The association between smoking cessation outpatient visits and total medical costs: a retrospective, observational analysis of Japanese employee-based public health insurance data

Abstract

Aims: The short-term effects of smoking cessation (SC) on overall healthcare costs are unclear. This study aimed to compare the short-term medical costs between patients with SC outpatient visits (SCOVs) and those without SCOVs, consisting of SCOV itself and overall medical costs.

Materials and methods: This study is a retrospective, observational study using a Japanese employee-based health insurance claims database (January 1, 2005–December 31, 2013). It analyzed individuals who were registered as smokers based on their medical checkup details. It compared the per-patient-per-year (PPPY) medical costs for male smokers who made ≥1 claim for SCOVs with those who made no claims. We also assessed whether the number of SCOVs by male and female smokers impacted medical costs. The Index Year was the year after the first SCOV claim and that after the first registration as a smoker (non-SCOV group). Medical costs were calculated using regression analysis and adjusted for baseline costs.

Results: In Index Year −1, PPPY medical costs for male smokers were ∼USD 323.01 (JPY 36,500, as of November 2017) higher in the SCOV (n = 5,608) vs the non-SCOV (n = 81,721) group; however, by Year 6 the costs were similar. From Year 4–6, PPPY medical costs for SCOVs were lower than those in the adjusted non-SCOV group. For 2,576 male and female smokers in the SCOV group, the average rates of increasing medical costs before and after the SCOV for 1, 2, 3, 4, and 5 SCOVs made were 58%, 44%, 50%, 41%, and 34%, respectively.

Limitations: The database includes limited data on individuals >65 years. Only SCOVs based on claims data and not on other outcomes were assessed.

Conclusions: Medical costs declined in the short-term following the first SCOV. Attendance at a greater number of SCOVs was associated with a lower increase ratio of medical costs.

Keywords:

• Smoking cessation
• health insurance
• outpatient visits
• medical costs
• claims data

Introduction

Smoking is a serious health issue associated with substantial medical, social, and economic burdens. In 2003, smoking rates in Japan were 46.8% for men and 11.3% for women. Although these rates had decreased to 32.2% for men and 8.5% for women by 2014¹, it was still higher in men than that in western countries in 2015². Among Japanese men, 38.6% of deaths associated with cancer, 23.4% with respiratory diseases, and 23.0% with cardiovascular diseases are attributable to a history of smoking³. The total social costs associated with smoking in Japan in 2005 were estimated to be USD 56.31 billion (JPY 6,363 billion, 1 USD is equivalent to 113 JPY, as of November 2017), of which USD 20.9 billion (JPY 2,366 billion) accounted for labor losses due to smoking, USD 15.65 billion (JPY 1,768 billion) for excessive medical costs, USD 13.81 billion (JPY 1,560 billion) for labor losses due to the time spent smoking, USD 4.21 billion (JPY 476 billion) for excessive nursing care costs, and USD 1.70 billion (JPY 192 billion) for facility/environment-related costs³. Data indicated that up to 4% of the total healthcare costs in Japan are associated with smoking⁴.

In the past, Japan has trailed behind western countries when it comes to taking measures against smoking, resulting in a higher smoking rate in men compared to Western countries. In recent years, policy measures have been implemented, such as the 2003 Health Promotion Act, which incorporates measures to prevent passive smoking. In addition, in 2004 Japan ratified the World Health Organization Framework Convention on Tobacco Control, and, in 2012, Health Japan 21 was established, setting specific numerical goals such as decreasing the adult smoking rate to 12% by 2022⁵. Furthermore, the number of insurers incorporating smoking cessation in their business plans is increasing in response to the 2013 Japan Revitalization Strategy, which requests all health insurance providers to engage in efforts such as analysis of claims data and preparation of healthcare business plans for maintaining and promoting the health of their subscribers. Regarding the higher smoking rate in men compared to women, which is probably due to the cultural background, urgent measures to promote smoking cessation are needed, especially for men.

