<b>Signal Detection of Adverse Drug Reactions through LASSO Logistic Regression </b><b>Using an Electronic Health Records Database</b>：<b>A Case-Control Study </b> / 薬剤疫学

<p><b>Objective</b>：The objective of this study was to apply Least Absolute Shrinkage and Selection Operator （LASSO）logistic regression to detection of adverse drug reaction （ADR） signals using an electronic health records database as a comprehensive and quantitative method to supplement the current pharmacovigilance activities in Japan.</p><p><b>Design</b>：case-control study</p><p><b>Methods</b>：We analyzed data from 40767 inpatients using a single-institution hospital database and identified two ADRs, suspected pancreatitis and thrombocytopenia, using abnormal laboratory test results. LASSO logistic regression analysis was applied to detect ADR signals with adjustment for age, sex, comorbidities and medical procedures. The positive predictive value （PPV） was calculated using reference standard of known drug-ADR associations based on drug product labels.</p><p><b>Results</b>：The number of case group was 6735 for suspected pancreatitis and 11561 for thrombocytopenia. The number of ADR signals detected using LASSO logistic regression was 27 for suspected pancreatitis and 40 for thrombocytopenia. The calculated PPV was 3.7% for suspected pancreatitis and 55.0% for thrombocytopenia.</p><p><b>Conclusion</b>：LASSO logistic regression analysis efficiently detects ADR signals by adjusting for confounding factors such as comorbidities and medical procedures. The false positive signals may contain unknown signals and further signal assessment will be needed.</p><p></p>

Signal Detection of Adverse Drug Reactions through LASSO Logistic Regression Using an Electronic Health Records Database：A Case-Control Study / 薬剤疫学

Objective：The objective of this study was to apply Least Absolute Shrinkage and Selection Operator （LASSO）logistic regression to detection of adverse drug reaction （ADR） signals using an electronic health records database as a comprehensive and quantitative method to supplement the current pharmacovigilance activities in Japan.Design：case-control studyMethods：We analyzed data from 40767 inpatients using a single-institution hospital database and identified two ADRs, suspected pancreatitis and thrombocytopenia, using abnormal laboratory test results. LASSO logistic regression analysis was applied to detect ADR signals with adjustment for age, sex, comorbidities and medical procedures. The positive predictive value （PPV） was calculated using reference standard of known drug-ADR associations based on drug product labels.Results：The number of case group was 6735 for suspected pancreatitis and 11561 for thrombocytopenia. The number of ADR signals detected using LASSO logistic regression was 27 for suspected pancreatitis and 40 for thrombocytopenia. The calculated PPV was 3.7% for suspected pancreatitis and 55.0% for thrombocytopenia.Conclusion：LASSO logistic regression analysis efficiently detects ADR signals by adjusting for confounding factors such as comorbidities and medical procedures. The false positive signals may contain unknown signals and further signal assessment will be needed.

Survey on Usage of Medical Referral Information in Japanese Physicians / 대한의료정보학회지

OBJECTIVES: The purpose of this survey was to explore physicians' opinions to identify an adequate time range for clinical information to be provided with a referral that would help minimize wasteful retesting. METHODS: In 2011, we conducted a questionnaire survey of 193 physicians. Examining the degree of utilization of provided medical information, we determined the range of clinical information of referral documents. RESULTS: Less than three months of prescription history and blood sample test results in patient referral was most frequent. Less than one year of image information was most frequent. Most doctors answered there is no need to repeat the same type of blood test in their institute when they had information less than half a month old. Less than half to one month of image information was most frequent. Also, it appeared many doctors think “fundamentally they do not change their mind from their own medical department standpoint.” At the actual site, those who would even review referral clinical notes accounted for about 30% of all participants. CONCLUSIONS: Medical referral eventually takes place after the establishment of mutual communication and should consider the workflow and system environment of the receiver of the information.

Current Status and Challenge of the Japanese Standard Disease Master / 日本東洋医学雑誌

Standardization of health care information is one of the major tasks for national IT strategies in Japan. In health care domain, because disease information is important, assigning one unique code to one disease concept is an essential requirement for reliable information processing of disease information, as well as standardization by assigning one unique disease name to one disease concept. The Japanese Standard Disease Master was developed to contribute on this purpose. The Committee for Controlled Medical Terminology of Japanese Association of Medical Sciences is responsible for standardizing disease names, and another committee dedicated for assigning codes to the disease names is set up in the Social Insurance Medical Fee Payment Fund together with Medical Information System Development Center. The first version of the master was released in July 2002,and the maintenances are continued to update four times a year according to various users' requirements received. The master contains about 22,000 terms and 2,000 modifiers. However, it is getting difficult to cover all of the disease descriptions for various clinical settings, and the author believes that a new framework of building ontology-based clinical concept database and an application of the database.

