- Email: [email protected]
The effect of using the health smart card vs. CPOE reminder system on the prescribing practices of non-obstetric physicians during outpatient visits for pregnant women in Taiwan
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
journal homepage: www.ijmijournal.com
The effect of using the health smart card vs. CPOE reminder system on the prescribing practices of non-obstetric physicians during outpatient visits for pregnant women in Taiwan An-Jim Long a,b , Polun Chang c,∗ a b c
Division of Isotope Application, Institute of Nuclear Energy Research, Taiwan, ROC Institute of Public Health, National Yang-Ming University, Taiwan, ROC Institute of Biomedical Informatics, National Yang-Ming University, Taiwan, ROC
a r t i c l e
i n f o
a b s t r a c t
Article history:
Introduction: There is an evidence that pregnant women have been prescribed a significant
Received 28 December 2009
number of improper medications that could lead to potential damage for a developing fetus
Received in revised form
due to discontinuity of care. The safety of pregnant women raises public concern and there
18 April 2012
is a need to identify ways to prevent potential adverse events to the pregnant woman. This
Accepted 18 April 2012
study used a health smart card with a clinical reminder system to keep continuous records of general outpatient visits of pregnant women to protect them from potential adverse events caused by improper prescription.
Keywords:
Method: The health smart card, issued to all 23 million citizens in Taiwan, was used to
Health smart card
work with a Computerized Physician Order Entry (CPOE) implemented at a 700-bed teaching
Reminder system
medical center in Taipei to provide the outpatient information of pregnant women. FDA
Computerized order entry
pregnancy risk classification was used to categorize the risk of pregnant women. The log
Medication safety
file, combined with the physicians’ and patients’ profiles, were statistically examined using
Pregnancy safety
the Mantel–Haenszel technique to evaluate the impact of system in changing physician’s prescription behavior. Results: A total of 441 patients ranged in age from 15 to 50 years with 1114 prescriptions involved in FDA pregnancy risk classification C, D, and X during the study period. 144 reminders (13.1%) were accepted by physicians for further assessment and 100 (69.4%) of them were modified. Non-obstetric physicians in non-emergency setting were more intended to accept reminders (27.8%, 4.9 folds than obstetricians). Reminders triggered on patients in second trimester (15.5%) were accepted by all physicians more than third trimester (OR 1.52, p < 0.05).
∗ Corresponding author at: Room 520, Library and Information Building, No. 155, Sec. 2, Li-Nong St., Beitou District, Taipei City 11221, Taiwan, ROC. Tel.: +886 2 28267238. E-mail addresses: [email protected] (A.-J. Long), [email protected] (P. Chang). 1386-5056/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijmedinf.2012.04.009
606
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Conclusion: A health smart card armed with CPOE reminder system and well-defined criteria had the potential to decrease harmful medication prescribed to pregnant patients. The results show better conformance for non-obstetric physicians (26%) and when physicians accepted the alerts they are more likely to went back and review their orders (69%). In sum, reminder criteria of FDA pregnancy risk classification C for obstetricians and reminder based on different trimesters is suggested to be refined to improve system acceptability and to decrease improper prescription. © 2012 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
Since the thalidomide tragedy occurred in the late 1950s and early 1960s [1–4], a number of drugs given to pregnant women have been reported to act as possible teratogens As a result, the use of drugs during pregnancy has received increasing attention [4,5] since their use poses a potential risk to both the mother and the fetus. These concerns may not always be fully understood or applied in current practice, due to limited knowledge on medication safety practice during pregnancy. The first classification system based on clinical and animal data for risks associated with drug use during pregnancy was implemented in Sweden in 1978. The United States Food and Drug Administration (FDA) then introduced its own system in 1979. In the latter, medications are rated as A, B, C, D, and X relative to pregnancy, based on evidence in humans and animals, with category A drugs having been proven safe and category X drugs being contraindicated in pregnancy. Among the classifications, D and X are rated as positive evidence of human fetal risk while category C remains debatable especially a significant number of drugs are rated as category C. This system is widely used and it is believed to be the most efficient approach for establishing risk and safety relative to drug use during pregnancy [6]. It should be noted, however, that it has suggested that the FDA abandon the current classification system in favor of more meaningful evidence-based narrative statements [7]. Prior studies highlight that many of pregnant women are prescribed potentially harmful medications [8–13]. One study in France even showed 99% of the women received at least one prescription for a drug during pregnancy with a mean of 13.6 medications per woman and 1.6% of women received one or more prescriptions of drugs from X category, while 59% of women had a prescription of drugs from the D category [14–16]. Improper prescription during pregnancy is clearly a serious international health problem. Although prescribing potential harmful drugs to the pregnant women may due to a trade-off between medication damage and disease damage, it may also possible that the physicians are lack of knowledge on medication safety practice during pregnancy (especially non-obstetric physicians) that is to say an error in the planning stage of medication use and therefore is often preventable. A number of studies have shown that use of a CPOE with decision support components can reduce medication error [8–13]. To alert prescribing physicians to prevent potential medication errors requires digital identity of the pregnant patient. Hospitals do not have patients’ pregnant history such as weeks if patients were not receiving pregnant regular health check or the patients
come from emergency setting. In Taiwan the health smart card issued by government provides real-time patient eligibility and allows the government to monitor potential fraud or abuse of the medical insurance system. All pregnancy health check records are included and it provides digital identity for the pregnant patient so that incorrect reminders can be minimized when reminder system is applied. Several studies have indicated the use of health smart card can potentially improve patient safety [17–20]. But papers on the practical uses to which a smart card can be put have not been fully evaluated and described. Thus, this leaves open the question as to how to better integrate health smart cards into areas such as the clinical decision support system; such an approach has the great potential to bridge gaps in care. Our literature review found many publications indicating the importance of safe drug administration during pregnancy, but very few had practical ways to solve the problem or to evaluate the results of a solution. We created a reminder system in which a health smart card was integrated into an existing CPOE and selected pregnant women as our study population. FDA classification system was implemented to determine the proportion of potential harmful prescriptions for pregnant patients. We analyzed the order log to assess physicians’ behavior in response to computerized drug reminders to see whether the reminder system armed with health smart card based on FDA classification system was effective in the reduction of potential harmful medications for pregnant patients being prescribed.
2.
System setting and design
For our study every participating outpatient clinic had a desktop installed with computerized physician order entry system. A smart card reader was connected to the desktop which was connected securely to both internet and healthcare insurance private network. Physicians were required to start the system by an authentication process via inserting their own healthcare professional card (HPC) into the card reader. The secure authentication module (SAM) authenticated the HPC, after which the clinical session would then be started. Each patient inserts their card into the reader. We used multi-threading approach to allow the clinician to work while the system read and updated the health smart card. Multi-threading technique empowers the software to deal with multiple jobs as well as multiple devices simultaneously. It means while clinicians are waiting smart card reader to get the required data from health smart card, the software is not suspending and they can edit the patient record while the software is reading the card. It
607
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Fig. 1 – CPOE with progress bar showing the amount read from the health smart card.
should be noted that it required 1–2 s to finish acquisition of one record for the current health smart card issued by Taiwan government each record requires so for acquire data from entire card will take some minutes. The progress of reading the health smart card is shown on the top right side of the system as shown in Fig. 1. When the data read from health smart card has been finished, it triggers the rule base in the CPOE back end and generates a warning message if any. The card contains four sections of patient data as described in Table 1 which are explicit basic data section, healthcare data section, medical data section, and healthcare administration data section. Hospitals record every regular health check for pregnant women at healthcare data section with a week code range from 41 to 50. Ten times free health checks are offered by the Taiwan government. The first prenatal health check is defined at the third month and the second at the fifth month,
four weeks per check after the third check. Although we do not have record of first day of pregnancy on pregnant patients who were not regularly checked by the study hospital, we can calculate pregnancy weeks by the week code on the health smart card.