In the UK, free smoking cessation support is offered through the National Health Service, and health insurance reimbursement for smoking cessation is provided in many other western countries. In Japan, smoking cessation outpatient visits (SCOVs) were included in the health insurance coverage in 2006. To be eligible for health insurance coverage for smoking cessation the subject must (i) wish to stop smoking immediately, (ii) be diagnosed through the Tobacco Dependence Screener (TDS) as being addicted to nicotine, (iii) have a Brinkman Index score of 200 or higher, and (iv) consent in writing to receive smoking cessation therapy. The smoking cessation therapy program comprises five SCOVs, and subjects must visit a smoking cessation clinic at weeks 0, 2, 4, 8, and 12, and receive therapy in accordance with the standard operating procedures for smoking cessation therapy⁶.

There are only two drugs, nicotine patches and varenicline, which are approved for smoking cessation therapy covered by the Japanese public health insurance system. Nicotine addiction administration fees are USD 20.35 (JPY 2,300) for the first outpatient visit, USD 16.28 (JPY 1,840) for the second through the fourth outpatient visits, and USD 15.93 (JPY 1,800) for the fifth (final) outpatient visit. The co-insurance rate for the patient is generally 30%, and the costs are ∼USD 115.04 (JPY 13,000) for 8 weeks’ therapy using nicotine patches and ∼USD 168.14 (JPY 19,000) for 12 weeks’ therapy using varenicline⁶. Hence, the cost of smoking cessation therapy to the individual is usually lower than the cost of purchasing cigarettes (1/3 to 1/2 of the cost of smoking a pack per day). For those subjects who fail to stop smoking, a second course of therapy is also covered by health insurance, provided that at least 1 year has elapsed since participation in the first program.

Previous research has established the efficacy of SCOVs in terms of their beneficial impact on clinical outcomes^7–12. Epidemiological data for smoking-related diseases, such as cancer, chronic obstructive pulmonary disease, stroke, and myocardial infarction span from years to tens of years, and model-based assessments of the impact of smoking cessation on these diseases and their associated medical costs, therefore, assume long-term perspectives^13,14. However, smoking cessation is also associated with short-term health benefits, including symptom relief (such as reduction of asthma symptoms and coughing)¹⁵, as well as a reduction in mortality within 5 years of quitting¹⁶. In Japan, assessments of the cost-effectiveness of smoking cessation to date have been based on long-term perspectives. On the other hand, for most insurers, employers, and local government officials in charge of healthcare and healthcare promotion, any measures to promote smoking cessation are considered as an investment for which they would expect to see a short-term effect. Therefore, there is an unmet need for short-term assessments of the impact of smoking cessation on healthcare costs.

To address this need, we undertook a database analysis to investigate the relationship between SCOVs and short-term medical costs. Using health insurance claims data, we compared the medical costs for smokers who made SCOVs with those who did not. Furthermore, we examined the impact of the number of SCOVs made on medical costs before and after the SCOV.

Methods

Data source and patient identification

The analyses were based on claims data from Japan Medical Data Center Company, Ltd. (JMDC), which holds data from health insurers for ∼2 million employees of general companies and their family members, commonly aged 65 years or younger. This figure represents 7% of all insured individuals in Japan. The JMDC database contains claims data (e.g. subject characteristics, examinations, diagnoses, prescriptions, etc.) for inpatient, outpatient, and pharmacy, as well as annual medical checkup data, received by insured individuals from all insurance-covered medical institutions.

Among all insureds having medical checkup data in the database, we extracted individuals who were registered as smokers based on their medical checkup details. Patients who received smoking cessation therapy were defined as those diagnosed as “nicotine dependence” (F17.2) under the International Classification of Diseases, 10th revision (ICD-10)¹⁷, since patients must be diagnosed with “nicotine dependence” to receive smoking cessation therapy under the Japanese health insurance system.

The extracted data were used for two analyses: Analysis 1 compared the medical costs between the SCOV group and non-SCOV group; and Analysis 2 assessed the relationship between the number of SCOVs and the increased ratio of medical costs. We did not have this study reviewed via Ethical Committees, as this study was a retrospective, observational study using a commercial database in which all data were anonymized.

