Abstract
Objectives:
This study examines the evolutionary impact of valsartan initiation on medical costs.
Methods:
A retrospective time series study design was used with a large, US national commercial claims database for the period of 2004–2008. Hypertensive patients who initiated valsartan between the ages of 18 and 63, and had continuous enrollment for 24-month pre-initiation period and 24-month post-initiation period were selected. Patients’ monthly medical costs were calculated based on individual claims. A novel time series model was devised with monthly medical costs as its dependent variables, autoregressive integrated moving average (ARIMA) as its stochastic components, and four indicative variables as its decomposed interventional components. The number of post-initiation months before a cost-offset point was also assessed.
Results:
Patients (n = 18,269) had mean age of 53 at the initiation date, and 53% of them were female. The most common co-morbid conditions were dyslipidemia (52%), diabetes (24%), and hypertensive complications (17%). The time series model suggests that medical costs were increasing by ∼$10 per month (p < 0.01) before the initiation, and decreasing by ∼$6 per month (p < 0.01) after the initiation. After the 4th post-initiation month, medical costs for patients with the initiation were statistically significantly lower (p < 0.01) than forecasted medical costs for the same patients without the initiation.
Limitations:
The study has its limitations in data representativeness, ability to collect unrecorded clinical conditions, treatments, and costs, as well as its generalizability to patients with different characteristics.
Conclusions:
Commercially insured hypertensive patients experienced monthly medical cost increase before valsartan initiation. Based on our model, the evolutionary impact of the initiation on medical costs included a temporary cost surge, a gradual, consistent, and statistically significant cost decrease, and a cost-offset point around the 4th post-initiation month.
Introduction
Hypertension is a chronic medical condition with elevated blood pressure. The condition often comes without symptoms, but can lead to strokes, chronic kidney failure, heart attacks, heart failure, arterial aneurysm, and other cardiovascular morbiditiesCitation1–10. Even moderate hypertension can lead to shortened life expectancyCitation11.
Despite medical advances in hypertension care, therapeutic choices to cure the disease have been limited by lack of efficacy or side-effectsCitation12–15. Thus, practical treatments for hypertension are still aiming at controlling blood pressure and preventing or delaying cardiovascular outcomes.
Existing classes of anti-hypertensive medications include oral diuretics, adrenergic receptor antagonists, adrenergic receptor agonists, calcium channel blockers, angiotensin-converting enzyme (ACE) inhibitors, angiotensin receptor blockers (ARBs), and direct rennin inhibitors (DRIs).
Over time, researchers have shown that ARBs are effective at reducing blood pressureCitation16–21. However, patients, providers, and payers still want to know whether an increase in medication costs for an ARB can be offset by a reduction in non-ARB medical costs over a long time period and in a real world setting. Our recent literature review over the last-10-year publications suggests that no real-world study has ever been published to answer this question. In fact, after reviewing the existing studies on the cost and use of ARBs for hypertension, a recently published manuscript has called for a longitudinal study to examine the long term cost-effectiveness of ARBsCitation16.
To fill this information void, and to provide patients, providers, and payers with real world evidence, we have conducted a time series analysis study that aims to examine the evolutionary impact of the most widely used ARB, valsartan, on medical cost among hypertensive patients.
Our research questions were: (1) whether valsartan initiation was associated with a post-initiation medical cost reduction?; and (2) if the post-initiation cost reduction did occur, what is the number of months after which the post-initiation costs will be at or lower than the predicted costs for the same patients as if no valsartan was initiated?
The rationale for us to select valsartan as a proxy of ARBs lies in the fact that valsartan was the most widely prescribed ARB (representing ∼40+% ARB use in the USCitation22).
Methods
Data source
This study was conducted using 2004–2008 MarketScan commercial insurance claims databases from Thomson Reuter, New York, USA. The database includes patient enrollment files, inpatient claim files, outpatient claim files, and medication claim files collected in the US. The enrollment files contain patients’ demographic information such as age, gender, geographic residential region, health plan type, and detailed enrollment records. The inpatient and outpatient claims files contain International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes, Current Procedural Terminology (CPT) codes, date of service, place of service, provider type, health plan, and patient paid amounts. The medication claims files contain National Drug Code (NDC), dispense dates, quantity, days supplied, and paid amounts by plans and patients. Linked through an encrypted unique personal identifier, these individual files can provide comprehensive information on patients’ medical conditions, healthcare utilization, and cost; therefore, enabling us to understand how the inpatient, outpatient, and pharmacy care are provided for hypertension patients in the real world.