3.
Materials and methods
The proposed use of health smart card with a clinical reminder system for pregnant women was implemented at a 700-bed teaching medical center in 2006, which serves 1.3 million outpatient visits and has 6000 emergency room visits per year. The computerized physician order entry system has been in use for 8 years and 100% of the outpatient orders are prescribed electronically.
Table 1 – Data sections of health smart card and their application to patient safety. Data section Core data section Healthcare data section
Medical data section
Administration data section
Content Social security number Remarks on the cardholder’s status, catastrophic diseases; number of visits and admissions with diagnosis, utilization of the national health insurance prevention programs such as pre-pregnancy check-ups, premium records of the cardholder, accumulated medical expenditure, cost sharing, etc. Prescriptions, drug allergic history, long-term prescriptions for ambulatory care, and certain medical treatments Personal immunization chart and the willingness to allow organ donation
Application Identification of the patient Medication safety check for pregnant women, diagnosis-medication interaction check
Drug–drug interaction check, drug allergy check, dosage and duplicated medication check Duplicated immunization check
608
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
The rule for a reminder that the physician is prescribing an improper drug to a pregnant woman involves the identification of the pregnancy status of the female patient by acquiring the date of the last health check or diagnosis. Sometimes patients go to hospital to have an abortion or have been aborted. These situations can result in changes to the patient’s status or a wrong diagnosis and may generate a false positive reminder when detecting an improper prescription. Thus, target cases for reminder system were identified that met the following criteria with the aim of minimizing the triggering of false positive reminders: (1) The prescription drugs are evaluated on the assumption of a gestation period of 270 days if no end-up signal (delivery, abortion, etc.) is detected. (2) Some cases are excluded when the diagnosis meets a defined code in International Classification of Diseases 9th (ICD-9) revision and its clinical modification version (ICD-9-CM). These include a legally induced abortion (ICD9 code 635.90), a mother with single live born (ICD-9-CM V27.0-9) and various other cases of abortion (ICD-9-CM 634.0–638.9, 640.0–676.90). (3) Some prescriptions involving female reproductive hormone such as Progestin in some cases is required for pregnancy were excluded [8] since they are the most frequently dispensed category D and X drugs within the FDA pregnancy risk classification [8] to pregnant women. (4) A reminder for each improper prescription only pops-up once for each patient during a clinical session. The prescription drugs are reminded when those are listed as a category C, D and X drug as defined by the FDA pregnancy risk classification system. Although prescribing category C drugs sometimes are needed and whether reminding this remains debatable. But after discussion with pharmacists of the target hospital we concluded there is a need to inform physicians their prescriptions are potentially harmful whatever the level is minor or significant, especially when physicians are not aware of their prescriptions are in category C. System diagram of the implemented reminder system is shown in Fig. 2. The system recorded the entire physicians’ behavior when modifying orders as part of the integrated log system after a reminder had been given. The system recorded the log of prescription behavior in a single table. Every time the system triggered a reminder for improper prescription, the first response (accept or cancel), the profile of prescribing physician, the current status of the patient and the dosage, frequency and duration of the current order were saved. Thus, this research was capable of analyzing the acceptance behavior of physicians when they received a reminder and the relationship between this and a wide range of different variables. The final order was not part of the log system, but was confirmed and saved to the hospital information system. The reminder only appears to the physician after he/she has finished the order and if the order involves a category C, D and X drug from the FDA pregnancy risk classification system. If they accepts the reminder and wish to review their order, they click No to go back to order entry system and review the order. If they click Yes, then the alerted prescription has been
Fig. 2 – Diagram of the implemented system. The health smart card is issued by Taiwan government and includes the person’s name, identification number, date of birth and a photograph, as shown in the left top side of the diagram; the card is blue. On the top of the card is an integrated circuit (IC chip) containing a 32-KB Electrically Erasable Programmable Read-Only Memory (EEPROM). The red card on the right of the machine is the healthcare professional card used to authorize the system. The privacy and confidentiality of the patient information within the card is protected by a PIN number and a healthcare professional card. Without a PIN number and healthcare professional card one can never read information from the health smart card. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.) accepted. According to experience with the drug–drug interaction reminders, we found that physicians exhibited fatigue with an online survey approach and for this reasons ignore clinical reminders. Therefore we only provide a Yes or No selection. We count that as an acceptance if users click “No” button, while count as an override if users click “Yes” button (Fig. 3). The reminder system log for a 1-year period in 2006 for outpatient orders were analyzed along with the pregnancy risk
Fig. 3 – The example of prescription reminder. The dialog box is translated into English; the original is in Chinese. One patient was prescribed Fluitran 2 mg for the treatment of hypertension during pregnancy. Fluitran is classified as D category.
609
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Table 2 – A comparison of the number of reminders accepted (and following actions) and overrode stratified by physician specialty, the risk classification, and the different trimesters of pregnancy. Accepted Overrode Accept Crude OR (CI) rate Specialty 49 Obstetric Non-obstetric 68 9 ER obstetric ER non-obstetric 18 Drug category 117 C 24 D 3 X Trimester 62 Third 49 Second 33 First 144
Total ∗
626 177 83 73
7.3% 27.8% 9.8% 19.8%
Reference 4.90* (3.28–7.35) 1.38 (0.66–2.92) 3.15* (1.74–5.69)
808 136 15
12.7% 15.0% 16.7%
Reference 1.22 (0.76–1.96) 1.38 (0.39–4.84)
515 268 176
10.8% 15.5% 15.8%
Reference 1.52* (1.02–2.27) 1.55 (0.99–2.46)
959
13.1%
p-Value < 0.05.
classification, the specialty of the various physicians involved, and trimester of the pregnant women. Odds ratios (OR) and 95% confidence interval (CI), as well as p values, were calculated for each variable using the Mantel–Haenszel technique. Level of significance was considered when p value <0.05. All patient information used in this study is de-identified. At the time this study was conducted, IRB committee of the hospital was just set up and only accepts clinical trials. The chief information officer of the hospital was well informed and all information was de-identified and obtained from the information office.
4.
