Doctor Shopping Behavior for Zolpidem Among Insomnia Patients in Taiwan: A Nationwide Population-Based Study

Abstract

Objectives:

Although zolpidem is listed as a controlled drug in Taiwan, patients' behavior has not been restricted and has led to the problem of doctor shopping behavior (DSB), leading to overutilization of medical resources and excess spending. The National Health Insurance Administration in Taiwan has instituted a new policy to regulate physicians' prescribing behavior and decrease DSB. This retrospective study aimed to analyze the DSB for zolpidem by insomnia patients and assess related factors.

Design and Participants:

Data were extracted from the Longitudinal Health Insurance Database in Taiwan. Individuals with a diagnosis of insomnia who received more than one prescription of zolpidem in 2008 were followed for 24 mo. Doctor shopping was defined as ≥ 2 prescriptions by different doctors within ≥ 1 day overlapping in the duration of therapy. The percentage of zolpidem obtained through doctor shopping was used as an indicator of the DSB of each patient.

Results:

Among the 6,947 insomnia patients who were prescribed zolpidem, 1,652 exhibited DSB (23.78%). The average dose of zolpidem dispensed for each patient during 24 mo was 244.21 daily defined doses. The doctor shopping indicator (DSI) was 0.20 (standard deviation, 0.23) among patients with DSB. Younger age, chronic diseases, high number of diseases, higher premium status, high socioeconomic status, and fewer people served per practicing physicians were all factors significantly related to doctor shopping behavior.

Conclusion:

Doctor shopping for zolpidem appears to be an important issue in Taiwan. Implementing a proper referral system with efficient data exchange by physician or pharmacist-led medication reconciliation process might reduce DSB.

Citation:

Lu TH, Lee YY, Lee HC, Lin YM. Doctor shopping behavior for zolpidem among insomnia patients in Taiwan: a nationwide population-based study. SLEEP 2015;38(7):1039–1044.

INTRODUCTION

Insomnia is the most common sleep problem. It is not life threatening but has a huge effect on quality of life; patients with insomnia require treatment (pharmacological or nonpharmacological) to function normally. A review study reported that 33% of the general population has experienced insomnia symptoms.¹ An estimated 9% to 15% of the population have had insomnia symptoms that affected their daytime lives; 8% to 18% indicated that they have sleep dissatisfaction; and 6% could receive an insomnia diagnosis as defined by the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV).¹ According to national telephone surveys conducted by the Taiwan Society of Sleep Medicine, the portion of the Taiwanese population with insomnia was approximately 11.5% in 2006, 21.8% in 2009, and 19.3% in 2013.² With the increase in persons complaining of insomnia, the quantity of medication prescribed for treatment also increased, with the largest increase in the use of zolpidem.³ The Taiwanese government reported that 136 million tablets and 150 million tablets of zolpidem were dispensed in 2007 and 2010, respectively. At the same time, 20 to 40 million tablets of zolpidem tablets were manufactured without being dispensed annually.⁴ There appears to be a need to strengthen the control and management of zolpidem.

Zolpidem is a schedule IV controlled drug in Taiwan. The Regulations of the National Health Insurance Payment System restrict the frequency and the amount of nonbenzodiazepine sleeping pills that physicians can prescribe. Although physicians' prescribing behaviors are limited, no restrictions have been established for patients regarding the frequency with which they seek medical care or the sites where they receive that care. Patients can choose different hospitals or clinics for the same symptoms at the same time, which may lead to doctor shopping behavior (DSB) and repeated acquisition of zolpidem. The drug may then be used for personal consumption or sold on the street market.

Studies about DSB have been instituted to investigate how patients choose doctors. A doctor shopper is defined as a person who has changed doctors without referral during the same episode of illness.⁵ This may result in waste of medical manpower,⁶ diversion of prescription drugs into illicit uses,⁷ and excess medical costs.⁸ Surveys have found that 23% to 40% of patients had DSB.^5,9,10 A study in 2004 found that 6.3% of Taiwanese patients with upper respiratory infection had DSB.¹¹ Pradel et al.^12,13 used the doctor shopping indicator (DSI) to assess the relative abuse potential of different benzodiazepines. DSI is defined as the percentage of each drug obtained through doctor shopping among the total reimbursed quantity.

Most studies have investigated DSB with a focus on specific disease states or patient populations. No overall evaluation of DSB in patients using zolpidem is available. This retrospective study aims to analyze doctor shopping for zolpidem by insomnia outpatients and assess the related factors in DSB.