Analysis 1: Comparison of medical costs between the SCOV group and non-SCOV group

The following inclusion criteria were applied:

SCOV group

1. Men;

2. Having at least one claim with ICD-10 F17.2, indicating a SCOV between January 1, 2005 and December 31, 2013;

3. Aged 21–60 years at the first SCOV; and

4. Enrolled in the database at least 1 year before the first SCOV.

Non-SCOV group

1. Men;

2. Having no claims with ICD-10 F17.2;

3. Aged 21–60 years the first time the individual was classified as a smoker; and

4. Enrolled in the database at least 1 year before the first medical checkup visit at which the individual was classified as a smoker.

The per-patient-per-year (PPPY) medical cost for the SCOV group was compared with that of the non-SCOV group. In this analysis, we used regression-based methods for the model to adjust the difference of age between groups. The PPPY calculation method is outlined below:

Definitions of index date/index year

For subjects in the SCOV group, the date for the first SCOV was set as the Index Date; for subjects in the non-SCOV group, the Index Date was set as the first date of being registered as a smoker, based on medical checkup details. The first-year period from the Index Date was set as the Index Year (i.e. Year 0), and each subsequent year was defined in sequential numerical order. The year prior to the Index Year was defined as Year −1.

Calculation of medical costs for person-years

Total medical costs for each year post Year 0, and the total number of claims data-captured months (member months; MMOS) post Year 0 were calculated. The total number of claims data-captured person-years for each year was then calculated by dividing MMOS by 12. The PPPY medical costs for each year was then calculated by dividing the total medical costs by the number of captured years.

Regression of PPPY medical costs with age and year

PPPY medical costs for each group from Year −1 through Year 6 can be generalized by the following equation: where α and β are regression coefficients, x is the age on the Index Date, i is the year, and j is the index variable for the relevant group [j = 0 (SCOV group) or j = 1 (non-SCOV group)].

Calculation of PPPY medical costs by age and year

Using the regression equation above, PPPY medical costs for each year and age were calculated. Then, PPPY_i,j was calculated for each year by the following equation: where MMOS indicates the claims data-capture period for each patient. MMOS_x,i,j is the total MMOS of all patients in the SCOV group for each age, in the index year.

Correction by baseline medical costs

When comparing medical costs between the SCOV group and non-SCOV group, the medical costs of the SCOV group may be higher at baseline (before the first SCOV), because this group includes individuals making SCOVs due to development of other diseases, which may result in a greater number of claims made. To eliminate this impact, we defined an adjusted non-SCOV group based on the non-SCOV group, and a correction was made so that the medical costs of the adjusted non-SCOV group matched those of the SCOV group in Year −1. The equation used for this correction is as follows:

PPPY medical costs of the adjusted non-SCOV group (j = 2): where k is an adjustment term for regression to the average value determined based on the ratios of PPPY_x,i,0 and PPPY_x,i,1.

With the purpose of eliminating the impact of gender from the constructed model, only male smokers were included in this analysis.

Analysis 2: Relationship between number of SCOVs and the increased ratio of medical costs

We compared the medical costs for those in the SCOV group who made 1–5 SCOVs. For this analysis, men and women between 20–65 years old, who had at least one claim with ICD-10 F17.2 at least l year after enrollment in the database, were included. While in Analysis 1 the upper age limit was set at 60 years for the purpose of tracking medical costs for 6 years post the first SCOV, in Analysis 2 the upper age limit was set at 65 years. This was because medical costs for the year following the first SCOV and thereafter were calculated, irrespective of the number of elapsed years, and because the receipt data used for this analysis was obtained from health insurance societies where people over the age of 65 years are not well represented.

The average PPPY increase ratio in medical costs before and after the Index Year during the insured period was calculated as follows for each number of SCOVs: where PPPYB is the average PPPY medical cost in the year before the Index Year (Year −1)—calculated by dividing the medical costs before the first SCOV by the insured period before the first SCOV—and PPPYA is the average PPPY medical costs from the year following the Index Year (Year 1) through Year 7—calculated by dividing the total medical costs for Year 1 through Year 7 by the total insured period during Year 1 through Year 7.

Results

Analysis 1

Table 1 presents the demographic data for Analysis 1. This analysis included 87,329 male smokers, of which 5,608 had made a claim for a SCOV and were, thereby, classified into the SCOV group. The remaining 81,721 smokers had not made a claim for a SCOV and were classified into the non-SCOV group.

Table 1. Demographic data for Analysis 1.

