Variability in distribution and use of tuberculosis diagnostic tests in Kenya: a cross-sectional survey

Globally, there is a significant TB case detection gap: 40% of all tuberculosis (TB) cases, 65% paediatric cases and 75% multi-drug resistant TB (MDR-TB) cases are missed due to a mixture of underreporting and under diagnosis [1–3]. Rapid and accurate diagnosis of TB is critical for timely initiation of treatment to prevent death [4–6]. A recent prevalence survey in Kenya found higher rates of TB than previously thought (558/100,000), with up to 55% of cases being missed probably due to under-detection [7]. Three quarters of the positive TB cases identified in the survey reported seeking care for TB-like symptoms but were not diagnosed. The survey recruited ≥15yr olds, but extrapolation from adult data showed that two thirds of paediatric TB cases were missed [7]. Notification data may underestimate child TB incidence, which may be explained in part by poor reporting of diagnosed paediatric cases [8]; and challenges of diagnosing TB in children due to paucibacillary disease and difficulty obtaining suitable samples [9, 10].

Quicker, more sensitive TB diagnostic technologies are being introduced globally [4, 11]. In 2010, Xpert MTB/RIF® was initially endorsed by the World Health Organisation (WHO) for children, the HIV infected and suspected MDR-TB cases [12], but is now recommended as the first line diagnostic test for all presumed TB cases [13, 14]. Kenya introduced Xpert® in 2011. According to the guidelines used in Kenya in 2015 (when this study was done), all presumptive paediatric, HIV-infected smear negative, drug resistant (DR) cases, or retreatment cases i.e. relapse (recurrence)/defaults/treatment failures should have had at least one Xpert® assay as part of their diagnostic work up (Additional file 1) [15, 16]. By 2015, there were 129 machines distributed throughout the country (146 machines to date), with more expected [17]. Optimal use of Xpert® is important for TB case detection [18, 19]. Few studies have specifically reviewed spatial distribution and practices in the utilisation of bacteriological TB diagnostic tests like Xpert® comparing adults and children, in countries that carry a high TB/MDR burden like Kenya.

A multi-country study on gaps in reporting paediatric TB found children were rarely considered for testing [20]. Age, gender, poverty and literacy are known to influence general health care utilisation and demand for services [21, 22]. One study found socioeconomic status and prior anti-TB treatment were strong determinants of utilisation of bacteriological tests [23]. Behavioural or health system issues may also play a role where services are available but underutilised- perhaps due to patients’ lack of awareness or perceived poor quality [22, 24–26].

Seeking to understand and improve TB case detection in Kenya, we set out to: describe the characteristics and spatial distribution of TB cases reported to the TB programme; the availability, distribution and patterns of use of TB diagnostic tests as per Kenyan guidelines [15, 16]; and to establish the determinants of use of Xpert®, noting differences in adults and children in the various counties. To the best of our knowledge, this is the first attempt to describe in detail the utilisation of TB diagnostics in Kenya, comparing adults and children to try to unmask the “hidden epidemic” of TB [27]. Findings will hopefully help guide policy makers on where the greatest needs are, how well guidelines are being implemented and offer suggestions to mitigate gaps identified. As new tests emerge [28], understanding patterns and determinants of use could help reduce TB deaths by guiding early diagnosis and linkage to appropriate treatment.

We set out to describe the spatial distribution of TB patients and TB diagnostic services, and patterns of use of TB diagnostic tests in Kenya as per the guidelines (with emphasis on Xpert MTB/RIF®), noting differences in children and adults. The TB case notification rate (CNR) in children 0-14yrs was nearly eight times less than that of those ≥15yrs, which may imply under detection. Children are thought to represent 10-20% of the total reported TB cases in high TB endemic settings like Kenya, but we observed only 7.8% in our data [40, 41]. Low CNR among children could additionally be explained by underreporting due to difficulties in confirming a diagnosis of TB in them and poor surveillance [2].

