Rita L. Hui, PharmD, MS

In the setting of changing temporal trends in the management of osteoporosis, we examined how select characteristics of new oral bisphosphonate (BP) initiators changed over time among 94,073 women within a large, integrated healthcare organization during the period 2004 to 2012. In the earlier era (2004-2007), approximately half of women younger than 65 years initiating BP therapy (47%-54%) had osteoporosis by bone mineral density (BMD) criteria, but this proportion increased sharply in the later era (2008-2012), with 55% to 81% having osteoporosis. This trend was not evident in older women (≥65 years). The proportion of younger women with prior fracture increased from 15% in 2008 to 32% in 2012, after remaining relatively stable (10%-15%) during the earlier era. Again, this trend was not observed among older women. Thus, among women younger than 65 years, we observed a marked temporal shift in initiation of BP treatment toward women at high risk (including those with prior fracture and those with osteoporosis by BMD testing) and away from those at lower risk (such as those with osteopenia and/or no prior fracture). 

Examination of osteoporosis treatment patterns within healthcare systems should consider the effect of temporal trends in disease management. In Kaiser Permanente Southern California (KPSC), a large-scale osteoporosis program was established in 2002 to reduce fracture rates by 1) increasing bone mineral density (BMD) screening in women 65 years or older and those older than 50 years with fracture and 2) increasing bisphosphonate (BP) treatment for those with osteoporosis by BMD criteria and those with fracture.

 In 2008, osteoporosis metrics, quality measures, and changes in disease management were promulgated nationally, focusing on women 65 years or older and those at higher fracture risk.

In this context, we examined characteristics of new oral BP initiators over time, including age, prior fracture status, and BMD findings. Of the 94,073 women 50 years or older who initiated oral BP within KPSC from 2004 to 2012, 65% were 65 years or older and 21% had a fracture diagnosis within the 5 years prior to BP initiation. There were 75% with accessible BMD data within 2 years of BP initiation, of whom 62% had osteoporosis based on the lowest T-score (≤—2.5) of the hip or spine. 

Examining time trends among women who initiated BP and had available BMD data, we found that during 2004 to 2007, approximately half of the women younger than 65 years had osteoporosis by BMD criteria (47%-54%); the remaining 43% to 50% had osteopenia and 2% to 3% had normal BMD. In contrast, during 2008 to 2012, the proportion of women younger than 65 years with osteoporosis by BMD increased from 55% to 81% (

). This temporal trend was not evident among women 65 years or older who initiated BP, of whom about two-thirds had osteoporosis by BMD in both the earlier (62%-66%) and later (59%-70%) time periods.

Initiation of BP among younger women (<65 years) was notable for a progressively increasing proportion with prior fracture, from 15% in 2008 to 32% in 2012 (

). This focus on treatment of higher-risk women was reflected in a declining proportion of women younger than 65 years (42% in 2007 to 21% in 2012) among those without prior fracture. 

Similar observations were seen in Kaiser Permanente Northern California,

 a healthcare delivery system equal in size to KPSC but providing care independently, where a regional osteoporosis program was not implemented until 2008. For both health plans, major shifts toward treatment of older, rather than younger, women were observed beginning in 2008.

In summary, beginning in 2008 there was a marked shift toward treating higher-risk subsets among younger postmenopausal women, including women with prior fracture and those with osteoporosis by BMD, and away from women at lower risk (younger age, osteopenia, and no prior fracture). Health plan attention toward assessment of integrated fracture risk and national quality metrics disseminated in 2008

 likely contributed to these observed patterns. These findings highlight the importance of considering changing initiatives and guidelines that may impact temporal trends and patterns of BP initiation within integrated healthcare systems.

The authors wish to thank Shireen Fatemi, Nora Strick, Shelley Cao, and Kathryn Williams of the KPSC Healthy Bones program for their input and review of this letter.

 Department of Medicine, Kaiser Permanente Oakland Medical Center (RJH, JCL), Oakland, CA; Division of Research, Kaiser Permanente Northern California (JCL, BE), Oakland, CA; Department of Research & Evaluation, Kaiser Permanente Southern California (BHL, ALA), Pasadena, CA; Pharmacy Outcomes Research Group, Kaiser Permanente California (FN), Downey, CA; Pharmacy Outcomes Research Group, Kaiser Permanente California (RLH), Oakland, CA.

 These analyses were supported by a grant from the National Institute of Aging at the National Institutes of Health, 1R01AG047230.

 Dr Lo has received past research funding from Amgen and current research funding from Sanofi, unrelated to the current study. Dr Ettinger has previously received payments for providing expert testimony on the relationship between bisphosphonate use and atypical femur fracture in litigation involving Merck and Fosamax. Dr Adams has received research funding from Amgen, Merck, and Otsuka, unrelated to the current study. The remaining authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article. 

 Concept and design (RJH, JCL, RLH, ALA); acquisition of data (BHL, FN, ALA); analysis and interpretation of data (RJH, JCL, BE, BHL, FN, RLH, ALA); drafting of the manuscript (RJH, JCL, ALA); critical revision of the manuscript for important intellectual content (BE, BHL, FN, RLH); statistical analysis (BHL); provision of patients or study materials (BHL, FN, ALA); and obtaining funding (JCL).

 Annette L. Adams, PhD, MPH, Kaiser Permanente Southern California, 100 S Los Robles Ave, 2nd Fl, Pasadena, CA 91101. E-mail: Annette.L.Adams@kp.org.

1. Dell R, Greene D, Schelkun SR, Williams K. Osteoporosis disease management: the role of the orthopaedic surgeon. 

2008;90(suppl 4):188-194. doi: 10.2106/JBJS.H.00628.

