Changes in drug utilization following the outpatient prescription drug cost-sharing program evidence from Taiwan s elderly

Health Policy 68 (2004) 277 287 s in drug utilization following the outpatient prescription drug program evidence from Taiwan s elderly Shuen-Zen Liu a, James C. Romeis b, a Department of Accounting, College of Management, National Taiwan University, Taiwan, ROC b Health Services Research, School of Public Health, Saint Louis University, St. Louis, MO 63104-1314, USA Accepted 12 November 2003 Abstract This paper examines changes in drug utilization following Taiwan s newly implemented National Health Insurance (NHI) outpatient prescription drug program for persons over 65 years old. The study is a hospital outpatient prescription level analysis that adopts a pretest posttest control group experiment design. Selected measures of outpatient prescription drug utilization are examined for and non groups in s and pre s. Additional analyses were conducted comparing older patients with and without chronic diseases and differences for essential and non-essential drugs. Patients over age 65 were drawn from 21 hospitals in the Taipei area using a stratified random sampling method. This paper yields several interesting findings. First, average prescription cost and prescription increased for both the and non groups. However, the rate of increase was significantly less in the group when compared with the non group. Second, the elderly with non-chronic diseases were more sensitive (i.e., reducing drug utilization) to the drug program when compared with those with chronic diseases. Third, for the elderly with non-chronic diseases average drug cost per prescription experienced a smaller decrease in essential drugs but a moderate increase in non-essential drugs for the group. By contrast, for the non group, average drug cost per prescription increased sharply in non-essential drugs as well as essential drugs. Finally, there was a significant increase in the number of prescriptions as well as drug costs above the upper bound of the schedule. The outpatient drug program implemented by the NHI in Taiwan did not reverse the trend of prescription drug cost increases in hospitals. The significant increase in the number of prescriptions above the upper bound of the schedule implies that the NHI should increase the upper bound. Further analysis needs to evaluate any adverse clinical impact for older patients resulting from policy changes. 2004 Elsevier Ireland Ltd. All rights reserved. Keywords: Outpatient prescription drugs; Drug program; Drug co-payments; Taiwan s elderly 1. Background Taiwan s National Health Insurance (NHI) program was implemented in March 1995 [1] and included Corresponding author. Tel.: +1-314-977-8148; fax: +1-314-977-8150. E-mail address: romeisjc@slu.edu (J.C. Romeis). a comprehensive and very extensive drug coverage benefit for all citizens. For a variety of reasons NHI soon found the program financially unsustainable; thus, cost containment measures needed to be taken. This paper takes advantage of a rare policy analysis opportunity and examines selected effects of the outpatient drug program in Taiwan by comparing pre- with s 0168-8510/$ see front matter 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.healthpol.2003.12.013

278 S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 for older Taiwanese affected and unaffected by a cost containment policy shift. Throughout the paper our reference to prescription cost refers to or co-payment to patients or price to the government. It does not refer to cost from the manufacturer or seller s perspective. In addition, because of the descriptive nature of our analysis, we focus on changes in drug utilization following the policy shift instead of comprehensive policy effects. The definition of a prescription requires special clarification. In the US, a prescription is a doctor s order for one drug. Multiple medications for one condition may result in prescriptions for each drug. For example, if a doctor puts a patient on four drugs, there exist four prescriptions. By contrast, a Taiwan prescription usually contains orders for multiple drugs. The number of drugs in a prescription is referred as drug items. The above-mentioned US example will become one prescription with four drug items in Taiwan. Throughout the paper our reference to a prescription is based upon the Taiwan definition. 