DATA COLLECTION TECHNIQUES

6. DATA COLLECTION TECHNIQUES

6.1. Abstracts of the medical record

Manual collection. Medical and nursing records are analysed in the wards. The following demographic data should be collected: identification of the patient (code number), age, sex, weight, identification of the prescriber (code number). The (presumed) diagnosis of the infection. The generic and brand name of antibiotics prescribed, the route of administration, the unit dose, the dates (and hours) of the start and stop of treatment. If the exact time of admin- istration is not known, the hours of the nursing rounds can be used as an approximation. For the audit of surgical prophylaxis, the diagnosis and type of surgical procedure can be retrieved from medical records or from surgical intervention reports. Details supporting the suspected or confirmed diagnosis of infection (history, clinical findings) and laboratory data are needed in order to study empiric or definitive treatments. Depending on the degree of comput- erisation of the hospital, much of this information can be retrieved from the central hospital computer.

6.2. Interviews

Generally there is no contact between the prescribers and the researchers. Few authors have used the interview technique (Moss et al., 1981). Interviews with prescribers can in itself act as an intervention.

6.3. Antibiotic Order Forms

This prescription sheet, filled in by the prescriber, forces the prescriber to review clinical information and laboratory results in order to specify whether therapy is started on an empiric, a definitive, or prophylactic basis; to name a (suspected) causative microorganism, the spectrum needed, the dose, the fre- quency, and the duration (Durbin et al., 1981). This approach has mainly been used for some specific restricted drugs (Thuong et al., 2000). By filling in the Antibiotic Order Form, the prescriber provides the data for the audit. In return, the pre-printed drug information on the form facilitates prescribing by provid- ing information on formulary antibiotics, standardised dosing regimens at the time the prescription is written. In this way, an antibiotic form acts in itself as

Audits for Monitoring the Quality of Antimicrobial Prescriptions 215 an intervention. The antibiotic order form may be compulsory (Durbin et al.,

1981; Echols et al., 1984) or voluntary (Gyssens et al., 1997). Some centres have reported successes with such a form (Echols et al., 1984; Gyssens et al., 1997; Lipsy et al., 1993; Soumerai et al., 1993; Thuong et al., 2000) while oth- ers have not. Recently, authors have developed a “Vancomycin Continuation Form” in order to monitor the use of vancomycin (Evans et al., 1999). Another advantage of the form is automatic stop order, for example after 24–48 hr in case of prophylaxis (Echols et al., 1984; Lipsy et al., 1993) or after 72 hr of empiric therapy (Lipsy et al., 1993).

6.4. Automated methods—computer-assisted prescribing

For 15 years, in the university hospital LDS of Salt Lake City, a computer program “Clinical-decision-support” has been in use for assisting clinicians with antimicrobial drug prescribing (Evans et al., 1986). Clinicians prescribe drugs with the help of a computer (Evans et al., 1998). The “errors” of prescribing are recorded and they allow a continuous evaluation of process parameters (aller- gies, dosages, costs . . .) and patient outcome parameters such as side effects, length of stay. The program is linked to microbiology laboratory results and provides alerts to identify patients with inappropriate therapy due to suscepti- bility mismatches (Pestotnik et al., 1996). The LDS pharmacy database has provided several evaluations of quality interventions in intensive care (Evans et al., 1998) and on the timing of surgical prophylaxis (Classen et al., 1992).

A similar program, “Computer-based expert system for quality assurance of antimicrobial therapy,” which produces reports of the discrepancies between prescribed therapy and microbiology results was described in 1993 by hospital pharmacists of a US reference hospital (Morrell et al., 1993). Other reports on automated prescribing programs are scarce. German authors have published their experience in intensive care with a program “Computer-assisted infection monitoring program” developed with the support of Bayer to improve the choice of empiric therapy (Heininger et al., 1999). Another group has reported an “Antimicrobial prescribing program” with a historic control (Frank et al., 1997).

6.5. Use of surveillance data as a quality measure

Data collected at individual patient level provide the most reliable informa- tion on antibiotic use regarding the population exposed. However, since data at the individual level are not often available, the majority of studies on antibiotic consumption present data that have been collected at the collective level. These

216 Inge C. Gyssens data are now preferentially expressed as defined daily dose (DDD) (World

Health Organization, 2003) per 100 or 1,000 patient-days. On the other hand, in quality audits, when individual patient data are retrieved from medication charts at the wards, consumption is traditionally expressed in grams or exactly prescribed daily doses (PDD). To allow comparison of quantitative audit data with surveillance reports, consumption measured in audits should also be con- verted in DDD. Combining quantitative and qualitative methods of evaluation allows one to calculate inappropriate or excessive consumption either in grams (Parret et al., 1993), in DDD/100 patient-days (Gyssens et al., 1996a, 1997; Røder et al., 1993) or both (Hamilton et al., 2000). Quantitative surveillance with interrupted time-series analysis can be used as a method for identifying wards with changes in use and for targeting more detailed audits (Ansari et al., 2003). Finally, by calculating the PDD/DDD ratio, audits can reveal to what extent the local dosage in a particular ward differs from the DDD.

Example: The DDD of cefazolin is 3 g. In a surgical department, an audit reveals that cefazolin is uniquely used for prophylaxis in a single preoperative dose of 1 g. The PDD/DDD ratio in this ward is 0.33. The PDD/DDD ratio shows that three times more patients are exposed to cefazolin than would be expected from the number of DDD/100 patient-days.