These findings highlight the importance of reducing stigma in HIV care delivery through services that recognise, address, and mitigate stigma to ensure equitable, non-discriminatory, and person-centred services(11, 20). They align with multi-level stigma intervention evidence in LMICs showing that organisational climate and community engagement jointly enhance sustainability. High agreement across both HP- and AP-PHCs reflects consensus on the value of engaging PWH and KPs as co-implementers, consistent with evidence supporting contact-based strategies for knowledge translation in stigma reduction interventions (3, 10).

A significant association was found between occupational role and PHC performance (p=0.012). HP-PHCs had a higher proportion of doctors and nurses, whereas respondents from AP-PHCs were predominantly midwives and laboratory technicians. Although this pattern may reflect differences in staffing composition across facilities, it could also indicate sampling bias, with higher engagement from clinicians directly involved in IPSD implementation in HP-PHCs. Nonetheless, the presence of more clinically trained staff in these facilities may contribute to stronger leadership and a greater capacity to sustain intervention fidelity.

The IPSD intervention leveraged Tension for Change by creating an urgency for transformation, and its compatibility with existing workflows facilitated acceptance. These findings align with implementation science literature on organisational alignment in intervention delivery(21, 22).

Significant differences were observed in the Learning Climate and Executing constructs, both of which are core components of implementation fidelity. High-performing PHCs reported greater perceived technical capacity (p=0.033) and knowledge retention, particularly in HIV and STI epidemiology (p=0.033). These findings suggest that fidelity is influenced by a facility’s internal capacity to absorb and operationalise intervention components. While stigma-reduction content was delivered by KP and PHIV experts, the technical modules were facilitated by clinical professionals(16), which may have contributed to variations in how technical information was retained and applied across facilities.

The Organisational Incentives and Rewards sub-construct was used to assess both adoption and fidelity outcomes. While no significant difference was observed in adoption scores between HP- and AP-PHCs (p=0.25), fidelity scores differed significantly (p=0.032), with HP-PHCs reporting higher perceptions of organisational support and recognition. This finding suggests that beyond initial adoption, sustained implementation fidelity may be reinforced by effective capacity-building efforts and non-financial incentives that foster a more supportive and motivated work environment(23).

External policy influences were evident, as national mandates ensured uniform IPSD implementation across PHCs (p=0.802), underscoring the importance of government accountability and policy alignment in sustaining intervention fidelity(23, 24).

A more integrated view of the findings using the CFIR framework suggests that organisational and policy factors work together to influence implementation fidelity. Organisational incentives and a supportive learning climate (Inner Setting) help improve staff motivation and competence, while alignment with national policies and mandates (Outer Setting) strengthens accountability and consistency. These results indicate that maintaining fidelity depends not only on individual knowledge but also on the combined support from within the organisation and from external policies(23). Embedding CBPA principles within national HIV service accreditation processes could institutionalise stigma-reduction standards(25).

The study acknowledges several limitations. The challenges encountered in achieving the intended sample size may have compromised statistical power, and the geographic scope confined to selected districts within Greater Jakarta may limit the generalisability of the findings. As the data were self-reported, responses may have been subject to social desirability or recall bias, potentially leading to an overestimation of positive behaviours or attitudes. Additionally, the cross-sectional design restricted causal inference. Future research should adopt longitudinal designs to evaluate sustained impacts and conduct comparative studies across diverse settings to identify additional contextual factors that influence implementation success.

Given the limited sample, multivariable modelling was not feasible without compromising model validity and findings are exploratory. The achieved sample size (n=120) limits the study’s statistical power to detect small associations, and results should therefore be interpreted as indicative rather than confirmatory. Although the achieved sample was lower than the estimated 286, the study’s analytical focus remained descriptive and exploratory. Therefore, associations were explored descriptively within the CFIR framework to generate insights for future hypothesis testing. The findings are therefore interpreted with appropriate caution.