Development of an epilepsy-specific risk adjustment comorbidity index

Authors

  • Christine St. Germaine-Smith,

    1. Department of Clinical Neurosciences and Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
  • MingFu Liu,

    1. Alberta Health Services, Calgary, Alberta, Canada;
    Search for more papers by this author
  • Hude Quan,

    1. Department of Community Health Sciences and Calgary Institute of Population and Public Health, University of Calgary, Alberta, Canada
    Search for more papers by this author
  • Samuel Wiebe,

    1. Department of Clinical Neurosciences and Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada
    2. Department of Community Health Sciences and Calgary Institute of Population and Public Health, University of Calgary, Alberta, Canada
    Search for more papers by this author
  • Nathalie Jette

    1. Department of Clinical Neurosciences and Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada
    2. Department of Community Health Sciences and Calgary Institute of Population and Public Health, University of Calgary, Alberta, Canada
    Search for more papers by this author

Address correspondence to Nathalie Jette, MD, MSc, FRCPC, Assistant Professor Neurology, University of Calgary, 1403 29 Street NW, Calgary, AB, Canada. E-mail: nathalie.jette@albertahealthservices.ca

Summary

Purpose:  To develop an epilepsy-specific comorbidity risk adjustment index for mortality outcomes research.

Methods:  Data were extracted from five linked administrative databases in Calgary, Canada from April 1, 1996 to March 31, 2004. Epilepsy patients were defined using a validated ICD-9-CM– and ICD-10-CA–based case definition. An epilepsy-specific comorbidity index was developed using comorbidities from the Charlson and Elixhauser indexes and other relevant epilepsy comorbidities. In the final model, 14 comorbidities significantly associated with mortality remained and each was assigned a value of 1–6 based on the hazard ratio from the survival analysis. Total prognostic scores were calculated and compared for all subjects using the epilepsy-specific index and the Charlson index. Crude mortality and survival curves of both indices were compared.

Key Findings:  We identified 7,253 subjects who met our case definition for epilepsy. The mean age of participants was 38 years (range 0.03–96), and 52% were male. The mortality rate was 7.9%. High rates of chronic pulmonary disease (20.3%), hypertension (19.6%), cerebrovascular disease (13.7%), fracture (12.1%), depression (28.2%), and alcohol abuse (10.1%) were noted. Patients with lower total prognostic scores were more likely to survive than patients with higher scores, using both indices. However, increasing prognostic scores were more strongly associated with reduced survival using the epilepsy-specific index compared to the Charlson index.

Significance:  A new comorbidity index for epilepsy, designed to include clinically relevant conditions, provided better discrimination of crude mortality in a population-based group of epilepsy patients compared with the Charlson index.

In outcomes research, risk adjustment is necessary to account for patient characteristics and disease severity. Various risk adjustment methodologies have been used in previous outcome studies including grouping patients according to clinical severity of disease, coexisting conditions, and comorbidities or disease-related groups, or by utilizing comorbidity indices such as the Charlson or Elixhauser (Charlson et al., 1987; Deyo et al., 1992; Elixhauser et al., 1998; Quan et al., 2005). These techniques have been applied to patients with various conditions or in various settings to predict mortality from cancer (Charlson et al., 1987), liver disease (Cockerell et al., 1994), renal disease (Barooni et al., 2007), and stroke (Annegers et al., 1984) and during an intensive care unit stay (Hauser et al., 1980; Kwon et al., 2008).

Epilepsy is one of the most common neurologic disorders affecting 50 million people worldwide, and has been found to be associated with an increased risk of mortality compared to the general population (Lhatoo et al., 2001; Hitiris et al., 2007). The unique comorbidity profile of those with epilepsy must be accounted for in risk adjustment, especially for mortality outcome studies. Although several general comorbidity indexes have been developed for risk adjustment, including the Charlson and Elixhauser, they were developed in hospitalized patients or patients with cancer and may not reflect the prevalent comorbidities associated with mortality in epilepsy patients (Charlson et al., 1987; Elixhauser et al., 1998; Mohanraj et al., 2006; Kwon et al., 2008). Therefore, we developed a risk adjustment comorbidity index designed specifically for epilepsy and then compared it to the commonly used Charlson index, by defining a population-based epilepsy cohort and following their mortality between April 1999 and December 2005. Our hypothesis was that our new epilepsy-specific comorbidity index would provide better discrimination of crude mortality in a population-based group of epilepsy patients compared with the Charlson index.