We observed wide county variation in distribution and use of TB diagnostic tests, as well as in case notification rates. Some of the northern and eastern counties had high TB CNRs, but a lower density/100,000 of facilities equipped with TB diagnostic services. Conversely, some counties had higher facility densities but reported lower rates of use. Kenya’s health system has been affected by devolution and decentralisation of health services, and this could be having a bearing on TB care [42]. A recently published patient pathway analysis in Kenya found distinct variation in diagnostic and treatment availability across counties and facility levels [32].

To comply with existing guidelines at the time, all presumed child TB cases, HIV infected adults and suspected MDR-TB cases and retreatment cases should have had at least an Xpert® done as part of the diagnostic work up [15, 16]. We noted underutilisation, with less than 5% of patients 0-14yrs and 12.2% of patients ≥15yrs having an Xpert MTB/RIF® test done, with gaps in guideline adherence. Many children, however, had smear microscopy done, on a specimen that presumably could have been used for Xpert®. The assay has much better sensitivity than smear microscopy and can identify rifampicin resistance with much faster turn-around time than traditional culture, and has been shown to be cost effective [43–46]. The Kenya prevalence survey found that when microscopy was used alone, up to 50% of TB cases were being missed, while Xpert MTB/RIF® detected 78% of TB cases [7]. While use of Xpert® is encouraged as the more sensitive test, it still misses approximately 22% of TB cases, and a negative Xpert®, especially in children does not rule out TB [43, 44]. However, if health workers can get sputum for microscopy especially in children, then they should ideally be able to send the sample for Xpert®, the recommended first-line test.

From the regression analysis, retreatment cases (i.e. failures, relapses/recurrences, defaulters) had the greatest odds of getting an Xpert® as expected from the guidelines, as did those who were HIV infected after adjusting for other covariates. Children had reduced odds of getting tested, despite the recommendation that all should have had an Xpert®. There is, however, no documentation of failed attempts to test.

Low utilisation of Xpert® has been illustrated in other low income, high TB burden countries both in adults and children [47–49]. From surveys of implementation in high TB burden countries including Kenya, reasons associated with low utilisation included operational issues like power outages and poor specimen referral; doubts about impact to TB morbidity and mortality; preference to trust clinical acumen; low sensitivity especially in children; challenges in getting good specimens and false negatives; and lack of awareness amongst health care workers and patients [45, 48, 50]. In Kenya, Xpert® sensitisation was initially done for lab staff only, which could have led to less demand for the test from clinicians. Some studies identified that training, workload, administrative support, staff motivation, role models and participation in the guideline development influence the implementation of TB guidelines in general [51–53].

Patients from higher level facilities had a tendency towards reduced odds of getting an Xpert® done. This could be because they get patients already being worked up from lower level facilities being referred to them to manage complications. The Kenya TB patient pathway analysis paper found that 58% of patients sought care in lower level facilities [32]. The Kenya SARAM report found only 60% of Kenya’s health facilities were ready to provide Kenya Essential Package for Health-defined TB services [34]. Readiness was found to be highest at the primary care facilities and in public facilities [34]. This is unlike in more well-resourced low TB burden countries, where TB care is at tertiary level facilities, and most patients have access to molecular diagnostic tests [54]. Private sector patients also had a tendency toward reduced odds of getting tested compared to public sector patients. This could be explained by private practice patients having to pay for tests to be done, and physicians may fail to comply with national guidelines, as seen in other settings [55–57].

Underuse of Xpert® despite wide scale roll out and considerable investments undermines assertions of the cost-effectiveness of this technology. There is need to increase use of Xpert®, to ensure the correct patients are targeted for lengthy and potentially toxic TB treatment, and so that its potential advantages are realised and investments justified. Current focus on increasing access to Xpert® is insufficient and more attention needs to be given to service delivery models that involve and educate staff and that address potential challenges of patient and specimen referral, with consideration needed for the private sector. Further research can support such efforts by identifying barriers and testing strategies to overcome these.