This article investigated the utilization of palliative care utilization (PCU) in COPD, heart failure and malignancy. Several factors were associated with PCU. I have several suggestions in the current manuscript.
Major:
1. The diagnosis might be overlapping. As we know, more than 30% COPD patients have cardiovascular co-morbidites. Did every patient receive pulmonary function test, echocardiography and cancer survey to confirm they have only COPD/malignancy/heart failure? How about the impact of other underlying diseases, such as myocardial infarction, stroke, diabetes, chronic renal or liver diseases, etc?
2. PCU is not an objective parameter, it might be affected by the physician in charge, patient's personal will, and availability of PCU resources. Since a large proportion of patinets were admitted to hospital in the previous year, why didn't they receive any end-of-life consultation or PCU during the hospitalization?
3. The statistic method is confusing. For example, if a patient had a PCU prior to specialist visit, it is not appropriate using logistic regression analysis for these 2 variables. We can only say the associations between these variables instead of causal relationships.
Other:
1. How many patients were screened and refuse to join the study? It might influence the characteristics of enrolled patients. Still, small case numbers and single center nature will biased the interpretation of this study.
2. Total specialist visit times in previous year, optimal and details disease treatment both pharmacological and non-pharmacologic should also be considered.
3. PCU is mainly self reported by patients. Although authors validated it by medical records, recall bias is still a concern.
4. Authors could discuss more about why 43% COPD patients (strikingly high percentage) would be intubated even in irreversible disease? Also in malignant and HF patients?

The authors made efforts for this study during the revision process. I still have concerns of this investigation.
Major:
1. Since not every patient received tests to exclude other potential diagnosis, the categorization of COPD/HF/malignancy is still not accurate. In real world, COPD is usually with comorbid HF, and vice versa.
2. Cuase of low PC utilization was not explored yet. We didn't know what we should do to improve PCU. It should be cautious about the term of "predictor" as these variables are "associated" instead of causual relationship.
3. Also in conclusion section, we also didn't know whether increased PCU will definitely improve patients' outcomes from this study.