Predictors of attrition among adults in a rural HIV clinic in southern Mozambique: 18-year retrospective study

HIV remains a major cause of morbidity and mortality for people living in many low-income countries. With an HIV prevalence of 12.4% among people aged over 15 years, Mozambique was ranked in 2019 as one of eight countries with the highest HIV rates in the world. We analyzed routinely collected data from electronical medical records in HIV-infected patients aged 15 years or older and enrolled at Carmelo Hospital of Chokwe in Chokwe from 2002 to 2019. Attrition was defined as individuals who were either reported dead or lost to follow-up (LTFU) (≥ 90 days since the last clinic visit with missed medical pick-up after 3 days of failed calls). Kaplan–Meier survival curves and Cox regression analyses were used to model the incidence and predictors of time to attrition. From January 2002 to December 2019, 16,321 patients were enrolled on antiretroviral therapy (ART): 59.2% were women, and 37.9% were aged 25–34 years old. At the time of the analysis, 7279 (44.6%) were active and on ART. Overall, the 16,321 adults on ART contributed a total of 72,987 person-years of observation. The overall attrition rate was 9.46 per 100 person-years. Cox regression showed a higher risk of attrition in those following an inpatient regimen (hazard ratio [HR] 3.18, 95% confidence interval [CI] 2.89–3.50; p < 0.001), having CD4 counts under 50 cells/µL (HR 1.91, 95% CI 1.63–2.24, p < 0.001), receiving anti-TB treatment within 90 days of ART initiation (HR 6.53, 95% CI 5.72–7.45; p < 0.001), classified as WHO clinical stage III (HR 3.75, 95% CI 3.21–4.37; p < 0.001), and having Kaposi’s sarcoma (HR 1.99, 95% CI 1.65–2.39, p < 0.001). Kaplan–Meier analysis showed that patients with CD4 counts of less than 50 cells/µL on ART initiation had a 40% lower chance of survival at 18 years. Low CD4 cell counts, ART initiation as an inpatient, WHO clinical stage III, and anti-tuberculosis treatment within 90 days of ART initiation were strongly associated with attrition. Strengthening HIV testing and ART treatment, improving the diagnosis of tuberculosis before ART initiation, and guaranteed psychosocial support systems are the best tools to reduce patient attrition after starting ART.


