Incidence and predictors of loss to follow-up after ART initiation at public health facilities in Sierra Leone (2020–2024): a retrospective cohort study Authors: Lovel Fornah, Augustus Osborne, Mulugeta Shegaze Shimbre, Wei Ma Published: 2026 ABSTRACT ObjectiveThis study aimed to estimate the incidence and predictors of loss to follow-up (LTFU) among adults after antiretroviral therapy (ART) initiation at five provincial referral hospitals in Sierra Leone.DesignA facility-based retrospective cohort study.SettingThis study was conducted at five provincial government referral hospitals in Sierra Leone from January 2020 to December 2024.ParticipantHIV patients aged 15 and above who initiated ART between 1 January 2020, and 31 December 2020, and attended at least one follow-up visit at the five national referral hospitals.Primary outcome measureThe primary outcome in this study was LTFU. LTFU was defined as absence from care, operationalised as no documented clinical encounter and/or ART refill for ≥90 days after the last expected clinical contact among patients not recorded as dead or transferred out.ResultsOf the 1544 patients (67.7% female), the median follow-up was 4.25 years (IQR not stated), totalling 5748.6 PY. Outcomes were 550 LTFU (35.6%), 853 in care (55.2%), 64 (4.1%) transferred and 77 (5.0%) died. The LTFU incidence was 9.57 per 100 PY (95% CI 8.78 to 10.37). Multivariable analyses revealed several factors significantly associated with higher LTFU: lack of tuberculosis (TB) screening (HR=4.47; 95% CI 3.66 to 5.46), WHO stage III/IV (HR=1.51; 95% CI 1.20 to 1.89), no regimen change (HR=5.12; 95% CI 2.39 to 11.00), non-disclosure of HIV status (HR=1.63; 95% CI 1.19 to 2.25), underweight body mass index (BMI) at follow-up (HR=1.37; 95% CI 1.05 to 1.78) and residence in the Southern (HR=1.97; 95% CI 1.57 to 2.48) or Eastern region (HR=2.13; 95% CI 1.49 to 3.03). Conversely, rural residence was associated with lower LTFU (HR=0.58; 95% CI 0.43 to 0.79).ConclusionThe study found a substantial incidence of LTFU. Risk factors of LTFU were gaps in TB screening, advanced clinical stage, absence of timely regimen adjustments, non-disclosure, low BMI and regional disparities. Strengthening integrated TB/HIV services, nutritional support, disclosure-enabling counselling and region-specific retention strategies, especially in urban, Southern and Eastern areas, may improve retention. STRENGTHS AND LIMITATIONS OF THIS STUDY This study uses a large, multi-facility cohort dataset spanning five provincial referral hospitals across all five regions of Sierra Leone with diverse catchment populations, enhancing external validity. This study uses a large, multi-facility cohort dataset spanning five provincial referral hospitals across all five regions of Sierra Leone with diverse catchment populations, enhancing external validity. The application of robust statistical methods, including Kaplan-Meier estimation and a stratified Cox proportional hazards regression model, to identify the predictors of loss to follow-up (LTFU) provides dependable risk estimates. The application of robust statistical methods, including Kaplan-Meier estimation and a stratified Cox proportional hazards regression model, to identify the predictors of loss to follow-up (LTFU) provides dependable risk estimates. Application of a computer-generated simple random sampling method from antiretroviral therapy (ART) registers, minimising selection bias. Application of a computer-generated simple random sampling method from antiretroviral therapy (ART) registers, minimising selection bias. Retrospective design relying on secondary data from routine ART registries, which carries a risk of incomplete or inconsistent documentation for certain variables. Retrospective design relying on secondary data from routine ART registries, which carries a risk of incomplete or inconsistent documentation for certain variables. Hospital routine records pose a risk of outcome misclassification (such as silent transfers and unrecorded deaths), which may skew LTFU estimates. Hospital routine records pose a risk of outcome misclassification (such as silent transfers and unrecorded deaths), which may skew LTFU estimates. Introduction As of 2024, HIV is still a major global health problem, with about 40.8 million people living with the virus around the world. Global antiretroviral therapy (ART) access reached 31.6 million by the end of 2024, yet the UNAIDS target of 34 million by 2025 remains elusive.1 Sub-Saharan Africa bears the greatest burden, accounting for approximately two-thirds of all people living with HIV (PLHIV) globally.1 Achieving the UNAIDS 95–95-95 targets requires not only initiating patients on treatment but also ensuring sustained retention in care, a challenge that loss to follow-up (LTFU) from ART programmes continues to undermine across the region.2 HIV remains a major public health issue in Sierra Leone. In 2021, the estimated national HIV prevalence among adults aged 15–49 years was 1.5%, with around 110 000 individuals living with HIV.3 Women constitute approximately 57% of those living with HIV in the country.3 Despite the growth of ART services, only about 60% of PLHIV knew their status, and ART coverage among those who had been diagnosed was not satisfactory enough, with major gaps in viral suppression, especially in rural and underserved areas.4 These gaps point out the importance of not only commencing patients on ART but also guaranteeing ongoing retention in care, which continues to be a substantial programmatic challenge in the context of Sierra Leone. In 2016, the WHO implemented the “treat all” policy, advising that all individuals diagnosed with HIV initiate ART after a confirmed diagnosis and clinical assessments, irrespective of WHO clinical stage or CD4 cell count.2 The free ART policy was implemented in Sierra Leone in 2005. Since then, there has been a notable rise in the utilisation of ART services and an expansion of ART sites; however, the rate of LTFU among adults and other demographic groups, including pregnant women, remains elevated.5 Retaining patients in ART programmes is important for achieving sustained viral suppression and decreasing HIV-related morbidity and mortality. Nonetheless, LTFU persists as an impediment to progress, especially in low- and middle-income nations, where healthcare systems confront the dual challenge of augmenting ART services while addressing additional health concerns such as tuberculosis (TB) and malnutrition. LTFU distorts programme monitoring, complicates outcome ascertainment, and can conceal unreported deaths, treatment interruptions, and silent transfers.6–11 Understanding the incidence and predictors of LTFU is essential for designing customised retention strategies that fit the specific context of healthcare systems.6 12–14 Evidence from Ethiopia, Uganda, Cameroon, Rwanda, Malawi, Eritrea, Liberia and other sub-Saharan African settings shows that LTFU after ART initiation is common but heterogeneous, reflecting differences in service organisation, cohort composition, follow-up periods and measurement practices.6–21 In Ethiopia, multi-site cohorts reported substantial LTFU under test and treat, with programmatic and clinical correlates including advanced WHO stage, suboptimal adherence and TB co-infection/screening gaps.7 12 13 15 17 20 22 23 In Uganda and Malawi, LTFU clustered in the early months after ART