Methodological issues in using drugs data in epidemiology.

Using drugs data in epidemiological studies is becoming fashionable with the large databases available today. We describe the philosophical basis and the main methodological issues. The process leading a patient affected by a disease to be treated with an appropriate drug appears to be complex, and influenced by external and internal many factors, described here. The disease-drug use relationship is disentangled describing its main characteristics. Drug use data means the information concerning the use of drugs, which encompasses any form of record, on paper or computerised, collected from patients, doctors prescribing the drug, pharmacists dispensing it, the drug company that produces it, or its distributor. The possible sources of drug use data are described, along with the possible strengths and weaknesses of each source. Various examples of uses of drug utilisation data as indicator in epidemiological researches are presented. Drug utilisation studies are a recent issue of public health research, merging knowledge of public health doctors, epidemiologists, pharmacologists and health economists. Automated databases demonstrate nowadays to be a valuable approach to drug utilisation investigation. In conclusion, the aim of any forward looking health care system should be to collect and freely circulate these detailed information, allowing not only useful health care planning but also indeep, population based, health research.

The role of drug use data in public health research.

The growing importance of drug consumption automated databases in public health research is summarised in this research. Dispensed drugs data, and drug consumption data more in general, are a relatively new source of information for health research. This analysis takes into consideration the known uses of drug consumption data in various fields of health related research. Pharmacoepidemiology examples of adverse drug reaction studies are presented along with drug prescription surveillance and analysis of prescribing patterns, useful tools for health services evaluation as well as health economics, from the simple evaluation of health care burden for drug expenditure to the various studies of cost of drugs. Other fields of use are prevalence and incidence epidemiology studies, or causal inference studies using drugs data as health indicators. Drug utilisation automated databases, organised and maintained in large areas, with individual information, seem to improve the feasibility and usefulness of this kind of studies. It reduces the need for time, money and people to perform valid scientific investigations, in epidemiology as well as in other areas of health research.

Aetiology and prognosis of bacteraemia in Italy.

A prospective multi-centre study was conducted to assess the microbiological pattern and prognostic factors of bacteraemia and their impact on clinical outcome. All patients admitted to 41 Italian hospitals over 2 months, from whom one or more clinically significant organisms were isolated from blood culture, were studied according to a standardized protocol and case definition. A total of 156 episodes of bacteraemia were identified in 20,601 patients. There were 3.9 episodes of nosocomially acquired bacteraemia and 3.7 episodes of community-acquired bacteraemia per 1000 admissions. The most frequent pathogens isolated were Gram-negative bacteria (44.9%) but Gram-positive species accounted for 40.4 % of episodes. Fungal infections due to Candida spp. were found in 3.8 % of episodes, and multiple pathogens were recovered from 9.6% of episodes. The clinical response to bacteraemia was classified as sepsis in 90 episodes (577%), severe sepsis in 21 (13.5%) and septic shock in 26 (167%); 19 episodes (12.2%) showed no clinical response. The total in-hospital mortality was 25.0%. By multivariate logistic regression, the variables which independently predicted mortality were increasing age, the presence of septic shock, infection with Gram-positive bacteria or fungi and nosocomial acquisition.

Three scenarios of clinical claim reimbursement for nosocomial infection: the good, the bad, and the ugly.

We studied the extent to which hospitals can expect to receive reimbursement for costs relating to nosocomial infections (NI) under the diagnosis-related groups (DRG) system of clinical claims and calculated the loss of reimbursement due to missed or incorrect registration of infective complications on hospital discharge records (HDR). We calculated clinical claim reimbursement in three scenarios: the good, in which all NI are recorded on HDR; the bad, in which a proportion of NI recorded on HDR observed at the 41 participating hospitals; the ugly, in which none of the NI are recorded on HDR. We analysed in which patients the recording of infective complications changed the DRG clinical claim and the economic consequences on reimbursements. Compared with the ugly scenario, the bad scenario, which is closest to what actually occurs, with only 55.9% of NI (180/322) properly recorded, produced an increased DRG clinical claim in 30 cases, of on average 403 for every NI. Compared with the ugly scenario, the good scenario, produced an increased DRG clinical claim in 45 cases with an average reimbursement of 618. The difference between the bad and the good scenarios shows an average loss of 215 for every case. Our calculated good scenario could cover only 3.8% of direct costs per case attributable to NI. Real, tangible benefits in health, both social and economic, will only accrue from the monitoring and control of NI in hospitals.

