Findings

View the most recent version.

Hospital care for smoking-related illnesses constitutes an important part of the health care burden.  However, because hospital administrative records contain only limited information, quantifying hospital use according to patients’ personal characteristics is challenging.  For example, although smoking may have contributed to the illness for which a person is hospitalized, no information on smoking history is captured in the administrative discharge abstract.

An indirect method of assessing the impact of smoking on hospital use has been to apply the “attributable fraction” for diseases that have been shown to be etiologically associated with smoking. The attributable fraction is the proportion by which the incidence of a disease would be reduced in the population if smoking were eliminated.  Smoking-attributable days in hospital are calculated by multiplying each disease- and age-specific fraction (based on the literature) by the total number of hospital days for that disease.  Since the early 1980s, studies in Canada and elsewhere have employed this approach to estimate health care use associated with smoking.^1-6

Estimates derived from the attributable-fraction approach are limited in several ways. First, calculations are based on disease-specific risk estimates in the literature, so their precision depends on the quality of the underlying evidence. Second, the attributable-fraction approach does not include conditions for which there is currently insufficient understanding of their true association with smoking. Third, this approach does not cover smokers’ higher risks of complications after surgery and slower recovery time.^7-15

A more accurate technique for quantifying smoking-related hospital use involves the collection of baseline data from a cohort, and monitoring their hospital use over a period of time.  Few studies have employed this method, probably because of the considerable resources required for follow-up.^16,17 A similar, but less expensive, method entails linking survey data from a specified baseline year with administrative data covering a subsequent follow-up period.^18-20  Recent linkage of health survey data to hospital administrative data offers an opportunity to study the association between smoking status and use of acute care hospitals in Canada.

The objective of this study was to quantify the likelihood of hospitalization, and estimate time spent in hospital, according to smoking status. The study is based on a sample of respondents aged 45 to 74 to the 2000/2001 Canadian Community Health Survey, whose hospitalization experience was monitored prospectively for four years through linkage of the survey data with the Hospital Person-Oriented Information database.

Methods

Data source

Canadian Community Health Survey

Data on smoking and other personal characteristics are from the 2000/2001 Canadian Community Health Survey (CCHS). The survey collects cross-sectional information about the health of Canadians on an ongoing basis. It covers the non-institutionalized household population aged 12 or older in all provinces and territories, except members of the regular Canadian Forces and residents of Indian reserves, of Canadian Forces bases (military and civilian) and of some remote areas.  A description of the methodology has been previously reported.²¹

The overall response rate to the 2000/2001 CCHS was 85%; the total sample numbered 131,535.  Insufficient information was available in the hospital data for Quebec residents’ records to be linked (see Hospitalization data); therefore, the 22,667 (17%) CCHS respondents in Quebec were dropped.  Of the 108,868 respondents who remained, 90,450 had given permission for their survey data to be linked to administrative data. Check-digit algorithms were used to verify the plausibility of the health numbers they provided; 72,363 provided a plausible health number. Respondents aged 45 to 74 numbered 28,288, but data on smoking status were not available for 33 of them (Figure 1). Approximately half of them had been surveyed by telephone, and the other half, in person (data not shown).

Figure 1
Creation of analysis file

Survey weights were produced by Statistics Canada to adjust for non-response to the CCHS, as well as for the exclusion of records of respondents who did not provide plausible health numbers or give permission for linkage to administrative health data.  These weights were applied to the analysis file; the weighted data were representative of the Canadian household population residing outside Quebec.

Hospitalization data

Statistics Canada’s Hospital Person-Oriented Information database (HPOI) is a person-level dataset derived from discharge records of inpatients in most of the acute care hospitals and some psychiatric, chronic and rehabilitation hospitals across Canada.²² The discharge records  contain demographic (for example, date of birth, postal code), administrative (health number, admission and separation dates) and clinical information,²³ and are compiled into the Hospital Morbidity Database by the Canadian Institute for Health Information.²⁴ During processing at Statistics Canada, about 3% of Hospital Morbidity Database records for patients aged 12 or older were excluded because of missing or invalid health numbers.²² Because this analysis was restricted to records for hospitalizations in acute care hospitals, a further 2% of records from non-acute care institutions were excluded. 

After processing to ensure the consistency of demographic information among records assumed to pertain to the same person, two stages of linkage were carried out. First, HPOI records were created by linking records for the same patient, based on health number, postal code and date of birth. Probabilistic linkage routines were then used to match the CCHS with HPOI records. 

