June 16th: Applied Methods Discussion Group

Hello everyone,

All students, faculty, research staff and post-doctoral fellows are invited to participate in the new EBOH Applied Methods Discussion Group.

The objective of this group is to give us an opportunity to address and discuss real methodological challenges in our area of practice. The goal of each session will be to tackle one methodological challenge (or a small number of related challenges) that we encountered in our research, all the while putting a strong emphasis on the application and the implementation of related methods to overcome this problem. The group is intended to be student-run and student-led for the most part.

Our next meeting is to take place onJune 16th from 12-1pm in Purvis Room 24where Bruno Riverin will share a current challenge in developing a study design to assess outpatient follow-up with a physician after discharge from hospital (see below and attached for more details).

Refreshments will be served! 

If you have any questions or have an interest in sharing a methods related issue at a later date, please feel free to contact Daniala Weir (daniala.weir [at] mail.mcgill.ca), Helen Cerigo (helen.cerigo [at] mail.mcgill.ca), Bruno Riverin (Bruno.riverin [at] mail.mcgill.ca) or Sahir Bhatnagar (sahir.bhatnagar [at] mail.mcgill.ca).

Thanks,

Daniala, Helen, Bruno and Sahir

Context

Timely outpatient follow-up with a physician after hospital discharge has been promoted as a strategy to reduce readmissions. The quality of evidence supporting this strategy is weak and inconsistent, mainly due to methodological challenges.

Study Question

We would like to know if the timing of physician follow-up plays a role in reducing the risk of hospital readmissions within 30 days. For instance, is the reduction in risk of readmission greater if follow-up is provided earlier; or is there a critical time after discharge at which the follow-up no longer has an effect?

Study Setting

This study uses health care utilization data on nearly all elderly or chronically ill patients hospitalized during a 5-year period. The unit of analysis is the index admission, and outpatient as well as inpatient billing data is collected over the 30 days following hospital discharge.

Target Trial Protocol

Eligibility criteria:                            Elderly or chronically ill patient hospitalized for any cause

Treatment strategies:                      Receive outpatient physician follow-up on any one day following hospital discharge, up to 30 days, or no follow-up at all

Assignment procedures:                 Participants will be randomly assigned at hospital discharge (t = 0) to one of the 30 days, or to no follow-up

Follow-up period:                             Starts at randomization (at discharge); ends at readmission, death, or 30 days after hospital discharge (no loss to follow-up)

Outcome:                                            Hospital readmission within 30 days of discharge

Causal contrast:                                Per-protocol?

Analysis plan:                                    Per-protocol effect adjusted for pre-baseline confounders; IPTW derived from time-specific PS; time-dependent covariates; time-dependent effect of exposure in flexible parametric survival model with competing risk.

Design issues

Follow-up begins after discharge for everyone. However, patients readmitted before their pseudo-assigned day have in fact incomplete follow-up; and, by design, are counted as unexposed, which may increase the risk of readmission among the unexposed. So, either those patients should have been assigned to earlier treatment, or they are doomed* (readmission cannot be prevented)); those who die before, however, are handled in a competing risk framework.