The Long-Run Impact of Changes in Prescription Drug Sales on Mortality and Hospital Utilization in Belgium, 1998–2019

Abstract

Objectives: We investigate the long-run impact of changes in prescription drug sales on mortality and hospital utilization in Belgium during the first two decades of the 21st century. Methods: We analyze the correlation across diseases between changes in the drugs used to treat the disease and changes in mortality or hospital utilization from that disease. The measure of the change in prescription drug sales we use is the long-run (1998–2018 or 2000–2019) change in the fraction of post-1999 drugs sold. A post-1999 drug is a drug that was not sold during 1989–1999. Results: The 1998–2018 increase in the fraction of post-1999 drugs sold is estimated to have reduced the number of years of life lost before ages 85, 75, and 65 in 2018 by about 438 thousand (31%), 225 thousand (31%), and 114 thousand (32%), respectively. The 1995–2014 increase in in the fraction of post-1999 drugs sold is estimated to have reduced the number of hospital days in 2019 by 2.66 million (20%). Conclusions: Even if we ignore the reduction in hospital utilization attributable to changes in pharmaceutical consumption, a conservative estimate of the 2018 cost per life-year before age 85 gained is EUR 6824. We estimate that previous changes in pharmaceutical consumption reduced 2019 expenditure on inpatient curative and rehabilitative care by EUR 3.55 billion, which is higher than the 2018 expenditure on drugs that were authorized during the period 1998–2018: EUR 2.99 billion.

1. Introduction

During the last few decades, there have been substantial changes in the prescription drugs sold in Belgium. According to data provided by IQVIA, the number of drugs (WHO ATC5 chemical substances1 (World Health Organization, 2023a) sold in Belgium increased from 1212 in 1989 to 1450 in 2019. Of the drugs sold in 1989, 51% were not sold in 2019; 60% of the drugs sold in 2019 had not been sold in 1989. Of the 1989 defined daily doses (DDDs), 25% were for drugs that were not sold in 2019; 51% of 2019 DDDs were for drugs that were not sold in 1989. The top 20 (ranked by DDD) antineoplastic and immunomodulating agents (ATC anatomical main group L) in 1989 and 2019 are shown in Appendix A 

Table A1. Top 20 (ranked by DDD) antineoplastic and immunomodulating agents in 1989 and 2019.

Top 20 (Ranked by DDD) Drugs in 1989
ATC	Molecule	Year Authorized	ddd1989
L01XX11	Estramustine	1978	58,327,163
L01BC02	Fluorouracil	1973	50,894,135
L02BB01	Flutamide	2000	29,799,702
L02BA01	Tamoxifene	1975	8,073,000
L01AA02	Chlorambucil	2001	6,483,074
L04AX01	Azathioprine	1971	6,239,931
L02AE03	Gosereline	1987	3,544,189
L01BB02	Mercaptopurine	1962	1,971,468
L01AD02	Lomustine		1,436,264
L01DB03	Epirubicine	1985	948,384
L01DB01	Doxorubicine	1989	899,964
L01BA01	Methotrexate	1968	801,600
L02AB01	Megestrol	1987	703,125
L04AD01	Ciclosporine	1983	635,000
L02AA02	Polyestradiol phosphate	1961	579,997
L01AA01	Cyclophosphamide	1962	525,625
L01BB03	Tioguanine	1974	506,692
L01CB01	Etoposide	1981	491,447
L01AA06	Ifosfamide	1981	314,281
L02AB02	Medroxyprogesterone		292,500
Top 20 (Ranked by DDD) Drugs in 2019
ATC	Molecule	Year Authorized	ddd2019
L04AB02	Infliximab	1999	800,380,242
L04AB04	Adalimumab	2003	448,143,808
L01BC02	Fluorouracil	1973	355,386,483
L02AE04	Triptoreline	1988	346,310,425
L04AX01	Azathioprine	1971	284,342,157
L04AC05	Ustekinumab	2009	216,678,846
L04AA06	Acide mycophenolique	1996	210,640,204
L02AE02	Leuproreline	2005	161,205,049
L02AE03	Gosereline	1987	134,882,903
L03AB07	Interferon beta-1a	1998	130,015,687
L02BX02	Degarelix	2009	104,666,264
L02BG06	Exemestane	2000	98,671,022
L04AB06	Golimumab	2009	86,281,752
L01XA01	Cisplatine	1998	80,811,504
L02BA03	Fulvestrant	2004	79,692,819
L04AX07	Dimethyl fumarate	2017	65,453,981
L01FF01	Nivolumab		64,716,505
L04AX05	Pirfenidone	2011	63,343,778
L01AA01	Cyclophosphamide	1962	63,264,313
L04AB05	Certolizumab pegol	2009	56,911,341

.2

The fact that there have been substantial changes in the prescription drug market is not surprising since the pharmaceutical industry is one of the most research-intensive industries. In Europe in 2021, the pharmaceutical industry’s ratio of R&D to sales (13.72%) was 3.46 times as high as it was for all industries (European Commission, 2022).

Many leading economists believe that economic growth is primarily driven by technological progress, which is generated by R&D investment. Jones (1998, pp. 89–90) argued that “technological progress is driven by research and development (R&D) in the advanced world.” Jones (2002) presented a model in which long-run growth is driven by the discovery (via research effort) of new ideas throughout the world. His model built upon a large collection of previous research, including Romer (1990), Grossman and Helpman (1991), and Aghion and Howitt (1992), as well as earlier contributions by Phelps (1966), Shell (1966), Nordhaus (1969), and Simon (1986).

Economists also recognize that mortality reduction is an important component of economic growth, broadly defined. Nordhaus (2005) argued that “to a first approximation, the economic value of increases in longevity in the last hundred years is about as large as the value of measured growth in non-health goods and services.” Cutler et al. (2006) concluded that “knowledge, science, and technology are the keys to any coherent explanation” of mortality.

In a nutshell:

The purpose of this study is to investigate the long-run impact of changes in prescription drug sales on mortality and hospital utilization in Belgium during the periods 1998–2018 and 2000–2019, respectively.3 To perform this assessment, we will analyze the correlation across diseases between changes in the drugs used to treat the disease and changes in mortality or hospital utilization from that disease. This “difference-in-differences” approach controls for the effects of aggregate macroeconomic and demographic trends, to the extent that those trends have similar effects on mortality or hospital utilization from different diseases.4

The measure of prescription drug sales change we will use is the long-run (1998–2018 or 2000–2019) change in the fraction of post-1999 drugs sold. A post-1999 drug is a drug that was not sold during 1989–1999. We have data on the sales of each drug in each year during the period 1989–2019.5 A total of 2270 drugs were sold in at least one year during that period. Figure 1a presents data on the number of drugs sold during 1989–2019 by the year first sold. Of the 2270 drugs, 53% (1212) were sold in the first year of the sample, 1989, and 366 (16%) other drugs were first sold sometime in the next decade, 1990–1999. Similar numbers (360 and 332) of drugs were first sold in the following two decades: 2000–2009 and 2010–2019.

The year in which a drug was first sold in Belgium is likely to be highly correlated with the vintage of the drug, defined as the year in which the drug was first sold anywhere in the world. Using data from two major databases, DrugCentral and Theriaque, we can estimate the year in which many drugs were first approved or sold in the USA and France; the pharmaceutical markets of both of those countries are much larger than Belgium’s. We define the vintage of a drug as min(year_USA, year_France), where year_USA = the drug’s FDA approval year, and year_France = the earliest year of sale in France. Figure 1b presents data on the mean vintage of drugs sold during 1989–2019, by the year first sold.6 As one would expect, mean vintage is strongly positively correlated with the year first sold in Belgium. The mean vintage of drugs first sold in 1989 is 1965.5. The mean vintage of drugs first sold during 1990–1999 is 1987.4, and so forth. We hypothesize that, in general, newer (later vintage) drugs tend to be of higher quality (more effective) than older (earlier vintage) drugs. Therefore, the higher the fraction of drugs that are post-1999 drugs, the higher the average quality of drugs.

