Two‐thirds had missed their medications before admission, and among those without insurance, 80% reported that the reason was financial or running out. Few patients reported other reasons for missing medications, such as adverse effects (6%) and forgetting (7%).

The monthly costs of the admission (mean, $55; range, $6–$161) and discharge (mean, $54; range, $8–$140) antihypertensive regimens were high (10% of the median household income) and showed substantial variation. On average, patients were also prescribed as many other additional medications as antihypertensives (mean, 2 additional medications; range, 0–12). Table I shows the cost of the prescribed admission and discharge regimens by insurance coverage, reported difficulty affording medications, and adherence. None of these variables was significantly associated with costs in either univariate or multivariate analyses.

Table I.

 Patient‐Reported Markers of Difficulty Affording Medications and Monthly Total Cost of Antihypertensive Regimen

Mean (range)	No. of patients (%)	Cost of Admission Regimen, a $ (n=141)	Cost of Discharge Regimen, $ (n=189)
No insurance	60/189 (32)	41	51
Patients with insurance
No medication coverage	26/128 (20)	61	54
Medication co‐pays	54/128 (42)	57	53
Difficulty affording medications	45/128 (35)	50	51
Reported missing medications before admission	122/189 (65)	54	54
Reported no medications for at least 2 weeks before admission	97/189 (51)	41	51

^aFor patients with an admission regimen reported in the medical record (n=141/189, 75%).

Medication Choices

Table II shows prescribed antihypertensive medications on admission and discharge, listed by the generic name, and their brand and generic cost ranges. Calcium channel blockers were the most frequently prescribed before admission (91 of the 141 with available information, or 65%, compared with 67 of 141, or 48%, for the next most frequently prescribed class, β‐blockers). The largest increases from the admission to the discharge regimen were in the two drug classes that are least expensive and best supported by guidelines, β‐blockers (67 to 109, or a 63% increase) and diuretics (63 to 102, or a 62% increase), despite the fact that diuretics would infrequently be used to acutely control blood pressure during hospitalization. For the most expensive classes, angiotensin‐converting enzyme (ACE) inhibitors and calcium channel blockers, housestaff preferred one drug within each class. Although the few patients who had been prescribed the other drugs within the class before admission tended to be prescribed the same drugs at discharge, new prescriptions were almost exclusively confined to one choice in the class. Although one option in each of these two drug classes was much less expensive than the others, this option was rarely prescribed.

Table II.

Antihypertensives, Frequency Prescribed, and Costs on Admission and Discharge, Listed by Generic Name

Class/Medication (In Order of Cost)	No. Prescribed		Average Monthly Cost
	Admission	Discharge	Generic	Brand Name
ACE Inhibitors
Captopril	0	1	$8	$28
Enalapril	7	3	$21	$26
Lisinopril	41	74	–	$28
Other	4	1	–	$26–$28
Total	52	79
Calcium channel blockers
Verapamil	13	13	$22	$38–$40
Nifedipine	53	91	$37	$36–$41
Diltiazem	5	5	$40	$37–$57
Amlodipine	17	14	–	$40
Other	3	0	–	$32–$46
Total	91	123
β‐Adrenergic blockers
Metoprolol	36	59	$8	$20–$22
Atenolol	14	24	$9	$32
Propranolol	1	1	$8	$34
Labetalol	15	22	–	$25
Carvedilol	1	3	–	$92
Total	67	109
Diuretics
Furosemide	38	39	$4	$8
Chlorthalidone	1	1	$5	–
Hydrochlorothiazide	22	43	$14	$16
Hydrochlorothiazide/triamterene	6	7	$11	$15
Other	6	12	$9–$14	$12–$23
Total	73	102
Other
Central α‐adrenergic agents (clonidine)	34	24	$6	$22–$40
α‐Adrenergic blockers	4	3	$5–$25	$15–$31
Direct vasodilators	10	10	$1–$8	–
Combination drugs	6	5	–	$31–$51
Total	54	42

Costs are from The Medical Letter, 1999–2000; they are very similar to costs from The Medical Letter, 2003. Admission regimen information was available for only 141 of 189 patients. For purposes of simplification, drugs prescribed under a brand name are included under the generic name, and varying costs are shown as a range. Where no generic price is shown, no generic alternative was available at the time of the study. The least expensive alternative in each class is shown in boldface. Nonthiazide diuretics were likely frequently prescribed for reasons other than hypertension. All of these drugs were available on the hospital formulary.