Because all identifying information was removed from the original claims database by the third party vendor prior to its release, approval from an institutional review board (IRB) or other ethic review entities was neither needed nor sought.
Study design and duration
Our study used a “quasi-experimental time series study design” defined by Campbell and StanleyCitation23, in which valsartan initiation was viewed as an intervention for hypertension for the select hypertensive patients. These patients’ clinical characteristics and medical costs were successively measured every month throughout a 24-month pre-initiation period and a 24-month post-initiation period (the maximum time length we can get from our 4-year claims databases). The study design and duration scheme can be expressed as follows:
Subject identification and selection
Our study cohort consisted of hypertensive patients who initiated valsartan in 2006 at age between 18–63, and were continuously enrolled in a commercial health plan throughout the pre- and post-initiation periods. All select patients must have had at least one inpatient or outpatient claim with a hypertension diagnosis code (ICD-9-CM: 401, 402, 404, 437.2X, 362.11) before the initiation, and, at least, 31 days of valsartan supply during the post-initiation period. Subjects with valsartan before the initiation or with only one valsartan prescription in the post-initiation period were excluded from the study. Finally, the measure of valsartan use was based on national drug codes with generic name containing “Valsartan”, “valsartan”, or “VALSARTAN”.
Measurement of demographic and clinical characteristics
We measured the cohort’s demographic characteristics on the initiation date. The measured demographic characteristics included cohort mean age, percent of female patients, percent of patients in each one of health insurance types (e.g., comprehensive health plan, HMO health plan, POS health plan, PPO health plan, other health plan), and percent of patients in each one of geographic residential regions (e.g., northeast, north central, Midwest, south).
Clinical characteristics of the cohort were measured through either prevalence rates or incidence rates of hypertension related co-morbid conditions. The prevalence rate of a hypertension-related co-morbid condition at a specific time point is the percent of patients with that co-morbid condition at that time point; while the incidence rate of that co-morbid condition in a specific time period is a ratio between the number of new patients with the 1st diagnosis of that co-morbid condition in that period and the number of patients without that condition in the prior period.
The co-morbid conditions examined in our study included hypertension complications (in renal, cardiovascular, cerebrovascular, or vision systems), paralysis, congestive heart failure, ischemic heart diseases (including myocardial infarction, angina, and coronary artery disease), diabetes, dyslipidemia, arterial fibrillation, chronic kidney diseases, stroke, cerebrovascular disease other than stroke, and peripheral vascular disease.
Measurement of medical costs
The primary outcome of our study was the per-member-per-month (PMPM) medical costs, which include inpatient, outpatient, and medication costs reimbursed by payers (not including patients’ copay). To derive it, we first measured each select patient’s monthly medical costs successively for each study month. We, then, used 10% winsorized meanCitation24 of that costs as the cohort’s monthly medical costs with a purpose of improving the robustness of our analysis. Further, to adjust for inflation effect, we applied 5% annual discount rate to standardize the dollars of medical costs in a non-initiation month to the dollars in the initiation month. Finally, we measured the cohort’s monthly inpatient costs, monthly outpatient costs, and monthly medication costs (including other anti-hypertensive medications).
Statistical analysis
Descriptive data analysis
Descriptive data analyses were performed to examine the cohort’s demographic and clinical characteristics, such as the cohort’s mean age, percent of female patients, percent of patients in each residential region, percent of patients in each health plan type in the initiation month, as well as prevalence or incidence rates of hypertension related co-morbid conditions. Additional descriptive data analyses were performed to assess and compare the cohort’s medical costs in the 1st month of pre-initiation period, the initiation month, and the last month of post-initiation period. All comparative statistics between different time points were tested using dependent t-test for paired means of continuous variables, and McNemar test for repeated measures of proportions across time.
Time series analysis
We examined the evolutionary impact of valsartan initiation on medical costs through using a novel time series model.