The hypothesis that later vintage goods tend to be of higher quality than earlier vintage goods has been advocated by many economists since it was first formulated in the 1950s. Johansen (1959) developed a theoretical model of vintage capital in which there are technological improvements in capital in later vintages. Intriligator (1992, p. S77) said that “the newer capital is more productive than the older capital as a result of technological improvements in the later vintages.” Bresnahan and Gordon (1996) said that “new goods are at the heart of economic progress.” As noted by Jovanovic and Yatsenko (2012), in “the Spence–Dixit–Stiglitz tradition…new goods [are] of higher quality than old goods.” Bohlmann et al. (2002, p. 1177) said that “technology improves over time…As technology advances, later entrants can utilize a more recent and efficient vintage of technology than an earlier entrant who has committed to older technology…vintage effects will benefit later entrants…We refer to ‘vintage effects’ as any technology shift that results in lower costs for a later entrant, enabling it to achieve higher product quality.” In 1987, the Royal Swedish Academy of Sciences (1987) awarded the Alfred Nobel Memorial Prize in Economic Sciences to Robert Solow for his contributions to the theory of economic growth. The Academy cited Solow’s (1960) article, Investment and Technical Progress, in which he presented a new method of studying the role played by capital formation in economic growth. His basic assumption was that technical progress is “built into” machines and other capital goods and that this must be taken into account when making empirical measurements of the role played by capital. This idea then gave birth to the “vintage approach” (a similar idea was discussed by Leif Johansen in Norway at about the same time). The most important aspect of Solow’s article was not so much the empirical outcome but the method of analyzing “vintage capital.” Nowadays, the vintage capital concept has many other applications and is no longer solely employed in analyses of the factors underlying economic growth. The vintage approach has proved invaluable, both from the theoretical point of view and in applications such as the analysis of the development of industrial structures.

Figure 2 presents data on the percent of drugs sold in 2019 that were post-1999 drugs by disease for diseases that caused more than 1000 deaths in 2019. Over 60% of the drugs sold in 2019 that are used to treat diabetes mellitus and malignant neoplasms of the colon, rectum, and anus were post-1999 drugs. Only 23–25% of the drugs sold in 2019 that are used to treat pneumonia and other respiratory system diseases were post-1999 drugs. We will estimate econometric models to test the hypothesis that diseases that experienced larger increases in the fraction of post-1999 drugs had larger reductions in mortality and hospital utilization.

In the next section, we describe the econometric models that we estimate. In Section 3, we explain how the variables included in those models were constructed using data from a number of reliable sources. Estimates of the models are presented in Section 4. Major implications of the estimates are discussed in Section 5. The final section provides a summary and conclusions.

2. Econometric Model of Mortality and Hospital Utilization

Our estimates of the effect of (current or lagged) attributes of drugs sold on mortality and hospital utilization are based on the following model:7

ln(outcome_dt) = β_k post1999%_d,t-k + α_d + δ_t + ε_dt

(1)

where outcome_dt is one of the following variables:

n_deathsdt	= the number of deaths due to disease d in year t;
yll85dt	= the number of years of life lost before age 85 due to disease d in year t;
yll75dt	= the number of years of life lost before age 75 due to disease d in year t;
yll65dt	= the number of years of life lost before age 65 due to disease d in year t8;
hosp_daysdt	= the total number of hospital days due to disease d in year t;
dischargesdt	= the number of hospital discharges due to disease d in year t;
losdt	= the average length (in days) of hospital discharges due to disease d in year t.

and post1999%_d,t-k is defined as follows:

post1999%d,t-k

= the fraction of WHO ATC5 chemical substances used to treat disease d sold in year t-k that were first sold in Belgium after 1999.

post1999%_d,t-k is computed as follows:

post1999%_d,t-k = ∑_s post1999_s treat_sd sold_s,t-k/∑_s treat_sd sold_s,t-k

where

post1999s	= 1 if chemical substance s was first sold in Belgium after 19999; = 0 if chemical substance s was sold in Belgium during 1989–1999;
treatsd	= 1 if chemical substance s is used to treat (indicated for) disease d10; = 0 if chemical substance s is not used to treat (indicated for) disease d;
solds,t-k	= 1 if chemical substance s was sold in year t-k; = 0 if chemical substance s was not sold in year t-k.

In Equation (1), α_d is a fixed effect for disease d, and δ_t is a fixed effect for year t. The year fixed effects (δ_t‘s) in Equation (1) control for the effects of changes in aggregate demographic and macroeconomic variables (e.g., population size and age structure, GDP, educational attainment), to the extent that those variables have similar effects on mortality and hospital utilization from different diseases.

Equation (1) allows the effect of the post-1999 share of drugs sold on mortality and hospital utilization to be subject to a lag. Some drugs may have to be utilized for several years to have their maximum impact on those outcomes. We present estimates of mortality and hospital utilization models with lags of 0–5 years.

post1999%_d,t-k is the only disease-specific, time-varying regressor in Equation (1). If the data are available, we control for disease incidence. Failure to control for incidence is unlikely to result in overestimation of the magnitude of β_k; exclusion of incidence may even result in underestimation of the magnitude of β_k. Higher growth in disease incidence is likely to result in both higher disease burden and a larger fraction of post-1999 drugs sold:

The arrows indicate that an increase in disease incidence will cause an increase in both disease burden and an increase in post1999%.

Previous studies (Acemoglu & Linn, 2004; Danzon et al., 2005) have shown that both innovation (the number of drugs developed) and diffusion (the number of drugs launched in a country) depend on market size.

Annual mortality data are available for the years 1998–2018. From Equation (1) (a model of the level of an outcome), we can derive the following model of the long-run (1998–2018) growth of a mortality outcome:

Δln(outcome_d) = β_k Δpost1999%_k_d + δ’ + ε’_d

(2)

where

Δln(outcomed)	= ln(outcomed,2018) − ln(outcomed,1998) = the log-change from 1998 to 2018 of a mortality outcome from disease d;
Δpost1999%_kd	= post1999%d,2018-k − post1999%d,1998-k = the change from 1998-k to 2018-k in post1999% of disease d;
δ’	= δ2018 − δ1998;
ε’d	= εd,2018 − εd,1998.

We estimate Equation (2) by weighted least squares, weighting by (outcome_d,1998 + outcome_d,2018)/2. The analysis is performed on about 70 diseases included in the U.S. National Center for Health Statistics List of 113 Selected Causes of Death (New Jersey Department of Health, 2023). We exclude deaths from external causes.

The estimate of δ’ in Equation (2) is an estimate of mortality growth in the absence of changes in post1999%, i.e., if mean(Δpost1999%_k_d) = 0. This can be compared with mortality growth in the presence of changes in drug sales: mean(Δln(outcome_d)).

Annual hospital utilization data are available for the years 2000–2019. We estimate the following model of the long-run (2000–2019) growth of a hospital outcome:

Δln(outcome_d) = β_k Δpost1999%_k_d + δ’ + ε’_d

(3)

where

Δln(outcomed)	= ln(outcomed,2019) − ln(outcomed,2000) = the log-change from 2000 to 2019 of a hospital outcome from disease d;
Δpost1999%_kd	= post1999%d,2019-k − post1999%d,2000-k = the change from 2000-k to 2019-k in post1999% of disease d;
δ’	= δ2019 − δ2000;
ε’d	= εd,2019 − εd,2000.

We estimate Equation (3) by weighted least squares. For the first measure of hospital utilization, the weight is (hosp_days_d,2000 + hosp_days_d,2019)/2. For the second and third measures, the weight is (discharges_d,2000 + discharges_d,2019)/2. The analysis of hospital outcomes is performed on about 115 diseases included in the OECD Health Statistics database.

