Correspondence to: Dr Rama Bhunia, 60A/1R, Dr S C Banerjee Road, Kolkata, West Bengal 700 010, India.
Email: [email protected]  

In India, many therapeutic injections are unnecessary and unsafe(16-23). We conducted a study to determine whether an "interactional group discussions" intervention was effective at reducing injections use among public sector medical officers in India.

Study population: Medical officers, formally trained with MBBS degree, working in government rural health care facilities of the North 24 Parganas district, West Bengal, India.

Operational definitions: A therapeutic injection was defined as a skin-piercing procedure performed with a syringe and needle to introduce a curative substance into a patient by the intramuscular, intravenous or subcutaneous route. We excluded blood transfusions, surgery, tattoos and vaccines.

Study design: We conducted a randomized controlled trial from June to November 2007. The study consisted of three stages: (i) a pre-intervention prescription survey; (ii) the "interactional group discussions" intervention four weeks later during which medical officers were confronted to community members (who according to past experience with the method usually express to prescribers their absence of preference for injections); and (iii) a post-intervention prescription survey two months later. We randomized medical officers to the intervention and control groups using a computer-generated random number list and used quality assurance procedures to prevent any mix-up after randomization. While we made use of qualitative methods for the intervention, the study was quantitative in nature as we aimed at measuring a reduction in injection use. We conducted seven interactional group discussions to distribute medical officers in groups small enough to interact with community members.

Sampling: We sampled medical officers on the employment list from the 135 working in rural facilities (i.e., block primary health centres and primary health centres) (Fig. 1). For each selected medical officer, we sampled every third prescription on the day of the survey. We calculated the sample size on the basis of the number of prescriptions to be analyzed before and after the intervention, using the Epi Table Software. Our assumptions included a proportion of prescriptions including at least one injection of 50% and 40% before and after the intervention, respectively; an alpha error of 5% and a power of 99%. As the software generated an estimate of 925 prescriptions, we anticipated 5% of non-response and therefore aimed at surveying 971 prescriptions before and after, for the intervention and the control group. To round up and obtain an even count, we included 30 prescriptions for 36 medical officers in the intervention and control groups, before and after (30×36 = 1,080, twice)

Fig. 1 Diagram summarizing the randomized controlled trial.

Data collection and analysis: Health workers unaware of the intervention/ control status surveyed prescriptions of medical officers from outpatient register without interviewing the medical officer. The principal investigator verified 15% prescriptions. We calculated the crude frequency of prescriptions, injections, the WHO "OT8 indicator" (i.e., the number of prescriptions including at least one injection), the effect size [(% after - % before) intervention – (% after - % before) control] and relative risks. We tested the significance of the differences in the ratio of injections per 100 prescriptions before and after in the two groups using a paired MacNemar Chi square(24).

Human subjects protection: We explained the voluntary nature of participation and collected written informed consent from participants, including medical officers (who were the target of the intervention) and community members who took part in the discussions. We ensured confidentiality through codes. The ethical committee of the National Institute of Epidemiology, Chennai cleared the protocol.

TABLE I



Proportion of Prescriptions Including at Least One Injection Before and After the Intervention

After the intervention, the ratio of injections per 100 prescriptions decreased to 11 and 21 per 100, in the intervention and control groups, respectively (P = 0.009). The proportion of prescriptions including at least one injection decreased in both the intervention and control group (Effect size: -11%). However, in the control group, the reduction was larger among the 15 medical officers who shared their assignment location with a medical officer who was part of the intervention group than among 21 others (RR: 0.47, 95% CI: 0.36 – 0.63 vs. RR: 1.1, 95% CI: 0.86-1.4, respectively).

Before the intervention, a high proportion of prescriptions included injections in both the intervention and control groups. After intervention, the proportion of prescriptions including at least one injection reduced significantly in both groups. However, the decrease was more marked in intervention group. Furthermore, in the control group, the reduction was more marked among medical officers who were assigned at a facility where a medical officer of the intervention group worked.

Our study had two main limitations. First, the post-intervention survey was conducted only once. Unlike the Indonesian study, we were unable to conduct those at regular intervals for a prolonged period to assess the sustainability of the intervention. As a result, we could not conclude about the sustainability of the effect we measured. Second, our post-intervention prescription survey was based on outpatient registers only and was not validated by observations. Thus, we could not exclude that the difference among groups exposed and unexposed to the intervention was in fact a consequence of a desire to satisfy the interviewer or of an observer-induced bias ("Hawthorne effect"). However, even if that was the case, the differences in two groups suggest that the intervention was at least successful in communicating the acceptable standards to the target audience.

Our results support the findings of three comparable trials in other countries and suggest that ‘interactional group discussions" were also effective as an intervention to reduce prescriptions of injections in the rural public health care facilities in West Bengal, India. Our study also suggested that prescriber-to-prescriber dissemination of the outcome of discussions could be effective by itself in reducing injection prescription habits. On the basis of these conclusions, we recommended scaling up similar intervention through "interactional group discussions" sessions for more medical officers. Dissemination of this message should help reducing injection use and prevent injection-associated infections.

Dr Tapas Sen, State Public Health Cell and Dr Kusum Kumar Adhikari, Chief Medical Officer of Health, North 24 Parganas District of the Department of Health and Family Welfare, Government of West Bengal.

Contributors: BR: Concept, design, acquisition of data, analysis, interpretation and writing of manuscript; HY: Concept, design, analysis and critical editing of manuscript; RR: Design, analysis and interpretation; GPK: Acquisition of data, analysis and critical revision of manuscript; DS: Acquisition of data, interpretation and writing of manuscript; MM: Analysis, interpretation and writing of manuscript.

Funding: ICMR, WHO, Government of West Bengal.

Competing interests: None stated.