Like traditional interrupted time series models and regime switching time series models used in assessing the impact of an intervention on medical costsCitation25–32, the model devised for this study included a stochastic component represented by autoregressive, integrated, moving average parts of ARIMA (p, d, q), and structural or interventional components. However, different from these traditional time series models, our novelly-devised models decomposed the structural/or interventional component into the following four sub-components: (1) marginal (or monthly) change of medical costs in the absence of a treatment initiation, (2) marginal (or monthly) change of medical costs in the presence of a treatment initiation, (3) short-term change of medical costs at initiation, and (4) additional impact of a treatment initiation over the observed time period. The conceptual framework of our novelly-devised time series model is presented in , and its specification for this study is expressed as follows:
where: Yt = monthly medical costs, p = the order of autoregressive part, d = the order of integrated part, q = the order of moving average part,
= sequence of pre-initiation month(s) (1–24, 0), Mta = sequence of post-initiation month(s) (0, 1–24), St = initiation month indicator (0 or 1), Dt = post-initiation indicator t (0 or 1), and εt = error terms.
Figure 1. A new conceptual framework of medical cost time series.
The coding scheme of time indicative variables ,
, St, and Dt was as follows:
Based on the above time series model and coding scheme, we estimated Yt in the presence of the initiation , and forecast Yt in the post-initiation period as if without the initiation
as follows:
and
If the valsartan initiation leads to a reduction in
, then, we should have:
In other words, if the valsartan initiation was associated with a post-initiation reduction in medical costs, we should be able to find a cost off-set time point , so that:
and the number of the post-initiation months after which the post-initiation medical costs to be at or lower than the predicted medical costs without the initiation is
. Similarly, the number of the post-initiation months after which the post-initiation medical costs were at or lower than that in the last pre-initiation month is
− 24, so that
In building this time series model, we used an augmented Dickey-Fuller (ADF) testCitation33–35 to check the stationarity of monthly medical cost time series, the Kolmogorov-Smirnov testCitation34,Citation36 and the Ljung-Box (Q) testCitation37 to check if the residuals were white noise, Levene’s testCitation38 to evaluate homoscedasticity in the residuals, and Akaike information criterion (AIC) testCitation39 to select the best fitted model.
Results
Study sample and demographic characteristics
Our data source contained 18,269 unique individual subjects who met our pre-specified inclusion and exclusion criteria. All of them were included in the study.
presents the cohort’s demographic characteristics at the initiation month with a mean age of 52.5 years, 52.6% female patients, 53.8% of PPO health plan members, and 54.7% southern region residents.
Table 1. Demographic characteristics of hypertensive patients with valsartan initiation.
Clinical characteristics
presents the cohort’s prevalence rates of the most common hypertension-related co-morbid conditions. At the first month of the pre-initiation period, 4.3% had hypertension complications, 4.5% had ischemic heart diseases, 13.3% had diabetes, and 21% had dyslipidemia. Twenty-four months later (at the initiation month), the prevalence rates of these co-morbid conditions increased significantly with 16.9% for hypertension complications, 13.4% for ischemic heart diseases, 23.5% for diabetes, and 52.4% for dyslipidemia. By the last month of the post-initiation period, these prevalence rates increased further with 23.4% for hypertension complications, 18.8% for ischemic heart diseases, 30.5% for diabetes, and 68.7% for dyslipidemia.
Table 2. Prevalence rate comparison of hypertension-related co-morbid conditions among hypertensive patients with valsartan initiation.
compares the cohort’s incidence rates of hypertension-related co-morbid conditions between the pre-initiation period and post-initiation period. The unit of the incidence rates here was the number of new cases of a co-morbid condition per 1000 patients without that condition in the previous period. All incidence rates in the post-initiation period were statistically significantly lower than the rates in the pre-initiation period (p < 0.05), except for chronic kidney disease and peripheral vascular disease. The co-morbid conditions with the most significant decreases in their post-initiation incidence rates were dyslipidemia (before: 2.85 vs after: 1.48, p < 0.05), hypertensive complications (before: 1.15 vs after: 0.59, p < 0.05), ischemic heart diseases (before: 0.81 vs after: 0.50, p < 0.05).
Table 3. Incidence rate comparison of hypertension-related co-morbid conditions among hypertensive patients with valsartan initiation.