A recently published evaluation of the linkage between the CCHS and HPOI reported high coverage for the population younger than age 75. The estimated number of people aged 12 to 74 who were hospitalized, based on the number of CCHS records that were matched to HPOI records, was 96.4% of the number hospitalized according to HPOI records alone.²⁵ To further assess the comparability of the linked file with the original CCHS file (excluding Quebec respondents, but including respondents who did not give permission for linkage), the population distribution by smoking status (percentage who are current smokers, never-smokers, former smokers who quit within five years, former smokers who quit for more than five years) was compared between the two files; the distributions were nearly identical (data not shown).

CCHS interviews were conducted from September 1, 2000 until November 3, 2001.  Starting from the date of the CCHS interview, HPOI records were searched prospectively for four years (1,462 days) for each respondent. Thus, in theory, each respondent had an equal period of eligibility for hospitalization. Censoring before the end of the four-year period because of events such as death or moving out of the province could not be accounted for because information about these events was not available or incomplete.  However, it was expected that such occurrences would be relatively rare in the age group studied and would have a minimal effect on the results of the analysis. Probabilistic linkage of the 28,288 CCHS records to HPOI records yielded 7,229 matches, indicating that 26% of CCHS respondents outside Quebec had been hospitalized at least once during the four-year follow-up.  No matching HPOI records within the four-year follow-up were found for the remaining 21,059 respondents, who were, therefore, categorized as not having been hospitalized. No information on smoking status was available for 13 of those who had been hospitalized and for 20 of those who had not. These records were deleted, yielding final samples of 7,216 and 21,039, respectively.

Analytical techniques

Frequencies, cross-tabulations and means were employed to estimate smoking prevalence, the proportions of persons hospitalized, and hospital days used.  All tabulations were produced by age group, and all computations were carried out with weighted data. Variance on estimates was calculated with the bootstrap technique to account for the complex design of the survey.^26,27    

Preliminary analysis revealed that relatively few (7%) CCHS respondents aged 75 or older were current daily smokers.²⁸ As well, for the population aged 75 or older, correspondence between the CCHS and HPOI is substantially lower than for younger people—largely because the CCHS does not cover the institutionalized population.²⁵ Therefore, the analysis excluded people aged 75 or older. 

Four multiple logistic regression models were fitted to assess the association between smoking status (as ascertained during the CCHS interview) and the odds of hospitalization, while controlling for potential confounders.  The first model contained control variables for age and sex; to that model were added variables reflecting socio-economic status (Model 2), urban/rural residence (Model 3), and physician consultations, leisure-time physical activity, body mass index and level of alcohol consumption (Model 4).  To maximize sample size, variables for missing values on household income, leisure-time physical activity and body mass index were included.  Cox proportional hazards regression, useful in assessing time-to-first-event, was also considered as a means of studying this association.  However, because time-to-hospitalization was unrelated to the purpose of the regression (to assess the possible influence of socio-economic status and other risk factors on the association between smoking and hospitalization), logistic regression was considered an appropriate approach.

Definitions

Respondents were assigned to one of four categories, based on daily smoking:  current daily smoker; former daily smoker who quit sometime in the five years before the date of the CCHS interview (recent quitters); former daily smoker who quit more than five years before the date of the interview (long-term quitters); and never a daily smoker.    

To be consistent with low-risk drinking guidelines,²⁸ seven categories of alcohol consumption were specified:  weekly binge (at least five drinks in one occasion, at least once a week during past year); heavy (ten drinks or more in past week for women; fifteen or more for men); moderate (two to nine drinks in past week for women; two to fourteen for men); light (one drink in past week); occasional (at least one drink in past year, but none in past week); former (at least one drink in lifetime, but none in past year); and lifetime abstainer.

Household income groups were derived by calculating the ratio between total household income from all sources in the previous 12 months and Statistics Canada’s low-income cut-off specific to the number of people in the household, the size of the community, and the survey year.  These adjusted income ratios were sorted and grouped into within-province quintiles (five groups, each containing one-fifth of the population in each province).

Days in hospital were calculated by summing the lengths of stay for each hospitalized CCHS respondent; totals were cross-tabulated by age group and smoking status category. Respondents categorized as not having been hospitalized were included and counted as contributing zero days.      

The average number of days in hospital was calculated by dividing the weighted total number of days for each age group and smoking category by the corresponding weighted population count.  For each age group and smoking category, the average number of excess hospital days was calculated by subtracting the average number of days for never-smokers from the corresponding average number of days for each smoking category. Then the total number of excess days in hospital was calculated by multiplying the average number of excess days in each age group and smoking category by the population in each of these groups.  Finally, the percentage of all days in hospital that were excess days was calculated by dividing the number of excess days by the number of total days.  All estimates were produced for each age group and smoking category.