Clinician’s Guide to Prevention and Treatment of Osteoporosis. 

Washington, DC: National Osteoporosis Foundation; 2008.

3. NQF-Endorsed national voluntary consensus standards for physician-focused ambulatory care: appendix A—NCQA measure technical specifications. National Committee for Quality Assurance website. ncqa.org/portals/0/hedisqm/NQF_Posting_Appendix.pdf. Published April 2008. Accessed March 21, 2017. 

4. Curtis JR, Adachi JD, Saag KG. Bridging the osteoporosis quality chasm. 

2009;24(1):3-7. doi: 10.1359/jbmr.081116.

5. Lee DR, Ettinger B, Chandra M, Hui RL, Lo JC. Changing patterns in oral bisphosphonate initiation in women between 2004 and 2012. 

2017;65(3):656-658. doi: 10.1111/jgs.14723.

Articles

The authors observed a marked shift toward treatment of higher-risk subsets of younger postmenopausal women (with prior fracture and/or with osteoporosis), and away from women at lower risk.

Trends in Bisphosphonate Initiation Within an Integrated Healthcare Delivery System

To assess the impact of a Medicare Medication Therapy Management (MTM) program in a large integrated health plan on patient mortality, hospitalization and emergency department (ED) utilization, and daily prescription costs.

Patients who received MTM services between 2006 and 2010 were matched to control patients who were enrolled in Medicare but did not receive MTM services. They were matched in a 1:4 ratio based on age, gender, geographic location, and prospective diagnostic-cost-group (DxCG) risk score. Multivariate regressions were used to analyze the outcomes. Subgroup analyses were conducted for patients enrolled in 2010 because the Centers for Medicare & Medicaid Services lowered the drug-cost threshold for MTM eligibility and changed from opt-in to optout participation.

We identified 34,532 members who received MTM services and 138,128 control members. The MTM group was found to have a significantly reduced mortality (hazard ratio 0.86, 95% confidence interval [CI], 0.84-0.88; 

<.001), lower odds for hospitalization (odds ratio [OR] = 0.97, 95% CI, 0.94-0.99; 

 = .018), higher odds for emergency department visits (OR = 1.17, 95% CI, 1.14-1.20; 

<.001), and no differences in change in daily medication costs when compared to the matched group. The subgroup analysis of the 2010 cohort found similar results with better outcomes than the overall cohort.

 Medicare MTM services resulted in lower mortality and odds for hospitalization for enrolled patients compared with matched controls. This study observed an increase in ED visits and no differences in change in daily medication costs in MTM services.

		Medicare Medication Therapy Management (MTM) programs are mandated by the Centers for Medicare & Medicaid Services to be provided by any health plans that have a Medicare Part D (MPD) drug plan.

		A pharmacy-led MTM program is useful in improving clinical outcomes in Medicare beneficiaries but it may not decrease medication costs.

Medicare Part D (MPD) is a prescription drug benefit for seniors and individuals with disabilities, which began in 2006 after enactment of the 2004 Medicare Modernization Act.

Along with the establishment of MPD, there was a mandate for the participating plans to provide medication therapy management (MTM) services to MPD beneficiaries who met specific eligibility requirements. 

The goal of the Medicare MTM program is to improve health outcomes. Although MTM services can be provided by a variety of healthcare professionals, pharmacists—with their extensive knowledge of medications&mdash;are the most common providers.

The Centers for Medicare & Medicaid Services (CMS) mandates provision of MTM services, targeted at MPD beneficiaries with multiple chronic conditions who are taking multiple Part D medications and likely to exceed a preset annual cost for covered Part D medications.

The eligibility criteria for MTM services have changed since the inception of the program for MPD beneficiaries, with the most significant updates imposed by CMS in 2010. The major changes included: (1) standardizing the minimum number of chronic conditions and Part D medications required; (2) decreasing the annual threshold for medication costs from $4000 in 2006 to $3000 in 2010; and (3) requiring all participating plans to enroll eligible Part D beneficiaries in an opt-out manner, versus the previous opt-in or opt-out option that MTM programs could offer at their discretion between 2006 and 2009. These changes resulted in a substantial increase in eligible beneficiaries for MTM programs in 2010.

Several small studies have evaluated various outcomes of MTM services in various settings.

These studies were limited in that they were either limited to specific disease states, lacked a comparison group or used patients who opted out of MTM services as the comparison group. The purpose of our study was to evaluate the impact of pharmacist-run telephone based Medicare MTM services on health-related outcomes for enrolled patients compared with a matched control group from January 2006 to December 2010 in the Kaiser Permanente (KP) California Regions. Our aim was to evaluate if there was any impact that our MTM program had on hospitalization utilization, emergency department (ED) visits, change in daily prescription costs, and mortality rate.

KP California is an integrated healthcare delivery system caring for 6.6 million members in California. Medicare beneficiaries comprise approximately 10% of the total membership, with over 90% of those members having purchased the MPD benefit. Over 40% of our MPD beneficiaries have employer-sponsored retiree benefits. KP California’s Medicare MTM program started in 2006 and was initially staffed by 16 full time (FT) clinical pharmacists and 10 FT support staff. By 2010, the number had grown to approximately 40 FT pharmacists and 22 FT support staff to meet the increased demand for services.