2. Drug literature The rapid growth in prescription drug spending has caught the attention of policy makers and health care researchers worldwide. Iglehart [2] indicates that since 1995 the rate of increase in Medicare drug expenditures has been approximately twice that of total health care expenditure. In 1999, total health care spending in the US reached 1.2 trillion, an increase of 5.6% over the previous year. By contrast, expenditures for prescription drugs increased by 16.6%, to 100 billion in 1999. Pharmaceutical costs, which accounted for 9.4% of personal health care expenditures in 1999, will account for 12.6% per year over the next decade, reaching 14% in 2010 [3]. In particular, Medicare beneficiaries without prescription drug coverage are placed at a further disadvantage because different purchasers pay different prices and persons without insurance pay the most. Although the elderly are a small percentage of the population, they usually consume a relatively large share of drug expenses. In the US, older adults constitute 13% of the population but account for 34% of all prescriptions dispensed and 42% of prescription drug expenses [4]. A similar situation exists in Taiwan; older adults constitute 8.6% of the population and account for 26.5% of health expenditures. As recent advances in the development of medications offer life-extending and life-enhancing benefits for the elderly, policy tradeoffs must be made between the welfare of the elderly who, with, must bear a fraction of an ever-increasing level of expenditure for prescription medications and the welfare of the rest of society that must bear the cost of paying for these services through taxes. Recent studies [4 6] have noted that 30 50% of the US Medicare beneficiaries are without reliable drug coverage and are most vulnerable to the escalating prices of prescription drugs. Prior studies [7 9] have generally found that drug co-payment or decreases drug exposure. Limited evidence is available concerning the drug utilization behavior among the elderly. Lillard et al. [10] estimated the effect of increasing the insurance coverage for prescription drugs under Medicare. The study reported an estimated increase of US$ 83 for each Medicare beneficiary, or US$ 2.6 billion (in 1990 dollars) in total prescription costs. Johnson et al. [11] provided rigorous analyses for the impact of increasing patient prescription drug for elderly Health Maintenance Organization (HMO) members. HMO is one type of managed-care plans that generally impose explicit criteria for the selection of health-care providers and provide significant financial incentives for members to use procedures associated with the plan. Because consumers pay a fixed annual capitation fee to an HMO in return for health care services, the HMO has an incentive to minimize the cost of servicing a patient in order to make higher profits. They found that moderate increases of from US$ 1 to 3, from US$ 3 to 5 per co-payment, and from 50% per dispensing to 70% per dispensing with a maximum payment per dispensing resulted in low annual capita prescription drug use and expenses. No consistent annual changes were observed in either medical care utilization or total medical expenses. In another study using the same data, Johnson et al. [12] found no consistent evidence that exposure, cost, and annual days of drug use for discretionary drug use would decrease; furthermore, they found that exposure, cost, and annual days of drug use for essential drug use would not decrease. Tamblyn et al. [13] used a random sample of 93,950 elderly and 55,333 adult welfare medication recipients in Quebec to examine the impact of introducing

S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 279 prescription drug on use of essential and non-essential drugs among elderly persons and welfare recipients and compared rates of adverse events (e.g. emergency department visits, hospitalization, mortality, etc.). They found that after was introduced, use of essential drugs decreased by 9.12% in the elderly and by 14.42% in welfare recipients; use of non-essential drugs decreased by 15.14 and 22.39% for the elderly and welfare recipients respectively. The rate (per 10,000 person-months) of serious adverse events associated with reductions in use of essential drugs increased from 5.8% in the pre-policy control cohort to 12.6% in the post-policy cohort and from 14.7 to 27.6% among welfare recipients. In addition, emergency department visit rates related to reductions in the use of essential drugs also increased by 14.2% per 10,000 person-month among the elderly and by 54.2% among welfare recipients. Schneeweiss et al. [14] found reference pricing for angiotensin-converting-enzyme inhibitors reduced the expenditure of prescription drugs and caused patients to switch to less effective medications. Little evidence suggests that patients stopped treatment for hypertension or increased health care utilization because of the program. 