Academic Articles Public Policy Supply of Medical Services

The Nursing Home as a public good

Medicaid currently accounts for roughly 50% of all nursing home expenditures and 70% of all bed days.  The government mandates that nursing homes provide a uniform level of quality to all residents, regardless of the payer type.  Yet one may ask: does this mandate hold in reality?  Nursing homes may have an incentive to segregate private insurance patients and Medicaid patients.  If private insurance patients demand a higher quality of care and their insurance pays more to the nursing homes, one would expect them to receive a higher quality of care than Medicaid patients.  For instance, Gottesman (Am J Public Health 1974) finds that as the number of public-pay residents increases, the frequency of care by staff members diminished.  Troyer (Medical Care 2004) found large cross-facility differences in mortality for Medicaid and private-paying residents in Florida, but this finding was not robust to including facility fixed effects. 

David Grabowski, Jon Gruber, and Joseph Angelelli’s 2006 NBER working paper (“Nursing home quality as a public good“) aims to explain these findings by investigating whether or not the nursing home is a public good.  Using a CMS nursing home quality repository of data–which was mandated by the OBRA 1987–the authors were able to access twelve process and outcome variables based on nursing home quality over time.  The data allow the authors to run the following regressions:

  1. General OLS: This specification uses two types of dependent variables.  The first are health outcome variables (eg: urinary tract infections, depression, presence of a fall) and the second group are outcome variables attributed to poor quality of care (eg: presence of a physical restraint, use of an indwelling catheter, bedfast, use of a feeding tube).  These outcome variables are regressed on dummies for the types of insurance, a vector of patient characteristics, a vector of nursing home specific characteristics, and time dummies.
  2. OLS + nursing home fixed effects: This is the same specification as (1) but includes a vector of nursing home dummies.
  3. OLS + patient fixed effects: From specification (2), a vector of patient dummy variables replaces the nursing home fixed effects.

One worry is that negative health shocks may increase a persons medical costs and thus ‘force’ the person onto Medicaid.  If this was true, one would erroneously conclude that Medicaid patients were worse off due to their insurance type, when in reality we have a case of reverse causality.  Grabowski, et al. look into this possibility and find that once patient fixed effects are included in the regression there does not seem to be any selection into Medicaid based on observable health attributes.  The authors also look at cross-state differences in Medicaid nursing home payment rates.  If Medicaid was much stingier in one state than another and nursing homes were not a public good, we would expect to see these states have the largest difference in quality between Medicaid and private insurance patients.  The conclusion reached from these tests finds that the nursing home is a public good; there was no finding of differential quality between Medicaid and non-Medicaid patients within the same facility. 

Grabowski, David; Gruber, Jon; Angelelli, Joseph; (2006) “Nursing home quality as a public good” NBER Working Paper #12361.