Clalit Health Services (CHS) is the largest health care payer/provider in Israel, with about 4,217,000 insured citizens that provides care to all ages including > 60% of adults older than 65 years of age in Israel. The system is characterized by extremely low annual member turnover of < 1% . Since 1998, with increasing comprehensiveness, CHS’s information is kept in a central computerized data warehouse that includes integrated demographic data, clinical diagnoses (based on hospital discharge diagnoses, primary care physician diagnoses, and specialist outpatient clinic diagnoses), laboratory data results, medical procedures, and medications (including date of prescription and quantity and time of medication dispensed). Death records, including date of death from the Israel Central Bureau of Statistics, were linked to the Clalit population using the unique identification number for all Israeli residents. The need for consent was waived by the Helsinki Ethics Committee of the CHS (no. 037/2015).
This is a retrospective cohort study of newly diagnosed gout between 1/1/2006–31/12/2009 and followed for a 5-year period. For example, patients identified on 1/1/2006 were followed for 5 years through 31/12/2010 and patients identified on 31/12/2009 were followed through 30/12/2014. Follow-up data were included for the partial year the patient left the health plan or died.
Included were patients with continuous enrollment in Clalit for 1 year prior to date of diagnosis (index date). Patients had to be at least 25 years old as of index date. Adults 18–24 years were excluded because the majority was serving in the Israeli military where they receive full healthcare coverage. The following criteria developed in other electronic health record (EHR) studies [33,34,35] to identify incident cases of gout were used (Fig. 1):
International Classification of Diseases 9th version (ICD-9) codes 274 diagnosis from at least one rheumatologist visit;
ICD-9 274 diagnosis or free-text diagnosis of ‘gout’ from at least two community diagnoses at least 30 days apart between and either
the purchase of at least two gout-related prescription medications (allopurinol, probenecid, colchicine, or sulfinpyrazone) at least 30 days apart with the first within 6 months prior to or any time after the first community diagnosis or
two sUA test results > 6 mg/dL with the first within 6 months prior to or any time after the first community diagnosis at least 30 days apart;
ICD-9 274 diagnosis from at least one hospital admission diagnosis;
Clalit Health Services internal chronic diagnosis registry, based on ICD-9 diagnostic codes, diagnostic free text, procedures and test results ; and
Clalit Health Services physician-determined diagnosis given a ‘permanent’ status in the patient’s medical record, based on ICD-9 diagnostic codes.
Subject with at least one of these were considered to have gout. The earliest diagnosis was considered the index date, and patients were required to have 12 months without any indication of gout (baseline period) to be considered newly diagnosed (incident cases). Gout patients who had documentation of at least one of the above criteria prior to the start of study eligibility were excluded. Patients whose free-text diagnosis also included pseudo, suspected, family history, or nephrolithiasis were also excluded. Finally, 98 subjects with the following diseases also known to affect sUA levels were excluded: Familial Mediterranean Fever (ICD-9 277.31) (n = 56), glycogen storage disease (ICD-9 271.0) (n = 42), Lesch-Nyhan syndrome (ICD-9 277.2) (n = 0), juvenile gout (ICD-9 277.2) (n = 0), tumor lysis syndrome (ICD-9 277.88) (n = 0), or lead toxicity associated with gout (ICD-9 984.9) (n = 1).
Gout patients were categorized by the presence of CKD at index date as defined by an indication of moderate/severe chronic kidney disease (based on laboratory values and using the CKD-EPI creatinine equation to identify CKD ≥ 3 , or a diagnosis or claim for chronic renal failure, kidney transplantation, or dialysis.
Demographic variables such as age, sex were collected at index date. Age was assessed continuously and by groups < 55 and 55+ years as quality of gout healthcare management has been shown to decrease with older ages . Socio-economic status (SES) (low, medium, high, or missing) is an area-level score calculated based on current or last place of residence thus it likely reflects the patient’s SES at the end of follow-up. Misclassification of SES as a result of this definition is considered minimal as SES is considered to be stable during the non-critical 5-year period of adulthood in comparison to potential shifts during one’s early life course . In addition, the use of the latter SES indicator, as an adjustment for the confounding effect on resource utilization, is perhaps a better indicator of the cumulative influence of SES . Individual-level SES data are not collected by any health plan in Israel due to Israeli law, therefore SES scores derived by the Israel Central Bureau of Statistics and based on small statistical areas were used [40, 41].
