Decreased risk of acute myocardial infarction in stroke patients receiving acupuncture treatment: a nationwide matched retrospective cohort study

Taiwan’s National Health Insurance Program, which was implemented in March 1995, covers more than 99 % of the 22.6 million residents of Taiwan. As described previously [11‐15], all beneficiaries’ inpatient and outpatient medical services such as simple demographics (gender, birth date, low-income status, and urbanization of living area), primary and secondary diagnoses, procedures, prescriptions, and medical expenditures are recorded in the National Health Insurance Research Database established by the National Health Research Institutes. The research articles based on Taiwan’s National Health Insurance Research Database have been accepted in distinguished scientific journals worldwide [11‐15].

The insurance reimbursement claims used in this study were from Taiwan’s National Health Insurance Research Database. To ensure privacy, the patients were de-identified in the electronic database. This study was evaluated and approved by the Institutional Review Board of Taiwan’s National Health Research Institutes (NHIRD-100-122) and E-DA Hospital, Kaohsiung, Taiwan (2014012). Informed consent was not required because the patients’ identities were decoded and scrambled. This study was conducted in accordance with the Declaration of Helsinki [11‐15].

We identified 182,619 new stroke survivors aged 40–79 years and older who were hospitalized (people who received inpatient care with a primary diagnosis of stroke) between January 1, 2000 and December 31, 2004 (admission date); 23475 of these patients had received at least one course (one course includes six consecutive treatments) of acupuncture after their stroke hospitalizations. To confirm that all patients with stroke in our study were incident cases, only new-onset stroke cases were included; patients with a previous medical history of stroke within five years of the index date were excluded. The diagnosis of stroke was validated as described in previous studies [11‐15]. Stroke patients who received only one course of acupuncture treatment were excluded from this study. The stroke patients who did not receive acupuncture treatment were selected via matching with propensity scores (exposure vs. non-exposure ratio = 1:2). Each hospitalized stroke patient was either followed from the index date (for the non-acupuncture group: discharge date of stroke hospitalization; for the acupuncture group: the date of the first acupuncture treatment after stroke hospitalization) until 31 December 2009 or censored. For the acupuncture group, the time period between the discharge date of the stroke hospitalization and the date of the first acupuncture treatment after the stroke hospitalization was referred to as the immortal time. Therefore, immortal time bias was avoided in this study [16]. The follow-up time in person-years was calculated for each stroke patient until the diagnosis of AMI (stroke patients who visited the emergency department with a primary diagnosis) or until the patient was censored due to death or withdrawal from the insurance system or was lost to follow-up. The non-acupuncture group included patients with stroke who did not receive acupuncture treatment before the end of the follow-up period. We compared the risk of AMI between the stroke patients who did and did not receive acupuncture treatment during the follow-up period.

In this study, low-income patients were defined as patients who qualified for waived medical copayment as verified by the Bureau of National Health Insurance. We defined a newly diagnosed stroke according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM 430-437). The primary outcome of AMI was defined via ICD-9-CM 410. Coexisting medical conditions were determined from medical claims for the follow-up period and included hypertension (ICD-9-CM 401-405), diabetes (ICD-9-CM 250), mental disorders (ICD-9-CM 290-319), traumatic brain injury (ICD-9-CM 800-804, 850-854), hyperlipidemia (ICD-9-CM 272.0-272.4), cardiac arrhythmia (ICD-9-CM 427.9), Parkinson’s disease (ICD-9-CM 332), depression (ICD-9-CM 296.2, 296.3), Alzheimer’s disease (ICD-9-CM 331.0), obesity (ICD-9-CM 278.0, 278.1), and malignant brain tumors (ICD-9-CM 191). Subtypes of stroke, lengths of hospital stay, admission to the intensive care unit or undergoing neurosurgery during the index stroke hospitalization were also identified as potential confounding factors in this retrospective cohort study. Cardiovascular medications such as anticoagulants, anti-platelet agents, and lipid-lowering agents were considered in this study.

To reduce confounding effects, we developed a non-parsimonious multivariable logistic regression model to estimate a propensity score for acupuncture treatment. Clinical significance guided the initial selections of the covariates, which included age, sex, low-income status, subtypes of stroke, hypertension, diabetes, mental disorders, traumatic brain injury, hyperlipidemia, Parkinson’s disease, Alzheimer’s disease, malignant brain tumors, stays in the intensive care unit, neurosurgery, and lengths of hospital stay. We used a structured iterative approach to refine this logistic regression model to achieve a balance of covariates within the matched pairs. We then matched (without replacement) patients who received acupuncture treatment with those who did not using a greedy matching algorithm with a caliper width of 0.2 SD of the log odds of the propensity score. The nearest-neighbor algorithm was applied to construct matched pairs, assuming that the proportion of 0.95 to 1.0 was perfect. This method removed 98 % of the bias from the measured covariates [17, 18].