Methods

Data sources

Data were extracted from five linked administrative health databases in Calgary, Alberta, Canada from April 1, 1996 to March 31, 2004 (except the vital statistics database, which included data up to December 31, 2005). The five linked clinical databases include: the Discharge Abstract Database (DAD), which includes data on hospital admissions; the Ambulatory Care Classification System (ACCS), which includes information on emergency room visits, some hospital based clinic visits, and day surgeries; the Alberta Physicians Claims Database, which captures all inpatient and outpatient encounters with a physician; the provincial vital statistics database, which includes information on births and deaths; and the Alberta Health Care Insurance Plan Registry (AHCIP), which includes for every Alberta resident a personal health number that is used for data linkage between each of the above databases.

Prior to April 1, 2002, the DAD contained up to 16 ICD-9-CM coding fields for diagnoses and 10 for procedures. Since April 1, 2002, the DAD contains up to 50 ICD-10-CA coding fields for diagnoses. The ACCS contains up to 15 coding fields for diagnoses (ICD-9-CM until March 31, 2002 and ICD-10-CA since April 1, 2002). The physician claims database codes for up to three diagnoses using the ICD-9-CM coding classification. The approval for access to the data used in this study was provided by the University of Calgary Conjoint Health Research Ethics Board.

Study population

Epilepsy patients of all ages were identified in the above-described health databases using the epilepsy International Classification of Disease, Ninth Edition, Clinical Modification (ICD-9-CM) code 345 or the ICD-10-CA codes G40-G41 found in any diagnostic field. We used an optimal case definition, with >95% sensitivity and specificity in our regional databases, that was developed and validated in a previous study (Jette et al., 2010; Reid et al., 2010). Epilepsy patients had to have two physician billing claims, one hospitalization, or one emergency visit coded for epilepsy over 2 years between April 1, 1996 and March 31, 2004 to meet the case definition for epilepsy. The index date was the date at which any of the above scenarios was first encountered. To increase the likelihood of capturing newly diagnosed cases, we then excluded all patients who met the case definition for epilepsy between April 1, 1996 and March 31, 1999, and only included those patients who met the case definition between April 1, 1999 and March 31, 2004.

Mortality

Patients with epilepsy were linked to the vital statistics database to identify those who died between April 1, 1999 and December 31, 2005 and to obtain the date of death.

Development of the comorbidity index

To develop the epilepsy-specific comorbidity index, we first compiled a list of all comorbidities from the Charlson and Elixhauser comorbidity indices (Charlson et al., 1987; Elixhauser et al., 1998). The Charlson index includes the following comorbidities: myocardial infarct, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic pulmonary disease, connective tissue disease, ulcer disease, mild liver disease, diabetes, hemiplegia, moderate or severe renal disease, diabetes with end organ damage, any tumor, leukemia, lymphoma, moderate or severe liver disease, metastatic solid tumor, and AIDS. The Elixhauser index includes the following comorbidities: congestive heart failure, cardiac arrhythmias, valvular disease, pulmonary circulation disorders, peripheral vascular disorders, hypertension, paralysis, other neurologic disorders, chronic pulmonary disease, diabetes (uncomplicated), diabetes (complicated), hypothyroidism, renal failure, liver disease, peptic ulcer disease excluding bleeding, AIDS, lymphoma, metastatic cancer, solid tumor without metastasis, rheumatoid arthritis/collagen vascular disease, coagulopathy, obesity, weight loss, fluid electrolyte disorders, blood loss anemia, deficiency anemias, alcohol abuse, drug abuse, psychoses, and depression. To expand the list, we identified relevant epilepsy comorbidities associated with mortality based on published population-based studies or large cohort studies (Hauser et al., 1980; Annegers et al., 1984; Cockerell et al., 1994; Jalava & Sillanpaa, 1996; Nilsson et al., 1997; Shackleton et al., 1999; Lindsten et al., 2000; Lhatoo et al., 2001; Gaitatzis et al., 2004; Mohanraj et al., 2006; Barooni et al., 2007; Kwon et al., 2008). This full list of comorbidities was then clinically evaluated by two fellowship-trained epileptologists, and comorbidities were removed or additional ones added based on their clinical relevance to patients with epilepsy. A total of 24 comorbidities from the Charlson and Elixhauser comorbidity indexes and 9 additional clinically relevant comorbidities that were either not included or were not listed as independent conditions in the Charlson or Elixhauser indices (brain tumor, aspiration pneumonia, traumatic brain and head injuries, multiple sclerosis, cerebral palsy, anoxic brain injury, encephalopathy, fracture, and CNS infections) were evaluated for their statistical significance (see statistical analysis section below and Table S1 for the full list of comorbidities tested).

For comorbidities already included in the Charlson or Elixhauser indices, the corresponding ICD-9-CM (http://www.chaps.ucalgary.ca/sas) and ICD-10-CA codes were used (Quan et al., 2005) (Table S1). For new comorbidities, ICD-9-CM and ICD-10-CA codes were reviewed and clinically relevant codes were selected to define each comorbidity (also included in Table S1). Comorbid conditions were defined using 3 years of linked administrative data prior to the epilepsy diagnosis date (i.e., index date).

Statistical analysis

Descriptive statistics were used to assess baseline demographics of our study population. A Cox proportional hazards regression model was used to identify significant comorbidities. Survival time was calculated from the date of diagnosis (met case definition for epilepsy) to the date at which the data were censored or mortality occurred. Data were censored if a patient stopped paying Alberta health premiums (usually indicating out of province migration), died, or was still alive at the end of the study (December 31, 2005). The cohort of epilepsy patients was fit to a basic model including age (<18, 18–34, 35–64, 65+ years) and sex only, and then the full model including age, sex, and all comorbidities. In an effort to focus the index on comorbidities significantly associated with mortality in patients with epilepsy, stepwise selection was used to develop a parsimonious model selecting the comorbidities with a p-value ≤0.25 to enter the model and a p-value ≤0.05 to stay in the model.

The prognostic score for each comorbidity included in the parsimonious model was calculated by rounding the hazard ratio to the nearest integer from 1–6. This prognostic score simplifies the epilepsy-specific comorbidity index providing comorbidity weights that can be summed for each patient to provide a total prognostic score. For example, a patient with metastatic cancer and dementia would have a total prognostic score of 6 (metastatic cancer) + 2 (dementia) = 8. A total prognostic score was calculated by identifying all comorbidities in the parsimonious model coded for each patient and adding together the corresponding prognostic scores. Crude mortality was calculated for total prognostic scores from 1 to 10+. Survival analysis was conducted to develop survival curves.

To assess the face validity of our new index, we compared crude mortality and survival curves using total prognostic scores from the epilepsy-specific comorbidity index and the Charlson index. The Charlson index was selected for comparison because, unlike the Elixhauser index, it is a comorbidity index where a score that is also weighted from 1 to 6 can be generated. This allows for simple comparison to our newly developed index using proportional hazards regression. Survival curves were compared using the Kolmogorov-Smirnov test and statistical significance set at p ≤ 0.05. SAS version 9.2 (SAS Institute, Cary, NC, U.S.A.) was used to perform all analyses in this study.

Results

A cohort of 7,253 patients met our case definition for epilepsy, and these patients were eligible for the survival analysis. The mean age with standard deviation was 38 years ± 22.9 years; 52% of the patients were male (Table 1). Prevalence by age distribution was as follows: 1,625 patients (22.4%) were younger than 18 years of age, 1,724 patients (23.8%) were 18–34 years, 2,843 patients (39.2%) were 35–64 years, and 1,061 patients (14.6%) were 65 or older. Mean follow-up time was 3.5 years (0–6.8 years). The mortality rate in this epilepsy cohort was 7.9%.

Table 1.   Baseline characteristics of the study population
 No. (%)
Number of subjects7,253
Follow-up time, years, mean (SD) [range]3.5 (1.7) [0–6.75]
Died during the study period576 (7.9%)
Male3,745 (51.6%)
Age, years, mean (SD) [range]38 (22.9) [0.03–96]
 <18 years1,625 (22.4%)
 18–34 years1,724 (23.8%)
 35–64 years2,843 (39.2%)
 65+ years1,061 (14.6%)

Table 2 shows the prevalence of comorbidities for the epilepsy study group. High rates of chronic pulmonary disease (20.3%), hypertension (19.6%), cerebrovascular disease (13.7%), fracture (12.1%), depression (28.2%), and alcohol abuse (10.1%) were noted.

