Cream skimming and hospital transfers in a mixed public-private system Cheng TC, Haisken-DeNew JP, Yong J. Soc Sci Med. 2015 May;132:156-64. doi: 10.1016/j.socscimed.2015.03.035. Epub 2015 Mar 19.
This paper investigates the phenomenon of cream skimming in Australia’s mixed public-private hospital setting, using the novel approach of analyzing hospital transfers. Cream skimming occurs when hospitals, surgical centres, and health care providers select patients with lower than expected treatment costs, thereby gaining financially. It can also occur through preferential selection of treatments with low cost relative to reimbursement rate, and by avoidance of treatments with high cost relative to reimbursement rate.
Australia has a mix of publicly-owned not-for-profit hospitals, privately-owned not-for-profit hospitals, and private/investor-owned for-profit hospital. Privately-owned hospitals are paid mostly on a fee-for-service basis, charging what the market will bear. They derive revenue predominantly from privately-insured patients (whose insurance costs are heavily subsidized by the government) or from patients paying privately out-of-pocket (with partial fixed government subsidy). In contrast, most publicly-owned hospitals in Australia are funded through global budgets, except in the Australian state of Victoria, where this study took place, which funds publicly-owned hospitals through activity-based funding (ABF).
Australia also allows physician dual-practice, in which physicians are permitted to work in both the publicly- and privately-funded sectors.
The study used data extracted from a hospital administrative database containing all hospital admission episodes in the Australia state of Victoria (the Victorian Admitted Episodes Dataset, VAED). The study population included all patients (1.77 million admission episodes) identified as having ischemic heart disease (IHD), admitted for all episodes of IHD- and non-IHD-related care between 1998–2005, and who did not die during the study period. The data were divided into two sub-samples according to whether the patient was initially admitted to one of 132 publicly-owned or 177 privately-owned facilities. For each sub-sample, three types of transfers were identified: (1) no transfer; (2) transfers between the same hospital type (i.e. private-to-private or public-to-public transfers); and (3) transfers across hospital type (i.e. private-to-public transfers and public-to-private transfers).
This was a two-part study:
Part 1 examined whether the probability of cross-type transfer was influenced by patient severity and complexity (measured using the Charlson comorbidity index), with separate multinomial logit regressions for each type of hospital admission. The study controlled for factors that could affect the probability of transfer, including patient and socioeconomic characteristics, geographic location, diagnosis-related group (DRG), and type of hospitalization (i.e. emergency or other).
Part 2 examined differences in immediate post-transfer hospital utilization (measured by length of stay and cost-weighted utilization) in matched and unmatched patients. Hospital transfers occur for a variety of legitimate reasons, including capacity issues or patient-initiated reasons. The study controlled for instances of appropriate transfers between hospitals with different specializations or available technology.
(1) Pricing and costing data were unavailable from private hospitals because the data access agreement explicitly prohibited identifying private hospitals. It was not possible, therefore, to identify cream skimming by private hospital providers using price and cost data alone. (2) The authors did not separately analyze outcomes in the sub-group of privately-owned for-profit vs. privately-owned not-for-profit hospitals, instead grouping all privately-owned hospital results together. (3) Results are specific to IHD, though generalizable to other diagnoses in which patient severity and complexity varies. (4) While the study was carefully designed to isolate and identify the effects of cream skimming behaviour, it cannot conclusively rule out other mechanisms that might influence hospital transfers. (5) Prevalence of cream skimming is likely understated in this study; anecdotal evidence suggests it is more common for patients to be turned away at privately-owned hospitals before admission, on the grounds that they could be better cared for in public hospitals. These patients did not appear in the transfer data because they were not admitted to privately-owned hospitals in the first place.
Effect of patient severity and complexity on probability of cross-type transfer:
(1) patients at all levels of illness severity and complexity are more likely to be transferred from private-to-public hospitals than from public-to-private hospitals;
(2) patients with more severe or complex medical conditions (Charlson index ≥3) are even more likely to be transferred from private-to-public hospitals than are patients with less severe or complex medical conditions (Charlson index <3); and
(3) patients with less severe or complex medical conditions (Charlson index <3) are more likely to be transferred from public-to-private hospitals than are patients with more severe and complex medical conditions (Charlson index ≥3).
Effect of hospital ownership on post-transfer hospital utilization:
Matched patients: After transfer from private-to-public hospitals, matched patients had 16.5% longer stays and incurred 7.3% higher levels of cost-weighted utilization than did patients in private-to-private transfers (p< .01). In contrast, after transfer from public-to-private hospitals, matched patients have 5.8% shorter stays and marginally lower levels of cost-weighted utilization than did patients in public-to-public transfers. This pattern is consistent with cream skimming in the form of patient selection, meaning private hospitals appear to keep low severity/complexity patients and transfer those with higher care needs to the public sector.
Unmatched patients: After transfer from private-to-public hospitals, unmatched patients stayed 21.8% longer and incurred 11.2% higher levels of cost-weighted utilization than did patients in private-to-private transfers. In contrast, after transfer from public-to-private hospitals, unmatched patients transfers had similar lengths of stay and 8.2% lower cost-weighted utilization than did patients in public-to-public transfers. This pattern is consistent with cream skimming in the form of treatment selection.
These effects were not explained by differences in efficiency between public and private hospitals.
This study provides some of the best available evidence on gaming, in the form of cream skimming of patients and of treatments. This behavior appears to be facilitated by the co-existence of publicly- and privately-owned hospitals, in combination with mixed public- and private- hospital financing, and a policy allowing physician dual practice. The hospitals in this study were remunerated through activity-based funding or fee-for-service, but the relative effect of those funding models on cream skimming was not considered, so remains uncertain.
Privately-owned/funded hospitals appeared to engage in patient selection by retaining those with low severity and complexity, and transferring to the public sector those with high care needs. Dual-practice doctors have an incentive to transfer less severe and complex patients from publicly-owned/funded hospitals to privately-owned/funded hospitals, where they are permitted to charge higher fees.
As Canada considers options to improve our health care system, some have proposed introducing private insurance and physician dual practice. Such changes might result in outcomes similar to those revealed in this study: shifting sicker more costly patients to publicly-funded hospitals, and healthier less costly patients to privately-funded hospitals. In the absence of higher levels of public funding to account for the higher levels of sicker patients, “public” hospitals could expect more strain on already limited budgets.
Read in PDF format: