Socioeconomic and marital status among liver cirrhosis patients and associations with mortality: a population-based cohort study in Sweden

This study is based on a cohort described earlier in detail [6], which was retrieved by a systematic inventory of cirrhosis data in a well-defined Swedish population. Essentially, a broad search was performed for all patients with cirrhosis diagnosed in the Region of Halland (310,665 inhabitants, year 2014) between January 1st 2011 and December 31st 2018 [6].

The search was conducted using a wide array of cirrhosis-related International Classification of Diseases 10th Revision – Swedish Edition (ICD-10-SE) codes (Supplementary Material 1) [6]. The use of ICD-10-SE has been mandatory in Sweden since January 1st 2011 and, according to the Swedish National Board of Health and Welfare, none of the codes described in the supplementary material have been changed since then [23]. In addition, patient data were retrieved from the pathology registry using the Systemized Nomenclature of Medicine (SNOMED) codes for liver (T-56), cirrhosis (M-495), and HCC (M-817) as described elsewhere [5, 6]. SNOMED is an international nomenclature system that is broadly used in pathology [24].

Liver cirrhosis was diagnosed by histology, or based on clinical and laboratory findings combined with standardised radiological features previously described [6, 25]. All patients without distinctive radiological features or typical histology confirming the diagnosis of cirrhosis, were excluded. Minor patients (age under 18 years), patients not registered as permanent residents in the Region of Halland, and those diagnosed before 2011 or after 2018, were also excluded.

All medical records were thoroughly reviewed and information was retrieved regarding age, sex, date of diagnosis, aetiology, marital status, employment status, occupation, length and weight, complications and comorbidities at diagnosis, laboratory results, use of warfarin, date of moving from the Region of Halland, liver transplantation or death and cause of death.

Each patients was included into one the following aetiological groups: alcohol, HCV, Non-alcoholic fatty liver disease (NAFLD), cryptogenic cirrhosis, primary biliary cholangitis, autoimmune hepatitis, and “Other causes” [6]. “Other causes” comprised patients with less common aetiologies in our cohort (observed in a previous study) [6], such as primary sclerosing cholangitis, hepatitis B and hemochromatosis. Alcohol-related cirrhosis was defined for patients with a history of long-lasting alcohol overconsumption and/or elevated concentrations of phosphatidylethanol, or (in some cases) carbohydrate-deficient transferrin [6]. If a patient had both alcohol overconsumption and HCV as possible causes of cirrhosis, only the latter was registered as the aetiology [6]. Patients without a well-defined cause of cirrhosis were regarded as having cryptogenic cirrhosis. NAFLD was only used when diagnosed by clinicians or verified through biopsy.

In addition to the electronic registries described earlier, medical journal data from scanned reports from 1990 to 2011 were revised when compiling SES. These medical records included data regarding sick-leave certificates, disability certificates and certificates of fitness. These certificates were in turn regulated by the Swedish Social Insurance Agency, and included employment status and main reported occupation. Patients were classified as married (including cohabiting), previously married (separated, divorced or widowed), and never married, at the time of diagnosis. Patients were also classified according to their employment status at the time of diagnosis: employed, pensioner, disability retired, or unemployed. Employment status is strongly associated with the mean disposable annual income (Supplementary Material 2).

Patients’ occupations were categorised according to the Swedish Standard Classification of Occupations 2012 (SSYK 2012) [26]. SSYK 2012 is based on the International Standard Classification of Occupations 2008 (ISCO-08), which grades occupations into four main occupational skill levels (level I-IV) [27]. Skill level refers to type of working task and its complexity. Translated into a Swedish context, level I represents elementary occupations – low-skilled (e.g. cleaners), level II includes most skilled manual-workers – medium-skilled (e.g. plumbers, drivers), level III represents occupations requiring up to 3 years of tertiary education – medium-high skilled (e.g. IT technicians, real estate agents), and level IV mainly stands for managers and professionals – high- and very high skilled [26, 27]. In Sweden, there are strong associations between occupational skill level and educational level (Table 1). In turn, occupational skill level is strongly linked to income level (Supplementary Material 2).

Each patient (represented once) was included in one of the main occupational skill level groups. If several occupations were registered for a patient, only the most representative (longest period of time or main income source during work-life) was registered. If a patient had various simultaneous occupations, only the one referred by the patient (or the treating physician) as the primary was registered. Pensioners and disability retired were classified according to the main occupation reported during their work-life. Unemployed, without known previous occupation, or with a history of longstanding unemployment (10–15 years prior to cirrhosis diagnosis), were included in the occupational skill level I.