Medical cost characteristics
presents various PMPM medical costs in the 1st pre-initiation month, initiation month, and last post-initiation month. All types of medical costs went through significant increases from the 1st month of pre-initiation period to the initiation month, and significant decreases from the initiation month to the last month of post-initiation period (Total: $292 → $756 → $450; Inpatient: $83 → $299 → $149; Outpatient: $124 →$265 → $167; Medication: $85 → $192 → $134, all p < 0.05).
Table 4. Evolutionary comparison of inpatient, outpatient, and medication costs at three different time points.
Time series analysis
presents a time series model of monthly medical costs with previous month’s medical costs as one of its covariates (i.e., ARIMA(1,0,0)). The model was selected out of a group of time series models (with different orders of p, d, q) because of its least AIC value and its statistical significance for all coefficients (all with p < 0.005). With the same notations used before, the final time series model can be expressed as follows:
Table 5. Time series model of monthly total medical costs with a decomposed intervention structure.
Based on the coefficients of this final model, the estimated monthly medical costs consistently increased from the 1st month of pre-initiation period at a rate of $9.9 per month throughout the pre-initiation period, surged $209.8 in the initiation month (β2 + β3 + β4 − 24β1 = −6.1 + 226.1 + 221.3 − 9.9 × 24 = 209.8, then consistently decreased at a rate of $6.1 per month for the rest of the post-initiation months.
Assuming that the medical costs would continue to increase at a rate of $9.9 per month in the post-initiation period if valsartan was not initiated, we predicted the monthly medical costs in the post-initiation period through setting
(as if valsartan had not be initiated). We found that at and after the 4th post-initiation month (i.e., t ≥ 28), the predicted monthly medical costs without initiation were higher than the estimated monthly medical costs with initiation, even with the consideration of 95% confidence intervals (see ). Hence, we conclude that valsartan initiation was associated with a reduction in monthly medical costs, and a cost-offset could occur around the 4th post-initiation month.
Figure 2. Time series of total medical costs.
Discussions
Our study was the first one that used a quasi-experimental time series study design to examine the evolutionary impact of valsartan initiation on medical costs. Through the novel use of time series analysis, the study was able to identify a consistent monthly cost increase before the initiation, a temporary cost surge in the initiation month, and a consistent monthly cost decrease after the initiation month. It also identified the maximum number of post-initiation months before a cost-offset point, after which monthly medical costs with the initiation were at or lower than the monthly medical costs for the same patients as if no valsartan was initiated.
Based on our recent literature review, this kind of evolutionary impact on medical costs has never been reported for valsartan or other ARBs in Medline-indexed literature. In fact, all published and Medline-indexed studies that examined the impact of valsartan or other ARBs on medical costsCitation21 did not use a time series study design, but rather traditional pre–post, parallel comparison, or Markov simulation model approaches. Therefore, they were not able to examine the evolutionary impact of valsartan or other ARBs on medical costs.
The discovery of this evolutionary impact has several real world implications for patients, clinicians, payers, and policy-makers in selecting or reimbursing anti-hypertensive medications. First it suggests that although ARBs are more expensive than other anti-hypertensive medications, the costs increase due to the use of an ARB can be offset over the time by potential reduction in inpatient, outpatient, and non-ARB medication costs. This cost-offset evidence should be incorporated into the medication coverage policy and clinical choice of anti-hypertensive medications. Second, it suggests that the medical costs for hypertensive patients without an ARB could increase significantly over time, and exceed the medical costs for those on an ARB, because the medical costs for hypertensive patients will increase over time regardless of the use of an ARB. Third, it suggests that after the initiation of valsartan or an ARB, hypertensive patients used progressively fewer medical resources over time than those without the initiation. If we assume that these patients had the same physical and financial access to medical resources, and had the same medical care consumption behavior, reduced use of medical resources may be associated with better disease control. In other words, hypertensive patients that have initiated valsartan therapy might have better control of blood pressure and related complications than those without the initiation. This assumption is supported by the finding that incidence rates of hypertension complications were much lower in the post-initiation period than in the pre-initiation period (see ), and by previously published clinical trials and cost-effectiveness studiesCitation16,Citation20,Citation21,Citation40–65.