3. Taiwan s prescription drug program To provide better insights into the current prescription drug policy debates, empirical evidence concerning changes in drug utilization associated with Taiwan s drug coverage programs and their cost containment initiatives should be instructive for other countries. Taiwan s NHI is a compulsory single payer program that offers medical care coverage to about 97% of the population. Prior to the implementation of NHI, 40% of the population did not have access to care because of financial constraints. Those without access were mostly older adults, women and children. In the pre-nhi era, uninsured older Taiwanese paid for their care, including prescriptions, out-of-pocket. Within the first year of implementation, the access problem was virtually solved. However, the NHI experienced its first operational loss in 1998, mainly because of the rapid increase in medical expenditures, especially its outpatient costs. Furthermore, drug costs have been the most important item of outpatient costs (about 33.3%), with an annual rate of increase of about 13%. To constrain the rapid increase in drug costs, the NHI implemented an outpatient prescription drug program beginning 1 August 1999. The program imposed no charge for outpatient prescriptions (a prescription usually contains several drug items in Taiwan) costing less than US$ 3.125 (Data are converted into US$. At the time of analysis, US$ 1 equals 32 New Taiwanese Dollars, NT$s). An additional charge of US$ 0.625 would be imposed for every increase in drug costs of US$ 3.125 until the upper bound of US$ 15.625 was met. In Taiwan, the number of drugs per prescription typically is for a few days and thus requires multiple visits for a 10 30 day supply. Refill visits also require an outpatient visit charge. Selected groups and situations were exempted from the programs. These included: veterans, people with low incomes (earning less than 60% of average personal consumption level in the community), outpatient visits in remote mountain areas or off-shore islands, emergency visits, major illness/injury mandated by the NHI (about 30 items, including renal failure, malignant neoplasm, etc.), preventive care, and people with continued prescriptions for chronic diseases. The provision to exempt continued prescriptions for chronic diseases was intended to reduce unnecessary outpatient visits for refills. Patients with chronic diseases in stable conditions, under physicians discretion, can be put under this exemption provision to receive free medications for up to 30 days; the prescriptions can be refilled once. Because most patients obtain outpatient prescription drugs directly from pharmacies in hospitals after their visits, Taiwan physicians have great influence on patients drug utilization behavior and associated prescription related profits to the hospitals. NHI s information system captures hospital outpatient prescription data for all age groups and permits us to describe changes in drug utilization following the program on older Taiwanese patients. 4. Methods The paper adopts a reasonable approximation of a pretest-posttest control group experimental design. Specifically, there are four groups in the study:

280 S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 (1) group (experiment) in the drug ; (2) group prior to the drug ; (3) non group (control) in the drug ; (4) non group prior to the drug costsharing. One methodological difficulty here is that patients are not randomly assigned between the (experiment) group and non (control) group and not matched for follow-up. Instead, Taiwan s health policy dictates which patient group would be exempted from the drug program (e.g., low income, veterans, etc.). As a result, there could be systematic bias of patients between the and non groups. The implication of this will be discussed in Section 6. There is a long list of patients/illness to be included in the non group; our sampling approach reduced the observations to the following categories (ranked by the amount in prescription drug costs): emergency visits, renal failures, continued prescriptions for chronic diseases, AIDS, and preventive care. By contrast, observations included in the group consist of the following categories (ranked by the amount in total prescription drug costs): ordinary chronic diseases, other non-chronic diseases, outpatient surgeries, and home care. Our samples consist of patients aged 65 or older and were drawn using stratified random sampling from 21 hospitals selected from the Taipei area, the largest metropolitan area in the country with a population of over 2.2 million. They consist of two medical centers out of six (one public hospital), four regional hospitals out of 25 (two public hospitals), and 15 district hospitals out of 102 (seven public hospitals). The distribution of hospitals was chosen to best represent the population of hospitals in Taipei and the prescriptions analyzed represent about 15% of the total prescriptions in the sample. Our drug is from 1 August 1999 