From 2006 to 2009, the KP California MTM program identified members that were likely to incur annual drug costs of $4000 or more, who were taking at least 2 medications covered by MPD and had at least 2 or more chronic conditions, such as diabetes, cardiovascular diseases, hypertension, hyperlipidemia, or stroke. MTM services were offered to these patients, and participation in the program was voluntary. In 2010, in addition to the changes in eligible requirements made by CMS, KP California MTM increased the number of chronic conditions from 2 to 3 and the number of required medications from 2 to 5. It was estimated that approximately 3% of MPD enrollees were eligible for the MTM program during the period from 2006 to 2009, while 5% of the MPD enrollees were eligible for MTM program during 2010.

Potential participants meeting the above criteria were identified several times per year during the study period by a computerized system and letters outlining the MTM program were mailed to the members. During 2006 to 2009, only patients who agreed to participate were contacted by the MTM staff, while in 2010, all patients who did not opt out were contacted by the MTM staff. All contacts were made by phone. The initial encounter consisted of the pharmacist or a support staff member obtaining a list of current medications, followed by the pharmacist performing a comprehensive medication review. Follow-up encounters generally focused on identifying opportunities to improve medication therapy. All encounters and interventions made by the MTM staff were documented in each patient's electronic medical record.

The MTM pharmacists worked under collaborative practice agreements with physicians from the Permanente Medical Group and Southern California Permanente Medical Group. Interventions may have included: performing or obtaining necessary assessments of the patient’s health status; initiating and adjusting doses of medications and ordering medication-associated laboratory tests; reviewing and adjusting medications to reduce the likelihood of adverse drug events, interactions, or duplication of therapy; ensuring optimal dosing; improving medication adherence; providing education to enhance each patient’s understanding and to encourage appropriate use of medications; simplifying drug therapy and reducing drug therapy costs; and ensuring therapies are in concordance with clinical practice guidelines. Any recommendations falling outside of the protocol were communicated to the primary care physician for approval. Coordination of services was performed with local population care management services, as well as ambulatory care pharmacist-managed services (eg, anticoagulation, oncology, etc).

This study was a retrospective, matched-control cohort study. The study group consisted of KP members participating in the Medicare MTM program who had a comprehensive medication review and as-needed follow-ups conducted by an MTM pharmacist. The matched-control group consisted of KP members who were Medicare beneficiaries, who may or may not have been enrolled in MPD, and were not eligible for MTM. Since patients that qualified for MTM (1) were enrolled in MPD, (2) had higher medication costs, and (3) had high disease burden, a likely control patient did not have all 3 conditions and hence may have had less spending in medications and less disease burden. Both the study and the control groups received the usual care, with scheduled visits with their primary care physician or other healthcare providers. Each study group member was matched to 4 control group members based on geographic location, age (same birth year), gender, and prospective diagnostic-cost-group (DxCG) score (within ± 0.01).

 The DxCG score predicts an individual’s total healthcare cost in the next year relative to the population mean using a model based on age, gender, diagnoses, and drug codes. This relative score is calculated at the individual level. DxCG scores for Medicare beneficiaries ranged from .05 to 15.59 in the national population. Individuals with a score of 1.0 have a relative risk that is equivalent to that of the national Medicare reference population.

The study group was enrolled in MTM during the 5-year period between January 2006 and December 2010. They were followed from their enrollment date for 365 days, or until death, or disenrollment from the health plan, whichever came first. If members were enrolled in the MTM program for multiple years, only the first year of enrollment was included in the study. The matched controls for each MTM member were assigned the same enrollment date as the MTM member and were followed in the same way as the patients enrolled in the study group. Each study and control group member was unique and was only used once in this study. All subjects must have had continuous health plan membership with drug benefits during the 12 months prior to the study period to ensure complete pre-period data.

Exclusions included subjects with a cancer diagnosis within 1 year of study entry and nursing home residents. This was because cancer patients were being followed by the specialized oncology team, which included physicians, pharmacists, nurses and social workers at KP and patients residing in the nursing homes did not obtain prescription drugs from our internal KP outpatient pharmacies.

All data were obtained from KP’s integrated electronic medical, pharmacy, and administrative databases. Other covariables included: age, gender, Charlson Comorbidity Index (CCI),

 and pre-period utilization, including hospitalization, ED visits, and daily medication costs during the 12-month period prior to enrollment.

This study was approved by Kaiser Permanente Institutional Review Boards of both Northern and Southern California regions.

The primary outcome of this study was all-cause mortality within 365 days of study enrollment. Deaths were identified using several administrative and clinical data sources within KP, including records indicating discharge status after hospitalization. Deaths that occurred outside of the health plan were identified from California vital statistics death tapes. Secondary outcomes included percentage of hospitalization and ED visits within each group, and median change in prescription cost per day for the same period. Since some patients did not have 365 days of follow-up records due to death or termination of membership, prescription cost were assessed on a cost-per-day metric. Daily medication costs were calculated by dividing total medication costs by the length of follow up time during study period.

 test, and Wilcoxon rank sum test, were used to compare baseline characteristics. In order to identify the direct impact MTM services had on the outcomes of interest, multiple logistic regressions were used to analyze proportion of patients who were hospitalized or who visited the ED; the Cox proportional hazards model was used to analyze death rate; and multiple ordinary least squares modeling was used to analyze changes in daily medication costs. Except for the Cox proportional hazards model, each model was adjusted for age, gender, region, CCI, and prior utilization of the same outcomes. All outcomes, except changes in daily medication costs, were analyzed as dichotomous variable. Changes in medication costs were analyzed as continuous variable. Test for proportional hazards assumption was assessed for the Cox proportional hazards models. Identical and separate analyses were done on study groups that were enrolled in 2010 because of the substantial changes in enrollment criteria. All analyses were performed using SAS 9.1.3 (SAS Institute, Cary, North Carolina).

