For persons living with serious illness, palliative care can relieve symptoms and stress. This type of care benefits both the patient and their family and is delivered by a multi-disciplinary team comprised of clinicians, social workers, and clergy. These team members work in tandem with other doctors to provide an extra layer of care.
While palliative care plays a pivotal role for patients, data demonstrating its effectiveness and cost-effectiveness can be muddled by confounding and selection bias.
For example, people with severe illness are more likely to receive palliative care. Illness severity is also correlated with poor outcomes. But problems can arise if an analysis doesn’t account for the ways in which illness severity is associated with both likelihood of palliative care and with poor outcomes. This might lead to an incorrect inference that palliative care is harmful
To better understand these relationships, a team of investigators from Trinity College (Dublin), Boston University, and the University of Southern California conducted a systematic review of ‘natural experiments’ in palliative care.
Natural experiments, also called quasi-experiments, are useful in contexts like palliative care where it would be unethical or impractical to conduct clinical trials. In natural experiments, investigators apply statistical techniques to previously collected data to make treatment and comparison groups look as similar as possible, other than receipt of treatment. To review the extent to which natural experiments exist in the palliative care literature, investigators considered four research designs.
- Interrupted time series analysis: repeated measures of a given outcome in a population over time.
- Difference-in-differences: controlled before-and-after studies that examines outcomes at two or more time points for two groups.
- Regression discontinuity: compare outcomes for two groups on either side of an arbitrary threshold.
- Instrumental variables: explain variation in the treatment variable but have no independent effect on the outcome of interest.
With these study designs in mind, investigators were guided by four research questions:
- How many studies have evaluated palliative care using causal frameworks and what specific methods were used?
- What results have been reported from these studies?
- What are these studies’ strengths and weaknesses?
- What are the implications for expanding use of this methodological framework given research challenges in palliative care?
To answer these questions, investigators identified peer-reviewed studies from seven databases, two hand-searched journals, and grey literature. The search returned nearly 750 articles which were narrowed down to 17 relevant studies that met inclusion criteria. These studies were available in English and included study designs investigating causal inference with observational data.
To ensure thoroughness and quality, multiple reviewers reached consensus for study inclusion during data extraction and screening. Individual studies were assessed for methodological quality using both a custom-built tool and the STROBE checklist.
Results
Most studies (seven of 17) used a difference-in-differences study design; five employed instrumental variables and another five employed interrupted time series analysis. Studies spanned from 2002 to 2021, and the median year of publication was 2018.
Nine studies examined intensity of care (e.g., acute hospitalizations length of stay, and intensive care unit admissions). Others examined palliative care’s effects on costs.
Overall, most studies showed that palliative care reduced health care costs, lowered health care utilization, and resulted in less aggressive care at end of life (consistent with prior reviews on the subject).
There were some differences from prior systematic reviews, though. For example, the magnitude of cost-savings was lower than previously reported in less rigorous research designs and null results were common (i.e., palliative care had no effect on health care utilization or costs).
Investigators noted a few limitations related to the individual studies they found, as well as the systematic review they undertook. Namely, reporting standards varied across all three experimental frameworks (difference-in-difference, interrupted time series, instrumental variables) used by individual studies. Regarding this systematic review, investigators noted that they included a wide variety of treatments and policy changes in their criteria, as well as a broad population. Therefore, investigators assert that only general conclusions can be made without deriving meta-analysis effect estimates.
Takeaway
Palliative care is a critical component of health care for patients with serious illnesses and their families, yet research about this care can be blurred by confounding and selection bias. This systematic review demonstrates that natural experiments are one important tool to strengthen the evidence base for palliative care being effective for patients and cost-effective for the wider system.