to 31 March 2000 and the corresponding of 1 August 1998 to 31 March 1999 refers to the pre. We did not use data after 31 March 2000, because the NHI reduced about 10,000 items of the drug reimbursement rates beginning 1 April 2000, the largest such action ever. Thus, including data after 31st March may have severe confounding effects, i.e., a reduction in drug costs may come from a decrease in reimbursement rates instead of the drug program. Data for the elderly patients were obtained from the Bureau of NHI Taipei office and are the most reliable and valid available for this level of analysis. Data for each group and refer to number of hospital outpatient prescriptions and associated costs for NHI. We focus on selected drug utilization measures (e.g., average prescription cost, average prescription duration, and the amount of drug prescribed, etc.) for outpatient visits instead of individual patient profiling (i.e., matched-patients) (e.g. Tamblyn et al. [14]). Given our sampling approach, focusing on patient profiling would encounter a bias in the study when interpreting why certain patients stop appearing in our dataset. They may seek outpatient treatments in local private clinics or hospitals that were not in our sample, instead of being influenced by the program. Similar to other older populations, older Taiwanese mainly suffer from chronic diseases (about 87% of our sample). We further break down the sample between chronic and non-chronic disease groups to investigate the changes in drug utilization. Note that in the non group patients with chronic diseases were subject to the influence of another program of NHI continued prescription for chronic diseases. As discussed earlier, the program aimed at reducing unnecessary outpatient visits for refills. Patients with chronic diseases in stable conditions can participate in the program to receive free medications for up to 30 days, and the prescription can be refilled once. We cannot observe health outcomes associated with the program directly. A useful approximation is to analyze the changes in essential and non-essential drugs. A decrease in the use of essential drug is generally perceived as more likely to result in adverse health outcomes. No formal essential versus non-essential drug classification is readily available in Taiwan. Thus we classify drugs in our database as essential drugs if they were in the WHO essential drug list [15]; otherwise, they are classified as non-essential drugs. In our study, local private clinics were excluded from the sample because of a lack of detailed drug utilization data; however, their share of drug costs was substantial (e.g. 36.13% of total outpatient prescription drug costs in 1998). Thus, our results should be viewed as estimating the change in drug utilization at

S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 281 the hospital level, not the overall change in outpatient prescription drug costs. 5. Results In the pre drug, 1,522,029 prescriptions based on 228,444 patients aged 65 and over were included in the analysis, and 1,604,928 prescriptions based on 241,585 patients aged 65 and over were included in the drug. Table 1 shows distribution of outpatient prescriptions across sex and age groups in our sample. The non group was much smaller in terms of the number of prescriptions (3.76% in the pre, 4.43% in the ) compared to the group (96.24% in the pre, 95.57% in the ) and expected because of NHI policy regarding exemptions. In addition, the sample had more prescriptions from male patients (e.g., 62.75 61.54% in the group) than female patients (e.g., 33.49 34.03% in the group). Patients were about equally distributed among the 65 84 age groups, with fewer prescriptions from patients aged over 85 (e.g., 3.57 4.15% in the group). Finally, more than 60% of prescriptions came from medical centers, showing older patients preference for medical centers and their more intensive use of medical services. Table 2 summarizes changes in drug utilization following the program. There were increases in total drug costs, total number of prescriptions, total number of patients, and average drug duration for both the group and non group. However, the increase rates were much smaller in the group when compared with the non group. Consistent with prior literature, the average drug items per prescription decreased for both the group and the non group. In Table 3 (Panel A), there were significant increases in drug costs, the number of prescriptions, and the number of patients (103.92, 96.47 and 62.20%, respectively) in the non group. The dramatic increases resulted from continued prescriptions for chronic diseases. However, we do not observe significant increases in drug utilization for patients in non group when compared with the regular chronic disease group with