Being one of the 50 million Americans without medical care insurance has long been associated with an increased risk of death in the ICU, but as Sarah Lyon, Jeremy Kahn et al point out, most of the studies were single-center and insufficiently controlled for patient factors.
They give the field a shot in the arm in this intriguing article. Using a database of all hospital discharges in Pennsylvania in 2005 and 2006 (471,112 at 169 hospitals), the authors analyzed the effect of insurance status on events and outcomes in critical illness. The outcome was 30-day mortality, the exposure, insurance status (private: 69%, Medicaid: 27%, or uninsured: 4%). Only patients under 65 were included (no Medicare).
The uninsured patients’ absolute 30-day mortality was 5.7%; those with private insurance was 4.6%; those with Medicaid was 6.4%.
But (you may say) being uninsured is highly associated with lower socioeconomic status, and low SES patients have worse health at baseline, worse self-care, lower access to preventive care, smoke more, are more often obese and diabetic, generally live near lower-quality hospitals, and are less healthy in a galaxy of ways, for a universe of reasons — the known risk factors representing just a few stars and planets, if you will. Is an absolute 1% difference in 30-day mortality that surprising?
More rigorously than others previously, Lyon, Kahn and the epi geniuses controlled for a multitude of patient- and hospital-level factors obtainable from the databases and elsewhere. They modeled patients’ expected mortality using validated epidemiologic tools. I’m not qualified to critique the methods, but after adjustment for all these factors, the odds ratios for death remained 1.25 for the uninsured. Impressive, although I’m always skeptical of the power of even the most sophisticated data crunching to control so finely for all these incredibly complex influences of behavior, social class, living conditions, etc. (which, in aggregate, are sometimes called “life.”)
But I suspect the authors anticipated this quibble. So they made their secondary outcomes life-sustaining procedures that anyone, rich or poor, should have received roughly equally: central lines, hemodialysis, and tracheostomies. Strikingly, the uninsured had stat.significant odds ratios of 0.84 for getting a central line, 0.59 for dialysis, and 0.43 for a tracheostomy, after multivariate analysis.
A 2010 ATS systematic review noted that the uninsured have an odds ratio of 0.56 for receiving critical care services, and get 8.5% fewer procedures, but couldn’t parse out the contribution to worse outcomes by the unmeasured confounders alluded to above.
Lyon SM et al. The Effect of Insurance Status on Mortality and Procedural Use in Critically Ill Patients. Am J Respir Crit Care Med 2011;184:809-815.