The results of our study also have implications for researchers. Most comparative therapeutic cost studies used classical pre–post or parallel comparative study design with one or two measures of medical costs, such as annual medical costs before a treatment initiation and annual medical costs after the initiation. By choosing only one or two measures of medical costs over a large time span, these classical studies either assume that medical costs are uniformly distributed over time within that large time span or the evolutionary impact of a disease and treatment are negligible. When these assumptions are violated, their study results are likely to be inaccurate. The medical cost time trend revealed in our study suggests that such assumptions are likely to be false in the real world, at least for chronic diseases like hypertension. Further, because the studies with the classical study designs cannot compare the same person (in parallel studies) at the same time (in pre–post studies), they have to adjust various confounding factors and selection biases through using different complicated methods to mimic counterfactuality required by causality estimation. With a reasonable assumption that the trend of medical costs before a treatment initiation will continue if the treatment was not initiated, time series study design can mimic the counterfactuality better than the classical study designs because it not only parsimoniously compares the medical costs of the same persons without the need of complicated adjustment for confounding factors and selection biases, but also recognizes the evolutionary impact of a disease and treatment better than the classical study design. This was the major reason why we chose a time series study design and time series analysis approach for our study.
The medical cost time series examined in this study had a significant cost surge in the initiation month. To explain this cost surge ($209.8), we hypothesized that the cost surge was associated with more intensive consumption of medical resources for new clinical conditions. To verify this hypothesis, we examined and compared monthly incidence rates of hypertension-related co-morbid conditions as well as monthly non-valsartan medication costs between the initiation month and other months. The results suggested a similar surge of monthly incidence rates of hypertension complications, congestive heart failure, ischemic heart disease, dyslipidemia, and diabetes, and of the monthly cost of medications other than valsartan. For example, the monthly incidence rate of hypertension complications was surged from 1.39 in the last pre-initiation month to 1.73 in the initiation month; and the monthly costs of medications other than valsartan were surged from $96 in the last pre-initiation month to $139 in the initiation month.
Our study has several limitations. First, our data source was from a geographically diverse commercial population covered by large employer-sponsored private health insurance programs, therefore may not be representative of other populations with different patients’ characteristics. In particular, generalization of the study results to the elderly, children and youth, Medicare and Medicaid populations cannot be assumed. Second, all medical conditions identified were based on ICD-9CM diagnosis codes recorded in a healthcare claims database, and were not validated with medical chart reviews. Medical conditions not recorded in the claims database were not recognized in our study. Third, only direct medical costs were examined. Indirect costs incurred by patients including productivity loss, disability claims, and opportunity costs associated with work-time lost while seeking treatment were not assessed. Fifth, because our data source did not capture over-the-counter medications, the actual medication costs may be under-estimated in our study. Sixth, our evolutionary cost analysis assumed that the cohort’s medical costs will continue its pre-initiation trend if patients’ anti-hypertensive treatment remained the same. Therefore, the study results may not be generalizable to a cohort of hypertensive patients who could not remain on the same anti-hypertensive treatment. Seventh, our analysis did not examine the impact of valsartan initiation among individual sub-groups of hypertensive patients. Future research in this area is needed. Eighth, if other anti-hypertension clinical intervention occurred parallel to valsartan initiation, our study results may contain the impact of these parallel interventions. Future research with a parallel control group of patients who did not initiate valsartan may be helpful to confirm the study findings. Finally, our time series analysis revealed the evolutionary impact of valsartan initiation on medical costs, but we did not examine the driving forces behind the evolutionary impact. In other words, we did not verify whether the post-initiation reduction in the medical costs and the incidence rates of the co-morbid conditions are purely from the valsartan initiation or from other medical treatments (though we found rather flat post-initiation cost trends for non-hypertension-related medications, hypertension-related medications, and non-valsartan medications individually). Further research in this direction using real world data will be fruitful and can provide further useful information for patients, clinicians, and policy-maker to improve the quality and efficiency of hypertension care.
Conclusions
Commercially insured hypertensive patients experienced increasing monthly medical costs prior to the valsartan initiation. The evolutionary impact of the valsartan initiation on these patients’ medical costs consisted of a temporary cost surge in the initiation month, and a gradual, consistent, and statistically significant cost decrease in the rest of post-initiation months. After the 4th post-initiation month, the actual medical costs for patients with valsartan initiation might be equal to or less than the projected medical costs as if patients did not initiate valsartan. Future research to examine the driving forces of the cost changes will be fruitful.