During the study period from 2006 to 2010, a total of 46,734 MPD members had received an MTM comprehensive medication review. After applying the matching and exclusion criteria, 34,532 members receiving MTM services were identified and matched to 138,128 control members in a 1:4 ratio (Figure) based on age, gender, geographic location, and DxCG risk score. Over 81% of the members in both groups had a 1-year follow-up period. It represented 31,549 personyears in the study group and 124,546 person-years in the control group. Study enrollment rate was similar during 2006 to 2009, however, enrollment significantly increased in 2010, due to CMS-mandated changes, and these enrollees comprised 39% of the entire study cohort (MTM: 13,402; control: 53,608). The study population had a mean age of 75 ± 8 years, 58% were female, and the median DxCG score was 1.5 (IQR 0.8-2.5). Based on the 12-month data prior to enrollment to MTM services, the study group had a significantly higher CCI (a score that predicted the relative risk of death from prognostic clinical covariables), a higher rate of inpatient hospitalizations and ED usage, and higher daily medication costs than the matched controls (

Unadjusted observed outcomes rates are presented in 

 for members that were enrolled in 2010. No differences were found in the unadjusted all-cause mortality rate between the 2 groups (5.7% in study group vs 5.6% in matched group), while the same mortality rate was significantly lower in the MTM group for the cohort enrolled in 2010 only (4.3% vs 5.0%,

 <.001). In the pre-intervention period, the MTM group had a higher percentage of hospitalization compared with the control group (absolute between-group difference of +5.75%) and a higher percentage of ED visits (absolute difference of +9.3%) (Table 1). In the postintervention period, the MTM groups showed an absolute reduction in hospitalization of 4.1% while the control group showed an increase of 2.1% (absolute between-group differences of + 0.5%). Similar change was seen with ED visits with an absolute between-group difference of 6.2%, down from 9.3%. In the 2010 cohort, a significantly lower proportion of the MTM group was hospitalized (Table 3: 24.1% vs 24.9%), despite higher percentage of hospitalization in pre-period (27.3% vs 22.1%). We also observed a narrowing in the absolute difference in the percentage of ED visits (3.4% during study vs 7.7% before study). A difference in change of daily median prescription cost was also observed between the 2 study groups, with a median decrease of $0.39 for the MTM group and a median increase of $0.10 for the matched group (

The adjusted outcomes comparing the MTM group with the matched group are presented in 

. The adjusted proportional hazard ratio for mortality was significantly lower at 0.86 (confidence interval [CI], 0.84-0.88, 

 <.001) for the entire cohort and at 0.71 (CI, 0.68-0.75, 

 <.001) for the 2010 cohort. The adjusted logistic regression analysis shows a significantly lower odds of hospital admission in the MTM group (odds ratio [OR] = 0.97 [CI = 0.94-0.99, 

= .016] for the entire cohort and OR = 0.91 [CI 0.87-0.95, 

 <.001] for the 2010 cohort). The OR of ED visits was significantly higher in the MTM group. The change in daily medication costs for the MTM group was not different from the matched group.

In this large, retrospective, matched cohort study, we observed a 14% reduction in the risk of mortality, a 3% reduction in the risk of hospitalizations, 17% increased risk for ED visits, and no differences in change in median daily medication costs for MPD beneficiaries within 12 months after receiving MTM services, compared with a matched group of Medicare patients who did not receive MTM services. The same trend, but with a more profound magnitude, was observed in the subgroup enrolled in 2010 when the criteria for MTM services had changed. The improvement in patient outcomes may be due to a combination of services and interventions provided by MTM ambulatory care pharmacists. These interventions included optimizing medication regimens under physician-approved protocol, providing education on medication use to enhance patient understanding and appropriate use which may lead to increased adherence, ordering necessary laboratory testing, and coordination of care. Optimizing drug therapy by the pharmacist usually consisted of simplifying treatment regimens, titrating existing medication regimens to achieve therapeutic goals or reduce the likelihood of adverse reactions, initiating new medications when gaps in therapy were identified, discontinuing duplicate or unnecessary therapies, including drugs on the Beer’s List recommended to be avoided in the elderly; and switching drug therapy to avoid drug-drug interactions. These interventions may help achieve targeted clinical goals that benefit the patient’s health status.

 Lastly, participating MPD beneficiaries and plans providing MTM services can benefit economically by managing prescription drug costs through appropriate drug use, use of generics, and elimination of unnecessary medications. 

These measures may reduce total drug spending and delay the time for MPD beneficiaries to reach the coverage gap (donut hole) and incur out-of-pocket costs. Although we did not observe a significant decrease in change of daily medication costs in this study, we were able to find improvement in other outcomes without an increase in medication costs. We do not have data on the number or type of interventions outlined above during 2006 to 2009 as they were not recorded in an analyzable format. In 2010, there were a total of 51,732 interventions recorded by the MTM program. The impacts of specific types of interventions warrant future study.

The results of this study support findings reported in the available literature. In a study by Welsh et al,

a reduction in mortality risk was observed in patients enrolled in an MTM program (OR 0.5; 95% CI, 0.3-0.9). The results of our study did not show as great a decrease in mortality for the entire study population or the 2010 subgroup. This may be due to differences in study design between the 2 studies. Our study design included matching, with the intention of obtaining groups with similar cost burden using DxCG scores. Conversely, the study by Welsh et al compared patients who agreed to enroll in the MTM program with those who declined participation.

A mixed impact of the MTM programs on hospital admissions and ED visits was observed in the literature. Some studies have shown no difference or a reduction in hospital admissions, while others have shown an increase associated with MTM programs.

