Objective: Standard pharmacotherapy for patients with gastroesophageal reflux disease (GERD) includes treatment with proton pump inhibitors (PPIs). This study examined the effect of GERD patients' compliance with PPI therapy on health-care resource utilization and costs.
Methods: This was a retrospective study of more than 25 million managed care lives in the United States from January 2000 through February 2005. Administrative claims data were obtained from the National Managed Care Benchmarks database, developed by Integrated Health Care Information Solutions. GERD-diagnosed patients who had at least two PPI dispensings were extracted and grouped into two treatment categories based on their PPI medication possession ratio (MPR): compliant (MPR > 0.8) and noncompliant. A regression-based difference-in-differences approach was used to estimate the effect of compliance on the frequency and costs of inpatient and outpatient visits and pharmacy costs. Statistical controls included health plan type, patient age, baseline use of nonsteroidal antiinflammatory drugs, and comorbidities.
Results: Of the total 41,837 patients studied, 68% were compliant. On an annual, per-patient basis, PPI compliance resulted in 0.47 fewer outpatient visits (P = 0.040), 0.03 fewer inpatient visits (P = 0.015), and 0.47 fewer hospitalization days (P = 0.001) from the pre-PPI use period, compared to noncompliance. PPI therapy increased pharmacy costs for both groups, but the total annual health-care costs were reduced for both groups. Compliant patients experienced a greater decline in total cost from the pre-PPI period compared to noncompliant patients (declines of $3261 vs. $2406 per patient per year, P = 0.012).
Conclusions: Both health-care resource use and costs were reduced after initiation of PPI therapy. Additional reductions from the pre-PPI period were further observed by compliance with PPI therapy.
Gastroesophageal reflux disease (GERD) is a common, chronic disorder. In the United States, approximately 20% of people aged 25 to 74 years reported weekly or more frequent occurrence of heartburn and/or acid regurgitation . GERD is characterized by symptoms including frequent heartburn and acid regurgitation which may lead to inflammation of the esophagus. This can lead to esophageal erosions, which are observed in approximately 28% of GERD patients . The aggregate direct cost of treating GERD makes it the most costly gastrointestinal disease in the United States. It is estimated that in 1998 GERD was associated with direct medical and drug costs in excess of $10 billion (2000 US dollars) .
The mainstay of therapy for GERD is acid suppressive therapy, with proton pump inhibitors (PPIs) being more effective in controlling gastric acid than other available agents . Adverse effects of PPIs are usually minor, infrequent, well-characterized, and manageable.
One factor that can influence both health outcomes and cost is the extent to which patients comply with their prescribed therapy. This has been confirmed by researchers in a number of disease areas, such as the use of statins for coronary heart disease and antipsychotics for schizophrenia [5,6]. In general, the effects and importance of compliance in evaluating various management strategies have been overlooked [7,8]. To the best of our knowledge, only one study has examined the effects of PPI therapy compliance in patients with GERD. Hall et al.  found that for several PPIs, compliance with therapy increased pharmacy costs by 43% (P < 0.001) and increased GERD-related total costs by 34% (P < 0.001), relative to patients who were noncompliant. The authors concluded that PPI compliance did not decrease GERD-related costs. That study limited its cost evaluation to services with a primary diagnosis of GERD, so any changes in the use of health-care services not specifically recorded with a GERD diagnosis were not captured.
The present study examined the effect of PPI therapy compliance on overall health-care resource utilization and costs in a managed care population. The costs were calculated from a private insurance payer's perspective. Better understanding this relationship may provide an opportunity to improve GERD management and reduce health-care resource utilization and costs.
This study used administrative claims data from the National Managed Care Benchmarks database, developed by Integrated Health Care Information Solutions (IHCIS). This database includes complete medical history on more than 25 million managed care lives in more than 30 health plans covering all regions of the United States. Data from all contributing plans go through a series of evaluation and reconciliation steps to minimize the presence of errors. In particular, each data field goes through a systematic evaluation and cleaning process to ensure that complete and appropriate values are recorded for each data element. Data elements include patient demographics, enrollment records, inpatient and outpatient diagnostics and procedures, and outpatient pharmacy dispensing claims. Data from January 2000 through February 2005 were extracted.
