BACKGROUND: Serious illness is often characterised by physical/psychological problems, family support needs, and high healthcare resource use. Hospital-based specialist palliative care (HSPC) has developed to assist in better meeting the needs of patients and their families and potentially reducing hospital care expenditure. There is a need for clarity on the effectiveness and optimal models of HSPC, given that most people still die in hospital and also to allocate scarce resources judiciously. OBJECTIVES: To assess the effectiveness and cost-effectiveness of HSPC compared to usual care for adults with advanced illness (hereafter patients) and their unpaid caregivers/families. SEARCH METHODS: We searched CENTRAL, CDSR, DARE and HTA database via the Cochrane Library; MEDLINE; Embase; CINAHL; PsycINFO; CareSearch; National Health Service Economic Evaluation Database (NHS EED) and two trial registers to August 2019, together with checking of reference lists and relevant systematic reviews, citation searching and contact with experts to identify additional studies. SELECTION CRITERIA: We included randomised controlled trials (RCTs) evaluating the impact of HSPC on outcomes for patients or their unpaid caregivers/families, or both. HSPC was defined as specialist palliative care delivered by a palliative care team that is based in a hospital providing holistic care, co-ordination by a multidisciplinary team, and collaboration between HSPC providers and generalists. HSPC was provided to patients while they were admitted as inpatients to acute care hospitals, outpatients or patients receiving care from hospital outreach teams at home. The comparator was usual care, defined as inpatient or outpatient hospital care without specialist palliative care input at the point of entry into the study, community care or hospice care provided outside of the hospital setting. DATA COLLECTION AND ANALYSIS: We used standard methodological procedures expected by Cochrane. We assessed risk of bias and extracted data. To account for use of different scales across studies, we calculated standardised mean differences (SMDs) with 95% confidence intervals (CIs) for continuous data. We used an inverse variance random-effects model. For binary data, we calculated odds ratio (ORs) with 95% CIs. We assessed the evidence using GRADE and created a 'Summary of findings' table. Our primary outcomes were patient health-related quality of life (HRQoL) and symptom burden (a collection of two or more symptoms). Key secondary outcomes were pain, depression, satisfaction with care, achieving preferred place of death, mortality/survival, unpaid caregiver burden, and cost-effectiveness. Qualitative data was analysed where available. MAIN RESULTS: We identified 42 RCTs involving 7779 participants (6678 patients and 1101 caregivers/family members). Twenty-one studies were with cancer populations, 14 were with non-cancer populations (of which six were with heart failure patients), and seven with mixed cancer and non-cancer populations (mixed diagnoses). HSPC was offered in different ways and included the following models: ward-based, inpatient consult, outpatient, hospital-at-home or hospital outreach, and service provision across multiple settings which included hospital. For our main analyses, we pooled data from studies reporting adjusted endpoint values. Forty studies had a high risk of bias in at least one domain. Compared with usual care, HSPC improved patient HRQoL with a small effect size of 0.26 SMD over usual care (95% CI 0.15 to 0.37; I = 3%, 10 studies, 1344 participants, low-quality evidence, higher scores indicate better patient HRQoL). HSPC also improved other person-centred outcomes. It reduced patient symptom burden with a small effect size of -0.26 SMD over usual care (95% CI -0.41 to -0.12; I = 0%, 6 studies, 761 participants, very low-quality evidence, lower scores indicate lower symptom burden). HSPC improved patient satisfaction with care with a small effect size of 0.36 SMD over usual care (95% CI 0.41 to 0.57; I = 0%, 2 studies, 337 participants, low-quality evidence, higher scores indicate better patient satisfaction with care). Using home death as a proxy measure for achieving patient's preferred place of death, patients were more likely to die at home with HSPC compared to usual care (OR 1.63, 95% CI 1.23 to 2.16; I = 0%, 7 studies, 861 participants, low-quality evidence). Data on pain (4 studies, 525 participants) showed no evidence of a difference between HSPC and usual care (SMD -0.16, 95% CI -0.33 to 0.01; I = 0%, very low-quality evidence). Eight studies (N = 1252 participants) reported on adverse events and very low-quality evidence did not demonstrate an effect of HSPC on serious harms. Two studies (170 participants) presented data on caregiver burden and both found no evidence of effect of HSPC (very low-quality evidence). We included 13 economic studies (2103 participants). Overall, the evidence on cost-effectiveness of HSPC compared to usual care was inconsistent among the four full economic studies. Other studies that used only partial economic analysis and those that presented more limited resource use and cost information also had inconsistent results (very low-quality evidence). Quality of the evidence The quality of the evidence assessed using GRADE was very low to low, downgraded due to a high risk of bias, inconsistency and imprecision. AUTHORS' CONCLUSIONS: Very low- to low-quality evidence suggests that when compared to usual care, HSPC may offer small benefits for several person-centred outcomes including patient HRQoL, symptom burden and patient satisfaction with care, while also increasing the chances of patients dying in their preferred place (measured by home death). While we found no evidence that HSPC causes serious harms, the evidence was insufficient to draw strong conclusions. Although these are only small effect sizes, they may be clinically relevant at an advanced stage of disease with limited prognosis, and are person-centred outcomes important to many patients and families. More well conducted studies are needed to study populations with non-malignant diseases and mixed diagnoses, ward-based models of HSPC, 24 hours access (out-of-hours care) as part of HSPC, pain, achieving patient preferred place of care, patient satisfaction with care, caregiver outcomes (satisfaction with care, burden, depression, anxiety, grief, quality of life), and cost-effectiveness of HSPC. In addition, research is needed to provide validated person-centred outcomes to be used across studies and populations.