 Explanations for the reduction in proportion of patients being hospitalized include improved medication management, while the increase in hospitalizations may be due, in part, to a reduction in mortality. Our study demonstrated that there was a reduction in hospital admission but an increased rate of ED visits in a sicker MTM study population compared with a control group.

Similarly, studies have shown the impact of MTM on medication costs can vary due to a multitude of reasons.

 Reductions in medication costs include discontinuation of duplicate medications and those to be avoided in the elderly, and medications with an inappropriate indication. Increased medication costs may be due to the addition of medications identified as gaps in therapy, titrating of medications, and increased adherence.

This study, as far as we know, is the largest in size with 172,660 participants and a long follow-up time of 12 months to evaluate MTM services. The validity of our results was strengthened, as the size of the study population allowed us to detect very small differences in outcomes. Our study matched MTM to control patients by age, gender, location, and DxCG score. The ability to match at a 1:4 ratio also increased the power and provided greater precision in estimates and tests. We decided to use DxCG as a matching criterion because all but 1 of our outcomes was related to resource utilization. Matching non-MTM-eligible patients may avoid some of the selection bias. Several previous MTM studies analyzed populations of patients who opted in versus those who opted out of MTM participation. Patients who opted in for MTM services may have been more engaged in their healthcare, favoring results for patients enrolled in MTM. The matching in our study was not perfect, as subjects in MTM services had a higher disease burden at baseline, with greater hospitalization, ED visits, and medication costs. This may be due to the fact that eligibility for MTM required meeting a threshold annual medication cost and having multiple chronic conditions. The control group in our study were Medicare members, without 2 of the chronic conditions as stated in KP MTM criteria; or having a lowerthan- threshold annual Part D medication cost. After matching for age, gender, and DxCG, we found that our control group did not have the same disease burden in terms of prior hospitalization and ED visit rate and medications used, when compared with MTM study group. We used CCI and specific prior utilization to adjust the outcomes and conducted a difference in differences comparison in order to account for these baseline differences. It is impossible to conduct a cohort study that avoids all selection bias because MTM eligible patients have a higher disease burden by definition when compared to all non-eligible Medicare enrollees. The current study matched with a control group with less disease burden, based on baseline utilization. Hence, the primary outcome of mortality would be expected to be lower in the control group. Yet, we observed a favorable mortality rate in our MTM intervention group.

Given that KP is an integrated care system and the current study was limited to patients in California, there might be limited generalizability of the results.

The retrospective nature of this study inherently requires a level of caution when interpreting the results. Our study did not examine into what specific components or processes from the MTM program were associated with the outcomes. For example, pharmacist-led discontinuation of skeletal muscle relaxants, a class of medications to be avoided in the elderly, may have been a significant factor in the reduction in inpatient hospitalization. We also did not evaluate any surrogate markers, such as blood pressure, lipid levels, or glycated hemoglobin. Since the 2 groups were not matched on disease burden and not everyone would have these surrogate markers measured during the time period, we decided to look at outcomes that we could observe for the entire cohort. We also did not estimate the return on investment for the MTM services provided at KP California, as other studies had investigated this matter extensively.

This study is the largest to date, helping to supplement and strengthen available literature. Studies such as ours are essential to ensure that MTM services continue to provide a positive impact on health outcomes. The KP California Medicare MTM program provides targeted services that, when combined with other healthcare services, are likely to improve patient outcomes. Although the direct effect of specific interventions was not investigated, it can be noted that there is a reduction in mortality and inpatient hospitalizations when usual care is supplemented with a pharmacist-led MTM program.

Pharmacy Outcomes Research Group, Kaiser Permanente Drug Information Services (RLH, MMS, JC); Kaiser Permanente Drug Information Services, PGY-2 Pharmacy Resident in Drug Information (BDY); Kaiser Permanente Southern California Medicare Medication Therapy Management (EWJ).

 This study was funded by Kaiser Permanente Drug Information Services.

 Concept and design (RLH, BDY, MMS, JC); acquisition of data (RLH, MMS); analysis and interpretation of data (RLH, BDY, MMS, EWJ, JC); drafting of the manuscript (RLH, BDY, EWJ); critical revision of the manuscript for important intellectual content (RLH, BDY, MMS, EWJ, JC); statistical analysis (RLH, BDY); provision of study materials or patients (EWJ); administrative, technical, or logistic support (RLH).

 The authors (RLH, BDY, MMS, EWJ, JC) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Rita L. Hui, PharmD, MS, 1800 Harrison St, Ste 1301, Oakland, CA 94612. E-mail: Rita.L.Hui@kp.org1. Centers for Medicare and Medicaid Services. Prescription Drug Coverage - General Information. 2012. 10-29-2012.

2. Centers for Medicare & Medicaid Services: Medication Therapy Management. 2012. 10-29-2012.

A pharmacist-led Medicare Medication Therapy Management program can improve clinical outcomes in Medicare beneficiaries without an increase in medication costs.

Impact of a Medicare MTM Program: Evaluating Clinical and Economic Outcomes

 To determine whether health services use bypatients with selected acid-related gastrointestinal disorders (pepticulcer disease, gastroesophageal reflux disease, and gastritis or dyspepsia)is lower after initial treatment with proton pump inhibitors(PPIs) than with histamine

 Retrospective, 2-year longitudinal study.

 Among continuous enrollees fromDecember 1, 1996, to June 1, 2002, in a group model health maintenanceorganization, 13 971 members were electronically selectedwho began receiving antisecretory therapy during that periodand who had no previous drug therapy, endoscopy, or hospitalizationfor gastrointestinal disease. Adjusted medical costs andhealthcare use related to gastrointestinal disease (measured byoffice visits, endoscopy, or imaging and hospital admissions) andfactors associated with initial and subsequent drug therapy wereanalyzed using a 2-stage model. This method adjusted for unobservableconfounders, primarily drug selection bias, an inherentlimitation of retrospective database studies.