start, with urban residence, mobility and system-level bottlenecks implicated.10 11 16 18 In Rwanda and Cameroon, urban concentration and facility-level dynamics were associated with attrition, echoing broader regional patterns.7 19 Meta-analytic and pooled analyses across the region further underscore how advanced disease, malnutrition (low body mass index (BMI)) and incomplete integration of TB/HIV services elevate attrition risks.9 11 21 24–26 Several predictors of LTFU recur across contexts. Advanced WHO clinical stage (III/IV) is linked to higher LTFU and unascertained mortality, likely reflecting severe immunosuppression, opportunistic infections and complex care trajectories.6 10 13 15 17 22 23 27 Nutritional compromise, often proxied by low BMI, is associated with poorer ART outcomes and higher disengagement, pointing to the role of food insecurity and wasting in undermining adherence and clinic attendance.10 11 21 24–26 28 TB remains a pivotal co-morbidity; absence of routine TB screening or co-located management is consistently associated with poorer retention, whereas integrated screening/management reduces LTFU.13 29 Social and behavioural determinants, particularly non-disclosure of HIV status, limit social support and may increase stigma, depression and drop-out.12 14 22 30 31 Geographic disparities also matter: regional differences in infrastructure, staffing and outreach capacity shape retention, while urban–rural contrasts vary by country and programme model.7 13 14 19 24 32 Social stigma, lack of disclosure and limited health system access have been identified as key barriers to ART retention across sub-Saharan Africa.33 In Sierra Leone specifically, HIV-related stigma remains a pervasive barrier to care engagement. Qualitative evidence from Lahai et al demonstrates that fear of stigma and discrimination from family members, community members and even healthcare providers is a primary reason why PLHIV delay or discontinue ART.34 Non-disclosure of HIV status, which is strongly shaped by stigma, has been identified as a key barrier to accessing the social support networks that facilitate treatment adherence.34 Structural barriers further compound these challenges: Sierra Leone’s health system is characterised by limited facility density, particularly in rural and remote areas, inadequate numbers of trained HIV counsellors and historically inconsistent supply chains for ART commodities.3 4 Co-infection burden also contributes to treatment complexity; Song et al reported significant hepatitis B virus co-infection among PLHIV on ART in Sierra Leone, highlighting the need for integrated management approaches to sustain retention, particularly in light of the challenges posed by limited healthcare resources and the need for comprehensive care strategies.35 These country-specific factors must be considered when interpreting patterns of ART retention and LTFU in the Sierra Leonean context. Despite these insights, there is limited multi-province, facility-based evidence from Sierra Leone,36 37 an essential gap given national ART scale-up, evolving service models and regional heterogeneity in access and outcomes. Programmatically, Sierra Leone’s primary referral hospitals serve large catchment populations and anchor ART services; however, routine monitoring often under-captures early disengagement, silent transfers and context-specific risks such as urban mobility, stigma and variable TB service integration. Against this backdrop, rigorous province-spanning analyses are needed to quantify the incidence of LTFU and identify actionable predictors to guide targeted retention interventions, improve service integration and inform differentiated care. This retrospective cohort study addresses that gap by estimating the incidence of LTFU and identifying its predictors among adults initiating ART at five provincial government referral hospitals in Sierra Leone between 2020 and 2024. Grounded in prior literature from sub-Saharan Africa, we focused on factors repeatedly implicated in attrition: TB screening status, WHO clinical stage, regimen change, residence, HIV status disclosure and BMI, while assessing regional differences that may reflect system and access inequities. By leveraging routinely collected programme data and robust time-to-event methods, the study aims to generate practical evidence for designing retention strategies tailored to Sierra Leone’s service delivery realities. Methods Study design, period and setting A facility-based retrospective follow-up study was conducted at five provincial government referral hospitals in Sierra Leone from January 2020 to December 2024. The selected facilities are Connaught Hospital (Western Area), Makeni Government Hospital (Northern Province), Bo Government Hospital (Southern Province), Kenema Government Hospital (Eastern Province) and Port Loko Government Hospital (Northwestern Province). These hospitals were chosen because they are the primary referral centres in their respective regions and began offering ART services in 2017. Sierra Leone’s national HIV programme follows a “treat all” policy, making all individuals living with HIV eligible for ART, regardless of age, CD4 count or clinical stage, with rapid initiation prioritised within 7 days of diagnosis, especially for pregnant women, young children and those with advanced HIV disease.38 39 ART initiation begins with HIV testing through facility- and community-based services, followed by clinical assessment and rapid initiation, even before receiving baseline lab results if there is no evidence of TB, meningitis or renal disease. ART services are decentralised, with 439 facilities across tertiary, secondary and primary care levels, and national guidelines support task-sharing and community-based ART refills to improve access and retention.39 40 Study population This study included HIV patients aged 15 and above who initiated ART between 1 January 2020, and 31 December 2020, and attended at least one follow-up visit at the five national referral hospitals. Patients with unknown initiation dates, undefined treatment outcomes or incomplete baseline data due to transfers were excluded. Sample size determination and sample procedure A double population proportion formula was used for sample size calculation using STATA software (V.18) based on specific parameters: a 95% CI, 80% power, a 5% margin of error (d), a 1:1 ratio of exposed to unexposed individuals and the predictors (not being able to work (AHR=1.84), not participating in clubs (AHR=1.68) and drug side effects (AHR=1.44)) associated with LTFU from research conducted in Central Ethiopia.15 The sample size was calculated for each predictor individually, and the predictor with the largest sample size (1181) was chosen. To account for missing data and ensure precision across multiple facilities, the sample size was increased by 30%, resulting in a final total of 1560 participants. A computer-generated simple random sampling method was employed to select eligible patients from the ART registers at five facilities. Patient lists were compiled from the registries and organised in Excel, including key details such as patient ID, ART initiation date and year. Each patient was assigned a random number using Excel’s “=RAND ()” function, and the lists were sorted based on these numbers. The required sample size for each hospital was then determined from the ranked lists using