Results
A total of 84 physicians were involved in the study and all of them accepted at least one drug duplication reminder. Over the 1-year period, 441 patients with 1114 prescriptions ranged in age from 15 to 50 years with an average of 30.13 years (SD 5.78) were prescribed drugs in FDA pregnancy risk classification C, D, and X, i.e., prescriptions that are potentially harmful to the developing fetus. 144 reminders (13.1%) were accepted by physicians for further assessment and 100 (69.4%) of them were modified. Patients were receiving regular pregnancy health check from 215 different clinics or hospitals and asking second opinion or receiving diagnosis or treatment of other specialties to the study hospital. All visits were collected for investigation, together with the complete prescription orders and the medical history. Obstetricians prescribe majority of orders to the pregnant women (Table 2). In the other hand physicians in the emergency room (ER) receive high pressure of workload and have the least time to review patient record. We identified these properties to take into account confounder effects. According to the result, non-obstetric physicians showed a high total acceptance rate in respond to safety reminders, whether in an ER or non-ER situation (19. 8% vs. 27.8%). Obstetricians had the lowest acceptance rate in respond to the safety reminders (they accepted 44 class C reminders while they
Table 3 – A comparison of the number of reminders accepted (and following actions) and overrode stratified by physician specialty, the risk classification, and the different trimesters of pregnancy.
Specialty Obstetric Non-obstetric ER obstetric ER non-obstetric Drug category C D X Trimester Third Second First Total
Deleted
Updated
Unchanged
Total
16 (33%) 41 (60%) 1 (11%) 5 (28%)
9 (18%) 8 (12%) 8 (89%) 12 (67%)
24 (49%) 19 (28%) 0 (0%) 1 (6%)
49 68 9 18
56 (48%) 7 (29%) 0 (0%)
31 (27%) 5 (21%) 1 (33%)
30 (26%) 12 (50%) 2 (67%)
117 24 3
31 (50%) 18 (37%) 14 (42%)
12 (19%) 15 (31%) 10 (30%)
19 (31%) 16 (33%) 9 (27%)
62 49 33
63 (44%)
37 (26%)
44 (31%)
144
only accepted 14 class D and X reminders; they overrode 617 class C reminders while they overrode 92 class D and X reminders), whether in an ER or non-ER situation (9. 8% vs. 7.3%). During the research we found that there was an effect in terms of the seriousness of the risk to the pregnant women but was not statistical significant. Physicians had the slightly higher acceptance rate for X category drugs then D and then C (16. 7% vs. 15% vs. 12.7%). There were more reminders accepted during the first trimester than during other trimesters (15.8% vs. 15.5% and 10.8%). The first three months of pregnancy is the most important period for teratogenic influence [21]. This point seemed to have been taken into account by the physicians. Due to the pregnancy common wealth enrollment policy (first free prenatal health check begin with the second month), cases of first trimester are less than other trimester so the OR in responding alerts of first trimester patient is not statistical significant while second trimester has higher acceptance rate which is statistically significant than third trimester (Table 3). Once physicians accepted the reminder, CPOE system led the physicians back to review their orders and 69% of orders were updated by physicians (deleted 44% and updated 26%). 51.02% of orders made by obstetricians were updated (deleted 33% and updated 18%) after alerts accepted and reviewed, while in ER setting all orders were updated by obstetricians. Only one (6%) prescription was denied to update prescribed by non-obstetrics physicians in ER setting.
5.
Discussion
The results of this study show that the use of smart card with a CPOE reminder system has potential to decrease proportion of potential harmful medication prescribed to pregnant patient. Implementation of reminder system integrating health smart card with CPOE and well-defined criteria reduced triggering of false positives. And potential harmful prescriptions were errors in planning stage in medication use can be identified without over-reminding prescribers. In total, the overall level of acceptance at 13.1% for the reminder system was low compared to other studies [9–12].
610
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
The possible reasons are (1) the alerts were considered falsely positive (due to the use of health smart card the false positive rate is dropped and many of the factors associated to diagnosis code in [22] were never happened in the study); (2) there is usually no better candidate than the current prescribed medication; (3) there is a trade-off between medication damage and disease damage. According to analysis of the transaction log of the reminder system, the most frequent reminders were trigger by obstetricians which had the lowest acceptance rate and hence the entire acceptance rate dropped. According to some publications [23–25] reliability of drug classification systems for teratogenic risk is questioned and since category C contribute majority of reminders it is largely questioned. X category drugs are classified with evidence to be harmful to pregnant women so are considered more relevant to the clinical settings. In most cases obstetricians are more familiar with classification systems so they overrode most of the reminders if those were with less risk (especially category C prescriptions). This goes some ways to explaining why obstetricians had an acceptance rate for class C reminders of only 6.7% (44/661, as described in Section 4), while the acceptance rate for class D and X reminders was 13.2% (14/106). Therefore, we suggest that obstetricians should receive fewer reminders (by filtering some common medications in class C but are not considered harmful reviewed by obstetricians) in this area than other physicians. During this study we did not design a special reminder rule set for obstetricians, since the major point for this study is to facilitate the use of health smart card in clinical setting and evaluate the results, but this may be implemented in the future. On the other hand, the acceptance rate for class C reminders among the other physicians was 27.7% (accepted 73 reminders while overrode 191 reminders), which is high. It is also noted that non-obstetrics physicians intended to delete (60%) the alerted prescriptions after reviewed carefully. It may due to physicians are unfamiliar with the pregnancy classification system. We reviewed all the 68 reminders accepted by this group of physicians. We found many of the prescriptions can be replaced by alternative medications while some prescriptions might not be considered as harmful but was classified as C category. For example, Dextromethorphan is rated as A in Austria system and is OTC drug in US. Physicians can prescribe Acetylcysteine instead but most of prescriptions were deleted. There are some more examples such Allegra and Butyscol are in the same case. It is possible in this case the reminder system has misled other physicians by implying that some prescriptions are dangerous; in this context, according to some publications [23–25], informing risks by current classification systems for drug use during pregnancy remains in debate. This aspect therefore requires further research that will analyze the pros and cons of the pregnancy risk classification system; once this is known, it will be possible in the future to provide physicians with a better quality of information. Although category C drugs were having poor acceptance rate in overall groups and some physicians were questioning validity of the category C alerts, we still suggest to alert physicians when they prescribe category C drugs. First the category D and X were not having statistically significant acceptance rate than category C drugs. Second there were 44 category C alerts accepted by obstetricians and 26 of these were modified
(59%), while D and X category prescriptions were only accepted by 20% (one out of five). It means prescription behavior to the pregnant women can be changed even for obstetricians the ordering process can be improved potentially. The trimester of the pregnant woman is an important factor that might affect the behavior of a physician. In the first trimester the fetus grows at its fastest pace during the first trimester. So the first trimester of pregnancy is the most crucial period in pregnancy but is least visible since the fetus weights less than an ounce. During the period the fetus is most vulnerable to the harmful effects of drugs so prescriptions are much more important than during other trimesters. During the study we did show the physicians the data on the week of the pregnancy, but we did not put any warning message up showing that the patient was in first trimester. This means that most of the physicians understand the importance of medication safety in first trimester for pregnant women and also shows how the system was able to improve patient care. The hospital that took part in this study serves about 1% of all patients in Taiwan. Therefore, according to the result of this study, we estimate that there might be around 50,000 pregnant women every year receive medications that are harmful to the developing fetus, which is about 25% of all total pregnancies. Although this figure is still lower than for the US and France [8–16], important problems still remains. During the study, we found that an emergency setting had a significant effect. In the previous research we have found that ER physicians have a poorer acceptance ratio than other specialty when responding to clinical reminders (7%, OR < 0.1 [13]). Although, in this research, ER physicians also had a lower acceptance ratio than other specialty, it was greater than in the previous study (19. 8%, OR 3.15). As might be expected, the acceptance of the safety reminders was significantly affected by the seriousness of risk to the pregnant woman. There are a number of limitations to this study. The results came from only one source of data and this was collected over a relatively short period of time. The rules used in this analysis may not have filtered out irrelevant category C medication reminders to obstetricians and therefore some false positives may have helped to confound the results of this research. Future qualitative research is thus needed in order to examine the why and the context of the reasons for overrode the reminders that were not considered during this research. The strengths of this study are related to its data source. Firstly, Taiwan healthcare is almost 100% coverage therefore is representative because of the volume of visits. Clinical data within health smart card as pregnant identification helps to minimize the false fails and it also shows how health smart card helps to improve patient safety. Secondly, this research hospital has more than 95% of prescriptions ordered electronically so there is no potential for selection bias relative to the data source. Lastly, there is no potential for recall bias between cases (accepting the reminder) and the controls (overriding the reminders) as all behavior information is based on the log analysis.