METHODS

Data Source

The National Health Insurance (NHI) program was implemented in Taiwan March 1, 1995; it covered more than 99% of residents by the end of 2004.¹⁴ With implementation of NHI, insured people can receive nearly universal medical services in any health care institution that has been contracted to the Administration of NHI (ANHI). The contracted medical care institutions must submit claim documents for each month's medical expenses to ANHI in order to receive reimbursement. The National Health Research Institute (NHRI) maintains a large computerized database derived from NHI program for research purposes, called the National Health Insurance Research Database (NHIRD). Data in NHIRD include detailed medical records of insured people, such as age, sex, diagnosis, medical institutions visited, physicians visited, prescriptions, dates of prescriptions, and other details of ambulatory visits and inpatient care.

The current study used the Longitudinal Health Insurance Database (LHID) 2005 released by the NHRI. LHID 2005 contains the claim data and registration files for exactly 1 million beneficiaries randomly sampled from NHIRD in 2005. Every year, NHRI releases data for one million beneficiaries from that year. All data files are deidentified by scrambling the identification codes of patients, medical staff, and medical institution. The NHRI reported that there was no statistically significant difference in sex distribution between the randomly sampled beneficiaries in the LHID 2005 and all beneficiaries under the NHI program.¹⁵ This study was approved by the Taipei Medical University Joint Institutional Review Board.

Patient Population

This retrospective study used ambulatory care expenditures from the one million patients of LHID 2005 to identify insomnia patients with zolpidem therapy started in 2008. The first prescription date of zolpidem was defined as the index date. Diagnosis of insomnia was identified by using the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) codes 307.42, 327.0, 327.00, 327.02, 327.01, 327.15, 307.41, 327.09, and 780.52. All patients with these ICD-9-CM codes in diagnoses for ambulatory visits in 2008 met the criterion to be included in the study. Other inclusion criteria were initiation of zolpidem therapy in 2008 and receiving more than one prescription for zolpidem in 2008. Exclusion criteria were incomplete data and interruption of insurance coverage because of death or emigration. All subjects were followed for 24 mo after the index date.

Study Variables

The quantity of zolpidem was evaluated by converting the total milligrams obtained by the patients into the defined daily dose (DDD). The DDD for a drug is the assumed average maintenance dose per day for a drug used for its main indication in adults. The DDD for zolpidem is defined as 10 mg by the World Health Organization Collaborating Centre for Drug Statistics Methodology.¹⁶ For example, a patient takes zolpidem 6.25 mg per day for 14 days. Then the dose consumed by this patient is calculated by taking 6.25 mg/day times 14 days divided by 10 mg/DDD equal to 8.75 DDD during these 14 days.

To investigate factors related to DSB, the following variables were included: age, sex, presence of chronic diseases, numbers of diseases, hospital admissions, premium status, socioeconomic status, and whether the patient had copayment exemptions. The premium status was categorized into three groups according to the premium paid: 0–19,999 New Taiwan Dollars (NTD); 20,000–39,999 NTD; and ≥ 40,000 NTD. Socioeconomic status was evaluated by occupation and whether a person was economically independent. For example, students and people with no regular occupation would be designated as low socioeconomic status by the LHID. The numbers of diseases were calculated by using the ICD-9-CM diagnosis codes.¹⁷ The number of people served per practicing physician of the habitual residence released by National Statistics of Taiwan¹⁸ was used to evaluate accessibility to health care.

Doctor Shopping Behavior

Doctor shopping is patient behavior defined as concomitant consultation of several physicians with regard to the same episode of illness. This allows patients to obtain quantities of medication greater than actual therapeutic needs.¹⁹ In this study, DSB was defined according to ANHI as ≥ 2 prescriptions of zolpidem by different doctors with ≥ 1 overlapping day in the duration of therapy.^20,21 For example, if a prescription of zolpidem was written by one doctor on January 7 for 14 days, and another prescription of zolpidem was written by another doctor on January 19 for the same patient, the patient was considered to have DSB. In Taiwan, because patients usually obtain medications from the pharmacy in the same hospital on the day the prescription is written, the dispensing date is usually the same as the date on which the prescription is written.

The definition of DSI is the percentage of zolpidem quantity (in DDD) obtained through DSB divided by the total reimbursed quantity (in DDD) of zolpidem during the study period by each patient. DSI was used as an indicator to access the intensity of shopping behavior for each patient.

Statistical Analysis

All data were managed and analyzed using SAS version 9.3.²² The χ² test was used to investigate the relationship between study variables and DSB. Logistic regression analysis was used to detect if the patients with particular characteristics had DSB. Because the logarithm of DSI was found to be normally distributed, multivariate regression analysis was used to explore the influences between the factors and logarithm of DSI. Statistical significance was set at P < 0.05.