co-payments concerning average drug cost per prescription, average prescription duration, and average drug cost per day. In Panel B, drug costs experienced a decrease by 5.93% in the group but an increase by 24.97% in the non group. The number of total prescriptions and patients decreased in the group ( 9.21 and 6.25%, respectively) but increased in the non group (6.2 and 5.78%, respectively). Similarly, average prescrip- Table 1 The distribution of outpatient prescriptions Total sample group Non group (% of total) (% of total) in number of prescriptions (%) (% of total) (% of total) Total prescriptions 96.24 95.57 4.72 3.76 4.43 24.07 Sex Male 62.75 61.54 3.41 2.55 2.91 20.57 Female 33.49 34.03 7.16 1.21 1.51 31.43 Age 65 69 28.97 25.89 5.74 0.99 1.07 14.64 70 74 32.99 33.01 5.52 1.21 1.42 23.37 75 84 30.71 32.52 11.66 1.31 1.62 30.50 Over 85 3.57 4.15 22.44 0.25 0.31 31.17 Note: A Taiwan prescription usually contains orders for multiple drugs. The number of drugs used is referred as drug items. in number of prescription (%)

282 S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 Table 2 s in drug utilization following the program group Non group Total drug costs (US$) 42,772,969 48,273,813 12.86 NA 1,542,438 2,335,594 51.42 NA Total number of prescriptions 1,464,772 1,533,887 4.72 NA 57,257 71,041 24.07 NA Total number of patients 190,822 198,970 4.27 NA 37,622 42,615 13.27 NA Average drug cost per 29.20 31.47 7.78 0.0001 26.94 32.88 22.05 0.0001 prescription (US$) Average drug item per 4.20 4.15 1.19 0.0001 4.92 4.82 2.03 0.0001 prescription Average prescription 19.39 19.70 1.60 0.0001 7.66 10.92 42.56 0.0001 duration Average drug cost per day (US$) 1.58 1.67 5.70 0.0001 9.70 9.37 3.40 0.2456 Note: (1) NA denotes that t-statistic is not available. (2) All numbers shown are in US$; US$ 1 equals 32 New Taiwanese Dollars (NT$s). (3) A Taiwan prescription usually contains orders for multiple drugs. The number of drugs used is referred as drug items. tion duration decreased in the group for non-chronic disease patients ( 17.2%) compared to an increase in the non group for the same measure (4.09%). In addition, the average prescription cost for the elderly with non-chronic diseases increased only 3.56% if they were in the group; those in the non group experienced a sharp increase rate of 17.68%. Table 3 s in drug utilization between chronic and non-chronic diseases group Non group Panel A: Chronic disease group Total drug costs (US$) 41,160,406 46,756,938 13.60 NA 516,719 1,053,719 103.92 NA Total number of prescriptions 1,215,178 1,307,289 7.58 NA 11,337 22,274 96.47 NA Total number of patients 169,200 178,217 5.33 NA 5,621 9,117 62.20 NA Average drug cost per prescription 33.87 35.77 5.61 0.0001 45.58 47.31 3.80 0.0001 (US$) Average drug item per prescription 4.35 4.25 2.30 0.0001 4.23 4.07 3.78 0.0001 Average prescription duration (in 22.15 22.27 0.54 0.0001 29.76 29.82 0.20 0.0001 days) Average drug cost per day (US$) 1.58 1.66 5.06 0.0001 1.54 1.59 3.25 0.0005 Panel B: Non-chronic disease group Total drug costs (US$) 1,612,563 1,516,875 5.93 NA 1,025,719 1,281,875 24.97 NA Total number of prescriptions 249,594 226,598 9.21 NA 45,920 48,767 6.20 NA Total number of patients 85,398 80,063 6.25 NA 32,571 34,454 5.78 NA Average drug cost per prescription 6.46 6.69 3.56 0.0001 22.34 26.29 17.68 0.0001 (US$) Average drug item per prescription 3.46 3.52 1.73 0.0001 5.09 5.16 1.38 0.0011 Average prescription duration (in 5.93 4.91 17.20 0.0001 2.20 2.29 4.09 0.0001 days) Average drug cost per day (US$) 1.61 1.76 9.32 0.0001 11.72 12.88 9.90 0.0057 Note: (1) NA denotes that t-statistic is not available. (2) All numbers shown are in US$; US$ 1 equals 32 New Taiwanese Dollars (NT$). (3) A Taiwan prescription usually contains orders for multiple drugs. The number of drugs used is referred as drug items.

S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 283 Table 4 s in utilization between essential and non-essential drugs Essential drugs Non-essential drugs group Average drug cost per prescription 8.38 8.23 1.79 0.0001 24.31 26.98 10.98 0.0001 (US$) Average drug item per prescription 1.99 1.94 2.51 0.0001 2.91 2.90 0.34 0.0001 Average prescription duration 20.97 21.22 1.18 0.0001 20.40 20.96 2.75 0.0001 Average drug costs per day (US$) 0.45 0.43 4.44 0.0001 1.32 1.44 9.09 0.0001 Non group Average drug cost per prescription 8.49 9.37 10.37 0.0001 22.58 28.49 26.17 0.0001 (US$) Average drug item per prescription 3.31 3.25 1.81 0.0053 3.66 3.58 2.19 0.0001 Average prescription duration 5.60 7.12 27.14 0.0001 6.78 9.53 40.56 0.0001 Average drug cost per day (US$) 3.24 3.31 2.16 0.6209 7.90 7.47 5.44 0.1653 Note: A Taiwan prescription usually contains orders for multiple drugs. The number of drugs used is referred as drug items. In Table 4, we observe a minor