Methods
Study design and population. This retrospective cohort study included data for all HIV-infected individuals aged 15 years or older who initiated ART at the CHC between January 2002 and December 2019. Individuals who were referred to the CHC after initiating ART in a center other than CHC were excluded from the analysis (n = 1681) (Fig. 1).
Setting. CHC is a state-owned reference hospital in the district of Chókwè, a large rural district in Gaza Province in southern Mozambique, administered by Catholic missionaries (the Daughters of Charity, Saint Vincent de Paul) since 1993. Chókwè is predominantly rural, with a surrounding catchment population of approximately 200,000 inhabitants. Agriculture is the sole source of income for roughly 80% of the population, and approximately 40% of the population migrates seasonally to South Africa, seeking work in the mines. According to the last national HIV prevalence study in 2015, Gaza Province had the highest provincial HIV prevalence in the country at 24.4% 2 .
Definitions. We classified the ART outcomes as retention or attrition. Retention (encompassing adherence) was defined as documented active and alive HIV-infected adults who were still on ART (assessed at intervals longer than 6 months post-initiation). Attrition was defined as death and LTFU.
Our reported mortality rate represents the number of documented deaths during ART per 100 person-years of observed therapy. The attrition rate represents the number of patients lost through attrition (death, LTFU) per 100 person-years. Patients were considered LTFU if they did not show up for their consultation in the 90 days preceding data abstraction, with missed treatment pick-up after 3 days of failed calls, unless the medical record stated that the patient had died, stopped ART, or was transferred out. The date of LTFU was recorded as the date of the last recorded visit, or 1 day after ART initiation if the patients never returned after ART initiation. To enable comparison with published literature, we report 1-year attrition 15,16 . Patients transferred to other facilities during the 18 treatment years were excluded from the 18-year attrition proportion. Transfers were censored in www.nature.com/scientificreports/ time-to-event analyses at the date of transfer. To facilitate survival analysis, we assumed that individuals who initiated ART but never returned over the 18 years of follow-up contributed 1 day of follow-up. Individuals remaining in care were censored at the end of the 18-year study period. Individuals who were either reported dead, LTFU or had transferred to other health care facilities were censored at their last clinic visit date. To allow comparison with other ART programs in low-resource settings, the first CD4 count was used to estimate the treatment outcome 17,18 . CD4 counts were determined using the FlowCount PLG CD4 assay and analyzed with Beckman Coulter Epics XL-MCL Flow Cytometers (both Beckman Coulter, Inc., Johannesburg, South Africa). Blood hemoglobin was determined with an automated hematology analyzer, Sysmex XT2000i (Sysmex Europe, Inc., Hamburg, Germany) or HemoCue photometer assay (HemoCue AB, Helsingborg, Sweden).
Similar to other cohort studies in resource-constrained settings 16 , adherence to ART was estimated by measuring timeliness of patient visits to scheduled medicine pick-up appointments at clinic-based pharmacies 19,20 during the first 6 months of ART.
Categories used to classify exposure to anti-tuberculosis treatment (ATT) were: 1) completed ATT before ART initiation: 2) initiated ATT < 90 days after ART initiation; and 3) unexposed to ATT at ART initiation. Data collection. Data collection was performed using DREAM software (Diseases Relief through Excellence and Advanced Means), version 6.1.0.757, and electronic medical record (EMRs), designed according to criteria of excellence to manage the prevention and treatment centers of the DREAM Program of Sant'Egidio Community, located in 11 African countries, including Mozambique. Since 2002, the Catholic missionary organizations Daughters of Charity Saint Vincent de Paul (http:// www. daugh tersi ps. org/ hivai ds) and the Community of Sant'Egidio (http:// www. dream. sante gidio. org) have a collaboration agreement that allows the Daughters of Charity health facilities to use the DREAM software to manage their HIV-infected patients. Carmelo Hospital of Chokwe is a health facility administered by the Daughters of Charity and uses the DREAM Dataset as main the main EMR system. It works through Microsoft Access and JavaScript and is supervised by the hospital's local IT department. All offices work electronically and use the database for routine consultations, so all patient information is automatically stored on a common server. The ART database collects and stores long-term EMRs, sociodemographic data, clinical histories, laboratory tests (hematology, biochemistry), identification codes, scheduling of new appointments, and pharmaceutical stocks.
Recorded data from eligible ART participants for the study were collected and anonymized to remove identifying details: each patient was assigned an alphanumeric code (ID), and the anonymous data were included in the data sheet.
To convert data from Microsoft Access to SPSS, the Access data were first exported to a Microsoft Excel spreadsheet, and in turn the spreadsheet data were imported to SPSS for subsequent data analysis.
For each patient, we included the following data: gender, age (age band and median), point of entry (inpatient or outpatient), CD4 count at baseline, previous ART exposure, previous ATT (ATT prior to ART initiation, ATT within 90 days of ART initiation, and no exposure to ATT), WHO clinical stage of HIV disease, Kaposi's sarcoma diagnosis, ART therapy, and final ART outcome. Data analysis. Statistical analysis was performed using IBM SPSS Statistics Software version 25 (IBM Corporation, International Business Machines Corp, Release 2017, https:// www. ibm. com/ legal/ copyt rade, USA). Patients' baseline characteristics, as described above, were compared according to outcomes.
Time-to-event methods were used to investigate predictors of attrition. Time was measured from ART initiation to death or LTFU. Patients still alive at the date of data collection were considered active in ART. Patients  www.nature.com/scientificreports/ who did not come to the clinic for at least 90 days after their last scheduled appointment were considered LTFU 18 . Attrition was calculated by summing the number of patients who experienced the event (deaths, LTFU) during a particular period divided by the total years of follow-up during this period. Multivariable analysis was performed by Cox regression models. The proportional hazard assumption was assessed with log survival curves based on Schoenfeld residuals. We used Kaplan-Meier survival estimates to calculate the cumulative incidence of attrition. The cumulative incidence of mortality and LTFU were calculated using competing risk analysis. The Mozambican National Bioethics Committee for Health (Comité Nacional de Bioética para a Saúde, 25/ CNBS/2019) approved this analysis. Analysis was performed on de-identified, aggregated patient level data, and no individual informed consent was obtained. The need for written informed consent was explicitly waived, and the research was performed in accordance with the Declaration of Helsinki.
Ethics approval and consent to participate. The Mozambican National Bioethics Committee for Health (Comité Nacional de Bioética para a Saúde, 25/CNBS/2019), approved this analysis (25/CNBS/2019). Analysis was performed on de-identified, aggregated patient level data, and the need for written informed consent was explicitly waived.