This retrospective cohort study of GERD patients compared compliant to noncompliant PPI users. PPIs included in this study were the generic and branded versions of esomeprazole, lansoprazole, omeprazole, pantoprazole, and rabeprazole. To be included in the study, patients had to be continuously enrolled in a health plan, be 18 years of age or older at the time a PPI was first dispensed, have at least two insurance claims with a GERD diagnosis to confirm the diagnosis and to avoid coding errors from patients with one claim (International Classification of Diseases, Ninth Revision [ICD-9] diagnosis code 530.1, 530.81, 787.1), have at least 90 days of eligibility before their first GERD claim (GERD index date), and be PPI users with the date of the first PPI dispensing (PPI index date) at least 30 days after the first GERD claim and 30 days before the end of eligibility. Patients had to have a continuous period of PPI therapy in which PPIs were dispensed multiple times. Specifically, a patient 1) had to have at least one PPI episode, which is defined as a continuous period of PPI therapy with no interruption of more than 56 days in supply; and 2) had to have at least two PPI dispensings within an episode, which enables measurement of patient compliance. For example, a patient who is first dispensed a 30-day pill supply and obtains a refill within 86 days of the first dispensing would satisfy these two criteria.
Compliance for each patient in the study was measured using the medication possession ratio (MPR), which is defined by this equation:
The numerator of this ratio is the total days of PPI supply prescribed, including all refills. The denominator captures the duration of time that a patient is supposed to be treated; it is calculated as the number of days between the first and last refills, plus the number of days of supply for the last refill . If a patient had more than one PPI episode, the MPR was calculated as the sum of the pill supply days in each episode, divided by the sum of the durations in each episode. A patient with an MPR > 0.8 was defined as compliant; this cutoff value is used widely in the literature [11,12].
Basic demographic information (age, sex) and health plan type (e.g., health maintenance organization [HMO], point-of-service, preferred provider organization [PPO]) were compared between the PPI noncompliant and compliant groups. Also extracted were data for use as markers of GERD severity, including cough, asthma, sleeplessness, chronic pain, dyspepsia, and dysphagia . As an indicator of general health state, the Charlson comorbidity index was also calculated . The use of nonsteroidal antiinflammatory drugs (NSAIDs) and Cox-2 inhibitors, which have been studied for their association with gastrointestinal adverse effects, was also assessed [15,16].
The average annualized numbers of outpatient visits, inpatient visits, and hospitalization days were calculated and compared between the two study groups. Multiple outpatient visits on the same day were considered as a single visit. Total health-care costs were categorized as medical (i.e., outpatient and inpatient) and pharmacy (i.e., prescription drugs), and were calculated on an average annualized cost basis. The reported costs are those incurred by third-party payers and therefore did not include costs incurred by patients (e.g., copayments and deductibles).
A regression-based difference-in-differences approach was used to estimate the effect of compliance with PPI therapy on health-care resource utilization and costs. The pre-PPI period was defined as starting with the first GERD claim until the day before the initiation of PPI therapy. The post-PPI period was defined as starting with the initiation of PPI therapy until the end of eligibility or the defined study end date (February 28, 2005), whichever occurred first. Within each patient group, the incremental effect of PPI therapy was calculated as the difference between the post- and pre-PPI periods for each outcome measure. Those incremental effects of PPI treatment were then compared between the compliant and noncompliant groups to estimate the effect of compliance.