a proportional double population proportion method. Measure variable Outcome variable The outcome variable in this study was LTFU. LTFU was defined as absence from care, operationalised as no documented clinical encounter and/or ART refill for ≥90 days after the last expected clinical contact, among patients not recorded as dead or transferred out.23 This definition is appropriate in the Sierra Leone context where appointment spacing and multi-month dispensing allow up to 3 month intervals between scheduled visits/refills.39 A patient was considered LFTU if they met the aforementioned criteria. Those classified as dead or transferring out of care during the follow-up period, as well as those who remained on ART at the end of the study period but did not meet the criteria for LTFU, were censored. In this study, we assigned a value of 1 to those classified as LTFU and a value of 0 to those who were censored.22 Independent variables The independent variables for this analysis were selected based on an in-depth review of previous studies.15 18 22 23 27 32 They were categorised into baseline socio-demographic factors (Age in years, Sex, Marital status, Educational status, Occupation, HIV disclosure status, Residence and Region), baseline clinical factors (BMI at baseline, CD4 count at baseline, WHO clinical stage at baseline, Regimen at baseline and Functional status at baseline), and patient follow-up infection, treatment and clinical characteristics (TB Screening, TB Status, Opportunistic Infection, Regimen change during follow-up, Viral load at follow-up, BMI at follow-up, CD4 count at follow-up, WHO clinical stage at follow-up and Adherence). The independent variables for this study were categorised as depicted in the online supplemental information 1,table S1. Operational definition Loss to follow-up LTFU was defined as absence from care, operationalised as no documented clinical encounter and/or ART refill for ≥90 days after the last expected clinical contact (missed scheduled appointment/expected refill date), among patients not recorded as dead or transferred out.23 Additionally, in this study, patient’s follow-up measures refer to the clinical assessment performed at the last known visit before the patient was classified as LTFU. (Online supplemental file 1,table S1) presents sensitivity analyses regarding the choice of the LTFU cut-off of ≥90 days. Educational status Educational status refers to the highest level of formal schooling that an individual has attended and/or completed. WHO staging WHO stage system for HIV-positive patients. Functional status Functional status was classified based on the criteria set by the WHO into three distinct categories: Working (patients in this category are fully capable of engaging in their regular work activities, whether working inside or outside the home), Ambulatory (this group includes those who can perform daily living activities but are unable to fulfil work obligations) and Bedridden (patient classified as bedridden are unable to perform basic daily living activities).23 27 32 TB screening All HIV-positive patients were screened for TB using the four cardinal questions: prolonged fever, cough for 2 weeks or more, night sweating and weight loss. The responses to these questions were recorded as YES (screened) or NO (not screened). This screening occurred at every scheduled clinic visit throughout follow-up. For the purpose of this study, responses were coded as 1 for “Yes” and 0 for “No”. ART adherence ART adherence level is the extent to which individuals follow their prescribed ART regimen as directed by healthcare providers. This includes taking medications and attending medical appointments regularly to monitor progress and address concerns. This variable was classified as follows: Good (≥95% adherence, which means at least 95% of pills were taken and only 1 out of 30 doses was missing), Fair (85%–94% adherence, which means missing 2–4 doses out of 30) or Poor (less than 85% adherence, meaning missing ≥5 doses out of 30).27 Viral load suppression According to the Sierra Leone HIV guideline, viral load suppression will be determined with an undetectable viral load of <1000 copies/ml at 6 months on ART. This variable was categorised as follows: patients were classified as viral load unsuppressed if patient on ART had a viral load of >1000 copies/mL in the past 6 months, patients were classified as viral load suppressed if patient on ART had a viral load of <1000 copies/mL in the past 6 months and not measured if the patient LTFU before the 6 months of assessment. Data collection method, procedure and quality control Data used in our study was extracted by trained HIV clinic staff who follow a standardised operating procedure (SOP) for data collection. This procedure was aligned with national guidelines and the standard ART entry and follow-up forms used in Sierra Leone’s ART clinics. To ensure data accuracy, a checklist was employed to record all relevant patient information. The checklist covered socio-demographic factors (eg, age, sex, HIV disclosure status), baseline clinical factors (eg, BMI, CD4 count, WHO clinical stage) and follow-up data (eg, TB screening, viral load, regimen change). A qualified public health professional supervised the entire data extraction process to ensure the completeness and consistency of the data. In cases of discrepancies, the data collector, supervisor and principal investigator worked together to resolve the issues. Regarding the clinical variables, such as WHO stage and BMI, these were recorded directly by nurses as part of routine patient care and then extracted from patient records by the trained data collectors. The data collection process was subject to strict quality control measures, including a pretest of the data collection instrument, a 2 day training session for the data collectors and supervisors and regular review of the data for consistency and completeness. Any incomplete or ambiguous records were cross-checked and corrected, and all errors were resolved before data compilation. Statistical analysis Data processing and description The dataset for this analysis was derived from five public health facilities in Sierra Leone. After exporting the combined records into R V.4.3.1, the initial cohort of 1560 patients was reviewed. Sixteen (16) patients were excluded due to missing critical variables, resulting in a final analytical sample of 1544 individuals who had initiated ART. A detailed explanation of the data processing and descriptions of all variables, including their attributes, can be found in the online supplemental information 1. Data analysis Categorical variables were summarised using frequencies and percentages, with group comparisons made via the χ2 test. Time-to-event data were analysed using incidence rates per 100 person-years with Poisson CIs. Survival probability and curves were generated with the Kaplan-Meier method and compared using the log-rank test. We employed a systematic approach to build a multivariable Cox proportional hazards model for identifying predictors of LTFU. First, candidate variables were screened using univariate Cox regression with a liberal threshold of p<0.20 to reduce the number of potential predictors. The proportional hazards assumption was tested using scaled Schoenfeld residuals. Violations observed for three predictors(online supplemental information 1,figure S1) were addressed