Comparing the cost of cirrhosis to other common chronic diseases: A longitudinal study in a large national insurance database

Data Availability - Comparing the cost of cirrhosis to other common chronic diseases: A longitudinal study in a large national insurance database 

Data Availability 

The claims data used for this publication is stored in a secure portal, the UnitedHealth Group (UHG) Enclave, and is accessible only by those with prior authorization to view the data (Table D1). This data is only available to those who have signed Data Use Agreements (DUA) and have undergone training on accessing the data. All statistical analyses were performed using Stata 14.1. 

Table D1: Access to Enclave 

Person 

Title 

Organization 

Access to Enclave 

Federico Crippa 

Research Analyst 

Northwestern University 

Yes 

Aditya Jain 

Research Analyst 

Northwestern University 

Yes 

Filip Obradović 

Research Analyst 

Northwestern University 

Yes 

Eleena Koep 

Director of Research 

UnitedHealth Group 

Yes 

Study Methods 

A retrospective, longitudinal cohort study was conducted between January 1, 2011, and December 31, 2020, by using claims data from UHG, a large national insurer in the United States. The study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. The Northwestern University Institutional Review Board deemed this study exempt from review and waived the need for patient-informed consent. The database included deidentified medical and pharmacy claims data for enrollees. Available data included claims for all medical services and prescription medications that were submitted to the insurer for payment, patient diagnoses, and, when available, linked laboratory results. For this study, all enrollees were defined and reported as patients. 

Patients enrolled in non-health maintenance organization (HMO), Medicare Advantage (MA) plans between January 1, 2011, and December 31, 2020, were included in the analysis. Because provider fees under HMO plans/capitated contracts are not available in large insurer data sets, this study focused on patients covered under non-HMO plans (ie, non-capitated claims). Previously validated and published codes were used. Patients diagnosed with cirrhosis were defined as those with at least 1 claim specifying a validated cirrhosis code from the International Classification of Diseases, 9th Revision (ICD-9) or 10th Revision (ICD-10) (571.2, 571.5, 571.6, K70.30, K70.31, K74.0, K74.60, K74.69, K74.3, K74.4, and K74.5), in any position on the claim (Supplemental Figure S1).  

Cirrhosis etiologies were defined by at least 1 claim specifying an ICD-9 or -10 code for: alcohol-associated (ETOH), metabolic dysfunction–associated steatohepatitis (MASH, previously NASH), HBV, HCV, biliary cirrhosis (eg, primary sclerosing cholangitis and primary biliary cirrhosis), cardiac cirrhosis, genetic cirrhosis (eg, hereditary hemochromatosis), and autoimmune hepatitis. Patients were included in the HCV cohort even if they had multiple etiologies, any one of which was HCV.

Other etiologies were analyzed separately unless otherwise specified. Of note, MASH/NASH did not have a dedicated ICD code before October 1, 2015. For data before October 1, 2015, MASH/NASH was identified using a previously published algorithm (patients without specific cirrhosis etiology plus obesity, dyslipidemia, diabetes, and/or hypertension). Patients were categorized into 5 mutually exclusive groups by etiology: MASH, ETOH (which includes alcohol-associated liver disease and metabolic dysfunction–associated steatotic liver disease with increased alcohol intake), HCV, biliary (primary biliary cirrhosis and primary sclerosing cholangitis), and “Other” (eg, hemochromatosis). Validated ICD-9/-10 codes were used to define comorbidities and quantify the Charlson Comorbidity Index. 

To provide a tangible comparison, costs associated with cirrhosis were compared with costs associated with HF and COPD. Patients with HF and COPD were identified by ICD-9/-10 codes as previously published. A sample of these patients was included in the cohort as detailed in Supplemental Figure S1, and the Supplemental Appendix. 