Table III summarizes the impact of individual drug choices on costs. Substituting the generic equivalent when a brand name was prescribed would have reduced average patient monthly costs by only $2, since many drugs were not available as generics and some generics were as costly as brand‐name drugs. However, substituting the least expensive alternative within each class would have reduced costs by 42%, from a mean of $54 to $31 per patient per month (range, $8–$81). Costs for calcium channel blockers would still have remained high, and substituting drugs from other classes would have also substantially reduced costs. In the sensitivity analyses, results were generally similar, although restricting the definition of therapeutic equivalence would have reduced cost savings somewhat for substituting within drug classes.

Table III.

Discharge Prescriptions and Monthly Drug Costs, by Most Commonly Prescribed Drug Classes

Drug Class	% of Patients Prescribed Drug in This Class	Mean Cost of Drug Chosen (for All Patients), $	Cost of Least Expensive Alternative in Same Class, $	Total Monthly Projected Population Spending, $	Total Monthly Spending if Least Expensive Alternative Had Been Used, $	Difference Between Projected and Alternative (Least Expensive) Costs, $ (%)
Calcium channel blockers	65	36	21	4423	2580	1843 (42)
Angiotensin‐converting enzyme inhibitors	43	27	8	2194	650	1544 (70)
β‐Blockers	59	15	8	1672	892	780 (47)
Diuretics	52	7	5	688	491	197 (29)

Discussion

In this population of patients admitted for severe, poorly controlled hypertension, despite high rates of nonadherence reportedly due to insurance issues and difficulty affording medications, indicators of insurance coverage were not associated with physicians’ cost sensitivity when prescribing discharge antihypertensive regimens. Particularly for ACE inhibitors and calcium channel blockers, few patients were prescribed the less expensive alternatives within the drug class, even though using these alternatives would have reduced costs nearly by half. Calcium channel blockers, the most expensive drug class with the least evidence to support their use, were the most frequently prescribed class, and reducing their use would have also significantly decreased costs.

Why were the least expensive medications rarely used, and why were markers of insurance status not associated with cost sensitivity? The complexity of medication costs, lack of information on and understanding of medication coverage, and inadequate communication about difficulty affording medications may complicate efforts to reduce patients’ financial burden. Physicians have difficulty accounting for medication costs for many reasons, including the relative inaccessibility of medication cost and medication coverage information. In one study, 80% of physicians reported that they often felt unaware of medication costs; when asked to estimate costs of specific medications, they underestimated in 40% of cases. Also, physicians often misunderstood complexities of insurance coverage, such as whether Medicare covered prescriptions, and intended to reduce costs only for uninsured patients. ¹⁴

Physicians may often be unaware of patients’ difficulty affording medications and self‐restriction. Other studies have found that most patients who self‐restricted medications did not discuss it with their physician. ¹⁷ , ¹⁸ Of those who did not discuss the issue, two‐thirds reported that nobody asked them about ability to pay, and more than half said they didn’t think providers could help them. Alexander and associates ¹⁸ also found that patients with greater financial burden were more likely to discuss costs. Among physicians, the most common barriers to discussing costs were lack of time and the belief that they could not offer a viable solution. ¹⁹