by incorporating these variables as stratification factors. After stratification, the global test confirmed the assumption was met (online supplemental information 1,figure S2). Subsequently, forward stepwise selection based on the Akaike Information Criterion was applied to identify the final multivariable model. This approach balances the risk of omitting important confounders while avoiding overfitting. A detailed explanation can be found in the online supplemental information 1. Patient and public involvement There was no patient involvement from the conception of the research questions to the write-up of this research finding. However, the authors will disseminate the findings through conference presentations. Results Baseline socio-demographic characteristics of study participants Table 1 presents the baseline socio-demographic characteristics of 1544 HIV patients enrolled in the study. The cohort was predominantly female (67.7%) and urban-dwelling (73.5%). Clinically, most patients presented at an early stage of disease, with 82.3% having a baseline CD4 count ≥200 cells/mm³ and 82.3% in WHO Stage I or II at ART initiation. Approximately 79.2% had not disclosed their HIV status at baseline. Table 1Baseline socio-demographic characteristics of HIV patients attending antiretroviral therapy care in a selected government hospital in Sierra Leone, 2020–2024 (n=1544)CharacteristicLTFU n=550*Censored n=994*Overall n=1544*Statistic†P value†‡Age in years3.960.3 15–24144 (26.18%)236 (23.74%)380 (24.61%) 25–34192 (34.91%)373 (37.53%)565 (36.59%) 35–44133 (24.18%)214 (21.53%)347 (22.47%) 45+81 (14.73%)171 (17.20%)252 (16.32%)Sex0.7760.4 Female380 (69.09%)665 (66.90%)1045 (67.68%) Male170 (30.91%)329 (33.10%)499 (32.32%)Marital status2.000.4 Single290 (52.73%)489 (49.20%)779 (50.45%) Married227 (41.27%)447 (44.97%)674 (43.65%) Widower33 (6.00%)58 (5.84%)91 (5.89%)Education status2.710.4 No education137 (24.91%)250 (25.15%)387 (25.06%) Primary education61 (11.09%)97 (9.76%)158 (10.23%) Secondary education176 (32.00%)354 (35.61%)530 (34.33%) Tertiary education176 (32.00%)293 (29.48%)469 (30.38%)Occupational5.880.015‡ Employed407 (74.00%)677 (68.11%)1084 (70.21%) Unemployed143 (26.00%)317 (31.89%)460 (29.79%)HIV disclosure status57.0<0.001‡ Not disclosed493 (89.64%)729 (73.34%)1222 (79.15%) Disclosed57 (10.36%)265 (26.66%)322 (20.85%)Type of residence42.3<0.001‡ Rural92 (16.73%)318 (31.99%)410 (26.55%) Urban458 (83.27%)676 (68.01%)1134 (73.45%)Region124<0.001‡ Eastern94 (17.09%)286 (28.77%)380 (24.61%) Northern57 (10.36%)215 (21.63%)272 (17.62%) Northwestern7 (1.27%)29 (2.92%)36 (2.33%) Southern264 (48.00%)221 (22.23%)485 (31.41%) Western128 (23.27%)243 (24.45%)371 (24.03%)*n (%).†Pearson’s χ2 test.‡*p<0.05; **p<0.01; ***p<0.001.LTFU, loss to follow-up. Baseline socio-demographic characteristics of HIV patients attending antiretroviral therapy care in a selected government hospital in Sierra Leone, 2020–2024 (n=1544) *n (%). †Pearson’s χ2 test. ‡*p<0.05; **p<0.01; ***p<0.001. LTFU, loss to follow-up. Additionally, comparing the two outcome groups, females constituted the majority in both groups (LTFU: 69.09%; censored: 66.90%). HIV status non-disclosure was markedly more common in the LTFU group (89.64%) compared with the censored group (73.34%). Urban residence predominated in both groups, but the proportion of rural residents was nearly twice as high in the censored group (31.99%) as in the LTFU group (16.73%; p<0.001). Regional differences were also evident: the Southern region accounted for nearly half of LTFU patients (48.00%) but only 22.23% of censored patients, whereas the Eastern region represented 28.77% of censored patients versus 17.09% of LTFU patients (see table 1). Baseline clinical characteristics of study participants In this study, more than half of the participants (69.17%) had a normal BMI at baseline, and the majority had a CD4 count of ≥200 (82.32%). A considerable portion of the participants were classified in WHO Clinical Stage I&II (66.84%), and nearly all patients (96.63%) had a working functional status. All participants were initiated on a first-line ART regimen at baseline (table 2). Table 2Baseline clinical characteristics of HIV patients attending antiretroviral therapy care in a selected government hospital in Sierra Leone, 2020–2024 (n=1544)CharacteristicLTFU n=550*Censored n=994*Overall n=1544*Statistic†P value†‡BMI (Baseline)11.10.004‡ Underweight103 (18.73%)157 (15.79%)260 (16.84%) Normal weight391 (71.09%)677 (68.11%)1068 (69.17%) Overweight/Obese56 (10.18%)160 (16.10%)216 (13.99%)CD4 count (Baseline)3.410.065 <20084 (15.27%)189 (19.01%)273 (17.68%) ≥200466 (84.73%)805 (80.99%)1271 (82.32%)WHO clinical stage (Baseline)0.2230.6 Stage I&II449 (81.64%)821 (82.60%)1270 (82.25%) Stage III&IV101 (18.36%)173 (17.40%)274 (17.75%)Regimen (Baseline)>0.9 First-line550 (100%)994 (100%)1544 (100%) Second-line0 (0.00%)0 (0.00%)0 (0.00%)Functional status at baseline3.690.055 Ambulatory/Bedridden12 (2.18%)40 (4.02%)52 (3.37%) Working538 (97.82%)954 (95.98%)1492 (96.63%)*n (%).†Pearson’s χ2 test; Fisher’s exact test.‡*p<0.05; **p<0.01; ***p<0.001.BMI, body mass index; LTFU, loss to follow-up. Baseline clinical characteristics of HIV patients attending antiretroviral therapy care in a selected government hospital in Sierra Leone, 2020–2024 (n=1544) *n (%). †Pearson’s χ2 test; Fisher’s exact test. ‡*p<0.05; **p<0.01; ***p<0.001. BMI, body mass index; LTFU, loss to follow-up. Comparing the two outcome groups, BMI distribution differed significantly (p=0.004). Normal weight was the most common category in both groups (LTFU: 71.1%; censored: 68.1%). However, overweight/obesity was less frequent among LTFU patients (10.2%) than censored patients (16.1%), while underweight was slightly more common in LTFU (18.7% vs 15.8%). Follow-up infection, treatment and clinical characteristics Table 3 presents follow-up clinical and treatment characteristics. In the overall cohort, most patients underwent TB screening (72.7%), were virally suppressed (98.6%), had normal BMI (67.6%), CD4 ≥200 (83.9%), WHO stage I/II (86.1%) and good adherence (89.8%). Regimen changes occurred in 9.97% of patients. Table 3Follow-up infection, treatment and clinical characteristics of HIV patients attending antiretroviral therapy care in a selected government hospital in Sierra Leone, 2020–2024 (n=1544)VariablesLTFU n=550*Censored n=994*Overall n=1544*Statistic†P value†‡TB screening319<0.001‡ Yes250 (45.45%)872 (87.73%)1122 (72.67%) No300 (54.55%)122 (12.27%)422 (27.33%)TB status267<0.001‡ Negative233 (42.36%)815 (81.99%)1048 (67.88%) Positive34 (6.18%)40 (4.02%)74 (4.79%) Not done283 (51.45%)139 (13.98%)422 (27.33%)Regimen change during follow-up72.0<0.001‡ Yes7 (1.27%)147 (14.79%)154 (9.97%) No543 (98.73%)847 (85.21%)1390 (90.03%)Viral load at follow-up0.6 >1000 copies/ml6 (1.09%)12 (1.21%)18 (1.17%) <1000 copies/ml542 (98.55%)981 (98.69%)1523 (98.64%) Not measured2 (0.36%)1 (0.10%)3 (0.19%)BMI at follow-up41.2<0.001‡ Underweight72 (13.09%)55 (5.53%)127 (8.23%) Normal weight382 (69.45%)661 (66.50%)1043 (67.55%) Overweight/Obese96 (17.45%)278 (27.97%)374 (24.22%)CD4 count at follow-up4.500.034‡ <200103 (18.73%)145 (14.59%)248 (16.06%) ≥200447 (81.27%)849 (85.41%)1296 (83.94%)WHO clinical stage at follow-up27.0<0.001‡ Stage I&II440 (80.00%)890 (89.54%)1330 (86.14%) Stage III&IV110 (20.00%)104 (10.46%)214 (13.86%)Regimen change during follow-up70.7<0.001‡ First-line543 (98.73%)849 (85.41%)1392 (90.16%) Second-line7 (1.27%)145 (14.59%)152 (9.84%)Adherence14.3<0.001‡ Good499 (90.73%)888 (89.34%)1387 (89.83%) Fair33 (6.00%)95 (9.56%)128 (8.29%) Poor18 (3.27%)11 (1.11%)29 (1.88%)*n (%).