We undertook to document the trend in notification and outcome of children TB cases in Kampala city. Our findings show a non-significant declining trend of TB notification rate over the study period. The outcomes of TB varied over the study period with no significant increase in the treatment success rate and the average treatment success rate not reaching the national target of 80%. The average mortality rate was highest among the under-fives at 9% almost double the ≤5% WHO acceptable case fatality ratio to achieve the 2025 milestone for reduction of TB deaths  .
Our results show a modest declining trend in TB notification rate among children reported in Kampala. We noted that this decrease was not uniform with children under 5 years showing a significant decline in notification rate but a stable trend in the 5 to 14 years. This finding is consistent with what we earlier reported for the 2009 and 2010 . Our results are also comparable to the reported global trend in world TB report 2017 that shows a modest overall decline . We note the higher proportionate TB notification in age group 0 to 4 years over the study period attributable to the increased focus by the national program on TB diagnosis in the under-fives. Since appointing of a paediatric TB focal person at the NTLP, there has been greater impetus in TB case notification among children. A Paediatric TB curriculum was developed and systematic training and mentorships undertaken. A cluster randomized study showed that systematic training increased TB notification in children by three times from the baseline . We note the documented steady decrease of notified TB cases among adults . The decrease in the number of adults with TB represents a lower transmission chance most especially in the home environment  and this best reflects in the under-fives that have the highest risk of infection. In this report, we show that children represent about 7% of reported TB in Kampala city but also observe a decrease in the proportion of children over the years though the numbers increased. Compared to the projected estimate of 15% of all TB among children, the current performance remains below the expected . The value addition of this paper above our previous work  is demonstration of trend of outcome of TB in children. Our previous work had 2-time points that could not allow us to show trend. This work also covers a period in which several interventions to increase TB case finding in children took place as opposed to of our previous work where minimal took place.
Most of the TB in children (80%) was pulmonary of which the clinically diagnosed represented 82%. This observation might be because of to the extra efforts to train health workers in childhood TB diagnosis and the emphasis on using algorithms to make a clinical diagnosis. Only about 7% of pulmonary bacteriologically confirmed cases had smear microscopy done of whom many (45%) were smear positive. We can speculate this to have resulted from the limited ability of health workers to collect sputum samples. The higher smear positivity rate suggests a selection bias towards older children who are more likely to provide a sputum sample with higher bacillary load. This low sputum collection rate implies the Xpert® MTB/RIF which is the recommended first test for TB has low uptake. We note that even with emphasis on Xpert® MTB/RIF as first test, the bacteriologically confirmed constituted 18% of the PTB cases. This compares with 15% reported in our previous work when the Xpert® MTB/RIF was not readily available . This small difference may suggest underutilization of the Xpert® MTB/RIF in children whose underlying reasons are outside the scope of this work.
The average proportion tested for HIV was still low at 79% over the study period compared to the national target of 100%. There was however a significant increase from 56% in 2011 to 100% in 2015 and all the tested children received their results. This suggests the efforts over the study period in TB and HIV integration were yielding results. The average TB and HIV co-infection was 35% and remained in the same range over the study period with greatest decline seen in the 5–9 years. Data on the prevalence of HIV in children with TB are rare but available literature shows 5 to 56% in different settings . Previous work in our setting reported an HIV prevalence of up to 49% among children with TB [11, 18] therefore our findings show a significant decline in HIV burden among children registered for TB treatment. This could be a reflection of reduced mother to child HIV transmission that has reduced new infections in children from 27,660 in 2011 to 9629 in 2013 . Alternatively, it could be HIV positive children with TB die before diagnosis since they have a higher hospital related mortality . Recent work by Dodd et al. showed the odds of HIV in cohorts of children with TB compared to that of children without TB was 7 times . We found good CPT uptake above 95% and ART uptake increasing from 24% in our previous work  to the reported 91%. This is a good sign of integrated TB and HIV care services in accordance to with the Uganda national guidelines.
The average treatment success rate over the study period was 77%. We noted variations in all age groups but 0 to 4 years and 5 to 9 years had declining treatment success rate over the study period. A report from Malawian hospital TB records in 1998 showed poor treatment success rate of less than 43% in children below 5 years and 54% in those above 5 years . In our study, we noted the direct linkage between many cases of loss to follow, high mortality and treatment success rate. We hypothesize this results from the drive to diagnose many more children without satisfactory human resources to support adherence leading to high loss to follow up and poor completion rates. The 10 to 14 years showed the highest loss to follow up in the earlier years of the study period. In most children, diagnosis is made on clinical grounds and where parents do not agree, adherence to treatment remains a challenge. We noted a significant decline in loss to follow up over the study period. The explanation for this significant decline in loss to follow is beyond the scope of this study. Our results show that mortality in the 0 to 4 years age-group increased as the notifications increased. Most of these deaths are likely from the hospital setting where TB diagnosis in children is commonly made late. There is evidence that many children die either as a result or with underlying undiagnosed TB . Since we document a high death rate, we hypothesize that TB in this 0 to 4 years age-group is more likely to be severe and diagnosed late usually at the referral health unit. There is evidence that children with TB present with other conditions most especially severe pneumonia where TB is an afterthought leading to late diagnosis of TB [24, 25]. Harries A et al. reported a higher mortality in the under 1-year and those below 4 years compared to other age groups . A recent systematic review by Jenkins et al.  showed a low mortality rate of 0.9% in children with TB in the context of available TB medicines. This may confirm our suggestion that most of deaths are due to late diagnosis of TB rather than failure of TB medicines. The Adolescent age group had a high average mortality rate of about 6%. This high mortality in adolescents has largely gone unnoticed whose causes our study could not find out. The adolescents commonly have adult type disease and are less likely to be compliant to chronic treatment . The poor TB treatment outcomes among children needs further exploration and there is need to examine the time points of these deaths on the path of care and other co-morbidities in the adolescents.
Strengths and limitations
In this paper, we showed a trend over a five-year study period. This report follows a previous study with similar methods and therefore allows for a fair assessment of the progress of childhood TB services. We admit the inherent weaknesses of retrospective data that includes missing and inability to verify the records. Our inability to report data on Xpert ® MTB/RIF use in the diagnosis of TB in children is a drawback. The routine reporting tools to the NTLP were not capturing this data during the period covered by this study. We are however confident that our results are a representation of the true picture of the trend of childhood TB notification and outcomes in Kampala city. The weakness of notification rate is that it does not reflect the prevailing TB incidence unless there are few numbers of undetected, unreported or retreatment cases alongside good access to health services. Besides, the denominator projected population assumed a uniform growth rate despite the migrations in the urban setting, limiting accurate measurement of TB notification rates. We are also aware that some of the patients registered in Kampala are nonresidents and that some registered in one division may be residing in another division. Nevertheless, for purposes of showing a trend, our paper still makes an important contribution.
Implications for practice and program implementation
The focused efforts by the NTLP to look for TB among the under 15 years age group did show good results. The declining trend of notification rates in children mirrored declining adult TB notifications. This study cannot explain drivers of the noted trend for which we recommend further study. The low proportion of children among the notified TB cases in Kampala is likely due to a weak case finding system. Weak household contact tracing, TB screening and diagnostic skills lead to under diagnosis and under reporting TB in children. The high loss to follow up may suggest weaknesses in the support services such as adherence counseling and human resources for health leading to inadequate quality of care. We hypothesize the high mortality rate is mainly among previously hospitalized children. There is high mortality of up to 11% among children diagnosed with TB after hospitalization with delays of up to 16 days to make a TB diagnosis . Thus, many children started on TB treatment while on the wards are commonly sicker.