Univariate and multivariate linear regression analyses were performed. The multivariate analyses controlled for potential confounding factors, including health plan type, PPI index year, age, sex, length of time since the first GERD diagnosis, baseline use of Cox-2 inhibitors and NSAIDs, and the Charlson comorbidity index. To assess the robustness of the results, sensitivity analyses excluding potentially endogenous covariates (i.e., health plan type, length of time since the first GERD diagnosis, baseline use of Cox-2 inhibitors and NSAIDs, and the Charlson comorbidity index) and a Variance Inflation Factor (VIF) test were performed.
All statistical analyses were performed using the SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA). To test for statistically significant differences between two groups, Student's t-test was used with continuous outcome variables following a normal distribution, the nonparametric Wilcoxon sum-rank test was used for continuous variables with strictly positive values, and the Poisson test was used for count variables. A two-sided alpha level of 0.05 was used.
Study Groups and Baseline Characteristics
Of the 41,837 patients who met the inclusion criteria, 28,321 and 13,516 were respectively identified as PPI compliant and noncompliant patients. Table 1 shows the baseline characteristics of the sample population. Compliant patients were older than noncompliant patients (mean age: 50.5 vs. 47.2 years; P < 0.001). Patients were enrolled in a variety of health plan types, though most were in HMOs and PPOs. At baseline, the compliant and noncompliant groups did not statistically differ with respect to some GERD-related symptoms, such as cough, asthma, and dysphagia, but the compliant group had a greater prevalence of sleeplessness and chronic pain. Compliant patients had a higher utilization of NSAID (6.2% vs. 5.7%, P = 0.041) and Cox-2 inhibitor drugs (5.3% vs. 3.8%, P < 0.001), and had slightly more comorbidities (Charlson comorbidity index: 0.5 vs. 0.4; P < 0.001) than the noncompliant patients.
Table 1. Baseline characteristics
|Age, mean (SD)||47.2 (12.7)||50.5 (12.7)||<0.0001|
|Female, n (%)||7,521 (55.6)||15,871 (56.0)||0.4472|
|Health plan type, n (%)|| || ||<0.0001|
| HMO||5,321 (39.4)||11,299 (39.9)|| |
| IND||749 (5.5)||2,155 (7.6)|| |
| POS||1,310 (9.7)||3,144 (11.1)|| |
| PPO||5,893 (43.6)||11,209 (39.6)|| |
| Other||243 (1.8)||514 (1.8)|| |
|GERD-related comorbidities, n (%)|| || || |
| Cough||686 (5.1)||1,515 (5.3)||0.2405|
| Asthma||1,088 (8.0)||2,322 (8.2)||0.6021|
| Sleeplessness||568 (4.2)||1,364 (4.8)||0.0052|
| Chronic pain||1,933 (14.3)||4,507 (15.9)||<0.0001|
| Dyspepsia||804 (5.9)||1,543 (5.4)||0.0376|
| Dysphagia||1,295 (9.6)||2,841 (10.0)||0.1491|
|NSAID utilization, n (%)||772 (5.7)||1,762 (6.2)||0.0409|
|COX-2 utilization, n (%)||513 (3.8)||1,496 (5.3)||<0.0001|
|Charlson comorbidity index, Mean (SD)||0.4 (0.9)||0.5 (1.1)||<0.0001|
Table 2 shows the average annualized number of visits and lengths of hospital stays of the compliant and noncompliant groups for the pre- and post-PPI periods as well as the incremental difference between the two periods for each group. For both compliant and noncompliant groups, visits and hospital stays decreased in the post-PPI period compared to the pre-PPI period. The relative decrease in outpatient visits between the pre-PPI and post-PPI periods was similar between compliant and noncompliant groups (6.59 vs. 6.65 fewer visits per year, P = 0.780). Nevertheless, the compliant group experienced significantly greater relative declines in the number of emergency room (ER) visits (0.21 vs. 0.16 fewer visits per year, P = 0.024), inpatient visits (0.12 vs. 0.08 fewer visits per year, P = 0.003), and hospital lengths of stay (0.71 vs. 0.17 fewer days per year, P < 0.001).