The long-run change in post1999% (Δpost1999%_d) might be correlated across diseases with other changes in disease treatment. Belgian data on treatment methods, by disease and year, are not available. However, U.S. data on the methods of treatment by office-based medical providers, by disease and year, are available from U.S. Medical Expenditure Panel Survey Office-Based Medical Provider Visits Event Files (Agency for Healthcare Research and Quality, 2023). The 1998 and 2015 files contain records of 104,740 and 172,388 patient visits, respectively. Each record indicates the patient’s diagnosis (coded by modified clinical classification code (CCS)), and whether the patient received each of the following: lab tests, a sonogram or ultrasound, x-rays, a mammogram, an mri/ctscan, an EKG or ECG, an EEG, a vaccination, anesthesia, or other diagnostic tests or exams. The fraction of visits in which at least one of those treatments was performed increased from 32% in 1998 to 40% in 2015. We computed the 1998–2015 change in the fraction of visits for each disease (157 CCS codes) in which any of those procedures were performed: Δany_procedure%_CCS = any_procedure%_CCS,2015 − any_procedure%_CCS,1998, where any_ procedure%_CCS,t = the fraction of visits with patient diagnosis CCS in year t in which any procedure was performed. We also computed the 1998–2015 change in the log of the number of and mean vintage of drugs sold using this disease classification: Δrx_measure_0_CCS = rx_measure_0_CCS,2015 − rx_measure_0_CCS,1998. Then, we estimated the following simple regression by weighted least-squares, weighting by the mean number of patient visits in 1998 and 2015: Δany_ procedure%_CCS = α + π Δrx_measure_0_CCS + ε. When Δrx_measure_0_CCS = Δln(n_drug_CCS,t), the estimate of π (=−0.117; t-value = −2.95; p-value = 0.0036). When Δrx_measure_0_CCS = Δvintage_CCS,t, the estimate of π (=−0.0066; t-value = −2.79; p-value = 0.0059). Diseases for which there were larger 1998–2015 increases in the number and vintage of drugs sold had significantly smaller 1998–2015 increases in the fraction of U.S. office-based medical provider visits in which any procedure was performed. This suggests that, to the extent that use of those procedures reduces mortality and hospital utilization, failure to control for them may bias estimates of β_k in Equations (2) and (3) towards zero.

3. Data Sources

Pharmaceutical sales data. Data on the number of DDDs (if any) of drugs sold in each year (1989–2019) were provided by IQVIA.

Drug indications and vintage. Data on the approved indications of each drug (treat_sd) were obtained by combining data from two sources. The first source is the Thériaque database, produced by the Centre National Hospitalier d’Information sur le Médicament (2025). This database contains information on over 30,000 drug products sold in France. For each product, the database provides (1) the WHO ATC code(s) of the substance(s) contained in the product; (2) the ICD-10 codes of the product’s approved indications; and (3) the year it was first sold in France. This enabled us to compute the approved indications of each ATC code, and the year it was first sold in France.

The second source is the DrugCentral (2025) database.11 Data on treat_sd (an indicator of whether chemical substance s is used to treat (indicated for) disease d) were constructed from the Omop_relationship and Synonyms tables of the DrugCentral (2025) database, and the SNOMED CT to ICD-10-CM Map (National Library of Medicine, 2025).12 Data on FDA_year_s (the FDA approval year of chemical substance s) were obtained from the Approval table of the DrugCentral (2025) database.

Neither of these databases is complete. Thériaque provides information about 2343 ATC codes, 925 of which are not included in DrugCentral. DrugCentral provides information about 2463 ATC codes, 1045 of which are not included in Thériaque. Information about 1418 ATC codes is available from both sources. We considered drugs to have indication d (treat_sd = 1) if either database said that was the case.

Mortality and population data were obtained from the WHO Mortality Database (World Health Organization, 2023b).

Hospital utilization data were obtained from the OECD Health Statistics database (OECD, 2023b). The disease classification is the International Shortlist for Hospital Morbidity Tabulation (OECD, 2023a).

Aggregate mortality and population data for 1998 and 2018 and aggregate hospital utilization data for 2000 and 2019 are shown in Appendix A 

Table A2. (A) Aggregate mortality and population data for 1998 and 2018; (B) Aggregate hospital utilization data for 2000 and 2019.

(A)
		Years of life lost (YLL)
Year	No. of deaths	before age 85	before age 75	before age 65
1998	104,382	1,196,074	613,414	313,316
2018	110,693	946,467	469,792	218,460
% change	6.0%	−20.9%	−23.4%	−30.3%
	Population
	total	below age 85	below age 75	below age 65
1998	10,203,008	10,023,391	9,500,058	8,514,986
2018	11,427,068	11,102,874	10,413,740	9,279,006
% change	12.0%	10.8%	9.6%	9.0%
	Rates per 100,000 people
	Deaths	YLL85	YLL75	YLL65
1998	1023	11933	6457	3680
2018	969	8525	4511	2354
% change	−5.3%	−28.6%	−30.1%	−36.0%
(B)
	days	discharges	los	population
2000	12,867,139	1,671,057	7.7	10,250,000
2019	11,525,244	1,920,874	6.0	11,490,000
% change	−10.4%	14.9%	−22.1%	12.1%
	Rates per 100,000 people
	days	discharges
2000	125,533	16,303
2019	100,307	16,718
% change	−20.1%	2.5%

.

Appendix A 

Table A3. Subset of data used in mortality analysis.