	SCOV group (n = 5,608)	Non-SCOV group (n = 81,721)
Age, mean, years (SD)^a	42.7 (8.7)	40.4 (9.5)
Number of observed years, mean (SD)	2.3 (1.5)	3.3 (1.5)

Abbreviations. SCOV, smoking cessation outpatient visit; SD, standard deviation.

^a Age is the age of the subject in the Index Year.

The medical costs for the SCOV, non-SCOV, and adjusted non-SCOV groups for Year −1 through Year 6 are shown in Figure 1. In the SCOV group, PPPY medical costs increased from Year −1 to Year 0, and decreased from Year 0 to Year 1; from Year 1 through Year 6 the tendency was to remain stable or decrease in the PPPY medical costs. In the non-SCOV and adjusted non-SCOV groups there was a trend towards an annual increase in PPPY medical costs from Year −1 through Year 6, with the exception of the period from Year 3 through Year 4, where no change in medical costs was observed. Although average PPPY medical costs in the SCOV group were ∼USD 323.01 (JPY 36,500) higher than in the non-SCOV group in Year −1, by Year 6 the costs were similar for the SCOV and adjusted non-SCOV group. In Years 4–5, the average PPPY medical costs in the SCOV group dropped to below that of the adjusted non-SCOV group. In Years 5–6, the average PPPY medical costs of the SCOV group remained lower than that of the adjusted non-SCOV group.

Figure 1. Annual medical costs for subjects with (SCOV group) and without (non-SCOV group) health insurance claims for SCOVs. Error bars represent 95% confidence interval. Abbreviations. SCOVs, smoking cessation outpatient visits; PPPYY, average per-patient-per-year medical costs. For the adjusted non-SCOV group, medical costs were matched to those of the SCOV group in Index Year –1. 1 USD is equivalent to 113 JPY, as of November 2017.

Analysis 2

Table 2 presents the demographic data classified according to the number of SCOVs. The analysis included 2,576 male and female smokers who had made at least one claim for a SCOV. As the number of SCOVs increased, the average age of the participants also increased (Table 2).

Table 2. Demographic data for Analysis 2.

	Number of SCOVs
	1 (n = 596)	2 (n = 680)	3 (n = 594)	4 (n = 403)	5 (n = 303)
Age, mean, years (SD)	37.9 (9.2)	39.2 (9.5)	41.4 (9.4)	42.4 (9.3)	44.5 (10.0)
Number of observed years, mean (SD)	2.8 (1.2)	2.6 (1.1)	2.6 (0.9)	2.6 (0.9)	3.0 (1.3)

Abbreviations. SCOVs, smoking cessation outpatient visits; SD, standard deviation.

The average PPPY medical costs in Year −1 (PPPYB), the average PPPY medical costs from Year 1 through Year 7 (PPPYA), and the average increase in the ratio of medical costs for each number of SCOVs attended are presented in Table 3, while Figure 2 depicts the average increased ratio of medical costs per SCOV. As the number of SCOVs increased, there was a trend towards a decrease in the average increased ratio of medical costs. The average increased ratio of medical costs for SCOV number 1, 2, 3, 4, and 5 was 58%, 44%, 50%, 41%, and 34%, respectively. For those subjects who made five SCOVs, which is the standard duration of therapy for smoking cessation under health insurance coverage, the increased ratio of medical costs became lower than those who made ≤4 SCOVs.

Figure 2. Mean increase ratios of medical costs for each number of SCOVs attended. Solid line: straight line approximation of mean increase ratios of medical costs; dashed line: 95% confidence interval for the straight line approximation of mean increase ratios of medical costs. Abbreviations. SCOVs, smoking cessation outpatient visits.

Table 3. Average per-patient-per-year medical costs and average increase ratio of medical costs for each number of smoking cessation outpatient visits.

	Number of SCOVs
	1	2	3	4	5
PPPYB, USD (JPY)	824 (93,100)	757 (85,500)	995 (112,500)	1,135 (128,300)	1,767 (199,700)
PPPYA, USD (JPY)	1,302 (147,100)	1,087 (122,900)	1,490 (168,400)	1,595 (180,300)	2,369 (267,700)
Mean increase ratio of medical cost, %	58	44	50	41	34

Abbreviations. SCOVs, smoking cessation outpatient visits; PPPYB, average per-patient-per-year medical costs in the year before the Index Year (Year −1); PPPYA, average per-patient-per-year medical costs from the year following the Index Year (Year 1) through Year 7.