Table 2. Pre-PPI vs. post-PPI annualized resource utilization per patient: univariate analysis
|Pre-PPI period, mean (SD)|
| Outpatient visits per year||24.39 (22.69)||26.90 (25.56)||<0.0001|
| ER visits per year||0.68 (2.24)||0.68 (2.26)||0.2691|
| Inpatient visits per year||0.26 (1.21)||0.31 (1.38)||<0.0001|
| Hospital length of stay, days per year||1.49 (9.93)||2.09 (14.02)||<0.0001|
|Post-PPI period, mean (SD)|
| Outpatient visits per year||17.74 (19.48)||20.31 (21.88)||<0.0001|
| ER visits per year||0.52 (1.42)||0.46 (1.34)||<0.0001|
| Inpatient visits per year||0.19 (0.72)||0.20 (0.78)||0.0053|
| Hospital length of stay, days per year||1.32 (7.61)||1.38 (7.92)||<0.0001|
|Incremental from pre- to post-PPI period, mean (SD)|
| Outpatient visits per year||−6.65 (20.92)||−6.59 (23.40)||0.7801|
| ER visits per year||−0.16 (2.23)||−0.21 (2.23)||0.0238|
| Inpatient visits per year||−0.08 (1.26)||−0.12 (1.41)||0.0031|
| Hospital length of stay, days per year||−0.17 (10.84)||−0.71 (14.72)||<0.0001|
Table 3 shows similar analyses but evaluates the impacts on health-care costs. Pharmacy costs increased for both compliant and noncompliant groups from pre- to post-PPI periods, and the increase was $412 larger for compliant than for noncompliant patients ($680 increase vs. $269 increase; P < 0.001). With respect to medical costs, both patient groups experienced a decline following initiation of PPI therapy. Medical costs in the post-PPI period decreased $1124 more for the compliant patients than for noncompliant patients ($3895 decrease vs. $2771 decrease; P < 0.001). The cost decrease in the medical service components offset the cost increase in the prescription drugs, and the net effect was that total annual health-care costs decreased by $3215 for compliant patients and by $2503 for noncompliant patients; the larger cost reduction observed among compliant patients was statistically significant (P = 0.022).
Table 3. Pre-PPI vs. post-PPI annualized costs per patient ($): univariate analysis
|Pre-PPI period, mean (SD)|
| Outpatient||8,006 (14,470)||9,188 (16,299)||<0.0001|
| Inpatient||2,340 (16,176)||3,452 (23,885)||<0.0001|
| Total, medical||10,346 (25,049)||12,640 (33,719)||<0.0001|
| Pharmacy||1,653 (2,097)||2,209 (2,679)||<0.0001|
| Total, medical and pharmacy||11,999 (25,605)||14,849 (34,302)||<0.0001|
|Post-PPI period, mean (SD)|
| Outpatient||5,458 (11,665)||6,427 (16,985)||<0.0001|
| Inpatient||2,116 (12,043)||2,318 (13,543)||0.9138|
| Total, medical||7,575 (20,143)||8,745 (25,449)||<0.0001|
| Pharmacy||1,922 (2,368)||2,889 (3,067)||<0.0001|
| Total, medical and pharmacy||9,497 (21,001)||11,634 (26,393)||<0.0001|
|Incremental annualized costs, mean (SD)|
| Outpatient||−2,548 (14,316)||−2,761 (18,551)||0.1976|
| Inpatient||−224 (18,410)||−1,135 (25,143)||<0.0001|
| Total, medical||−2,771 (26,286)||−3,895 (34,971)||0.0003|
| Pharmacy||269 (1,900)||680 (2,444)||<0.0001|
| Total, medical and pharmacy||−2,503 (26,589)||−3,215 (35,282)||0.0217|
Table 4 reports the incremental number of visits and costs adjusted for baseline differences between the compliant and noncompliant groups. Consistent with the univariate findings, compliant patients experienced larger decreases from pre- to post-PPI periods than noncompliant patients with respect to outpatient visits (6.76 decrease vs. 6.29 decrease, P = 0.040), inpatient visits (0.11 decrease vs. 0.08 decrease, P = 0.015), and hospital lengths of stay (0.69 decrease vs. 0.22 decrease, P = 0.001). Both patient groups experienced reductions in annual medical costs following commencement of PPI therapy, but the reduction was greater by $1204 for compliant patients ($3921 decrease vs. $2717 decrease; P < 0.001). Pharmacy costs in the post-PPI period increased for both compliant and noncompliant patients, but compliant patients had a $349 higher increase ($660 increase vs. $311 increase; P < 0.001). The net impact of PPI therapy on total annual health-care costs was a reduction of $3261 for compliant patients and $2406 for noncompliant patients; the