Consent for publication.
We performed analysis on routine administrative data; consent for publication is not applicable.

Results
ART treatment outcomes. A total of 25,275 HIV-infected adults aged 15 years or older were enrolled in HIV care from 1 January 2002 to 31 December 2019. Of these, 16,321 (64.6%) initiated ART at CHC during the 18-year study period (Fig. 1).
Women, patients who were ART-naïve, those with CD4 count of more than 200 cells/µL at ART enrollment year, and certain drug combinations (ABC/AZT/D4T + 3TC + LPV/RTV,) were not associated with higher mortality (Table 4).

Survival estimates.
Kaplan-Meier estimates for the probability of 18-year survival by CD4 cell count on ART initiation are presented in Fig. 3.
PLWHIV who had a CD4 count of more than 500 cells/µL had the highest probability of approximately 90% of survival at 18 years following ART initiation, while those with a CD4 count of less than 50 cells/µL had a survival probability of approximately 40% at 18 years in comparison.

Discussion
In our cohort, patients with CD4 counts of less than 50 cells/µL had almost double the attrition risk compared to those with CD4 counts of more than 500 cells/µL. These findings are a little higher than observed in other Mozambican literature 21 . An in-depth analysis of our cohort showed that the frequency of attrition in the patients with low CD4 cell counts was 26.3%; of these, 43.9% died, and 14.5% were lost to follow-up. Other studies have reported high mortality and LTFU in severely immunosuppressed patients 22,23 , highlighting LTFU as a fundamental problem for the health care in PLHIV. Risk factors for ART interruption are still poorly understood in many settings, including Mozambique. The survival analysis shows that in our cohort, patients with CD4 counts under 50 cells/µL had an 18-year survival rate of less than 50%. These data are consistent with other reports 24 . The baseline CD4 cell count is the dominant prognostic factor for survival in PLHIV starting ART 25 . Our study showed that patients who started treatment at an advanced disease stage are at a higher risk of mortality than those who start treatment with higher CD4 cell counts, even if they survive the first 5 years of therapy. These findings are in keeping with studies from high-income settings 25 .
On analyzing the first 2 years of follow-up, we observed a considerable drop in the survival curve among people with CD4 counts of less than 50 cells/µL, from 100 to 60%; by 5 years, survival stood at just 50%. This finding is strongly related to the fact that severe immunosuppression in this population is not restricted to HIV infection alone; rather, multiple factors are at play, including hunger, adult malnutrition 26 , and poor nutritional supplementation 27 . These people are considered "immunological non-responders, " that is, HIV-infected individuals who fail to achieve normalization of CD4 + T-cell counts despite persistent virological suppression, which leads to AIDS and non-AIDS comorbidities 28,29 . This population continues to have a substantially increased mortality ratio even after 3 years of treatment. Thus, clinical factors such as immunological non-response, AIDS and non-AIDS-related multiple events predict a substantially increased risk of death after 3 years of ART treatment 30,31 .
Recovery of CD4 lymphocytes occurs in the first 2 years after starting ART and is associated with age and the pre-existing degree of HIV-1-related immunodeficiency, so long-term exposure to HIV-1 infection damages the immune system in ways that are difficult to correct due to exhaustion [32][33][34] .
Better understanding of patterns and risk factors that cause-specific mortality in the ART era can inform the development of appropriate care for HIV-infected individuals and guidelines for risk factor management 35 . This may include screening programs and intensive adherence counseling 34 .
We observed that patients who received ATT within the first 90 days of their ART initiation had a six-fold higher risk of attrition than those who were unexposed to ATT. These results are a little higher than those observed in other Mozambican studies 36 . In our cohort, the attrition rate on active TB patients was 24.13 per 100 person-years, and the frequency of attrition was 49.7%. The overwhelming majority of patients receiving ATT within 90 days (97.4%) died, while 25.3% of those LTFU had a fatal outcome. These results show high mortality but low LTFU, so having TB is a favorable predictor against LTFU.