	Deaths		YLL85		YLL75		YLL65		Post1999%
Cause of Death/Year	1998	2018	1998	2018	1998	2018	1998	2018	2013	2018
Certain other intestinal infections (A04,A07–A09)	94	477	876	1837	526	732	369	334	4.2%	11.1%
Respiratory tuberculosis (A16)	63	14	925	128	468	53	208	10	0.0%	10.0%
Other tuberculosis (A17–A19)	12	12	224	219	134	154	81	104	0.0%	9.4%
Meningococcal infection (A39)	24	8	1753	232	1513	177	1273	137	14.3%	28.6%
Septicemia (A40–A41)	696	825	6338	6154	2750	2586	1233	991	5.3%	13.2%
Syphilis (A50–A53)	4	2	80	23	45	13	28	3	0.0%	8.3%
Viral hepatitis (B15–B19)	169	1	2567	33	1259	23	552	13	60.9%	74.3%
Human immunodeficiency virus (HIV) disease (B20–B24)	57	53	2376	678	1813	358	1283	160	45.7%	54.5%
Other and unspecified infectious and parasitic diseases and their sequelae (A00,A05,A20–A36,A42–A44,A48–A49,A54–A79,A81–A82,A85.0-A85.1,A85.8,A86–B04,B06–B09,B25–B49,B55–B99)	223	910	2926	4995	1544	2063	824	775	17.1%	22.2%
Malignant neoplasms of lip, oral cavity, and pharynx (C00–C14)	566	522	11,968	8880	7163	4653	3445	1623	35.0%	35.0%
Malignant neoplasm of esophagus (C15)	621	739	10,770	10,115	5690	4645	2335	1410	30.0%	36.4%
Malignant neoplasm of stomach (C16)	1115	641	12,864	7575	5814	3825	2204	1593	30.4%	40.7%
Malignant neoplasms of colon, rectum, and anus (C18–C21)	3138	2628	36,128	25,295	15,768	11,180	5523	3878	51.6%	61.5%
Malignant neoplasms of liver and intrahepatic bile ducts (C22)	619	938	8328	11,685	3628	5288	1215	1763	38.9%	52.2%
Malignant neoplasm of pancreas (C25)	1310	1849	16,868	21,170	7348	9265	2553	3008	28.0%	30.8%
Malignant neoplasm of larynx (C32)	314	160	5760	2363	3140	1118	1358	338	40.0%	40.0%
Malignant neoplasms of trachea, bronchus, and lung (C33–C34)	6379	5861	99,083	78,020	46,055	34,943	15,413	10,163	30.6%	46.8%
Malignant melanoma of skin (C43)	220	355	4600	5193	2768	2915	1473	1365	36.0%	44.8%
Malignant neoplasm of breast (C50)	2540	2237	42,195	28,613	23,068	14,798	10,515	6163	28.6%	35.5%
Malignant neoplasm of cervix uteri (C53)	178	174	3493	3440	2060	2108	1088	1068	40.9%	43.5%
Malignant neoplasms of corpus uteri and uterus, part unspecified (C54–C55)	388	382	4648	3365	2068	1348	748	358	27.3%	34.8%
Malignant neoplasm of ovary (C56)	695	574	9770	6678	4760	3083	1973	1103	25.8%	27.3%
Malignant neoplasm of prostate (C61)	1646	1598	12,408	8933	3608	2473	590	395	28.6%	35.5%
Malignant neoplasms of kidney and renal pelvis (C64–C65)	602	522	7712	4891	3482	2004	1275	589	53.3%	60.6%
Malignant neoplasm of bladder (C67)	876	798	8863	5590	3350	2033	865	558	25.0%	24.0%
Malignant neoplasms of meninges, brain, and other parts of the central nervous system (C70–C72)	706	636	15,269	13,355	9259	8062	5126	4355	36.0%	33.3%
Hodgkin’s disease (C81)	61	41	1473	650	983	380	620	215	13.3%	22.0%
Non-Hodgkin’s lymphoma (C82–C85)	657	662	10,020	6113	5103	2623	2298	1093	22.5%	30.4%
Leukemia (C91–C95)	892	972	12,522	9494	6442	4501	3274	2144	24.0%	35.3%
Multiple myeloma and immunoproliferative neoplasms (C88,C90)	420	517	4923	3960	2140	1385	735	380	26.7%	40.5%
Other and unspecified malignant neoplasms of lymphoid, hematopoietic, and related tissue (C96)	4	38	70	385	33	175	5	83	38.1%	47.6%
All other and unspecified malignant neoplasms (C17,C23–C24,C26–C31,C37–C41,C44–C49,C51–C52,C57–C60,C62–C63,C66,C68–C69,C73–C80,C97)	3531	3293	44,699	35,962	21,509	17,407	9129	7092	40.4%	52.3%
In situ neoplasms, benign neoplasms and neoplasms of uncertain or unknown behavior (D00–D48)	447	1477	4223	11,929	1936	5421	918	2256	45.0%	50.0%
Anemias (D50–D64)	122	201	765	823	308	398	143	225	29.4%	32.7%
Diabetes mellitus (E10–E14)	1671	1499	14,828	9295	5930	3573	2188	1128	53.4%	64.0%
Malnutrition (E40–E46)	25	176	180	475	100	160	63	25	0.0%	0.0%
Other nutritional deficiencies (E50–E64)	3	5	20	80	8	40	0	10	25.9%	25.9%
Meningitis (G00,G03)	24	29	547	680	369	475	249	323	18.9%	25.0%
Parkinson’s disease (G20–G21)	643	1333	3648	5513	785	1108	110	183	25.0%	33.3%
Alzheimer’s disease (G30)	1002	2345	4920	5553	1005	855	120	128	33.3%	33.3%
Acute rheumatic fever and chronic rheumatic heart diseases (I00–I09)	211	255	1400	1358	403	515	85	200	0.0%	6.9%
Hypertensive heart disease (I11)	345	290	2023	748	823	265	308	93	0.0%	6.1%
Hypertensive heart and renal disease (I13)	56	204	245	308	55	30	13	0	0.0%	6.1%
Acute myocardial infarction (I21–I22)	7735	3879	83,040	34,370	36,103	15,408	13,378	5535	25.6%	34.8%
Other acute ischemic heart diseases (I24)	437	324	3083	2520	1003	1148	268	415	53.8%	69.2%
Atherosclerotic cardiovascular disease, so described (I25.0)	4742	2717	36,153	15,605	12,125	5705	3243	1773	29.2%	38.5%
All other forms of chronic ischemic heart disease (I20,I25.1–I25.9)	5070	2770	38,428	15,965	12,925	5880	3453	1828	32.7%	43.1%
Acute and subacute endocarditis (I33)	18	59	240	675	85	300	18	135	0.0%	4.5%
Diseases of pericardium and acute myocarditis (I30–I31,I40)	34	43	408	323	183	128	90	38	0.0%	0.0%
Heart failure (I50)	4513	5043	18,733	14,575	5805	5063	1820	1710	22.4%	33.9%
All other forms of heart disease (I26–I28,I34–I38,I42–I49,I51)	7571	7497	55,873	39,823	23,178	16,923	9455	6718	27.1%	32.3%
Essential hypertension and hypertensive renal disease (I10,I12,I15)	499	423	4180	2230	1625	915	603	288	21.9%	28.4%
Cerebrovascular diseases (I60–I69)	8627	6478	57,479	34,635	20,719	13,377	7619	4842	47.2%	57.1%
Atherosclerosis (I70)	918	111	4195	410	1273	143	313	43	46.7%	60.0%
Aortic aneurysm and dissection (I71)	637	488	6275	3610	2375	1430	723	463	0.0%	0.0%
Other diseases of arteries, arterioles, and capillaries (I72–I78)	503	581	4543	4230	1920	1815	690	633	27.3%	35.3%
Other disorders of circulatory system (I80–I99)	397	210	3470	1948	1520	938	645	375	16.4%	25.0%
Influenza (J09–J11)	368	543	1443	3918	513	2053	188	1141	33.3%	20.0%
Pneumonia (J12–J18)	3913	4718	21,770	17,455	8145	6318	3440	2318	18.6%	22.5%
Acute bronchitis and bronchiolitis (J20–J21)	109	192	785	452	442	195	310	105	17.5%	20.9%
Other and unspecified acute lower respiratory infections (J22,U04)	19	538	65	1601	13	531	0	204	5.1%	11.9%
Bronchitis, chronic and unspecified (J40–J42)	534	272	2623	720	815	223	265	45	17.0%	25.9%
Emphysema (J43)	702	231	7960	2195	2978	813	678	213	38.1%	53.3%
Asthma (J45–J46)	344	109	5718	1013	3138	500	1560	218	18.5%	29.5%
Other chronic lower respiratory diseases (J44,J47)	3715	3935	32,063	31,845	10,610	11,648	2365	2665	28.0%	41.0%
Pneumoconioses and chemical effects (J60–J66,J68)	327	55	3255	180	880	55	115	15	0.0%	8.0%
Pneumonitis due to solids and liquids (J69)	491	923	3218	3740	1290	1253	573	368	0.0%	13.6%
Other diseases of respiratory system (J00–J06,J30–J39,J67,J70–J98)	1398	1400	11,191	9707	4214	3787	1534	1302	18.5%	24.9%
Peptic ulcer (K25–K28)	414	195	3229	1258	1279	545	499	178	27.8%	40.0%
Diseases of appendix (K35–K38)	14	11	103	58	33	30	18	13	50.0%	33.3%
Hernia (K40–K46)	88	111	473	413	143	133	25	48	0.0%	0.0%
Alcoholic liver disease (K70)	849	686	22,600	15,693	14,570	9263	7733	4100	50.0%	54.5%
Other chronic liver disease and cirrhosis (K73–K74)	425	644	7780	10,833	4328	5700	2038	2183	31.6%	51.9%
Cholelithiasis and other disorders of gallbladder (K80–K82)	267	252	1455	788	453	218	125	33	14.3%	0.0%
Acute and rapidly progressive nephritic and nephrotic syndrome (N00–N01,N04)	4	8	33	130	8	93	0	63	0.0%	9.4%
Chronic glomerulonephritis, nephritis and nephropathy not specified as acute or chronic, and renal sclerosis unspecified (N02–N03,N05–N07,N26)	29	16	310	128	175	40	90	5	23.1%	35.7%
Renal failure (N17–N19)	961	1508	6069	4985	2129	1580	707	488	33.3%	45.9%
Other disorders of kidney (N25,N27)	2	3	8	43	0	18	0	8	18.2%	25.0%
Infections of kidney (N10–N12,N13.6,N15.1)	113	433	680	1505	230	423	78	100	20.0%	23.3%
Hyperplasia of prostate (N40)	25	22	118	50	23	8	3	0	43.8%	47.1%
Inflammatory diseases of female pelvic organs (N70–N76)	7	12	55	198	33	125	23	83	10.7%	20.0%
Pregnancy with abortive outcome (O00–O07)	1	1	58	48	48	38	38	28	12.5%	22.2%
Other complications of pregnancy, childbirth, and the puerperium (O10–O99)	12	3	640	143	520	113	400	83	20.4%	25.0%
Certain conditions originating in the perinatal period (P00–P96)	152	186	12,785	15,717	11,265	13,857	9745	11,997	22.8%	31.6%
Congenital malformations, deformations, and chromosomal abnormalities (Q00–Q99)	287	296	20,170	16,733	17,390	13,871	14,663	11,233	13.7%	23.2%
Symptoms, signs, and abnormal clinical and laboratory findings, not elsewhere classified (R00–R99)	3392	7722	29,948	59,598	16,623	31,985	9306	15,878	17.5%	20.7%
All other diseases (Residual)	8333	13,712	77,264	91,924	40,837	43,577	22,717	19,929	29.6%	36.8%
Other and unspecified non-transport accidents and their sequelae (W20–W31,W35–W64,W75–W99,X10–X39,X50–X59,Y86)	703	2000	7417	14,657	4592	8067	3009	4615	0.0%	20.0%
Complications of medical and surgical care (Y40–Y84,Y88)	91	554	1070	5305	558	2358	305	915	17.6%	20.6%

contains a subset of the data used in the mortality analysis. Appendix A 

Table A4. Subset of data used in hospital utilization analysis.