Results

Of Canadians (minus Quebec residents) aged 45 to 74 in 2000/2001, an estimated 19% were current daily smokers; 6% were former daily smokers who had quit in the past five years (recent quitter); 30% were former daily smokers who had quit for more than five years (long-term quitter); and 44% had never smoked daily (Table 1).  The percentage of current daily smokers in the 45-to-54 age group was higher than in the older age groups.

Table 1
Smoking status, by age group, household population aged 45 to 74, Canada excluding Quebec, 2000/2001

Over the subsequent four-year period, 14% of people aged 45 to 54 were hospitalized—a figure that rose in successively older age groups to 35% of 65- to 74-year-olds (Table 2).  When smoking status was considered, the likelihood of hospitalization during the four years ranged from a low of 12% of never-smokers aged 45 to 54 to a high of 43% of 65- to 74-year-olds who were current daily smokers or recent quitters. 

Table 2
Percentage hospitalized over subsequent four years, by age group and smoking status, household population aged 45 to 74 in 2000/2001, Canada excluding Quebec

In each age group, the percentage of current daily smokers who were hospitalized substantially exceeded the percentage among those who had never smoked daily.  In all age groups, recent quitters were also significantly more likely to have been hospitalized than were people who had never smoked daily.  This pattern held for long-term quitters except at ages 55 to 64:  the percentage of long-term quitters hospitalized was not statistically different from the percentage for people who had never smoked daily. 

Adjusted for age and sex, the odds of hospitalization for current daily smokers were 80% higher than those for never daily smokers (Table 3, Model 1).  The odds ratios for former daily smokers were also significantly elevated, at 1.6 for recent quitters, and 1.3 for long-term quitters. Even as other control variables were added (level of education and income in Model 2; urban/rural residence in Model 3; and family doctor consultation, level of leisure-time physical activity, body mass index category and level of alcohol consumption in Model 4), the odds ratios for all smoking categories remained at nearly the same levels.

Table 3
Adjusted odds ratios relating smoking status to hospitalization over subsequent four years, controlling for selected characteristics, household population aged 45 to 74 in 2000/2001, Canada excluding Quebec

More hospital days

The time current and former daily smokers spent in hospital was disproportionate to their share of the population.  Current daily smokers aged 45 to 74 accounted for 19% of the population in that age range in 2000/2001, but they used 29% of hospital days for that age group (Table 4). Recent quitters constituted 6% of the population, but accounted for 10% of hospital days. Long-term quitters used a percentage of hospital days (32%) that more closely reflected their share of the population (30%).  Finally, people who had never smoked daily made up 44% of the population, but used 30% of hospital days.

Table 4
Number of days and percentage distribution of days in hospital over subsequent four years, by age group and smoking status, household population aged 45 to 74 in 2000/2001, Canada excluding Quebec

The mean number of days in hospital per person (including those who were not hospitalized during the study period) was significantly higher for current daily smokers than for people who had never smoked daily.  In each age group, the average for current daily smokers was more than double that for never-daily smokers (Table 5).  Recent quitters also averaged significantly more days in hospital, compared with never-daily smokers.

Table 5
Average number of days and excess days hospitalized over subsequent four years, by age group and smoking status, household population aged 45 to 74 in 2000/2001, Canada excluding Quebec

For long-term quitters aged 45 to 64, average days in hospital did not statistically exceed averages for never-daily smokers.  However, average days in hospital for long-term quitters aged 65 to 74 (6.8 days) significantly surpassed the average for never-daily smokers in this age group (4.3 days).

Current daily smokers’ average number of excess hospital days ranged from 1.5 days at ages 45 to 54 to 6.0 days at ages 65 to 74.  Recent quitters’ excess days averaged 0.9 at ages 45 to 54, 2.6 at ages 55 to 64, and 8.4 days at ages 65 to 74.  For long-term quitters, average excess days at ages 45 to 64 did not differ significantly from zero; those aged 65 to 74 averaged 2.6 excess days, which was significantly lower than the average of 8.4 excess days for recent quitters.

In total, current and former daily smokers aged 45 to 74 used an excess of 7.1 million hospital days over the four-year study period, which accounted for nearly one-third (32%) of all days spent in hospital by people these ages (Table 6). This proportion was consistent in all three age groups.  When the analysis was repeated for the sexes separately, the excess was 36% of all days for men, and 28% of all days for women (data not shown).