 Drug costs were more than 4-fold higher (P < .001)when PPIs rather than H

RAs were prescribed initially, but nondrugcosts and health services use showed no decrease. A historyof physicians' prescribing PPIs in the prior 12 months was associatedwith prescribing PPIs as initial therapy (odds ratio, 4.29; 95%confidence interval, 3.74-4.90) and with step-up therapy (changefrom H

RAs to PPIs). A history of physicians' prescribing H

RAs inthe prior 12 months was associated with step-down therapy(change from PPIs to H

RA therapy asinitial therapy for acid-related gastrointestinal disease produced nodecrease in nondrug costs or health services use.

Proton pump inhibitors (PPIs) are among the mostwidely prescribed drugs. Since the 1989 introductionof omeprazole, sales of PPIs haveincreased greatly. By 2000, worldwide sales of omeprazoletotaled more than $6 billion.

 In 2001, US sales ofPPIs (more than $10 billion) were second only to salesof cholesterol-reducing agents (statins).

 Heavy use ofdirect-to-consumer advertising has made PPIs one ofthe most widely promoted drug classes.

 At the sametime, the cost of managing acid-related gastrointestinaldisorders continues to escalate, driven by the primarydirect cost of drugs, especially PPIs.

 report the superiority of PPIs overhistamine

RAs) for efficacious,cost-effective treatment of acid-related gastrointestinaldisorders, especially gastroesophageal reflux disease(GERD) and peptic ulcer disease (PUD). Although thesestudies indicated that PPIs are superior to H

RAs forhealing and preventing recurrence of erosive esophagitis,the study populations lacked the variety of lesssevere and functional acid-related gastrointestinal diseasesand symptoms that primary care physicians treat.Evidence of superiority of PPIs over H

RAs is much lesscompelling in patients who have functional dyspepsia(gastroesophageal symptoms without endoscopic abnormalityor evidence of other disease known to producedyspepsia) or who have gastroesophageal symptomswithout evidence of a lesion.

 Therefore, although thedata suggest that PPI therapy may be more cost effectivethan H

RA therapy for selected patients who need moreintense acid suppression, these data are lacking forpatients who have the less serious gastrointestinal disorderstypically seen in primary care practice.

Treatment of acid-related gastrointestinal disease isnot limited to the initial selection of antisecretory drugs(ie, PPIs or H

RAs) but also includes their sequentialuse. Initial therapy with H

RAs may be changed to PPItherapy (step-up therapy) if needed, or initial therapywith PPIs may be changed to H

RA therapy (step-downtherapy) after symptoms are controlled. These strategieshave produced mixed results in modeling investigationsand in protocol-driven prospective trials.

These models are sensitive to the assumptions madeand to the drug costs used in the calculations. Moreimportant, the models do not reflect clinical practice,because the need for step-up or step-down therapy orendoscopy is predetermined. Protocol-driven prospectivestudies are further limited by the characteristicsand small size of the study populations. A recent study

showing a 58% success rate for step-down therapy after1 year included only 71 patients after rigorous exclusioncriteria were applied. Step-down therapy was studiedonly in patients who remained asymptomatic whilereceiving PPIs; this criterion excluded from analysis thesubstantial number of patients who remained symptomatic,despite PPI therapy. The study did not addressstep-up therapy.

In the present retrospective study of patients withdifferent acid-related gastrointestinal disorders (PUD,GERD, and gastritis or dyspepsia), we assessed theeffect of initial medication choice (ie, PPIs, H

RAs, orboth) and of subsequent stepwise treatment on cost anduse of healthcare in a large health maintenance organization(HMO). In particular, we examined return oninvestment by determining whether initial therapy withPPIs results in lower associated healthcare costs thaninitial therapy with H

RAs. We also assessed factorsaffecting initial and subsequent treatment strategies forthese patients.

This study was a 2-year, retrospective longitudinalanalysis of data from the administrative databases ofKaiser Permanente (KP) in northern California,

 agroup model HMO with about 3 million members. Thestudy population (n = 13 971) included all patients whomet the following criteria: age 18 years or older;received a new prescription for 1 or more antisecretorydrugs (PPIs or H

RAs) between December 1, 1997, andJune 1, 1998 (index period); received no prescriptionfor PPIs or H

RAs and had no hospital admission orendoscopic examination for gastrointestinal disease inthe 12 months before the index period; was diagnosedby a physician as having an acid-related gastrointestinaldisorder within 180 days of the index date; and wascontinuously enrolled at KP for 12 months before and24 months after the index date (ie, 36-month minimumenrollment). Data for each patient were collected for 24months after the index period. This study wasapproved by the KP Northern California InstitutionalReview Board.

International Classification ofDiseases, Ninth Revision, Clinical Modification (ICD-9-CM)

 coded entries in the KP Outpatient SummaryClinical Records, patients were assigned to 1 of3 diagnostic groups: PUD, GERD, and dyspepsiaor gastritis (

). Patient selection wasaccomplished without bias from electronicadministrative databases using an SAS (SASInstitute, Cary, NC) algorithm. Patients withdiagnoses in more than 1 category wereassigned to a single diagnostic group. Patientswhose medical record contained the diagnosticcode for PUD were assigned to this diseasegroup even if the record contained additionalcodes from different diagnostic groups; patientswere grouped in the GERD category if theirrecords contained the code for GERD and thecode for gastritis or dyspepsia but not the codefor PUD.