Transparency
Declaration of funding:
This study was supported by a research grant from Novartis Pharmaceutical Corporation.
Declaration of financial/other relationship:
The first author is employed by Kailo Research Group. The rest of the authors are employed by Novartis Pharmaceutical Corporation, and own Novartis stock.
Acknowledgments
The authors acknowledge help and comments from the journal editors and from article reviewers. No other assistance in preparation of this article is to be declared.
References
- Alcocer L, Cueto L. Hypertension, a health economics perspective. Ther Adv Cardiovasc Dis 2008;2:147–55
- Boivin JM, Zannad F. [Official recommendations and guidelines for the management and reimbursement of severe hypertension as a chronic disease]. Presse Med Apr 2009;38:652–60
- McManus F, Freel EM, Connell JM. Hypertension. Scott Med J 2007;52:36–42; quiz 42, 54
- Mendis S, Lindholm LH, Mancia G, et al. World Health Organization (WHO) and International Society of Hypertension (ISH) risk prediction charts: assessment of cardiovascular risk for prevention and control of cardiovascular disease in low and middle-income countries. J Hypertens 2007;25:1578–82
- Zuccala A. [Guidelines for management of hypertension: why doesn't evidence lead to unanimity?]. G Ital Nefrol 2009;26:226–35
- Volpe M, Tocci G. 2007 ESH/ESC Guidelines for the management of hypertension, from theory to practice: global cardiovascular risk concept. J Hypertens 2009;27(3 Suppl):3–11
- Clark CE, Campbell JL. Hypertension guidelines. Br J Gen Pract 2009;59:448–9; author reply 449
- Segura J, Ruilope LM. New guidelines of the European Society of Hypertension. Curr Hypertens Rep 2008;10:337–8
- Kaplan NM. Guidelines for the management of hypertension. Can J Cardiol 2000;16:1147–52
- Donner-Banzhoff N, Echterhoff HH, Hense HW, et al. [Guidelines Clearing House Statement “Hypertension”. Summary and recommendations for a rational hypertension guideline in Germany]. Z Arztl Fortbild Qualitatssich 2000;94:341–9
- Ferdinand KC, Saunders E. Hypertension-related morbidity and mortality in African Americans–why we need to do better. J Clin Hypertens (Greenwich) 2006;8(1 Suppl):21–30
- Bardage C, Isacson DG. Self-reported side-effects of antihypertensive drugs: an epidemiological study on prevalence and impact on health-state utility. Blood Press 2000;9:328–34
- Cheung BM, Wong YL, Lau CP. Queen Mary Utilization of Antihypertensive Drugs Study: side-effects of antihypertensive drugs. J Clin Pharm Ther 2005;30:391–9
- Cordoba Garcia R. [Side effects of antihypertensive drugs]. Aten Primaria 1996;17:420–4
- Juergens JL. Side effects of antihypertensive drugs. Proc Staff Meet Mayo Clin 1958;33:317–22
- Coca A. Economic benefits of treating high-risk hypertension with angiotensin II receptor antagonists (blockers). Clin Drug Investig 2008;28:211–20
- Dina R, Jafari M. Angiotensin II-receptor antagonists: an overview. Am J Health Syst Pharm 2000;57:1231–41
- Holwerda NJ, Fogari R, Angeli P, et al. Valsartan, a new angiotensin II antagonist for the treatment of essential hypertension: efficacy and safety compared with placebo and enalapril. J Hypertens 1996;14:1147–51
- McInnes GT. Clinical advantage of valsartan. Cardiology 1999;91(1 Suppl):14–18
- Muller-Nordhorn J, Willich SN. Angiotensin II antagonists in the treatment of hypertension-effective and efficient? Herz 2003;28:733–7
- Theodoratou D, Maniadakis N, Fragoulakis V, et al. Analysis of published economic evaluations of angiotensin receptor blockers. Hellenic J Cardiol 2009;50:105–18
- Marketwire. Novartis reports higher sales and earnings in first quarter of 2008 showing healthcare portfolio well on the way to new growth cycle. 2008; Available at: http://www.marketwire.com/press-release/Novartis-International-Ag-845918.html. Accessed May 21, 2010
- Campbell D, Stanley J. Experimental and quasi-experimental designs for research. Boston, MA: Houghton Mifflin Company, 1966