difference is the effect of compliance, which corresponds to an additional reduction of $855 per year (P = 0.012). Adjustment for the covariates increased the estimate of the effect of compliance ($855 vs. $712) and increased the statistical significance of the estimate slightly (P-values of 0.012 vs. 0.022). Exclusion of covariates showed that the results were robust to the model specification. For instance, the impact of PPI compliance on total annual health-care costs ranged from $776 to $937 across the different specifications. Also the VIF test showed variance inflation factors of less than two for all variables, which implies that no material collinearity of covariates was present.
Table 4. Adjusted* annual incremental number of visits and costs: multivariate analysis
|Incremental annualized resource utilization, mean (SD)|
| Outpatient visits||−6.29||−6.76||−0.47||0.0396|
| ER visits||−0.15||−0.22||−0.07||0.0044|
| Inpatient visits||−0.08||−0.11||−0.03||0.0151|
| Hospital length of stay (days)||−0.22||−0.69||−0.47||0.0010|
|Incremental annualized costs ($), mean (SD)|
| Total, medical||−2717||−3921||−1204||0.0004|
| Total, medical and pharmacy||−2406||−3261||−855||0.0123|
The high prevalence of GERD and the high aggregate cost associated with this disease make the management of GERD a high priority for patients, health-care providers, and third-party payers. Significant literature exists on the pharmacology of PPIs, which have emerged as a leading choice for controlling GERD . Studies of other diseases and therapies have shown that higher compliance rates can lead to beneficial outcomes with respect to health and/or cost [7–10]. This retrospective cohort study adds to this literature by examining the relationship between PPI compliance and health-care resource utilization and cost.
Patients in this study were observed to experience reductions in health-care utilization following initiation of PPI therapy, which is consistent with existing literature on the beneficial effects of PPI therapy . This study was designed to measure the incremental impact of compliance versus noncompliance. This study has found that a sizeable fraction (32%) of GERD patients who take PPIs were noncompliant and that patients who were compliant with PPI therapy experienced a significantly greater reduction in the number of outpatient, ER, and inpatient visits. Compliant patients experienced approximately a threefold reduction in the number of hospitalization days per year (0.69 decrease vs. 0.22 decrease, P = 0.001). The inpatient utilization reduction accounted for 67% of the overall medical cost reduction associated with compliance with PPI therapies.
This study also finds that pharmacy costs increased more for compliant patients than for noncompliant patients, as expected, while medical costs declined more for compliant patients than for noncompliant patients. Because medical costs formed the bulk of total health-care costs, the net effect of PPI therapy was that total costs declined for both patient groups. Moreover, compliant patients experience a greater relative decrease in total costs than noncompliant patients. This study estimated the total annual cost-savings associated with compliance at $855 per patient. This estimate was based on a multivariate regression model that explicitly controls for a number of potentially confounding factors.
A simple calculation can put into perspective the magnitude of this compliance effect. Using the results shown in Table 3, the average GERD patient was estimated to incur $10,944 (weighted average) in total annual costs after initiation of PPI therapy. This study estimates that compliance induced a decrease of $855 in total annual costs. This implies that a noncompliant PPI user that would become compliant would experience an 8% (855/10,944) reduction in total cost. Considering the high prevalence of GERD and the substantial fraction of PPI users that are noncompliant, the aggregate cost-savings associated with greater compliance may be substantial at a system level.