Conflict of interest There is no conflict of interest of this research.
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Summary points What was already known: • Many of pregnant women are prescribed potentially harmful medications. Improper prescription during pregnancy is clearly a serious international health problem. • Many publications indicating the importance of safe drug administration during pregnancy, but very few had practical ways to solve the problem or to evaluate the results of a solution. • A number of studies have shown that use of a CPOE with decision support components can reduce medication error. • Several studies have indicated the use of health smart card can potentially improve patient safety. What this study has added to our knowledge: • Implementation of alerting system based on CPOE shows good conformance for non-obstetric physicians (27.8%) and when physicians accepted the alerts and go back to review their orders (69%). • This study reveals sometimes even obstetricians neglect to carefully review their orders since 59% of risk category C drugs alerts accepted were modified. The ordering process can be improved potentially. • The trimester of the pregnant woman is an important factor that might affect the behavior of a physician. • It is possible reminder system has misled nonobstetric physicians by implying that some prescriptions are dangerous.
Authors’ contribution An-Jim Long, Ph.D., collected the major data for this work and has written this paper. Polun Chang, Ph.D., is the supervisor of the first author and provided guidance and assistance for organizing and paper writing.
Acknowledgments We thank Prof. Yu-Chuan (Jack) Li and Prof. Min-Huei (Marc) Hsu for help with preparation of the manuscript and involvement at all stages of this study.
references
[1] W.J. Curran, The thalidomide tragedy in Germany: the end of a historic medicolegal trial, N. Engl. J. Med. 284 (1971) 481–482. [2] W. Lenz, A short history of thalidomide embryopathy, Teratology 38 (1988) 203–215. [3] C.J. van Boxtel, Lessons still to be learned from Thalidomide, Int. J. Risk Saf. Med. 16 (2004) 103–106.
611
[4] A. Ornoy, J. Arnon, Clinical teratology, West J. Med. 159 (1993) 382–390. [5] J.M. Opitz, Associations and syndromes: terminology in clinical genetics and birth defects epidemiology, Am. J. Med. Genet. 49 (1994) 14–20. [6] R. Sannerstedt, P. Lundborg, B.R. Danielsson, et al., Drugs during pregnancy. An issue of risk classification and information to prescribers, Drug Saf. 14 (1996) 69–77. [7] Teratology Society Public Affairs Committee, FDA Classification of drugs for teratogenic risk, Teratology 49 (1994) 446–447. [8] S.S. Singer, A. Falwell, M. David, et al., Patient safety climate in US Hospitals: variation by management level, Med. Care 46 (2008) 1149–1156. [9] J.R. Spina, P.A. Glassman, P. Belperio, et al., Clinical relevance of automated drug alerts from the perspective of medical providers, Am. J. Med. Qual. 20 (2005) 7–14. [10] P.G. Nightingale, D. Adu, N.T. Richards, et al., Implementation of rules based computerised bedside prescribing and administration: intervention study, Br. Med. J. 320 (2000) 750–753. [11] D. Magnus, S. Rodgers, A.J. Avery, GPs’ views on computerized drug interaction alerts: questionnaire survey, J. Clin. Pharm. Ther. 27 (2002) 377–382. [12] H. van der Sijs, J. Aarts, A. Vulto, et al., Overriding of drug safety alerts in computerized physician order entry, J. Am. Med. Inform. Assoc. 13 (2006) 138–147. [13] A.-J. Long, P. Chang, Y.-C. Li, W.-T. Chiu, The use of a CPOE Log for the analysis of physicians’ behavior when responding to drug-duplication reminders, Int. J. Med. Inform. 77 (2008) 499–506. [14] S.E. Ndrade, Jerry H. Gurwitz, Robert L. Davis, Prescription drug use in pregnancy, Am. J. Obstet. Gynecol. 191 (2004) 398–407. [15] J.D. Rubin, C. Ferencz, C. Loffredo, Use of prescription and non-prescription drugs in pregnancy, J. Clin. Epidemiol. 46 (1993) 581–589. [16] I. Lacroix, C. Damase-Michel, M. Lapeyre-Mestre, J.L. Montastruc, Prescription of drugs during pregnancy in France, J. Lancet 356 (2000) 1735–1736. [17] R. Koshy, Navigating the information technology highway: computer solutions to reduce errors and enhance patient safety, Transfusion 45 (2005) 189–205. [18] C. Lambrinoudakis, S. Gritzalis, Managing medical and insurance information through a smart-card-based information system, J. Med. Syst. 24 (2000) 213–234. [19] J.T. Lai, T.W. Hou, C.L. Yeh, et al., Using healthcare IC cards to manage the drug doses of chronic disease patients, Comput. Biol. Med. 37 (2007) 206–213. [20] C. Abrahamsen, Optimal patient safety a computer chip away, Nurs. Manage. 35 (2004) 47–48. [21] A. Queiner-Luft, I. Eggers, G. Stolz, et al., Serial examination of 20,248 newborn fetuses and infants: correlations between drug exposure and major malformations, Am. J. Med. Genet. 63 (1996) 268–276. [22] M. Raebel, N. Carroll, J. Kelleher, Randomized trial to improve prescribing safety during pregnancy, J. Am. Med. Inform. Assoc. 14 (2007) 440–450. [23] S.S. Addis, M. Bonati, Risk classification systems for drug use during pregnancy: are they a reliable source of information? Drug Saf. 23 (2000) 245–253. [24] P. Merlob, B. Stahl, Classification of drugs for teratogenic risk: an anachronistic way of counseling? Teratology 66 (2002) 61–62. [25] P.L. Doering, L.A. Boothby, M. Cheok, Review of pregnancy labeling of prescription drugs: is the current system adequate to inform of risks? Am. J. Obstet. Gynecol. 187 (2002) 333–339.
journal homepage: www.ijmijournal.com
The effect of using the health smart card vs. CPOE reminder system on the prescribing practices of non-obstetric physicians during outpatient visits for pregnant women in Taiwan An-Jim Long a,b , Polun Chang c,∗ a b c
Division of Isotope Application, Institute of Nuclear Energy Research, Taiwan, ROC Institute of Public Health, National Yang-Ming University, Taiwan, ROC Institute of Biomedical Informatics, National Yang-Ming University, Taiwan, ROC
a r t i c l e
i n f o
a b s t r a c t
Article history:
Introduction: There is an evidence that pregnant women have been prescribed a significant
Received 28 December 2009
number of improper medications that could lead to potential damage for a developing fetus
Received in revised form
due to discontinuity of care. The safety of pregnant women raises public concern and there
18 April 2012
is a need to identify ways to prevent potential adverse events to the pregnant woman. This
Accepted 18 April 2012
study used a health smart card with a clinical reminder system to keep continuous records of general outpatient visits of pregnant women to protect them from potential adverse events caused by improper prescription.
