Author Affiliations: Division of Geriatric Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia (Dr Casarett); and Health Equity Research and Promotion, Department of Veterans Affairs Medical Center, Philadelphia (Dr Richardson and Mss Smith and Johnson).
Drs Bruera and Hui make a valid point regarding the challenges of measuring the impact of palliative care interventions in observational studies. There is always the concern that assignment to receive an intervention (in this case, either palliative care consultation or care in a palliative care unit) is not random. If patients with greater needs or more complex problems are more likely to receive the intervention, the result may be an underestimate of the intervention's effectiveness. Although to some extent this risk can be mitigated by propensity score adjustment, it is still possible that other, unmeasured characteristics differ among the treatment groups. Nevertheless, in the absence of randomized controlled trials, which would be difficult or impossible in the setting of an evaluation of palliative care units, observational studies with appropriate adjustment for nonrandom selection offer an efficient alternative design. Further research is needed to determine how these methods can be made more robust, ensuring that their results are sufficient to guide clinical care and policy.
Casarett D, Richardson D, Smith D, Johnson M. Palliative Care Units: The Best Option for the Most Distressed—Reply. Arch Intern Med. 2011;171(17):1601-1602. doi:10.1001/archinternmed.2011.416