We found the design of our QC protocol to be practical in field settings and easily understood and implemented by laboratory staff with limited training. We attribute this to a combination of a small QC sample size, a fixed number of slides independent of the workload, and use of simple percentage agreement for statistical analysis. The small sample size considerably decreased the QC workload while maintaining statistical reliability by using targeted sampling of only weakly positive slides and 4-month cohort analysis.
Our findings show a significant improvement in the accuracy of malaria and AFB microscopy comparing the periods May–December 2005 and January–August 2008. We attribute this improvement to the strengthening of our protocols, field support, and training over this period. However our QC protocol also played a central role by providing key information on a timely basis allowing us to prioritize those laboratory support activities. Also, and we believe critically, the reporting of compiled data back to the field provided the laboratories with clear performance indicators and enabled field laboratories to directly compare their performance against other laboratories working in similar circumstances. In our experience, this generated an environment of positive “competition” among laboratories that we believe has also contributed significantly to the improvement in laboratory quality performance.
For malaria microscopy, the number of FP and FN results decreased markedly. We attribute this to active follow-up of poorly performing laboratories identified by the QC protocol. In contrast, the frequency of FN results for AFB microscopy did not change significantly, and the improvement in percentage agreement reflects the decrease in the frequency of FP results. Laboratories for AFB also entered the analysis period at a higher level of performance compared with malaria microscopy (59.1% of AFB cohorts achieving ≥95% percentage agreement for May–December 2005 compared with 32.3% for malaria). This may be because AFB microscopy is relatively easier to perform than malaria microscopy as accurate malaria microscopy requires greater microscopy resolution and has a technically more demanding staining procedure.
However we also speculate that the random selection of negative AFB smears, which is the standard methodology for AFB QC protocols and is used in our protocol, may be problematic. Saliva smears are in general more likely to be negative or have an AFB density below the threshold of microscopy detection than sputum smears 
. Therefore there is less opportunity for QC to detect FN results by reexamining saliva slides as they have a higher prior probability of being truly microscopically negative than a sputum smear. With random selection, laboratories with a high proportion of saliva samples in routine practice will also have a high proportion of saliva slides in their QC sample, and therefore the QC FN frequency for such laboratories may be lower than their true FN frequency.
For the future, we are currently incorporating clerical error monitoring into our laboratory QC protocol, as this can also be a major source of error. With the increasing emphasis on disease eradication, we are also developing QC protocols to accommodate low positivity. Finally, we have implemented a pilot study to exclude saliva smears from the AFB QC sample.