Educational outreach to general practitioners reduces children's asthma symptoms: a cluster randomised controlled trial

The design was a cluster randomised controlled trial with the general practices as the unit of randomisation, intervention, and analysis. The study conforms to the Consort group recommendations for such trials [11] and was analysed on an intention to treat basis (see additional file 2).

A list of practices was composed in 1998, after identifying practitioners named in the baseline data from the medical register and telephone directory. Practices within the study area were numbered and randomised to two groups using a computer-generated list of random numbers.

We included all general practitioners, both practice principals and their hired doctors, working in private practice in Mitchells Plain. There were no multipartner practices. We included all schoolchildren up to age 17 living in the study area and their preschool siblings two years old or older with moderate to severe asthma. We determined their eligibility based on their answers to parent self-administered questionnaires and confirmatory face-to-face interviews conducted three months later.

The intervention was a tailored, multifaceted educational outreach intervention based on qualitative and survey research that identified barriers to the appropriate medical diagnosis and treatment of children with asthma in a similar nearby community (Table 1) [12]. It was aimed at improving the diagnosis, prescribing and follow-up care provided by private general practitioners to children with asthma.

The intervention contained eight key messages to convey to practitioners. We included only messages related to clinical behaviours that we believed to be largely under the control of the practitioner; in other words, free of external constraints and thus amenable to change by the practitioner (Table 2).

The intervention was delivered during 1998 to individual practitioners by a pharmacist trained in the methods of academic detailing. A first visit took 30 minutes with a repeat visit of similar duration conducted three months later. At the first visit, the pharmacist used a visual aid, a set of printed glossy materials similar to those used by pharmaceutical company representatives, structured as a plastic laminated desk blotter, on which the key messages were outlined (Figure 1). The blotter was left behind in the practice, along with instructions for modifying a 500 ml plastic soft drink bottle to attach to a pressurised metered dose inhaler as a volume increasing spacer and an actual example of one such spacer. (Spacers reduce the difficulties children have in coordinating their breathing with triggering of the inhaler).

The principal outcome for the trial was the change in an individual child's asthma symptom score reported by the parent or guardian before and after the intervention. The South African Consensus guidelines use a severity grading of childhood asthma based on frequency of attacks of tight chest, nocturnal coughing, and nocturnal waking, hospital admissions, and peak flow rate. The three attack frequency questions were amenable for use in a brief interview, and were weighted by zero to three points, according to frequency of episodes in the previous 12 months, using weights developed in a previous study in this community [5], where one to two episodes equalled one point, three episodes equalled two points, and four or more episodes equalled three points (see additional file 1).

The score obtained for these three frequency questions was added to give a total score. The maximum score attainable was nine points and the minimum score attainable was zero points. Children with the highest pre-intervention symptom scores were included until the desired sample size for the trial was obtained. In contrast with the specific symptom questions in the score, we also asked parents a number of questions designed to measure their perception of the child's asthma severity and the effect of asthma on participation in usual activities such as school (see additional file 1).

Initially, parents completed a self-administered asthma screening survey for their primary school children and younger siblings, distributed via the primary schools in the study area. In later rounds, parents of those children identified from the returned screening questionnaires as having moderate or severe asthma symptom scores and a regular private doctor were interviewed face-to-face in their homes by experienced and trained fieldworkers before the intervention and one year later (1999). Self administered questionnaires and interviews were completed in the respondents chosen language. Interviewers and parents were blinded as to the allocation of practitioners.

The process was designed to obtain a group of children whose parents consistently reported their usual source of care as one or more participating private practitioners, and whose pre-intervention symptoms were compatible with moderate to severe asthma. We thus excluded at screening children whose parents reported no usual private general practitioner, or whose home address was outside Mitchells Plain. At the baseline face-to-face interview, we also excluded children whose parents reported that their child's usual family practitioner was outside the study area, and at follow-up face-to-face interview, we further excluded children whose parents identified as their practitioner a doctor who was not on our list of randomised practitioners.

We wished to detect a clinically meaningful improvement, 0.5 standard deviations, in the symptom score between intervention and control groups [13]. We assumed a standard deviation of one, 5% significance, 80% power, and an intracluster correlation coefficient of 0.17 (similar to primary care practices in other countries [14]). With 43 available clusters, we needed 15 patients per cluster, for a total of 280 patients [15].

Data were collected on paper and entered into a computer. They were managed and analysed using SAS (Version 8.2. SAS Institute Inc., Cary, NC, USA). We report univariate descriptions, cross-tabulations with chi-squares, and, for adjusted analysis of the principal outcome, the asthma symptom score, we report a linear regression analysis on the change in asthma symptom score from pre-to post-intervention evaluation. The clustered design was accommodated by fitting the generalised estimating equation version of the linear regression model with an exchangeable working correlation model. To check the findings of the linear model on the change scores, an ordinal logistic regression analysis using the proportional odds model was carried out on the post intervention score [16]. Since the results were corroborated, only the linear model with change scores is reported. The variables included in the model, decided a priori, were the baseline score and the number of visits to the usual practitioner during the study period. We conducted the adjusted analysis using Proc GENMOD in SAS.

Permission to contact parents of schoolchildren to complete a survey was obtained from the Department of Education for the province, and then, during an information meeting at each of the primary schools in the study area, permission was sought from, and granted by, each school principal and class teacher.

Plain language explanations of the purpose of the study and its voluntary nature were included with the self-administered questionnaire sent to parents. Return of the questionnaire indicated consent to use the data. During home visits, the study was again explained and confirmatory verbal consent was obtained.

Intervention group practitioners were contacted to obtain an appointment for the academic detailing visit. At the beginning of the visit, some well-known problems of asthma care and the academic detailing intervention were outlined, and the doctor's participation in an evaluation of the intervention was invited. No monetary reward was offered. Control group practices received a hand-delivered copy of the then current South African childhood asthma guideline. Because practitioner records were not used for identification or follow-up of children with asthma, no permission was required from practitioners for this element of the study.

Ethical approval for the study was granted by the Medical Research Council of South Africa Ethics Committee.

The principal outcome measure for the trial was the change in asthma symptom score over the one year between baseline and follow up surveys, during which period the intervention took place. There was substantial decline in reported symptoms over one year in the intervention group (4.08) and the control group (3.24) (Table 5). The decline in symptom score was 0.84 points greater in the intervention group than in the control (p = 0.03).

At baseline, the mean asthma symptom score was higher in the intervention group, suggesting a slightly more severe distribution of disease in that group (Table 5). Adjusting for the baseline difference using ordinal logistic regression produces an odds ratio (rather than a mean difference) that is consistent with the unadjusted results (OR = 1.48, 95% CI 1.00 – 2.20, p = 0.049).

To investigate the effect on the principal outcome measure of the number of visits to the specified physician, this factor was added to the model as a linear effect. The estimated intervention effect (Table 5) was close to the unadjusted intention to treat analysis, but with a narrower confidence interval. Children with more frequent physician visits had a non-significant tendency towards smaller symptom changes (slope of -1.41, p = 0.10). The slope was similar in both intervention and control groups.

There were no significant differences between the intervention and control group respondents in their subjective assessment of their children's overall asthma severity in comparison with the previous year, nor in their ability to undertake normal school activities (Table 6).