reports

Analysis of practice variation due to payment methods across health insurance schemes

Khon Kaen, Thailand: Khon Kaen University.

There are three public insurance schemes, the non-contributory Civil Servant Medical Benefit  Scheme (CS) for government employees, pensioners and their dependents (9% of total  population); the Social Security Scheme (SS) for formal private sector employees (12% of total  population) financed by tripartite contributions; and the Universal Health Coverage (UC) scheme  for the rest (74%) of the population financed by non-contributory general tax revenue. Three  schemes are different in the provider payment methods, whereby CS uses an open-end,  retrospective, fee-for-service (FFS) system, and UC and SS relies on a close-end, prospective,  capitation payment. This study aimed to assess variation in health care provided to the  beneficiaries of public insurance schemes in Thailand that cover 95% of population, and to  examine mortality rates after hospitalization across years of hospitalization in the UC schemes.

The analyses relied on: (1) outpatient prescribing databases (2003-2007) from six provincial  hospitals; (2) nation-wide inpatient databases (2004-2007); and (3) outpatient visit, laboratory,  and drug databases (2003-2007) from 13 district hospitals in 2 neighboring provinces.  Comparison of hospitalization outcomes across insurance schemes or across years of  hospitalization in the UC schemes was controlled for underlying differences in patient  demographics, disease severity, and hospital types.

This analysis reiterates a close link between propensity to receive six groups of innovative drugs  (angiotensin-2 receptor blockers, single-source statins and new antilipidemics, clopidogrel, singlesource  proton pump inhibitors, coxibs, and inhaled corticosteroids), and two expensive  interventions (laparoscopic cholecystectomy-LC and Cesarean section), and health insurance  payment methods. Between 20% and 29% of the CS patients were prescribed four out of six  selected innovative drug groups whereas only 1-2% of the UC and SS patients were so.  Approximately half of the CS patients with cholecystectomy underwent the laparoscopic method,  whereas only 20-30% of the UC and SS patients did. Cesarean section rate in CS increased from  45% in the first quarter of 2004 to 59% in the third quarter of 2007, whereas that in UC and SS  were relatively stable at 16-21% over the same period.

An intensive use of health resources in CS is evident by longer LOS of 1-4 days for in pneumonia,  congestive heart failure (CHF), and acute myocardial infarction (AMI), and 3-5 days for stroke  subtypes in almost all types of hospitals, as compared with UC. The SS-UC gap in LOS is much  less in both magnitude (1-3 days) and in the number of hospital types found statistically  significant difference. A longer stay for two major cardiac interventions (coronary artery bypass  graft -CABG and percutaneous coronary intervention -PCI) in university hospitals by UC as  compared with CS and SS suggests that the UC patients may have other poor prognostic  conditions (such as low socio-economic status) remained uncaptured by the multivariable analysis.

Hospital readmissions within 28 days after LC, appendectomy, hip replacement, and CABG are  less than 1% and similar across insurance schemes. The predicted in-hospital mortality rates  from CABG and PCI for UC patients, is slightly higher in an absolute term (1-4 percentage points)  than that for CS patients. Interpretation of in-hospital death is limited. Notably, the magnitude  of 30 day mortality for the UC patients hospitalized with the studied diseases was quite large as  compared to that of the in-hospital mortality.

Across hospitalization years (2005-2007), the UC patients admitted to general, regional, and  university hospitals with pneumonia, CHF, and ischemic stroke had a comparable risk of death  within 30 days and similar survivals over a three-year follow up. Despite increasing burdens to  health care providers due to cost pressure from capitation payment and service load from better  access of the UC scheme, except for hemorrhagic stroke there is no obvious sign of worsening  patient outcomes as indicated by the 30-day mortality. For those admitted in 2006 and 2007, a  better overall survival of AMI was found in regional and university hospitals, whereas a higher  mortality of hemorrhagic stroke was found in general hospitals. The noticeably high mortality of  those admitted with CHF, AMI, and strokes over a three-year follow up raises a particular concern  to the continuity of care for chronic non-communicable diseases among the UC scheme.

For UC patients, survivals from breast, cervical, and colon cancers are higher than that from lung,  liver, and pancreatic cancers. The survivals in UC are lower than in US SEER for the former group  of cancers, but indifferent for the latter deadliest cancers. Worsening survivals in liver and  pancreatic cancers during the recent years reflect a delayed proper investigation or intervention  probably due to the UC service crowding.

Scheme variation in the quality of care for diabetic patients during ambulatory visits to district  hospitals in 2 neighboring provinces showed no clear pattern in both process and outcome  indicators. Disease management program implemented in one province, however, helped  improve adherence to the indicators over time. After three years of follow up among those  whose kidney function already deteriorated to CKD stage 3, further progression to stage 4  occurred more frequently about 5 percentage points in the UC (35%) than that of the CS  counterparts (30%).

Based on the ACG casemix analysis, the insurance gap in health status for hypertension and  diabetes reveals the same patterns that the CS cohorts started with the more severe conditions  than the UC and SS. This can be the influence of better access to care, hence investigation and  utilization biases were stronger for the CS patients since the beginning of the universal coverage reform.