A comorbidity was included if diagnosed for up to 10 years prior to, and observed at the time of cirrhosis diagnosis. Body mass index (BMI) values (kg/m²) were calculated as defined by WHO [28]. The following definitions were applied: underweight (BMI < 18.5), normal weight (BMI 18.5–24.9), pre-obesity (BMI 25.0–29.9), and obesity (BMI > 29.9) [28]. Ascites was registered if detected clinically and/or radiologically. Variceal bleeding was assumed upon evident signs of bleeding according to the Baveno IV classification of significant bleeding [29]. Hepatic encephalopathy was registered if observed at diagnosis or under the initial follow-up. Only HCC cases diagnosed at the time of cirrhosis diagnosis, or within the 6 months after, were considered a complication.

Model of end-stage liver disease (MELD) scores and Child-Pugh class were calculated and used as indicators of cirrhosis severity [30, 31]. Patients were divided into the following groups: MELD < 10, MELD 10–14, and MELD ≥15. Patients with advanced chronic kidney disease and/or treatment with warfarin were excluded from the calculation of MELD score and Child-Pugh class, being thereby considered as missing data for these variables.

Each patient was followed until December 31st 2019, date of transplantation, death or moving from the Region of Halland (whichever occurred first). End-point data were automatically linked to our medical record system via the Swedish Civil Registration System. For deceased patients, the cause of death was obtained from our medical system as all death certificates for residents in the Region of Halland are routinely scanned and integrated into the medical journal. These death certificates are duplicates of the ones sent to the Swedish National Board of Health and Welfare. If the cause of death was missing, the data were obtained from the Swedish National Cause of Death Register.

Data were expressed as medians and percentiles, or as numbers and percentages, depending on the type of variable presented. Chi-square test, or Fisher’s test (when appropriate), was performed for the comparison of sex, aetiology, comorbidities, stratified MELD score, Child-Pugh class, and complications between the different marital status, employment status, and occupational skill levels. Median values for age at diagnosis and MELD score were compared among the different marital status, employment status, and occupational skill levels, by the Kruskal-Wallis test [32]. Chi-square test was used for the comparison of causes of death between the different sexes, marital status, employment, and occupational skill level. Missing data was expressed as percentages.

Primarily, we considered transplant-free survival, i.e. each patient was followed from date of diagnosis until date of death, transplant surgery, moving from the Region of Halland, or December 31st 2019, whichever occurred first. Moreover, we analysed survival among the patients who underwent a transplant, i.e. each such patient was followed from date of transplant surgery until date of death or December 31st 2019, whichever occurred first. Kaplan-Meier estimates with Greenwood confidence intervals (CI) were used to determine mean survival [33]. Survival curves were compared using the log-rank test. Cox regression models were used to determine hazard ratios (HRs) for the variables of interest: sex, age, marital status, employment status, occupational skill level, aetiology, MELD score, and Child-Pugh class at diagnosis. For a variable with > 2 categories, a reference category was selected and the HRs corresponding to each other category were estimated. Unadjusted HRs were estimated from univariate Cox models, while the adjusted HRs were obtained from a multivariate Cox model that included all variables, together with the registered comorbidities, as covariates. A p-value < 5% from a two-tailed test was referred to as statistically significant. All tests were conducted using IBM SPSS Statistics for Macintosh (version 26.0, IBM Statistics, Amorak, NY, USA).

A total of 598 patients with cirrhosis were identified. From this total, we excluded 16 patients as they were not diagnosed with cirrhosis before autopsy. The final study cohort comprised 582 patients (Fig. 1). Most patients had alcohol-related cirrhosis (51%), and the median age at diagnosis was 66 years. Missing BMI values were registered for 3.3% of the cohort. Normal weight and pre-obesity (35% each) were common, and obesity was registered in 23% of the cohort. Due to missing data, MELD scores and Child-Pugh class were not available for 4.5% of the cohort. The associations between marital status, employment status and occupational skill level are shown in Supplementary Material 3.

Previously married individuals had the highest median age at diagnosis (71 years; p < 0.0001) (Table 2). There were no statistically significant differences regarding the aetiology of cirrhosis upon marital status (p = 0.083). Arterial hypertension and diabetes mellitus were more common among the married individuals (p = 0.047 and 0.025, respectively). The never married group had the highest rate of obesity (p = 0.010) and MELD ≥15 at diagnosis (p = 0.041). There were no statistically significant differences within marital status groups regarding median MELD, Child-Pugh class or complications observed at diagnosis.

Employed had the lowest median age at diagnosis (49 years; p < 0.0001) (Table 2). These patients also had the highest percent of alcohol-related cirrhosis (p < 0.0001). Cryptogenic cirrhosis was common in pensioners (24%), and HCV was very common among the unemployed (40%). Arterial hypertension was uncommon among the unemployed, and cardiac artery disease was more common among the disability retired (p = 0.033 and < 0.0001; respectively). Unemployed had the highest rate of Child-Pugh C, at diagnosis (29%; p = 0.001). These patients also had the highest median MELD and MELD ≥15 rate at diagnosis, but these associations were not statistically significant (p = 0.050, and 0.156; respectively). Employed had the lowest rate of ascites (39%; p = 0.016). Pensioners and disability retired had the highest rate of prevalent HCC at diagnosis (17 and 14%, respectively; p = 0.026).