increase in average prescription cost for the group (1.79%) in terms of essential drugs while a much higher increase rate for non-essential drugs (10.98%). For the non group, average prescription cost increased sharply for both essential drugs (10.37%) and non-essential drugs (26.17%). Concerning prescription duration, there was a large increase rate for non-essential drugs in the non group (40.56%) compared to essential drugs (27.14%). A further investigation (Panel A in Table 5) reveals that for elderly persons with chronic diseases average drug cost per prescription decreased for essential drugs ( 3.55 and 4.13%) but increases in non-essential drugs (8.87 and 6.73%) in both the and non groups. By contrast, for patients with non-chronic diseases (Panel B in Table 5) average drug cost per prescription experienced a smaller decrease ( 1.49%) in essential drugs but a moderate increase in non-essential drug (5.44%) for the group. For the non group, average drug cost per prescription increased sharply in non-essential drugs (21.47%) as well as essential drugs (9.59%). The program also had an impact on the distribution of outpatient prescriptions and drug costs (Table 6). For the group (Panel A), there was a minor increase (from 9.11 up to 9.12%) in the proportion of prescription below US$ 3.125 (i.e., the lower bound of co-payments), and a much higher increase (from 54.9 to 56.4%) above US$ 15.625 (i.e., the upper bound of co-payments). By contrast, for the non group (Panel B) there was a minor decrease in the proportion of prescription below US$ 3.125 (from 30.27 to 26.26%), and an even higher rate of increase above US$ 15.625 (from 32.89 to 41.83%). That is, for the group we observe the distribution of average prescription cost shifts towards the lower bound (physicians seemed to help save patients out-of-pocket costs) and upper bound (no further financial burden for patients beyond the upper bound) of the schedule. For the non group, the distribution clearly shifts toward prescriptions with higher costs. We found prescriptions with costs above US$ 15.625 constituted 88.99% of outpatient drug costs in the group before co-payments (Panel B); the percentage went up to 90.14% despite the program, reflecting an increase in the total number of high cost prescriptions. Drug costs in the non group had similar pattern. The percentage of drug costs above the upper bound went up from 88.06% before the program to 91.46% in the program, mainly because

284 S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 Table 5 s in essential and non-essential drugs between chronic vs. non-chronic diseases Essential drugs Non-essential drugs % Panel A: Chronic diseases group Average drug cost per 9.58 9.24 3.55 0.0001 28.06 30.55 8.87 0.0001 prescription (US$) Average drug item per 2.02 1.96 2.97 0.0001 3.00 2.97 0.01 0.0001 prescription Average prescription duration 23.64 23.70 0.25 0.0001 22.85 23.26 1.79 0.0001 Average drug cost per day (US$) 0.43 0.42 2.32 0.0001 1.32 1.43 8.33 0.0001 Non group Average drug cost per 12.82 12.29 4.13 0.0028 37.88 40.43 6.73 0.0001 prescription (US$) Average drug item per 2.13 2.00 6.10 0.0001 2.72 2.72 0 0.5250 prescription Average prescription duration 29.82 29.89 0.23 0.0001 29.81 29.85 1.34 0.0001 Average drug cost per day (US$) 0.43 0.41 4.65 0.0007 1.28 1.36 6.25 0.0001 Panel B: Non-chronic diseases group Average drug cost per 2.02 1.99 1.49 0.0518 5.51 5.81 5.44 0.0001 prescription (US$) Average drug item per 1.80 1.80 0 0.9179 2.46 2.49 1.22 0.0001 prescription Average prescription duration 5.09 4.46 12.38 0.0001 5.42 4.73 12.73 0.0001 Average drug cost per day (US$) 0.54 0.55 1.85 0.3960 1.34 1.50 11.94 0.0001 Non group Average drug cost per 7.51 8.23 9.59 0.0020 18.49 22.46 21.47 0.0001 prescription (US$) Average drug item per 3.58 3.74 4.47 0.0001 3.91 4.01 2.56 0.0004 prescription Average prescription duration 2.33 2.36 1.29 0.0001 2.51 2.57 2.39 0.0001 Average drug cost per day (US$) 3.88 4.44 14.43 0.0007 9.68 10.55 8.99 0.0403 Note: (1) NA denotes that t-statistic is not available. (2) All numbers shown are in US$; US$ 1 equals 32 New Taiwanese Dollars (NT$s). (3) A Taiwan prescription usually contains orders for multiple drugs. The number of drugs used is referred as drug items. of the effect of continued prescription for chronic diseases. 6. Discussion Because Taipei has more medical centers than other parts of the country, our sample may exaggerate the impact of medical centers in terms of the overall effect of NHI s program. Taiwan s health policy dictates which patient group would be exempted from the drug program. Thus, systematic bias of patients between the and non groups may exist. For example, in the non-chronic disease group the average drug cost for patients subject to co-payments (US$