In our sample, men were underrepresented (39%), as published elsewhere 37 . There were few men who went to health facilities to test for HIV and to initiate and adhere to ART 38,39 . We fully acknowledge that more women than men take an active role in accessing HIV services, and studies show that their adherence to ART is also higher 40 . This tendency may be rooted in different dimensions of gender norms in rural areas 41 .
Our cohort shows a higher ART initiation rate, of around 37.9%, and a higher death rate among 25-34-year olds (35.4%) compared to an Ethiopian University Teaching Hospital study, which reported an ART initiation rate of 27.9% in the same age group 24 25 .
With this advanced immune suppression, then, a high proportion of men started ART with advanced AIDS and AIDS Clinical Stage IV (45.0% in men vs. 25.8% in women). Immune reconstitution syndrome was closely associated with tuberculosis, confirming previous findings from our cohort and other resource-limited settings 42,43 . This syndrome was more frequent in men (40.8% vs 22.2%), and over the course of 18 years it was observed that the male population was always on a smaller scale.
Women comprised the majority of the population at 60% compared to men, but half (49%) were classified in the attrition group. Being a woman was not a statistically significant favorable predictor in our cohort.
Among the patient group with CD4 counts of less than 50 cells/µL, over half were men, 39.55% were aged 25-34 years, and 65% had started ATT within 90 days of initiating ART. These findings are consistent with data from other studies 44,45 , which reveal a greater predominance of men among patients with late HIV diagnosis as well as late-onset ART. The care, ART initiation and treatment monitoring are in accordance with the WHO and national ART guidelines, which were updated successively in 2002 46 Table S1). This is partly due to community awareness campaigns on ART treatment, as well as the influx of patients into inpatient services. At that time, many patients were diagnosed at the advanced stage of the AIDS disease, consequently with severe immunosuppression.
The years 2014/2015 48 were the first following implementation of other eligibility criteria that defined coinfection with TB, Hepatitis B, cancer, and HTLV as criteria for starting ART, regardless of WHO stage or CD4 cell count (Supplementary Table S1). Thus, these data contributed to a substantial increase in patients with CD4 counts of 350-399 cells/µL.
The now outdated D4T-based regime was used as a first-line treatment from 2002 to 2018 [46][47][48][49][50] , and as an alternative line in patients with Hgb anemia of less than 8 g/dL at the time of screening. Another outdated initial regime, based on AZT, was used from 2009 to 2018 [47][48][49][50] . The updated TDF-based regime has been used since 2013 48 , and the current DTG-based regime since 2019 51 , a period characterized by the definitive discontinuation of INNRT (NVP/EFV) drugs.
The strengths of this study reside in the large sample size and the long study period. Limitations include those of many retrospective studies in which data collection is often inadequate and incomplete, without baseline viral load and data to monitor therapeutic failure. This may have an influence on the reported results, mainly for death and LTFU.

Conclusion
The results of this cohort study indicate that attrition after ART initiation is high. Nearly 40% of patients die or are LTFU. Low CD4 cell counts, starting ART as an inpatient, WHO clinical stage IV, and being a man were strongly associated with attrition. Although the mortality of patients diagnosed with tuberculosis after ART initiation was extremely high, patients who started anti-tuberculosis treatment before ART had similar attrition as patients without tuberculosis. The results of this study indicate that interventions aimed at initiating patients sooner after HIV seroconversion, improving the diagnosis and treatment of TB and other comorbidities (like Kaposi's sarcoma), and giving extra support to patient groups at higher risk of attrition could potentially reduce the mortality and LTFU in HIV programs in Mozambique.   www.nature.com/scientificreports/

Data availability
The datasets analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.