Data on medication use by patients and eachphysician's drug-prescribing history in the 12months preceding the index date were obtainedfrom the KP Pharmacy Information ManagementSystem; data on office visits, from theOutpatient Registration System Plus; and laboratorydata, from the Laboratory UtilizationReview System. Hospitalization data wereobtained from the Admission Discharge Transfer databaseusing selected 

 codes for diagnosis-relatedgroups 174 through 178, 182, and 183. Physicianoffice visits comprised visits to the following departmentsor divisions: medicine (including gastroenterology)and pulmonary medicine and allergy because ofpossible association of GERD with asthma), urgentcare, and emergency. Results of endoscopy and uppergastrointestinal imaging were obtained from outpatientand inpatient records using 

Drug costs were calculated on the basis of acquisitioncost. Procedure and hospital cost data were obtainedfrom the KP Cost Management Information System.

Data on patient demographics and length of enrollmentin the health plan were obtained from the KPPatient Demographics database and from the Length ofEnrollment database. Age bands contained 10-yearranges, except for patients aged 18 through 24 years andpatients older than 74 years. Descriptive statistics wereused to describe age, sex, acid-related disorder, rate ofendoscopy, hospital admissions, and acid-suppressiontherapy continuing longer than 6 months.

Patients were categorized based on diagnoses andinitial prescription. They were followed up longitudinallyfor up to 2 years after the indexdate and identified as to whetherthey refilled their initial prescription,did not continue their initial prescription,stepped down from PPIs toH

 illustrates the variousprescribing pathways.

Combination therapy was definedas concurrent initial prescription ofPPIs and H

RAs or concurrent prescriptionsof PPIs and H

RAs for 100or more days or for 50% or longer ofthe study. Because few patients startedcombination therapy, thesepatients were not included in subsequentanalyses.

Predictors of PPI initiation andsubsequent step-up or step-downtherapy were analyzed by logisticregression. Independent variablesincluded age bands, sex, diagnosticcategory (GERD, PUD, and gastritis or dyspepsia),physicians' drug-prescribing history, and current orprevious nonsteroidal anti-inflammatory drug use.

During our study, prescribing physicians at KP wereencouraged, through information in fliers or presentationsby pharmacists, to start therapy with the lessexpensive H

RAs. There were no formal clinical protocolsfor step-up or step-down implementation. However,if patients had initially received PPIs, physicians wereencouraged to step down therapy, as appropriate.

When comparing direct medical costs and use forpatients initially receiving PPIs and H

RAs, a 2-stagestatistical model was used to control for drug selectionbias. This method provided an adjustment factor(inverse Mills ratio) estimated from observable variablesto adjust for unobservable confounders that may be correlatedwith treatment selection and that may affecttotal medical cost or use.

 Correlation analysis wasused to test multicolinearity between the inverse Millsratio and other variables contained in the cost or useequations.

Independent variables included in the first stage ofanalysis were the patient's age band, sex, diagnostic category,physicians' prescribing history, and presence of20 chronic diseases (based on drug prescriptions filled).Chronic diseases included human immunodeficiencyvirus infection, cardiovascular disease, hypertension,hyperlipidemia, seizure disorders, Parkinson's disease,rheumatoid arthritis, pain disorders, asthma, diabetesmellitus, thyroid disease, gout, Crohn's disease, depression,psychosis, anxiety, mania, cancer, previous organtransplantation, and renal disease. This use of population-based pharmacy data to measure chronic diseasestatus has been described as a means of predictinghealthcare use.

Dependent variables for the second stage of analysisincluded drug costs and nondrug costs of treating acid-relatedgastrointestinal disease and frequency of physicianoffice visits, endoscopy, upper gastrointestinalimaging, or hospital admission for gastrointestinal disease.Independent or explanatory variables used in thesecond stage of analysis included the inverse Millsratio, sex, age bands, initial antisecretory drug prescription,acid-related gastrointestinal disease diagnosis,hospital admission for gastrointestinal disease,endoscopy, upper gastrointestinal imaging, current useof nonsteroidal anti-inflammatory drugs, duration ofantisecretory therapy longer than 6 months,

 eradication, visits to the departmentof medicine or gastroenterology, and presence ofchronic disease.

Severity of acid-related gastrointestinal disease wasmeasured by the frequency of endoscopy, upper gastrointestinalimaging, hospital admission for acid-relatedgastrointestinal disease, and proportion of patientswho required therapy for longer than 6 months duringthe 2-year follow-up. Depending on the outcome ofinterest, a variable could be dependent or independent,but never both in the same model. For example,endoscopy is an independent variable in the cost modelbut is a dependent variable when frequency ofendoscopy is the outcome of interest.

Return on investment was computed by comparingdrug and nondrug costs of treatment for acid-relatedgastrointestinal disease between patients who initiallyreceived a prescription for PPIs and patients who initiallyreceived a prescription for H

RAs. Drug costsincluded cost of antisecretory drugs (PPIs and H

RAs),prokinetic drugs (eg, cisapride and metoclopramidehydrochloride), other gastrointestinal drugs (eg, misoprostoland sucralfate), and drugs for 

 eradication.Drugs such as bismuth subsalicylate, tetracycline,amoxicillin, metronidazole, or clarithromycin were consideredprescribed for 

 eradication when prescribedthe same day as any of the other classes ofdrugs. Nondrug costs included costs of endoscopy, 

 serology, upper gastrointestinal imaging, and hospitaladmissions.