S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 285 Table 6 The distribution of outpatient prescriptions and drug costs Drug cost per prescription % of number of prescriptions % of drug costs Panel A: group Under US$ 3.125 9.11 9.12 0.57 0.53 US$ 3.125 15.625 35.99 34.48 10.44 9.33 Over US$ 15.625 54.90 56.40 88.99 90.14 Panel B: Non group Under US$ 3.125 30.27 26.26 1.86 1.30 US$ 3.125 15.625 36.84 31.91 10.08 7.24 Over US$ 15.625 32.89 41.83 88.06 91.46 Note: A Taiwan prescription usually contains orders for multiple drugs. The number of drugs used is referred as drug items. 1.61 1.76) is much lower than that for patients not subject to co-payments (US$ 11.72 12.88; see Panel B, Table 3). Prior research on policy effect of drug utilization was based upon patient profiling [14]. Our sampling method drew patients from selected hospitals; the main purpose was to reduce database administrative difficulties on the part of NHI during the study. Given the approach, patient profiling is not appropriate for our study. It would incorrectly count a decrease in drug utilization if patients switched to local private clinics or hospitals outside our samples. The selected measures used in our study were intended to capture the average drug utilization behavior of older patients associated with the program. The outpatient drug program implemented by the NHI in Taiwan did not reverse the trend of drug cost increases at the hospital level; the total outpatient prescription costs increased by 13.54% in our sample. However, including the fees collected from the program (covered 57.04% of the drug cost increase), the net increase rate of outpatient prescription drug costs was 5.82% and thus somewhat helped the financing for the NHI program. Our results indicated that a higher increase rate of total drug costs in the non group mainly came from changes in volume (i.e., increases in the number of prescriptions and patients as well as prescription duration) instead of higher intensity of drug uses. The average drug cost per day in the non group (US$ 9.37 9.7 for all patients; US$ 11.72 12.88 for patients in the non-chronic disease group) was much higher than that in the group (US$ 1.58 1.67 for all patients; US$ 1.61 1.76 for patients in the non-chronic disease group). Based upon analyses on drug utilization associated with different causes in the non group (detailed statistics are available upon request from the authors), we found that the result was mainly driven by more intensive use of prescription drugs in emergency visits and renal failures that were exempted from co-payments. We found several differential results of the drug program. For example, elderly Taiwanese with non-chronic diseases were more sensitive to the drug program (i.e. more likely to reduce drug utilization) when compared with those patients with chronic diseases. As elderly patients mainly suffer from chronic diseases, our results suggest that the drug programs may not be a very effective way for cost control unless significant out-of-pocket costs are imposed. In addition, costs of non-essential drugs tend to have a much higher increase rate when compared with that of essential drugs, especially when they were not constrained by the program. For patients with chronic diseases, we found a general tendency of substituting essential drugs for non-essential drugs, regardless whether the drug program existed or not. For patients with non-chronic diseases, our results indicated that the drug program generally reduced the utilization of essential drugs instead of non-essential drugs. The significant increase in the number of prescriptions above the upper bound of schedule implies that the NHI should increase the upper bound. In our analysis, the most significant increase in

286 S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 drug costs comes from prescriptions over US$ 78.125. Thus, revising the upper bound beyond US$ 78.125 can be an option for policy-makers to address the cost control issue more effectively. However, further analysis needs to assess the program s adverse impact on medical service accessibility for the elderly. Our results may also have implications for the debates of Medicare prescription drug policy or similar concerns in other countries. Several recent influential US proposals for prescription drug reform include stop-loss coverage (Fuchs, et al., 2000); the provision would cover all costs for covered drugs after an enrollee s total prescription drug expenses reaches a certain amount. Our analysis concerning the change in the distribution of prescriptions suggests that the upper bound of stop-loss may trigger a further surge in prescription drug costs, especially for non-essential drugs. In the last few years, a significant change occurred in the cost structure concerning outpatient drug