Amounts less than USD 1 and JPY 100 were rounded for PPPYB and PPPYA; the mean increase ratios of medical care costs were calculated based on rounded PPPYA and PPPYB values.

Discussion

A decrease in the long-term medical costs for smokers who enroll in smoking cessation programs has previously been reported^13,14. Our study showed that smokers who enrolled in smoking cessation programs (SCOV group) were associated with higher medical costs in the year prior to their first SCOV (Year −1) than those smokers who did not seek counseling/therapy (non-SCOV group). This finding is consistent with the results reported by Tsuji et al.¹⁸, and suggests that many of those smokers who were seeking SCOVs had previously received medical services. In the SCOV group, medical costs for the Index Year also increased compared with Year −1. This increase is thought to be due to the costs associated with enrolling in smoking cessation programs. However, the difference between the SCOV group and non-SCOV group is higher than the costs for the programs. A comparison of the medical costs during the Index Year by ICD-10 code classification did not identify any particular diagnoses, meaning that the costs covered various types of diagnoses other than F00–F99, which includes smoking cessation programs (F17.2). A possible reason for this difference is that they may see a doctor more often than the non-SCOV group for diseases/symptoms, including mild ones, while receiving smoking cessation programs.

The SCOV group was associated with a tendency to decrease the short-term medical costs (over a period of ≤6 years) after attending smoking cessation programs. In contrast, in the non-SCOV group, medical costs increased annually, and the initial ∼ USD 323.01 (JPY 36,500) difference in medical costs in Year −1 compared with the SCOV group shrank to ∼USD 23.01 (JPY 2,600) in Year 6. Furthermore, considering that the average medical costs of the SCOV group in Years 5–6 dropped to below those of the adjusted non-SCOV group, whose medical costs were set to be the same as those of the SCOV group in Year −1, this suggests that a decrease in medical costs for those attending smoking cessation programs is evident, even over a relatively shorter period of time.

While the medical costs in the non-SCOV group were associated with a trend towards an annual increase, the medical costs of the SCOV group were associated with a trend towards a decrease from the year following a smoking cessation program. To identify the cause of this, we calculated the difference in medical costs between the SCOV and non-SCOV groups for each ICD-10 code classification. Although this was an exploratory analysis, it gave an indication of trends that are consistent with conventional knowledge. In and after the Index Year, there was a trend towards a decrease in medical costs among smokers diagnosed with “Diseases of the circulatory system” (I00–I99). This is consistent with a report indicating that the risk of coronary artery diseases starts to decrease in the short-term (2–4 years) after smoking cessation¹⁵. Smokers with “Neoplasms” (C00-D48) who attended smoking cessation programs also showed a trend towards decreasing medical costs. This may be attributable to a drop in the rate of post-surgery complications in those who stop smoking prior to their surgery¹⁹. Among smokers diagnosed with “Diseases of the respiratory system” (J00–J99), no impact on medical costs was identified through this short-term analysis.

In general, medical costs increased with the increase in age²⁰. In this analysis of a sub-population of smokers, the increased ratios for medical costs were suppressed—even though the average age of individuals rose as the number of SCOVs increased. Therefore, there is a possibility that the observed results are related to the effect of an increase in the number of SCOVs. Smoking cessation has been reported to improve lung function in mild to moderate chronic obstructive pulmonary disease patients¹⁵, and to decrease the risk of reinfarction of coronary artery diseases¹⁵. There have also been reports of a decrease in the risk of post-surgery complications such as respiratory complications after lung cancer resection²¹, a decrease in post-surgery complications as a result of short-term smoking cessation¹⁹, as well as a lower risk of ICU hospitalization for overall post-surgery complications, infectious diseases, lung complications, and cranial nerve complications in non-smokers compared with smokers²². In addition to health conditions, labor market outcomes, such as lifetime income, which may be related to work productivities, are also reported to be affected by smoking²³. It is desirable to investigate the effect of smoking cessation on work productivity. The Ministry of Health, Labor and Welfare conducted a survey to assess the efficacy of the five-visit smoking cessation program in Japan. The result of this survey showed that 78.5% of the smokers who had completed the five therapy sessions did not smoke any cigarettes in the first 4 weeks after completing the program²⁴. The survey also found that, among the smokers who failed to complete the five-visit program, only 11.5%, 40.9%, 56.0%, and 66.9% of those who dropped out of the smoking cessation program after the first, second, third, or fourth visit, respectively, succeeded with smoking cessation²⁴.