	Days		Discharges		Los		Post1999%
	2000	2019	2000	2019	2000	2019	2014	2019
101 Intestinal infectious diseases except diarrhea	68,350	60,643	13,670	14,791	5.0	4.1	5.3%	8.3%
102 Diarrhea and gastroenteritis of presumed infectious origin	19,289	15,464	4104	3682	4.7	4.2	0.0%	7.1%
103 Tuberculosis	24,869	15,514	1256	929	19.8	16.7	3.1%	9.4%
104 Septicemia	61,537	211,935	4130	14,925	14.9	14.2	7.9%	13.9%
105 Human immunodeficiency virus [HIV] disease	6380	5039	417	442	15.3	11.4	47.9%	56.6%
106 Other infectious and parasitic diseases	146,676	164,723	21,570	25,738	6.8	6.4	25.4%	32.8%
201 Malignant neoplasm of colon, rectum, and anus	157,474	105,074	9661	9820	16.3	10.7	54.5%	62.5%
202 Malignant neoplasms of trachea, bronchus, and lung	125,939	101,521	11,449	9953	11.0	10.2	34.2%	47.9%
203 Malignant neoplasms of skin	12,946	9326	2232	2743	5.8	3.4	43.8%	52.6%
204 Malignant neoplasm of breast	90,368	48,107	11,439	12,335	7.9	3.9	31.6%	36.5%
205 Malignant neoplasm of uterus	27,695	13,530	2978	2460	9.3	5.5	34.5%	41.9%
206 Malignant neoplasm of ovary	24,162	12,968	1873	1425	12.9	9.1	25.8%	27.3%
207 Malignant neoplasm of prostate	72,597	34,887	6914	6709	10.5	5.2	31.0%	35.5%
208 Malignant neoplasm of bladder	46,178	36,743	5701	6335	8.1	5.8	25.0%	24.0%
209 Other malignant neoplasms	509,189	467,168	39,472	44,920	12.9	10.4	41.8%	53.9%
210 Carcinoma in situ	11,830	14,737	2232	4754	5.3	3.1	52.0%	53.8%
211 Benign neoplasm of colon, rectum, and anus	14,063	11,612	3430	4004	4.1	2.9	40.0%	40.0%
212 Leiomyoma of uterus	45,589	12,803	7353	4742	6.2	2.7	57.1%	57.1%
213 Other benign neoplasms and neoplasms of uncertain or unknown behavior	101,486	67,830	17,201	15,416	5.9	4.4	48.0%	54.2%
301 Anemias	70,832	88,870	8638	12,174	8.2	7.3	30.8%	33.3%
302 Other diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism	33,696	37,012	5265	5213	6.4	7.1	28.1%	37.0%
401 Diabetes mellitus	196,032	129,049	17,348	15,363	11.3	8.4	55.7%	65.4%
402 Other endocrine, nutritional, and metabolic diseases	194,691	203,615	26,670	40,723	7.3	5.0	28.6%	37.0%
501 Dementia	133,692	60,602	5641	3503	23.7	17.3	15.4%	23.1%
502 Mental and behavioral disorders due to alcohol	75,215	73,135	11,061	11,081	6.8	6.6	15.4%	15.4%
503 Mental and behavioral disorders due to the use of other psychoactive subst.	16,446	13,798	2164	2379	7.6	5.8	30.4%	30.4%
504 Schizophrenia, schizotypal, and delusional disorders	22,955	18,192	1739	1895	13.2	9.6	23.1%	28.2%
505 Mood [affective] disorders	178,522	72,095	13,947	4938	12.8	14.6	26.2%	29.8%
506 Other mental and behavioral disorders	146,198	74,176	15,229	11,071	9.6	6.7	26.3%	27.6%
601 Alzheimer’s disease	47,752	85,500	2221	4750	21.5	18.0	33.3%	33.3%
602 Multiple sclerosis	13,083	14,097	1869	1905	7.0	7.4	15.8%	28.6%
603 Epilepsy	81,619	73,922	12,182	16,070	6.7	4.6	30.3%	34.4%
604 Transient cerebral ischemic attacks and related syndromes	82,895	42,437	8127	8161	10.2	5.2	36.4%	40.0%
605 Other diseases of the nervous system	217,241	331,607	21,509	106,970	10.1	3.1	21.2%	28.3%
701 Cataract	32,137	3466	16,914	2666	1.9	1.3	0.0%	0.0%
702 Other diseases of the eye and adnexa	36,593	20,782	10,455	9896	3.5	2.1	21.8%	28.1%
800 Diseases of the ear and mastoid process	50,504	36,045	12,318	12,015	4.1	3.0	7.4%	9.6%
901 Hypertensive diseases	64,031	225,194	6885	21,447	9.3	10.5	22.7%	30.5%
902 Angina pectoris	40,091	6507	8909	2324	4.5	2.8	34.7%	45.1%
903 Acute myocardial infarction	150,604	111,507	15,853	17,423	9.5	6.4	25.6%	34.8%
904 Other ischemic heart disease	307,562	116,464	50,420	35,292	6.1	3.3	32.4%	41.2%
905 Pulmonary heart disease and diseases of pulmonary circulation	69,714	55,902	5126	6655	13.6	8.4	56.5%	60.0%
906 Conduction disorders and cardiac arrhythmias	203,266	156,702	28,629	40,180	7.1	3.9	19.5%	18.4%
907 Heart failure	264,786	161,424	18,136	14,160	14.6	11.4	22.4%	34.4%
908 Cerebrovascular diseases	497,162	379,870	30,131	28,778	16.5	13.2	47.4%	57.8%
909 Atherosclerosis	119,354	92,518	12,179	17,792	9.8	5.2	46.7%	63.6%
910 Varicose veins of lower extremities	39,758	6550	14,725	2047	2.7	3.2	0.0%	0.0%
911 Other diseases of the circulatory system	347,319	254,325	38,591	32,193	9.0	7.9	13.7%	21.1%
1001 Acute upper respiratory infections and influenza	47,309	98,482	11,002	24,020	4.3	4.1	13.1%	17.3%
1002 Pneumonia	406,654	364,934	32,020	36,862	12.7	9.9	18.3%	21.4%
1003 Other acute lower respiratory infections	140,760	107,210	17,595	21,442	8.0	5.0	12.7%	16.1%
1005 Other diseases of upper respiratory tract	50,853	25,400	20,341	14,111	2.5	1.8	19.2%	24.1%
1006 Chronic obstructive pulmonary disease and bronchiectasis	365,719	381,498	27,706	35,654	13.2	10.7	25.4%	34.2%
1007 Asthma	52,483	24,374	7392	5078	7.1	4.8	21.4%	31.0%
1008 Other diseases of the respiratory system	215,490	163,333	16,325	16,013	13.2	10.2	12.9%	21.0%
1101 Disorders of teeth and supporting structures	17,569	6670	9247	2668	1.9	2.5	7.0%	11.6%
1102 Other diseases of oral cavity, salivary glands, and jaws	12,250	10,526	3141	2339	3.9	4.5	10.0%	15.0%
1103 Diseases of esophagus	79,438	37,556	13,464	8942	5.9	4.2	29.4%	31.3%
1104 Peptic ulcer	91,321	33,314	8953	4217	10.2	7.9	30.0%	40.0%
1105 Dyspepsia and other diseases of stomach and duodenum	39,001	31,920	5821	6650	6.7	4.8	0.0%	0.0%
