The surgical pathology laboratory in Mwanza, Tanzania: A survey on the reproducibility of diagnoses after the first years of autonomous activity

Background: In 2000, an Italian non-governmental organisation (NGO) began a 9-year project to establish a surgical pathology laboratory at the Bugando Medical Centre (BMC) in Mwanza, Tanzania, a country with a low Human Development Index (HDI), and as of 2009, the laboratory was operating autonomously. The present survey aims to evaluate the reproducibility of histological and cytological diagnoses assigned in the laboratory's early years of autonomous activity. We selected a random sample of 196 histological and cytological diagnoses issued in 2010-2011 at the BMC surgical pathology laboratory. The corresponding samples were sent to Italy for review by Italian senior pathologists, who were blinded to the local results. Samples were classified into four diagnostic categories: malignant, benign, inflammatory, and suspicious. The two-observer kappa-statistic for categorised (qualitative) data was then calculated to measure diagnostic concordance between the local Tanzanian pathologists and Italian senior pathologists. The k-Cohen was calculated for concordance in the overall study sample. Concordance and discordance rates were also stratified by subset: general adult, paediatric/adolescent, and lymphoproliferative histopathological diagnoses; fluid and fine needle aspiration (FNA) cytological diagnoses; and PAP tests. Discordance was also categorised by the corresponding hypothetical clinical implications: high, intermediate, and not significant. Results: Overall concordance was 85.2% (167 of 196 diagnoses), with a k-Cohen of 0.7691 (P = 0.0000). Very high concordance was observed in the subsets of adult general pathological diagnoses (90%) and paediatric/adolescent pathological diagnoses (91.18%). Concordance in the subset of PAP tests was 75%, and for fluid/FNA cytological diagnoses it was 56.52%. Concordance among 12 histological subtypes of lymphoma was 75.86%, with substantial discordance observed in the diagnosis of Burkitt lymphoma (five cases diagnosed by Italian pathologists versus 2 by local pathologists). The overall proportion of discordance with high hypothetical clinical implications was 6.1% (12 diagnoses). Conclusion: This blind review of diagnoses assigned in Tanzania, a country with low HDI, and in Italy, a country with a very high HDI, seemed to be a sensitive and effective method to identify areas of potential error and may represent a reference point for future, more detailed quality control processes or audits of surgical pathology services located in limited-resource regions.