RESULTS

Of a total of 7,211 patients who received zolpidem and had diagnostic codes of insomnia in 2008, 264 met exclusion criteria and were not included in the study (Figure 1). Of the remainder, 5,295 patients (76.22%) did not have DSB, and 1,652 patients (23.78%) did display DSB. The predictors of DSB and DSI were identical; the only exception was that higher premium status and high socioeconomic status predicted only DSB.

Figure 1.

Flow chart of patient recruitment. DSB, doctor shopping behavior.

Patient demographics and characteristics are listed in Table 1. Over half of patients (61.08%) who used zolpidem were age 30 to 59 y; 63.52% of them were female. Most patients (76.77%) had chronic diseases; the average number of diseases was 2.58. Fewer than 50% of the patients had admissions to hospitals during the study period. With regard to insurance, nearly 47% were at 0–19,999 NTD for premium status, and 40% patients were at 20,000–39,999 NTD. Most patients (72.78%) were considered to have high socioeconomic status, and most (71.18%) had no copayment exemptions. The average number of people served per practicing physician was 594.90.

Table 1.

Patient demographics and characteristics.

During the 24-mo follow-up period, the average total dose of zolpidem obtained by patients without DSB was 160.63 DDD; the average for patients with DSB was 736.49 DDD (Table 2). It is obvious that the patients with DSB used much more zolpidem than those without DSB; however, only 244.21 DDD of zolpidem were obtained by DSB; therefore, the DSI was 20%.

Table 2.

Total quantity of zolpidem obtained per person during the study period (24 mo).

Using logistic regression analysis, younger age, chronic diseases, high number of diseases, higher premium status, high socioeconomic status, and fewer people served per practicing physician were all factors significantly related to DSB (Table 3). Younger age, chronic diseases, higher number of diseases, and fewer people served per practicing physician all led to higher DSI (Table 4).

Table 3.

Logistic regression analysis of factors associated with doctor shopping behavior.

Table 4.

Regression analysis of factors associated with doctor shopping indicator.

DISCUSSION

The current study used LHID provided by NHRI to investigate the health care- seeking behavior for 24 mo by Taiwanese insomnia patients who first started to use zolpidem in 2008. We identified 7,211 zolpidem users in the current study, which is drastically different from another study that reported 77,036 zolpidem users between January 2005 and December 2007 in Taiwan.²³ This difference is because the current analysis was limited to first-time users who started in 2008. The patients in the current study were also required to have a concurrent insomnia diagnosis in the diagnoses that were registered for the patient in LHID. These factors resulted in a much smaller patient population than the other study.

We found that 23.78% of the patients who used zolpidem and were coded for insomnia among their diagnoses had DSB. A survey study in 1974 investigated the residents of Salt Lake City (USA). Among a total of 750 families, 37% had DSB.⁵ Surveys in Hong Kong in 1989 and Japan in 1990 found 40% and 23% of patients, respectively, had DSB.^9,10 A Taiwanese study in 2002 of outpatient services found 23.5% of patients exhibited doctor-switching behavior.²⁴ Another Taiwanese study in 2004 investigated health care usage by patients with upper respiratory tract infections and found 6.3% of them exhibited doctor-switching behavior.¹¹ From the aforementioned study results, it can be concluded that the ratio of DSB may be different because of different countries, different study methods, different patient populations, or different drugs.

The average quantity of zolpidem used in the 24 mo of study period was much higher in patients with DSB than patients without DSB (736.49 DDD versus 160.63 DDD). Overuse by patients with DSB has been reported: total zolpidem used by one patient was 14,015 DDD, which was 19.2 times the recommended daily dose; the amount obtained by DSB from the same patient was 10,928 DDD, which was15.0 times the recommended daily dose. The average DSI was 20% in the current study, which means that one-fifth of zolpidem was obtained by DSB in patients with DSB. Patients with DSB may use zolpidem more regularly. There is a concern that patients with DSB might build up tolerance for zolpidem as a result of chronic use, and in turn require higher amounts than patients without DSB. Therefore, it is recommended that zolpidem be prescribed in the smallest effective dose for the shortest amount of time.

This study found that the older the patients were, the lower their chances of having DSB and the lower their DSI. This is consistent with a Hong Kong study, which indicated that young people are more prone to feel dissatisfied with the medical process, and they also have more vigor to change physicians.⁹ It is also consistent with a study from France that found that individuals who showed more DSB and pharmacy shopping behavior were younger.²⁵ Another potential problem was found with few patients younger than 18 y who were prescribed zolpidem and exhibited DSB. However, zolpidem is only recommended for use in patients older than 18 y, according to the package insert. The safety and efficacy of zolpidem has not been established in children. We cannot be sure what happened with these cases by LHID data only.

In the current study, patient with chronic disease and those with more diseases had higher chances of exhibiting DSB; their DSIs were also higher. Previous research has indicated that people with poor health status tend to have DSB,^9,10,26 which is consistent with the results of current study. NHI regulations allow physicians to write refillable prescriptions for patients with chronic illnesses. The allowance of the refillable prescriptions is usually for 90 days at most. They can usually be refilled two more times, each time for 28- to 30-day supply of medications. Therefore, if zolpidem was prescribed with refillable prescriptions, patients could more easily obtain a large quantity of medication. Patients who have more diseases have higher chances of going to different hospitals seeking help from doctors with different specialties. If the patient complains about insomnia to one doctor who does not know of medications obtained from other hospitals or clinics, he/she could prescribe zolpidem again while the patient is already obtaining zolpidem from other institutions. The drug is paid for by the NHI, and the patient does not need to pay out of pocket. This may contribute to the problem of repeated prescriptions of zolpidem.

Patients with high socioeconomic status were more likely to have DSB in the current study. Other studies have indicated that socioeconomic status affects health care-seeking behavior. People with higher socioeconomic status may have more money to spend on health care and travels related to health care, and are more willing to spend their resources on the cost of health care.²⁷ Another study found that 48% of families with high income and 37% of families with low income had DSB.⁴ The authors speculated that families with high income were more prone to feel dissatisfied with their doctors. DSI can be treated as an indicator for the trend of drug abuse.^12,13 The results of this study showed that patients with high socioeconomic status had lower DSI, which may be an indication of lower tendency of drug abuse.

However, the more people served per practicing physician in a residential area, the less the chance of DSB and the lower the DSI. In other words, if people have more access to medical care, the probability of DSB and the DSI will both increase. One study has suggested that people in the health care system act increasingly like consumers of other goods and services: the more convenient it is for people to seek health care and the more choices people have for their health care, the more likely they are to have DSB.²⁶

This study is an analysis of a secondary database; therefore, there are a few limitations. First, because a maximum of three diagnoses are registered for each ambulatory visit in LHID, the number of patients with insomnia may be underestimated. In Taiwan, one prescription may contain multiple drugs for multiple medical problems for one patient. A patient may have more than three medical problems when he/she visits the clinic, but the physician can only record up to three diagnoses on the prescriptions. If insomnia was not one of the three diagnoses, even if the patient was prescribed zolpidem, he/she would not be included in the study. This may result in selection bias in patient population because those with multiple comorbidities may be excluded from the study. However, we did try to find the patients with zolpidem prescriptions but no insomnia diagnosis in 2008; there were only 43 patients. The age of these patients varied widely. Because the sample size was large (6,947 patients) and the reason for insomnia not appearing diagnoses may only be by chance, it should not affect the results of the study. The fact that both chronic diseases and high number of diseases were found to be significantly related to DSB is consistent with this theory. Second, conditions of zolpidem use that cannot be included in the current study include purchasing the drug out of pocket, drug abuse, underground transaction, and other illicit transactions. Therefore, not all conditions of zolpidem uses were represented in the study. Third, because socioeconomic status and premium status were both estimated, they were imprecise measures and may not describe the actual situations. Past studies have shown that DSB is affected by the doctor-patient relationship.^5,9,10,28,29 However, the current study was based on a research database and did not investigate the doctor-patient relationships in these Taiwanese patients. Finally, if patients did not intend to go doctor shopping but physicians prescribe medications due to duplication or prescription errors, then DSB may be overestimated.

France has established a program to monitor high-dose buprenorphine prescriptions since 2004. If a patient is using twice the recommended daily dose of buprenorphine, the patient and the prescribing doctor are notified that the patient has been identified as a subject of concern. The government may refuse health insurance for patients who continue to use high doses of buprenorphine. Since the establishment of the prescription monitoring program, the doctor shopping ratio decreased from 21.7% in 2004 to 16.9% in 2005.³⁰ The National Health Insurance Bureau in Taiwan has established the “Specific Medical Information Query System of the Insured.” Physicians can access the system to look up all medications prescribed to the patient from other health care institutions when prescribing zolpidem for them. It is suggested that following the French example may be helpful in recognizing when a patient demonstrates excess use of zolpidem to help physicians prescribe appropriately. Additionally, pharmacists can also educate patients with DSB to avoid drug abuse while doing medication reconciliations.

CONCLUSION

In conclusion, doctor shopping for zolpidem appears to be an important issue in Taiwan. Implementing a proper referral system with efficient data exchange by physician or pharmacist-led medication reconciliation process might reduce DSB.

DISCLOSURE STATEMENT

This was not an industry supported study. This work was supported by a research grant from Taipei Medical University-Shuang Ho Hospital (grant no. 102SHH-HCP-08). The sponsoring organization was not involved in the study design, data analysis, or interpretation. The authors have indicated no financial conflicts of interest.