Table 2.   Prevalence of comorbidities
ConditionN (%)
Congestive heart failure356 (4.9)
Peripheral vascular disease238 (3.3)
Chronic pulmonary disease1,475 (20.3)
Renal disease206 (2.8)
Mild liver disease170 (2.3)
Moderate or severe liver disease49 (0.7)
Diabetes complicated208 (2.9)
Diabetes uncomplicated182 (2.5)
Peptic ulcer disease excluding bleeding104 (1.4)
Myocardial infarction272 (3.8)
Valvular disease223 (3.1)
Metastatic cancer115 (1.6)
Brain tumor215 (3.0)
Solid tumor without metastasis251 (3.5)
Rheumatoid arthritis/collagen vascular disease269 (3.7)
Paraplegia and hemiplegia274 (3.8)
Cerebrovascular disease992 (13.7)
Aspiration pneumonia122 (1.7)
Dementia296 (4.1)
Pulmonary circulation disorders172 (2.4)
Cardiac arrhythmias556 (7.7)
Hypertension1,424 (19.6)
Traumatic brain and head injuries584 (8.0)
Multiple sclerosis115 (1.6)
Cerebral palsy144 (2.0)
Anoxic brain injury76 (1.1)
Encephalopathy108 (1.5)
Alcohol abuse729 (10.1)
Drug abuse535 (7.4)
Psychoses398 (5.5)
Depression2,044 (28.2)
Fracture878 (12.1)
CNS infections156 (2.2)

Of the 33 comorbidities assessed for inclusion in the epilepsy-specific comorbidity index, 14 were significantly associated with mortality and remained in the final parsimonious model (Table 3).

Table 3.   Hazard ratio estimates for the risk of mortality using a basic, full, and parsimonious model
 Basic model
HR (95% CI)
Full model
HR (95% CI)
Parsimonious model
HR (95% CI)a
  1. a Parsimonious model used p ≤ 0.25 to enter the model and p ≤ 0.05 to stay in the model.

<18 years111
18–34 years1.08 (0.65–1.82)1.08 (0.64–1.82)1.12 (0.67–1.88)
35–64 years4.31 (2.95–6.53)3.04 (2.04–4.68)3.26 (2.22–4.97)
65+ years21.32 (14.77–32.10)7.92 (5.23–12.40)8.29 (5.54–12.83)
Male sex1.23 (1.05–1.45) 1.31 (1.10–1.56) 1.30 (1.10–1.54)
Congestive heart failure 1.71 (1.32–2.21) 1.69 (1.32–2.14)
Peripheral vascular disease 1.65 (1.27–2.12)1.66 (1.28–2.12)
Chronic pulmonary disease 0.98 (0.81–1.19)
Renal disease 1.62 (1.20–2.16)1.63 (1.23–2.13)
Mild liver disease 0.86 (0.54–1.33)
Moderate or severe liver disease 2.80 (1.62–4.58)3.25 (1.97–5.05)
Diabetes complicated 1.16 (0.84–1.57)
Diabetes uncomplicated 1.08 (0.75–1.51)
Peptic ulcer disease excluding bleeding 0.86 (0.54–1.30)
Myocardial infarction 0.88 (0.67–1.17)
Valvular disease 1.10 (0.82–1.47)
Metastatic cancer 6.00 (4.40–8.04)6.09 (4.54–8.02)
Brain tumor 3.27 (2.40–4.35)3.29 (2.43–4.36)
Solid tumor without metastasis 1.56 (1.16–2.05)1.52 (1.14–1.99)
Rheumatoid arthritis/collagen vascular disease 1.01 (0.73–1.37)
Paraplegia and hemiplegia 1.54 (1.15–2.03)1.63 (1.24–2.13)
Cerebrovascular disease 1.13 (0.93–1.38)
Aspiration pneumonia 2.21 (1.56–3.07)2.32 (1.67–3.15)
Dementia 1.70 (1.33–2.15)1.76 (1.39–2.21)
Pulmonary circulation disorders 1.38 (1.00–1.88)1.43 (1.04–1.92)
Cardiac arrhythmias 1.32 (1.04–1.65)1.32 (1.05–1.64)
Hypertension 1.28 (1.04–1.57)1.31 (1.08–1.60)
Traumatic brain and head injuries 0.86 (0.63–1.15)
Multiple sclerosis 0.99 (0.47–1.82)
Cerebral palsy 1.47 (0.79–2.51)
Anoxic brain injury 2.98 (1.92–4.46)3.19 (2.11–4.63)
Encephalopathy 1.03 (0.67–1.53)
Alcohol abuse 1.27 (0.96–1.67)
Drug abuse 1.09 (0.80–1.47)
Psychoses 1.00 (0.74–1.34)
Depression 1.08 (0.89–1.30)
Fracture 1.05 (0.83–1.32)
CNS infections 1.06 (0.67–1.62)

