The National Board of Health and Welfare (NBHW)

NBHW is a government agency responsible for several national registries. We used data from two of these registries: the National Patient Registry (NPR) and the National Prescription Registry (NPrR). The registries under the NBHW were both accessed January 8, 2018.

Reporting to the NPR has been mandatory since 2001. The NPR contains information about all contact with specialist health care with diagnoses coded according to the 10th revision of the International Classification of Diseases (ICD-10). We used the ICD-10 codes to classify comorbidities according to the Elixhauser comorbidity index [24], with the exclusion of conditions associated with LGG, as defined in Table ​Table11.

Definition of variables

The underreporting to the NPR has been estimated to be 1% [25]. We received data on inpatient and outpatient visits, including diagnostic and procedural codes in the period January 1, 2003 - December 31, 2016.

The NPrP covers all prescribed drugs delivered by a pharmacy. Registration is mandatory. We received information on type of drug according to the Anatomical Therapeutic Chemical (ATC) classification system and date of dispensing between July 1, 2005 until December 31, 2017. We retrieved information on all anti-epileptic (N03A), sedative (anxiolytics, hypnotics and sedatives: N05B and N05C) and antidepressant drugs (N06A). We excluded drugs and diagnoses commonly used for other conditions, when creating variables to the regression models. Use was defined as having purchased at least one of the specified drugs within the last 90 days for AEDs and antidepressants, and 30 days for sedatives (prescriptions of sedatives are usually for shorter periods). Definitions of variables are presented in Table ​Table11.

Since the NPrR was established July 1, 2005, all patients with index dates before October 1, 2006 were excluded (n = 58). Patients that had moved abroad (n = 4), were not able to match with controls and thus excluded. Patients in the remaining group (n = 485) were available for analyses at index date. All patients alive at 1 year postoperative were analysed in the regression models (n = 435). Patient selection is presented in Supplement Fig. 1.