Patients’ and physicians’ perspectives of nonadherence may differ in other ways. Most of the nonadherent patients in our study reported difficulty affording medications or running out as a cause of their nonadherence. Few reported reasons such as taking them too many times a day or experiencing adverse effects, although these are frequent targets of physician interventions to improve antihypertensive adherence. ²⁰ Physicians’ perceptions of antihypertensive medications may also not accurately reflect clinical evidence. Despite extensive clinical trial data showing that diuretics are as or more effective and well‐tolerated as more expensive medications, ²¹ , ²² a recent survey found that physicians commonly believed that diuretics are less effective than other medications. Physicians who used free samples were significantly more likely to hold these beliefs. ²³

Will physicians prescribe less costly drugs? In an observational study, a third of patients who brought up the issue of costs reported that their medications were not changed, 30% were told about programs that help pay for drugs, and 28% reported they were told how to find less expensive medications. ²⁴ Much of the cost burden is also due to variation among physicians. A survey of physician attitudes found that those with the highest prescribing costs were significantly more likely to see drug company representatives and to use more newly available medications. ²³

Interventions to change physician prescribing have had mixed results. ²⁵ Structural interventions to keep medication costs low by altering drug choices can be effective, although they are usually oriented toward insurer costs. In British Columbia, reference‐based pricing, which provides financial incentives for patients to use the least expensive therapeutically equivalent medication, substantially reduced expenditures for ACE inhibitors ²⁶ without decreasing overall antihypertensive agent utilization or increasing health care utilization. ²⁷ Alternatively, a generic‐only program in a Medicare HMO did reduce pharmacy costs but may have also reduced drug utilization and increased hospital admissions. ²⁸

Does the cost of the discharge regimen matter? A review of the literature ²⁹ and more recent research ³⁰ have concluded that patients are more likely to self‐restrict when costs are high or they increase. However, the relationship between costs and adherence is complex and is likely affected by patient factors, such as other financial needs, and care factors, such as physician‐patient trust. Reducing costs or decreasing co‐pays alone may not improve adherence. The transition to Medicare for previously uninsured hypertensive patients did not increase their use of antihypertensives. ³¹ In a national survey of managed care Medicare beneficiaries, although some patients who exceeded pharmacy coverage gaps stopped taking their medications, many others tried to find ways to decrease their drug costs, such as calling different pharmacies. ⁵ Medications for chronic illnesses are most vulnerable to self‐restriction, and patients with less education or support may have more difficulty coping with costs and be more likely to stop medications.

Our study has a number of limitations and strengths and suggests several areas for further exploration. The concept of difficulty affording medications is complex, and the relationships with insurance and adherence are not well‐defined. Actual costs may not matter for a subset of patients with fixed co‐pays, but we used this variable because our primary comparisons were for patients with and without insurance coverage. Since we surveyed only patients, not their physicians, we could not determine whether physicians were aware of patients’ insurance coverage, medication costs, or other factors influencing decision making; although we did review the admission histories and discharge summaries, these issues were rarely documented there. Further research should interview physicians to determine how factors related to insurance and costs may influence their prescribing. A number of specific factors, such as the presence of housestaff, local practices, or the inclusion of only hospitalized patients, may have affected our findings. Although housestaff may be less aware of these issues, with their important role in choosing medication, potentially more resources for them on costs might be beneficial. Finally, continuing changes in drug costs, including the availability of several new generics, may change the relative price of different classes of antihypertensives. However, generic calcium channel blockers and most ACE inhibitors continue to be more expensive than β‐blockers and diuretics. ³²

Conclusions

We found that, although minor changes in medications could result in markedly reduced costs, insurance status or reported difficulty affording medications were not associated with antihypertensive discharge regimen costs in patients admitted for poorly controlled hypertension. Medication costs continue to rise, and changing pharmacy benefits may further complicate these issues and make costs more challenging for physicians and patients to address. We believe that our results provide justification for the need for more transparency and availability of information about costs and more attention to difficulty affording medications, especially in vulnerable populations. In addition, we need better ways to improve prescribing and partner with vulnerable patients to minimize medication cost burden while maintaining a high quality of care.