Comorbid conditions at or prior to index date include cancer (ICD-9 140–208), cardiovascular disease (CVD) (ICD-9 410, 411, 413, 414, 429–434, 436, 438, V45.81/2, and coronary artery stent insertion and aortic bypass surgery procedures), diabetes (ICD-9 250), and hypertension (ICD-9 401–405) were identified using CHS algorithms [36, 42]. The Charlson Comorbidity Index (CCI) , was used to represent a weighted sum of multiple comorbid conditions predictive of higher resource utilization. Greater scores indicated a greater comorbid burden on the patient.
Clinical characteristics included smoking habits (current smoker, former smoker, and never smoker) and body mass index (BMI) (continuous and categorical coded according to the World Health Organization as: underweight [< 18.5 kg/m2], normal weight [18.5 to < 25.0 kg/m2], overweight [25.0 to < 30.0 kg/m2], obese [≥ 30.0 kg/m2], or missing).
Healthcare resource utilization for the five follow-up years following index date was calculated as the mean of the total number per year of general practitioner visits, specialist visits (e.g., rheumatologist or orthopedist), hospital admissions, use of imaging services (x-ray, MRI, ultrasound, and CT), or allopurinol (ATC M04AA01) purchase similar to others’ methods [3, 19, 44]. Mean total number of tests and test values for sUA levels (last test value prior to index date) ≤ 6 or > 6 mg/dL were reported. Survival was examined using date of death.
Age-adjusted incidence was calculated using the 2009 Clalit population distribution and direct standardization according to the Israeli population in 2009 (Central Bureau of Satistics, 2010) was used to calculate age-standardized incidence of gout . Standardized rates and their 95% confidence intervals (CI) were used to assess age-standardized rate ratios by sex. Descriptive analyses were performed to characterize the patient population’s demographic, medical history, and clinical characteristics of patients at index date.
Generalized estimating equations for repeated measures were used to assess change in healthcare utilization over the 5-year follow-up period for gout patients with and without pre-existing CKD at index date and stratified by age groups < 55 and 55+ years. Model distributions differed depending on outcome variable (general practitioners visits, allopurinol purchase = normal; specialist visits, hospitalizations, and sUA testing = negative binomial; imaging [data were restructured to binomial data, i.e., yes/no annual testing, to account for correlated data due to multiple testing for a single event] = binomial) with a first order autoregressive (AR (1)) correlation structure. The AR (1) order is used since the model is fitting longitudinal repeated measures of correlated data and similar estimates were observed when using the unstructured correlation structure. Annual unit change and 95% CI were presented and when appropriate, data were transformed from the logarithmic scale. Data prior to index date was indexed as year 0 and all subsequent years as years 1 through 5. All models were adjusted for age, sex, smoking status (current vs non-current), SES (low vs other), and CCI. During the 5-year follow-up 1673 (20.2%) healthy gout patients developed CKD. Change in patient kidney disease status from index date was not adjusted for in models since the goal of the analysis was to examine utilization based on characteristics at index date.
Time-to-death was examined using Kaplan Meier survival curves and the log rank test were used to test equality of survival distributions between the subgroups with and without CKD and stratified by ages < 55 and 55+ years. Patients were right-censored according to the month they left the health plan. Cox proportional hazard models were used to assess the risk of death in patients with and without CKD at index date and stratified by age group < 55 and 55+ years accounting for age, sex, SES, CCI, smoking status, BMI, sUA control, and gout medication use. Proportional hazard assumptions were used examining the effect of age within each age group. Hazard ratios (HR) and 95% CI were reported.
Analyses were conducted using SPSS version 23.