Background and aim There are persistent concerns regarding the potential adverse effects of in utero ART exposure. Whereas the association between untreated, advanced HIV disease and adverse pregnancy (APOs)outcomes is well documented, non or few have focused on APOs subject to immune reconstitution inflammatory syndrome (IRIS) as a predictor, devoid of differential aspect of either paradoxical or unmasking IRIS. The current study sought to investigate the incidence and demographic indicators of IRIS, IRIS type, survival-time to, and its prediction of APOs. Methods An active records study was conducted between June 2019 and March 2020 among ART-naive pregnant women attending the antenatal care units (ANCu) at the Kenyatta National and Mbagathi Hospitals, Nairobi, Kenya. Participants were aged between 20 and 49 years and had a confirmed HIV-positive test. IRIS diagnosis was adjudicated for accuracy and consistency by an independent review committee. Baseline demographic characteristics including: age, education level, religion, marital status, residence, occupation and economic status were recorded. IRIS incidence was assessed using the International Network for Studies Against HIV-Associated IRIS (INSHI) during the first three months after ART initiation. Bivariate analysis via test of Pearsons for demographics relative to IRIS type was performed. The association and its strength between the IRIS type and APOs were further established through Pearson Chi-Square test and Phi and Cramers V tests respectively. Kaplan Meier analysis test estimated the survival time to APOs using log rank test statistic. Multivariate Cox-regression analysis for Pre-ART demographics on IRIS type incidence was performed using survival package in SPSS. Decision tree analysis for predictive modelling on APOs relative to demographics was conducted. Results The incidence of IRIS was 25% (n=133) among the 532 ART-naive pregnant women with 97 (72.9%) presenting with unmasking IRIS, significantly associated with APOs [X (1) = 4.911a, P = 0 .027]. Maternal age, between 40-49 years had a positive co-efficient with unmasking IRIS [(Wald test = 0.329, Wald test = 1.011, (HR = 1.389, 95% C.I 0.732 - 2.638, P = 0.325]. The cumulative survival function evaluating all the demographic characteristics (as covariates) indicated that, over 80% of the ART naive pregnant women survived IRIS diagnosis for six weeks, while half of them had been diagnosed with it at approximately two months. Kaplan-Meier survival function showed that, women diagnosed with unmasking IRIS compared to those diagnosed with paradoxical IRIS survived longer before experiencing an APO (χ2 = 5.292, Log Rank test = 0.021), cumulative hazard, [HR = 0.18 and 0.4] respectively. Decision tree analysis demonstrated that, women aged 30-39 had most of APOs (P = 0.688). Conclusion Unmasking IRIS was the most common, associated highly with APOs that were experienced much later compared to those predicted by paradoxical IRIS with older age being a plausible predictor. The survival time to experiencing an APO was longer for women presenting with unmasking IRIS as opposed to paradoxical IRIS, supporting the need to conduct further research on possible APOs due to paradoxical IRIS.