- Wilcox RR, Keselman HJ. Modern robust data analysis methods: measures of central tendency. Psychol Methods 2003;8:254–74
- Brufsky JW, Ross-Degnan D, Calabrese D, et al. Shifting physician prescribing to a preferred histamine-2-receptor antagonist. Effects of a multifactorial intervention in a mixed-model health maintenance organization. Med Care 1998;36:321–32
- Adams AS, Zhang F, LeCates RF, et al. Prior authorization for antidepressants in Medicaid: effects among disabled dual enrollees. Arch Intern Med 2009;169:750–6
- Jacobson Vann JC, Christofferson S, Humble CG, et al. Pharmacist and physician satisfaction and rates of switching to preferred medications associated with an instant prior authorization program for proton pump inhibitors in the North Carolina Medicaid program. J Manag Care Pharm 2010;16:250–63
- Martin BC, Chisholm MA, Kotzan JA. Isolating the cost of osteoporosis-related fracture for postmenopausal women. A population-based study. Gerontology 2001;47:21–9
- Siriwardena AN, Fairchild P, Gibson S, et al. Investigation of the effect of a countywide protected learning time scheme on prescribing rates of ramipril: interrupted time series study. Fam Pract 2007;24:26–33
- Law MR, Lu CY, Soumerai SB, et al. Impact of two Medicaid prior-authorization policies on antihypertensive use and costs among Michigan and Indiana residents dually enrolled in Medicaid and Medicare: results of a longitudinal, population-based study. Clin Ther 2010;32:729–41; discussion 716
- Hartung DM, Touchette DR, Ketchum KL, et al. Effects of a prior-authorization policy for celecoxib on medical service and prescription drug use in a managed care Medicaid population. Clin Ther 2004;26:1518–32
- Zhang Y. Cost-saving effects of olanzapine as long-term treatment for bipolar disorder. J Ment Health Policy Econ 2008;11:135–46
- Banerjee A, Dolado J, Galbraith J, et al. Cointegration, error correction, and the econometric analysis of non-stationary data. Oxford: Oxford University Press, 1993
- Brockwell PJ, Davis RA. Time series: theory and methods, 2nd edn. New York: Springer, 2009
- Said S, Dickey D. Testing for unit roots in autoregressive-moving average models of unknown Order. Biometrika 1984;71(3):599–607
- Chakravarti I, Laha R, Roy J. Handbook of methods of applied statistics. Vol. 1: John Wiley and Sons, New York, 1967
- Ljung G, Box G. On a measure of a lack of fit in time series models. Biometrika 1978;65(2):297–303
- Levene H. Robust tests for equality of variances. In: Olkin I, al. HHe, eds. Contributions to probability and statistics: essays in honor of Harold Hotelling. Standford University Press, 1960. p 278–292
- Bozdogan H. Model selection and Akaike’s Information Criterion (AIC): the general theory and its analytical extensions. Psychometrika 1987;52:345–70
- Bramlage P, Durand-Zaleski I, Desai N, et al. The value of irbesartan in the management of hypertension. Expert Opin Pharmacother 2009;10:1817–31
- Smith DG, Nguyen AB, Peak CN, et al. Markov modeling analysis of health and economic outcomes of therapy with valsartan versus amlodipine in patients with type 2 diabetes and microalbuminuria. J Manag Care Pharm 2004;10:26–32
- Taylor M, Scuffham PA, Chaplin S, et al. An economic evaluation of valsartan for post-MI patients in the UK who are not suitable for treatment with ACE inhibitors. Value Health 2009;12(4)
- Coyle D, Rodby RA. Economic evaluation of the use of irbesartan and amlodipine in the treatment of diabetic nephropathy in patients with hypertension in Canada. Can J Cardiol 2004;20:71–9
- Szucs TD, Sandoz MS, Keusch GW. The cost-effectiveness of losartan in type 2 diabetics with nephropathy in Switzerland–an analysis of the RENAAL study. Swiss Med Wkly 2004;134:440–7
- Ekman M, Bienfait-Beuzon C, Jackson J. Cost-effectiveness of irbesartan/hydrochlorothiazide in patients with hypertension: an economic evaluation for Sweden. J Hum Hypertens 2008;22:845–55
