The National Health Insurance (NHI) in Taiwan started in 1995 and covered 23,025,773 beneficiaries  (about 99% of 23,119,772 inhabitants ) at the end of 2009. Since 1999, the Bureau of National Health Insurance (BNHI) has released the claims data in electronic form to the National Health Research Institutes for research use under the project of National Health Insurance Research Database (NHIRD). Some dozen kinds of extracted datasets are available on application.
From the NHIRD, we obtained the ambulatory part of systematic sampling datasets in 2009. This kind of datasets was randomly sampled in a ratio of 1/500 from the entire datasets of the NHI in the year. It was composed of claims of 576,416 visits and 2,765,389 prescriptions in 2009. The data of each visit included the date of the consultation, the consulted physician’s specialty, and the patient’s age, sex and birthday. The data of each prescribed drug item included the drug identification number, dosage, frequency, and total amount. On the web site of the BNHI we could look up the ingredient, dosage form and strength of a drug according to its identification number. An interface for searching drug items according to the ingredient was also available.
This study has been approved by the institutional review board of Taipei Veterans General Hospital, Taipei, Taiwan.
This study is the cross-sectional, observational database analysis. We limited our analysis to prescriptions given to children not older than 18 years who were then grouped into infant (younger than 2 years), preschool child (2 to 5 years), child (6 to 11 years), and adolescent (12 to 17 years). To simplify the analysis, we further limited the analysis to drug items containing acetaminophen which was the most frequently prescribed ingredient as antipyretics and analgesics in children . It was generally recommended that in patients younger than 13 years the dosage of acetaminophen should be adjusted according to body weight .
To calculate the dosage of acetaminophen in each prescription, we at first identified 288 drug items containing acetaminophen from the web site of the BNHI. The exact single dosage in unit of milligram was computed by multiplying the strength with the prescribed quantity, in either single-ingredient or compound preparations. In the analysis, the various drug forms were grouped into liquid (22 items, including syrup and elixir), tablet (200 items, including sugar-coated tablet and chewable tablet), capsule (62 items), and others (4 items, including oral granules and suppository). According to the reimbursement regulations of the NHI, liquid forms of drugs could only be prescribed to children under 13 years or on special situations. Furthermore, we analyzed only the regular prescriptions. The prescriptions as statim, pro re nata and ut dictum as well as outliers with dosages greater or less than 1.5 interquartile ranges were excluded from analysis.
To make a cross-specialty comparison of dosing variability, we chose only three major specialties: pediatrics, family medicine and otolaryngology. A recent study about child care within the NHI in Taiwan revealed that these three specialties covered 93.3% of ambulatory visits by children younger than 7 years . Because people in Taiwan could freely choose physicians and healthcare settings without referral and the copayment was relatively low within the NHI, otolaryngologists became the second largest specialty in providing ambulatory health care, especially in treating diseases of the respiratory system that accounted for one third of ambulatory visits . In Taiwan, the share of otolaryngology in ambulatory care could reach 9.5% in general population  and as high as 20.7% in children .
Although the NHI claims were lacking in the data of body weight, we supposed that the patients in each of three specialties have similar distributions of body weights. On this assumption, we would compare the distributions of acetaminophen dosages prescribed by physicians in three specialties. Our hypothesis was that pediatricians should be more likely to prescribe on the basis of a patient’s body weight. The dosing variability by pediatricians should be more marked than that by family physicians or otolaryngologists. The extent of difference in dosing variability between family physicians/otolaryngologists and pediatricians might serve as a quality indicator of prescribing.
The database management software of Microsoft SQL 2008 (Microsoft Corp., Redmond, WA, USA) was used for data linkage and processing. At first, descriptive statistics were presented. Besides the mean and standard deviation (SD), we used the coefficient of variation (CV) to represent the dosing variability of acetaminophen prescriptions. The CV, defined as the ratio of the standard deviation to the mean, was thought to be more accurate than standard deviations to compare the variability  and usually applied to represent laboratory values, e.g. red blood cell distribution width in complete blood count. The results were stratified by child’s age group, prescriber’s specialty and dosage form of acetaminophen preparation. Then, the Kolmogorov-Smirnov test was used to explore whether the single dosages of acetaminophen prescriptions in each age group and prescriber’s specialty followed the normal distribution. The Wilcoxon rank-sum test was used to compare the distribution of acetaminophen dosages in family medicine or otolaryngology with that in pediatrics. A value of p < 0.05 was regarded as statistically significant. Statistical analyses were performed using the IBM SPSS statistical package, version 19.0.