†Pearson’s χ2 test; Fisher’s exact test.‡*p<0.05; **p<0.01; ***p<0.001.BMI, body mass index; LTFU, loss to follow-up; TB, tuberculosis. Follow-up infection, treatment and clinical characteristics of HIV patients attending antiretroviral therapy care in a selected government hospital in Sierra Leone, 2020–2024 (n=1544) *n (%). †Pearson’s χ2 test; Fisher’s exact test. ‡*p<0.05; **p<0.01; ***p<0.001. BMI, body mass index; LTFU, loss to follow-up; TB, tuberculosis. Comparing LTFU (n=550) and censored (n=994) groups, differences were observed for nearly all variables (p<0.001), except viral load (p=0.6). TB screening was markedly lower among LTFU patients (45.5%) than censored (87.7%). Consequently, TB status showed that LTFU had higher proportions of “Not done” (51.5% vs 14.0%) and positive TB (6.2% vs 4.0%). Regimen changes were rare in LTFU (1.3%) compared with censored (14.8%). At follow-up, LTFU patients had higher rates of underweight (13.1% vs 5.5%) and lower rates of overweight/obesity (17.5% vs 28.0%). CD4<200 was more common in LTFU (18.7% vs 14.6%; p=0.034). WHO stage III/IV was twice as frequent in LTFU (20.0% vs 10.5%). Adherence differed greatly: poor adherence was three times higher in LTFU (3.3% vs 1.1%), while fair adherence was lower (6.0% vs 9.6%). Incidence of loss to follow-up (LTFU) among HIV patients attending ART care This study reported a total follow-up time of 5748.6 person-years, with participants followed for a minimum of 0.25 years and a maximum of 5 years after the initiation of ART, resulting in a median follow-up time of 4.25 years. The study revealed the following outcomes for patients: 550 (35.6%) were LTFU, 853 (55.2%) remained on follow-up, 64 (4.1%) were transferred out and 77 (5.0%) died. The cumulative incidence rate of LTFU was 9.57% per 100 person-years. The total follow-up time for the cohort was 5748.6 person-years, with an overall median follow-up of 4.25 years (range: 0.08–5.00 years). For cause-specific incidence rate calculations, the total cohort person-years were used as the denominator for all outcome groups. As shown in table 4, the median follow-up time varied substantially by outcome. Patients still on ART had the longest median follow-up (4.58 years), while those who died, transferred or were LTFU had shorter medians of 3.17, 3.21 and 3.00 years, respectively. Among patients classified as LTFU, the minimum follow-up duration was 0.25 years (3 months), which is consistent with the study definition of LTFU requiring at least 90 days without a documented clinical encounter and/or ART refill after the last expected clinical contact. Table 4Outcome of HIV patients attending antiretroviral therapy care in selected government hospitals in Sierra Leone, 2020–2024 (n=1544)OutcomeNPercentagePerson-years*Incidence rate per 100 PY (95% CI)Follow-up yearsmedian (Min-Max) LTFU55035.6%5748.69.57 (8.78 to 10.37)3.00 (0.25–4.75) Patients on follow-up85355.2%5748.614.84 (13.85 to 15.85)4.58 (3.17–5.00) Transfer644.1%5748.61.11 (0.85 to 1.39)3.21 (0.25–4.75) Death775.0%5748.61.34 (1.04 to 1.65)3.17 (0.25–4.75)*Person-years represent total cohort follow-up time used as the denominator for all cause-specific incidence rates. Median follow-up is calculated separately for each outcome group.LTFU, loss to follow-up. Outcome of HIV patients attending antiretroviral therapy care in selected government hospitals in Sierra Leone, 2020–2024 (n=1544) *Person-years represent total cohort follow-up time used as the denominator for all cause-specific incidence rates. Median follow-up is calculated separately for each outcome group. LTFU, loss to follow-up. Survival analysis The Kaplan-Meier retention curve shows a gradual decline in the probability of remaining in care during the first 20 months of follow-up (from 100% to 98%), followed by a steeper decline between 20 and 60 months (reaching 80% at 60 months). This pattern suggests that the risk of LTFU increases after the initial period of engagement, possibly reflecting late disengagement or cumulative barriers to retention. The ‘Number at risk’ table illustrates progressive attrition, consistent with the observed decline in retention probability (figure 1). Kaplan-Meier estimates of loss to follow-up among adult patients attending antiretroviral therapy care in selected government hospitals in Sierra Leone, 2020–2024 (n=1544). The Kaplan-Meier curve indicates that the survival probability was lower for patients who are employed across all time points (refer to online supplemental information 1,table S2. Similarly, patients who did not disclose their HIV status exhibited lower survival probabilities across all time points than those who disclosed their status. Additionally, a patient’s residence was associated with probability at baseline, with patients living in urban areas exhibiting a lower survival probability. Furthermore, patients from the southern, western, northwestern and eastern regions also demonstrated reduced survival probabilities at baseline. The variance in survival probability was lower among patients classified as underweight or of normal weight at baseline (refer to online supplemental information 2 table S1,Figure S1). Moreover, the Kaplan-Meier curve indicates that infection characteristics were associated with survival probability at follow-up. Patients who were not screened for TB and those who tested positive exhibited lower survival probabilities at follow-up. Additionally, survival probability was lower regarding treatment-related characteristics; patients who did not change their regimen, those on first-line regimens and those who had good or poor levels of adherence had lower survival probability (refer to online supplemental information 2,table S1 and figure S2). Finally, patients who had a viral load exceeding 1000 copies/ml, were categorised as underweight or of normal weight, possessed CD4 counts below 200 and were in WHO clinical Stages III&IV at follow-up all showed lower survival probabilities at follow-up (refer to online supplemental information 2,table S1 and figure S3). Predictors of loss to follow-up In this study, to identify the predictors of LTFU, both univariate and multivariate Cox regression analyses were performed. In the univariate Cox regression analysis, we found that TB screening, region, WHO clinical stage at follow-up, regimen change, residence, HIV disclosure status and BMI at follow-up were statistically significant predictors of LTFU (table 5). Table 5Univariate and Multivariate Cox regression model to identify predictors of loss to follow-up among patients who attend antiretroviral therapy in selected government hospitals in Sierra Leone, 2020–2024 (n=1544)CharacteristicCategoriesN (%)Univariable HR (95% CI, P value)Multivariable HR (95% CI, P value)TB screeningYes1122 (72.7)11No422 (27.3)5.61 (4.71 to 6.68, p<0.001)4.47 (3.66 to 5.46, p<0.001)RegionWestern371 (24.0)11Eastern380 (24.6)0.63 (0.48 to 0.82, p=0.001)2.13 (1.49 to 3.03, p<0.001)Northern272 (17.6)0.58 (0.43 to 0.80, p=0.001)0.99 (0.72 to 1.38, p=0.976)Northwestern36 (2.3)0.50 (0.24 to 1.08, p=0.077)1.78 (0.78 to 4.06, p=0.171)Southern485 (31.4)1.86 (1.50 to 2.30, p<0.001)1.97 (1.57 to 2.48, p<0.001)WHO clinical