- Saito I, Fujikawa K, Saruta T. Cost-effectiveness analysis: controlled-release nifedipine and valsartan combination therapy in patients with essential hypertension: the adalat CR and valsartan cost-effectiveness combination (ADVANCE-Combi) study. Hypertens Res 2008;31:1399–405
- Saito I, Saruta T. Controlled release nifedipine and valsartan combination therapy in patients with essential hypertension: the adalat CR and valsartan cost-effectiveness combination (ADVANCE-combi) study. Hypertens Res 2006;29:789–96
- Anderson AN, Wessels F, Moodley I, et al. AT1 receptor blockers–cost-effectiveness within the South African context. S Afr Med J 2000;90:494–8
- Gellad WF, Huskamp HA, Phillips KA, et al. Angiotensin receptor blockers on the formularies of Medicare drug plans. J Gen Intern Med 2007;22:1172–5
- Belsey JD. Switch strategies in the management of hypertension: a cost minimisation analysis of angiotensin receptor blocker based regimen. Curr Med Res Opin 2008;24:581–9
- Dollar A, Brown C, Putnam D, et al. A retrospective electronic chart review of blood pressure changes in elderly patients treated with amlodipine or an angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker. Clin Ther 2002;24:930–41
- Cheng JW. Nebivolol: a third-generation beta-blocker for hypertension. Clin Ther 2009;31:447–62
- Usher-Smith J, Ramsbottom T, Pearmain H, et al. Evaluation of the clinical outcomes of switching patients from atorvastatin to simvastatin and losartan to candesartan in a primary care setting: 2 years on. Int J Clin Pract 2008;62:480–4
- Signorovitch J, Zhang J, Wu EQ, et al. Economic impact of switching from valsartan to other angiotensin receptor blockers in patients with hypertension. Curr Med Res Opin 2010;26:849–60
- Maniadakis N, Ekman M, Fragoulakis V, et al. Economic evaluation of irbesartan in combination with hydrochlorothiazide in the treatment of hypertension in Greece. Eur J Health Econ 2010;12(3):253–61
- Miller LA, Wade R, Dai D, et al. Economic evaluation of four angiotensin II receptor blockers in the treatment of hypertension. Curr Med Res Opin 2010;26:1307–20
- Garcia AJ, Martos F, Martin A, et al. [Do generic drugs help control expenditure on hypertension? Apropos of a case]. Gac Sanit 2004;18:137–44
- Boersma C, Voors AA, Visser ST, et al. Cost effectiveness of angiotensin receptor blocker monotherapy in patients with hypertension in the Netherlands: a comparative analysis using clinical trial and drug utilization data. Am J Cardiovasc Drugs 2010;10:49–54
- Simons WR. Comparative cost effectiveness of angiotensin II receptor blockers in a US managed care setting: olmesartan medoxomil compared with losartan, valsartan, and irbesartan. Pharmacoeconomics 2003;21:61–74
- Bramlage P, Hasford J. Blood pressure reduction, persistence and costs in the evaluation of antihypertensive drug treatment–a review. Cardiovasc Diabetol 2009;8:18
- Nixon RM, Muller E, Lowy A, et al. Valsartan vs. other angiotensin II receptor blockers in the treatment of hypertension: a meta-analytical approach. Int J Clin Pract 2009;63:766–75
- Tzemos N, Lim PO, MacDonald TM. Valsartan improves endothelial dysfunction in hypertension: a randomized, double-blind study. Cardiovasc Ther 2009;27:151–8
- Kosch M, Levers A, Lang D, et al. A randomized, double-blind study of valsartan versus metoprolol on arterial distensibility and endothelial function in essential hypertension. Nephrol Dial Transplant 2008;23:2280–5
- Okura T, Watanabe S, Kurata M, et al. Long-term effects of angiotensin II receptor blockade with valsartan on carotid arterial stiffness and hemodynamic alterations in patients with essential hypertension. Clin Exp Hypertens 2008;30:415–22
- Georgescu EF, Ionescu R, Niculescu M, et al. Angiotensin-receptor blockers as therapy for mild-to-moderate hypertension-associated non-alcoholic steatohepatitis. World J Gastroenterol 2009;15:942–54
Appendix: List of ICD-9-CM codes of hypertension and its co-morbid conditions.