Occupational skill level IV consisted of only 55 patients (9%; p < 0.0001), and there was a marked male predominance (80%; p = 0.001) (Table 3). Occupational skill level I had the highest frequency of HCV (30%; p < 0.0001). Arterial hypertension was lowest in occupational skill level I (16%; p < 0.0001), and cardiac artery disease was uncommon in occupational skill level IV (p = 0.027). Child-Pugh C was more common in occupational skill level I (28%; p = 0.010). Although MELD ≥15 was 2.15 times more frequent in occupational skill level I compared to occupational skill level IV, this association was not statistically significant (p = 0.178). The prevalence of complications at diagnosis was significantly different between these two occupational skill levels, as ascites and hepatic encephalopathy were 1.6 and 6 times higher in occupational skill level I (p = 0.045 and p = 0.022, respectively). There were no statistically significant differences regarding the prevalence of other comorbidities or complications at diagnosis among the different occupational skill levels.

During an accumulated follow-up time of 1684 person-years, a total of 319 patients (55%) died, including one of the 18 patients who previously underwent a transplant. The follow-up was censored for the remaining 246 patients. The mean follow-up for surviving patients was 4.52 years (range 1.01–8.95), and the mean survival time was 4.41 years (95% CI, 4.08–4.73).

Men had lower mean survival compared to women (3.81 vs 4.82 years; p = 0.003). This association was consistent with the calculated HR in the univariate analysis (Table 4). Married patients had better mean survival compared to those previously married (4.56 vs 3.35 years; p = 0.010) (Fig. 2). Previously married (HR 1.50, 95% CI 1.14–1.96) had higher HR compared to married individuals in the univariate analysis. However, marital status was not a significant predictor of mortality in the multivariate analysis. Employed had the highest mean survival (5.74 years, 95% CI 5.17–6.32), followed by unemployed (4.30 years, 95% CI 3.15–5.45). There were only minor differences regarding mean survival between disability retired and pensioners (3.63 vs 3.28 years; p < 0.0001) (Fig. 2). This association was consistent with the calculated HR in univariate analysis (p < 0.0001) (Table 4). However, as with marital status, employment was not a statistically significant predictor for mortality in the multivariate analysis.

The mean survival in occupational skill level IV was significantly greater than the mean survival for occupational skill level I and II (6.39 vs 3.00 and 4.04 years, respectively; p < 0.0001) (Table 4). The calculated HRs in univariate analysis were consistent with these findings (p < 0.0001) (Fig. 2). Occupational skill level was also a strong predictor of mortality in the multivariate analysis (Table 4). When aggregated, occupational skill levels I-II also had poorer mean survival (3.79 vs 5.64 years; p < 0.0001) and higher HR compared to the aggregated occupational skill levels III-IV (Fig. 2). This association was also significant in multivariate analysis (Table 4).

Occupational skill level and Child-Pugh at cirrhosis diagnosis were statistically significant predictors of mortality in multivariate analysis (Table 4). Since there were statistically significant differences regarding the prevalence of a given Child-Pugh class between the different occupational skill levels (Table 3), we examined the interaction between the two variables in the Cox models. There were no statistically significant interactions between occupational skill level and Child-Pugh, neither in the original model (occupational skill level I through IV; p = 0.381) nor in the modified one (aggregated occupational skill level I-II and III-IV; p = 0.530). Likewise, there were no statistically significant interactions between occupational skill level and MELD-score in the original or in the modified Cox model (p = 0.334 and 0.483; respectively).

Upon a given Child-Pugh class, occupational skill level was not associated with mean survival among patients with Child-Pugh A (Fig. 3). However, there were distinct associations between mean survival and occupational skill level among patients with Child-Pugh B and C (Fig. 3). These associations were more evident when the aggregated occupational skill levels (I-II vs III-IV) were compared upon the observed Child-Pugh class att cirrhosis diagnosis (Fig. 3).

A total of 18 patients had undergone a transplant and the median age at transplantation was 58 years (range 37–72). The median time from diagnosis to transplantation was 1.75 years (range 0.08–5.40). There were no statistically significant differences among these patients with regard to sex (10 women; p = 0.093), aetiology (p = 0.340), marital status (13 married; p = 0.170) or occupational skill level (p = 0.065). However, most transplant patients (72%; p < 0.0001) were employed at the time of their transplant surgery.

Among the 319 deceased patients, most died due to acute on chronic liver failure and related complications (42%). HCC was the second common cause of death (21%), followed by infection (12%), heart disease (8%), and other cancer forms (8%). Other causes of death, such as stroke, lung or kidney failure, accounted for 9% of all deaths. There were no statistically significant differences regarding the cause of death upon sex, marital status, employment status, or occupational skill level.