Study design
This was a retrospective analysis of routine programme data from the AMPATH electronic Medical Record System (AMRS) database.
Setting
General setting
Kenya is a lower-middle-income country with a population of 48 million, 74% of whom reside in rural areas [31]. The crude death rate is eight per 1000 population per year and life expectancy is about 60 years [32, 33]. HIV prevalence among adults in Kenya is estimated at 5.9% and is higher in western Kenya [34]. Kenya’s Gross Domestic Product stands at USD 70.5 billion with a largely service and agricultural economy [31]. Fourty-6 % of the population still live below the poverty line [35].
The national TB programme carries out various activities aimed at improving TB care, prevention and control in the country. These include case finding approaches targeting both drug sensitive and drug resistant TB patients, children and other special populations, TB/HIV integrated approaches including HIV testing, early co-trimoxazole and antiretroviral therapy (ART) uptake, TB preventive therapy with isoniazid, public private partnerships, laboratory diagnostics, among others. In 2016, the country notified a total of 77,376 TB patients, 31% of whom were co-infected with HIV [1]. A total of 445 drug resistant TB cases were notified in the same year [36]. Treatment success rate was 87 and 82% amongst all notified drug sensitive TB patients and HIV co-infected TB patients respectively while it was 72% among those started on second-line treatment in 2014. Mortality from HIV-positive TB was 50 per 100,000 population in 2016, and ART uptake among TB/HIV co-infected patients was reported to be 95% [1]. In 2014, about 0.5 million PLHIV were reported to have been screened for TB [26].
TB/HIV care at AMPATH
The Academic Model Providing Access to Healthcare (AMPATH) programme supports 29 subcounties within eight counties in western Kenya. These facilities have their TB and HIV care and prevention services integrated at various levels [37]. These include screening for TB among PLHIV and HIV testing among TB patients, co-trimoxazole and ART uptake among those co-infected, TB treatment, TB preventive therapy and infection prevention and control. In the fully integrated HIV-TB care set ups, patients are clinically reviewed for both diseases by the same clinical officers (equivalent of physician assistants) with difficult cases handled by medical officers and consultants. The team also has the support of pharmaceutical technologists or pharmacists, counsellors, nutritionists, social workers, records officers and laboratory staff. In the minimally or non-integrated TB-HIV set ups, patients are reviewed for their TB status by nurses and HIV status by clinical officers in different settings or facilities. Difficult cases are referred to higher facilities.
In 2016, over 6000 TB patients were notified in 244 TB or TB/HIV treatment facilities in these supported regions. About one-third of the TB patients were co-infected with HIV. ART and co-trimoxazole therapy uptakes were high at over 95%. Tuberculosis treatment success rate was 88% overall and 83% among HIV positive TB patients.
The AMPATH HIV programme has over 80,000 PLHIV in care at any particular time, of whom about 90% are in 34 of the 140 HIV clinics. The AMPATH electronic Medical Record System (AMRS) [38, 39] – based on Open MRS platform (OpenMRS, Inc. USA) – is used to capture patients’ clinical details for majority of the PLHIV. A few clinics still rely fully on paper-based recording systems. However, all the high and medium volume clinics are supported by an AMRS system. Typically, clinicians enter patients details into structured paper-based encounter forms after which the data entry officers transfer the data into the AMRS electronic database. Appropriate queries are used to extract details from the database.
The Kenya national TB/HIV guidelines require that every PLHIV is screened for TB at every clinical visit/ encounter. The screening questions include the presence of cough of any duration, unintentional weight loss, fever, and night sweats [18]. Any affirmative response leads to testing for TB using Xpert MTB/RIF (preferably, if available) or sputum smear microscopy for acid-fast bacilli where Xpert is not available. Any positive test result leads to the diagnosis of TB and subsequent initiation of TB therapy. Those screening negative are offered isoniazid preventive therapy if eligible. Those screening positive but testing negative for TB are investigated further by, for example, chest radiography or given antibiotics and reviewed at subsequent visits. The pulmonary form of TB may still be diagnosed clinically if sputum microscopy and/or Xpert are negative but the chest radiograph or clinical symptoms point to TB disease [19]. Extra-pulmonary TB is usually diagnosed on the basis of clinical presentation though some may access histology (depending on the anatomical site affected) thus strengthening the clinical diagnosis.