There are some limitations in this research. First, the JMDC database has limited data on individuals >65 years of age. This is because the JMDC database is based on those insured by health insurance societies as company employees as well as their family members, and does not include data from the National Health Insurance, which also covers those who are self-employed and who tend to be relatively elderly. In this study, we targeted individuals of working-age (≤65 years of age). Furthermore, the JMDC database is only representative of ∼7% of the 30 million Japanese individuals who have health insurance²⁵. Nevertheless, this database is considered as a large-scale database and often used to analyze data of the working-age population in Japan^26–28. Moreover, a comparative study between the JMDC database and National Database, which is also a claims database including all claims data constructed by Japanese government, concluded that the results of analysis using the database can contribute back to society if the results are appropriately interpreted²⁹. Second, our research was centered on whether or not a smoker had made a SCOV based on insurance claims data, and did not capture the outcome of the smoking cessation counseling/therapy (i.e. whether the smoker stopped smoking as a result). Our analysis also provides no insight into those individuals in either group who have attempted smoking cessation without making an insurance claim for SCOVs, or those smokers in the SCOV group who failed to stop smoking despite attending smoking cessation counseling. However, it is well established that smoking cessation counseling programs and/or the use of oral anti-smoking therapy significantly increase the smoking cessation success rate compared with other measures such as quitting smoking all at once or with the help of over the counter adjuvant drugs. Combined with the knowledge that smokers who adhere to a smoking cessation program (i.e. attend a greater number of smoking cessation outpatient counseling/therapy visits) are more likely to successfully quit smoking, the resulting increase in the number of ex-smokers is entirely likely to impact the medical costs.

Conclusions

Most health economic research on smoking cessation is based on life-time medical costs, which is useful from the perspectives of public health, public policy, and healthcare payers. However, measures to promote smoking cessation are considered as an investment for which a short-term effect is expected. Through an analysis of health insurance claims data, we demonstrated that short-term (≤6 years) medical costs decreased for those smokers who sought smoking cessation counseling/therapy on an outpatient basis. Whilst the reason for this cannot be fully explained, this research showed that there is a relationship between smoking cessation outpatient visits and short-term medical costs. In addition, a greater number of smoking cessation visits was associated with a lower increase in the ratio of medical costs. Basically, smoking cessation should be promoted as a policy from the aspect of public health. Our results provide additional supportive evidence for smoking cessation from the point of view of medical costs. We expect to conduct a prospective cohort study to examine the motivation for smoking cessation and future effects on the medical costs shown in this study as a further study.

Transparency

Declaration of funding

This study was funded by Pfizer.

Declaration of financial interests/other relationships

KS and RY are employees of Pfizer. KI is an employee of Milliman, Inc., who was hired as a consultant for the development of this manuscript. AI reports grants and personal fees from Pfizer Japan during the conduct of the study; grants from Gilead Sciences, AbbVie, Abbott Japan, Beckton Dickinson and Company, Eli Lilly Japan, CSL Behring Japan, Fuji Film, Takeda Pharmaceutical, Milliman, Bristol-Myers Squibb, Intuitive Surgical, Boston Scientific Japan, Sanofi Japan, and Terumo Corporation outside of the submitted work; and personal fees from Chugai Pharmaceuticals, Astellas Pharma, and Creativ-Ceuticals outside of the submitted work. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Previous presentations

Parts of this study were presented at the ISPOR 18th Annual European Congress.

Acknowledgments

Editorial support was provided by Michelle Jenvey, PhD; Katy Beck, PhD; and Anne Jakobsen, MSc, at Engage Scientific Solutions. The editorial support was funded by Pfizer.