Table 6
Number of excess days† hospitalized over subsequent four years, by age group and smoking status, household population aged 45 to 74 in 2000/2001, Canada excluding Quebec

Discussion

For the first time, population-based estimates of acute care hospital use in Canada (excluding Quebec) in relation to the smoking status of patients have been prospectively estimated—using linked survey and administrative data.  The results indicate that excess days spent in hospital by current and former daily smokers aged 45 to 74 account for nearly one-third of all days spent in hosptial by people in this age range.  Furthermore, the elevated likelihood of hospitalization for current and former smokers is not explained by other characteristics, including socio-economic status. 

The relatively lower averages of excess hospital days for former daily smokers who had quit more than five years before their CCHS interview underscores the benefit of long-term cessation.  This finding is consistent with research reported by the United States Surgeon General showing improvements in mortality risk with time since quitting.²⁹  

A previous study reported increases in hospital admissions for former smokers during the year in which they quit, leading the investigators to speculate that the onset of illness may have motivated their quitting.³⁰ This may explain the high average number of excess hospital days for quitters of five years or less in the 65 to 74 age group in the current study. 

Previous Canadian studies of various smoking-related outcomes (for example, health care use, direct and indirect costs, and mortality) have been based on the attributable-fraction methodology, making comparisons with the current study problematic.^2,3,6 The aim of the current study was to quantify hospital use according to the patient’s smoking status, without regard to the age-specific, smoking-attributable fraction of the patient’s diagnosis. 

Comparisons with studies from other countries based on methodology similar to the current study are also limited—by differences in study endpoints, smoking prevalence, and follow-up time.  For example, in a 16-year follow-up of a nationally representative cohort in Finland, male smokers used 70% more hospital days, and female smokers, 49% more, than did never smokers.¹⁷ In a 30-month prospective study in Japan, the per capita cost of hospital inpatient care was 33% higher for male smokers, compared with non-smokers, but did not differ among females.¹⁶ In Scotland, the odds of hospitalization for a cohort followed for 7.5 years were significantly higher for smokers and former smokers.¹⁹ Despite the variety of methodological approaches among studies, the consistent finding is that smoking is a factor in an appreciable proportion of hospital care.

Limitations

An important limitation of this study is the exclusion of residents of Quebec where the prevelance of smoking was higher in 2000/2001. ³¹Exclusion of the institutionalized and on-reserve populations further added to the undercounting of hospital days used.  Although CCHS respondents for whom a valid health number was not available and those who did not give permission to link their survey data with administrative data were also omitted, Statistics Canada produced special survey weights to adjust for these exclusions.

The hospital care received by smokers is underestimated in this analysis. Data on use of the emergency and outpatient departments were not available. As well, because data were available only from acute care hospitals, specialized facilities such as psychiatric and rehabilitation centres were excluded. And the extent to which exposure to second-hand smoke may have been related to days in hospital could not be estimated.

The maximum response category for the 2000/2001 CCHS question asked of former daily smokers about time since quitting was “more than 5 years ago.”  A more detailed breakdown (for example, 5 to 10 years; 10 years or more) would have permitted an examination of the length of cessation time that is required before hospital use by former smokers might fully drop to the level of people who have never smoked.  As well, information on intensity of exposure to smoking (pack-years) was not available. 

Smoking status was assessed only at the beginning of the period and may have changed during the subsequent four years. Current smokers could have quit, and former smokers could have resumed.  Previous research based on longitudinal data indicates that within a two-year period, 13% of daily smokers had quit.³² However, among former smokers, recidivism was particularly high in the first two years—approximately 20% had relapsed.  Although the full extent to which such changes may have diluted or biased the observed associations cannot be quantified, it is likely that the association with hospitalization observed for people classified as recent quitters (quit in past five years) was stronger than it would have been if recidivists had been excluded.

Factors other than smoking that could not be included in the analysis may account for some of the excess hospitalization among smokers.  Such factors might include a propensity for risk-taking, poorer nutritional status and less frequent primary and preventive care.

Data from the CCHS were self-reported by respondents; no independent verification of the information was undertaken.  The degree to which the data may be biased because of reporting error is unknown.  For example, any tendency to deny or underreport smoking would contribute to misclassification, which would weaken the strength of the association between hospital use and smoking.

Probabilistic linkage was used to match hospitalization records to survey records; some false links or missed links may have resulted.

Conclusion

This analysis illustrates the value of linking administrative to survey data by being the first Canadian study to directly quantify time spent in hospital according to smoking status.  The precision of the association between hospitalization and smoking status was enhanced by controlling for other possible influences such as obesity and socio-economic status.