Keywords:
Method: The health smart card, issued to all 23 million citizens in Taiwan, was used to
Health smart card
work with a Computerized Physician Order Entry (CPOE) implemented at a 700-bed teaching
Reminder system
medical center in Taipei to provide the outpatient information of pregnant women. FDA
Computerized order entry
pregnancy risk classification was used to categorize the risk of pregnant women. The log
Medication safety
file, combined with the physicians’ and patients’ profiles, were statistically examined using
Pregnancy safety
the Mantel–Haenszel technique to evaluate the impact of system in changing physician’s prescription behavior. Results: A total of 441 patients ranged in age from 15 to 50 years with 1114 prescriptions involved in FDA pregnancy risk classification C, D, and X during the study period. 144 reminders (13.1%) were accepted by physicians for further assessment and 100 (69.4%) of them were modified. Non-obstetric physicians in non-emergency setting were more intended to accept reminders (27.8%, 4.9 folds than obstetricians). Reminders triggered on patients in second trimester (15.5%) were accepted by all physicians more than third trimester (OR 1.52, p < 0.05).
∗ Corresponding author at: Room 520, Library and Information Building, No. 155, Sec. 2, Li-Nong St., Beitou District, Taipei City 11221, Taiwan, ROC. Tel.: +886 2 28267238. E-mail addresses: [email protected] (A.-J. Long), [email protected] (P. Chang). 1386-5056/$ – see front matter © 2012 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijmedinf.2012.04.009
606
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Conclusion: A health smart card armed with CPOE reminder system and well-defined criteria had the potential to decrease harmful medication prescribed to pregnant patients. The results show better conformance for non-obstetric physicians (26%) and when physicians accepted the alerts they are more likely to went back and review their orders (69%). In sum, reminder criteria of FDA pregnancy risk classification C for obstetricians and reminder based on different trimesters is suggested to be refined to improve system acceptability and to decrease improper prescription. © 2012 Elsevier Ireland Ltd. All rights reserved.
1.
Introduction
Since the thalidomide tragedy occurred in the late 1950s and early 1960s [1–4], a number of drugs given to pregnant women have been reported to act as possible teratogens As a result, the use of drugs during pregnancy has received increasing attention [4,5] since their use poses a potential risk to both the mother and the fetus. These concerns may not always be fully understood or applied in current practice, due to limited knowledge on medication safety practice during pregnancy. The first classification system based on clinical and animal data for risks associated with drug use during pregnancy was implemented in Sweden in 1978. The United States Food and Drug Administration (FDA) then introduced its own system in 1979. In the latter, medications are rated as A, B, C, D, and X relative to pregnancy, based on evidence in humans and animals, with category A drugs having been proven safe and category X drugs being contraindicated in pregnancy. Among the classifications, D and X are rated as positive evidence of human fetal risk while category C remains debatable especially a significant number of drugs are rated as category C. This system is widely used and it is believed to be the most efficient approach for establishing risk and safety relative to drug use during pregnancy [6]. It should be noted, however, that it has suggested that the FDA abandon the current classification system in favor of more meaningful evidence-based narrative statements [7]. Prior studies highlight that many of pregnant women are prescribed potentially harmful medications [8–13]. One study in France even showed 99% of the women received at least one prescription for a drug during pregnancy with a mean of 13.6 medications per woman and 1.6% of women received one or more prescriptions of drugs from X category, while 59% of women had a prescription of drugs from the D category [14–16]. Improper prescription during pregnancy is clearly a serious international health problem. Although prescribing potential harmful drugs to the pregnant women may due to a trade-off between medication damage and disease damage, it may also possible that the physicians are lack of knowledge on medication safety practice during pregnancy (especially non-obstetric physicians) that is to say an error in the planning stage of medication use and therefore is often preventable. A number of studies have shown that use of a CPOE with decision support components can reduce medication error [8–13]. To alert prescribing physicians to prevent potential medication errors requires digital identity of the pregnant patient. Hospitals do not have patients’ pregnant history such as weeks if patients were not receiving pregnant regular health check or the patients
come from emergency setting. In Taiwan the health smart card issued by government provides real-time patient eligibility and allows the government to monitor potential fraud or abuse of the medical insurance system. All pregnancy health check records are included and it provides digital identity for the pregnant patient so that incorrect reminders can be minimized when reminder system is applied. Several studies have indicated the use of health smart card can potentially improve patient safety [17–20]. But papers on the practical uses to which a smart card can be put have not been fully evaluated and described. Thus, this leaves open the question as to how to better integrate health smart cards into areas such as the clinical decision support system; such an approach has the great potential to bridge gaps in care. Our literature review found many publications indicating the importance of safe drug administration during pregnancy, but very few had practical ways to solve the problem or to evaluate the results of a solution. We created a reminder system in which a health smart card was integrated into an existing CPOE and selected pregnant women as our study population. FDA classification system was implemented to determine the proportion of potential harmful prescriptions for pregnant patients. We analyzed the order log to assess physicians’ behavior in response to computerized drug reminders to see whether the reminder system armed with health smart card based on FDA classification system was effective in the reduction of potential harmful medications for pregnant patients being prescribed.
2.
System setting and design
For our study every participating outpatient clinic had a desktop installed with computerized physician order entry system. A smart card reader was connected to the desktop which was connected securely to both internet and healthcare insurance private network. Physicians were required to start the system by an authentication process via inserting their own healthcare professional card (HPC) into the card reader. The secure authentication module (SAM) authenticated the HPC, after which the clinical session would then be started. Each patient inserts their card into the reader. We used multi-threading approach to allow the clinician to work while the system read and updated the health smart card. Multi-threading technique empowers the software to deal with multiple jobs as well as multiple devices simultaneously. It means while clinicians are waiting smart card reader to get the required data from health smart card, the software is not suspending and they can edit the patient record while the software is reading the card. It
607
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Fig. 1 – CPOE with progress bar showing the amount read from the health smart card.
should be noted that it required 1–2 s to finish acquisition of one record for the current health smart card issued by Taiwan government each record requires so for acquire data from entire card will take some minutes. The progress of reading the health smart card is shown on the top right side of the system as shown in Fig. 1. When the data read from health smart card has been finished, it triggers the rule base in the CPOE back end and generates a warning message if any. The card contains four sections of patient data as described in Table 1 which are explicit basic data section, healthcare data section, medical data section, and healthcare administration data section. Hospitals record every regular health check for pregnant women at healthcare data section with a week code range from 41 to 50. Ten times free health checks are offered by the Taiwan government. The first prenatal health check is defined at the third month and the second at the fifth month,
four weeks per check after the third check. Although we do not have record of first day of pregnancy on pregnant patients who were not regularly checked by the study hospital, we can calculate pregnancy weeks by the week code on the health smart card.
3.
Materials and methods
The proposed use of health smart card with a clinical reminder system for pregnant women was implemented at a 700-bed teaching medical center in 2006, which serves 1.3 million outpatient visits and has 6000 emergency room visits per year. The computerized physician order entry system has been in use for 8 years and 100% of the outpatient orders are prescribed electronically.
Table 1 – Data sections of health smart card and their application to patient safety. Data section Core data section Healthcare data section
Medical data section
Administration data section
Content Social security number Remarks on the cardholder’s status, catastrophic diseases; number of visits and admissions with diagnosis, utilization of the national health insurance prevention programs such as pre-pregnancy check-ups, premium records of the cardholder, accumulated medical expenditure, cost sharing, etc. Prescriptions, drug allergic history, long-term prescriptions for ambulatory care, and certain medical treatments Personal immunization chart and the willingness to allow organ donation
Application Identification of the patient Medication safety check for pregnant women, diagnosis-medication interaction check
Drug–drug interaction check, drug allergy check, dosage and duplicated medication check Duplicated immunization check
608
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
The rule for a reminder that the physician is prescribing an improper drug to a pregnant woman involves the identification of the pregnancy status of the female patient by acquiring the date of the last health check or diagnosis. Sometimes patients go to hospital to have an abortion or have been aborted. These situations can result in changes to the patient’s status or a wrong diagnosis and may generate a false positive reminder when detecting an improper prescription. Thus, target cases for reminder system were identified that met the following criteria with the aim of minimizing the triggering of false positive reminders: (1) The prescription drugs are evaluated on the assumption of a gestation period of 270 days if no end-up signal (delivery, abortion, etc.) is detected. (2) Some cases are excluded when the diagnosis meets a defined code in International Classification of Diseases 9th (ICD-9) revision and its clinical modification version (ICD-9-CM). These include a legally induced abortion (ICD9 code 635.90), a mother with single live born (ICD-9-CM V27.0-9) and various other cases of abortion (ICD-9-CM 634.0–638.9, 640.0–676.90). (3) Some prescriptions involving female reproductive hormone such as Progestin in some cases is required for pregnancy were excluded [8] since they are the most frequently dispensed category D and X drugs within the FDA pregnancy risk classification [8] to pregnant women. (4) A reminder for each improper prescription only pops-up once for each patient during a clinical session. The prescription drugs are reminded when those are listed as a category C, D and X drug as defined by the FDA pregnancy risk classification system. Although prescribing category C drugs sometimes are needed and whether reminding this remains debatable. But after discussion with pharmacists of the target hospital we concluded there is a need to inform physicians their prescriptions are potentially harmful whatever the level is minor or significant, especially when physicians are not aware of their prescriptions are in category C. System diagram of the implemented reminder system is shown in Fig. 2. The system recorded the entire physicians’ behavior when modifying orders as part of the integrated log system after a reminder had been given. The system recorded the log of prescription behavior in a single table. Every time the system triggered a reminder for improper prescription, the first response (accept or cancel), the profile of prescribing physician, the current status of the patient and the dosage, frequency and duration of the current order were saved. Thus, this research was capable of analyzing the acceptance behavior of physicians when they received a reminder and the relationship between this and a wide range of different variables. The final order was not part of the log system, but was confirmed and saved to the hospital information system. The reminder only appears to the physician after he/she has finished the order and if the order involves a category C, D and X drug from the FDA pregnancy risk classification system. If they accepts the reminder and wish to review their order, they click No to go back to order entry system and review the order. If they click Yes, then the alerted prescription has been
Fig. 2 – Diagram of the implemented system. The health smart card is issued by Taiwan government and includes the person’s name, identification number, date of birth and a photograph, as shown in the left top side of the diagram; the card is blue. On the top of the card is an integrated circuit (IC chip) containing a 32-KB Electrically Erasable Programmable Read-Only Memory (EEPROM). The red card on the right of the machine is the healthcare professional card used to authorize the system. The privacy and confidentiality of the patient information within the card is protected by a PIN number and a healthcare professional card. Without a PIN number and healthcare professional card one can never read information from the health smart card. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of the article.) accepted. According to experience with the drug–drug interaction reminders, we found that physicians exhibited fatigue with an online survey approach and for this reasons ignore clinical reminders. Therefore we only provide a Yes or No selection. We count that as an acceptance if users click “No” button, while count as an override if users click “Yes” button (Fig. 3). The reminder system log for a 1-year period in 2006 for outpatient orders were analyzed along with the pregnancy risk
Fig. 3 – The example of prescription reminder. The dialog box is translated into English; the original is in Chinese. One patient was prescribed Fluitran 2 mg for the treatment of hypertension during pregnancy. Fluitran is classified as D category.
609
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Table 2 – A comparison of the number of reminders accepted (and following actions) and overrode stratified by physician specialty, the risk classification, and the different trimesters of pregnancy. Accepted Overrode Accept Crude OR (CI) rate Specialty 49 Obstetric Non-obstetric 68 9 ER obstetric ER non-obstetric 18 Drug category 117 C 24 D 3 X Trimester 62 Third 49 Second 33 First 144
Total ∗
626 177 83 73
7.3% 27.8% 9.8% 19.8%
Reference 4.90* (3.28–7.35) 1.38 (0.66–2.92) 3.15* (1.74–5.69)
808 136 15
12.7% 15.0% 16.7%
Reference 1.22 (0.76–1.96) 1.38 (0.39–4.84)
515 268 176
10.8% 15.5% 15.8%
Reference 1.52* (1.02–2.27) 1.55 (0.99–2.46)
959
13.1%
p-Value < 0.05.
classification, the specialty of the various physicians involved, and trimester of the pregnant women. Odds ratios (OR) and 95% confidence interval (CI), as well as p values, were calculated for each variable using the Mantel–Haenszel technique. Level of significance was considered when p value <0.05. All patient information used in this study is de-identified. At the time this study was conducted, IRB committee of the hospital was just set up and only accepts clinical trials. The chief information officer of the hospital was well informed and all information was de-identified and obtained from the information office.
4.
Results
A total of 84 physicians were involved in the study and all of them accepted at least one drug duplication reminder. Over the 1-year period, 441 patients with 1114 prescriptions ranged in age from 15 to 50 years with an average of 30.13 years (SD 5.78) were prescribed drugs in FDA pregnancy risk classification C, D, and X, i.e., prescriptions that are potentially harmful to the developing fetus. 144 reminders (13.1%) were accepted by physicians for further assessment and 100 (69.4%) of them were modified. Patients were receiving regular pregnancy health check from 215 different clinics or hospitals and asking second opinion or receiving diagnosis or treatment of other specialties to the study hospital. All visits were collected for investigation, together with the complete prescription orders and the medical history. Obstetricians prescribe majority of orders to the pregnant women (Table 2). In the other hand physicians in the emergency room (ER) receive high pressure of workload and have the least time to review patient record. We identified these properties to take into account confounder effects. According to the result, non-obstetric physicians showed a high total acceptance rate in respond to safety reminders, whether in an ER or non-ER situation (19. 8% vs. 27.8%). Obstetricians had the lowest acceptance rate in respond to the safety reminders (they accepted 44 class C reminders while they
Table 3 – A comparison of the number of reminders accepted (and following actions) and overrode stratified by physician specialty, the risk classification, and the different trimesters of pregnancy.
Specialty Obstetric Non-obstetric ER obstetric ER non-obstetric Drug category C D X Trimester Third Second First Total
Deleted
Updated
Unchanged
Total
16 (33%) 41 (60%) 1 (11%) 5 (28%)
9 (18%) 8 (12%) 8 (89%) 12 (67%)
24 (49%) 19 (28%) 0 (0%) 1 (6%)
49 68 9 18
56 (48%) 7 (29%) 0 (0%)
31 (27%) 5 (21%) 1 (33%)
30 (26%) 12 (50%) 2 (67%)
117 24 3
31 (50%) 18 (37%) 14 (42%)
12 (19%) 15 (31%) 10 (30%)
19 (31%) 16 (33%) 9 (27%)
62 49 33
63 (44%)
37 (26%)
44 (31%)
144
only accepted 14 class D and X reminders; they overrode 617 class C reminders while they overrode 92 class D and X reminders), whether in an ER or non-ER situation (9. 8% vs. 7.3%). During the research we found that there was an effect in terms of the seriousness of the risk to the pregnant women but was not statistical significant. Physicians had the slightly higher acceptance rate for X category drugs then D and then C (16. 7% vs. 15% vs. 12.7%). There were more reminders accepted during the first trimester than during other trimesters (15.8% vs. 15.5% and 10.8%). The first three months of pregnancy is the most important period for teratogenic influence [21]. This point seemed to have been taken into account by the physicians. Due to the pregnancy common wealth enrollment policy (first free prenatal health check begin with the second month), cases of first trimester are less than other trimester so the OR in responding alerts of first trimester patient is not statistical significant while second trimester has higher acceptance rate which is statistically significant than third trimester (Table 3). Once physicians accepted the reminder, CPOE system led the physicians back to review their orders and 69% of orders were updated by physicians (deleted 44% and updated 26%). 51.02% of orders made by obstetricians were updated (deleted 33% and updated 18%) after alerts accepted and reviewed, while in ER setting all orders were updated by obstetricians. Only one (6%) prescription was denied to update prescribed by non-obstetrics physicians in ER setting.
5.
Discussion
The results of this study show that the use of smart card with a CPOE reminder system has potential to decrease proportion of potential harmful medication prescribed to pregnant patient. Implementation of reminder system integrating health smart card with CPOE and well-defined criteria reduced triggering of false positives. And potential harmful prescriptions were errors in planning stage in medication use can be identified without over-reminding prescribers. In total, the overall level of acceptance at 13.1% for the reminder system was low compared to other studies [9–12].
610
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
The possible reasons are (1) the alerts were considered falsely positive (due to the use of health smart card the false positive rate is dropped and many of the factors associated to diagnosis code in [22] were never happened in the study); (2) there is usually no better candidate than the current prescribed medication; (3) there is a trade-off between medication damage and disease damage. According to analysis of the transaction log of the reminder system, the most frequent reminders were trigger by obstetricians which had the lowest acceptance rate and hence the entire acceptance rate dropped. According to some publications [23–25] reliability of drug classification systems for teratogenic risk is questioned and since category C contribute majority of reminders it is largely questioned. X category drugs are classified with evidence to be harmful to pregnant women so are considered more relevant to the clinical settings. In most cases obstetricians are more familiar with classification systems so they overrode most of the reminders if those were with less risk (especially category C prescriptions). This goes some ways to explaining why obstetricians had an acceptance rate for class C reminders of only 6.7% (44/661, as described in Section 4), while the acceptance rate for class D and X reminders was 13.2% (14/106). Therefore, we suggest that obstetricians should receive fewer reminders (by filtering some common medications in class C but are not considered harmful reviewed by obstetricians) in this area than other physicians. During this study we did not design a special reminder rule set for obstetricians, since the major point for this study is to facilitate the use of health smart card in clinical setting and evaluate the results, but this may be implemented in the future. On the other hand, the acceptance rate for class C reminders among the other physicians was 27.7% (accepted 73 reminders while overrode 191 reminders), which is high. It is also noted that non-obstetrics physicians intended to delete (60%) the alerted prescriptions after reviewed carefully. It may due to physicians are unfamiliar with the pregnancy classification system. We reviewed all the 68 reminders accepted by this group of physicians. We found many of the prescriptions can be replaced by alternative medications while some prescriptions might not be considered as harmful but was classified as C category. For example, Dextromethorphan is rated as A in Austria system and is OTC drug in US. Physicians can prescribe Acetylcysteine instead but most of prescriptions were deleted. There are some more examples such Allegra and Butyscol are in the same case. It is possible in this case the reminder system has misled other physicians by implying that some prescriptions are dangerous; in this context, according to some publications [23–25], informing risks by current classification systems for drug use during pregnancy remains in debate. This aspect therefore requires further research that will analyze the pros and cons of the pregnancy risk classification system; once this is known, it will be possible in the future to provide physicians with a better quality of information. Although category C drugs were having poor acceptance rate in overall groups and some physicians were questioning validity of the category C alerts, we still suggest to alert physicians when they prescribe category C drugs. First the category D and X were not having statistically significant acceptance rate than category C drugs. Second there were 44 category C alerts accepted by obstetricians and 26 of these were modified
(59%), while D and X category prescriptions were only accepted by 20% (one out of five). It means prescription behavior to the pregnant women can be changed even for obstetricians the ordering process can be improved potentially. The trimester of the pregnant woman is an important factor that might affect the behavior of a physician. In the first trimester the fetus grows at its fastest pace during the first trimester. So the first trimester of pregnancy is the most crucial period in pregnancy but is least visible since the fetus weights less than an ounce. During the period the fetus is most vulnerable to the harmful effects of drugs so prescriptions are much more important than during other trimesters. During the study we did show the physicians the data on the week of the pregnancy, but we did not put any warning message up showing that the patient was in first trimester. This means that most of the physicians understand the importance of medication safety in first trimester for pregnant women and also shows how the system was able to improve patient care. The hospital that took part in this study serves about 1% of all patients in Taiwan. Therefore, according to the result of this study, we estimate that there might be around 50,000 pregnant women every year receive medications that are harmful to the developing fetus, which is about 25% of all total pregnancies. Although this figure is still lower than for the US and France [8–16], important problems still remains. During the study, we found that an emergency setting had a significant effect. In the previous research we have found that ER physicians have a poorer acceptance ratio than other specialty when responding to clinical reminders (7%, OR < 0.1 [13]). Although, in this research, ER physicians also had a lower acceptance ratio than other specialty, it was greater than in the previous study (19. 8%, OR 3.15). As might be expected, the acceptance of the safety reminders was significantly affected by the seriousness of risk to the pregnant woman. There are a number of limitations to this study. The results came from only one source of data and this was collected over a relatively short period of time. The rules used in this analysis may not have filtered out irrelevant category C medication reminders to obstetricians and therefore some false positives may have helped to confound the results of this research. Future qualitative research is thus needed in order to examine the why and the context of the reasons for overrode the reminders that were not considered during this research. The strengths of this study are related to its data source. Firstly, Taiwan healthcare is almost 100% coverage therefore is representative because of the volume of visits. Clinical data within health smart card as pregnant identification helps to minimize the false fails and it also shows how health smart card helps to improve patient safety. Secondly, this research hospital has more than 95% of prescriptions ordered electronically so there is no potential for selection bias relative to the data source. Lastly, there is no potential for recall bias between cases (accepting the reminder) and the controls (overriding the reminders) as all behavior information is based on the log analysis.