Methodological Issues of Pharmacoepidemiology Studies Using Hospital-based Database in Japan / 薬剤疫学

Objective : To know how to conduct good pharmacoepidemiology studies using hospital-based database in Japan.<BR>Methods : Medical records during 15 months January 1996 and March 1997 in the University of Tokyo Hospital Information System (HIS) are examined know whether it is possible to conduct pharmacoepidemiology studies similar to previous studies on asthma drugs (Spitzer et al, 1992) and calcium antagonists (Psaty et al, 1995). To know the stability of population covered by HIS, the following two intervals are calculated for ambulatory patients with asthma and hypertension ; 1) average intervals of successive two outpatient visits and 2) intervals between the last day of outpatient visit and the last day of observation.<BR>Results : The size of possible pharmacoepidemiology studies attainable using HIS is judged to be more than 5% of previous studies in Canada and America. Average intervals of successive two outpatient visits are estimated to be 30 days or less for 59% of 693 asthmatics and 77% of 2842 hypertensives. For 48% of asthmatics and 71% of hypertensives, intervals between the last day of outpatient visit and the last day of observation are estimated to be 30 days or less.<BR>Discussion : To attain a size appropriate for pharmacoepidemiology study, researchers must cooperate across hospitals. Although a patient can visit any hospital anywhere under Japanese comprehensive medical care plan, it seems that patients tend to become to visit one particular hospital. However, additional information on medical care in other hospitals is needed for each study subject.<BR>Conclusion : Japanese hospital-based database is suitable for pharmacoepidemiology studies as a record during a long time period is usually available for a large fraction of patients with a particular disease. The study may be free from some of biases closely associated with referral processes known to occur in hospital case-control studies. A design of case-control study selecting patients with long medical records across 5-10 hospitals is probably the most promising when using Japanese hospital-based databases.

Comparative Study on Data Sources of Detecting Adverse Drug Reactions / 薬剤疫学

Objective : To find the effective means to detect adverse drug reactions (ADRs) from hospital information system, three data sources, i.e. diagnosis data (Dx), laboratory data (Lab), and prescription data (Rx), are compared in diuretics induced hyperuricemia and/or gout (H/G).<BR>Design : Retrospective cohort study.<BR>Methods : Cohort entry period was three months. Hypertensive outpatients who already had H/G prior to that period were excluded. Then, they were surveyed for 9 months. The patients using diuretics were separated into two groups, i.e. Thiazide-treated group, and Loop-treated group.<BR>Controls were randomly selected from non-diuretic-treated hypertensive outpatients matched to each diuretic group by age and sex. Signals of ADRs were the new prescription of drugs employed in the treatment of H/G from Rx, abnormal serum uric acid level from Lab, and diagnosis of H/G from Dx. The interrelationship of them were examined by the Venn diagram and scatter plot. Finally the incidence of ADRs detected by the above signals and relative risks were calculated and compared. Moreover, prevalence of renal disease in each group was surveyed to examine the possibility that renal disease caused H/G.<BR>Results : Eighteen patients in 240 outpatients treated with Thiazide diuretics and 70 patients in 523 outpatients treated with Loop diuretics were found having developed H/G from Dx, Lab, and/or Rx data sources. More than 90% of total patients were detected from Lab while, a few patients were identified from Dx and Rx. It was rare and coincidental that the three data sources agreed with one another.<BR>The risk of Loop diuretics is approximately twice that of Thiazide diuretics. The incidence and risk of H/G in diuretics estimated in the current study were compatible with the prior report. However, the prevalence of renal disease were high (though not statistically significant) in Loop-treated group so that we possibly overestimated the risk of it.<BR>Conclusion : The order of three data sources, arranged according to the number of ADR signals detected, was Lab, Rx, and Dx. It may be possible to assess the risk of ADR even by Lab only. If Lab is not available, Rx and Dx are useful provided that more subjects and longer research period are involved. However it is necessary to combine three data sources, Dx, Lab, and Rx to detect as many suspected adverse events as possible when using the present clinical database.