During the study period, Xpert MTB/RIF machines were still limited in the study area. Only five sites had resident Xpert machines with some of the remaining sites networked with these in a less-than-optimal systems of sample transport, results relay and recording. As a result Xpert uptake among PLHIV was still low in the setting and when performed results did not get recorded in the HIV database. In contrast, more sites had X-ray capabilities while microscopy services were the most common diagnostics available in the sites. HIV and TB treatment services, Xpert and microscopy are provided free of charge in the country’s public sector. In addition, Xray services for PLHIV are also free for PLHIV in the AMPATH-supported sites.
Study population
The study population consisted of all PLHIV in the AMRS database and who were in care in the AMPATH programme between January 2015 – December 2016. For programmatic and logistical reasons, all patients in the AMRS were included. Non-clinical encounters such as food and social support were excluded from analysis.
Data and analysis
Study variables included patient demographics (e.g. gender, age, clinic, encounter type [e.g. youth]) and clinical details (e.g. use of ART, TB screening status and the presence of TB-suggestive symptoms, TB testing results [sputum smear microscopy and chest radiograph] and dates of clinical encounters, screening, and TB test results). Xpert MTB/RIF testing procedures were dropped from the analysis due to the lack of data in the database. Data on whether the PLHIV were on TB therapy during the encounters were also dropped due to significant erroneous recording.
Screening for TB during a clinical encounter was defined as the recorded presence or absence of individual symptoms (cough ≥2 weeks, fever, night sweats, significant weight loss, chest pain and/or breathlessness) by documented checking of the appropriate box; when no ticking was recorded, the encounter was regarded as ‘no screening’. A patient was regarded to have had ‘Undesirable’ screening if s/he had been screened for TB in < 90% of the clinical encounters which the patient had experienced during the study period; otherwise, the patient was categorized as having had ‘Desirable’ screening (i.e. if screened ≥90% of the times in the programme in the study period). The cut-off of ≥90% for desirable screening was chosen as this is also a programmatic target in the country.
At the basic level each observation for each patient was assessed for TB screening and scored as Yes/No. These were then aggregated to the total number of times a patient was screened and divided by the total encounters the patient had to generate the proportion of times the patient was screened for TB. Clinics were categorized as high, medium, low and very low volume based on the number of patients enrolled in the clinics and ensuring equal numbers of clinics per subcategory.
The data were analyzed per-encounter/ observation and per-patient as necessary. Categorical data have been presented in frequencies and proportions and compared using the chi-square test while means (standard deviation) or medians (interquartile range) have been used for continuous variables depending on normality of the data. Any comparisons of continuous variables were by the t-test or mann whitney U test, as appropriate.
Factors influencing the main outcome measure (Undesirable screening) were assessed by log-binomial regression at both bi- and multivariable levels and effects presented as relative risks and their 95% confidence levels. The multivariable model was built sequentially by adding age and gender (selected a priori) and factors identified at bivariable level while assessing confounding, interactions and improvement of the model fit to the data. Variables were also dropped if collinearity or sparsity existed. The final regression model included over 92% of the PLHIV, the difference mainly due to the 8% of PLHIV who did not have data on use of ART. This did not significantly affect the fit of the model. Grossly missing or inaccurate data (Xpert uptake and TB therapy) were not included in the analyses. Level of significance during the analyses was set at P < 0.05.
Analyses were carried out using Stata/SE 14.1 software (StataCorp, College Station, TX, USA).