stage at follow-upStage I&II1330 (86.1)––Stage III&IV214 (13.9)1.94 (1.58 to 2.39, p<0.001)1.51 (1.20 to 1.89, p<0.001)Regimen changeYes154 (10.0)11No1390 (90.0)11.16 (5.29 to 23.52, p<0.001)5.12 (2.39 to 11.00, p<0.001)ResidenceUrban1134 (73.4)11Rural410 (26.6)0.47 (0.37 to 0.58, p<0.001)0.58 (0.43 to 0.79, p=0.001)HIV disclosure statusDisclosed322 (20.9)11Not disclosed1222 (79.1)2.69 (2.04 to 3.54, p<0.001)1.63 (1.19 to 2.25, p=0.003)BMI at follow-upNormal weight1043 (67.6)––Underweight127 (8.2)1.79 (1.39 to 2.31, p<0.001)1.37 (1.05 to 1.78, p=0.019)Overweight/Obese374 (24.2)0.62 (0.50 to 0.78, p<0.001)0.99 (0.79 to 1.25, p=0.934)1 refers to the reference categories for each characteristic. P value from the Wald test.BMI, body mass index; TB, tuberculosis. Univariate and Multivariate Cox regression model to identify predictors of loss to follow-up among patients who attend antiretroviral therapy in selected government hospitals in Sierra Leone, 2020–2024 (n=1544) 1 refers to the reference categories for each characteristic. P value from the Wald test. BMI, body mass index; TB, tuberculosis. In the multivariate analysis, TB screening, region, WHO clinical stage at follow-up, regimen change, residence, HIV disclosure status and BMI at follow-up remained statistically significant predictors of LTF. Accordingly, patients who were not screened for TB were 4.47 times higher at LTFU compared with those who were screened for TB (HR=4.47, 95% CI 3.66 to 5.46, p<0.001). Patients at WHO clinical stages III & IV were 1.51 times more likely to be LTFU than those who were at WHO stage I & II (HR=1.51, 95% CI 1.20 to 1.89, p<0.001). Regimen change was a strong predictor of LTFU, with patients who did not undergo regimen change having a 5.12 times higher risk of LTFU compared with those who did undergo regimen change (HR=5.12, 95% CI 2.39 to 11.00, p<0.001). Patients who did not disclose their HIV status were 1.63 times more likely to LTFU than those who did (HR=1.63, 95% CI 1.19 to 2.25, p=0.003). The risk of LTFU was 1.37 times higher in underweight patients compared with normal weight patients, with HRs of HR=1.37 (95% CI 1.05 to 1.78, p=0.019). Patients from the Southern area had a 1.97 times higher risk of LTFU, while those from the Eastern region had a 2.13 times higher risk, both compared with patients from the Western region (HR=1.97, 95% CI 1.49 to 3.03, p<0.001; HR=2.13, 95% CI 1.57 to 2.48, p<0.001). Conversely, the risk of LTFU was 0.58 times lower for patients who reside in the rural areas compared with those residing in urban areas, with HRs of HR=0.58 (95% CI 0.43 to 0.79, p=0.001). Discussion Numerous research studies have identified LTFU as a major barrier to the effective implementation of ART programmes in resource-limited settings.6 This barrier complicates the assessment of ART care, making it necessary to implement targeted interventions that improve patients’ outcomes, such as enhanced patient tracking systems and community outreach programmes.16 Our study, conducted from 2020 to 2024, coincided with the COVID-19 pandemic, during which Sierra Leone declared a state of emergency and implemented travel and clinic restrictions. These disruptions, including clinic closures, mobility limitations, and patient fear of infection, likely contributed to the elevated LTFU rates observed.41 Understanding how these pandemic-era conditions affected ART retention is important for developing resilient HIV care models that can better withstand future health crises. This study aimed to assess the incidence and predictors of LTFU after ART initiation in public health facilities in Sierra Leone. Among the 1544 HIV patients enrolled in ART care during the study period, over half remained engaged in care, while 4% were transferred to other facilities, 5% died, and 35.6% were LTFU. The proportion of LTFU in this study is consistent with results from earlier research in Central Ethiopia,15 Cameroon7 and Uganda.10 The rate of patients LTFU in this study was higher than that reported in studies from Nigeria,42 Northwest Ethiopia,6 Eritrea32 and Sub-Saharan African settings where studies were conducted prior to 2020. This discrepancy may be due to variations in healthcare infrastructure, patient population characteristics and regional challenges, such as cost of care and social stigma. The 35.6% LTFU rate noted in this cohort (2020–2024) should be understood within the framework of the COVID-19 pandemic. Sierra Leone declared a state of emergency in March 2020, imposing curfews and clinic closures, which strained healthcare services during the early stages of the pandemic. Although case counts remained low, and restrictions were gradually lifted, many health services faced major disruptions. This pandemic context significantly impacts our study. Since all patients-initiated ART during the pandemic, we could not perform a clear pre- and post-pandemic analysis, making it difficult to separate the effects of COVID-19 from other factors. The high LTFU rate observed aligns with findings from other studies, such as those in Northwest Ethiopia43 and Chad,44 which documented significantly higher LTFU during the COVID-19 period. The overall incidence rate of LTFU in this study was 9.57% per 100 person-years. This rate is consistent with findings from Northwest Ethiopia,12 Southern Ethiopia,17 Central Ethiopia,15 Cameroon7 and Rwanda.19 It was, however, higher than those observed in Uganda,18 Southwest Ethiopia22 and Central Ethiopia.15 Furthermore, the LTFU incidence rate in our study was lower compared with research conducted in Liberia,8 Malawi11 and Uganda.10 These variations may be attributed to differences in healthcare infrastructure, time, patient population characteristics, transfer-out rates, follow-up periods and the standard protocol of ART service based on WHO guidelines across the different study settings.27 Methodological factors also matter: differing LTFU definitions (eg, ≥60 vs ≥90 days late), analytic windows and censoring rules can shift incidence estimates; studies incorporating competing risks or more granular time-updated covariates often report lower crude LTFU but higher refined estimates of disengagement pathways. In this study, after making adjustments and stratifications, we found that TB screening, region, WHO clinical stage at follow-up, regimen change, residence, HIV disclosure status and BMI at follow-up are statistically significant predictors of LTFU. Patients who were not screened for TB had a 4.47 times higher risk of LTFU than those screened for TB. This strong association demonstrates the importance of integrated TB screening in safeguarding individuals; early detection of TB or TB-related symptoms can expedite treatment and enhance patient involvement in HIV care services.29 The absence of TB screening might also reflect broader systemic weaknesses in patient monitoring and integrated management services that are essential for retaining patients in treatment, such as inadequate follow-up procedures and insufficient resources for patient engagement.29 Operationally, no screening may serve as a proxy for missed clinical touchpoints or congestion at triage; embedding hard-stop TB screening and same-day evaluation within ART clinics could therefore function as both a clinical safeguard and a retention intervention. In comparison to those in WHO clinical stages I&II, patients in advanced WHO clinical stages III&IV had