Frequency of endoscopy, upper gastrointestinalimaging, physician office visits, and hospital admissionfor acid-related gastrointestinal disease was also evaluated.Because we focused on overall return on investmentand because we assumed that patients withunresolved symptoms would continue to take prescribedmedication or return for diagnostic evaluation,additional outcome data were not part of this study.

Multivariate (ordinary least squares and logistic)regression analysis was used to evaluate the relationsbetween dependent and independent variables.Ordinary least squares regression was used to analyzecontinuous variables, and logistic regression analysiswas used for dichotomous variables. Stepwise proceduresused to select independent variables having a significanteffect for the model were then repeated instandard stepwise analysis. Variables were estimated byusing maximum likelihood or analysis of variance.Statistical significance was defined at 

 < .001. All confidenceintervals (CIs) were 95%, and all 

All costs were natural log transformed to reduceskewness of the data. To ensure inclusion of all subjectsin the analysis, $1 was added in the log-transformedanalysis when a cost variable was equal to zero. SAS statisticalsoftware version 8.0 was used for data analysis.

Patient demographics and characteristics are summarizedin 

. The 13 971 patients meeting thecriteria for newly started treatment had a female-maleratio of 59:41. The mean &#177; SD age of the patients was52.2 &#177; 15.9 years. Most patients with gastritis or dyspepsiawere younger, and more women than men had adiagnosis of GERD and gastritis or dyspepsia. Of thesewomen, 56.3% had GERD, 35.6% had gastritis or dyspepsia,and 8.0% had PUD. Patients with PUD were morelikely to have endoscopy or hospital admission for gastrointestinaldisease, and patients with gastritis or dyspepsiawere less likely to have treatment for longer than6 months. There was a concordance of 69% betweenpreendoscopic diagnoses and postendoscopic diagnoses.About 30% of all patients required treatment for longerthan 6 months.

 summarizes the disposition of patients alongtreatment pathways (initial and subsequent therapy) bydiagnoses and for the overall cohort. Step-up or step-downtherapy was used infrequently, for 12.6% ofpatients who initiallyreceived therapy with H

RAs and for 9.4% ofpatients who initially received therapy withPPIs. About one third (27.0% of patients whoinitially received PPI therapy and 35.6% ofpatients who initially received H

RA therapy)of patients discontinued therapy.

The adjusted odds ratios (ORs) for initiationof treatment with PPIs are shown in 

. After controlling for other variables, physicians'history of prescribing PPIs was thestrongest predictor.Physicians with a PPIprescribing ratio (PPI prescriptions/[PPI + H

RA prescriptions])greater than 50% for the past year weremore than 4 times more likely to prescribe PPIs thanH

RAs for a new patient (OR, 4.29; 95% CI, 3.74-4.90;

 < .001). Other factors associated with initial PPI prescriptionwere diagnosis of PUD (OR, 2.47; 95% CI, 2.05-2.96) or GERD (OR, 2.00; 95% CI, 1.79-2.20) comparedwith gastritis or dyspepsia, as well as use of medicationsfor rheumatoid arthritis (OR, 1.54; 95% CI, 1.33-1.78).Younger patients were less likely (and older patients,more likely) to receive PPIs initially. The type of PPIsprescribed was almost exclusively (approximately 95%)omeprazole; of prescriptions for H

RAs, 60% were writtenfor cimetidine, and 40% were written for ranitidinehydrochloride.

Physicians' prescribing history largely predicted useof step-up or step-down therapy (Table 4). The OR forprescribing step-up therapy was 3.47 (95% CI, 2.80-4.32)for physicians whose PPI prescribing ratio was greaterthan 50% for the past year. Use of step-up therapy wasassociated with patients' receiving or having any of thefollowing: endoscopy (OR, 3.80; 95% CI, 3.05-4.75),hospital admission related to gastrointestinal disease(OR, 3.54; 95% CI, 2.54-4.93), upper gastrointestinalimaging (OR, 2.78; 95% CI, 2.33-3.32), orlonger than 6 months of therapy (OR, 4.39;95% CI, 3.87-4.90). Gastroenterologistswere more likely to use the step-down strategythan physicians practicing other medicalspecialties (OR, 2.35; 95% CI, 1.42-3.89). Step-down therapy was morethan twice as likely (OR, 2.25; 95% CI, 1.44-3.53) to be prescribed by a physicianwhose H

RA + PPI prescriptions]) was greater than 50% for the past year.

No statistically significant difference infrequency of endoscopy, physician officevisits, upper gastrointestinal imaging, orhospital admission for gastrointestinal diseasewas found between patients who initiallyreceived PPI therapy and patientswho initially received H

RA therapy.During the study, 7.7% of all patients hadendoscopy, 10.4% had upper gastrointestinalimaging, and 30.5% received therapyfor longer than 6 months.

 depicts the natural log—transformedrelative cost ratio of drug and nondrugmedical costs for patients in thisstudy. Drug costs were 4.2 times higher (

< .001) for patients who initially receivedPPI therapy than for patients who initiallyreceived H

RA therapy; this cost ratioranged from 3.7 times higher for PUD to5.0 times higher for gastritis or dyspepsia.Costs of antisecretory drugs (PPIs andH

RAs) accounted for 91% of drug costs.The remaining costs were for prokineticdrugs, other gastrointestinal drugs, anddrugs for treating 

Rita L. Hui, PharmD, MS

Articles

Trends in Bisphosphonate Initiation Within an Integrated Healthcare Delivery System

Impact of a Medicare MTM Program: Evaluating Clinical and Economic Outcomes

Prescribing Proton Pump Inhibitors for Initial Treatment of Acid-related Gastrointestinal Diseases in a Managed Care Population