A comparision of the yield of three tuberculosis screening modalities among people living with HIV: a retrospective quasi-experiemental study

We conducted a retrospective quasi-experimental study of People Living with HIV (PLHIV) attending a Private Not for Profit (PNFP) HIV/AIDS clinic offering comprehensive services in Jinja district, Eastern Uganda. The HIV clinic was started in September 2004 following with the PEPFAR initiative.

The study population was HIV-positive patients enrolled in HIV care between 2006 and 2011. Patients that were found to be HIV positive were enrolled into the clinic and then they were treatment according to the Ugandan HIV/AIDS treatment guidelines [18]. By 31st December 2011, the clinic had 2800 PLHIV clients registered in care and alive; 1,040 of these (37.14 %) were males, 313 (11.17 %) were children under the age of 14 years and 1,934 (69.07 %) were on Antiretroviral therapy (ART).

Over the study period the clinic had two doctors, four Nurses, two Laboratory, two Counselors, two data clerks, two pharmacy technicians and a receptionist, the same staff was kept over the whole study period. All the patients who came to the clinic got their files from the reception and were directed to the Nurse in-charge for the assessment of vital signs. If the patient had no complaints, they were sent to another nurse who gave them a drug refill. The patients were allowed to get a drug re-fill without seeing the doctor twice but it was a must every 6 months. Here the patients had their vitals taken by the nurse, who sent them to the doctor. The doctor assessed the patient and also requested for laboratory tests mainly a Cluster for Differentiation 4 (CD4), Liver Function Tests (LFT’s) and Renal Function Tests (RFT’s). Then after passing by the laboratory, the patient would get the medicine from the pharmacy. However, if the patient was very ill, the patient was sent to the doctor immediately. The doctor then did the vitals as well as assessing the patient and offered the necessary care and treatment. Patients were often given bi-monthly appointments. Any patient who missed their appointment was given a call the following day. A home visit was done if the patient did not return to the clinic two weeks after the appointment. This was done to ascertain why the patient missed their appointment. But otherwise patients were allowed to visit the clinic any day in case of emergencies. Sometimes the patient was sent to the counselor but was mainly for either patients who are starting Anti-retroviral Treatment (ART) or those who were thought to be non-adherent. TB screening could be done by any staff along the process.

Three different strategies were used to identify potential TB patients: passive case finding, embedded intensified case finding, and an independent paper-based intensified case-finding tool, the different modalities were used exclusively. Whenever there was a change in intervention, the clinic staffs were taught about the new intervention at the same time, and they were oriented on the new tools to use.

With Passive Case Finding (PCF), no screening questions were administered since the identification of suspected TB cases was based on patients’ self-reports of any of the TB-related symptoms (i.e., current cough, fever, weight loss or night sweats) without prompting by a clinician. This was the standard of care between 2006 and 2007.

Between 2008 and 2009, a data capture field was added to the patient clinical encounter forms to compel clinicians to screen for TB symptoms for each patient. The data capture field necessitated the clinician to ask if the patients had any of the following symptoms and signs, current cough, drenching night sweats, night fevers, unexplained weight loss > 10 % of the body weight, and all respondents reporting any of the above-mentioned TB-related symptoms were considered for full TB evaluation. This underlies what we referred to as the embedded Intensified Case Finding (e- ICF) strategy.

Finally, the National TB and Leprosy Program of Uganda (NTLP) in collaboration with the AIDS Control Program (ACP) introduced a screening data collection form on which details and symptom questionnaire results are recorded each time a PLHIV attends the clinic (independent paper based Intensified Case Finding tool (i-ICF)). This i-ICF approach used the same screening questions as the e-ICF.

We included all patients who had at least one clinic visit between 1st January 2006 and 31st December 2011 (6 years).

The following patients were excluded, patients known to have TB or on TB treatment and cases missing significant variables. These were Basal Mass Index (BMI) ≤ 15 or ≥ 40, Cluster for differentiation 4 (CD4⁺) > 1500 cells/mm³ and were the age variable could not be determined.

During patient visits clinic staff prospectively entered into an electronic database the clinical parameters, ART initiation and adherence, WHO stage, toxicities, and opportunistic infections.

We extracted data of all the patients who had a cough for two weeks or more, fevers, weight loss or night sweats. We then corroborated this data with the health facility paper–based TB recording and reporting tools of the Uganda NTLP. These were: the TB laboratory registers for persons undergoing sputum examination for suspected TB; the unit TB treatment register and TB pharmacy dispensing logs for those who received treatment for TB. Where data was unclear or inconsistent, we validated these by reviewing clinical notes from the patient charts.

The study main outcome of interest was yield of tuberculosis on using different screening modalities. The TB yield was defined as a percentage of the patients that were found with TB when a particular screening modality was used. The screening modality used was dependant on the time period. The different yields were determined by getting a proportion of patients with TB indentified by a particular TB screening modality over the number of active patients when the modality was used.

We defined a patient with Tuberculosis according to the guidelines of the Uganda NTLP as follows: a person with a sputum smear microscopy result positive for TB or a patient with signs and symptoms of Tuberculosis with a chest–radiograph suggestive of TB or a patient with signs and symptoms of Tuberculosis who is diagnosed to have Tuberculosis by a competent Medical Officer [19].