Analyses were performed at the patient-month level. Costs are reported per patient-month unless otherwise specified. For patients with cirrhosis, the index date was the earliest listed date of service on a claim with a diagnosis of cirrhosis or cirrhosis decompensation. For patients with HF and COPD, the index date was the earliest listed date of service on a claim for the appropriate corresponding diagnosis. Diagnoses on claims were defined and identified by validated inclusion ICD codes. If the index code for a patient with cirrhosis indicated decompensation, the patient was considered to be decompensated at the time of inclusion. To most closely represent the epidemiological reality, cirrhosis, HF, and COPD were not considered mutually exclusive. All months of coverage for all included patients under non-HMO MA plans, including and following the index month until the end of observation, were analyzed. The end of observation was defined as disenrollment, transplant, or the end of the cohort study period (December 31, 2020). In the case of a patient receiving a transplant, the entire month of transplantation was censored.

Months of coverage were subcategorized into compensated and decompensated cirrhosis patient-months. The date of decompensation was defined as the date of service on the first claim under any plan that listed an ICD or CPT code for a decompensation event (ie, ascites, spontaneous bacterial peritonitis, HE, variceal bleeding, hepatorenal syndrome, or hepatopulmonary syndrome) (see the Supplemental Appendix). Patients were considered to have decompensated cirrhosis starting from and including the month of the decompensation date until the end of observation. The total number of patient-months with compensated cirrhosis was calculated from the index month until the decompensation month. Inpatient costs were calculated by identifying claims between admission and discharge. The same method of categorizing months of coverage was applied to decompensated HF and exacerbated COPD. Decompensated HF and exacerbated COPD were defined as inpatient admissions with the corresponding condition as the primary diagnosis. 

Cost was defined as the total amount paid by the insurer and patient, taking the health care sector cost perspective. These data were analyzed and reported over time using mean costs in a specific calendar month during the study period. Data reported per patient-year were estimated by multiplying patient-month data by 12. Following common practice, denied service lines were excluded from cost calculation. Average costs were adjusted for inflation to 2021 US dollars using the Bureau of Labor Statistics consumer price index to adjust for inflation between 2011 and 2020. Mean costs were reported with corresponding standard deviations (±SD). Comparisons between relevant subgroups were performed using unequal variance, 2-sided Student t tests. Contained in the Supplemental Appendix is a discussion of the process of generating the patient data and the assumptions underlying the standard inference methods. To identify predictors of increased cost, a multivariable linear regression model of costs for patients with cirrhosis was performed. The outcome of interest was the cost per patient per month, which was modeled continuously. The reference group was non-female patients with ETOH. All statistical analyses were performed using Stata 14.1.

数据可用性——肝硬化与其他常见慢性疾病的成本比较：基于大型全国性保险数据库的纵向研究

数据可用性

本研究使用的索赔数据存储于安全门户UnitedHealth Group (UHG) Enclave中，仅授权人员可访问（表D1）。数据仅向签署了数据使用协议（Data Use Agreements, DUA）并完成数据访问培训的人员开放。所有统计分析均采用Stata 14.1完成。

表D1：Enclave访问权限

人员 职称 机构 Enclave访问权限
Federico Crippa Research Analyst Northwestern University Yes
Aditya Jain Research Analyst Northwestern University Yes
Filip Obradović Research Analyst Northwestern University Yes
Eleena Koep Director of Research UnitedHealth Group Yes

研究方法

本研究为回顾性纵向队列研究，时间跨度为2011年1月1日至2020年12月31日，使用美国大型全国性保险公司UHG的索赔数据。研究遵循流行病学观察性研究报告强化（Strengthening the Reporting of Observational Studies in Epidemiology, STROBE）队列研究报告指南。西北大学机构审查委员会认定本研究豁免审查，并免除患者知情同意要求。数据库包含参保人员的去标识化医疗和 pharmacy 索赔数据。可用数据包括提交给保险公司的所有医疗服务和处方药索赔、患者诊断，以及（若有）关联的实验室结果。本研究中，所有参保人员均被定义为患者。

2011年1月1日至2020年12月31日期间参加非健康维护组织（health maintenance organization, HMO）的联邦医疗保险优势（Medicare Advantage, MA）计划的患者被纳入分析。由于HMO计划/按人头付费合同下的提供者费用在大型保险公司数据集中不可得，本研究聚焦于非HMO计划覆盖的患者（即非按人头付费索赔）。研究使用经预先验证并发表的编码。肝硬化患者定义为至少有1份索赔中（任意诊断位置）包含国际疾病分类第9版（International Classification of Diseases, 9th Revision, ICD-9）或第10版（10th Revision, ICD-10）中经验证的肝硬化编码（571.2、571.5、571.6、K70.30、K70.31、K74.0、K74.60、K74.69、K74.3、K74.4和K74.5）的患者（补充图S1）。