1109 Crohn’s disease and ulcerative colitis	33,997	29,471	3434	4534	9.9	6.5	21.6%	34.1%
1110 Other noninfective gastroenteritis and colitis	44,706	49,358	9512	11,752	4.7	4.2	8.0%	12.5%
1111 Paralytic ileus and intestinal obstruction without hernia	99,900	94,160	9250	11,919	10.8	7.9	0.0%	0.0%
1113 Diseases of anus and rectum	32,432	20,920	7371	6153	4.4	3.4	16.7%	18.2%
1114 Other diseases of intestine	75,460	55,395	9316	9389	8.1	5.9	14.5%	19.6%
1115 Alcoholic liver disease	52,877	55,740	4099	4645	12.9	12.0	54.5%	55.6%
1116 Other diseases of liver	39,708	30,015	3514	3335	11.3	9.0	32.4%	50.0%
1117 Cholelithiasis	123,157	101,946	19,864	27,553	6.2	3.7	0.0%	0.0%
1118 Other diseases of gall bladder and biliary tract	47,510	36,320	5001	6262	9.5	5.8	0.0%	0.0%
1119 Diseases of pancreas	65,792	60,378	6036	8271	10.9	7.3	0.0%	0.0%
1120 Other diseases of the digestive system	45,075	89,802	4553	16,036	9.9	5.6	7.5%	13.2%
1201 Infections of the skin and subcutaneous tissue	49,620	52,240	7406	8564	6.7	6.1	17.3%	18.6%
1202 Dermatitis, eczema and papulosquamous disorders	13,598	6985	1478	1145	9.2	6.1	18.0%	27.3%
1203 Other diseases of the skin and subcutaneous tissue	106,814	60,754	7854	8438	13.6	7.2	24.8%	33.8%
1301 Coxarthrosis [arthrosis of hip]	174,496	147,200	11,556	21,647	15.1	6.8	18.2%	18.2%
1302 Gonarthrosis [arthrosis of knee]	155,805	195,456	10,599	26,413	14.7	7.4	25.0%	25.0%
1303 Internal derangement of knee	30,769	3590	13,378	2112	2.3	1.7	10.0%	10.0%
1304 Other arthropathies	96,528	106,690	13,223	29,636	7.3	3.6	22.4%	30.6%
1305 Systemic connective tissue disorders	22,963	19,509	2208	2192	10.4	8.9	23.8%	37.0%
1306 Deforming dorsopathies and spondylopathies	85,815	142,319	9535	18,246	9.0	7.8	20.8%	28.0%
1307 Intervertebral disc disorders	153,289	100,321	22,879	23,886	6.7	4.2	33.3%	20.0%
1308 Dorsalgia	58,053	23,214	10,555	4380	5.5	5.3	25.0%	25.0%
1309 Soft tissue disorders	24,191	67,109	4244	21,648	5.7	3.1	19.8%	19.6%
1310 Other disorders of the musculoskeletal system and connective tissue	267,794	162,996	41,199	18,953	6.5	8.6	25.8%	29.7%
1401 Glomerular and renal tubulo-interstitial diseases	74,570	98,974	10,215	22,494	7.3	4.4	6.9%	14.0%
1402 Renal failure	64,963	40,105	4848	3518	13.4	11.4	33.3%	43.8%
1403 Urolithiasis	62,099	24,478	18,818	12,883	3.3	1.9	12.5%	16.7%
1404 Other diseases of the urinary system	114,270	188,402	15,236	26,167	7.5	7.2	17.6%	21.8%
1405 Hyperplasia of prostate	84,037	36,758	10,774	9425	7.8	3.9	47.1%	52.9%
1406 Other diseases of male genital organs	25,782	25,766	8057	5992	3.2	4.3	10.7%	11.5%
1407 Disorders of breast	18,773	7778	6953	4575	2.7	1.7	16.7%	20.0%
1408 Inflammatory diseases of female pelvic organs	14,551	6198	3307	1675	4.4	3.7	10.8%	14.7%
1409 Menstrual, menopausal, and other female genital conditions	21,347	4770	6099	1908	3.5	2.5	36.1%	38.5%
1410 Other disorders of the genitourinary system	105,106	46,165	21,897	17,098	4.8	2.7	26.5%	31.0%
1501 Medical abortion	1600	316	842	186	1.9	1.7	16.7%	16.7%
1502 Other pregnancy with abortive outcome	11,472	4622	6038	3081	1.9	1.5	12.5%	12.5%
1503 Complications of pregnancy predominantly in the antenatal period	324,429	294,986	50,692	77,628	6.4	3.8	17.6%	27.8%
1504 Complications of pregnancy predominantly during labor and delivery	145,228	132,318	26,405	36,755	5.5	3.6	33.3%	36.4%
1505 Single spontaneous delivery	202,630	14,445	41,353	4815	4.9	3.0	33.3%	33.3%
1506 Other delivery	40,456	672	6224	160	6.5	4.2	20.0%	25.0%
1507 Complications predominantly related to the puerperium	12,740	6683	2498	2025	5.1	3.3	10.0%	0.0%
1601 Disorders related to short gestation and low birth weight	14,388	21,532	723	1321	19.9	16.3	25.0%	25.0%
1602 Other conditions originating in the perinatal period	22,967	19,828	3281	3605	7.0	5.5	17.0%	21.9%
1700 Congenital malformations, deformations, and chromosomal abnormalities	58,919	43,065	10,911	9570	5.4	4.5	18.9%	23.6%
1801 Pain in throat and chest	29,755	19,414	8042	8089	3.7	2.4	8.3%	8.3%
1802 Abdominal and pelvic pain	26,347	16,621	7749	6156	3.4	2.7	12.5%	9.1%
1804 Other symptoms, signs, and abnormal clinical and laboratory findings	317,655	327,257	58,825	64,168	5.4	5.1	19.4%	22.9%
1901 Intracranial injury	100,723	93,632	20,984	13,376	4.8	7.0	0.0%	0.0%
1904 Fracture of femur	387,258	381,486	16,340	20,510	23.7	18.6	75.0%	75.0%
1905 Fracture of lower leg, including ankle	100,013	95,303	10,418	13,423	9.6	7.1	100.0%	.
1906 Other injuries	459,480	503,055	61,264	60,609	7.5	8.3	15.8%	16.7%
1907 Burns and corrosions	18,042	10,922	1841	1126	9.8	9.7	0.0%	7.7%
1908 Poisonings by drugs, medicaments, and biological substances and toxic effects of substances chiefly nonmedicinal as to source	46,855	68,775	11,428	13,226	4.1	5.2	23.3%	25.8%
1909 Complications of surgical and medical care, not elsewhere classified	245,731	288,708	23,628	35,643	10.4	8.1	22.6%	32.6%
1911 Other and unspecified effects of external causes	15,535	21,533	2390	4141	6.5	5.2	13.0%	17.3%
2102 Contraceptive management	4120	560	2575	467	1.6	1.2	38.5%	45.5%
2104 Other medical care (including radiotherapy and chemotherapy sessions)	130,191	192,654	30,277	14,595	4.3	13.2	0.0%	0.0%
2105 Other factors influencing health status and contact with health services	281,418	254,280	34,743	50,856	8.1	5.0	10.7%	18.5%