costs between hospitals and local clinics. Specifically, in medical centers their share of outpatient drug costs increased from 24.62% in 1998 to 29.07% in 2000; in metropolitan hospitals their share of outpatient drug costs increased from 19.41% in 1998 to 21.16% in 2000. By contrast, local community hospitals incurred 17.79% of outpatient drug costs in 1998; the percentage decreased to 15.81% in 2000. Local clinics used to incur 36.13% of outpatient drug costs in 1998; the percentage was down to 30.68% in 2000 [16]. Although the statistics do not provide detailed breakdown among age groups, we suspect that the elderly are increasingly seeking for outpatient drug prescriptions in health institutions of higher accreditation status under the NHI. The behavior may contribute to the rapid increase in the number of more expensive outpatient drug prescriptions as shown in our analysis. 7. Conclusion and limitations Taiwan s outpatient drug program did not reverse the trend of prescription drug cost increases in hospitals. However, it still achieved moderate success from a cost-control perspective. The rate of increase in net drug costs (including co-payments collected from patients) in the elderly was reduced when compared with the rates in prior years. However, overall effects of the program require further analyses on how the increased financial burdens affected accessibility of health services. We suspect that elderly patients who did not qualify for co-payment exemptions (e.g., not in the low income group or not under continued prescriptions for chronic diseases) but need close and frequent physician supervision for drugs uses may be adversely affected most. This study has two major limitations. First, this is a descriptive study so causal conclusion cannot be made. Multivariate analyses are needed to further examine effect of drug co-payments on utilization. Second, the study focuses on the change in drug utilization for the elderly in Taipei hospitals. Because outpatient drug costs in small private clinics and hospitals outside Taipei city were excluded in our analysis, the study cannot directly address the overall impact of the program in the nation. Acknowledgements This research was supported by grants from the Supervisory Committee of the National Health Insurance and National Science Council (90-2416-H-002-012) in Taiwan. We deeply appreciate assistance from Taipei Office of the Bureau of National Health Insurance for providing research data. Excellent research assistance provided by Mei-Ling Huang is also acknowledged. References [1] Chiang TL. Taiwan s 1995 health care reform. Health Policy 1997;39:225 39. [2] Iglehart JK. Medicare and prescription drugs. New England Journal of Medicine 2001;344(13):1010 5. [3] Health Care Financing Administration, National Health Accounts, 1999. [4] Families USA, Cost overdose: Growth in drug spending for the elderly, 1992 2010, p. 2000. [5] Kreling D, Mott D, Wiederholt JB. scription drug trends: a chartbook. The Kaiser Family Foundation; 2000. [6] Moon M, Storeygard M. Targeting Medicare Drug Benefits: Costs and issues. The Urban Institute; 2000. [7] Leibowitz A, Manning WG, Newhouse JP. The demand for prescription drugs as a function of. Social Science & Medicine 1985;21(10):1063 9. [8] Harris BL, Stergachis A, Ried LD. The effect of drug co-payments on utilization and cost of pharmaceuticals in a health maintenance organization. Medical Care 1990;28(10):907 17.

S.-Z. Liu, J.C. Romeis / Health Policy 68 (2004) 277 287 287 [9] Smith DG. The effects of copayments and generic substitution on the use and costs of prescription drugs. Inquiry 1993;30:189 98. [10] Lillard LA, Rogowski J, Kington R. Insurance coverage for prescription drugs. Medical Care 1999;37(9):926 36. [11] Johnson RE, Goodman MJ, Hornbrook MC, Michael B, Eldredge MB. The impact of increasing patient prescription drug cost sharing on therapeutic classed of drugs received and on the health status of elderly HMO members. Health Services Research 1997;32(1):103 22. [12] Johnson RE, Goodman MJ, Hornbrook MC, Michael B, Eldredge MB. The effect of increased prescription drug costsharing on medical care utilization and expenses of elderly Health Maintenance Organization members. Medical Care 1997;35(11):1119 31. [13] Tamblyn R, Laprise R, Hanley J, Abraahaamowiez M, Scott S, Mayo N, et al. Adverse events associated with prescription drug among poor and elderly persons. JAMA 2001;285(4):421 9. [14] Schneeweiss S, Walker AM, Glynn RJ, Maclure M, Dormuth C, Soumerai SB. Outcomes of reference pricing for angiotensin-converting-inhibitors. New England Journal of Medicine 2002;346:822 9. [15] WHO, Essential drugs, Drug Information 1999;13(4):249 62. [16] ROC Bureau of National Health Insurance, National Health Insurance Annual Statistical Report, 2001.