In this group of epilepsy patients, the distribution of the total prognostic score calculated using the new epilepsy-specific index (Table 4) was 0 (67%), 1 (11%), 2 (6%), 3 (6%), and 4+ (10%), and using the Charlson index was 0 (57%), 1 (20%), 2 (9%), 3 (5%), and 4+ (9%).

Table 4.   Conditions associated with each prognostic score in the epilepsy-specific comorbidity index
Prognostic scoreConditions
1Pulmonary circulation disorders
Hypertension
Cardiac arrhythmias
2Congestive heart failure
Peripheral vascular disease
Renal disease
Solid tumor without metastases
Paraplegia and hemiplegia
Aspiration pneumonia
Dementia
3Brain tumor
Anoxic brain injury
Moderate or severe liver disease
6Metastatic cancer

Mortality was higher using the epilepsy-specific comorbidity index, ranging from 0.0047/person-year for score 0 to 0.5356/person-year for score >10 compared to the Charlson index ranging from 0.0058/person-year for score 0 to 0.3704/person-year for score >10 (Fig. 1; Table S2). The slope for the epilepsy-specific comorbidity index was also steeper, suggesting better discrimination of mortality using this new epilepsy-specific index versus the Charlson index. Patients with lower total prognostic scores were more likely to survive than patients with higher scores, using both the epilepsy-specific comorbidity index and the Charlson index (p < 0.001 for both measures) (Fig. 2A,B, Table S3). However, increasing prognostic score was more strongly associated with reduced survival using the epilepsy-specific comorbidity index. Although survival curves using the Kolmogorov-Smirnov test were not statistically different for prognostic scores of 2 or less [score 0 (p = 0.87), 1 (p = 0.37), and 2 (p = 1.00)], they were statistically different for scores 3 (p = 0.0005) and 4+ (p < 0.0001), with better performance of the epilepsy-specific index compared to the Charlson index.

Figure 1.


Crude mortality by total prognostic score using the new epilepsy-specific comorbidity index versus the Charlson index.

Figure 2.


Comparison of survival curves for the epilepsy-specific comorbidity score and the Charlson comorbidity score. (A) Survival curves by epilepsy-specific comorbidity score. (B) Survival curves by Charlson comorbidity score.

Discussion

Using population-based health data, we developed a clinically relevant comorbidity index for risk adjustment in epilepsy outcomes research, notably for mortality research. This new index provides better discrimination of mortality in our epilepsy cohort than the commonly used Charlson index. The final epilepsy-specific comorbidity index includes 11 comorbidities from the Charlson and Elixhauser indices and 3 new comorbidities included as independent variables: aspiration pneumonia, anoxic brain injury and brain tumor. Although these new comorbidities may be rare in a general patient population, they are more prevalent and clinically relevant in those with epilepsy (DeToledo et al., 2004; San-Juan et al., 2009; Vecht & Wilms, 2010). Most of the comorbidities in our model are complex disease processes where prevention may be difficult; however, aspiration pneumonia is a comorbidity in which preventive measures may reduce the occurrence of this condition (DeToledo et al., 2004).