Conflict of interest There is no conflict of interest of this research.
i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 8 1 ( 2 0 1 2 ) 605–611
Summary points What was already known: • Many of pregnant women are prescribed potentially harmful medications. Improper prescription during pregnancy is clearly a serious international health problem. • Many publications indicating the importance of safe drug administration during pregnancy, but very few had practical ways to solve the problem or to evaluate the results of a solution. • A number of studies have shown that use of a CPOE with decision support components can reduce medication error. • Several studies have indicated the use of health smart card can potentially improve patient safety. What this study has added to our knowledge: • Implementation of alerting system based on CPOE shows good conformance for non-obstetric physicians (27.8%) and when physicians accepted the alerts and go back to review their orders (69%). • This study reveals sometimes even obstetricians neglect to carefully review their orders since 59% of risk category C drugs alerts accepted were modified. The ordering process can be improved potentially. • The trimester of the pregnant woman is an important factor that might affect the behavior of a physician. • It is possible reminder system has misled nonobstetric physicians by implying that some prescriptions are dangerous.
Authors’ contribution An-Jim Long, Ph.D., collected the major data for this work and has written this paper. Polun Chang, Ph.D., is the supervisor of the first author and provided guidance and assistance for organizing and paper writing.
Acknowledgments We thank Prof. Yu-Chuan (Jack) Li and Prof. Min-Huei (Marc) Hsu for help with preparation of the manuscript and involvement at all stages of this study.
references
[1] W.J. Curran, The thalidomide tragedy in Germany: the end of a historic medicolegal trial, N. Engl. J. Med. 284 (1971) 481–482. [2] W. Lenz, A short history of thalidomide embryopathy, Teratology 38 (1988) 203–215. [3] C.J. van Boxtel, Lessons still to be learned from Thalidomide, Int. J. Risk Saf. Med. 16 (2004) 103–106.
611
[4] A. Ornoy, J. Arnon, Clinical teratology, West J. Med. 159 (1993) 382–390. [5] J.M. Opitz, Associations and syndromes: terminology in clinical genetics and birth defects epidemiology, Am. J. Med. Genet. 49 (1994) 14–20. [6] R. Sannerstedt, P. Lundborg, B.R. Danielsson, et al., Drugs during pregnancy. An issue of risk classification and information to prescribers, Drug Saf. 14 (1996) 69–77. [7] Teratology Society Public Affairs Committee, FDA Classification of drugs for teratogenic risk, Teratology 49 (1994) 446–447. [8] S.S. Singer, A. Falwell, M. David, et al., Patient safety climate in US Hospitals: variation by management level, Med. Care 46 (2008) 1149–1156. [9] J.R. Spina, P.A. Glassman, P. Belperio, et al., Clinical relevance of automated drug alerts from the perspective of medical providers, Am. J. Med. Qual. 20 (2005) 7–14. [10] P.G. Nightingale, D. Adu, N.T. Richards, et al., Implementation of rules based computerised bedside prescribing and administration: intervention study, Br. Med. J. 320 (2000) 750–753. [11] D. Magnus, S. Rodgers, A.J. Avery, GPs’ views on computerized drug interaction alerts: questionnaire survey, J. Clin. Pharm. Ther. 27 (2002) 377–382. [12] H. van der Sijs, J. Aarts, A. Vulto, et al., Overriding of drug safety alerts in computerized physician order entry, J. Am. Med. Inform. Assoc. 13 (2006) 138–147. [13] A.-J. Long, P. Chang, Y.-C. Li, W.-T. Chiu, The use of a CPOE Log for the analysis of physicians’ behavior when responding to drug-duplication reminders, Int. J. Med. Inform. 77 (2008) 499–506. [14] S.E. Ndrade, Jerry H. Gurwitz, Robert L. Davis, Prescription drug use in pregnancy, Am. J. Obstet. Gynecol. 191 (2004) 398–407. [15] J.D. Rubin, C. Ferencz, C. Loffredo, Use of prescription and non-prescription drugs in pregnancy, J. Clin. Epidemiol. 46 (1993) 581–589. [16] I. Lacroix, C. Damase-Michel, M. Lapeyre-Mestre, J.L. Montastruc, Prescription of drugs during pregnancy in France, J. Lancet 356 (2000) 1735–1736. [17] R. Koshy, Navigating the information technology highway: computer solutions to reduce errors and enhance patient safety, Transfusion 45 (2005) 189–205. [18] C. Lambrinoudakis, S. Gritzalis, Managing medical and insurance information through a smart-card-based information system, J. Med. Syst. 24 (2000) 213–234. [19] J.T. Lai, T.W. Hou, C.L. Yeh, et al., Using healthcare IC cards to manage the drug doses of chronic disease patients, Comput. Biol. Med. 37 (2007) 206–213. [20] C. Abrahamsen, Optimal patient safety a computer chip away, Nurs. Manage. 35 (2004) 47–48. [21] A. Queiner-Luft, I. Eggers, G. Stolz, et al., Serial examination of 20,248 newborn fetuses and infants: correlations between drug exposure and major malformations, Am. J. Med. Genet. 63 (1996) 268–276. [22] M. Raebel, N. Carroll, J. Kelleher, Randomized trial to improve prescribing safety during pregnancy, J. Am. Med. Inform. Assoc. 14 (2007) 440–450. [23] S.S. Addis, M. Bonati, Risk classification systems for drug use during pregnancy: are they a reliable source of information? Drug Saf. 23 (2000) 245–253. [24] P. Merlob, B. Stahl, Classification of drugs for teratogenic risk: an anachronistic way of counseling? Teratology 66 (2002) 61–62. [25] P.L. Doering, L.A. Boothby, M. Cheok, Review of pregnancy labeling of prescription drugs: is the current system adequate to inform of risks? Am. J. Obstet. Gynecol. 187 (2002) 333–339.