a 1.51 times higher risk of LTFU. This finding is consistent with studies indicating an association between advanced WHO clinical stage and LTFU,6 13 17 23 as these patients may face more intricate health challenges that hinder adherence and mobility.22 23 This condition is evident in advanced WHO clinical disease, where individuals with weakened immune systems are at higher risk of opportunistic infections, potentially resulting in severe health complications and unreported deaths.22 Another plausible explanation could be the psychological distress and stigma associated with advanced disease stages, which may lead to patients avoiding healthcare facilities and subsequently being LTFU. Programmatically, the evidence underscores the importance of intensified early retention packages, frequent follow-up in the first 3–6 months, rapid tracing of missed visits and expedited management of intercurrent illness for patients presenting with advanced disease. Regimen change emerged as another important predictor. Patients who did not experience a regimen change had a significantly higher risk of LTFU (HR=5.12, 95% CI 2.39 to 11.00, p<0.001) than those whose regimen was adjusted. A change in the treatment plan may show proactive clinical management, such as lowering the risk of drug toxicity or addressing ineffective treatment. It may also mean better counselling and follow-up with patients which lowers the number of people who stop taking their medicine.25 The magnitude of this association suggests potential missed opportunities for timely optimisation; routine multidisciplinary case reviews and clear pathways for switching per national guidelines may reduce disengagement by addressing side effects and perceived treatment failure before patients drop out. Non-disclosure of HIV status was independently associated with a 63% higher risk of LTFU (HR=1.63; 95% CI 1.19 to 2.25), a finding that is consistent with studies conducted in Northwest Ethiopia,12 Northeast Ethiopia,30 Southwest Ethiopia,22 the Accra Regional Hospital in Ghana31 and in Sierra Leone where qualitative evidence demonstrating that stigma-driven secrecy fundamentally undermines the social support networks critical for sustained ART engagement.34 Patients who have not disclosed their status to a trusted person lack a treatment supporter who can provide appointment reminders, accompany them to the facility and offer emotional reinforcement during periods of treatment fatigue or side effects. Importantly, the relationship between non-disclosure and LTFU in Sierra Leone is not simply a matter of individual choice; it is deeply embedded in the structural reality of HIV-related stigma, where disclosure carries genuine risks of discrimination, relationship breakdown and social exclusion.34 Effective disclosure counselling in this context must therefore focus on helping patients identify safe, trustworthy confidants rather than mandating broad disclosure and must be delivered within a framework that actively addresses stigma at the facility and community levels. Integration of peer support programmes in which PLHIV who have successfully navigated disclosure support newly initiated patients represents a promising, low-cost strategy for improving both disclosure rates and retention in Sierra Leone’s ART programme. This study identified BMI during follow-up as a significant predictor of LTFU among adults living with HIV. The risk of LTFU was 1.37 times higher in underweight patients compared with normal-weight patients, with HRs of HR=1.37 (95% CI 1.05 to 1.78, p=0.019). This result is consistent with earlier research conducted in Guinea-Bissau,9 Tanzania,21 Uganda,10 Malawi11 and various studies in Ethiopia.23 24 26 28 A lower BMI can indicate severe disease and malnutrition, both linked to poor treatment outcomes and difficulties in retention.21 25 We reinforced the notion that comprehensive nutritional interventions, including therapeutic feeding programmes and regular malnutrition assessment, are essential throughout the ART treatment regimen.26 Given resource constraints, integrating routine BMI screening with simple triage to targeted supplementation, food support and social protection referral paired with EMR flags for concerning weight trends could strengthen retention among undernourished patients. Significant regional disparities in LTFU were observed, with patients in the Eastern (HR=2.13; 95% CI 1.49 to 3.03) and Southern (HR=1.97; 95% CI 1.57 to 2.48) regions having substantially higher LTFU hazard compared with the Western region. This finding is consistent with a previous study conducted in Eritrea,30 which highlighted regional disparities in HIV treatment outcomes and mortality based on geographic location. These disparities likely reflect a combination of genuine geographic and structural barriers rather than individual-level factors alone. The Eastern and Southern regions are characterised by lower facility density, greater distances between communities and ART service delivery points, weaker community health worker infrastructure and historically lower levels of health system investment compared with the Western region.4 It is important to note, however, that the extreme magnitude of some regional HRs may partly reflect the statistical instability associated with smaller subgroup sample sizes in a multi-facility cohort, and should be interpreted as indicative of the direction and approximate magnitude of regional disparities rather than as precise effect sizes. Geographically targeted interventions including community ART groups, multi-month dispensing and mobile ART refill services are urgently needed in the Eastern and Southern regions to address these structural barriers and reduce the regional equity gap in ART retention. Residential status was a significant determinant. Patients residing in rural areas had a 42% lower risk of LTFU compared with urban residents (HR=0.58, 95% CI 0.43 to 0.79, p=0.001). This finding contradicts a study done in Ethiopia, whereby rural residents had an increased risk of LTFU due to the availability of transport expenses, long distances to ART health facilities, social stigma and low levels of HIV knowledge in these communities.24 One possible reason for our study’s finding is that living in a rural area may protect by limiting mobility and fostering stronger community support networks, which encourage individuals to remain engaged in their care.14 The observed protection may stem from the higher HIV burden in urban areas of Sierra Leone, social interaction, the ART care model in rural settings or other factors that were not accounted for in our study. Further research is necessary to explore these potential explanations and develop targeted interventions to reduce the risk of LTFU in urban areas. In practice, urban mobility and clinic congestion may drive missed appointments; multi-month dispensing, fast-track refills and flexible clinic hours could mitigate these urban-specific barriers. Contrary to findings from some other African settings where rural residence is a risk factor for LTFU due to distance barriers, rural residence was independently associated with a 42% lower risk of LTFU in this cohort (HR=0.58; 95% CI 0.43 to 0.79). This urban-rural paradox warrants careful interpretation. Several mechanisms may explain this finding in the Sierra Leonean context. First, rural communities in Sierra Leone may provide stronger