contains a subset of the data used in the hospital utilization analysis.

4. Estimates of Models of Mortality and Hospital Utilization

4.1. Mortality Growth Estimates

Estimates of β_k from Equation (2) for mortality outcomes are presented in

Table 1. Estimates of β_k from Equation (2) for mortality outcomes: ln(outcome_d,2018) − ln(outcome_d,1998) = β_k (post1999%_d,2018-k − post1999%_d,1998-k) + δ’ + ε’_d.

Row	k (lag)	Estimate	Std. Err.	t Value	Pr > |t|
		Dep. Var.: Δln(n_deaths)
1	0	−1.521	0.415	−3.67	0.0004
2	1	−1.579	0.429	−3.68	0.0004
3	2	−1.420	0.450	−3.15	0.0022
4	3	−1.349	0.484	−2.79	0.0065
5	4	−1.355	0.488	−2.77	0.0068
6	5	−1.373	0.495	−2.77	0.0068
		Dep. Var.: Δln(yll85)
7	0	−0.954	0.409	−2.33	0.0220
8	1	−1.007	0.426	−2.37	0.0202
9	2	−0.815	0.440	−1.85	0.0674
10	3	−0.712	0.467	−1.53	0.1306
11	4	−0.756	0.473	−1.60	0.1136
12	5	−0.823	0.463	−1.78	0.0789
		Dep. Var.: Δln(yll75)
13	0	−0.960	0.399	−2.41	0.0181
14	1	−1.014	0.416	−2.44	0.0169
15	2	−0.838	0.428	−1.96	0.0538
16	3	−0.758	0.452	−1.68	0.0971
17	4	−0.818	0.459	−1.78	0.0782
18	5	−0.835	0.437	−1.91	0.0596
		Dep. Var.: Δln(yll65)
19	0	−1.064	0.410	−2.60	0.0112
20	1	−1.103	0.429	−2.57	0.0120
21	2	−0.925	0.439	−2.11	0.0382
22	3	−0.872	0.460	−1.90	0.0615
23	4	−0.947	0.468	−2.02	0.0465
24	5	−0.885	0.430	−2.06	0.0430

Note: each estimate is from a separate regression. Estimates in bold are statistically significant (p-value < 0.05).

and plotted in Figure 3. Each estimate shown is from a separate regression. In rows 1–6 of the table, the dependent variable is the 1998–2018 change in the log of the number of deaths. The assumed lag from post1999% to outcome ranges between 0 years (row 1) to 5 years (row 6).

The estimates in rows 1–6 indicate that the 1998–2018 growth in the number of deaths is significantly inversely related across diseases to the change in post1999% 0–5 years earlier. It is most significantly inversely related to the change in post1999% one year earlier.

In rows 7–12 of the table, the dependent variable is the 1998–2018 change in the log of the number of years of life lost before age 85 (YLL85). The 1998–2018 growth in YLL85 is significantly inversely related across diseases to the change in post1999% 0–1 years earlier.

In rows 13–18 of the table, the dependent variable is the 1998–2018 change in the log of the number of years of life lost before age 75 (YLL75). The 1998–2018 growth in YLL75 is significantly inversely related across diseases to the change in post1999% 0–1 years earlier, and marginally significantly (p-value = 0.0538) inversely related across diseases to the change in post1999% 2 years earlier.

In rows 19–24 of the table, the dependent variable is the 1998–2018 change in the log of the number of years of life lost before age 65 (YLL65). The 1998–2018 growth in YLL65 is significantly inversely related across diseases to the change in post1999% 0–2 and 4–5 years earlier.

As discussed above, the estimate of δ’ in Equation (2) is an estimate of mortality growth in the absence of changes in drug sales, i.e., if mean(Δpost1999%_d) = 0. This can be compared with mortality growth in the presence of changes in drug sales: mean(Δln(outcome_d)). These calculations, based on estimates for which the lag is most significant, are shown in Figure 4.

Panel A of Figure 4 shows the calculations when the outcome measure is the number of deaths. The actual 1998–2018 decline in the number of deaths per 100,000 people was 7.1% (this decline is not statistically significantly different from zero). The estimates in row 1 of Table 1 indicate that, if post1999% had not increased between 1998 and 2018, the number of deaths per 100,000 people would have increased by 65.6%. We do not think that estimate is implausible: the population by age figures in Appendix A Table A2 indicate that the fraction of the population over age 74 increased from 6.9% to 8.9%, and the fraction of the population over age 84 increased from 1.8% to 2.8%.

Panel B of Figure 4 shows the calculations when the outcome measure is YLL85. The number of years of life lost before age 85 per 100,000 people below age 85 (POP_LT_85) declined by 28.0%. This decline is statistically significant. The estimates in row 7 of Table 1 indicate that, if post1999% had not increased between 1998 and 2018, YLL85/POP_LT_85 would have increased by 5.0%, but this increase is not significantly different from zero.

Panel C of Figure 4 shows the calculations when the outcome measure is YLL75. The number of years of life lost before age 75 per 100,000 people below age 85 (POP_LT_75) declined by 26.5%. This decline is statistically significant. The estimates in row 13 of Table 1 indicate that, if post1999% had not increased between 1998 and 2018, YLL75/POP_LT_75 would have increased by 6.9%, but this increase is not significantly different from zero.

Panel D of Figure 4 shows the calculations when the outcome measure is YLL65. The number of years of life lost before age 65 per 100,000 people below age 85 (POP_LT_65) declined by 29.2%. This decline is statistically significant. The estimates in row 19 of Table 1 indicate that, if post1999% had not increased between 1998 and 2018, YLL65/POP_LT_65 would have increased by 4.4%, but this increase is not significantly different from zero.

4.2. Hospital Utilization Growth Estimates

Estimates of β_k from Equation (3) for hospital utilization outcomes are presented in

Table 2. Estimates of β_k from Equation (3) for hospital utilization outcomes: ln(outcome_d,2019) − ln(outcome_d,2000) = β_k (post1999%_d,2019-k − post1999%_d,2000-k) + δ’ + ε’_d.

Row	k (lag)	Estimate	Std. Err.	t Value	Pr > |t|
		Dep. Var.: Δln(days)
1	0	−0.135	0.276	−0.49	0.6254
2	1	−0.437	0.287	−1.52	0.1301
3	2	−0.706	0.341	−2.07	0.0405
4	3	−0.771	0.347	−2.22	0.0285
5	4	−0.888	0.370	−2.40	0.0179
6	5	−0.917	0.379	−2.42	0.0171
		Dep. Var.: Δln(discharges)
7	0	0.441	0.384	1.15	0.2537
8	1	0.046	0.409	0.11	0.9102
9	2	0.037	0.454	0.08	0.9347
10	3	−0.077	0.463	−0.17	0.8686
11	4	−0.172	0.489	−0.35	0.7261
12	5	−0.179	0.503	−0.35	0.7235
		Dep. Var.: Δln(los)
13	0	−0.333	0.200	−1.67	0.0985
14	1	−0.496	0.209	−2.37	0.0195
15	2	−0.595	0.231	−2.57	0.0114
16	3	−0.569	0.237	−2.41	0.0177
17	4	−0.609	0.250	−2.44	0.0164
18	5	−0.644	0.257	−2.51	0.0136

Note: each estimate is from a separate regression. Estimates in bold are statistically significant (p-value < 0.05).

and plotted in Figure 5. Each estimate shown is from a separate regression. In rows 1–6 of the table, the dependent variable is the 2000–2019 change in the log of the number of hospital days. The assumed lag from post1999% to outcome ranges between 0 years (row 1) to 5 years (row 6).

The estimates in rows 1–6 of Table 2 indicate that the 2000–2019 change in the log of the number of hospital days is not significantly related to change in post1999% 0–1 years earlier, but it is significantly inversely related to the change in post1999% 2–5 years earlier.

The estimates in rows 7–12 of Table 2 indicate that the 2000–2019 change in the log of the number of hospital discharges is not significantly related to the change in post1999% 0–5 years earlier.

The estimates in rows 13–18 of Table 2 indicate that the 2000–2019 change in the log of the average length of hospital stays is significantly inversely related to the change in post1999% 1–5 years earlier.

Comparisons of hospital utilization growth in the presence of changes in post1999% to hospital utilization growth in the absence of changes in post1999%, are shown in Figure 6. As shown in Panel A, between 2000 and 2019 the number of hospital days per 100,000 people declined by 24.6%. This decline is statistically significant. The estimates in row 6 of Table 2 imply that if post1999% had not increased between 1995 and 2014, the number of hospital days per 100,000 people would have declined by only 6.2% between 2000 and 2019; that estimate is not statistically significantly different from zero.

ln(outcome_d,2018) − ln(outcome_d,1998) = β_k (post1999%_d,2018-k − post1999%_d,1998-k) + δ’ + ε’_d

As shown in Panel B, between 2000 and 2019, the number of hospital discharges per 100,000 people declined by 2.3%. That decline is not statistically significant. The estimates in row 7 of Table 2 imply that if post1999% had not increased between 2000 and 2019, the number of hospital discharges per 100,000 people would have declined more, by 12.6% between 2000 and 2019; this estimate is also not statistically significantly different from zero.