[Antiasthmatic drug consumption as an indicator of the prevalence of respiratory pathology in a pediatric population]. / Il consumo di farmaci antiasmatici come indicatore di prevalenza di patologie respiratorie nella popolazione infantile.

The study of disease prevalence requires specific and sensitive indicators, which are hardly gathered at population level. The use of anti-asthma drugs, which are relatively specific for diseases characterised by bronchospasms, has been already experimented in the USA with algorithms describing patients affected by asthmatic disease by their use of drugs. We retrospectively analyse spatial variations of drug use as dispensed by the SSN (National Health Service) to Lombardy children less than 15 years old, between 1st January 1995 and 31st December 1997, estimating prevalence of use as a proxy of disease prevalence. Two algorithms already experimented and a new one have been employed to select children treated with antiasthmatic therapy (R03 of ATC classification) dispensed. Local communities' data were used for basic demography. Number of daily defined doses (DDD) dispensed has been used as indicator of drug quantity. Euro/year per person has been used as indicator of costs. Small area distribution analysis has been performed with a Geographic Information System at local Communities level. Frequency of patients (of a total of 1,252,958 children): with at least one dispensed drug in the three years was 26.6% (SD 10.5), DDD per person per month were 1.21 (SD 0.78), Euro/Year per person 6.84 (SD 3.61); with the association of a anti-inflammatory and a b2-agonist was 8.2% (SD 5.6), DDD per person per month were 3.01 (SD 3.07), Euro/Year per person 15.87 (SD 15.07); with at least 90 days of DDD in the three years was 2.6% (SD 1.3) DDD per person per month were 7.79 (SD 4.48), Euro/Year per person 37.66 (SD 21.83). It must be taken into account that, in the best of cases, these approximation refers to the prevalence of all diseases, characterised by bronchospasm, treated with the selected drugs. Estimated data of prevalence are comparable with those of other authors. Prevalence of drug use appears to have relevant geographical differences. Taking into consideration these comments, prevalence of drug use and consequently the estimated disease prevalence appears to be worryingly high, having as well wide geographical differences.

Prevalence and patterns of antibiotic prescribing in Italian hospitals.

Use of antimicrobial drugs is associated with the emergence of resistant bacteria, increasing hospital expenditures and requires an assessment of appropriateness. The aim of this study was to investigate prevalence and patterns of antibiotic prescribing and expenditures in Italian hospitals. A prevalence survey was performed in 15 Italian hospitals. Data on antimicrobial use and expenditures in the year 2000 and 1-day point prevalence information on antibiotic prescription, type and sources of infections and expenditures were obtained through a review of pharmacy records and charts of 2,165 inpatients. 45.5% of the patients surveyed were receiving at least one antibiotic: 148 for hospital-acquired infections (HAI), 262 for community-acquired infections (CAI). No active infection was recorded for 575 patients (58.4%). 57.3% of prescriptions for HAI and 65.9% for CAI were based on empiric and presumptive criteria, while 39% of HAI and 15.8% of CAI were oriented by microbiology and susceptibility testing results. Guidelines were followed in choosing the antimicrobial drugs for 22.8% of therapeutic and 46.8% of prophylactic courses. High rates of antibiotic usage were observed in Italian hospitals. Areas for improvement, both for the use of diagnostic resources and the prescribing habits, were identified. The 1-day prevalence survey of antimicrobial use may be a tool of feedback to prescribers for more appropriate drug selection.

What relevance do advances in molecular biology and genetics have for epidemiology: high tech needle and thread to sew the web of causation.

Epidemiology has changed, during the past, from miasma versus germ theory to the studies of relationship between exposure and outcome. New definitions such as risk factors or multifactorial aetiology, and more recently web of causation, have been helpful to understand many diseases. Molecular biology and genetics advances in the past decade can be instruments that allow to investigate more deeply the web of causation. Molecular biomarkers are the new investigating mean. Examples of studies using these new techniques are presented, in many epidemiological fields: in non communicable disease with linkage and association studies, allowing experimental studies; in cancer, investigating susceptibility to the disease, early exposure to environmental carcinogens detection and specific mutational patterns in aetiological studies; in communicable disease, in particular with genetic subtyping of particular agents and causal mechanisms research. To mix up all the knowledge required to handle these techniques and epidemiology shall be profitable integrate multidisciplinary team of scientists, epidemiologists and molecular biologists together. This might be the way to approach the distribution and determinants of disease in deep but also with more clear and explicit explanation of the elaborate relationships.