Prescription surveillance for early detection system of emerging and reemerging infectious disease outbreaks.

Based on prescriptions filled at external pharmacies, prescription surveillance (PS) in Japan has been reporting the estimated numbers of influenza and varicella patients and people prescribed certain drugs since 2009. Every morning, this system estimates the numbers of patients from the numbers of prescriptions filled nationwide for neuraminidase inhibitors, anti-herpes virus drugs, antibiotic drugs, antipyretic analgesics, and multi-ingredient cold medications. Moreover, it can detect "unexplained" infectious diseases that are not explained as infectious diseases monitored by other surveillance systems. Such "unexplained" infectious diseases might be emerging and re-emerging infectious diseases including bioterrorism attacks, which are reportedly difficult to diagnose, at least in early outbreak stages. To ascertain the system's potential benefits, this study examined schemes to detect "unexplained" infectious diseases using PS information. The numbers of patients prescribed the respective drugs are first regressed on the known infectious diseases, time trends, and dummies for day-of-the-week, holidays, and days following a holiday. Known infectious diseases are defined as covered by the National Official Sentinel Surveillance for Infectious Diseases under the Infection Control Law. After the numbers of patients from PS are compared with the predicted numbers of patients, their probabilities of occurrence are calculated. We examined the system's prospective operation from January 2017 through July 2018. The criterion we used to define aberrations varied, from 0.01 to 10-7. For criteria of 0.01 and 10-7 we found 254 and 15 aberrations, respectively. We confirmed its feasibility and effectiveness.

Evaluation of the global action plan on antimicrobial resistance in Japan during its first eighteen months.

Antimicrobial resistance (AMR) has garnered the most attention among public health concerns worldwide. Japan formulated a national action plan for AMR in April 2016. The plan seeks to reduce the amount of antimicrobials used in 2020 to two-thirds of the use recorded in 2013. Prescription surveillance (PS) is being used to monitor trends in the amount of antimicrobials used. PS estimates the number of patients prescribed an antimicrobial each day. The number of patients who were prescribed an antimicrobial under the action plan was analyzed by including dummy variables with other control variables. Data from April 1, 2011 to 30 September 30, 2017 were analyzed. When the number of patients with an infectious disease (1 of 13 specified diseases) served as a dummy variable, estimates indicated that the coefficient of that dummy variable was not significant. If the number of patients with an infectious disease (1 of 13 specified diseases) was excluded as an explanatory variable, then the estimated coefficient was significant. The global action plan in Japan might not reduce the amount of antimicrobials used. The current results indicated that the number of patients who were prescribed an antimicrobial did not decrease significantly after initiation of the action plan. This finding does not exclude the possibility that the average amount of antimicrobials used per patient has decreased.

[Association between Diarrheal Infectious Diseases and Antidiarrheal Drugs Assessed with Prescription Surveillance].

OBJECT: It appears to be possible to measure the outbreak of infectious diseases with accompanying diarrhea by early detection of an outbreak. However, because some diseases are reported weekly from sentinel medical institutions, early detection of outbreaks might be inadequate. In this study, we examined the number of patients with diarrhea assessed with prescription surveillance and validated its association with the number of patients who have infectious diseases with diarrhea. METHODS: The number of patients who were prescribed an antidiarrheal drug or intestinal drug was estimated for each prefecture using the prescription surveillance system from April 1, 2011 through March 31, 2014. Moreover, we examined the association between the prevalence of diarrhea in prescription surveillance and the number of patients with diarrheal infectious diseases. RESULTS: Results showed a positive correlation between the prevalence of diarrhea in prescription surveillance and the number of reported cases per sentinel with gastrointestinal infections. However, no positive correlation was found with the others infectious diseases. CONCLUSION: Estimation of the number of patients with diarrhea in prescription surveillance might facilitate early detection of an outbreak of gastrointestinal infections.

[Evaluation of the Number of Varicella Patients Estimated by Prescription Surveillance].

