OBJECTIVES

Patients can freely choose health care providers without referral under Taiwan’s National Health Insurance program (NHI). This study aimed at developing an algorithm for identifying patients with diabetes whose conditions are treatable in primary care settings but bypass to go directly to tertiary hospitals.

METHODS

Claim data from one million of randomly sampled beneficiaries enrolled in Taiwan’s NHI in 2010 was used to develop an algorithm of classifying ambulatory care visits made by patients with diabetes into various primary care groups. Patients had similar probability of visiting office-based physicians, similar comorbidities, and same level of previous health care utilization pattern such as hospitalization, the numbers of emergency department visits, the numbers of drugs prescribed and continuity of care were classified into same group using Classification and regression tree (CART) tool in SAS enterprise miner 13.1. Groups with probability of being treated in primary care settings higher than 50% were classified as primary care treatable groups.

RESULTS

There were 148,751 visits made by patients with simple diabetes. About 44% of these visits were treated in tertiary hospitals. After classification tree analysis, patients with simple diabetes can be classified into 22 groups with various probability of getting care from primary care settings. The most predictive factors to identify visits treatable in primary care setting were the absence of previous hospitalization, comorbidities, and the number of emergency visits in 6 months. The classification error rate was 43%. The proportion of diabetic patients treatable in primary care setting who were instead treated in tertiary hospitals ranged from 35% to 56% among hospitals.

CONCLUSIONS

Near half of patients with simple diabetes who were treated in tertiary hospitals can be treated in primary care settings. A health policy to redirect these patients into primary care settings is warranted to improve the efficiency of health care system.