Detecting Incident Delirium within Routinely Collected Inpatient Rehabilitation Data: Validation of a Chart-Based Method

Round 1

Reviewer 1 Report

This is a study on the retrospective detection of a possible delirium in medical records, an important topic for research due to the low diagnostic rate observed in routine clinical practice.

 

I have some comments:

 

I found very interesting the way you adapted the initial list of words, based on particular social and clinical characteristics of your specific setting.

 

I think that the Introduction part is not well focused, the 3rd paragraph describes the difficulties in the use of screening tools, specifically one of them, but the article is about the retrospective review of the medical chart for delirium episodes, an aspect that, on the contrary, is little commented on.

 

In Methods numeration, I find subsections 2.4 and 2.5 belong to the 2.3 subsection, so they should be renumbered as 2.3.1 and 2.3.2, respectively.

 

I’m not sure what is the reason why you discarded patients with prevalent delirium. Would it not have been better to have more delirium episodes for the purpose of the study?

 

What kind of professionals did the initial review? Do they have any specific previous training?

 

I think the big problem of this study is the “gold standard” against you compare the chart review method. You use, as you state in limitations, a very subjective measure, which is another chart review, although in this occasion, made for experts. This choice of a “gold standard” requires further justification.

 

Also, if you maintain it, it is notable that there was a high internal disagreement rate between the experts (more than 25% the first time between the junior physician and the senior neurologist and 67% in the second time between the two senior neurologists). What is the reason for these disagreements and how did they finally get an agreement -did they use any scale or a specific protocol?

 

You state a variety of possible reasons for the finding that only 10.8% of patients with discharged diagnosis of delirium were identified by the reviewers as incident delirium. I think that the mixture of possible causes derived from clinical and administrative procedures particularities, makes this finding of little real interest.

 

Although in this case, we would like to have a tool with a high PPV, I think it should be also useful to determine and report the NPV  

 

It is not clear neither in the text, nor in table 3, the results of PPV for the subgroups you analysed: on the main text you say the PPV of 0.55 and 0.69 correspond to “subjects with ≥6 recorded key words only OR those with ≥1 administered antipsychotic drug”, but in the table it seems to say that the PPV of 0.55 corresponds to subjects with ≥6 recorded key words AND  ≥1 administered antipsychotic drug and the PPV of 0.69 to subjects with ≥6 recorded key words AND without any administered antipsychotic drug (which is also contradictory with what you say in discussion).

 

You should state the proportion of confirmed incident delirium episodes in expert review (1.2%) in the text, as you do in the discussion.

 

As you say correctly in the second paragraph of the discussion part, it is more difficult to identify delirium in patients with other conditions sharing similar symptoms. Maybe it would be useful to do a separate analysis for patients with dementia, including PPV/NPV, since you have them identified? Also, did you also consider to analyse the data for patients in each rehabilitation subgroup (angiology, cardiology, neurology, etc)? Probably this analysis could help to know how much other conditions may worsen the performance of your screening method.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

thank you for the opportunity to read the article, I found it interesting. The topic is very important. The authors designed the research well, however, I missed the introduction and discussion of well-known authors who have been studying the problem of delirium for a long time, eg Kotfis Katarzyna or Krupa Sabina, I suggest that the authors read the publications and take them into account in the introduction. It is also worth adding restrictions to work as well as therapeutic implications. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Thanks to the authors for considering all my recommendations. Congratulations on the article.

Reviewer 2 Report

recommends an article for publication