OBJECT: It is important to know the precise number of varicella patients infected for evaluation of routine immunization and anti-bioterrorism attack using smallpox. Prescription Surveillance (PS) has been providing the estimated number of varicella patients up to the present. However, the estimated number of varicella patients cannot be validated because to date there has been no other comparable precise method of estimation. Recently, all electronic medical claims nationwide (NDB) have been disclosed. In this paper, we compare the number of varicella patients estimated by PS with NDB data, and adjust the number estimated with PS, if necessary. METHOD: For both NDB and PS, we used the monthly data from April, 2010 to March, 2013. The estimation of the number of varicella patients from the PS data was adjusted by the proportion of estimated number based on PS to the one based on NDB in the entire study period. Moreover, we adjusted it month by month, if the former method may not be enough to compensate for the discrepancy between the two datasets. RESULTS: The average discrepancy between NDB and PS was 48.00% in three years. By the adjustment using NDB in the three years, the discrepancy was improved to 11.49%. However, seasonal patterns of overestimation or underestimation were found. Conversely, by the adjustment using NDB month by month, the discrepancy was greatly reduced to 4.33%. Moreover, the seasonal patterns of overestimation or underestimation disappeared. CONCLUSION: The number of patients based on NDB would appear the most precise number, however, there may be a delay of about one year before it becomes available. On the other hand, PS data are updated every day and provide us with the up-to-date situation. This paper found that combining the timeliness of the PS data and preciseness of the NDB data will provide substantial benefit for public health.

Proposition of real-time precise prediction model of infectious disease patients from Prescription Surveillance using the National Database of Electronic Medical Claims.

The incidence of common pediatric infectious diseases has been monitored officially at sentinel medical institutions in Japan. However, the numbers of affected patients are not provided. Prescription Surveillance (PS), which infers the number of patients with influenza, varicella, and gastrointestinal infections from data related to prescriptions at external pharmacies, provides estimates to the public the following morning. This study assessed the prediction ability of the incidence of common pediatric infectious diseases from PS information using the National Database of Electronic Medical Claims (NDBEMC): the number of patients prescribed neuraminidase inhibitors, anti-herpes virus drugs, antibiotic drugs, antipyretic analgesics, and multi-ingredient cold medications. The diseases include RS virus infection, pharyngoconjunctival fever, hand, foot and mouth disease, erythema infectiosum, exanthem subitum, pertussis, herpangina, influenza, varicella, and gastrointestinal infection. For comparison, we used the estimated number of patients who were prescribed neuraminidase inhibitor in PS, which had been confirmed already for precision, and provided estimates to the general public via the internet. The discrepancy rates of all considered diseases between the reported number in NDBEMC and the predicted numbers of patients from PS were less than the value in NI counts and the coefficients of determination in the estimation were from .8109 to .9825. These predictions were sufficiently precise to provide to the general public.

Evaluation of estimated number of influenza patients from national sentinel surveillance using the national database of electronic medical claims.

Officially, the national official sentinel surveillance of infectious diseases (NOSSID) has been used to estimate the number of influenza patients nationwide; NOSSID is based on the Law Concerning the Prevention of Infectious Diseases and Medical Care for Patients of Infections (the Infectious Diseases Control Law). Prescription Surveillance (PS) has also provided a numerical estimate of influenza patients. This study compared these 2 estimations using NOSSID and PS with the numbers of influenza patients from all electronic medical claims (NDBEMC), which had the nearly-comprehensive data from surveys. Results showed that the estimate from NOSSID was about twice the estimate from the NDBEMC. However, the estimated number from the PS was almost equivalent to that from the NDBEMC. The estimated number of patients from NOSSID might not be precise, but NOSSID itself may be useful to monitor influenza trends.

Real-time prescription surveillance and its application to monitoring seasonal influenza activity in Japan.

BACKGROUND: Real-time surveillance is fundamental for effective control of disease outbreaks, but the official sentinel surveillance in Japan collects information related to disease activity only weekly and updates it with a 1-week time lag. OBJECTIVE: To report on a prescription surveillance system using electronic records related to prescription drugs that was started in 2008 in Japan, and to evaluate the surveillance system for monitoring influenza activity during the 2009-2010 and 2010-2011 influenza seasons. METHODS: We developed an automatic surveillance system using electronic records of prescription drug purchases collected from 5275 pharmacies through the application service provider's medical claims service. We then applied the system to monitoring influenza activity during the 2009-2010 and 2010-2011 influenza seasons. The surveillance system collected information related to drugs and patients directly and automatically from the electronic prescription record system, and estimated the number of influenza cases based on the number of prescriptions of anti-influenza virus medication. Then it shared the information related to influenza activity through the Internet with the public on a daily basis. RESULTS: During the 2009-2010 influenza season, the number of influenza patients estimated by the prescription surveillance system between the 28th week of 2009 and the 12th week of 2010 was 9,234,289. In the 2010-2011 influenza season, the number of influenza patients between the 36th week of 2010 and the 12th week of 2011 was 7,153,437. The estimated number of influenza cases was highly correlated with that predicted by the official sentinel surveillance (r = .992, P < .001 for 2009-2010; r = .972, P < .001 for 2010-2011), indicating that the prescription surveillance system produced a good approximation of activity patterns. CONCLUSIONS: Our prescription surveillance system presents great potential for monitoring influenza activity and for providing early detection of infectious disease outbreaks.