informal social support networks including extended family structures and community health worker presence that facilitate consistent ART attendance in ways that are less available in the more fragmented social environments of urban centres.34 Second, the COVID-19 pandemic (2020–2021), which coincided with the study period, likely had a disproportionate effect on urban ART retention due to travel restrictions, fears of clinic congestion and economic disruption, whereas rural areas may have experienced relatively less mobility disruption related to the pandemic. Third, urban patients in Sierra Leone may face greater competing demands on their time, including informal employment and caregiving responsibilities that make consistent clinic attendance more difficult. These findings suggest that urban-specific retention strategies, including multi-month dispensing, fast-track refill services and flexible clinic hours, are particularly needed to address the paradoxically higher LTFU risk among urban patients. Finally, several findings from this study are consistent with pandemic-related effects on ART retention. First, the high overall LTFU rate (35.6%) during 2020–2024 coincides with a period when Sierra Leone’s health system, like others globally, experienced significant strain from COVID-19 response. Second, the protective effect of rural residence (HR=0.58) may reflect differential pandemic impacts: urban areas typically experienced stricter and longer lockdowns, greater COVID-19 caseloads diverting health workers, and higher population density increasing fear of facility-based infection. Third, the exceptionally strong association between lack of TB screening and LTFU (HR=4.47) may partially reflect the suspension or reduction of integrated TB/HIV services during pandemic peaks, when resources and attention shifted to COVID-19. Fourth, the strong link between not changing the regimen and LTFU (HR=5.12) could mean that clinical reviews and regimen optimisation were put off when the health system was under a lot of stress. Policy implications Taken together, these findings highlight modifiable service-level levers: TB screening, proactive regimen optimisation, disclosure-supportive counselling, nutrition support and geographically tailored retention strategies that align with national guidelines and are feasible within Sierra Leone’s referral-hospital–anchored ART model, which serve as central hubs for ART initiation and care, coordinating treatment for patients referred from district health centres and peripheral health units while providing specialised care. To reduce LTFU in Sierra Leone’s referral-hospital–anchored ART programme, make TB screening a hard-stop at every visit (electronic/paper checklist with WHO symptom screen and on-site diagnostics), ensure same-day evaluation for presumptive/confirmed TB, institutionalise rapid regimen reviews (weekly/mid-weekly multidisciplinary huddles, timely regimen switches with documented counselling), standardise disclosure-supportive counselling that identifies a treatment supporter and mitigates stigma, and embed routine nutrition care (BMI screening, malnutrition triage, linkage to supplementation/food support and social protection and EMR weight-trend flags); prioritise high-burden geographies (Southern and Eastern regions) for intensified tracing, decentralised refills, community ART groups and peer navigation, while adapting urban services with multi-month dispensing, flexible clinic hours, fast-track refills and SMS/voice reminders; concentrate early-phase retention supports in the first 3–6 months (frequent counselling, proactive reminders, rapid tracing, differentiated care for advanced disease); and strengthen accountability through cohort dashboards tracking TB screening coverage, on-time visits, regimen changes, BMI trends and tracing outcomes, coupled with SOP-driven tracing (24–72-hour call-backs, 7–14 day CHW/home visits) and standardised outcome coding to distinguish true LTFU, transfers and deaths. Strengths and limitations This study has several strengths. It draws on a large, multi-facility cohort spanning five provincial referral hospitals with diverse catchment populations, enhancing the generalisability of findings within Sierra Leone’s public ART programme. The study applied robust time-to-event methods, including incidence rate estimation, Kaplan-Meier survival analysis and a stratified Cox proportional hazards model, with systematic data harmonisation and standardised operational definitions. Furthermore, it identified several potentially modifiable factors associated with LTFU, including TB screening, regimen change, HIV disclosure status, BMI, clinical stage, region and residence, which may inform future retention interventions. Nevertheless, several limitations should be considered. First, the retrospective use of routine programmatic data limited the availability and completeness of some variables, raising the possibility of residual confounding from unmeasured sociodemographic, clinical, treatment and behavioural factors. In addition, routine records may be subject to outcome misclassification. Some patients classified as LTFU may have transferred to other ART facilities without formal documentation (silent transfers), while others may have experienced prolonged interruptions in care and subsequently re-engaged after the study-defined LTFU period. These factors could lead to overestimation of LTFU incidence and may influence the magnitude of observed associations. Although sensitivity analyses using a more conservative≥180 day definition of LTFU yielded broadly similar findings, some degree of outcome misclassification may remain because patient tracing data were unavailable to verify long-term disengagement from care. Finally, although the study period overlapped with the COVID-19 pandemic, the study was not designed to directly evaluate the impact of pandemic-related disruptions on retention in care. Because all participants initiated ART during 2020, a pre-pandemic comparison group was unavailable, limiting our ability to isolate the independent contribution of COVID-19 to the observed LTFU rates. Consequently, observations regarding potential pandemic-related effects should be interpreted cautiously and not as evidence of a causal relationship. Future prospective studies incorporating patient tracing and qualitative investigations are warranted to better understand the determinants of LTFU and retention in care among PLHIV in Sierra Leone. Conclusion In a multi-facility cohort from Sierra Leone’s provincial referral hospitals, LTFU after ART initiation was high. Lack of TB screening, advanced WHO stage, absence of timely regimen changes, non-disclosure of HIV status, low BMI and region of residence independently associated LTFU, while rural residence was associated with a lower risk of LTFU. These findings identify potentially modifiable factors that may inform interventions aimed at improving retention in care. Universal and integrated TB screening, institutionalise rapid regimen reviews, scale disclosure-supportive counselling, strengthen nutrition services in ART clinics and concentrate retention resources in high-risk regions and urban settings may help improve retention in care. Implementing these focused strategies can meaningfully improve retention and treatment outcomes in Sierra Leone. Supplementary Material Reviewer comments Author's manuscript