These results on the cost-saving effect of PPI therapy compliance appear inconsistent with the findings of Hall et al., which reported that PPI compliance increased GERD-related total costs by 34%, compared to noncompliance . Nevertheless, that study did not take into account the possibility that PPIs may also have other benefits, such as reducing the adverse gastric side effects that may accompany use of NSAIDs . Such effects may not be recorded with a GERD diagnosis and therefore would not be considered in that study. In contrast, the present study made no limitation on the types of medical costs considered. Moreover, consideration of all-cause medical costs is appropriate because GERD is associated with a suite of symptoms (e.g., asthma, cough, sleeplessness), and to the extent that medical costs are not attributed directly to GERD, evaluation of the effects of therapy may be understated.
This study has several limitations. First, it does not evaluate the causes of noncompliance. Possible reasons include adverse side effects, patients who may be self-regulating their dosage after initially achieving control of GERD, poorly educating patients about the need for their PPI therapy, as well as inattentive patient behavior. Noncompliance may also tend to occur with patients with less severe cases of GERD. Although this study has controlled for severity in estimating the effect of compliance by including a Charlson comorbidity index in the multivariate regression models, it cannot ascertain whether this or other factors are the primary causes of noncompliance. Identifying the reasons for noncompliance would enhance the design of strategies to improve compliance.
Second, this study relies on administrative claims data. In some cases, administrative claims provide only proxies for the information ultimately sought. For example, regular dispensings that match the days of pill supply indicate compliance, but it is impossible to know from claims data that the patient actually took the medication as directed by the physician. A drug claim indicates the fact that a prescription has been filled but is not equal to patient ingestion. Therefore, the MPR method using pharmacy dispensing claims can be considered as only a surrogate marker for patient compliance. Additionally, while the multivariate analysis indirectly controlled for GERD severity through the length of time since the first GERD diagnosis and the Charlson comorbidity index (the assumption being that the more severe the GERD is, the more complications there are), such measures are only approximations for GERD severity. As a result, this study is limited in its ability to examine whether the effect of compliance varies across GERD severity level. Use of claims data, nevertheless, offers a valuable avenue for analysis, and future research should explore the use of more detailed clinical data. In addition, administrative claims data can be subject to errors in coding. Nevertheless, there is no reason to believe that these inherent limitations of the claims data have a differential impact on the compliant and noncompliant groups.
Third, not all costs have been considered in this study. IHCIS cost information includes only the portion incurred by the third-party payers; it also is standardized across health plans and therefore does not reflect the actual amounts paid by the various payers. Other cost items, such as patient copayments and deductibles, indirect productivity losses, transportation and time opportunity costs, and over-the-counter drug costs, are not included in this study. Such cost items may have positive or negative effects on the observed difference in cost reduction between the compliant and noncompliant groups.
Fourth, this study does not include a dynamic aspect of the cost-savings observed. Indeed, an interesting analysis would have been to see how the cost-savings associated with compliance might have been compounded over time; this is a topic for future research.
Fifth, although various confounding effects are controlled for to isolate the cost reductions only due to compliance, it is possible that some of the confounders used may not have captured completely the aspects intended to be controlled. Indeed, elements such as GERD severity, health insurance status, and copayments may not be perfectly captured by the corresponding variables used in the multivariate regressions. Moreover, this study does not control for additional confounding effects because of potential differences in lifestyle characteristics, which are not available in the IHCIS data.
This study finds that PPI therapy for GERD patients reduces health-care resource utilization and cost, taking into account both pharmacy and medical costs, and that this reduction is greater for compliant patients. Because a sizeable fraction of patients using PPI therapy are noncompliant, this suggests that opportunities exist to achieve reductions in health-care resource utilization and costs through more effective management of GERD and better compliance with treatment.
Source of financial support: This study was funded by AstraZeneca LP.