[The real-time pharmacy surveillance and its estimation of patients in 2009 influenza A (H1N1)].

OBJECT: Detecting of disease spread is an important task of public health and medical staff, especially in pandemics such as A/H1N1 flu (2009). This requires daily observation and estimation of the infected population. The fully automated real-time pharmacy survey we developed collects information electronically at pharmaceutical prescription. We used the data to analyze the pandemic A/H1N1 flu spread (2009) and to determine the system's and capability in estimating the infected population. METHOD: Automatic collection of prescription information on antiinfluenza virus drugs from 3959 pharmacies provided the basis for calculating the number of influenza sufferers and determining shape of the epidemic curve compared to that of official influenza sentinel surveys and mandatory reports of A/H1N1 (2009) patients. We also compared infection estimates from the pharmacy survey to those of official sentinel survey and a one-week survey of all hospitals and clinics in Gifu prefecture not reported in sentinel, RESULTS: Fully automated real-time pharmacy surveillance began on April 20, 2009, and provided feedback at 07:00 daily. It estimated the infected population at 22,708 when official sentinel surveillance recorded an average of 0.99 influenza visits per week in epidemic week 32 when publicly announced that the pandemic had began in Japan. By the end of March, epidemic week 12 in 2010, infected-population estimates totaled 9,234,289, and peaked on November 24 at 234,519 in one day. All A/H1N1 (2009) sufferers reported mandatorily until mid-July numbered 25,526. The pharmacy survey indicated that there were influenza nationalwide by the time the very first outbreak emerged in the Kansai (western Japan) area. The correlation coefficient for the pharmacy and official sentinel survey was 0.992 nationwide, exceeding 0.95 in which only 33 of Japan's 47 prefectures were counted. The estimated infected population in the pharmacy survey was half of that of the official sentinel survey. The pharmacy survey yielded almost the same number as the complete survey in Gifu prefecture, however. DISCUSSION: Fully automated real-time pharmacy surveys are useful in long-term observation e.g. detection of rapid emergence, identifying the peak, and careful monitoring of reemergence. It was demonstrated as the leading indicator for the official sentinel surveillance because of high correlation among them. Information collected daily is very useful in early detection and estimating the affected population. The survey consistently uses the same estimation criterion and operates automatically and routinely, facilitating the comparison of the latest and past results. The pharmacy survey indicated that official sentinel survey estimates overestimate actual cases and thus require modification to ensure accuracy. The pharmacy survey thus appears to be very valuable as a tool in measuring for the second wave of A/H1N1 (2009) or whatever the next pandemic may be. It can, of course, be applied to diseases other than influenza, e.g., varicella, by following antivaricellazostervirus prescriptions and antibiotic drugs.

[Chickenpox case estimation in acyclovir pharmacy survey and early bioterrorism detection].

OBJECTIVE: Early potential health hazards and bioterrorism threats require early detection. Smallpox cases caused by terrorist could, for example, be treated by prescribing acyclovir to those having fever and vesicle exanthema diagnosed as chicken pox. We have constructed real-time pharmacy surveillance scenarios using information technology (IT) to monitor acyclovir prescription. METHODS: We collected the number of acyclovir prescriptions from 5138 pharmacies using the Application Server Provider System (ASP) to estimate the number of cases. We then compared the number of those given acyclovir under 15 years old from pharmacy surveillance and sentinel surveillance for chickenpox under the Infection Disease Control Law. RESULTS: The estimated number of under 15 years old prescribed acyclovir in pharmacy surveillance resembled sentinel surveillance results and showed a similar seasonal chickenpox pattern. The correlation coefficient was 0.8575. The estimated numbers of adults, older than 15 but under 65 years old, and elderly, older than 65, prescribed acyclovir showed no clear seasonal pattern. DISCUSSION: Pharmacy surveillance for acyclovir identified the baseline and can be used to detect unusual chickenpox outbreak. Bioterrorism attack could potentially be detected using smallpox virus when acyclovir prescription for adults suddenly increases without outbreaks in children or the elderly. This acyclovir prescription monitoring such as an application is, to our knowledge, the first of its kind anywhre.