As shown in Panel C, between 2000 and 2019 the average length of hospital stays declined by 26.9%, and that decline is statistically significant. The estimates in row 15 of Table 2 imply that if post1999% had not increased between 1998 and 2017, the average length of hospital stays would have declined by only 14.8% between 2000 and 2019; that decline is also statistically significant.

5. Discussion

The estimates shown in Figure 4 and Figure 6 indicate the estimated effects of changes in pharmaceutical consumption in the long-run (1998–2018 or 2000–2019) growth in mortality and hospital utilization. Now, we will use these estimates to calculate the percentage and absolute differences in the levels of mortality in 2018 and hospital utilization in 2019 attributable to changes in pharmaceutical consumption. These calculations are shown in

Table 3. Estimated percentage and absolute differences in levels of mortality in 2018 and hospital utilization in 2019 attributable to changes in pharmaceutical consumption.

Column	1	2	3	4	5	6
	2018 Deaths	2018 YLL85	2018 YLL75	2018 YLL65	2019 Days	2019 Discharges
actual or estimated values in 2018 or 2019
actual	109,853	957,772	495,759	242,255	10,869,130	1,827,788
estimated, if constant post1999%	195,795	1,395,907	720,906	357,145	13,527,798	1,633,813
estimated % reduction in 2018 or 2019 due to change in post1999%	44%	31%	31%	32%	20%	−12%
estimated reduction in 2018 or 2019 due to change in post1999%	85,941	438,135	225,148	114,890	2,658,668	−193,975

.

The number of deaths per 100,000 people was 1023 in 1998. It declined by 7.1% to 950 by 2018. The estimates in Figure 5 imply that if post1999% had remained constant, the number of deaths per 100,000 people in 2018 would have been 73% (=(1 + 65.6%)/(1 − 7.1%)) higher. As shown in column 1 of Table 3, the 1998–2018 increase in post1999% is estimated to have reduced the number of deaths in 2018 by 85,941 (44%).

As shown in columns 2–4 of Table 3, the 1998–2018 increase in post1999% is estimated to have reduced the number of years of life lost before ages 85, 75, and 65 in 2018 by about 438 thousand (31%), 225 thousand (31%), and 114 thousand (32%), respectively.

Data from the marketing research firm IQVIA and from Agence Fédérale des Médicaments et des Produits de Santé (2023) indicate that 2018 expenditure on drugs that were authorized during the period 1998–2018 was EUR 2.99 billion (62% of total drug expenditure). This implies that, even if we ignore the reduction in hospital utilization attributable to changes in pharmaceutical consumption, a conservative estimate of the cost per life-year before age 85 gained is EUR 6824 (=EUR 2.99 billion/438,135 life-years). As noted by Bertram et al. (2016), authors writing on behalf of the WHO’s Choosing Interventions that are Cost–Effective project (WHO-CHOICE) suggested in 2005 that “interventions that avert one disability-adjusted life-year (DALY) for less than average per capita income for a given country or region are considered very cost–effective; interventions that cost less than three times average per capita income per DALY averted are still considered cost–effective.” Belgium’s per capita GDP in 2018 was EUR 40,260.

Moreover, as shown in column 5 of Table 3, the 1995–2014 increase in post1999% is estimated to have reduced the number of hospital days in 2019 by 2.66 million (20%). We estimate that, in the absence of previous changes in pharmaceutical consumption, the number of hospital days in 2019 would have been 24% (=13.53 million/10.87 million) higher than it actually was. It is reasonable to assume that expenditure on inpatient curative and rehabilitative care would have increased proportionately, by 24%. According to the OECD, expenditure on inpatient curative and rehabilitative care in 2019 was EUR 14.5 billion. Hence, we estimate that previous changes in pharmaceutical consumption reduced 2019 expenditure on inpatient curative and rehabilitative care by EUR 3.55 billion (=24% × EUR 14.5 billion). This estimate of hospital cost reduction is higher than 2018 expenditure on drugs that were authorized during the period 1998–2018 (EUR 2.99 billion).13

This study has several limitations. First, it relies on aggregate data, which may mask important variations in treatment effects across different population subgroups. Second, the long-run change in the fraction of post-1999 drugs is the only disease-specific, time-varying regressor in the models we estimated. This variable might be correlated across diseases with other changes in disease treatment. U.S. data on methods of treatment by office-based medical providers indicate that the diseases for which there were larger 1998–2015 increases in the number and vintage of drugs sold had significantly smaller 1998–2015 increases in the fraction of U.S. office-based medical provider visits in which any procedure was performed. However, that may not be the case in Belgium. If detailed data on nonpharmaceutical treatments in Belgium become available, they should be incorporated into the analysis.

6. Summary and Conclusions

We have investigated the long-run impact of changes in prescription drug sales on mortality and hospital utilization in Belgium during the first two decades of the 21st century, by analyzing the correlation across diseases between changes in the drugs used to treat the disease and changes in mortality or hospital utilization from that disease. This “difference-in-differences” approach controls for the effects of aggregate macroeconomic and demographic trends, to the extent that those trends have similar effects on mortality or hospital utilization from different diseases.

The measure of prescription drug sales change we used is the long-run (1998–2018 or 2000–2019) change in the fraction of post-1999 drugs sold. A post-1999 drug is a drug that was not sold during 1989–1999.

The 1998–2018 growth of the number of deaths and the number of years of life lost before ages 85, 75, and 65 is significantly inversely related across diseases to the change in the fraction of post-1999 drugs sold. The 1998–2018 increase in post1999% is estimated to have reduced the number of years of life lost before ages 85, 75, and 65 in 2018 by about 438 thousand (31%), 225 thousand (31%), and 114 thousand (32%), respectively. Even if we ignore the reduction in hospital utilization attributable to changes in pharmaceutical consumption, a conservative estimate of the 2018 cost per life-year before age 85 gained is EUR 6824. This figure is 17% of Belgium’s 2018 per capita income; authors writing on behalf of the WHO-CHOICE project suggested in 2005 that “interventions that avert one disability-adjusted life-year (DALY) for less than average per capita income for a given country or region are considered very cost–effective.”

Moreover, the 2000–2019 growth of the number of hospital days is highly significantly inversely related to the change in the fraction of post-1999 drugs sold 2–5 years earlier. The 1995–2014 increase in the fraction of post-1999 drugs sold is estimated to have reduced the number of hospital days in 2019 by 2.66 million (20%).

We estimated that previous changes in pharmaceutical consumption reduced 2019 expenditure on inpatient curative and rehabilitative care by EUR 3.55 billion. This estimate of hospital cost reduction is higher than 2018 expenditure on drugs that were authorized during the period 1998–2018: EUR 2.99 billion. Our estimates imply that the long-run change in pharmaceutical consumption was cost-saving, i.e., that its net cost per life-year gained was negative.

In many countries (including Belgium), there are discussions of increased public expenditure for medicines, even if their share in total healthcare expenditure decreases, as it has in Belgium. Increased expenditures are partly due to an aging population—most medicine expenditure is concentrated in the last 5 years of life—and to the arrival of new, highly valuable and therefore more expensive medicines.

For many other healthcare expenditures, spending increases are mainly due to increasing volumes, not to increasing value: paying more for the same outcome versus paying more for a better outcome. Whereas other healthcare expenditures are often viewed as an investment, pharmaceutical expenditure is sometimes mostly considered as a cost that needs to be reduced. Such a view neglects the savings that investments in medicines can realize elsewhere in the healthcare budget, like reduced hospital costs, as demonstrated in this study.

This study contributes to a broader perspective on the value of medicines and to a more nuanced view on their expenses. The value of medicines extends beyond the value to patients (like increased life expectancy and quality of life), because there are also broader societal benefits to be taken into account, like reduced health system costs and increased worker productivity. This broader perspective is becoming even more important in the context of recent cell and gene therapies providing cure instead of care.