肝硬化病因由至少1份索赔中包含以下ICD-9或ICD-10编码定义：酒精相关性（ETOH）、代谢功能障碍相关性脂肪性肝炎（metabolic dysfunction–associated steatohepatitis, MASH，前称NASH）、乙型肝炎病毒（HBV）、丙型肝炎病毒（HCV）、胆汁性肝硬化（如原发性硬化性胆管炎和原发性胆汁性肝硬化）、心源性肝硬化、遗传性肝硬化（如遗传性血色病）和自身免疫性肝炎。即使患者存在多种病因（其中任一为HCV），仍被纳入HCV队列。

除非另有说明，其他病因单独分析。值得注意的是，2015年10月1日前MASH/NASH无专用ICD编码。对于2015年10月1日前的数据，MASH/NASH通过先前发表的算法识别（无特定肝硬化病因的患者加肥胖、血脂异常、糖尿病和/或高血压）。患者按病因分为5个互斥组：MASH、ETOH（包括酒精相关性肝病和酒精摄入增加的代谢功能障碍相关性脂肪性肝病）、HCV、胆汁性（原发性胆汁性肝硬化和原发性硬化性胆管炎）和“其他”（如血色病）。使用经验证的ICD-9/ICD-10编码定义合并症并量化查尔森合并症指数（Charlson Comorbidity Index）。

为提供具体比较，将肝硬化相关成本与心力衰竭（HF）和慢性阻塞性肺疾病（COPD）相关成本进行对比。HF和COPD患者通过先前发表的ICD-9/ICD-10编码识别。这些患者的样本按补充图S1和补充附录中的详细说明纳入队列。

分析在患者-月水平进行。成本按患者-月报告，除非另有说明。对于肝硬化患者，索引日期为包含肝硬化或肝硬化失代偿诊断的索赔中最早列出的服务日期。对于HF和COPD患者，索引日期为对应诊断索赔中最早列出的服务日期。索赔中的诊断通过经验证的纳入ICD编码定义和识别。若肝硬化患者的索引编码显示失代偿，则该患者在纳入时被视为失代偿状态。为最接近流行病学实际，肝硬化、HF和COPD不被视为互斥。分析所有纳入患者在非HMO MA计划下的所有覆盖月份，包括索引月及之后直至观察结束。观察结束定义为退出参保、移植或队列研究期结束（2020年12月31日）。若患者接受移植，则移植当月整月被删失。

覆盖月份细分为代偿期和失代偿期肝硬化患者-月。失代偿日期定义为任何计划下首次出现包含失代偿事件ICD或CPT编码（即腹水、自发性细菌性腹膜炎、肝性脑病（HE）、静脉曲张出血、肝肾综合征或肝肺综合征）的索赔服务日期（见补充附录）。患者从失代偿日期所在月份（含）起至观察结束被视为失代偿期肝硬化。代偿期肝硬化患者-月总数从索引月计算至失代偿月。住院成本通过识别入院至出院期间的索赔计算。相同的覆盖月份分类方法应用于失代偿HF和急性加重期COPD。失代偿HF和急性加重期COPD定义为以相应疾病为主要诊断的住院入院。

成本定义为保险公司和患者支付的总金额，采用医疗保健部门成本视角。这些数据通过研究期间特定日历月的平均成本进行时序分析和报告。按患者-年报告的数据通过将患者-月数据乘以12估算。遵循常规做法，拒绝的服务项目排除在成本计算之外。平均成本使用美国劳工统计局消费者价格指数调整至2021年美元，以校正2011至2020年的通胀。平均成本报告时附相应标准差（±SD）。相关亚组间比较采用不等方差双侧Student t检验。补充附录包含患者数据生成过程及标准推断方法基础假设的讨论。为识别成本增加的预测因素，对肝硬化患者成本进行多变量线性回归模型分析。关注的结局为每月每位患者成本，采用连续模型。参考组为非女性ETOH患者。所有统计分析均采用Stata 14.1完成。