Patients that were identified as TB suspects by the various screening modalities were sent for further investigations. The investigations often used were sputum microscopy, chest radiography, abdominal ultrasonography, lymph node biopsy and Fine needle aspirate for microscopy. Mycobaterial culture and Xpert MTB/RIF tests were either not used because of the high costs or not being readily available at the time. The diagnosis was done based on the investigation and occasionally on clinical presentation alone.

We exported the data from spreadsheet (Microsoft Excel version 2010) into Intercooled Stata version 11.2 (StataCorp, College Station, Texas, USA) for a complete cases analysis cases missing significant variables were excluded). We summarized demographic and outcome data into frequencies, percentages and measures of central tendency the median and mean. We explored the association between baseline characteristics and TB by univariate analysis (logistic regression) with crude odds ratios (OR) and 95 % confidence intervals. Statistical tests for association were two–sided and considered significant if P < 0.05. We included potential determinants of TB with P < 0.25 at univariate level, in a multivariable logistic regression model to estimate their adjusted odds ratios (aOR). In the time–to–event analysis were the start and end point was the beginning of a two year period when a screening modality was used, the censorships were death, lost to follow up, having TB or being on TB treatment. We compared the TB Case Detection Rate (CDR) of the screening interventions across the distinct cohorts enrolled during the three study periods: between 1st January and 31st December of 2006 to 2007, 2008 to 2009 and from 2010 to 2011.

The yield per screening modality was calculated using the number of TB cases detected over the number of active patients at the time. We then compared the yield of PCF to both e-ICF and i-ICF and that of e-ICF to i-ICF. The Case Detection Rate per 1,000 person years of observation (PYO) was also computed, and its corresponding 95 % confidence interval using the Wald statistic. We plotted survival curves using the Kaplan–Meier technique to depict this difference in yield of TB using the three different screening strategies. We report our findings in accordance with ‘The Reporting Quality of Non Randomized Evaluations of Behavioral and Public Health Interventions’ (TREND) statement [20].

Between 2006 and 2011, a total of 2800 patients were enrolled in HIV care at the HIV care clinic in Jinja district, Eastern Uganda. Of the 2800, 342 patients (63 patients had prevalent TB, while 279 had extreme variables, 246 had a BMI ≤ 15 or ≥ 40, while 21 had a CD4⁺ > 1500 cells/mm³ and 12 had an age <0 years of age) were excluded, leaving 2458 patients who were considered for the final analysis (Fig. 1). The cohort had 535 patients between 2006/07, 876 patients between 2008/09 and 1047 between 2010/11. The mean age was 34.8 years (SD: 11.5). Majority of the patients (61.2 %) were females and 149 (6.06 %) had a formal education (Table 1). The baseline median CD4+ cell count was 505 cells/mm3 (IQR 200 – 720 cells/mm3), while the mean BMI was 24.9 (SD: 5), and 1,912 (78 %) PLHIV were on ART. Participants diagnosed with Tuberculosis were more likely to be older (aOR: 1.02, 95 % CI: 1.01 – 1.04, p < 0.001), have a tertiary education (aOR: 2.38, 95 % CI: 1.41 – 4.04, p = 0.001), to be on ART (aOR: 3.38, 95 % CI: 2.16 – 5.32, p < 0.001) and have a higher BMI (aOR: 1.05, 95 % CI: 1.02 – 1.07, p < 0.001), (Table 1). Differences in the CD4+ cell counts were not statistically significant (p = 0.034).

When PCF was introduced in 2006, the yield of TB was 1.87 % (10/535) increasing to 14.95 % (131/876) when e–ICF was introduced in 2008. The yield then decreased to 12.79 % (134/1,047) when i–ICF was introduced in 2011. Each of the modalities was used for a period of two years. This translated into a TB Case Detection Rate (CDR) of 15.4/1,000 PYO (95 % CI 5.9 – 24.9) for PCF, 137.8/1,000 PYO (95 % CI 115.8 – 159.7) for e-ICF and 123.9/1,000 (95 % CI 104.3 – 143.6) for i–ICF, respectively (Table 2).

This increment in the yield of TB was statistically significant when either e–ICF (OR: 9.2, 95 % CI: 4.81-17.73, P = 0.0001) or i–ICF (OR: 7.7, 95 % CI: 4.0-14.78, P = 0.0001) were introduced. However, there was no statistically significant difference in the yield of TB when i-ICF was compared to e-ICF (OR: 0.98, 95 % CI: 0.76-1.3, P = 0.89). These findings are further illustrated in Fig. 2.

Our retrospective quasi-experimental study was not without limitations particularly the completeness and integrity of the data set which we circumvented by corroborating information from patient clinical notes and excluding missing data. We did not adjust for differences in time on HIV care, since longer duration on ART has been found to be protective. However, we adjusted for age, sex, CD4+ cell counts, body mass index and education status. Additionally, there could have been a detection bias due to the use of sputum smear microscopy, which has even lower sensitivity in patients with HIV, could have led to the underestimation of the burden of TB in this cohort particularly between 2006 and 2007 with very few events. Nonetheless, sputum smear microscopy remains the mainstay of TB diagnosis in peripheral HIV clinics due to implementation difficulties of scaling up the more robust Xpert MTB/RIF as per WHO recommendations [29].