Several comorbidities known to be associated with higher mortality were prevalent in our epilepsy cohort but failed to reach the statistical significance necessary to be included in the final index. These include depression (28.2%), cerebrovascular disease (13.7%), and traumatic brain and head injuries (8.0%). For cerebrovascular disease and traumatic brain and head injuries, the nonsignificance may be explained by the fact that mortality from these conditions tends to be the highest in the first few hours to weeks after the event (Mosenthal et al., 2002; Palnum et al., 2009; Shi et al., 2009). Therefore, these patients with early mortality may not have survived long enough to meet our case definition for epilepsy, thereby decreasing our sample size for these conditions. It is unclear why depression was not a significant predictor of mortality in our study, given that psychiatric disorders are associated with an increased risk of suicide in those with epilepsy (Christensen et al., 2007). However, the number of completed suicides in those with depression in our cohort may have been too small to significantly impact mortality, or the follow-up period of our cohort may have been too short. Although these comorbidities are prevalent in patients with epilepsy, their association with mortality was not statistically significant. Therefore, they were likely not strong enough to significantly improve the predictive value of our epilepsy-specific risk-adjustment index and were excluded from the final model.

In general, risk-adjustment indices perform better with an increasing number of independent variables in the statistical model (Southern et al., 2004). In contrast to the Charlson index, which has 17 conditions, our index includes only 14 conditions. However, our list discriminates mortality better than the Charlson index. This finding suggests that the epilepsy-specific index has high face validity.

There are strengths and limitations to our study. One of the major strengths of the new epilepsy-specific comorbidity index is that it was developed for use in administrative health data, which are used internationally. We provide both the ICD-9 and ICD-10 codes used to define the conditions in our new index; therefore, it is widely applicable. As a result, the index can be used for international comparison studies as well as quality assessment studies across local healthcare settings (i.e., comparing outcomes between hospitals in one health region). In addition, the index is user-friendly, as simple calculations provide one numerical value that reflects the comorbidity burden for each patient. Although it was developed in a large population-based sample, the use of weights for each condition and a summary score to predict mortality ensures that it can also be used for small studies, even when frequency of comorbidities is low. One limitation is the use of administrative data, which can be associated with coding inaccuracy. However, epilepsy coding has been validated in our health region and has been found to be excellent, with high sensitivity (95.6%), specificity (87.4%), positive predictive value (84.3%),and negative predictive values (96.9%) (Jette et al., 2010; Reid et al., 2010). Secondly, our group has also validated comorbidity coding in hospital discharge data in our health region (Quan et al., 2008) and has developed validated ICD-9-CM and ICD-10-CA coding algorithms for the Charlson and the Elixhauser indices (Quan et al., 2005). Another limitation was our inability to adjust for epilepsy severity or the social determinants of health in this study, both of which could potentially affect mortality. Finally, our data may not be generalizable to other administrative databases or to other regions. In the future, the performance of the new index should be validated in other health databases or other large patient samples to confirm its predictive ability in other populations.

Despite the above-noted limitations, our study is truly population-based as we captured data from >99% of the population in our health region due to the existence of universal health care coverage, thereby eliminating selection bias. In addition, all contacts with the health care system are captured in our data sources, thus avoiding recall bias. We assessed this index using mortality as the only outcome variable. In addition to mortality, the applicability of this epilepsy-specific index could be assessed in other areas of outcomes research including health resource utilization, medical costs, and quality of life studies. Furthermore, additional validations in epilepsy subgroups (stratifying patients by epilepsy type, surgical candidacy, developmental delay, and so on) would be valuable to confirm the validity of the index in specific epilepsy populations.

In conclusion, we developed an epilepsy-specific index using a large population-based cohort and our comparison with the Charlson index provides face validity. We encourage researchers to use or test this comorbidity risk adjustment index in their own epilepsy-specific research.

Acknowledgments

This study was supported by a University of Calgary starter grant to N. Jette. N. Jette and H. Quan are the recipients of population health investigator salary awards from Alberta Innovates Health Solutions. H. Quan holds a new investigator award from the Canadian Institutes of Health Research. N. Jette also holds a Canada Research Chair Tier II in Neurological Population Health and Health Services Research. S. Wiebe holds the Hopewell Professorship of Clinical Neurosciences Research at the University of Calgary.

Disclosure

None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary